diff --git a/ext/README.md b/ext/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..933a33d053bf6704be350bc3464a6fd6f02fd2a1
--- /dev/null
+++ b/ext/README.md
@@ -0,0 +1,8 @@
+## Loadable Extensions
+
+Various [loadable extensions](https://www.sqlite.org/loadext.html) for
+SQLite are found in subfolders.
+
+Most subfolders are dedicated to a single loadable extension (for
+example FTS5, or RTREE). But the misc/ subfolder contains a collection
+of smaller single-file extensions.
diff --git a/ext/expert/README.md b/ext/expert/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..28886fd1f2d75d5074c9287fc3aef0cbfb189db4
--- /dev/null
+++ b/ext/expert/README.md
@@ -0,0 +1,83 @@
+## SQLite Expert Extension
+
+This folder contains code for a simple system to propose useful indexes
+given a database and a set of SQL queries. It works as follows:
+
+ 1. The user database schema is copied to a temporary database.
+
+ 1. All SQL queries are prepared against the temporary database.
+ Information regarding the WHERE and ORDER BY clauses, and other query
+ features that affect index selection are recorded.
+
+ 1. The information gathered in step 2 is used to create candidate
+ indexes - indexes that the planner might have made use of in the previous
+ step, had they been available.
+
+ 1. A subset of the data in the user database is used to generate statistics
+ for all existing indexes and the candidate indexes generated in step 3
+ above.
+
+ 1. The SQL queries are prepared a second time. If the planner uses any
+ of the indexes created in step 3, they are recommended to the user.
+
+# C API
+
+The SQLite expert C API is defined in sqlite3expert.h. Most uses will proceed
+as follows:
+
+ 1. An sqlite3expert object is created by calling **sqlite3\_expert\_new()**.
+ A database handle opened by the user is passed as an argument.
+
+ 1. The sqlite3expert object is configured with one or more SQL statements
+ by making one or more calls to **sqlite3\_expert\_sql()**. Each call may
+ specify a single SQL statement, or multiple statements separated by
+ semi-colons.
+
+ 1. Optionally, the **sqlite3\_expert\_config()** API may be used to
+ configure the size of the data subset used to generate index statistics.
+ Using a smaller subset of the data can speed up the analysis.
+
+ 1. **sqlite3\_expert\_analyze()** is called to run the analysis.
+
+ 1. One or more calls are made to **sqlite3\_expert\_report()** to extract
+ components of the results of the analysis.
+
+ 1. **sqlite3\_expert\_destroy()** is called to free all resources.
+
+Refer to comments in sqlite3expert.h for further details.
+
+# sqlite3_expert application
+
+The file "expert.c" contains the code for a command line application that
+uses the API described above. It can be compiled with (for example):
+
+
+ gcc -O2 sqlite3.c expert.c sqlite3expert.c -o sqlite3_expert
+
+
+Assuming the database is named "test.db", it can then be run to analyze a
+single query:
+
+
+ ./sqlite3_expert -sql <sql-query> test.db
+
+
+Or an entire text file worth of queries with:
+
+
+ ./sqlite3_expert -file <text-file> test.db
+
+
+By default, sqlite3\_expert generates index statistics using all the data in
+the user database. For a large database, this may be prohibitively time
+consuming. The "-sample" option may be used to configure sqlite3\_expert to
+generate statistics based on an integer percentage of the user database as
+follows:
+
+
+ # Generate statistics based on 25% of the user database rows:
+ ./sqlite3_expert -sample 25 -sql <sql-query> test.db
+
+ # Do not generate any statistics at all:
+ ./sqlite3_expert -sample 0 -sql <sql-query> test.db
+
diff --git a/ext/expert/expert.c b/ext/expert/expert.c
new file mode 100644
index 0000000000000000000000000000000000000000..051480f896025877c00bfd7c7f569980eb5556d3
--- /dev/null
+++ b/ext/expert/expert.c
@@ -0,0 +1,156 @@
+/*
+** 2017 April 07
+**
+** The author disclaims copyright to this source code. In place of
+** a legal notice, here is a blessing:
+**
+** May you do good and not evil.
+** May you find forgiveness for yourself and forgive others.
+** May you share freely, never taking more than you give.
+**
+*************************************************************************
+*/
+
+
+#include
+#include
+#include
+#include
+#include "sqlite3expert.h"
+
+
+static void option_requires_argument(const char *zOpt){
+ fprintf(stderr, "Option requires an argument: %s\n", zOpt);
+ exit(-3);
+}
+
+static int option_integer_arg(const char *zVal){
+ return atoi(zVal);
+}
+
+static void usage(char **argv){
+ fprintf(stderr, "\n");
+ fprintf(stderr, "Usage %s ?OPTIONS? DATABASE\n", argv[0]);
+ fprintf(stderr, "\n");
+ fprintf(stderr, "Options are:\n");
+ fprintf(stderr, " -sql SQL (analyze SQL statements passed as argument)\n");
+ fprintf(stderr, " -file FILE (read SQL statements from file FILE)\n");
+ fprintf(stderr, " -verbose LEVEL (integer verbosity level. default 1)\n");
+ fprintf(stderr, " -sample PERCENT (percent of db to sample. default 100)\n");
+ exit(-1);
+}
+
+static int readSqlFromFile(sqlite3expert *p, const char *zFile, char **pzErr){
+ FILE *in = fopen(zFile, "rb");
+ long nIn;
+ size_t nRead;
+ char *pBuf;
+ int rc;
+ if( in==0 ){
+ *pzErr = sqlite3_mprintf("failed to open file %s\n", zFile);
+ return SQLITE_ERROR;
+ }
+ fseek(in, 0, SEEK_END);
+ nIn = ftell(in);
+ rewind(in);
+ pBuf = sqlite3_malloc64( nIn+1 );
+ nRead = fread(pBuf, nIn, 1, in);
+ fclose(in);
+ if( nRead!=1 ){
+ sqlite3_free(pBuf);
+ *pzErr = sqlite3_mprintf("failed to read file %s\n", zFile);
+ return SQLITE_ERROR;
+ }
+ pBuf[nIn] = 0;
+ rc = sqlite3_expert_sql(p, pBuf, pzErr);
+ sqlite3_free(pBuf);
+ return rc;
+}
+
+int main(int argc, char **argv){
+ const char *zDb;
+ int rc = 0;
+ char *zErr = 0;
+ int i;
+ int iVerbose = 1; /* -verbose option */
+
+ sqlite3 *db = 0;
+ sqlite3expert *p = 0;
+
+ if( argc<2 ) usage(argv);
+ zDb = argv[argc-1];
+ if( zDb[0]=='-' ) usage(argv);
+ rc = sqlite3_open(zDb, &db);
+ if( rc!=SQLITE_OK ){
+ fprintf(stderr, "Cannot open db file: %s - %s\n", zDb, sqlite3_errmsg(db));
+ exit(-2);
+ }
+
+ p = sqlite3_expert_new(db, &zErr);
+ if( p==0 ){
+ fprintf(stderr, "Cannot run analysis: %s\n", zErr);
+ rc = 1;
+ }else{
+ for(i=1; i<(argc-1); i++){
+ char *zArg = argv[i];
+ int nArg;
+ if( zArg[0]=='-' && zArg[1]=='-' && zArg[2]!=0 ) zArg++;
+ nArg = (int)strlen(zArg);
+ if( nArg>=2 && 0==sqlite3_strnicmp(zArg, "-file", nArg) ){
+ if( ++i==(argc-1) ) option_requires_argument("-file");
+ rc = readSqlFromFile(p, argv[i], &zErr);
+ }
+
+ else if( nArg>=3 && 0==sqlite3_strnicmp(zArg, "-sql", nArg) ){
+ if( ++i==(argc-1) ) option_requires_argument("-sql");
+ rc = sqlite3_expert_sql(p, argv[i], &zErr);
+ }
+
+ else if( nArg>=3 && 0==sqlite3_strnicmp(zArg, "-sample", nArg) ){
+ int iSample;
+ if( ++i==(argc-1) ) option_requires_argument("-sample");
+ iSample = option_integer_arg(argv[i]);
+ sqlite3_expert_config(p, EXPERT_CONFIG_SAMPLE, iSample);
+ }
+
+ else if( nArg>=2 && 0==sqlite3_strnicmp(zArg, "-verbose", nArg) ){
+ if( ++i==(argc-1) ) option_requires_argument("-verbose");
+ iVerbose = option_integer_arg(argv[i]);
+ }
+
+ else{
+ usage(argv);
+ }
+ }
+ }
+
+ if( rc==SQLITE_OK ){
+ rc = sqlite3_expert_analyze(p, &zErr);
+ }
+
+ if( rc==SQLITE_OK ){
+ int nQuery = sqlite3_expert_count(p);
+ if( iVerbose>0 ){
+ const char *zCand = sqlite3_expert_report(p,0,EXPERT_REPORT_CANDIDATES);
+ fprintf(stdout, "-- Candidates -------------------------------\n");
+ fprintf(stdout, "%s\n", zCand);
+ }
+ for(i=0; i0 ){
+ fprintf(stdout, "-- Query %d ----------------------------------\n",i+1);
+ fprintf(stdout, "%s\n\n", zSql);
+ }
+ fprintf(stdout, "%s\n%s\n", zIdx, zEQP);
+ }
+ }else{
+ fprintf(stderr, "Error: %s\n", zErr ? zErr : "?");
+ }
+
+ sqlite3_expert_destroy(p);
+ sqlite3_free(zErr);
+ return rc;
+}
diff --git a/ext/expert/expert1.test b/ext/expert/expert1.test
new file mode 100644
index 0000000000000000000000000000000000000000..0c3b512af0ab45a2c4c87ff4518fd614ad7b608b
--- /dev/null
+++ b/ext/expert/expert1.test
@@ -0,0 +1,609 @@
+# 2009 Nov 11
+#
+# The author disclaims copyright to this source code. In place of
+# a legal notice, here is a blessing:
+#
+# May you do good and not evil.
+# May you find forgiveness for yourself and forgive others.
+# May you share freely, never taking more than you give.
+#
+#***********************************************************************
+# TESTRUNNER: shell
+#
+# The focus of this file is testing the CLI shell tool. Specifically,
+# the ".recommend" command.
+#
+#
+
+# Test plan:
+#
+#
+if {![info exists testdir]} {
+ set testdir [file join [file dirname [info script]] .. .. test]
+}
+source $testdir/tester.tcl
+set testprefix expert1
+
+if {[info commands sqlite3_expert_new]==""} {
+ finish_test
+ return
+}
+
+
+set CLI [test_binary_name sqlite3]
+set CMD [test_binary_name sqlite3_expert]
+
+proc squish {txt} {
+ regsub -all {[[:space:]]+} $txt { }
+}
+
+proc do_setup_rec_test {tn setup sql res} {
+ reset_db
+ if {[info exists ::set_main_db_name]} {
+ dbconfig_maindbname_icecube db
+ }
+ db eval $setup
+ uplevel [list do_rec_test $tn $sql $res]
+}
+
+foreach {tn setup} {
+ 1 {
+ if {![file executable $CMD]} { continue }
+
+ proc do_rec_test {tn sql res} {
+ set res [squish [string trim $res]]
+ set tst [subst -nocommands {
+ squish [string trim [exec $::CMD -verbose 0 -sql {$sql;} test.db]]
+ }]
+ uplevel [list do_test $tn $tst $res]
+ }
+ }
+ 2 {
+ if {[info commands sqlite3_expert_new]==""} { continue }
+
+ proc do_rec_test {tn sql res} {
+ set expert [sqlite3_expert_new db]
+ $expert sql $sql
+ $expert analyze
+
+ set result [list]
+ for {set i 0} {$i < [$expert count]} {incr i} {
+ set idx [string trim [$expert report $i indexes]]
+ if {$idx==""} {set idx "(no new indexes)"}
+ lappend result $idx
+ lappend result [string trim [$expert report $i plan]]
+ }
+
+ $expert destroy
+
+ set tst [subst -nocommands {set {} [squish [join {$result}]]}]
+ uplevel [list do_test $tn $tst [string trim [squish $res]]]
+ }
+ }
+ 3 {
+ if {[info commands sqlite3_expert_new]==""} { continue }
+ set ::set_main_db_name 1
+ }
+ 4 {
+ if {![file executable $CLI]} { continue }
+
+ proc do_rec_test {tn sql res} {
+ set res [squish [string trim $res]]
+ set tst [subst -nocommands {
+ squish [string trim [exec $::CLI test.db ".expert" {$sql;}]]
+ }]
+ uplevel [list do_test $tn $tst $res]
+ }
+ }
+} {
+
+ eval $setup
+
+
+do_setup_rec_test $tn.1 { CREATE TABLE t1(a, b, c) } {
+ SELECT * FROM t1
+} {
+ (no new indexes)
+ SCAN t1
+}
+
+do_setup_rec_test $tn.2 {
+ CREATE TABLE t1(a, b, c);
+} {
+ SELECT * FROM t1 WHERE b>?;
+} {
+ CREATE INDEX t1_idx_00000062 ON t1(b);
+ SEARCH t1 USING INDEX t1_idx_00000062 (b>?)
+}
+
+do_setup_rec_test $tn.3 {
+ CREATE TABLE t1(a, b, c);
+} {
+ SELECT * FROM t1 WHERE b COLLATE nocase BETWEEN ? AND ?
+} {
+ CREATE INDEX t1_idx_3e094c27 ON t1(b COLLATE NOCASE);
+ SEARCH t1 USING INDEX t1_idx_3e094c27 (b>? AND b)
+}
+
+do_setup_rec_test $tn.4 {
+ CREATE TABLE t1(a, b, c);
+} {
+ SELECT a FROM t1 ORDER BY b;
+} {
+ CREATE INDEX t1_idx_00000062 ON t1(b);
+ SCAN t1 USING INDEX t1_idx_00000062
+}
+
+do_setup_rec_test $tn.5 {
+ CREATE TABLE t1(a, b, c);
+} {
+ SELECT a FROM t1 WHERE a=? ORDER BY b;
+} {
+ CREATE INDEX t1_idx_000123a7 ON t1(a, b);
+ SEARCH t1 USING COVERING INDEX t1_idx_000123a7 (a=?)
+}
+
+if 0 {
+do_setup_rec_test $tn.6 {
+ CREATE TABLE t1(a, b, c);
+} {
+ SELECT min(a) FROM t1
+} {
+ CREATE INDEX t1_idx_00000061 ON t1(a);
+ SEARCH t1 USING COVERING INDEX t1_idx_00000061
+}
+}
+
+do_setup_rec_test $tn.7 {
+ CREATE TABLE t1(a, b, c);
+} {
+ SELECT * FROM t1 ORDER BY a, b, c;
+} {
+ CREATE INDEX t1_idx_033e95fe ON t1(a, b, c);
+ SCAN t1 USING COVERING INDEX t1_idx_033e95fe
+}
+
+#do_setup_rec_test $tn.1.8 {
+# CREATE TABLE t1(a, b, c);
+#} {
+# SELECT * FROM t1 ORDER BY a ASC, b COLLATE nocase DESC, c ASC;
+#} {
+# CREATE INDEX t1_idx_5be6e222 ON t1(a, b COLLATE NOCASE DESC, c);
+# 0|0|0|SCAN t1 USING COVERING INDEX t1_idx_5be6e222
+#}
+
+do_setup_rec_test $tn.8.1 {
+ CREATE TABLE t1(a COLLATE NOCase, b, c);
+} {
+ SELECT * FROM t1 WHERE a=?
+} {
+ CREATE INDEX t1_idx_00000061 ON t1(a);
+ SEARCH t1 USING INDEX t1_idx_00000061 (a=?)
+}
+do_setup_rec_test $tn.8.2 {
+ CREATE TABLE t1(a, b COLLATE nocase, c);
+} {
+ SELECT * FROM t1 ORDER BY a ASC, b DESC, c ASC;
+} {
+ CREATE INDEX t1_idx_5cb97285 ON t1(a, b DESC, c);
+ SCAN t1 USING COVERING INDEX t1_idx_5cb97285
+}
+
+
+# Tables with names that require quotes.
+#
+do_setup_rec_test $tn.9.1 {
+ CREATE TABLE "t t"(a, b, c);
+} {
+ SELECT * FROM "t t" WHERE a=?
+} {
+ CREATE INDEX "t t_idx_00000061" ON "t t"(a);
+ SEARCH t t USING INDEX t t_idx_00000061 (a=?)
+}
+
+do_setup_rec_test $tn.9.2 {
+ CREATE TABLE "t t"(a, b, c);
+} {
+ SELECT * FROM "t t" WHERE b BETWEEN ? AND ?
+} {
+ CREATE INDEX "t t_idx_00000062" ON "t t"(b);
+ SEARCH t t USING INDEX t t_idx_00000062 (b>? AND b)
+}
+
+# Columns with names that require quotes.
+#
+do_setup_rec_test $tn.10.1 {
+ CREATE TABLE t3(a, "b b", c);
+} {
+ SELECT * FROM t3 WHERE "b b" = ?
+} {
+ CREATE INDEX t3_idx_00050c52 ON t3('b b');
+ SEARCH t3 USING INDEX t3_idx_00050c52 (b b=?)
+}
+
+do_setup_rec_test $tn.10.2 {
+ CREATE TABLE t3(a, "b b", c);
+} {
+ SELECT * FROM t3 ORDER BY "b b"
+} {
+ CREATE INDEX t3_idx_00050c52 ON t3('b b');
+ SCAN t3 USING INDEX t3_idx_00050c52
+}
+
+# Transitive constraints
+#
+do_setup_rec_test $tn.11.1 {
+ CREATE TABLE t5(a, b);
+ CREATE TABLE t6(c, d);
+} {
+ SELECT * FROM t5, t6 WHERE a=? AND b=c AND c=?
+} {
+ CREATE INDEX t5_idx_000123a7 ON t5(a, b);
+ CREATE INDEX t6_idx_00000063 ON t6(c);
+ SEARCH t6 USING INDEX t6_idx_00000063 (c=?)
+ SEARCH t5 USING COVERING INDEX t5_idx_000123a7 (a=? AND b=?)
+}
+
+# OR terms.
+#
+do_setup_rec_test $tn.12.1 {
+ CREATE TABLE t7(a, b);
+} {
+ SELECT * FROM t7 WHERE a=? OR b=?
+} {
+ CREATE INDEX t7_idx_00000062 ON t7(b);
+ CREATE INDEX t7_idx_00000061 ON t7(a);
+ MULTI-INDEX OR
+ INDEX 1
+ SEARCH t7 USING INDEX t7_idx_00000061 (a=?)
+ INDEX 2
+ SEARCH t7 USING INDEX t7_idx_00000062 (b=?)
+}
+
+# rowid terms.
+#
+do_setup_rec_test $tn.13.1 {
+ CREATE TABLE t8(a, b);
+} {
+ SELECT * FROM t8 WHERE rowid=?
+} {
+ (no new indexes)
+ SEARCH t8 USING INTEGER PRIMARY KEY (rowid=?)
+}
+do_setup_rec_test $tn.13.2 {
+ CREATE TABLE t8(a, b);
+} {
+ SELECT * FROM t8 ORDER BY rowid
+} {
+ (no new indexes)
+ SCAN t8
+}
+do_setup_rec_test $tn.13.3 {
+ CREATE TABLE t8(a, b);
+} {
+ SELECT * FROM t8 WHERE a=? ORDER BY rowid
+} {
+ CREATE INDEX t8_idx_00000061 ON t8(a);
+ SEARCH t8 USING INDEX t8_idx_00000061 (a=?)
+}
+
+# Triggers
+#
+do_setup_rec_test $tn.14 {
+ CREATE TABLE t9(a, b, c);
+ CREATE TABLE t10(a, b, c);
+ CREATE TRIGGER t9t AFTER INSERT ON t9 BEGIN
+ UPDATE t10 SET a=new.a WHERE b = new.b;
+ END;
+} {
+ INSERT INTO t9 VALUES(?, ?, ?);
+} {
+ CREATE INDEX t10_idx_00000062 ON t10(b);
+ SEARCH t10 USING INDEX t10_idx_00000062 (b=?)
+}
+
+do_setup_rec_test $tn.15 {
+ CREATE TABLE t1(a, b);
+ CREATE TABLE t2(c, d);
+
+ WITH s(i) AS ( VALUES(1) UNION ALL SELECT i+1 FROM s WHERE i<100)
+ INSERT INTO t1 SELECT (i-1)/50, (i-1)/20 FROM s;
+
+ WITH s(i) AS ( VALUES(1) UNION ALL SELECT i+1 FROM s WHERE i<100)
+ INSERT INTO t2 SELECT (i-1)/20, (i-1)/5 FROM s;
+} {
+ SELECT * FROM t2, t1 WHERE b=? AND d=? AND t2.rowid=t1.rowid
+} {
+ CREATE INDEX t2_idx_00000064 ON t2(d);
+ SEARCH t2 USING INDEX t2_idx_00000064 (d=?)
+ SEARCH t1 USING INTEGER PRIMARY KEY (rowid=?)
+}
+
+do_setup_rec_test $tn.16 {
+ CREATE TABLE t1(a, b);
+} {
+ SELECT * FROM t1 WHERE b IS NOT NULL;
+} {
+ (no new indexes)
+ SCAN t1
+}
+
+do_setup_rec_test $tn.17.1 {
+ CREATE TABLE example (A INTEGER, B INTEGER, C INTEGER, PRIMARY KEY (A,B));
+} {
+ SELECT * FROM example WHERE a=?
+} {
+ (no new indexes)
+ SEARCH example USING INDEX sqlite_autoindex_example_1 (A=?)
+}
+do_setup_rec_test $tn.17.2 {
+ CREATE TABLE example (A INTEGER, B INTEGER, C INTEGER, PRIMARY KEY (A,B));
+} {
+ SELECT * FROM example WHERE b=?
+} {
+ CREATE INDEX example_idx_00000042 ON example(B);
+ SEARCH example USING INDEX example_idx_00000042 (B=?)
+}
+do_setup_rec_test $tn.17.3 {
+ CREATE TABLE example (A INTEGER, B INTEGER, C INTEGER, PRIMARY KEY (A,B));
+} {
+ SELECT * FROM example WHERE a=? AND b=?
+} {
+ (no new indexes)
+ SEARCH example USING INDEX sqlite_autoindex_example_1 (A=? AND B=?)
+}
+do_setup_rec_test $tn.17.4 {
+ CREATE TABLE example (A INTEGER, B INTEGER, C INTEGER, PRIMARY KEY (A,B));
+} {
+ SELECT * FROM example WHERE a=? AND b>?
+} {
+ (no new indexes)
+ SEARCH example USING INDEX sqlite_autoindex_example_1 (A=? AND B>?)
+}
+do_setup_rec_test $tn.17.5 {
+ CREATE TABLE example (A INTEGER, B INTEGER, C INTEGER, PRIMARY KEY (A,B));
+} {
+ SELECT * FROM example WHERE a>? AND b=?
+} {
+ CREATE INDEX example_idx_0000cb3f ON example(B, A);
+ SEARCH example USING INDEX example_idx_0000cb3f (B=? AND A>?)
+}
+
+do_setup_rec_test $tn.18.0 {
+ CREATE TABLE SomeObject (
+ a INTEGER PRIMARY KEY,
+ x TEXT GENERATED ALWAYS AS(HEX(a)) VIRTUAL
+ );
+} {
+ SELECT x FROM SomeObject;
+} {
+ (no new indexes)
+ SCAN SomeObject
+}
+do_setup_rec_test $tn.18.1 {
+ CREATE TABLE SomeObject (
+ a INTEGER PRIMARY KEY,
+ x TEXT GENERATED ALWAYS AS(HEX(a)) VIRTUAL
+ );
+} {
+ SELECT * FROM SomeObject WHERE x=?;
+} {
+ CREATE INDEX SomeObject_idx_00000078 ON SomeObject(x);
+ SEARCH SomeObject USING COVERING INDEX SomeObject_idx_00000078 (x=?)
+}
+
+
+do_setup_rec_test $tn.19.0 {
+ CREATE TABLE t1("index");
+} {
+ SELECT * FROM t1 ORDER BY "index";
+} {
+ CREATE INDEX t1_idx_01a7214e ON t1('index');
+ SCAN t1 USING COVERING INDEX t1_idx_01a7214e
+}
+
+ifcapable fts5 {
+ do_setup_rec_test $tn.20.0 {
+ CREATE VIRTUAL TABLE ft USING fts5(a);
+ CREATE TABLE t1(x, y);
+ } {
+ SELECT * FROM ft, t1 WHERE a=x
+ } {
+ CREATE INDEX t1_idx_00000078 ON t1(x);
+ SCAN ft VIRTUAL TABLE INDEX 0:
+ SEARCH t1 USING INDEX t1_idx_00000078 (x=?)
+ }
+}
+
+}
+
+proc do_candidates_test {tn sql res} {
+ set res [squish [string trim $res]]
+
+ set expert [sqlite3_expert_new db]
+ $expert sql $sql
+ $expert analyze
+
+ set candidates [squish [string trim [$expert report 0 candidates]]]
+ $expert destroy
+
+ uplevel [list do_test $tn [list set {} $candidates] $res]
+}
+
+
+reset_db
+do_execsql_test 5.0 {
+ CREATE TABLE t1(a, b);
+ CREATE TABLE t2(c, d);
+
+ WITH s(i) AS ( VALUES(1) UNION ALL SELECT i+1 FROM s WHERE i<100)
+ INSERT INTO t1 SELECT (i-1)/50, (i-1)/20 FROM s;
+
+ WITH s(i) AS ( VALUES(1) UNION ALL SELECT i+1 FROM s WHERE i<100)
+ INSERT INTO t2 SELECT (i-1)/20, (i-1)/5 FROM s;
+
+ CREATE INDEX i1 ON t1( lower(a) );
+}
+do_candidates_test 5.1 {
+ SELECT * FROM t1,t2 WHERE (b=? OR a=?) AND (c=? OR d=?)
+} {
+ CREATE INDEX t1_idx_00000062 ON t1(b); -- stat1: 100 20
+ CREATE INDEX t1_idx_00000061 ON t1(a); -- stat1: 100 50
+ CREATE INDEX t2_idx_00000063 ON t2(c); -- stat1: 100 20
+ CREATE INDEX t2_idx_00000064 ON t2(d); -- stat1: 100 5
+}
+
+do_candidates_test 5.2 {
+ SELECT * FROM t1,t2 WHERE a=? AND b=? AND c=? AND d=?
+} {
+ CREATE INDEX t1_idx_000123a7 ON t1(a, b); -- stat1: 100 50 17
+ CREATE INDEX t2_idx_0001295b ON t2(c, d); -- stat1: 100 20 5
+}
+
+do_execsql_test 5.3 {
+ CREATE INDEX t1_idx_00000061 ON t1(a); -- stat1: 100 50
+ CREATE INDEX t1_idx_00000062 ON t1(b); -- stat1: 100 20
+ CREATE INDEX t1_idx_000123a7 ON t1(a, b); -- stat1: 100 50 16
+
+ CREATE INDEX t2_idx_00000063 ON t2(c); -- stat1: 100 20
+ CREATE INDEX t2_idx_00000064 ON t2(d); -- stat1: 100 5
+ CREATE INDEX t2_idx_0001295b ON t2(c, d); -- stat1: 100 20 5
+
+ ANALYZE;
+ SELECT * FROM sqlite_stat1 ORDER BY 1, 2;
+} {
+ t1 i1 {100 50}
+ t1 t1_idx_00000061 {100 50}
+ t1 t1_idx_00000062 {100 20}
+ t1 t1_idx_000123a7 {100 50 17}
+ t2 t2_idx_00000063 {100 20}
+ t2 t2_idx_00000064 {100 5}
+ t2 t2_idx_0001295b {100 20 5}
+}
+
+do_catchsql_test 5.4 {
+ SELECT sqlite_expert_rem(123, 123);
+} {1 {no such function: sqlite_expert_rem}}
+do_catchsql_test 5.5 {
+ SELECT sqlite_expert_sample();
+} {1 {no such function: sqlite_expert_sample}}
+
+if 0 {
+do_test expert1-6.0 {
+ catchcmd :memory: {
+.expert
+select base64('');
+.expert
+select name from pragma_collation_list order by name collate uint;
+}
+} {0 {(no new indexes)
+
+SCAN CONSTANT ROW
+
+(no new indexes)
+
+SCAN pragma_collation_list VIRTUAL TABLE INDEX 0:
+USE TEMP B-TREE FOR ORDER BY
+}}
+}
+
+do_execsql_test 6.0 {
+ CREATE TABLE x1(a, b, c, d);
+ CREATE INDEX x1ab ON x1(a, lower(b));
+ CREATE INDEX x1dcba ON x1(d, b+c, a);
+}
+
+do_candidates_test 6.1 {
+ SELECT * FROM x1 WHERE b=? ORDER BY a;
+} {
+ CREATE INDEX x1_idx_0001267f ON x1(b, a);
+ CREATE INDEX x1_idx_00000062 ON x1(b);
+}
+
+#-------------------------------------------------------------------------
+ifcapable fts5 {
+ reset_db
+ do_execsql_test 7.0 {
+ CREATE VIRTUAL TABLE ft USING fts5(a);
+ CREATE TABLE t1(x, y);
+ }
+
+ do_candidates_test 7.1 {
+ SELECT * FROM ft, t1 WHERE a=x
+ } {
+ CREATE INDEX t1_idx_00000078 ON t1(x);
+ }
+
+ register_tcl_module db
+ proc vtab_command {method args} {
+ global G
+
+ switch -- $method {
+ xConnect {
+ return "CREATE TABLE t1(a, b, c);"
+ }
+
+ xBestIndex {
+ return [list]
+ }
+
+ xFilter {
+ return [list sql "SELECT rowid, * FROM t0"]
+ }
+ }
+
+ return {}
+ }
+
+ do_execsql_test 7.2 {
+ CREATE TABLE t0(a, b, c);
+ INSERT INTO t0 VALUES(1, 2, 3), (11, 22, 33);
+ CREATE VIRTUAL TABLE t2 USING tcl(vtab_command);
+ }
+
+ do_execsql_test 7.3 {
+ SELECT * FROM t2
+ } {
+ 1 2 3
+ 11 22 33
+ }
+
+ do_candidates_test 7.4 {
+ SELECT * FROM ft, t1 WHERE a=x
+ } {
+ CREATE INDEX t1_idx_00000078 ON t1(x);
+ }
+
+ do_test 7.5 {
+ set expert [sqlite3_expert_new db]
+ list [catch { $expert sql "SELECT * FROM ft, t2 WHERE b=1" } msg] $msg
+ } {1 {no such table: t2}}
+ $expert destroy
+
+ reset_db
+ do_execsql_test 7.6 {
+ BEGIN TRANSACTION;
+ CREATE TABLE IF NOT EXISTS 'bfts_idx_data'(id INTEGER PRIMARY KEY, block BLOB);
+ CREATE TABLE IF NOT EXISTS 'fts_idx_data'(id INTEGER PRIMARY KEY, block BLOB);
+ INSERT INTO fts_idx_data VALUES(1,X'');
+ INSERT INTO fts_idx_data VALUES(10,X'00000000ff000001000000');
+ CREATE TABLE IF NOT EXISTS 'fts_idx_idx'(segid, term, pgno, PRIMARY KEY(segid, term)) WITHOUT ROWID;
+ CREATE TABLE IF NOT EXISTS 'fts_idx_docsize'(id INTEGER PRIMARY KEY, sz BLOB, origin INTEGER);
+ CREATE TABLE IF NOT EXISTS 'fts_idx_config'(k PRIMARY KEY, v) WITHOUT ROWID;
+ INSERT INTO fts_idx_config VALUES('version',4);
+ PRAGMA writable_schema=ON;
+ INSERT INTO sqlite_schema(type,name,tbl_name,rootpage,sql)VALUES('table','fts_idx','fts_idx',0,'CREATE VIRTUAL TABLE fts_idx USING fts5(Title, Description, Channel, Tags, content='''', contentless_delete=1)');
+
+ CREATE TABLE f(x BLOB, y);
+ COMMIT;
+ PRAGMA writable_schema = RESET;
+ }
+
+ do_candidates_test 7.4 {
+ SELECT * FROM fts_idx, f WHERE x = fts_idx.Channel
+ } {
+ CREATE INDEX f_idx_00000078 ON f(x);
+ }
+}
+
+finish_test
diff --git a/ext/expert/sqlite3expert.c b/ext/expert/sqlite3expert.c
new file mode 100644
index 0000000000000000000000000000000000000000..93693cfae99169b92cf924b4d9b8357a09044d61
--- /dev/null
+++ b/ext/expert/sqlite3expert.c
@@ -0,0 +1,2234 @@
+/*
+** 2017 April 09
+**
+** The author disclaims copyright to this source code. In place of
+** a legal notice, here is a blessing:
+**
+** May you do good and not evil.
+** May you find forgiveness for yourself and forgive others.
+** May you share freely, never taking more than you give.
+**
+*************************************************************************
+*/
+#include "sqlite3expert.h"
+#include
+#include
+#include
+
+#if !defined(SQLITE_AMALGAMATION)
+#if defined(SQLITE_COVERAGE_TEST) || defined(SQLITE_MUTATION_TEST)
+# define SQLITE_OMIT_AUXILIARY_SAFETY_CHECKS 1
+#endif
+#if defined(SQLITE_OMIT_AUXILIARY_SAFETY_CHECKS)
+# define ALWAYS(X) (1)
+# define NEVER(X) (0)
+#elif !defined(NDEBUG)
+# define ALWAYS(X) ((X)?1:(assert(0),0))
+# define NEVER(X) ((X)?(assert(0),1):0)
+#else
+# define ALWAYS(X) (X)
+# define NEVER(X) (X)
+#endif
+#endif /* !defined(SQLITE_AMALGAMATION) */
+
+
+#ifndef SQLITE_OMIT_VIRTUALTABLE
+
+typedef sqlite3_int64 i64;
+typedef sqlite3_uint64 u64;
+
+typedef struct IdxColumn IdxColumn;
+typedef struct IdxConstraint IdxConstraint;
+typedef struct IdxScan IdxScan;
+typedef struct IdxStatement IdxStatement;
+typedef struct IdxTable IdxTable;
+typedef struct IdxWrite IdxWrite;
+
+#define STRLEN (int)strlen
+
+/*
+** A temp table name that we assume no user database will actually use.
+** If this assumption proves incorrect triggers on the table with the
+** conflicting name will be ignored.
+*/
+#define UNIQUE_TABLE_NAME "t592690916721053953805701627921227776"
+
+/*
+** A single constraint. Equivalent to either "col = ?" or "col < ?" (or
+** any other type of single-ended range constraint on a column).
+**
+** pLink:
+** Used to temporarily link IdxConstraint objects into lists while
+** creating candidate indexes.
+*/
+struct IdxConstraint {
+ char *zColl; /* Collation sequence */
+ int bRange; /* True for range, false for eq */
+ int iCol; /* Constrained table column */
+ int bFlag; /* Used by idxFindCompatible() */
+ int bDesc; /* True if ORDER BY DESC */
+ IdxConstraint *pNext; /* Next constraint in pEq or pRange list */
+ IdxConstraint *pLink; /* See above */
+};
+
+/*
+** A single scan of a single table.
+*/
+struct IdxScan {
+ IdxTable *pTab; /* Associated table object */
+ int iDb; /* Database containing table zTable */
+ i64 covering; /* Mask of columns required for cov. index */
+ IdxConstraint *pOrder; /* ORDER BY columns */
+ IdxConstraint *pEq; /* List of == constraints */
+ IdxConstraint *pRange; /* List of < constraints */
+ IdxScan *pNextScan; /* Next IdxScan object for same analysis */
+};
+
+/*
+** Information regarding a single database table. Extracted from
+** "PRAGMA table_info" by function idxGetTableInfo().
+*/
+struct IdxColumn {
+ char *zName;
+ char *zColl;
+ int iPk;
+};
+struct IdxTable {
+ int nCol;
+ char *zName; /* Table name */
+ IdxColumn *aCol;
+ IdxTable *pNext; /* Next table in linked list of all tables */
+};
+
+/*
+** An object of the following type is created for each unique table/write-op
+** seen. The objects are stored in a singly-linked list beginning at
+** sqlite3expert.pWrite.
+*/
+struct IdxWrite {
+ IdxTable *pTab;
+ int eOp; /* SQLITE_UPDATE, DELETE or INSERT */
+ IdxWrite *pNext;
+};
+
+/*
+** Each statement being analyzed is represented by an instance of this
+** structure.
+*/
+struct IdxStatement {
+ int iId; /* Statement number */
+ char *zSql; /* SQL statement */
+ char *zIdx; /* Indexes */
+ char *zEQP; /* Plan */
+ IdxStatement *pNext;
+};
+
+
+/*
+** A hash table for storing strings. With space for a payload string
+** with each entry. Methods are:
+**
+** idxHashInit()
+** idxHashClear()
+** idxHashAdd()
+** idxHashSearch()
+*/
+#define IDX_HASH_SIZE 1023
+typedef struct IdxHashEntry IdxHashEntry;
+typedef struct IdxHash IdxHash;
+struct IdxHashEntry {
+ char *zKey; /* nul-terminated key */
+ char *zVal; /* nul-terminated value string */
+ char *zVal2; /* nul-terminated value string 2 */
+ IdxHashEntry *pHashNext; /* Next entry in same hash bucket */
+ IdxHashEntry *pNext; /* Next entry in hash */
+};
+struct IdxHash {
+ IdxHashEntry *pFirst;
+ IdxHashEntry *aHash[IDX_HASH_SIZE];
+};
+
+/*
+** sqlite3expert object.
+*/
+struct sqlite3expert {
+ int iSample; /* Percentage of tables to sample for stat1 */
+ sqlite3 *db; /* User database */
+ sqlite3 *dbm; /* In-memory db for this analysis */
+ sqlite3 *dbv; /* Vtab schema for this analysis */
+ IdxTable *pTable; /* List of all IdxTable objects */
+ IdxScan *pScan; /* List of scan objects */
+ IdxWrite *pWrite; /* List of write objects */
+ IdxStatement *pStatement; /* List of IdxStatement objects */
+ int bRun; /* True once analysis has run */
+ char **pzErrmsg;
+ int rc; /* Error code from whereinfo hook */
+ IdxHash hIdx; /* Hash containing all candidate indexes */
+ char *zCandidates; /* For EXPERT_REPORT_CANDIDATES */
+};
+
+
+/*
+** Allocate and return nByte bytes of zeroed memory using sqlite3_malloc().
+** If the allocation fails, set *pRc to SQLITE_NOMEM and return NULL.
+*/
+static void *idxMalloc(int *pRc, int nByte){
+ void *pRet;
+ assert( *pRc==SQLITE_OK );
+ assert( nByte>0 );
+ pRet = sqlite3_malloc(nByte);
+ if( pRet ){
+ memset(pRet, 0, nByte);
+ }else{
+ *pRc = SQLITE_NOMEM;
+ }
+ return pRet;
+}
+
+/*
+** Initialize an IdxHash hash table.
+*/
+static void idxHashInit(IdxHash *pHash){
+ memset(pHash, 0, sizeof(IdxHash));
+}
+
+/*
+** Reset an IdxHash hash table.
+*/
+static void idxHashClear(IdxHash *pHash){
+ int i;
+ for(i=0; iaHash[i]; pEntry; pEntry=pNext){
+ pNext = pEntry->pHashNext;
+ sqlite3_free(pEntry->zVal2);
+ sqlite3_free(pEntry);
+ }
+ }
+ memset(pHash, 0, sizeof(IdxHash));
+}
+
+/*
+** Return the index of the hash bucket that the string specified by the
+** arguments to this function belongs.
+*/
+static int idxHashString(const char *z, int n){
+ unsigned int ret = 0;
+ int i;
+ for(i=0; i=0 );
+ for(pEntry=pHash->aHash[iHash]; pEntry; pEntry=pEntry->pHashNext){
+ if( STRLEN(pEntry->zKey)==nKey && 0==memcmp(pEntry->zKey, zKey, nKey) ){
+ return 1;
+ }
+ }
+ pEntry = idxMalloc(pRc, sizeof(IdxHashEntry) + nKey+1 + nVal+1);
+ if( pEntry ){
+ pEntry->zKey = (char*)&pEntry[1];
+ memcpy(pEntry->zKey, zKey, nKey);
+ if( zVal ){
+ pEntry->zVal = &pEntry->zKey[nKey+1];
+ memcpy(pEntry->zVal, zVal, nVal);
+ }
+ pEntry->pHashNext = pHash->aHash[iHash];
+ pHash->aHash[iHash] = pEntry;
+
+ pEntry->pNext = pHash->pFirst;
+ pHash->pFirst = pEntry;
+ }
+ return 0;
+}
+
+/*
+** If zKey/nKey is present in the hash table, return a pointer to the
+** hash-entry object.
+*/
+static IdxHashEntry *idxHashFind(IdxHash *pHash, const char *zKey, int nKey){
+ int iHash;
+ IdxHashEntry *pEntry;
+ if( nKey<0 ) nKey = STRLEN(zKey);
+ iHash = idxHashString(zKey, nKey);
+ assert( iHash>=0 );
+ for(pEntry=pHash->aHash[iHash]; pEntry; pEntry=pEntry->pHashNext){
+ if( STRLEN(pEntry->zKey)==nKey && 0==memcmp(pEntry->zKey, zKey, nKey) ){
+ return pEntry;
+ }
+ }
+ return 0;
+}
+
+/*
+** If the hash table contains an entry with a key equal to the string
+** passed as the final two arguments to this function, return a pointer
+** to the payload string. Otherwise, if zKey/nKey is not present in the
+** hash table, return NULL.
+*/
+static const char *idxHashSearch(IdxHash *pHash, const char *zKey, int nKey){
+ IdxHashEntry *pEntry = idxHashFind(pHash, zKey, nKey);
+ if( pEntry ) return pEntry->zVal;
+ return 0;
+}
+
+/*
+** Allocate and return a new IdxConstraint object. Set the IdxConstraint.zColl
+** variable to point to a copy of nul-terminated string zColl.
+*/
+static IdxConstraint *idxNewConstraint(int *pRc, const char *zColl){
+ IdxConstraint *pNew;
+ int nColl = STRLEN(zColl);
+
+ assert( *pRc==SQLITE_OK );
+ pNew = (IdxConstraint*)idxMalloc(pRc, sizeof(IdxConstraint) * nColl + 1);
+ if( pNew ){
+ pNew->zColl = (char*)&pNew[1];
+ memcpy(pNew->zColl, zColl, nColl+1);
+ }
+ return pNew;
+}
+
+/*
+** An error associated with database handle db has just occurred. Pass
+** the error message to callback function xOut.
+*/
+static void idxDatabaseError(
+ sqlite3 *db, /* Database handle */
+ char **pzErrmsg /* Write error here */
+){
+ *pzErrmsg = sqlite3_mprintf("%s", sqlite3_errmsg(db));
+}
+
+/*
+** Prepare an SQL statement.
+*/
+static int idxPrepareStmt(
+ sqlite3 *db, /* Database handle to compile against */
+ sqlite3_stmt **ppStmt, /* OUT: Compiled SQL statement */
+ char **pzErrmsg, /* OUT: sqlite3_malloc()ed error message */
+ const char *zSql /* SQL statement to compile */
+){
+ int rc = sqlite3_prepare_v2(db, zSql, -1, ppStmt, 0);
+ if( rc!=SQLITE_OK ){
+ *ppStmt = 0;
+ idxDatabaseError(db, pzErrmsg);
+ }
+ return rc;
+}
+
+/*
+** Prepare an SQL statement using the results of a printf() formatting.
+*/
+static int idxPrintfPrepareStmt(
+ sqlite3 *db, /* Database handle to compile against */
+ sqlite3_stmt **ppStmt, /* OUT: Compiled SQL statement */
+ char **pzErrmsg, /* OUT: sqlite3_malloc()ed error message */
+ const char *zFmt, /* printf() format of SQL statement */
+ ... /* Trailing printf() arguments */
+){
+ va_list ap;
+ int rc;
+ char *zSql;
+ va_start(ap, zFmt);
+ zSql = sqlite3_vmprintf(zFmt, ap);
+ if( zSql==0 ){
+ rc = SQLITE_NOMEM;
+ }else{
+ rc = idxPrepareStmt(db, ppStmt, pzErrmsg, zSql);
+ sqlite3_free(zSql);
+ }
+ va_end(ap);
+ return rc;
+}
+
+
+/*************************************************************************
+** Beginning of virtual table implementation.
+*/
+typedef struct ExpertVtab ExpertVtab;
+struct ExpertVtab {
+ sqlite3_vtab base;
+ IdxTable *pTab;
+ sqlite3expert *pExpert;
+};
+
+typedef struct ExpertCsr ExpertCsr;
+struct ExpertCsr {
+ sqlite3_vtab_cursor base;
+ sqlite3_stmt *pData;
+};
+
+static char *expertDequote(const char *zIn){
+ int n = STRLEN(zIn);
+ char *zRet = sqlite3_malloc(n);
+
+ assert( zIn[0]=='\'' );
+ assert( zIn[n-1]=='\'' );
+
+ if( zRet ){
+ int iOut = 0;
+ int iIn = 0;
+ for(iIn=1; iIn<(n-1); iIn++){
+ if( zIn[iIn]=='\'' ){
+ assert( zIn[iIn+1]=='\'' );
+ iIn++;
+ }
+ zRet[iOut++] = zIn[iIn];
+ }
+ zRet[iOut] = '\0';
+ }
+
+ return zRet;
+}
+
+/*
+** This function is the implementation of both the xConnect and xCreate
+** methods of the r-tree virtual table.
+**
+** argv[0] -> module name
+** argv[1] -> database name
+** argv[2] -> table name
+** argv[...] -> column names...
+*/
+static int expertConnect(
+ sqlite3 *db,
+ void *pAux,
+ int argc, const char *const*argv,
+ sqlite3_vtab **ppVtab,
+ char **pzErr
+){
+ sqlite3expert *pExpert = (sqlite3expert*)pAux;
+ ExpertVtab *p = 0;
+ int rc;
+
+ if( argc!=4 ){
+ *pzErr = sqlite3_mprintf("internal error!");
+ rc = SQLITE_ERROR;
+ }else{
+ char *zCreateTable = expertDequote(argv[3]);
+ if( zCreateTable ){
+ rc = sqlite3_declare_vtab(db, zCreateTable);
+ if( rc==SQLITE_OK ){
+ p = idxMalloc(&rc, sizeof(ExpertVtab));
+ }
+ if( rc==SQLITE_OK ){
+ p->pExpert = pExpert;
+ p->pTab = pExpert->pTable;
+ assert( sqlite3_stricmp(p->pTab->zName, argv[2])==0 );
+ }
+ sqlite3_free(zCreateTable);
+ }else{
+ rc = SQLITE_NOMEM;
+ }
+ }
+
+ *ppVtab = (sqlite3_vtab*)p;
+ return rc;
+}
+
+static int expertDisconnect(sqlite3_vtab *pVtab){
+ ExpertVtab *p = (ExpertVtab*)pVtab;
+ sqlite3_free(p);
+ return SQLITE_OK;
+}
+
+static int expertBestIndex(sqlite3_vtab *pVtab, sqlite3_index_info *pIdxInfo){
+ ExpertVtab *p = (ExpertVtab*)pVtab;
+ int rc = SQLITE_OK;
+ int n = 0;
+ IdxScan *pScan;
+ const int opmask =
+ SQLITE_INDEX_CONSTRAINT_EQ | SQLITE_INDEX_CONSTRAINT_GT |
+ SQLITE_INDEX_CONSTRAINT_LT | SQLITE_INDEX_CONSTRAINT_GE |
+ SQLITE_INDEX_CONSTRAINT_LE;
+
+ pScan = idxMalloc(&rc, sizeof(IdxScan));
+ if( pScan ){
+ int i;
+
+ /* Link the new scan object into the list */
+ pScan->pTab = p->pTab;
+ pScan->pNextScan = p->pExpert->pScan;
+ p->pExpert->pScan = pScan;
+
+ /* Add the constraints to the IdxScan object */
+ for(i=0; inConstraint; i++){
+ struct sqlite3_index_constraint *pCons = &pIdxInfo->aConstraint[i];
+ if( pCons->usable
+ && pCons->iColumn>=0
+ && p->pTab->aCol[pCons->iColumn].iPk==0
+ && (pCons->op & opmask)
+ ){
+ IdxConstraint *pNew;
+ const char *zColl = sqlite3_vtab_collation(pIdxInfo, i);
+ pNew = idxNewConstraint(&rc, zColl);
+ if( pNew ){
+ pNew->iCol = pCons->iColumn;
+ if( pCons->op==SQLITE_INDEX_CONSTRAINT_EQ ){
+ pNew->pNext = pScan->pEq;
+ pScan->pEq = pNew;
+ }else{
+ pNew->bRange = 1;
+ pNew->pNext = pScan->pRange;
+ pScan->pRange = pNew;
+ }
+ }
+ n++;
+ pIdxInfo->aConstraintUsage[i].argvIndex = n;
+ }
+ }
+
+ /* Add the ORDER BY to the IdxScan object */
+ for(i=pIdxInfo->nOrderBy-1; i>=0; i--){
+ int iCol = pIdxInfo->aOrderBy[i].iColumn;
+ if( iCol>=0 ){
+ IdxConstraint *pNew = idxNewConstraint(&rc, p->pTab->aCol[iCol].zColl);
+ if( pNew ){
+ pNew->iCol = iCol;
+ pNew->bDesc = pIdxInfo->aOrderBy[i].desc;
+ pNew->pNext = pScan->pOrder;
+ pNew->pLink = pScan->pOrder;
+ pScan->pOrder = pNew;
+ n++;
+ }
+ }
+ }
+ }
+
+ pIdxInfo->estimatedCost = 1000000.0 / (n+1);
+ return rc;
+}
+
+static int expertUpdate(
+ sqlite3_vtab *pVtab,
+ int nData,
+ sqlite3_value **azData,
+ sqlite_int64 *pRowid
+){
+ (void)pVtab;
+ (void)nData;
+ (void)azData;
+ (void)pRowid;
+ return SQLITE_OK;
+}
+
+/*
+** Virtual table module xOpen method.
+*/
+static int expertOpen(sqlite3_vtab *pVTab, sqlite3_vtab_cursor **ppCursor){
+ int rc = SQLITE_OK;
+ ExpertCsr *pCsr;
+ (void)pVTab;
+ pCsr = idxMalloc(&rc, sizeof(ExpertCsr));
+ *ppCursor = (sqlite3_vtab_cursor*)pCsr;
+ return rc;
+}
+
+/*
+** Virtual table module xClose method.
+*/
+static int expertClose(sqlite3_vtab_cursor *cur){
+ ExpertCsr *pCsr = (ExpertCsr*)cur;
+ sqlite3_finalize(pCsr->pData);
+ sqlite3_free(pCsr);
+ return SQLITE_OK;
+}
+
+/*
+** Virtual table module xEof method.
+**
+** Return non-zero if the cursor does not currently point to a valid
+** record (i.e if the scan has finished), or zero otherwise.
+*/
+static int expertEof(sqlite3_vtab_cursor *cur){
+ ExpertCsr *pCsr = (ExpertCsr*)cur;
+ return pCsr->pData==0;
+}
+
+/*
+** Virtual table module xNext method.
+*/
+static int expertNext(sqlite3_vtab_cursor *cur){
+ ExpertCsr *pCsr = (ExpertCsr*)cur;
+ int rc = SQLITE_OK;
+
+ assert( pCsr->pData );
+ rc = sqlite3_step(pCsr->pData);
+ if( rc!=SQLITE_ROW ){
+ rc = sqlite3_finalize(pCsr->pData);
+ pCsr->pData = 0;
+ }else{
+ rc = SQLITE_OK;
+ }
+
+ return rc;
+}
+
+/*
+** Virtual table module xRowid method.
+*/
+static int expertRowid(sqlite3_vtab_cursor *cur, sqlite_int64 *pRowid){
+ (void)cur;
+ *pRowid = 0;
+ return SQLITE_OK;
+}
+
+/*
+** Virtual table module xColumn method.
+*/
+static int expertColumn(sqlite3_vtab_cursor *cur, sqlite3_context *ctx, int i){
+ ExpertCsr *pCsr = (ExpertCsr*)cur;
+ sqlite3_value *pVal;
+ pVal = sqlite3_column_value(pCsr->pData, i);
+ if( pVal ){
+ sqlite3_result_value(ctx, pVal);
+ }
+ return SQLITE_OK;
+}
+
+/*
+** Virtual table module xFilter method.
+*/
+static int expertFilter(
+ sqlite3_vtab_cursor *cur,
+ int idxNum, const char *idxStr,
+ int argc, sqlite3_value **argv
+){
+ ExpertCsr *pCsr = (ExpertCsr*)cur;
+ ExpertVtab *pVtab = (ExpertVtab*)(cur->pVtab);
+ sqlite3expert *pExpert = pVtab->pExpert;
+ int rc;
+
+ (void)idxNum;
+ (void)idxStr;
+ (void)argc;
+ (void)argv;
+ rc = sqlite3_finalize(pCsr->pData);
+ pCsr->pData = 0;
+ if( rc==SQLITE_OK ){
+ rc = idxPrintfPrepareStmt(pExpert->db, &pCsr->pData, &pVtab->base.zErrMsg,
+ "SELECT * FROM main.%Q WHERE sqlite_expert_sample()", pVtab->pTab->zName
+ );
+ }
+
+ if( rc==SQLITE_OK ){
+ rc = expertNext(cur);
+ }
+ return rc;
+}
+
+static int idxRegisterVtab(sqlite3expert *p){
+ static sqlite3_module expertModule = {
+ 2, /* iVersion */
+ expertConnect, /* xCreate - create a table */
+ expertConnect, /* xConnect - connect to an existing table */
+ expertBestIndex, /* xBestIndex - Determine search strategy */
+ expertDisconnect, /* xDisconnect - Disconnect from a table */
+ expertDisconnect, /* xDestroy - Drop a table */
+ expertOpen, /* xOpen - open a cursor */
+ expertClose, /* xClose - close a cursor */
+ expertFilter, /* xFilter - configure scan constraints */
+ expertNext, /* xNext - advance a cursor */
+ expertEof, /* xEof */
+ expertColumn, /* xColumn - read data */
+ expertRowid, /* xRowid - read data */
+ expertUpdate, /* xUpdate - write data */
+ 0, /* xBegin - begin transaction */
+ 0, /* xSync - sync transaction */
+ 0, /* xCommit - commit transaction */
+ 0, /* xRollback - rollback transaction */
+ 0, /* xFindFunction - function overloading */
+ 0, /* xRename - rename the table */
+ 0, /* xSavepoint */
+ 0, /* xRelease */
+ 0, /* xRollbackTo */
+ 0, /* xShadowName */
+ 0, /* xIntegrity */
+ };
+
+ return sqlite3_create_module(p->dbv, "expert", &expertModule, (void*)p);
+}
+/*
+** End of virtual table implementation.
+*************************************************************************/
+/*
+** Finalize SQL statement pStmt. If (*pRc) is SQLITE_OK when this function
+** is called, set it to the return value of sqlite3_finalize() before
+** returning. Otherwise, discard the sqlite3_finalize() return value.
+*/
+static void idxFinalize(int *pRc, sqlite3_stmt *pStmt){
+ int rc = sqlite3_finalize(pStmt);
+ if( *pRc==SQLITE_OK ) *pRc = rc;
+}
+
+/*
+** Attempt to allocate an IdxTable structure corresponding to table zTab
+** in the main database of connection db. If successful, set (*ppOut) to
+** point to the new object and return SQLITE_OK. Otherwise, return an
+** SQLite error code and set (*ppOut) to NULL. In this case *pzErrmsg may be
+** set to point to an error string.
+**
+** It is the responsibility of the caller to eventually free either the
+** IdxTable object or error message using sqlite3_free().
+*/
+static int idxGetTableInfo(
+ sqlite3 *db, /* Database connection to read details from */
+ const char *zTab, /* Table name */
+ IdxTable **ppOut, /* OUT: New object (if successful) */
+ char **pzErrmsg /* OUT: Error message (if not) */
+){
+ sqlite3_stmt *p1 = 0;
+ int nCol = 0;
+ int nTab;
+ int nByte;
+ IdxTable *pNew = 0;
+ int rc, rc2;
+ char *pCsr = 0;
+ int nPk = 0;
+
+ *ppOut = 0;
+ if( zTab==0 ) return SQLITE_ERROR;
+ nTab = STRLEN(zTab);
+ nByte = sizeof(IdxTable) + nTab + 1;
+ rc = idxPrintfPrepareStmt(db, &p1, pzErrmsg, "PRAGMA table_xinfo=%Q", zTab);
+ while( rc==SQLITE_OK && SQLITE_ROW==sqlite3_step(p1) ){
+ const char *zCol = (const char*)sqlite3_column_text(p1, 1);
+ const char *zColSeq = 0;
+ if( zCol==0 ){
+ rc = SQLITE_ERROR;
+ break;
+ }
+ nByte += 1 + STRLEN(zCol);
+ rc = sqlite3_table_column_metadata(
+ db, "main", zTab, zCol, 0, &zColSeq, 0, 0, 0
+ );
+ if( zColSeq==0 ) zColSeq = "binary";
+ nByte += 1 + STRLEN(zColSeq);
+ nCol++;
+ nPk += (sqlite3_column_int(p1, 5)>0);
+ }
+ rc2 = sqlite3_reset(p1);
+ if( rc==SQLITE_OK ) rc = rc2;
+
+ nByte += sizeof(IdxColumn) * nCol;
+ if( rc==SQLITE_OK ){
+ pNew = idxMalloc(&rc, nByte);
+ }
+ if( rc==SQLITE_OK ){
+ pNew->aCol = (IdxColumn*)&pNew[1];
+ pNew->nCol = nCol;
+ pCsr = (char*)&pNew->aCol[nCol];
+ }
+
+ nCol = 0;
+ while( rc==SQLITE_OK && SQLITE_ROW==sqlite3_step(p1) ){
+ const char *zCol = (const char*)sqlite3_column_text(p1, 1);
+ const char *zColSeq = 0;
+ int nCopy;
+ if( zCol==0 ) continue;
+ nCopy = STRLEN(zCol) + 1;
+ pNew->aCol[nCol].zName = pCsr;
+ pNew->aCol[nCol].iPk = (sqlite3_column_int(p1, 5)==1 && nPk==1);
+ memcpy(pCsr, zCol, nCopy);
+ pCsr += nCopy;
+
+ rc = sqlite3_table_column_metadata(
+ db, "main", zTab, zCol, 0, &zColSeq, 0, 0, 0
+ );
+ if( rc==SQLITE_OK ){
+ if( zColSeq==0 ) zColSeq = "binary";
+ nCopy = STRLEN(zColSeq) + 1;
+ pNew->aCol[nCol].zColl = pCsr;
+ memcpy(pCsr, zColSeq, nCopy);
+ pCsr += nCopy;
+ }
+
+ nCol++;
+ }
+ idxFinalize(&rc, p1);
+
+ if( rc!=SQLITE_OK ){
+ sqlite3_free(pNew);
+ pNew = 0;
+ }else if( ALWAYS(pNew!=0) ){
+ pNew->zName = pCsr;
+ if( ALWAYS(pNew->zName!=0) ) memcpy(pNew->zName, zTab, nTab+1);
+ }
+
+ *ppOut = pNew;
+ return rc;
+}
+
+/*
+** This function is a no-op if *pRc is set to anything other than
+** SQLITE_OK when it is called.
+**
+** If *pRc is initially set to SQLITE_OK, then the text specified by
+** the printf() style arguments is appended to zIn and the result returned
+** in a buffer allocated by sqlite3_malloc(). sqlite3_free() is called on
+** zIn before returning.
+*/
+static char *idxAppendText(int *pRc, char *zIn, const char *zFmt, ...){
+ va_list ap;
+ char *zAppend = 0;
+ char *zRet = 0;
+ int nIn = zIn ? STRLEN(zIn) : 0;
+ int nAppend = 0;
+ va_start(ap, zFmt);
+ if( *pRc==SQLITE_OK ){
+ zAppend = sqlite3_vmprintf(zFmt, ap);
+ if( zAppend ){
+ nAppend = STRLEN(zAppend);
+ zRet = (char*)sqlite3_malloc(nIn + nAppend + 1);
+ }
+ if( zAppend && zRet ){
+ if( nIn ) memcpy(zRet, zIn, nIn);
+ memcpy(&zRet[nIn], zAppend, nAppend+1);
+ }else{
+ sqlite3_free(zRet);
+ zRet = 0;
+ *pRc = SQLITE_NOMEM;
+ }
+ sqlite3_free(zAppend);
+ sqlite3_free(zIn);
+ }
+ va_end(ap);
+ return zRet;
+}
+
+/*
+** Return true if zId must be quoted in order to use it as an SQL
+** identifier, or false otherwise.
+*/
+static int idxIdentifierRequiresQuotes(const char *zId){
+ int i;
+ int nId = STRLEN(zId);
+
+ if( sqlite3_keyword_check(zId, nId) ) return 1;
+
+ for(i=0; zId[i]; i++){
+ if( !(zId[i]=='_')
+ && !(zId[i]>='0' && zId[i]<='9')
+ && !(zId[i]>='a' && zId[i]<='z')
+ && !(zId[i]>='A' && zId[i]<='Z')
+ ){
+ return 1;
+ }
+ }
+ return 0;
+}
+
+/*
+** This function appends an index column definition suitable for constraint
+** pCons to the string passed as zIn and returns the result.
+*/
+static char *idxAppendColDefn(
+ int *pRc, /* IN/OUT: Error code */
+ char *zIn, /* Column defn accumulated so far */
+ IdxTable *pTab, /* Table index will be created on */
+ IdxConstraint *pCons
+){
+ char *zRet = zIn;
+ IdxColumn *p = &pTab->aCol[pCons->iCol];
+ if( zRet ) zRet = idxAppendText(pRc, zRet, ", ");
+
+ if( idxIdentifierRequiresQuotes(p->zName) ){
+ zRet = idxAppendText(pRc, zRet, "%Q", p->zName);
+ }else{
+ zRet = idxAppendText(pRc, zRet, "%s", p->zName);
+ }
+
+ if( sqlite3_stricmp(p->zColl, pCons->zColl) ){
+ if( idxIdentifierRequiresQuotes(pCons->zColl) ){
+ zRet = idxAppendText(pRc, zRet, " COLLATE %Q", pCons->zColl);
+ }else{
+ zRet = idxAppendText(pRc, zRet, " COLLATE %s", pCons->zColl);
+ }
+ }
+
+ if( pCons->bDesc ){
+ zRet = idxAppendText(pRc, zRet, " DESC");
+ }
+ return zRet;
+}
+
+/*
+** Search database dbm for an index compatible with the one idxCreateFromCons()
+** would create from arguments pScan, pEq and pTail. If no error occurs and
+** such an index is found, return non-zero. Or, if no such index is found,
+** return zero.
+**
+** If an error occurs, set *pRc to an SQLite error code and return zero.
+*/
+static int idxFindCompatible(
+ int *pRc, /* OUT: Error code */
+ sqlite3* dbm, /* Database to search */
+ IdxScan *pScan, /* Scan for table to search for index on */
+ IdxConstraint *pEq, /* List of == constraints */
+ IdxConstraint *pTail /* List of range constraints */
+){
+ const char *zTbl = pScan->pTab->zName;
+ sqlite3_stmt *pIdxList = 0;
+ IdxConstraint *pIter;
+ int nEq = 0; /* Number of elements in pEq */
+ int rc;
+
+ /* Count the elements in list pEq */
+ for(pIter=pEq; pIter; pIter=pIter->pLink) nEq++;
+
+ rc = idxPrintfPrepareStmt(dbm, &pIdxList, 0, "PRAGMA index_list=%Q", zTbl);
+ while( rc==SQLITE_OK && sqlite3_step(pIdxList)==SQLITE_ROW ){
+ int bMatch = 1;
+ IdxConstraint *pT = pTail;
+ sqlite3_stmt *pInfo = 0;
+ const char *zIdx = (const char*)sqlite3_column_text(pIdxList, 1);
+ if( zIdx==0 ) continue;
+
+ /* Zero the IdxConstraint.bFlag values in the pEq list */
+ for(pIter=pEq; pIter; pIter=pIter->pLink) pIter->bFlag = 0;
+
+ rc = idxPrintfPrepareStmt(dbm, &pInfo, 0, "PRAGMA index_xInfo=%Q", zIdx);
+ while( rc==SQLITE_OK && sqlite3_step(pInfo)==SQLITE_ROW ){
+ int iIdx = sqlite3_column_int(pInfo, 0);
+ int iCol = sqlite3_column_int(pInfo, 1);
+ const char *zColl = (const char*)sqlite3_column_text(pInfo, 4);
+
+ if( iIdxpLink){
+ if( pIter->bFlag ) continue;
+ if( pIter->iCol!=iCol ) continue;
+ if( sqlite3_stricmp(pIter->zColl, zColl) ) continue;
+ pIter->bFlag = 1;
+ break;
+ }
+ if( pIter==0 ){
+ bMatch = 0;
+ break;
+ }
+ }else{
+ if( pT ){
+ if( pT->iCol!=iCol || sqlite3_stricmp(pT->zColl, zColl) ){
+ bMatch = 0;
+ break;
+ }
+ pT = pT->pLink;
+ }
+ }
+ }
+ idxFinalize(&rc, pInfo);
+
+ if( rc==SQLITE_OK && bMatch ){
+ sqlite3_finalize(pIdxList);
+ return 1;
+ }
+ }
+ idxFinalize(&rc, pIdxList);
+
+ *pRc = rc;
+ return 0;
+}
+
+/* Callback for sqlite3_exec() with query with leading count(*) column.
+ * The first argument is expected to be an int*, referent to be incremented
+ * if that leading column is not exactly '0'.
+ */
+static int countNonzeros(void* pCount, int nc,
+ char* azResults[], char* azColumns[]){
+ (void)azColumns; /* Suppress unused parameter warning */
+ if( nc>0 && (azResults[0][0]!='0' || azResults[0][1]!=0) ){
+ *((int *)pCount) += 1;
+ }
+ return 0;
+}
+
+static int idxCreateFromCons(
+ sqlite3expert *p,
+ IdxScan *pScan,
+ IdxConstraint *pEq,
+ IdxConstraint *pTail
+){
+ sqlite3 *dbm = p->dbm;
+ int rc = SQLITE_OK;
+ if( (pEq || pTail) && 0==idxFindCompatible(&rc, dbm, pScan, pEq, pTail) ){
+ IdxTable *pTab = pScan->pTab;
+ char *zCols = 0;
+ char *zIdx = 0;
+ IdxConstraint *pCons;
+ unsigned int h = 0;
+ const char *zFmt;
+
+ for(pCons=pEq; pCons; pCons=pCons->pLink){
+ zCols = idxAppendColDefn(&rc, zCols, pTab, pCons);
+ }
+ for(pCons=pTail; pCons; pCons=pCons->pLink){
+ zCols = idxAppendColDefn(&rc, zCols, pTab, pCons);
+ }
+
+ if( rc==SQLITE_OK ){
+ /* Hash the list of columns to come up with a name for the index */
+ const char *zTable = pScan->pTab->zName;
+ int quoteTable = idxIdentifierRequiresQuotes(zTable);
+ char *zName = 0; /* Index name */
+ int collisions = 0;
+ do{
+ int i;
+ char *zFind;
+ for(i=0; zCols[i]; i++){
+ h += ((h<<3) + zCols[i]);
+ }
+ sqlite3_free(zName);
+ zName = sqlite3_mprintf("%s_idx_%08x", zTable, h);
+ if( zName==0 ) break;
+ /* Is is unique among table, view and index names? */
+ zFmt = "SELECT count(*) FROM sqlite_schema WHERE name=%Q"
+ " AND type in ('index','table','view')";
+ zFind = sqlite3_mprintf(zFmt, zName);
+ i = 0;
+ rc = sqlite3_exec(dbm, zFind, countNonzeros, &i, 0);
+ assert(rc==SQLITE_OK);
+ sqlite3_free(zFind);
+ if( i==0 ){
+ collisions = 0;
+ break;
+ }
+ ++collisions;
+ }while( collisions<50 && zName!=0 );
+ if( collisions ){
+ /* This return means "Gave up trying to find a unique index name." */
+ rc = SQLITE_BUSY_TIMEOUT;
+ }else if( zName==0 ){
+ rc = SQLITE_NOMEM;
+ }else{
+ if( quoteTable ){
+ zFmt = "CREATE INDEX \"%w\" ON \"%w\"(%s)";
+ }else{
+ zFmt = "CREATE INDEX %s ON %s(%s)";
+ }
+ zIdx = sqlite3_mprintf(zFmt, zName, zTable, zCols);
+ if( !zIdx ){
+ rc = SQLITE_NOMEM;
+ }else{
+ rc = sqlite3_exec(dbm, zIdx, 0, 0, p->pzErrmsg);
+ if( rc!=SQLITE_OK ){
+ rc = SQLITE_BUSY_TIMEOUT;
+ }else{
+ idxHashAdd(&rc, &p->hIdx, zName, zIdx);
+ }
+ }
+ sqlite3_free(zName);
+ sqlite3_free(zIdx);
+ }
+ }
+
+ sqlite3_free(zCols);
+ }
+ return rc;
+}
+
+/*
+** Return true if list pList (linked by IdxConstraint.pLink) contains
+** a constraint compatible with *p. Otherwise return false.
+*/
+static int idxFindConstraint(IdxConstraint *pList, IdxConstraint *p){
+ IdxConstraint *pCmp;
+ for(pCmp=pList; pCmp; pCmp=pCmp->pLink){
+ if( p->iCol==pCmp->iCol ) return 1;
+ }
+ return 0;
+}
+
+static int idxCreateFromWhere(
+ sqlite3expert *p,
+ IdxScan *pScan, /* Create indexes for this scan */
+ IdxConstraint *pTail /* range/ORDER BY constraints for inclusion */
+){
+ IdxConstraint *p1 = 0;
+ IdxConstraint *pCon;
+ int rc;
+
+ /* Gather up all the == constraints. */
+ for(pCon=pScan->pEq; pCon; pCon=pCon->pNext){
+ if( !idxFindConstraint(p1, pCon) && !idxFindConstraint(pTail, pCon) ){
+ pCon->pLink = p1;
+ p1 = pCon;
+ }
+ }
+
+ /* Create an index using the == constraints collected above. And the
+ ** range constraint/ORDER BY terms passed in by the caller, if any. */
+ rc = idxCreateFromCons(p, pScan, p1, pTail);
+
+ /* If no range/ORDER BY passed by the caller, create a version of the
+ ** index for each range constraint. */
+ if( pTail==0 ){
+ for(pCon=pScan->pRange; rc==SQLITE_OK && pCon; pCon=pCon->pNext){
+ assert( pCon->pLink==0 );
+ if( !idxFindConstraint(p1, pCon) && !idxFindConstraint(pTail, pCon) ){
+ rc = idxCreateFromCons(p, pScan, p1, pCon);
+ }
+ }
+ }
+
+ return rc;
+}
+
+/*
+** Create candidate indexes in database [dbm] based on the data in
+** linked-list pScan.
+*/
+static int idxCreateCandidates(sqlite3expert *p){
+ int rc = SQLITE_OK;
+ IdxScan *pIter;
+
+ for(pIter=p->pScan; pIter && rc==SQLITE_OK; pIter=pIter->pNextScan){
+ rc = idxCreateFromWhere(p, pIter, 0);
+ if( rc==SQLITE_OK && pIter->pOrder ){
+ rc = idxCreateFromWhere(p, pIter, pIter->pOrder);
+ }
+ }
+
+ return rc;
+}
+
+/*
+** Free all elements of the linked list starting at pConstraint.
+*/
+static void idxConstraintFree(IdxConstraint *pConstraint){
+ IdxConstraint *pNext;
+ IdxConstraint *p;
+
+ for(p=pConstraint; p; p=pNext){
+ pNext = p->pNext;
+ sqlite3_free(p);
+ }
+}
+
+/*
+** Free all elements of the linked list starting from pScan up until pLast
+** (pLast is not freed).
+*/
+static void idxScanFree(IdxScan *pScan, IdxScan *pLast){
+ IdxScan *p;
+ IdxScan *pNext;
+ for(p=pScan; p!=pLast; p=pNext){
+ pNext = p->pNextScan;
+ idxConstraintFree(p->pOrder);
+ idxConstraintFree(p->pEq);
+ idxConstraintFree(p->pRange);
+ sqlite3_free(p);
+ }
+}
+
+/*
+** Free all elements of the linked list starting from pStatement up
+** until pLast (pLast is not freed).
+*/
+static void idxStatementFree(IdxStatement *pStatement, IdxStatement *pLast){
+ IdxStatement *p;
+ IdxStatement *pNext;
+ for(p=pStatement; p!=pLast; p=pNext){
+ pNext = p->pNext;
+ sqlite3_free(p->zEQP);
+ sqlite3_free(p->zIdx);
+ sqlite3_free(p);
+ }
+}
+
+/*
+** Free the linked list of IdxTable objects starting at pTab.
+*/
+static void idxTableFree(IdxTable *pTab){
+ IdxTable *pIter;
+ IdxTable *pNext;
+ for(pIter=pTab; pIter; pIter=pNext){
+ pNext = pIter->pNext;
+ sqlite3_free(pIter);
+ }
+}
+
+/*
+** Free the linked list of IdxWrite objects starting at pTab.
+*/
+static void idxWriteFree(IdxWrite *pTab){
+ IdxWrite *pIter;
+ IdxWrite *pNext;
+ for(pIter=pTab; pIter; pIter=pNext){
+ pNext = pIter->pNext;
+ sqlite3_free(pIter);
+ }
+}
+
+
+
+/*
+** This function is called after candidate indexes have been created. It
+** runs all the queries to see which indexes they prefer, and populates
+** IdxStatement.zIdx and IdxStatement.zEQP with the results.
+*/
+static int idxFindIndexes(
+ sqlite3expert *p,
+ char **pzErr /* OUT: Error message (sqlite3_malloc) */
+){
+ IdxStatement *pStmt;
+ sqlite3 *dbm = p->dbm;
+ int rc = SQLITE_OK;
+
+ IdxHash hIdx;
+ idxHashInit(&hIdx);
+
+ for(pStmt=p->pStatement; rc==SQLITE_OK && pStmt; pStmt=pStmt->pNext){
+ IdxHashEntry *pEntry;
+ sqlite3_stmt *pExplain = 0;
+ idxHashClear(&hIdx);
+ rc = idxPrintfPrepareStmt(dbm, &pExplain, pzErr,
+ "EXPLAIN QUERY PLAN %s", pStmt->zSql
+ );
+ while( rc==SQLITE_OK && sqlite3_step(pExplain)==SQLITE_ROW ){
+ /* int iId = sqlite3_column_int(pExplain, 0); */
+ /* int iParent = sqlite3_column_int(pExplain, 1); */
+ /* int iNotUsed = sqlite3_column_int(pExplain, 2); */
+ const char *zDetail = (const char*)sqlite3_column_text(pExplain, 3);
+ int nDetail;
+ int i;
+
+ if( !zDetail ) continue;
+ nDetail = STRLEN(zDetail);
+
+ for(i=0; ihIdx, zIdx, nIdx);
+ if( zSql ){
+ idxHashAdd(&rc, &hIdx, zSql, 0);
+ if( rc ) goto find_indexes_out;
+ }
+ break;
+ }
+ }
+
+ if( zDetail[0]!='-' ){
+ pStmt->zEQP = idxAppendText(&rc, pStmt->zEQP, "%s\n", zDetail);
+ }
+ }
+
+ for(pEntry=hIdx.pFirst; pEntry; pEntry=pEntry->pNext){
+ pStmt->zIdx = idxAppendText(&rc, pStmt->zIdx, "%s;\n", pEntry->zKey);
+ }
+
+ idxFinalize(&rc, pExplain);
+ }
+
+ find_indexes_out:
+ idxHashClear(&hIdx);
+ return rc;
+}
+
+static int idxAuthCallback(
+ void *pCtx,
+ int eOp,
+ const char *z3,
+ const char *z4,
+ const char *zDb,
+ const char *zTrigger
+){
+ int rc = SQLITE_OK;
+ (void)z4;
+ (void)zTrigger;
+ if( eOp==SQLITE_INSERT || eOp==SQLITE_UPDATE || eOp==SQLITE_DELETE ){
+ if( sqlite3_stricmp(zDb, "main")==0 ){
+ sqlite3expert *p = (sqlite3expert*)pCtx;
+ IdxTable *pTab;
+ for(pTab=p->pTable; pTab; pTab=pTab->pNext){
+ if( 0==sqlite3_stricmp(z3, pTab->zName) ) break;
+ }
+ if( pTab ){
+ IdxWrite *pWrite;
+ for(pWrite=p->pWrite; pWrite; pWrite=pWrite->pNext){
+ if( pWrite->pTab==pTab && pWrite->eOp==eOp ) break;
+ }
+ if( pWrite==0 ){
+ pWrite = idxMalloc(&rc, sizeof(IdxWrite));
+ if( rc==SQLITE_OK ){
+ pWrite->pTab = pTab;
+ pWrite->eOp = eOp;
+ pWrite->pNext = p->pWrite;
+ p->pWrite = pWrite;
+ }
+ }
+ }
+ }
+ }
+ return rc;
+}
+
+static int idxProcessOneTrigger(
+ sqlite3expert *p,
+ IdxWrite *pWrite,
+ char **pzErr
+){
+ static const char *zInt = UNIQUE_TABLE_NAME;
+ static const char *zDrop = "DROP TABLE " UNIQUE_TABLE_NAME;
+ IdxTable *pTab = pWrite->pTab;
+ const char *zTab = pTab->zName;
+ const char *zSql =
+ "SELECT 'CREATE TEMP' || substr(sql, 7) FROM sqlite_schema "
+ "WHERE tbl_name = %Q AND type IN ('table', 'trigger') "
+ "ORDER BY type;";
+ sqlite3_stmt *pSelect = 0;
+ int rc = SQLITE_OK;
+ char *zWrite = 0;
+
+ /* Create the table and its triggers in the temp schema */
+ rc = idxPrintfPrepareStmt(p->db, &pSelect, pzErr, zSql, zTab, zTab);
+ while( rc==SQLITE_OK && SQLITE_ROW==sqlite3_step(pSelect) ){
+ const char *zCreate = (const char*)sqlite3_column_text(pSelect, 0);
+ if( zCreate==0 ) continue;
+ rc = sqlite3_exec(p->dbv, zCreate, 0, 0, pzErr);
+ }
+ idxFinalize(&rc, pSelect);
+
+ /* Rename the table in the temp schema to zInt */
+ if( rc==SQLITE_OK ){
+ char *z = sqlite3_mprintf("ALTER TABLE temp.%Q RENAME TO %Q", zTab, zInt);
+ if( z==0 ){
+ rc = SQLITE_NOMEM;
+ }else{
+ rc = sqlite3_exec(p->dbv, z, 0, 0, pzErr);
+ sqlite3_free(z);
+ }
+ }
+
+ switch( pWrite->eOp ){
+ case SQLITE_INSERT: {
+ int i;
+ zWrite = idxAppendText(&rc, zWrite, "INSERT INTO %Q VALUES(", zInt);
+ for(i=0; inCol; i++){
+ zWrite = idxAppendText(&rc, zWrite, "%s?", i==0 ? "" : ", ");
+ }
+ zWrite = idxAppendText(&rc, zWrite, ")");
+ break;
+ }
+ case SQLITE_UPDATE: {
+ int i;
+ zWrite = idxAppendText(&rc, zWrite, "UPDATE %Q SET ", zInt);
+ for(i=0; inCol; i++){
+ zWrite = idxAppendText(&rc, zWrite, "%s%Q=?", i==0 ? "" : ", ",
+ pTab->aCol[i].zName
+ );
+ }
+ break;
+ }
+ default: {
+ assert( pWrite->eOp==SQLITE_DELETE );
+ if( rc==SQLITE_OK ){
+ zWrite = sqlite3_mprintf("DELETE FROM %Q", zInt);
+ if( zWrite==0 ) rc = SQLITE_NOMEM;
+ }
+ }
+ }
+
+ if( rc==SQLITE_OK ){
+ sqlite3_stmt *pX = 0;
+ rc = sqlite3_prepare_v2(p->dbv, zWrite, -1, &pX, 0);
+ idxFinalize(&rc, pX);
+ if( rc!=SQLITE_OK ){
+ idxDatabaseError(p->dbv, pzErr);
+ }
+ }
+ sqlite3_free(zWrite);
+
+ if( rc==SQLITE_OK ){
+ rc = sqlite3_exec(p->dbv, zDrop, 0, 0, pzErr);
+ }
+
+ return rc;
+}
+
+static int idxProcessTriggers(sqlite3expert *p, char **pzErr){
+ int rc = SQLITE_OK;
+ IdxWrite *pEnd = 0;
+ IdxWrite *pFirst = p->pWrite;
+
+ while( rc==SQLITE_OK && pFirst!=pEnd ){
+ IdxWrite *pIter;
+ for(pIter=pFirst; rc==SQLITE_OK && pIter!=pEnd; pIter=pIter->pNext){
+ rc = idxProcessOneTrigger(p, pIter, pzErr);
+ }
+ pEnd = pFirst;
+ pFirst = p->pWrite;
+ }
+
+ return rc;
+}
+
+/*
+** This function tests if the schema of the main database of database handle
+** db contains an object named zTab. Assuming no error occurs, output parameter
+** (*pbContains) is set to true if zTab exists, or false if it does not.
+**
+** Or, if an error occurs, an SQLite error code is returned. The final value
+** of (*pbContains) is undefined in this case.
+*/
+static int expertDbContainsObject(
+ sqlite3 *db,
+ const char *zTab,
+ int *pbContains /* OUT: True if object exists */
+){
+ const char *zSql = "SELECT 1 FROM sqlite_schema WHERE name = ?";
+ sqlite3_stmt *pSql = 0;
+ int rc = SQLITE_OK;
+ int ret = 0;
+
+ rc = sqlite3_prepare_v2(db, zSql, -1, &pSql, 0);
+ if( rc==SQLITE_OK ){
+ sqlite3_bind_text(pSql, 1, zTab, -1, SQLITE_STATIC);
+ if( SQLITE_ROW==sqlite3_step(pSql) ){
+ ret = 1;
+ }
+ rc = sqlite3_finalize(pSql);
+ }
+
+ *pbContains = ret;
+ return rc;
+}
+
+/*
+** Execute SQL command zSql using database handle db. If no error occurs,
+** set (*pzErr) to NULL and return SQLITE_OK.
+**
+** If an error does occur, return an SQLite error code and set (*pzErr) to
+** point to a buffer containing an English language error message. Except,
+** if the error message begins with "no such module:", then ignore the
+** error and return as if the SQL statement had succeeded.
+**
+** This is used to copy as much of the database schema as possible while
+** ignoring any errors related to missing virtual table modules.
+*/
+static int expertSchemaSql(sqlite3 *db, const char *zSql, char **pzErr){
+ int rc = SQLITE_OK;
+ char *zErr = 0;
+
+ rc = sqlite3_exec(db, zSql, 0, 0, &zErr);
+ if( rc!=SQLITE_OK && zErr ){
+ int nErr = STRLEN(zErr);
+ if( nErr>=15 && memcmp(zErr, "no such module:", 15)==0 ){
+ sqlite3_free(zErr);
+ rc = SQLITE_OK;
+ zErr = 0;
+ }
+ }
+
+ *pzErr = zErr;
+ return rc;
+}
+
+static int idxCreateVtabSchema(sqlite3expert *p, char **pzErrmsg){
+ int rc = idxRegisterVtab(p);
+ sqlite3_stmt *pSchema = 0;
+
+ /* For each table in the main db schema:
+ **
+ ** 1) Add an entry to the p->pTable list, and
+ ** 2) Create the equivalent virtual table in dbv.
+ */
+ rc = idxPrepareStmt(p->db, &pSchema, pzErrmsg,
+ "SELECT type, name, sql, 1, "
+ " substr(sql,1,14)=='create virtual' COLLATE nocase "
+ "FROM sqlite_schema "
+ "WHERE type IN ('table','view') AND "
+ " substr(name,1,7)!='sqlite_' COLLATE nocase "
+ " UNION ALL "
+ "SELECT type, name, sql, 2, 0 FROM sqlite_schema "
+ "WHERE type = 'trigger'"
+ " AND tbl_name IN(SELECT name FROM sqlite_schema WHERE type = 'view') "
+ "ORDER BY 4, 5 DESC, 1"
+ );
+ while( rc==SQLITE_OK && SQLITE_ROW==sqlite3_step(pSchema) ){
+ const char *zType = (const char*)sqlite3_column_text(pSchema, 0);
+ const char *zName = (const char*)sqlite3_column_text(pSchema, 1);
+ const char *zSql = (const char*)sqlite3_column_text(pSchema, 2);
+ int bVirtual = sqlite3_column_int(pSchema, 4);
+ int bExists = 0;
+
+ if( zType==0 || zName==0 ) continue;
+ rc = expertDbContainsObject(p->dbv, zName, &bExists);
+ if( rc || bExists ) continue;
+
+ if( zType[0]=='v' || zType[1]=='r' || bVirtual ){
+ /* A view. Or a trigger on a view. */
+ if( zSql ) rc = expertSchemaSql(p->dbv, zSql, pzErrmsg);
+ }else{
+ IdxTable *pTab;
+ rc = idxGetTableInfo(p->db, zName, &pTab, pzErrmsg);
+ if( rc==SQLITE_OK && ALWAYS(pTab!=0) ){
+ int i;
+ char *zInner = 0;
+ char *zOuter = 0;
+ pTab->pNext = p->pTable;
+ p->pTable = pTab;
+
+ /* The statement the vtab will pass to sqlite3_declare_vtab() */
+ zInner = idxAppendText(&rc, 0, "CREATE TABLE x(");
+ for(i=0; inCol; i++){
+ zInner = idxAppendText(&rc, zInner, "%s%Q COLLATE %s",
+ (i==0 ? "" : ", "), pTab->aCol[i].zName, pTab->aCol[i].zColl
+ );
+ }
+ zInner = idxAppendText(&rc, zInner, ")");
+
+ /* The CVT statement to create the vtab */
+ zOuter = idxAppendText(&rc, 0,
+ "CREATE VIRTUAL TABLE %Q USING expert(%Q)", zName, zInner
+ );
+ if( rc==SQLITE_OK ){
+ rc = sqlite3_exec(p->dbv, zOuter, 0, 0, pzErrmsg);
+ }
+ sqlite3_free(zInner);
+ sqlite3_free(zOuter);
+ }
+ }
+ }
+ idxFinalize(&rc, pSchema);
+ return rc;
+}
+
+struct IdxSampleCtx {
+ int iTarget;
+ double target; /* Target nRet/nRow value */
+ double nRow; /* Number of rows seen */
+ double nRet; /* Number of rows returned */
+};
+
+static void idxSampleFunc(
+ sqlite3_context *pCtx,
+ int argc,
+ sqlite3_value **argv
+){
+ struct IdxSampleCtx *p = (struct IdxSampleCtx*)sqlite3_user_data(pCtx);
+ int bRet;
+
+ (void)argv;
+ assert( argc==0 );
+ if( p->nRow==0.0 ){
+ bRet = 1;
+ }else{
+ bRet = (p->nRet / p->nRow) <= p->target;
+ if( bRet==0 ){
+ unsigned short rnd;
+ sqlite3_randomness(2, (void*)&rnd);
+ bRet = ((int)rnd % 100) <= p->iTarget;
+ }
+ }
+
+ sqlite3_result_int(pCtx, bRet);
+ p->nRow += 1.0;
+ p->nRet += (double)bRet;
+}
+
+struct IdxRemCtx {
+ int nSlot;
+ struct IdxRemSlot {
+ int eType; /* SQLITE_NULL, INTEGER, REAL, TEXT, BLOB */
+ i64 iVal; /* SQLITE_INTEGER value */
+ double rVal; /* SQLITE_FLOAT value */
+ int nByte; /* Bytes of space allocated at z */
+ int n; /* Size of buffer z */
+ char *z; /* SQLITE_TEXT/BLOB value */
+ } aSlot[1];
+};
+
+/*
+** Implementation of scalar function sqlite_expert_rem().
+*/
+static void idxRemFunc(
+ sqlite3_context *pCtx,
+ int argc,
+ sqlite3_value **argv
+){
+ struct IdxRemCtx *p = (struct IdxRemCtx*)sqlite3_user_data(pCtx);
+ struct IdxRemSlot *pSlot;
+ int iSlot;
+ assert( argc==2 );
+
+ iSlot = sqlite3_value_int(argv[0]);
+ assert( iSlotnSlot );
+ pSlot = &p->aSlot[iSlot];
+
+ switch( pSlot->eType ){
+ case SQLITE_NULL:
+ /* no-op */
+ break;
+
+ case SQLITE_INTEGER:
+ sqlite3_result_int64(pCtx, pSlot->iVal);
+ break;
+
+ case SQLITE_FLOAT:
+ sqlite3_result_double(pCtx, pSlot->rVal);
+ break;
+
+ case SQLITE_BLOB:
+ sqlite3_result_blob(pCtx, pSlot->z, pSlot->n, SQLITE_TRANSIENT);
+ break;
+
+ case SQLITE_TEXT:
+ sqlite3_result_text(pCtx, pSlot->z, pSlot->n, SQLITE_TRANSIENT);
+ break;
+ }
+
+ pSlot->eType = sqlite3_value_type(argv[1]);
+ switch( pSlot->eType ){
+ case SQLITE_NULL:
+ /* no-op */
+ break;
+
+ case SQLITE_INTEGER:
+ pSlot->iVal = sqlite3_value_int64(argv[1]);
+ break;
+
+ case SQLITE_FLOAT:
+ pSlot->rVal = sqlite3_value_double(argv[1]);
+ break;
+
+ case SQLITE_BLOB:
+ case SQLITE_TEXT: {
+ int nByte = sqlite3_value_bytes(argv[1]);
+ const void *pData = 0;
+ if( nByte>pSlot->nByte ){
+ char *zNew = (char*)sqlite3_realloc(pSlot->z, nByte*2);
+ if( zNew==0 ){
+ sqlite3_result_error_nomem(pCtx);
+ return;
+ }
+ pSlot->nByte = nByte*2;
+ pSlot->z = zNew;
+ }
+ pSlot->n = nByte;
+ if( pSlot->eType==SQLITE_BLOB ){
+ pData = sqlite3_value_blob(argv[1]);
+ if( pData ) memcpy(pSlot->z, pData, nByte);
+ }else{
+ pData = sqlite3_value_text(argv[1]);
+ memcpy(pSlot->z, pData, nByte);
+ }
+ break;
+ }
+ }
+}
+
+static int idxLargestIndex(sqlite3 *db, int *pnMax, char **pzErr){
+ int rc = SQLITE_OK;
+ const char *zMax =
+ "SELECT max(i.seqno) FROM "
+ " sqlite_schema AS s, "
+ " pragma_index_list(s.name) AS l, "
+ " pragma_index_info(l.name) AS i "
+ "WHERE s.type = 'table'";
+ sqlite3_stmt *pMax = 0;
+
+ *pnMax = 0;
+ rc = idxPrepareStmt(db, &pMax, pzErr, zMax);
+ if( rc==SQLITE_OK && SQLITE_ROW==sqlite3_step(pMax) ){
+ *pnMax = sqlite3_column_int(pMax, 0) + 1;
+ }
+ idxFinalize(&rc, pMax);
+
+ return rc;
+}
+
+static int idxPopulateOneStat1(
+ sqlite3expert *p,
+ sqlite3_stmt *pIndexXInfo,
+ sqlite3_stmt *pWriteStat,
+ const char *zTab,
+ const char *zIdx,
+ char **pzErr
+){
+ char *zCols = 0;
+ char *zOrder = 0;
+ char *zQuery = 0;
+ int nCol = 0;
+ int i;
+ sqlite3_stmt *pQuery = 0;
+ int *aStat = 0;
+ int rc = SQLITE_OK;
+
+ assert( p->iSample>0 );
+
+ /* Formulate the query text */
+ sqlite3_bind_text(pIndexXInfo, 1, zIdx, -1, SQLITE_STATIC);
+ while( SQLITE_OK==rc && SQLITE_ROW==sqlite3_step(pIndexXInfo) ){
+ const char *zComma = zCols==0 ? "" : ", ";
+ const char *zName = (const char*)sqlite3_column_text(pIndexXInfo, 0);
+ const char *zColl = (const char*)sqlite3_column_text(pIndexXInfo, 1);
+ if( zName==0 ){
+ /* This index contains an expression. Ignore it. */
+ sqlite3_free(zCols);
+ sqlite3_free(zOrder);
+ return sqlite3_reset(pIndexXInfo);
+ }
+ zCols = idxAppendText(&rc, zCols,
+ "%sx.%Q IS sqlite_expert_rem(%d, x.%Q) COLLATE %s",
+ zComma, zName, nCol, zName, zColl
+ );
+ zOrder = idxAppendText(&rc, zOrder, "%s%d", zComma, ++nCol);
+ }
+ sqlite3_reset(pIndexXInfo);
+ if( rc==SQLITE_OK ){
+ if( p->iSample==100 ){
+ zQuery = sqlite3_mprintf(
+ "SELECT %s FROM %Q x ORDER BY %s", zCols, zTab, zOrder
+ );
+ }else{
+ zQuery = sqlite3_mprintf(
+ "SELECT %s FROM temp."UNIQUE_TABLE_NAME" x ORDER BY %s", zCols, zOrder
+ );
+ }
+ }
+ sqlite3_free(zCols);
+ sqlite3_free(zOrder);
+
+ /* Formulate the query text */
+ if( rc==SQLITE_OK ){
+ sqlite3 *dbrem = (p->iSample==100 ? p->db : p->dbv);
+ rc = idxPrepareStmt(dbrem, &pQuery, pzErr, zQuery);
+ }
+ sqlite3_free(zQuery);
+
+ if( rc==SQLITE_OK ){
+ aStat = (int*)idxMalloc(&rc, sizeof(int)*(nCol+1));
+ }
+ if( rc==SQLITE_OK && SQLITE_ROW==sqlite3_step(pQuery) ){
+ IdxHashEntry *pEntry;
+ char *zStat = 0;
+ for(i=0; i<=nCol; i++) aStat[i] = 1;
+ while( rc==SQLITE_OK && SQLITE_ROW==sqlite3_step(pQuery) ){
+ aStat[0]++;
+ for(i=0; ihIdx, zIdx, STRLEN(zIdx));
+ if( pEntry ){
+ assert( pEntry->zVal2==0 );
+ pEntry->zVal2 = zStat;
+ }else{
+ sqlite3_free(zStat);
+ }
+ }
+ sqlite3_free(aStat);
+ idxFinalize(&rc, pQuery);
+
+ return rc;
+}
+
+static int idxBuildSampleTable(sqlite3expert *p, const char *zTab){
+ int rc;
+ char *zSql;
+
+ rc = sqlite3_exec(p->dbv,"DROP TABLE IF EXISTS temp."UNIQUE_TABLE_NAME,0,0,0);
+ if( rc!=SQLITE_OK ) return rc;
+
+ zSql = sqlite3_mprintf(
+ "CREATE TABLE temp." UNIQUE_TABLE_NAME " AS SELECT * FROM %Q", zTab
+ );
+ if( zSql==0 ) return SQLITE_NOMEM;
+ rc = sqlite3_exec(p->dbv, zSql, 0, 0, 0);
+ sqlite3_free(zSql);
+
+ return rc;
+}
+
+/*
+** This function is called as part of sqlite3_expert_analyze(). Candidate
+** indexes have already been created in database sqlite3expert.dbm, this
+** function populates sqlite_stat1 table in the same database.
+**
+** The stat1 data is generated by querying the
+*/
+static int idxPopulateStat1(sqlite3expert *p, char **pzErr){
+ int rc = SQLITE_OK;
+ int nMax =0;
+ struct IdxRemCtx *pCtx = 0;
+ struct IdxSampleCtx samplectx;
+ int i;
+ i64 iPrev = -100000;
+ sqlite3_stmt *pAllIndex = 0;
+ sqlite3_stmt *pIndexXInfo = 0;
+ sqlite3_stmt *pWrite = 0;
+
+ const char *zAllIndex =
+ "SELECT s.rowid, s.name, l.name FROM "
+ " sqlite_schema AS s, "
+ " pragma_index_list(s.name) AS l "
+ "WHERE s.type = 'table'";
+ const char *zIndexXInfo =
+ "SELECT name, coll FROM pragma_index_xinfo(?) WHERE key";
+ const char *zWrite = "INSERT INTO sqlite_stat1 VALUES(?, ?, ?)";
+
+ /* If iSample==0, no sqlite_stat1 data is required. */
+ if( p->iSample==0 ) return SQLITE_OK;
+
+ rc = idxLargestIndex(p->dbm, &nMax, pzErr);
+ if( nMax<=0 || rc!=SQLITE_OK ) return rc;
+
+ rc = sqlite3_exec(p->dbm, "ANALYZE; PRAGMA writable_schema=1", 0, 0, 0);
+
+ if( rc==SQLITE_OK ){
+ int nByte = sizeof(struct IdxRemCtx) + (sizeof(struct IdxRemSlot) * nMax);
+ pCtx = (struct IdxRemCtx*)idxMalloc(&rc, nByte);
+ }
+
+ if( rc==SQLITE_OK ){
+ sqlite3 *dbrem = (p->iSample==100 ? p->db : p->dbv);
+ rc = sqlite3_create_function(dbrem, "sqlite_expert_rem",
+ 2, SQLITE_UTF8, (void*)pCtx, idxRemFunc, 0, 0
+ );
+ }
+ if( rc==SQLITE_OK ){
+ rc = sqlite3_create_function(p->db, "sqlite_expert_sample",
+ 0, SQLITE_UTF8, (void*)&samplectx, idxSampleFunc, 0, 0
+ );
+ }
+
+ if( rc==SQLITE_OK ){
+ pCtx->nSlot = nMax+1;
+ rc = idxPrepareStmt(p->dbm, &pAllIndex, pzErr, zAllIndex);
+ }
+ if( rc==SQLITE_OK ){
+ rc = idxPrepareStmt(p->dbm, &pIndexXInfo, pzErr, zIndexXInfo);
+ }
+ if( rc==SQLITE_OK ){
+ rc = idxPrepareStmt(p->dbm, &pWrite, pzErr, zWrite);
+ }
+
+ while( rc==SQLITE_OK && SQLITE_ROW==sqlite3_step(pAllIndex) ){
+ i64 iRowid = sqlite3_column_int64(pAllIndex, 0);
+ const char *zTab = (const char*)sqlite3_column_text(pAllIndex, 1);
+ const char *zIdx = (const char*)sqlite3_column_text(pAllIndex, 2);
+ if( zTab==0 || zIdx==0 ) continue;
+ if( p->iSample<100 && iPrev!=iRowid ){
+ samplectx.target = (double)p->iSample / 100.0;
+ samplectx.iTarget = p->iSample;
+ samplectx.nRow = 0.0;
+ samplectx.nRet = 0.0;
+ rc = idxBuildSampleTable(p, zTab);
+ if( rc!=SQLITE_OK ) break;
+ }
+ rc = idxPopulateOneStat1(p, pIndexXInfo, pWrite, zTab, zIdx, pzErr);
+ iPrev = iRowid;
+ }
+ if( rc==SQLITE_OK && p->iSample<100 ){
+ rc = sqlite3_exec(p->dbv,
+ "DROP TABLE IF EXISTS temp." UNIQUE_TABLE_NAME, 0,0,0
+ );
+ }
+
+ idxFinalize(&rc, pAllIndex);
+ idxFinalize(&rc, pIndexXInfo);
+ idxFinalize(&rc, pWrite);
+
+ if( pCtx ){
+ for(i=0; inSlot; i++){
+ sqlite3_free(pCtx->aSlot[i].z);
+ }
+ sqlite3_free(pCtx);
+ }
+
+ if( rc==SQLITE_OK ){
+ rc = sqlite3_exec(p->dbm, "ANALYZE sqlite_schema", 0, 0, 0);
+ }
+
+ sqlite3_create_function(p->db, "sqlite_expert_rem", 2, SQLITE_UTF8, 0,0,0,0);
+ sqlite3_create_function(p->db, "sqlite_expert_sample", 0,SQLITE_UTF8,0,0,0,0);
+
+ sqlite3_exec(p->db, "DROP TABLE IF EXISTS temp."UNIQUE_TABLE_NAME,0,0,0);
+ return rc;
+}
+
+/*
+** Define and possibly pretend to use a useless collation sequence.
+** This pretense allows expert to accept SQL using custom collations.
+*/
+int dummyCompare(void *up1, int up2, const void *up3, int up4, const void *up5){
+ (void)up1;
+ (void)up2;
+ (void)up3;
+ (void)up4;
+ (void)up5;
+ assert(0); /* VDBE should never be run. */
+ return 0;
+}
+/* And a callback to register above upon actual need */
+void useDummyCS(void *up1, sqlite3 *db, int etr, const char *zName){
+ (void)up1;
+ sqlite3_create_collation_v2(db, zName, etr, 0, dummyCompare, 0);
+}
+
+#if !defined(SQLITE_OMIT_SCHEMA_PRAGMAS) \
+ && !defined(SQLITE_OMIT_INTROSPECTION_PRAGMAS)
+/*
+** dummy functions for no-op implementation of UDFs during expert's work
+*/
+void dummyUDF(sqlite3_context *up1, int up2, sqlite3_value **up3){
+ (void)up1;
+ (void)up2;
+ (void)up3;
+ assert(0); /* VDBE should never be run. */
+}
+void dummyUDFvalue(sqlite3_context *up1){
+ (void)up1;
+ assert(0); /* VDBE should never be run. */
+}
+
+/*
+** Register UDFs from user database with another.
+*/
+int registerUDFs(sqlite3 *dbSrc, sqlite3 *dbDst){
+ sqlite3_stmt *pStmt;
+ int rc = sqlite3_prepare_v2(dbSrc,
+ "SELECT name,type,enc,narg,flags "
+ "FROM pragma_function_list() "
+ "WHERE builtin==0", -1, &pStmt, 0);
+ if( rc==SQLITE_OK ){
+ while( SQLITE_ROW==(rc = sqlite3_step(pStmt)) ){
+ int nargs = sqlite3_column_int(pStmt,3);
+ int flags = sqlite3_column_int(pStmt,4);
+ const char *name = (char*)sqlite3_column_text(pStmt,0);
+ const char *type = (char*)sqlite3_column_text(pStmt,1);
+ const char *enc = (char*)sqlite3_column_text(pStmt,2);
+ if( name==0 || type==0 || enc==0 ){
+ /* no-op. Only happens on OOM */
+ }else{
+ int ienc = SQLITE_UTF8;
+ int rcf = SQLITE_ERROR;
+ if( strcmp(enc,"utf16le")==0 ) ienc = SQLITE_UTF16LE;
+ else if( strcmp(enc,"utf16be")==0 ) ienc = SQLITE_UTF16BE;
+ ienc |= (flags & (SQLITE_DETERMINISTIC|SQLITE_DIRECTONLY));
+ if( strcmp(type,"w")==0 ){
+ rcf = sqlite3_create_window_function(dbDst,name,nargs,ienc,0,
+ dummyUDF,dummyUDFvalue,0,0,0);
+ }else if( strcmp(type,"a")==0 ){
+ rcf = sqlite3_create_function(dbDst,name,nargs,ienc,0,
+ 0,dummyUDF,dummyUDFvalue);
+ }else if( strcmp(type,"s")==0 ){
+ rcf = sqlite3_create_function(dbDst,name,nargs,ienc,0,
+ dummyUDF,0,0);
+ }
+ if( rcf!=SQLITE_OK ){
+ rc = rcf;
+ break;
+ }
+ }
+ }
+ sqlite3_finalize(pStmt);
+ if( rc==SQLITE_DONE ) rc = SQLITE_OK;
+ }
+ return rc;
+}
+#endif
+
+/*
+** Allocate a new sqlite3expert object.
+*/
+sqlite3expert *sqlite3_expert_new(sqlite3 *db, char **pzErrmsg){
+ int rc = SQLITE_OK;
+ sqlite3expert *pNew;
+
+ pNew = (sqlite3expert*)idxMalloc(&rc, sizeof(sqlite3expert));
+
+ /* Open two in-memory databases to work with. The "vtab database" (dbv)
+ ** will contain a virtual table corresponding to each real table in
+ ** the user database schema, and a copy of each view. It is used to
+ ** collect information regarding the WHERE, ORDER BY and other clauses
+ ** of the user's query.
+ */
+ if( rc==SQLITE_OK ){
+ pNew->db = db;
+ pNew->iSample = 100;
+ rc = sqlite3_open(":memory:", &pNew->dbv);
+ }
+ if( rc==SQLITE_OK ){
+ rc = sqlite3_open(":memory:", &pNew->dbm);
+ if( rc==SQLITE_OK ){
+ sqlite3_db_config(pNew->dbm, SQLITE_DBCONFIG_TRIGGER_EQP, 1, (int*)0);
+ }
+ }
+
+ /* Allow custom collations to be dealt with through prepare. */
+ if( rc==SQLITE_OK ) rc = sqlite3_collation_needed(pNew->dbm,0,useDummyCS);
+ if( rc==SQLITE_OK ) rc = sqlite3_collation_needed(pNew->dbv,0,useDummyCS);
+
+#if !defined(SQLITE_OMIT_SCHEMA_PRAGMAS) \
+ && !defined(SQLITE_OMIT_INTROSPECTION_PRAGMAS)
+ /* Register UDFs from database [db] with [dbm] and [dbv]. */
+ if( rc==SQLITE_OK ){
+ rc = registerUDFs(pNew->db, pNew->dbm);
+ }
+ if( rc==SQLITE_OK ){
+ rc = registerUDFs(pNew->db, pNew->dbv);
+ }
+#endif
+
+ /* Copy the entire schema of database [db] into [dbm]. */
+ if( rc==SQLITE_OK ){
+ sqlite3_stmt *pSql = 0;
+ rc = idxPrintfPrepareStmt(pNew->db, &pSql, pzErrmsg,
+ "SELECT sql, name, substr(sql,1,14)=='create virtual' COLLATE nocase"
+ " FROM sqlite_schema WHERE substr(name,1,7)!='sqlite_' COLLATE nocase"
+ " ORDER BY 3 DESC, rowid"
+ );
+ while( rc==SQLITE_OK && SQLITE_ROW==sqlite3_step(pSql) ){
+ const char *zSql = (const char*)sqlite3_column_text(pSql, 0);
+ const char *zName = (const char*)sqlite3_column_text(pSql, 1);
+ int bExists = 0;
+ rc = expertDbContainsObject(pNew->dbm, zName, &bExists);
+ if( rc==SQLITE_OK && zSql && bExists==0 ){
+ rc = expertSchemaSql(pNew->dbm, zSql, pzErrmsg);
+ }
+ }
+ idxFinalize(&rc, pSql);
+ }
+
+ /* Create the vtab schema */
+ if( rc==SQLITE_OK ){
+ rc = idxCreateVtabSchema(pNew, pzErrmsg);
+ }
+
+ /* Register the auth callback with dbv */
+ if( rc==SQLITE_OK ){
+ sqlite3_set_authorizer(pNew->dbv, idxAuthCallback, (void*)pNew);
+ }
+
+ /* If an error has occurred, free the new object and reutrn NULL. Otherwise,
+ ** return the new sqlite3expert handle. */
+ if( rc!=SQLITE_OK ){
+ sqlite3_expert_destroy(pNew);
+ pNew = 0;
+ }
+ return pNew;
+}
+
+/*
+** Configure an sqlite3expert object.
+*/
+int sqlite3_expert_config(sqlite3expert *p, int op, ...){
+ int rc = SQLITE_OK;
+ va_list ap;
+ va_start(ap, op);
+ switch( op ){
+ case EXPERT_CONFIG_SAMPLE: {
+ int iVal = va_arg(ap, int);
+ if( iVal<0 ) iVal = 0;
+ if( iVal>100 ) iVal = 100;
+ p->iSample = iVal;
+ break;
+ }
+ default:
+ rc = SQLITE_NOTFOUND;
+ break;
+ }
+
+ va_end(ap);
+ return rc;
+}
+
+/*
+** Add an SQL statement to the analysis.
+*/
+int sqlite3_expert_sql(
+ sqlite3expert *p, /* From sqlite3_expert_new() */
+ const char *zSql, /* SQL statement to add */
+ char **pzErr /* OUT: Error message (if any) */
+){
+ IdxScan *pScanOrig = p->pScan;
+ IdxStatement *pStmtOrig = p->pStatement;
+ int rc = SQLITE_OK;
+ const char *zStmt = zSql;
+
+ if( p->bRun ) return SQLITE_MISUSE;
+
+ while( rc==SQLITE_OK && zStmt && zStmt[0] ){
+ sqlite3_stmt *pStmt = 0;
+ /* Ensure that the provided statement compiles against user's DB. */
+ rc = idxPrepareStmt(p->db, &pStmt, pzErr, zStmt);
+ if( rc!=SQLITE_OK ) break;
+ sqlite3_finalize(pStmt);
+ rc = sqlite3_prepare_v2(p->dbv, zStmt, -1, &pStmt, &zStmt);
+ if( rc==SQLITE_OK ){
+ if( pStmt ){
+ IdxStatement *pNew;
+ const char *z = sqlite3_sql(pStmt);
+ int n = STRLEN(z);
+ pNew = (IdxStatement*)idxMalloc(&rc, sizeof(IdxStatement) + n+1);
+ if( rc==SQLITE_OK ){
+ pNew->zSql = (char*)&pNew[1];
+ memcpy(pNew->zSql, z, n+1);
+ pNew->pNext = p->pStatement;
+ if( p->pStatement ) pNew->iId = p->pStatement->iId+1;
+ p->pStatement = pNew;
+ }
+ sqlite3_finalize(pStmt);
+ }
+ }else{
+ idxDatabaseError(p->dbv, pzErr);
+ }
+ }
+
+ if( rc!=SQLITE_OK ){
+ idxScanFree(p->pScan, pScanOrig);
+ idxStatementFree(p->pStatement, pStmtOrig);
+ p->pScan = pScanOrig;
+ p->pStatement = pStmtOrig;
+ }
+
+ return rc;
+}
+
+int sqlite3_expert_analyze(sqlite3expert *p, char **pzErr){
+ int rc;
+ IdxHashEntry *pEntry;
+
+ /* Do trigger processing to collect any extra IdxScan structures */
+ rc = idxProcessTriggers(p, pzErr);
+
+ /* Create candidate indexes within the in-memory database file */
+ if( rc==SQLITE_OK ){
+ rc = idxCreateCandidates(p);
+ }else if ( rc==SQLITE_BUSY_TIMEOUT ){
+ if( pzErr )
+ *pzErr = sqlite3_mprintf("Cannot find a unique index name to propose.");
+ return rc;
+ }
+
+ /* Generate the stat1 data */
+ if( rc==SQLITE_OK ){
+ rc = idxPopulateStat1(p, pzErr);
+ }
+
+ /* Formulate the EXPERT_REPORT_CANDIDATES text */
+ for(pEntry=p->hIdx.pFirst; pEntry; pEntry=pEntry->pNext){
+ p->zCandidates = idxAppendText(&rc, p->zCandidates,
+ "%s;%s%s\n", pEntry->zVal,
+ pEntry->zVal2 ? " -- stat1: " : "", pEntry->zVal2
+ );
+ }
+
+ /* Figure out which of the candidate indexes are preferred by the query
+ ** planner and report the results to the user. */
+ if( rc==SQLITE_OK ){
+ rc = idxFindIndexes(p, pzErr);
+ }
+
+ if( rc==SQLITE_OK ){
+ p->bRun = 1;
+ }
+ return rc;
+}
+
+/*
+** Return the total number of statements that have been added to this
+** sqlite3expert using sqlite3_expert_sql().
+*/
+int sqlite3_expert_count(sqlite3expert *p){
+ int nRet = 0;
+ if( p->pStatement ) nRet = p->pStatement->iId+1;
+ return nRet;
+}
+
+/*
+** Return a component of the report.
+*/
+const char *sqlite3_expert_report(sqlite3expert *p, int iStmt, int eReport){
+ const char *zRet = 0;
+ IdxStatement *pStmt;
+
+ if( p->bRun==0 ) return 0;
+ for(pStmt=p->pStatement; pStmt && pStmt->iId!=iStmt; pStmt=pStmt->pNext);
+ switch( eReport ){
+ case EXPERT_REPORT_SQL:
+ if( pStmt ) zRet = pStmt->zSql;
+ break;
+ case EXPERT_REPORT_INDEXES:
+ if( pStmt ) zRet = pStmt->zIdx;
+ break;
+ case EXPERT_REPORT_PLAN:
+ if( pStmt ) zRet = pStmt->zEQP;
+ break;
+ case EXPERT_REPORT_CANDIDATES:
+ zRet = p->zCandidates;
+ break;
+ }
+ return zRet;
+}
+
+/*
+** Free an sqlite3expert object.
+*/
+void sqlite3_expert_destroy(sqlite3expert *p){
+ if( p ){
+ sqlite3_close(p->dbm);
+ sqlite3_close(p->dbv);
+ idxScanFree(p->pScan, 0);
+ idxStatementFree(p->pStatement, 0);
+ idxTableFree(p->pTable);
+ idxWriteFree(p->pWrite);
+ idxHashClear(&p->hIdx);
+ sqlite3_free(p->zCandidates);
+ sqlite3_free(p);
+ }
+}
+
+#endif /* ifndef SQLITE_OMIT_VIRTUALTABLE */
diff --git a/ext/expert/sqlite3expert.h b/ext/expert/sqlite3expert.h
new file mode 100644
index 0000000000000000000000000000000000000000..6048137237a29a7031132ed6199269f4c76dc9d8
--- /dev/null
+++ b/ext/expert/sqlite3expert.h
@@ -0,0 +1,168 @@
+/*
+** 2017 April 07
+**
+** The author disclaims copyright to this source code. In place of
+** a legal notice, here is a blessing:
+**
+** May you do good and not evil.
+** May you find forgiveness for yourself and forgive others.
+** May you share freely, never taking more than you give.
+**
+*************************************************************************
+*/
+#if !defined(SQLITEEXPERT_H)
+#define SQLITEEXPERT_H 1
+#include "sqlite3.h"
+
+typedef struct sqlite3expert sqlite3expert;
+
+/*
+** Create a new sqlite3expert object.
+**
+** If successful, a pointer to the new object is returned and (*pzErr) set
+** to NULL. Or, if an error occurs, NULL is returned and (*pzErr) set to
+** an English-language error message. In this case it is the responsibility
+** of the caller to eventually free the error message buffer using
+** sqlite3_free().
+*/
+sqlite3expert *sqlite3_expert_new(sqlite3 *db, char **pzErr);
+
+/*
+** Configure an sqlite3expert object.
+**
+** EXPERT_CONFIG_SAMPLE:
+** By default, sqlite3_expert_analyze() generates sqlite_stat1 data for
+** each candidate index. This involves scanning and sorting the entire
+** contents of each user database table once for each candidate index
+** associated with the table. For large databases, this can be
+** prohibitively slow. This option allows the sqlite3expert object to
+** be configured so that sqlite_stat1 data is instead generated based on a
+** subset of each table, or so that no sqlite_stat1 data is used at all.
+**
+** A single integer argument is passed to this option. If the value is less
+** than or equal to zero, then no sqlite_stat1 data is generated or used by
+** the analysis - indexes are recommended based on the database schema only.
+** Or, if the value is 100 or greater, complete sqlite_stat1 data is
+** generated for each candidate index (this is the default). Finally, if the
+** value falls between 0 and 100, then it represents the percentage of user
+** table rows that should be considered when generating sqlite_stat1 data.
+**
+** Examples:
+**
+** // Do not generate any sqlite_stat1 data
+** sqlite3_expert_config(pExpert, EXPERT_CONFIG_SAMPLE, 0);
+**
+** // Generate sqlite_stat1 data based on 10% of the rows in each table.
+** sqlite3_expert_config(pExpert, EXPERT_CONFIG_SAMPLE, 10);
+*/
+int sqlite3_expert_config(sqlite3expert *p, int op, ...);
+
+#define EXPERT_CONFIG_SAMPLE 1 /* int */
+
+/*
+** Specify zero or more SQL statements to be included in the analysis.
+**
+** Buffer zSql must contain zero or more complete SQL statements. This
+** function parses all statements contained in the buffer and adds them
+** to the internal list of statements to analyze. If successful, SQLITE_OK
+** is returned and (*pzErr) set to NULL. Or, if an error occurs - for example
+** due to a error in the SQL - an SQLite error code is returned and (*pzErr)
+** may be set to point to an English language error message. In this case
+** the caller is responsible for eventually freeing the error message buffer
+** using sqlite3_free().
+**
+** If an error does occur while processing one of the statements in the
+** buffer passed as the second argument, none of the statements in the
+** buffer are added to the analysis.
+**
+** This function must be called before sqlite3_expert_analyze(). If a call
+** to this function is made on an sqlite3expert object that has already
+** been passed to sqlite3_expert_analyze() SQLITE_MISUSE is returned
+** immediately and no statements are added to the analysis.
+*/
+int sqlite3_expert_sql(
+ sqlite3expert *p, /* From a successful sqlite3_expert_new() */
+ const char *zSql, /* SQL statement(s) to add */
+ char **pzErr /* OUT: Error message (if any) */
+);
+
+
+/*
+** This function is called after the sqlite3expert object has been configured
+** with all SQL statements using sqlite3_expert_sql() to actually perform
+** the analysis. Once this function has been called, it is not possible to
+** add further SQL statements to the analysis.
+**
+** If successful, SQLITE_OK is returned and (*pzErr) is set to NULL. Or, if
+** an error occurs, an SQLite error code is returned and (*pzErr) set to
+** point to a buffer containing an English language error message. In this
+** case it is the responsibility of the caller to eventually free the buffer
+** using sqlite3_free().
+**
+** If an error does occur within this function, the sqlite3expert object
+** is no longer useful for any purpose. At that point it is no longer
+** possible to add further SQL statements to the object or to re-attempt
+** the analysis. The sqlite3expert object must still be freed using a call
+** sqlite3_expert_destroy().
+*/
+int sqlite3_expert_analyze(sqlite3expert *p, char **pzErr);
+
+/*
+** Return the total number of statements loaded using sqlite3_expert_sql().
+** The total number of SQL statements may be different from the total number
+** to calls to sqlite3_expert_sql().
+*/
+int sqlite3_expert_count(sqlite3expert*);
+
+/*
+** Return a component of the report.
+**
+** This function is called after sqlite3_expert_analyze() to extract the
+** results of the analysis. Each call to this function returns either a
+** NULL pointer or a pointer to a buffer containing a nul-terminated string.
+** The value passed as the third argument must be one of the EXPERT_REPORT_*
+** #define constants defined below.
+**
+** For some EXPERT_REPORT_* parameters, the buffer returned contains
+** information relating to a specific SQL statement. In these cases that
+** SQL statement is identified by the value passed as the second argument.
+** SQL statements are numbered from 0 in the order in which they are parsed.
+** If an out-of-range value (less than zero or equal to or greater than the
+** value returned by sqlite3_expert_count()) is passed as the second argument
+** along with such an EXPERT_REPORT_* parameter, NULL is always returned.
+**
+** EXPERT_REPORT_SQL:
+** Return the text of SQL statement iStmt.
+**
+** EXPERT_REPORT_INDEXES:
+** Return a buffer containing the CREATE INDEX statements for all recommended
+** indexes for statement iStmt. If there are no new recommeded indexes, NULL
+** is returned.
+**
+** EXPERT_REPORT_PLAN:
+** Return a buffer containing the EXPLAIN QUERY PLAN output for SQL query
+** iStmt after the proposed indexes have been added to the database schema.
+**
+** EXPERT_REPORT_CANDIDATES:
+** Return a pointer to a buffer containing the CREATE INDEX statements
+** for all indexes that were tested (for all SQL statements). The iStmt
+** parameter is ignored for EXPERT_REPORT_CANDIDATES calls.
+*/
+const char *sqlite3_expert_report(sqlite3expert*, int iStmt, int eReport);
+
+/*
+** Values for the third argument passed to sqlite3_expert_report().
+*/
+#define EXPERT_REPORT_SQL 1
+#define EXPERT_REPORT_INDEXES 2
+#define EXPERT_REPORT_PLAN 3
+#define EXPERT_REPORT_CANDIDATES 4
+
+/*
+** Free an (sqlite3expert*) handle and all associated resources. There
+** should be one call to this function for each successful call to
+** sqlite3-expert_new().
+*/
+void sqlite3_expert_destroy(sqlite3expert*);
+
+#endif /* !defined(SQLITEEXPERT_H) */
diff --git a/ext/expert/test_expert.c b/ext/expert/test_expert.c
new file mode 100644
index 0000000000000000000000000000000000000000..cae5d0f258f7b9cfa3b5fdfdcd852b3a65ae3070
--- /dev/null
+++ b/ext/expert/test_expert.c
@@ -0,0 +1,212 @@
+/*
+** 2017 April 07
+**
+** The author disclaims copyright to this source code. In place of
+** a legal notice, here is a blessing:
+**
+** May you do good and not evil.
+** May you find forgiveness for yourself and forgive others.
+** May you share freely, never taking more than you give.
+**
+*************************************************************************
+*/
+
+#if defined(SQLITE_TEST)
+
+#include "sqlite3expert.h"
+#include
+#include
+#include "tclsqlite.h"
+
+#ifndef SQLITE_OMIT_VIRTUALTABLE
+
+/*
+** Extract an sqlite3* db handle from the object passed as the second
+** argument. If successful, set *pDb to point to the db handle and return
+** TCL_OK. Otherwise, return TCL_ERROR.
+*/
+static int dbHandleFromObj(Tcl_Interp *interp, Tcl_Obj *pObj, sqlite3 **pDb){
+ Tcl_CmdInfo info;
+ if( 0==Tcl_GetCommandInfo(interp, Tcl_GetString(pObj), &info) ){
+ Tcl_AppendResult(interp, "no such handle: ", Tcl_GetString(pObj), 0);
+ return TCL_ERROR;
+ }
+
+ *pDb = *(sqlite3 **)info.objClientData;
+ return TCL_OK;
+}
+
+
+/*
+** Tclcmd: $expert sql SQL
+** $expert analyze
+** $expert count
+** $expert report STMT EREPORT
+** $expert destroy
+*/
+static int SQLITE_TCLAPI testExpertCmd(
+ void *clientData,
+ Tcl_Interp *interp,
+ int objc,
+ Tcl_Obj *CONST objv[]
+){
+ sqlite3expert *pExpert = (sqlite3expert*)clientData;
+ struct Subcmd {
+ const char *zSub;
+ int nArg;
+ const char *zMsg;
+ } aSub[] = {
+ { "sql", 1, "TABLE", }, /* 0 */
+ { "analyze", 0, "", }, /* 1 */
+ { "count", 0, "", }, /* 2 */
+ { "report", 2, "STMT EREPORT", }, /* 3 */
+ { "destroy", 0, "", }, /* 4 */
+ { 0 }
+ };
+ int iSub;
+ int rc = TCL_OK;
+ char *zErr = 0;
+
+ if( objc<2 ){
+ Tcl_WrongNumArgs(interp, 1, objv, "SUBCOMMAND ...");
+ return TCL_ERROR;
+ }
+ rc = Tcl_GetIndexFromObjStruct(interp,
+ objv[1], aSub, sizeof(aSub[0]), "sub-command", 0, &iSub
+ );
+ if( rc!=TCL_OK ) return rc;
+ if( objc!=2+aSub[iSub].nArg ){
+ Tcl_WrongNumArgs(interp, 2, objv, aSub[iSub].zMsg);
+ return TCL_ERROR;
+ }
+
+ switch( iSub ){
+ case 0: { /* sql */
+ char *zArg = Tcl_GetString(objv[2]);
+ rc = sqlite3_expert_sql(pExpert, zArg, &zErr);
+ break;
+ }
+
+ case 1: { /* analyze */
+ rc = sqlite3_expert_analyze(pExpert, &zErr);
+ break;
+ }
+
+ case 2: { /* count */
+ int n = sqlite3_expert_count(pExpert);
+ Tcl_SetObjResult(interp, Tcl_NewIntObj(n));
+ break;
+ }
+
+ case 3: { /* report */
+ const char *aEnum[] = {
+ "sql", "indexes", "plan", "candidates", 0
+ };
+ int iEnum;
+ int iStmt;
+ const char *zReport;
+
+ if( Tcl_GetIntFromObj(interp, objv[2], &iStmt)
+ || Tcl_GetIndexFromObj(interp, objv[3], aEnum, "report", 0, &iEnum)
+ ){
+ return TCL_ERROR;
+ }
+
+ assert( EXPERT_REPORT_SQL==1 );
+ assert( EXPERT_REPORT_INDEXES==2 );
+ assert( EXPERT_REPORT_PLAN==3 );
+ assert( EXPERT_REPORT_CANDIDATES==4 );
+ zReport = sqlite3_expert_report(pExpert, iStmt, 1+iEnum);
+ Tcl_SetObjResult(interp, Tcl_NewStringObj(zReport, -1));
+ break;
+ }
+
+ default: /* destroy */
+ assert( iSub==4 );
+ Tcl_DeleteCommand(interp, Tcl_GetString(objv[0]));
+ break;
+ }
+
+ if( rc!=TCL_OK ){
+ if( zErr ){
+ Tcl_SetObjResult(interp, Tcl_NewStringObj(zErr, -1));
+ }else{
+ extern const char *sqlite3ErrName(int);
+ Tcl_SetObjResult(interp, Tcl_NewStringObj(sqlite3ErrName(rc), -1));
+ }
+ }
+ sqlite3_free(zErr);
+ return rc;
+}
+
+static void SQLITE_TCLAPI testExpertDel(void *clientData){
+ sqlite3expert *pExpert = (sqlite3expert*)clientData;
+ sqlite3_expert_destroy(pExpert);
+}
+
+/*
+** sqlite3_expert_new DB
+*/
+static int SQLITE_TCLAPI test_sqlite3_expert_new(
+ void * clientData,
+ Tcl_Interp *interp,
+ int objc,
+ Tcl_Obj *CONST objv[]
+){
+ static int iCmd = 0;
+ sqlite3 *db;
+ char *zCmd = 0;
+ char *zErr = 0;
+ sqlite3expert *pExpert;
+ int rc = TCL_OK;
+
+ if( objc!=2 ){
+ Tcl_WrongNumArgs(interp, 1, objv, "DB");
+ return TCL_ERROR;
+ }
+ if( dbHandleFromObj(interp, objv[1], &db) ){
+ return TCL_ERROR;
+ }
+
+ zCmd = sqlite3_mprintf("sqlite3expert%d", ++iCmd);
+ if( zCmd==0 ){
+ Tcl_AppendResult(interp, "out of memory", (char*)0);
+ return TCL_ERROR;
+ }
+
+ pExpert = sqlite3_expert_new(db, &zErr);
+ if( pExpert==0 ){
+ Tcl_AppendResult(interp, zErr, (char*)0);
+ rc = TCL_ERROR;
+ }else{
+ void *p = (void*)pExpert;
+ Tcl_CreateObjCommand(interp, zCmd, testExpertCmd, p, testExpertDel);
+ Tcl_SetObjResult(interp, Tcl_NewStringObj(zCmd, -1));
+ }
+
+ sqlite3_free(zCmd);
+ sqlite3_free(zErr);
+ return rc;
+}
+
+#endif /* ifndef SQLITE_OMIT_VIRTUALTABLE */
+
+int TestExpert_Init(Tcl_Interp *interp){
+#ifndef SQLITE_OMIT_VIRTUALTABLE
+ struct Cmd {
+ const char *zCmd;
+ Tcl_ObjCmdProc *xProc;
+ } aCmd[] = {
+ { "sqlite3_expert_new", test_sqlite3_expert_new },
+ };
+ int i;
+
+ for(i=0; izCmd, p->xProc, 0, 0);
+ }
+#endif
+ return TCL_OK;
+}
+
+#endif
diff --git a/ext/fts3/README.content b/ext/fts3/README.content
new file mode 100644
index 0000000000000000000000000000000000000000..b6a75399bec9623fbb1edb6ddb2bb498b8820fd4
--- /dev/null
+++ b/ext/fts3/README.content
@@ -0,0 +1,176 @@
+
+FTS4 CONTENT OPTION
+
+ Normally, in order to create a full-text index on a dataset, the FTS4
+ module stores a copy of all indexed documents in a specially created
+ database table.
+
+ As of SQLite version 3.7.9, FTS4 supports a new option - "content" -
+ designed to extend FTS4 to support the creation of full-text indexes where:
+
+ * The indexed documents are not stored within the SQLite database
+ at all (a "contentless" FTS4 table), or
+
+ * The indexed documents are stored in a database table created and
+ managed by the user (an "external content" FTS4 table).
+
+ Because the indexed documents themselves are usually much larger than
+ the full-text index, the content option can sometimes be used to achieve
+ significant space savings.
+
+CONTENTLESS FTS4 TABLES
+
+ In order to create an FTS4 table that does not store a copy of the indexed
+ documents at all, the content option should be set to an empty string.
+ For example, the following SQL creates such an FTS4 table with three
+ columns - "a", "b", and "c":
+
+ CREATE VIRTUAL TABLE t1 USING fts4(content="", a, b, c);
+
+ Data can be inserted into such an FTS4 table using an INSERT statements.
+ However, unlike ordinary FTS4 tables, the user must supply an explicit
+ integer docid value. For example:
+
+ -- This statement is Ok:
+ INSERT INTO t1(docid, a, b, c) VALUES(1, 'a b c', 'd e f', 'g h i');
+
+ -- This statement causes an error, as no docid value has been provided:
+ INSERT INTO t1(a, b, c) VALUES('j k l', 'm n o', 'p q r');
+
+ It is not possible to UPDATE or DELETE a row stored in a contentless FTS4
+ table. Attempting to do so is an error.
+
+ Contentless FTS4 tables also support SELECT statements. However, it is
+ an error to attempt to retrieve the value of any table column other than
+ the docid column. The auxiliary function matchinfo() may be used, but
+ snippet() and offsets() may not. For example:
+
+ -- The following statements are Ok:
+ SELECT docid FROM t1 WHERE t1 MATCH 'xxx';
+ SELECT docid FROM t1 WHERE a MATCH 'xxx';
+ SELECT matchinfo(t1) FROM t1 WHERE t1 MATCH 'xxx';
+
+ -- The following statements all cause errors, as the value of columns
+ -- other than docid are required to evaluate them.
+ SELECT * FROM t1;
+ SELECT a, b FROM t1 WHERE t1 MATCH 'xxx';
+ SELECT docid FROM t1 WHERE a LIKE 'xxx%';
+ SELECT snippet(t1) FROM t1 WHERE t1 MATCH 'xxx';
+
+ Errors related to attempting to retrieve column values other than docid
+ are runtime errors that occur within sqlite3_step(). In some cases, for
+ example if the MATCH expression in a SELECT query matches zero rows, there
+ may be no error at all even if a statement does refer to column values
+ other than docid.
+
+EXTERNAL CONTENT FTS4 TABLES
+
+ An "external content" FTS4 table is similar to a contentless table, except
+ that if evaluation of a query requires the value of a column other than
+ docid, FTS4 attempts to retrieve that value from a table (or view, or
+ virtual table) nominated by the user (hereafter referred to as the "content
+ table"). The FTS4 module never writes to the content table, and writing
+ to the content table does not affect the full-text index. It is the
+ responsibility of the user to ensure that the content table and the
+ full-text index are consistent.
+
+ An external content FTS4 table is created by setting the content option
+ to the name of a table (or view, or virtual table) that may be queried by
+ FTS4 to retrieve column values when required. If the nominated table does
+ not exist, then an external content table behaves in the same way as
+ a contentless table. For example:
+
+ CREATE TABLE t2(id INTEGER PRIMARY KEY, a, b, c);
+ CREATE VIRTUAL TABLE t3 USING fts4(content="t2", a, c);
+
+ Assuming the nominated table does exist, then its columns must be the same
+ as or a superset of those defined for the FTS table.
+
+ When a users query on the FTS table requires a column value other than
+ docid, FTS attempts to read this value from the corresponding column of
+ the row in the content table with a rowid value equal to the current FTS
+ docid. Or, if such a row cannot be found in the content table, a NULL
+ value is used instead. For example:
+
+ CREATE TABLE t2(id INTEGER PRIMARY KEY, a, b, c, d);
+ CREATE VIRTUAL TABLE t3 USING fts4(content="t2", b, c);
+
+ INSERT INTO t2 VALUES(2, 'a b', 'c d', 'e f');
+ INSERT INTO t2 VALUES(3, 'g h', 'i j', 'k l');
+ INSERT INTO t3(docid, b, c) SELECT id, b, c FROM t2;
+
+ -- The following query returns a single row with two columns containing
+ -- the text values "i j" and "k l".
+ --
+ -- The query uses the full-text index to discover that the MATCH
+ -- term matches the row with docid=3. It then retrieves the values
+ -- of columns b and c from the row with rowid=3 in the content table
+ -- to return.
+ --
+ SELECT * FROM t3 WHERE t3 MATCH 'k';
+
+ -- Following the UPDATE, the query still returns a single row, this
+ -- time containing the text values "xxx" and "yyy". This is because the
+ -- full-text index still indicates that the row with docid=3 matches
+ -- the FTS4 query 'k', even though the documents stored in the content
+ -- table have been modified.
+ --
+ UPDATE t2 SET b = 'xxx', c = 'yyy' WHERE rowid = 3;
+ SELECT * FROM t3 WHERE t3 MATCH 'k';
+
+ -- Following the DELETE below, the query returns one row containing two
+ -- NULL values. NULL values are returned because FTS is unable to find
+ -- a row with rowid=3 within the content table.
+ --
+ DELETE FROM t2;
+ SELECT * FROM t3 WHERE t3 MATCH 'k';
+
+ When a row is deleted from an external content FTS4 table, FTS4 needs to
+ retrieve the column values of the row being deleted from the content table.
+ This is so that FTS4 can update the full-text index entries for each token
+ that occurs within the deleted row to indicate that that row has been
+ deleted. If the content table row cannot be found, or if it contains values
+ inconsistent with the contents of the FTS index, the results can be difficult
+ to predict. The FTS index may be left containing entries corresponding to the
+ deleted row, which can lead to seemingly nonsensical results being returned
+ by subsequent SELECT queries. The same applies when a row is updated, as
+ internally an UPDATE is the same as a DELETE followed by an INSERT.
+
+ Instead of writing separately to the full-text index and the content table,
+ some users may wish to use database triggers to keep the full-text index
+ up to date with respect to the set of documents stored in the content table.
+ For example, using the tables from earlier examples:
+
+ CREATE TRIGGER t2_bu BEFORE UPDATE ON t2 BEGIN
+ DELETE FROM t3 WHERE docid=old.rowid;
+ END;
+ CREATE TRIGGER t2_bd BEFORE DELETE ON t2 BEGIN
+ DELETE FROM t3 WHERE docid=old.rowid;
+ END;
+
+ CREATE TRIGGER t2_bu AFTER UPDATE ON t2 BEGIN
+ INSERT INTO t3(docid, b, c) VALUES(new.rowid, new.b, new.c);
+ END;
+ CREATE TRIGGER t2_bd AFTER INSERT ON t2 BEGIN
+ INSERT INTO t3(docid, b, c) VALUES(new.rowid, new.b, new.c);
+ END;
+
+ The DELETE trigger must be fired before the actual delete takes place
+ on the content table. This is so that FTS4 can still retrieve the original
+ values in order to update the full-text index. And the INSERT trigger must
+ be fired after the new row is inserted, so as to handle the case where the
+ rowid is assigned automatically within the system. The UPDATE trigger must
+ be split into two parts, one fired before and one after the update of the
+ content table, for the same reasons.
+
+ FTS4 features a special command similar to the 'optimize' command that
+ deletes the entire full-text index and rebuilds it based on the current
+ set of documents in the content table. Assuming again that "t3" is the
+ name of the external content FTS4 table, the command is:
+
+ INSERT INTO t3(t3) VALUES('rebuild');
+
+ This command may also be used with ordinary FTS4 tables, although it may
+ only be useful if the full-text index has somehow become corrupt. It is an
+ error to attempt to rebuild the full-text index maintained by a contentless
+ FTS4 table.
diff --git a/ext/fts3/README.syntax b/ext/fts3/README.syntax
new file mode 100644
index 0000000000000000000000000000000000000000..01bc80c5fbe9fce601884bf9e6dec46dab62c97e
--- /dev/null
+++ b/ext/fts3/README.syntax
@@ -0,0 +1,209 @@
+
+1. OVERVIEW
+
+ This README file describes the syntax of the arguments that may be passed to
+ the FTS3 MATCH operator used for full-text queries. For example, if table
+ "t1" is an Fts3 virtual table, the following SQL query:
+
+ SELECT * FROM t1 WHERE MATCH
+
+ may be used to retrieve all rows that match a specified for full-text query.
+ The text "" should be replaced by either the name of the fts3 table
+ (in this case "t1"), or by the name of one of the columns of the fts3
+ table. should be replaced by an SQL expression that
+ computes to a string containing an Fts3 query.
+
+ If the left-hand-side of the MATCH operator is set to the name of the
+ fts3 table, then by default the query may be matched against any column
+ of the table. If it is set to a column name, then by default the query
+ may only match the specified column. In both cases this may be overriden
+ as part of the query text (see sections 2 and 3 below).
+
+ As of SQLite version 3.6.8, Fts3 supports two slightly different query
+ formats; the standard syntax, which is used by default, and the enhanced
+ query syntax which can be selected by compiling with the pre-processor
+ symbol SQLITE_ENABLE_FTS3_PARENTHESIS defined.
+
+ -DSQLITE_ENABLE_FTS3_PARENTHESIS
+
+2. STANDARD QUERY SYNTAX
+
+ When using the standard Fts3 query syntax, a query usually consists of a
+ list of terms (words) separated by white-space characters. To match a
+ query, a row (or column) of an Fts3 table must contain each of the specified
+ terms. For example, the following query:
+
+ MATCH 'hello world'
+
+ matches rows (or columns, if is the name of a column name) that
+ contain at least one instance of the token "hello", and at least one
+ instance of the token "world". Tokens may be grouped into phrases using
+ quotation marks. In this case, a matching row or column must contain each
+ of the tokens in the phrase in the order specified, with no intervening
+ tokens. For example, the query:
+
+ MATCH '"hello world" joe"
+
+ matches the first of the following two documents, but not the second or
+ third:
+
+ "'Hello world', said Joe."
+ "One should always greet the world with a cheery hello, thought Joe."
+ "How many hello world programs could their be?"
+
+ As well as grouping tokens together by phrase, the binary NEAR operator
+ may be used to search for rows that contain two or more specified tokens
+ or phrases within a specified proximity of each other. The NEAR operator
+ must always be specified in upper case. The word "near" in lower or mixed
+ case is treated as an ordinary token. For example, the following query:
+
+ MATCH 'engineering NEAR consultancy'
+
+ matches rows that contain both the "engineering" and "consultancy" tokens
+ in the same column with not more than 10 other words between them. It does
+ not matter which of the two terms occurs first in the document, only that
+ they be seperated by only 10 tokens or less. The user may also specify
+ a different required proximity by adding "/N" immediately after the NEAR
+ operator, where N is an integer. For example:
+
+ MATCH 'engineering NEAR/5 consultancy'
+
+ searches for a row containing an instance of each specified token seperated
+ by not more than 5 other tokens. More than one NEAR operator can be used
+ in as sequence. For example this query:
+
+ MATCH 'reliable NEAR/2 engineering NEAR/5 consultancy'
+
+ searches for a row that contains an instance of the token "reliable"
+ seperated by not more than two tokens from an instance of "engineering",
+ which is in turn separated by not more than 5 other tokens from an
+ instance of the term "consultancy". Phrases enclosed in quotes may
+ also be used as arguments to the NEAR operator.
+
+ Similar to the NEAR operator, one or more tokens or phrases may be
+ separated by OR operators. In this case, only one of the specified tokens
+ or phrases must appear in the document. For example, the query:
+
+ MATCH 'hello OR world'
+
+ matches rows that contain either the term "hello", or the term "world",
+ or both. Note that unlike in many programming languages, the OR operator
+ has a higher precedence than the AND operators implied between white-space
+ separated tokens. The following query matches documents that contain the
+ term 'sqlite' and at least one of the terms 'fantastic' or 'impressive',
+ not those that contain both 'sqlite' and 'fantastic' or 'impressive':
+
+ MATCH 'sqlite fantastic OR impressive'
+
+ Any token that is part of an Fts3 query expression, whether or not it is
+ part of a phrase enclosed in quotes, may have a '*' character appended to
+ it. In this case, the token matches all terms that begin with the characters
+ of the token, not just those that exactly match it. For example, the
+ following query:
+
+ MATCH 'sql*'
+
+ matches all rows that contain the term "SQLite", as well as those that
+ contain "SQL".
+
+ A token that is not part of a quoted phrase may be preceded by a '-'
+ character, which indicates that matching rows must not contain the
+ specified term. For example, the following:
+
+ MATCH '"database engine" -sqlite'
+
+ matches rows that contain the phrase "database engine" but do not contain
+ the term "sqlite". If the '-' character occurs inside a quoted phrase,
+ it is ignored. It is possible to use both the '-' prefix and the '*' postfix
+ on a single term. At this time, all Fts3 queries must contain at least
+ one term or phrase that is not preceded by the '-' prefix.
+
+ Regardless of whether or not a table name or column name is used on the
+ left hand side of the MATCH operator, a specific column of the fts3 table
+ may be associated with each token in a query by preceding a token with
+ a column name followed by a ':' character. For example, regardless of what
+ is specified for , the following query requires that column "col1"
+ of the table contains the term "hello", and that column "col2" of the
+ table contains the term "world". If the table does not contain columns
+ named "col1" and "col2", then an error is returned and the query is
+ not run.
+
+ MATCH 'col1:hello col2:world'
+
+ It is not possible to associate a specific table column with a quoted
+ phrase or a term preceded by a '-' operator. A '*' character may be
+ appended to a term associated with a specific column for prefix matching.
+
+3. ENHANCED QUERY SYNTAX
+
+ The enhanced query syntax is quite similar to the standard query syntax,
+ with the following four differences:
+
+ 1) Parenthesis are supported. When using the enhanced query syntax,
+ parenthesis may be used to overcome the built-in precedence of the
+ supplied binary operators. For example, the following query:
+
+ MATCH '(hello world) OR (simple example)'
+
+ matches documents that contain both "hello" and "world", and documents
+ that contain both "simple" and "example". It is not possible to forumlate
+ such a query using the standard syntax.
+
+ 2) Instead of separating tokens and phrases by whitespace, an AND operator
+ may be explicitly specified. This does not change query processing at
+ all, but may be used to improve readability. For example, the following
+ query is handled identically to the one above:
+
+ MATCH '(hello AND world) OR (simple AND example)'
+
+ As with the OR and NEAR operators, the AND operator must be specified
+ in upper case. The word "and" specified in lower or mixed case is
+ handled as a regular token.
+
+ 3) The '-' token prefix is not supported. Instead, a new binary operator,
+ NOT, is included. The NOT operator requires that the query specified
+ as its left-hand operator matches, but that the query specified as the
+ right-hand operator does not. For example, to query for all rows that
+ contain the term "example" but not the term "simple", the following
+ query could be used:
+
+ MATCH 'example NOT simple'
+
+ As for all other operators, the NOT operator must be specified in
+ upper case. Otherwise it will be treated as a regular token.
+
+ 4) Unlike in the standard syntax, where the OR operator has a higher
+ precedence than the implicit AND operator, when using the enhanced
+ syntax implicit and explict AND operators have a higher precedence
+ than OR operators. Using the enhanced syntax, the following two
+ queries are equivalent:
+
+ MATCH 'sqlite fantastic OR impressive'
+ MATCH '(sqlite AND fantastic) OR impressive'
+
+ however, when using the standard syntax, the query:
+
+ MATCH 'sqlite fantastic OR impressive'
+
+ is equivalent to the enhanced syntax query:
+
+ MATCH 'sqlite AND (fantastic OR impressive)'
+
+ The precedence of all enhanced syntax operators, in order from highest
+ to lowest, is:
+
+ NEAR (highest precedence, tightest grouping)
+ NOT
+ AND
+ OR (lowest precedence, loosest grouping)
+
+ Using the advanced syntax, it is possible to specify expressions enclosed
+ in parenthesis as operands to the NOT, AND and OR operators. However both
+ the left and right hand side operands of NEAR operators must be either
+ tokens or phrases. Attempting the following query will return an error:
+
+ MATCH 'sqlite NEAR (fantastic OR impressive)'
+
+ Queries of this form must be re-written as:
+
+ MATCH 'sqlite NEAR fantastic OR sqlite NEAR impressive'
diff --git a/ext/fts3/README.tokenizers b/ext/fts3/README.tokenizers
new file mode 100644
index 0000000000000000000000000000000000000000..70bdceff061d02a6a5ebb144aca876421d4782a0
--- /dev/null
+++ b/ext/fts3/README.tokenizers
@@ -0,0 +1,135 @@
+
+1. FTS3 Tokenizers
+
+ When creating a new full-text table, FTS3 allows the user to select
+ the text tokenizer implementation to be used when indexing text
+ by specifying a "tokenize" clause as part of the CREATE VIRTUAL TABLE
+ statement:
+
+ CREATE VIRTUAL TABLE USING fts3(
+ [, tokenize []]
+ );
+
+ The built-in tokenizers (valid values to pass as ) are
+ "simple", "porter" and "unicode".
+
+ should consist of zero or more white-space separated
+ arguments to pass to the selected tokenizer implementation. The
+ interpretation of the arguments, if any, depends on the individual
+ tokenizer.
+
+2. Custom Tokenizers
+
+ FTS3 allows users to provide custom tokenizer implementations. The
+ interface used to create a new tokenizer is defined and described in
+ the fts3_tokenizer.h source file.
+
+ Registering a new FTS3 tokenizer is similar to registering a new
+ virtual table module with SQLite. The user passes a pointer to a
+ structure containing pointers to various callback functions that
+ make up the implementation of the new tokenizer type. For tokenizers,
+ the structure (defined in fts3_tokenizer.h) is called
+ "sqlite3_tokenizer_module".
+
+ FTS3 does not expose a C-function that users call to register new
+ tokenizer types with a database handle. Instead, the pointer must
+ be encoded as an SQL blob value and passed to FTS3 through the SQL
+ engine by evaluating a special scalar function, "fts3_tokenizer()".
+ The fts3_tokenizer() function may be called with one or two arguments,
+ as follows:
+
+ SELECT fts3_tokenizer();
+ SELECT fts3_tokenizer(, );
+
+ Where is a string identifying the tokenizer and
+ is a pointer to an sqlite3_tokenizer_module
+ structure encoded as an SQL blob. If the second argument is present,
+ it is registered as tokenizer and a copy of it
+ returned. If only one argument is passed, a pointer to the tokenizer
+ implementation currently registered as is returned,
+ encoded as a blob. Or, if no such tokenizer exists, an SQL exception
+ (error) is raised.
+
+ SECURITY: If the fts3 extension is used in an environment where potentially
+ malicious users may execute arbitrary SQL (i.e. gears), they should be
+ prevented from invoking the fts3_tokenizer() function. The
+ fts3_tokenizer() function is disabled by default. It is only enabled
+ by SQLITE_DBCONFIG_ENABLE_FTS3_TOKENIZER. Do not enable it in
+ security sensitive environments.
+
+ See "Sample code" below for an example of calling the fts3_tokenizer()
+ function from C code.
+
+3. ICU Library Tokenizers
+
+ If this extension is compiled with the SQLITE_ENABLE_ICU pre-processor
+ symbol defined, then there exists a built-in tokenizer named "icu"
+ implemented using the ICU library. The first argument passed to the
+ xCreate() method (see fts3_tokenizer.h) of this tokenizer may be
+ an ICU locale identifier. For example "tr_TR" for Turkish as used
+ in Turkey, or "en_AU" for English as used in Australia. For example:
+
+ "CREATE VIRTUAL TABLE thai_text USING fts3(text, tokenizer icu th_TH)"
+
+ The ICU tokenizer implementation is very simple. It splits the input
+ text according to the ICU rules for finding word boundaries and discards
+ any tokens that consist entirely of white-space. This may be suitable
+ for some applications in some locales, but not all. If more complex
+ processing is required, for example to implement stemming or
+ discard punctuation, this can be done by creating a tokenizer
+ implementation that uses the ICU tokenizer as part of its implementation.
+
+ When using the ICU tokenizer this way, it is safe to overwrite the
+ contents of the strings returned by the xNext() method (see
+ fts3_tokenizer.h).
+
+4. Sample code.
+
+ The following two code samples illustrate the way C code should invoke
+ the fts3_tokenizer() scalar function:
+
+ int registerTokenizer(
+ sqlite3 *db,
+ char *zName,
+ const sqlite3_tokenizer_module *p
+ ){
+ int rc;
+ sqlite3_stmt *pStmt;
+ const char zSql[] = "SELECT fts3_tokenizer(?, ?)";
+
+ rc = sqlite3_prepare_v2(db, zSql, -1, &pStmt, 0);
+ if( rc!=SQLITE_OK ){
+ return rc;
+ }
+
+ sqlite3_bind_text(pStmt, 1, zName, -1, SQLITE_STATIC);
+ sqlite3_bind_blob(pStmt, 2, &p, sizeof(p), SQLITE_STATIC);
+ sqlite3_step(pStmt);
+
+ return sqlite3_finalize(pStmt);
+ }
+
+ int queryTokenizer(
+ sqlite3 *db,
+ char *zName,
+ const sqlite3_tokenizer_module **pp
+ ){
+ int rc;
+ sqlite3_stmt *pStmt;
+ const char zSql[] = "SELECT fts3_tokenizer(?)";
+
+ *pp = 0;
+ rc = sqlite3_prepare_v2(db, zSql, -1, &pStmt, 0);
+ if( rc!=SQLITE_OK ){
+ return rc;
+ }
+
+ sqlite3_bind_text(pStmt, 1, zName, -1, SQLITE_STATIC);
+ if( SQLITE_ROW==sqlite3_step(pStmt) ){
+ if( sqlite3_column_type(pStmt, 0)==SQLITE_BLOB ){
+ memcpy(pp, sqlite3_column_blob(pStmt, 0), sizeof(*pp));
+ }
+ }
+
+ return sqlite3_finalize(pStmt);
+ }
diff --git a/ext/fts3/README.txt b/ext/fts3/README.txt
new file mode 100644
index 0000000000000000000000000000000000000000..517a2a043466cac040cb45686e397c9c650bb93e
--- /dev/null
+++ b/ext/fts3/README.txt
@@ -0,0 +1,4 @@
+This folder contains source code to the second full-text search
+extension for SQLite. While the API is the same, this version uses a
+substantially different storage schema from fts1, so tables will need
+to be rebuilt.
diff --git a/ext/fts3/fts3.c b/ext/fts3/fts3.c
new file mode 100644
index 0000000000000000000000000000000000000000..e58f256a4887267f80f672fcfa81c592aad9eed0
--- /dev/null
+++ b/ext/fts3/fts3.c
@@ -0,0 +1,6194 @@
+/*
+** 2006 Oct 10
+**
+** The author disclaims copyright to this source code. In place of
+** a legal notice, here is a blessing:
+**
+** May you do good and not evil.
+** May you find forgiveness for yourself and forgive others.
+** May you share freely, never taking more than you give.
+**
+******************************************************************************
+**
+** This is an SQLite module implementing full-text search.
+*/
+
+/*
+** The code in this file is only compiled if:
+**
+** * The FTS3 module is being built as an extension
+** (in which case SQLITE_CORE is not defined), or
+**
+** * The FTS3 module is being built into the core of
+** SQLite (in which case SQLITE_ENABLE_FTS3 is defined).
+*/
+
+/* The full-text index is stored in a series of b+tree (-like)
+** structures called segments which map terms to doclists. The
+** structures are like b+trees in layout, but are constructed from the
+** bottom up in optimal fashion and are not updatable. Since trees
+** are built from the bottom up, things will be described from the
+** bottom up.
+**
+**
+**** Varints ****
+** The basic unit of encoding is a variable-length integer called a
+** varint. We encode variable-length integers in little-endian order
+** using seven bits * per byte as follows:
+**
+** KEY:
+** A = 0xxxxxxx 7 bits of data and one flag bit
+** B = 1xxxxxxx 7 bits of data and one flag bit
+**
+** 7 bits - A
+** 14 bits - BA
+** 21 bits - BBA
+** and so on.
+**
+** This is similar in concept to how sqlite encodes "varints" but
+** the encoding is not the same. SQLite varints are big-endian
+** are are limited to 9 bytes in length whereas FTS3 varints are
+** little-endian and can be up to 10 bytes in length (in theory).
+**
+** Example encodings:
+**
+** 1: 0x01
+** 127: 0x7f
+** 128: 0x81 0x00
+**
+**
+**** Document lists ****
+** A doclist (document list) holds a docid-sorted list of hits for a
+** given term. Doclists hold docids and associated token positions.
+** A docid is the unique integer identifier for a single document.
+** A position is the index of a word within the document. The first
+** word of the document has a position of 0.
+**
+** FTS3 used to optionally store character offsets using a compile-time
+** option. But that functionality is no longer supported.
+**
+** A doclist is stored like this:
+**
+** array {
+** varint docid; (delta from previous doclist)
+** array { (position list for column 0)
+** varint position; (2 more than the delta from previous position)
+** }
+** array {
+** varint POS_COLUMN; (marks start of position list for new column)
+** varint column; (index of new column)
+** array {
+** varint position; (2 more than the delta from previous position)
+** }
+** }
+** varint POS_END; (marks end of positions for this document.
+** }
+**
+** Here, array { X } means zero or more occurrences of X, adjacent in
+** memory. A "position" is an index of a token in the token stream
+** generated by the tokenizer. Note that POS_END and POS_COLUMN occur
+** in the same logical place as the position element, and act as sentinals
+** ending a position list array. POS_END is 0. POS_COLUMN is 1.
+** The positions numbers are not stored literally but rather as two more
+** than the difference from the prior position, or the just the position plus
+** 2 for the first position. Example:
+**
+** label: A B C D E F G H I J K
+** value: 123 5 9 1 1 14 35 0 234 72 0
+**
+** The 123 value is the first docid. For column zero in this document
+** there are two matches at positions 3 and 10 (5-2 and 9-2+3). The 1
+** at D signals the start of a new column; the 1 at E indicates that the
+** new column is column number 1. There are two positions at 12 and 45
+** (14-2 and 35-2+12). The 0 at H indicate the end-of-document. The
+** 234 at I is the delta to next docid (357). It has one position 70
+** (72-2) and then terminates with the 0 at K.
+**
+** A "position-list" is the list of positions for multiple columns for
+** a single docid. A "column-list" is the set of positions for a single
+** column. Hence, a position-list consists of one or more column-lists,
+** a document record consists of a docid followed by a position-list and
+** a doclist consists of one or more document records.
+**
+** A bare doclist omits the position information, becoming an
+** array of varint-encoded docids.
+**
+**** Segment leaf nodes ****
+** Segment leaf nodes store terms and doclists, ordered by term. Leaf
+** nodes are written using LeafWriter, and read using LeafReader (to
+** iterate through a single leaf node's data) and LeavesReader (to
+** iterate through a segment's entire leaf layer). Leaf nodes have
+** the format:
+**
+** varint iHeight; (height from leaf level, always 0)
+** varint nTerm; (length of first term)
+** char pTerm[nTerm]; (content of first term)
+** varint nDoclist; (length of term's associated doclist)
+** char pDoclist[nDoclist]; (content of doclist)
+** array {
+** (further terms are delta-encoded)
+** varint nPrefix; (length of prefix shared with previous term)
+** varint nSuffix; (length of unshared suffix)
+** char pTermSuffix[nSuffix];(unshared suffix of next term)
+** varint nDoclist; (length of term's associated doclist)
+** char pDoclist[nDoclist]; (content of doclist)
+** }
+**
+** Here, array { X } means zero or more occurrences of X, adjacent in
+** memory.
+**
+** Leaf nodes are broken into blocks which are stored contiguously in
+** the %_segments table in sorted order. This means that when the end
+** of a node is reached, the next term is in the node with the next
+** greater node id.
+**
+** New data is spilled to a new leaf node when the current node
+** exceeds LEAF_MAX bytes (default 2048). New data which itself is
+** larger than STANDALONE_MIN (default 1024) is placed in a standalone
+** node (a leaf node with a single term and doclist). The goal of
+** these settings is to pack together groups of small doclists while
+** making it efficient to directly access large doclists. The
+** assumption is that large doclists represent terms which are more
+** likely to be query targets.
+**
+** TODO(shess) It may be useful for blocking decisions to be more
+** dynamic. For instance, it may make more sense to have a 2.5k leaf
+** node rather than splitting into 2k and .5k nodes. My intuition is
+** that this might extend through 2x or 4x the pagesize.
+**
+**
+**** Segment interior nodes ****
+** Segment interior nodes store blockids for subtree nodes and terms
+** to describe what data is stored by the each subtree. Interior
+** nodes are written using InteriorWriter, and read using
+** InteriorReader. InteriorWriters are created as needed when
+** SegmentWriter creates new leaf nodes, or when an interior node
+** itself grows too big and must be split. The format of interior
+** nodes:
+**
+** varint iHeight; (height from leaf level, always >0)
+** varint iBlockid; (block id of node's leftmost subtree)
+** optional {
+** varint nTerm; (length of first term)
+** char pTerm[nTerm]; (content of first term)
+** array {
+** (further terms are delta-encoded)
+** varint nPrefix; (length of shared prefix with previous term)
+** varint nSuffix; (length of unshared suffix)
+** char pTermSuffix[nSuffix]; (unshared suffix of next term)
+** }
+** }
+**
+** Here, optional { X } means an optional element, while array { X }
+** means zero or more occurrences of X, adjacent in memory.
+**
+** An interior node encodes n terms separating n+1 subtrees. The
+** subtree blocks are contiguous, so only the first subtree's blockid
+** is encoded. The subtree at iBlockid will contain all terms less
+** than the first term encoded (or all terms if no term is encoded).
+** Otherwise, for terms greater than or equal to pTerm[i] but less
+** than pTerm[i+1], the subtree for that term will be rooted at
+** iBlockid+i. Interior nodes only store enough term data to
+** distinguish adjacent children (if the rightmost term of the left
+** child is "something", and the leftmost term of the right child is
+** "wicked", only "w" is stored).
+**
+** New data is spilled to a new interior node at the same height when
+** the current node exceeds INTERIOR_MAX bytes (default 2048).
+** INTERIOR_MIN_TERMS (default 7) keeps large terms from monopolizing
+** interior nodes and making the tree too skinny. The interior nodes
+** at a given height are naturally tracked by interior nodes at
+** height+1, and so on.
+**
+**
+**** Segment directory ****
+** The segment directory in table %_segdir stores meta-information for
+** merging and deleting segments, and also the root node of the
+** segment's tree.
+**
+** The root node is the top node of the segment's tree after encoding
+** the entire segment, restricted to ROOT_MAX bytes (default 1024).
+** This could be either a leaf node or an interior node. If the top
+** node requires more than ROOT_MAX bytes, it is flushed to %_segments
+** and a new root interior node is generated (which should always fit
+** within ROOT_MAX because it only needs space for 2 varints, the
+** height and the blockid of the previous root).
+**
+** The meta-information in the segment directory is:
+** level - segment level (see below)
+** idx - index within level
+** - (level,idx uniquely identify a segment)
+** start_block - first leaf node
+** leaves_end_block - last leaf node
+** end_block - last block (including interior nodes)
+** root - contents of root node
+**
+** If the root node is a leaf node, then start_block,
+** leaves_end_block, and end_block are all 0.
+**
+**
+**** Segment merging ****
+** To amortize update costs, segments are grouped into levels and
+** merged in batches. Each increase in level represents exponentially
+** more documents.
+**
+** New documents (actually, document updates) are tokenized and
+** written individually (using LeafWriter) to a level 0 segment, with
+** incrementing idx. When idx reaches MERGE_COUNT (default 16), all
+** level 0 segments are merged into a single level 1 segment. Level 1
+** is populated like level 0, and eventually MERGE_COUNT level 1
+** segments are merged to a single level 2 segment (representing
+** MERGE_COUNT^2 updates), and so on.
+**
+** A segment merge traverses all segments at a given level in
+** parallel, performing a straightforward sorted merge. Since segment
+** leaf nodes are written in to the %_segments table in order, this
+** merge traverses the underlying sqlite disk structures efficiently.
+** After the merge, all segment blocks from the merged level are
+** deleted.
+**
+** MERGE_COUNT controls how often we merge segments. 16 seems to be
+** somewhat of a sweet spot for insertion performance. 32 and 64 show
+** very similar performance numbers to 16 on insertion, though they're
+** a tiny bit slower (perhaps due to more overhead in merge-time
+** sorting). 8 is about 20% slower than 16, 4 about 50% slower than
+** 16, 2 about 66% slower than 16.
+**
+** At query time, high MERGE_COUNT increases the number of segments
+** which need to be scanned and merged. For instance, with 100k docs
+** inserted:
+**
+** MERGE_COUNT segments
+** 16 25
+** 8 12
+** 4 10
+** 2 6
+**
+** This appears to have only a moderate impact on queries for very
+** frequent terms (which are somewhat dominated by segment merge
+** costs), and infrequent and non-existent terms still seem to be fast
+** even with many segments.
+**
+** TODO(shess) That said, it would be nice to have a better query-side
+** argument for MERGE_COUNT of 16. Also, it is possible/likely that
+** optimizations to things like doclist merging will swing the sweet
+** spot around.
+**
+**
+**
+**** Handling of deletions and updates ****
+** Since we're using a segmented structure, with no docid-oriented
+** index into the term index, we clearly cannot simply update the term
+** index when a document is deleted or updated. For deletions, we
+** write an empty doclist (varint(docid) varint(POS_END)), for updates
+** we simply write the new doclist. Segment merges overwrite older
+** data for a particular docid with newer data, so deletes or updates
+** will eventually overtake the earlier data and knock it out. The
+** query logic likewise merges doclists so that newer data knocks out
+** older data.
+*/
+
+#include "fts3Int.h"
+#if !defined(SQLITE_CORE) || defined(SQLITE_ENABLE_FTS3)
+
+#if defined(SQLITE_ENABLE_FTS3) && !defined(SQLITE_CORE)
+# define SQLITE_CORE 1
+#endif
+
+#include
+#include
+#include
+#include
+#include
+#include
+
+#include "fts3.h"
+#ifndef SQLITE_CORE
+# include "sqlite3ext.h"
+ SQLITE_EXTENSION_INIT1
+#endif
+
+typedef struct Fts3HashWrapper Fts3HashWrapper;
+struct Fts3HashWrapper {
+ Fts3Hash hash; /* Hash table */
+ int nRef; /* Number of pointers to this object */
+};
+
+static int fts3EvalNext(Fts3Cursor *pCsr);
+static int fts3EvalStart(Fts3Cursor *pCsr);
+static int fts3TermSegReaderCursor(
+ Fts3Cursor *, const char *, int, int, Fts3MultiSegReader **);
+
+/*
+** This variable is set to false when running tests for which the on disk
+** structures should not be corrupt. Otherwise, true. If it is false, extra
+** assert() conditions in the fts3 code are activated - conditions that are
+** only true if it is guaranteed that the fts3 database is not corrupt.
+*/
+#ifdef SQLITE_DEBUG
+int sqlite3_fts3_may_be_corrupt = 1;
+#endif
+
+/*
+** Write a 64-bit variable-length integer to memory starting at p[0].
+** The length of data written will be between 1 and FTS3_VARINT_MAX bytes.
+** The number of bytes written is returned.
+*/
+int sqlite3Fts3PutVarint(char *p, sqlite_int64 v){
+ unsigned char *q = (unsigned char *) p;
+ sqlite_uint64 vu = v;
+ do{
+ *q++ = (unsigned char) ((vu & 0x7f) | 0x80);
+ vu >>= 7;
+ }while( vu!=0 );
+ q[-1] &= 0x7f; /* turn off high bit in final byte */
+ assert( q - (unsigned char *)p <= FTS3_VARINT_MAX );
+ return (int) (q - (unsigned char *)p);
+}
+
+#define GETVARINT_STEP(v, ptr, shift, mask1, mask2, var, ret) \
+ v = (v & mask1) | ( (*(const unsigned char*)(ptr++)) << shift ); \
+ if( (v & mask2)==0 ){ var = v; return ret; }
+#define GETVARINT_INIT(v, ptr, shift, mask1, mask2, var, ret) \
+ v = (*ptr++); \
+ if( (v & mask2)==0 ){ var = v; return ret; }
+
+int sqlite3Fts3GetVarintU(const char *pBuf, sqlite_uint64 *v){
+ const unsigned char *p = (const unsigned char*)pBuf;
+ const unsigned char *pStart = p;
+ u32 a;
+ u64 b;
+ int shift;
+
+ GETVARINT_INIT(a, p, 0, 0x00, 0x80, *v, 1);
+ GETVARINT_STEP(a, p, 7, 0x7F, 0x4000, *v, 2);
+ GETVARINT_STEP(a, p, 14, 0x3FFF, 0x200000, *v, 3);
+ GETVARINT_STEP(a, p, 21, 0x1FFFFF, 0x10000000, *v, 4);
+ b = (a & 0x0FFFFFFF );
+
+ for(shift=28; shift<=63; shift+=7){
+ u64 c = *p++;
+ b += (c&0x7F) << shift;
+ if( (c & 0x80)==0 ) break;
+ }
+ *v = b;
+ return (int)(p - pStart);
+}
+
+/*
+** Read a 64-bit variable-length integer from memory starting at p[0].
+** Return the number of bytes read, or 0 on error.
+** The value is stored in *v.
+*/
+int sqlite3Fts3GetVarint(const char *pBuf, sqlite_int64 *v){
+ return sqlite3Fts3GetVarintU(pBuf, (sqlite3_uint64*)v);
+}
+
+/*
+** Read a 64-bit variable-length integer from memory starting at p[0] and
+** not extending past pEnd[-1].
+** Return the number of bytes read, or 0 on error.
+** The value is stored in *v.
+*/
+int sqlite3Fts3GetVarintBounded(
+ const char *pBuf,
+ const char *pEnd,
+ sqlite_int64 *v
+){
+ const unsigned char *p = (const unsigned char*)pBuf;
+ const unsigned char *pStart = p;
+ const unsigned char *pX = (const unsigned char*)pEnd;
+ u64 b = 0;
+ int shift;
+ for(shift=0; shift<=63; shift+=7){
+ u64 c = p=0 );
+ return 5;
+}
+
+/*
+** Return the number of bytes required to encode v as a varint
+*/
+int sqlite3Fts3VarintLen(sqlite3_uint64 v){
+ int i = 0;
+ do{
+ i++;
+ v >>= 7;
+ }while( v!=0 );
+ return i;
+}
+
+/*
+** Convert an SQL-style quoted string into a normal string by removing
+** the quote characters. The conversion is done in-place. If the
+** input does not begin with a quote character, then this routine
+** is a no-op.
+**
+** Examples:
+**
+** "abc" becomes abc
+** 'xyz' becomes xyz
+** [pqr] becomes pqr
+** `mno` becomes mno
+**
+*/
+void sqlite3Fts3Dequote(char *z){
+ char quote; /* Quote character (if any ) */
+
+ quote = z[0];
+ if( quote=='[' || quote=='\'' || quote=='"' || quote=='`' ){
+ int iIn = 1; /* Index of next byte to read from input */
+ int iOut = 0; /* Index of next byte to write to output */
+
+ /* If the first byte was a '[', then the close-quote character is a ']' */
+ if( quote=='[' ) quote = ']';
+
+ while( z[iIn] ){
+ if( z[iIn]==quote ){
+ if( z[iIn+1]!=quote ) break;
+ z[iOut++] = quote;
+ iIn += 2;
+ }else{
+ z[iOut++] = z[iIn++];
+ }
+ }
+ z[iOut] = '\0';
+ }
+}
+
+/*
+** Read a single varint from the doclist at *pp and advance *pp to point
+** to the first byte past the end of the varint. Add the value of the varint
+** to *pVal.
+*/
+static void fts3GetDeltaVarint(char **pp, sqlite3_int64 *pVal){
+ sqlite3_int64 iVal;
+ *pp += sqlite3Fts3GetVarint(*pp, &iVal);
+ *pVal += iVal;
+}
+
+/*
+** When this function is called, *pp points to the first byte following a
+** varint that is part of a doclist (or position-list, or any other list
+** of varints). This function moves *pp to point to the start of that varint,
+** and sets *pVal by the varint value.
+**
+** Argument pStart points to the first byte of the doclist that the
+** varint is part of.
+*/
+static void fts3GetReverseVarint(
+ char **pp,
+ char *pStart,
+ sqlite3_int64 *pVal
+){
+ sqlite3_int64 iVal;
+ char *p;
+
+ /* Pointer p now points at the first byte past the varint we are
+ ** interested in. So, unless the doclist is corrupt, the 0x80 bit is
+ ** clear on character p[-1]. */
+ for(p = (*pp)-2; p>=pStart && *p&0x80; p--);
+ p++;
+ *pp = p;
+
+ sqlite3Fts3GetVarint(p, &iVal);
+ *pVal = iVal;
+}
+
+/*
+** The xDisconnect() virtual table method.
+*/
+static int fts3DisconnectMethod(sqlite3_vtab *pVtab){
+ Fts3Table *p = (Fts3Table *)pVtab;
+ int i;
+
+ assert( p->nPendingData==0 );
+ assert( p->pSegments==0 );
+
+ /* Free any prepared statements held */
+ sqlite3_finalize(p->pSeekStmt);
+ for(i=0; iaStmt); i++){
+ sqlite3_finalize(p->aStmt[i]);
+ }
+ sqlite3_free(p->zSegmentsTbl);
+ sqlite3_free(p->zReadExprlist);
+ sqlite3_free(p->zWriteExprlist);
+ sqlite3_free(p->zContentTbl);
+ sqlite3_free(p->zLanguageid);
+
+ /* Invoke the tokenizer destructor to free the tokenizer. */
+ p->pTokenizer->pModule->xDestroy(p->pTokenizer);
+
+ sqlite3_free(p);
+ return SQLITE_OK;
+}
+
+/*
+** Write an error message into *pzErr
+*/
+void sqlite3Fts3ErrMsg(char **pzErr, const char *zFormat, ...){
+ va_list ap;
+ sqlite3_free(*pzErr);
+ va_start(ap, zFormat);
+ *pzErr = sqlite3_vmprintf(zFormat, ap);
+ va_end(ap);
+}
+
+/*
+** Construct one or more SQL statements from the format string given
+** and then evaluate those statements. The success code is written
+** into *pRc.
+**
+** If *pRc is initially non-zero then this routine is a no-op.
+*/
+static void fts3DbExec(
+ int *pRc, /* Success code */
+ sqlite3 *db, /* Database in which to run SQL */
+ const char *zFormat, /* Format string for SQL */
+ ... /* Arguments to the format string */
+){
+ va_list ap;
+ char *zSql;
+ if( *pRc ) return;
+ va_start(ap, zFormat);
+ zSql = sqlite3_vmprintf(zFormat, ap);
+ va_end(ap);
+ if( zSql==0 ){
+ *pRc = SQLITE_NOMEM;
+ }else{
+ *pRc = sqlite3_exec(db, zSql, 0, 0, 0);
+ sqlite3_free(zSql);
+ }
+}
+
+/*
+** The xDestroy() virtual table method.
+*/
+static int fts3DestroyMethod(sqlite3_vtab *pVtab){
+ Fts3Table *p = (Fts3Table *)pVtab;
+ int rc = SQLITE_OK; /* Return code */
+ const char *zDb = p->zDb; /* Name of database (e.g. "main", "temp") */
+ sqlite3 *db = p->db; /* Database handle */
+
+ /* Drop the shadow tables */
+ fts3DbExec(&rc, db,
+ "DROP TABLE IF EXISTS %Q.'%q_segments';"
+ "DROP TABLE IF EXISTS %Q.'%q_segdir';"
+ "DROP TABLE IF EXISTS %Q.'%q_docsize';"
+ "DROP TABLE IF EXISTS %Q.'%q_stat';"
+ "%s DROP TABLE IF EXISTS %Q.'%q_content';",
+ zDb, p->zName,
+ zDb, p->zName,
+ zDb, p->zName,
+ zDb, p->zName,
+ (p->zContentTbl ? "--" : ""), zDb,p->zName
+ );
+
+ /* If everything has worked, invoke fts3DisconnectMethod() to free the
+ ** memory associated with the Fts3Table structure and return SQLITE_OK.
+ ** Otherwise, return an SQLite error code.
+ */
+ return (rc==SQLITE_OK ? fts3DisconnectMethod(pVtab) : rc);
+}
+
+
+/*
+** Invoke sqlite3_declare_vtab() to declare the schema for the FTS3 table
+** passed as the first argument. This is done as part of the xConnect()
+** and xCreate() methods.
+**
+** If *pRc is non-zero when this function is called, it is a no-op.
+** Otherwise, if an error occurs, an SQLite error code is stored in *pRc
+** before returning.
+*/
+static void fts3DeclareVtab(int *pRc, Fts3Table *p){
+ if( *pRc==SQLITE_OK ){
+ int i; /* Iterator variable */
+ int rc; /* Return code */
+ char *zSql; /* SQL statement passed to declare_vtab() */
+ char *zCols; /* List of user defined columns */
+ const char *zLanguageid;
+
+ zLanguageid = (p->zLanguageid ? p->zLanguageid : "__langid");
+ sqlite3_vtab_config(p->db, SQLITE_VTAB_CONSTRAINT_SUPPORT, 1);
+ sqlite3_vtab_config(p->db, SQLITE_VTAB_INNOCUOUS);
+
+ /* Create a list of user columns for the virtual table */
+ zCols = sqlite3_mprintf("%Q, ", p->azColumn[0]);
+ for(i=1; zCols && inColumn; i++){
+ zCols = sqlite3_mprintf("%z%Q, ", zCols, p->azColumn[i]);
+ }
+
+ /* Create the whole "CREATE TABLE" statement to pass to SQLite */
+ zSql = sqlite3_mprintf(
+ "CREATE TABLE x(%s %Q HIDDEN, docid HIDDEN, %Q HIDDEN)",
+ zCols, p->zName, zLanguageid
+ );
+ if( !zCols || !zSql ){
+ rc = SQLITE_NOMEM;
+ }else{
+ rc = sqlite3_declare_vtab(p->db, zSql);
+ }
+
+ sqlite3_free(zSql);
+ sqlite3_free(zCols);
+ *pRc = rc;
+ }
+}
+
+/*
+** Create the %_stat table if it does not already exist.
+*/
+void sqlite3Fts3CreateStatTable(int *pRc, Fts3Table *p){
+ fts3DbExec(pRc, p->db,
+ "CREATE TABLE IF NOT EXISTS %Q.'%q_stat'"
+ "(id INTEGER PRIMARY KEY, value BLOB);",
+ p->zDb, p->zName
+ );
+ if( (*pRc)==SQLITE_OK ) p->bHasStat = 1;
+}
+
+/*
+** Create the backing store tables (%_content, %_segments and %_segdir)
+** required by the FTS3 table passed as the only argument. This is done
+** as part of the vtab xCreate() method.
+**
+** If the p->bHasDocsize boolean is true (indicating that this is an
+** FTS4 table, not an FTS3 table) then also create the %_docsize and
+** %_stat tables required by FTS4.
+*/
+static int fts3CreateTables(Fts3Table *p){
+ int rc = SQLITE_OK; /* Return code */
+ int i; /* Iterator variable */
+ sqlite3 *db = p->db; /* The database connection */
+
+ if( p->zContentTbl==0 ){
+ const char *zLanguageid = p->zLanguageid;
+ char *zContentCols; /* Columns of %_content table */
+
+ /* Create a list of user columns for the content table */
+ zContentCols = sqlite3_mprintf("docid INTEGER PRIMARY KEY");
+ for(i=0; zContentCols && inColumn; i++){
+ char *z = p->azColumn[i];
+ zContentCols = sqlite3_mprintf("%z, 'c%d%q'", zContentCols, i, z);
+ }
+ if( zLanguageid && zContentCols ){
+ zContentCols = sqlite3_mprintf("%z, langid", zContentCols, zLanguageid);
+ }
+ if( zContentCols==0 ) rc = SQLITE_NOMEM;
+
+ /* Create the content table */
+ fts3DbExec(&rc, db,
+ "CREATE TABLE %Q.'%q_content'(%s)",
+ p->zDb, p->zName, zContentCols
+ );
+ sqlite3_free(zContentCols);
+ }
+
+ /* Create other tables */
+ fts3DbExec(&rc, db,
+ "CREATE TABLE %Q.'%q_segments'(blockid INTEGER PRIMARY KEY, block BLOB);",
+ p->zDb, p->zName
+ );
+ fts3DbExec(&rc, db,
+ "CREATE TABLE %Q.'%q_segdir'("
+ "level INTEGER,"
+ "idx INTEGER,"
+ "start_block INTEGER,"
+ "leaves_end_block INTEGER,"
+ "end_block INTEGER,"
+ "root BLOB,"
+ "PRIMARY KEY(level, idx)"
+ ");",
+ p->zDb, p->zName
+ );
+ if( p->bHasDocsize ){
+ fts3DbExec(&rc, db,
+ "CREATE TABLE %Q.'%q_docsize'(docid INTEGER PRIMARY KEY, size BLOB);",
+ p->zDb, p->zName
+ );
+ }
+ assert( p->bHasStat==p->bFts4 );
+ if( p->bHasStat ){
+ sqlite3Fts3CreateStatTable(&rc, p);
+ }
+ return rc;
+}
+
+/*
+** Store the current database page-size in bytes in p->nPgsz.
+**
+** If *pRc is non-zero when this function is called, it is a no-op.
+** Otherwise, if an error occurs, an SQLite error code is stored in *pRc
+** before returning.
+*/
+static void fts3DatabasePageSize(int *pRc, Fts3Table *p){
+ if( *pRc==SQLITE_OK ){
+ int rc; /* Return code */
+ char *zSql; /* SQL text "PRAGMA %Q.page_size" */
+ sqlite3_stmt *pStmt; /* Compiled "PRAGMA %Q.page_size" statement */
+
+ zSql = sqlite3_mprintf("PRAGMA %Q.page_size", p->zDb);
+ if( !zSql ){
+ rc = SQLITE_NOMEM;
+ }else{
+ rc = sqlite3_prepare(p->db, zSql, -1, &pStmt, 0);
+ if( rc==SQLITE_OK ){
+ sqlite3_step(pStmt);
+ p->nPgsz = sqlite3_column_int(pStmt, 0);
+ rc = sqlite3_finalize(pStmt);
+ }else if( rc==SQLITE_AUTH ){
+ p->nPgsz = 1024;
+ rc = SQLITE_OK;
+ }
+ }
+ assert( p->nPgsz>0 || rc!=SQLITE_OK );
+ sqlite3_free(zSql);
+ *pRc = rc;
+ }
+}
+
+/*
+** "Special" FTS4 arguments are column specifications of the following form:
+**
+** =
+**
+** There may not be whitespace surrounding the "=" character. The
+** term may be quoted, but the may not.
+*/
+static int fts3IsSpecialColumn(
+ const char *z,
+ int *pnKey,
+ char **pzValue
+){
+ char *zValue;
+ const char *zCsr = z;
+
+ while( *zCsr!='=' ){
+ if( *zCsr=='\0' ) return 0;
+ zCsr++;
+ }
+
+ *pnKey = (int)(zCsr-z);
+ zValue = sqlite3_mprintf("%s", &zCsr[1]);
+ if( zValue ){
+ sqlite3Fts3Dequote(zValue);
+ }
+ *pzValue = zValue;
+ return 1;
+}
+
+/*
+** Append the output of a printf() style formatting to an existing string.
+*/
+static void fts3Appendf(
+ int *pRc, /* IN/OUT: Error code */
+ char **pz, /* IN/OUT: Pointer to string buffer */
+ const char *zFormat, /* Printf format string to append */
+ ... /* Arguments for printf format string */
+){
+ if( *pRc==SQLITE_OK ){
+ va_list ap;
+ char *z;
+ va_start(ap, zFormat);
+ z = sqlite3_vmprintf(zFormat, ap);
+ va_end(ap);
+ if( z && *pz ){
+ char *z2 = sqlite3_mprintf("%s%s", *pz, z);
+ sqlite3_free(z);
+ z = z2;
+ }
+ if( z==0 ) *pRc = SQLITE_NOMEM;
+ sqlite3_free(*pz);
+ *pz = z;
+ }
+}
+
+/*
+** Return a copy of input string zInput enclosed in double-quotes (") and
+** with all double quote characters escaped. For example:
+**
+** fts3QuoteId("un \"zip\"") -> "un \"\"zip\"\""
+**
+** The pointer returned points to memory obtained from sqlite3_malloc(). It
+** is the callers responsibility to call sqlite3_free() to release this
+** memory.
+*/
+static char *fts3QuoteId(char const *zInput){
+ sqlite3_int64 nRet;
+ char *zRet;
+ nRet = 2 + (int)strlen(zInput)*2 + 1;
+ zRet = sqlite3_malloc64(nRet);
+ if( zRet ){
+ int i;
+ char *z = zRet;
+ *(z++) = '"';
+ for(i=0; zInput[i]; i++){
+ if( zInput[i]=='"' ) *(z++) = '"';
+ *(z++) = zInput[i];
+ }
+ *(z++) = '"';
+ *(z++) = '\0';
+ }
+ return zRet;
+}
+
+/*
+** Return a list of comma separated SQL expressions and a FROM clause that
+** could be used in a SELECT statement such as the following:
+**
+** SELECT FROM %_content AS x ...
+**
+** to return the docid, followed by each column of text data in order
+** from left to write. If parameter zFunc is not NULL, then instead of
+** being returned directly each column of text data is passed to an SQL
+** function named zFunc first. For example, if zFunc is "unzip" and the
+** table has the three user-defined columns "a", "b", and "c", the following
+** string is returned:
+**
+** "docid, unzip(x.'a'), unzip(x.'b'), unzip(x.'c') FROM %_content AS x"
+**
+** The pointer returned points to a buffer allocated by sqlite3_malloc(). It
+** is the responsibility of the caller to eventually free it.
+**
+** If *pRc is not SQLITE_OK when this function is called, it is a no-op (and
+** a NULL pointer is returned). Otherwise, if an OOM error is encountered
+** by this function, NULL is returned and *pRc is set to SQLITE_NOMEM. If
+** no error occurs, *pRc is left unmodified.
+*/
+static char *fts3ReadExprList(Fts3Table *p, const char *zFunc, int *pRc){
+ char *zRet = 0;
+ char *zFree = 0;
+ char *zFunction;
+ int i;
+
+ if( p->zContentTbl==0 ){
+ if( !zFunc ){
+ zFunction = "";
+ }else{
+ zFree = zFunction = fts3QuoteId(zFunc);
+ }
+ fts3Appendf(pRc, &zRet, "docid");
+ for(i=0; inColumn; i++){
+ fts3Appendf(pRc, &zRet, ",%s(x.'c%d%q')", zFunction, i, p->azColumn[i]);
+ }
+ if( p->zLanguageid ){
+ fts3Appendf(pRc, &zRet, ", x.%Q", "langid");
+ }
+ sqlite3_free(zFree);
+ }else{
+ fts3Appendf(pRc, &zRet, "rowid");
+ for(i=0; inColumn; i++){
+ fts3Appendf(pRc, &zRet, ", x.'%q'", p->azColumn[i]);
+ }
+ if( p->zLanguageid ){
+ fts3Appendf(pRc, &zRet, ", x.%Q", p->zLanguageid);
+ }
+ }
+ fts3Appendf(pRc, &zRet, " FROM '%q'.'%q%s' AS x",
+ p->zDb,
+ (p->zContentTbl ? p->zContentTbl : p->zName),
+ (p->zContentTbl ? "" : "_content")
+ );
+ return zRet;
+}
+
+/*
+** Return a list of N comma separated question marks, where N is the number
+** of columns in the %_content table (one for the docid plus one for each
+** user-defined text column).
+**
+** If argument zFunc is not NULL, then all but the first question mark
+** is preceded by zFunc and an open bracket, and followed by a closed
+** bracket. For example, if zFunc is "zip" and the FTS3 table has three
+** user-defined text columns, the following string is returned:
+**
+** "?, zip(?), zip(?), zip(?)"
+**
+** The pointer returned points to a buffer allocated by sqlite3_malloc(). It
+** is the responsibility of the caller to eventually free it.
+**
+** If *pRc is not SQLITE_OK when this function is called, it is a no-op (and
+** a NULL pointer is returned). Otherwise, if an OOM error is encountered
+** by this function, NULL is returned and *pRc is set to SQLITE_NOMEM. If
+** no error occurs, *pRc is left unmodified.
+*/
+static char *fts3WriteExprList(Fts3Table *p, const char *zFunc, int *pRc){
+ char *zRet = 0;
+ char *zFree = 0;
+ char *zFunction;
+ int i;
+
+ if( !zFunc ){
+ zFunction = "";
+ }else{
+ zFree = zFunction = fts3QuoteId(zFunc);
+ }
+ fts3Appendf(pRc, &zRet, "?");
+ for(i=0; inColumn; i++){
+ fts3Appendf(pRc, &zRet, ",%s(?)", zFunction);
+ }
+ if( p->zLanguageid ){
+ fts3Appendf(pRc, &zRet, ", ?");
+ }
+ sqlite3_free(zFree);
+ return zRet;
+}
+
+/*
+** Buffer z contains a positive integer value encoded as utf-8 text.
+** Decode this value and store it in *pnOut, returning the number of bytes
+** consumed. If an overflow error occurs return a negative value.
+*/
+int sqlite3Fts3ReadInt(const char *z, int *pnOut){
+ u64 iVal = 0;
+ int i;
+ for(i=0; z[i]>='0' && z[i]<='9'; i++){
+ iVal = iVal*10 + (z[i] - '0');
+ if( iVal>0x7FFFFFFF ) return -1;
+ }
+ *pnOut = (int)iVal;
+ return i;
+}
+
+/*
+** This function interprets the string at (*pp) as a non-negative integer
+** value. It reads the integer and sets *pnOut to the value read, then
+** sets *pp to point to the byte immediately following the last byte of
+** the integer value.
+**
+** Only decimal digits ('0'..'9') may be part of an integer value.
+**
+** If *pp does not being with a decimal digit SQLITE_ERROR is returned and
+** the output value undefined. Otherwise SQLITE_OK is returned.
+**
+** This function is used when parsing the "prefix=" FTS4 parameter.
+*/
+static int fts3GobbleInt(const char **pp, int *pnOut){
+ const int MAX_NPREFIX = 10000000;
+ int nInt = 0; /* Output value */
+ int nByte;
+ nByte = sqlite3Fts3ReadInt(*pp, &nInt);
+ if( nInt>MAX_NPREFIX ){
+ nInt = 0;
+ }
+ if( nByte==0 ){
+ return SQLITE_ERROR;
+ }
+ *pnOut = nInt;
+ *pp += nByte;
+ return SQLITE_OK;
+}
+
+/*
+** This function is called to allocate an array of Fts3Index structures
+** representing the indexes maintained by the current FTS table. FTS tables
+** always maintain the main "terms" index, but may also maintain one or
+** more "prefix" indexes, depending on the value of the "prefix=" parameter
+** (if any) specified as part of the CREATE VIRTUAL TABLE statement.
+**
+** Argument zParam is passed the value of the "prefix=" option if one was
+** specified, or NULL otherwise.
+**
+** If no error occurs, SQLITE_OK is returned and *apIndex set to point to
+** the allocated array. *pnIndex is set to the number of elements in the
+** array. If an error does occur, an SQLite error code is returned.
+**
+** Regardless of whether or not an error is returned, it is the responsibility
+** of the caller to call sqlite3_free() on the output array to free it.
+*/
+static int fts3PrefixParameter(
+ const char *zParam, /* ABC in prefix=ABC parameter to parse */
+ int *pnIndex, /* OUT: size of *apIndex[] array */
+ struct Fts3Index **apIndex /* OUT: Array of indexes for this table */
+){
+ struct Fts3Index *aIndex; /* Allocated array */
+ int nIndex = 1; /* Number of entries in array */
+
+ if( zParam && zParam[0] ){
+ const char *p;
+ nIndex++;
+ for(p=zParam; *p; p++){
+ if( *p==',' ) nIndex++;
+ }
+ }
+
+ aIndex = sqlite3_malloc64(sizeof(struct Fts3Index) * nIndex);
+ *apIndex = aIndex;
+ if( !aIndex ){
+ return SQLITE_NOMEM;
+ }
+
+ memset(aIndex, 0, sizeof(struct Fts3Index) * nIndex);
+ if( zParam ){
+ const char *p = zParam;
+ int i;
+ for(i=1; i=0 );
+ if( nPrefix==0 ){
+ nIndex--;
+ i--;
+ }else{
+ aIndex[i].nPrefix = nPrefix;
+ }
+ p++;
+ }
+ }
+
+ *pnIndex = nIndex;
+ return SQLITE_OK;
+}
+
+/*
+** This function is called when initializing an FTS4 table that uses the
+** content=xxx option. It determines the number of and names of the columns
+** of the new FTS4 table.
+**
+** The third argument passed to this function is the value passed to the
+** config=xxx option (i.e. "xxx"). This function queries the database for
+** a table of that name. If found, the output variables are populated
+** as follows:
+**
+** *pnCol: Set to the number of columns table xxx has,
+**
+** *pnStr: Set to the total amount of space required to store a copy
+** of each columns name, including the nul-terminator.
+**
+** *pazCol: Set to point to an array of *pnCol strings. Each string is
+** the name of the corresponding column in table xxx. The array
+** and its contents are allocated using a single allocation. It
+** is the responsibility of the caller to free this allocation
+** by eventually passing the *pazCol value to sqlite3_free().
+**
+** If the table cannot be found, an error code is returned and the output
+** variables are undefined. Or, if an OOM is encountered, SQLITE_NOMEM is
+** returned (and the output variables are undefined).
+*/
+static int fts3ContentColumns(
+ sqlite3 *db, /* Database handle */
+ const char *zDb, /* Name of db (i.e. "main", "temp" etc.) */
+ const char *zTbl, /* Name of content table */
+ const char ***pazCol, /* OUT: Malloc'd array of column names */
+ int *pnCol, /* OUT: Size of array *pazCol */
+ int *pnStr, /* OUT: Bytes of string content */
+ char **pzErr /* OUT: error message */
+){
+ int rc = SQLITE_OK; /* Return code */
+ char *zSql; /* "SELECT *" statement on zTbl */
+ sqlite3_stmt *pStmt = 0; /* Compiled version of zSql */
+
+ zSql = sqlite3_mprintf("SELECT * FROM %Q.%Q", zDb, zTbl);
+ if( !zSql ){
+ rc = SQLITE_NOMEM;
+ }else{
+ rc = sqlite3_prepare(db, zSql, -1, &pStmt, 0);
+ if( rc!=SQLITE_OK ){
+ sqlite3Fts3ErrMsg(pzErr, "%s", sqlite3_errmsg(db));
+ }
+ }
+ sqlite3_free(zSql);
+
+ if( rc==SQLITE_OK ){
+ const char **azCol; /* Output array */
+ sqlite3_int64 nStr = 0; /* Size of all column names (incl. 0x00) */
+ int nCol; /* Number of table columns */
+ int i; /* Used to iterate through columns */
+
+ /* Loop through the returned columns. Set nStr to the number of bytes of
+ ** space required to store a copy of each column name, including the
+ ** nul-terminator byte. */
+ nCol = sqlite3_column_count(pStmt);
+ for(i=0; i module name ("fts3" or "fts4")
+** argv[1] -> database name
+** argv[2] -> table name
+** argv[...] -> "column name" and other module argument fields.
+*/
+static int fts3InitVtab(
+ int isCreate, /* True for xCreate, false for xConnect */
+ sqlite3 *db, /* The SQLite database connection */
+ void *pAux, /* Hash table containing tokenizers */
+ int argc, /* Number of elements in argv array */
+ const char * const *argv, /* xCreate/xConnect argument array */
+ sqlite3_vtab **ppVTab, /* Write the resulting vtab structure here */
+ char **pzErr /* Write any error message here */
+){
+ Fts3Hash *pHash = &((Fts3HashWrapper*)pAux)->hash;
+ Fts3Table *p = 0; /* Pointer to allocated vtab */
+ int rc = SQLITE_OK; /* Return code */
+ int i; /* Iterator variable */
+ sqlite3_int64 nByte; /* Size of allocation used for *p */
+ int iCol; /* Column index */
+ int nString = 0; /* Bytes required to hold all column names */
+ int nCol = 0; /* Number of columns in the FTS table */
+ char *zCsr; /* Space for holding column names */
+ int nDb; /* Bytes required to hold database name */
+ int nName; /* Bytes required to hold table name */
+ int isFts4 = (argv[0][3]=='4'); /* True for FTS4, false for FTS3 */
+ const char **aCol; /* Array of column names */
+ sqlite3_tokenizer *pTokenizer = 0; /* Tokenizer for this table */
+
+ int nIndex = 0; /* Size of aIndex[] array */
+ struct Fts3Index *aIndex = 0; /* Array of indexes for this table */
+
+ /* The results of parsing supported FTS4 key=value options: */
+ int bNoDocsize = 0; /* True to omit %_docsize table */
+ int bDescIdx = 0; /* True to store descending indexes */
+ char *zPrefix = 0; /* Prefix parameter value (or NULL) */
+ char *zCompress = 0; /* compress=? parameter (or NULL) */
+ char *zUncompress = 0; /* uncompress=? parameter (or NULL) */
+ char *zContent = 0; /* content=? parameter (or NULL) */
+ char *zLanguageid = 0; /* languageid=? parameter (or NULL) */
+ char **azNotindexed = 0; /* The set of notindexed= columns */
+ int nNotindexed = 0; /* Size of azNotindexed[] array */
+
+ assert( strlen(argv[0])==4 );
+ assert( (sqlite3_strnicmp(argv[0], "fts4", 4)==0 && isFts4)
+ || (sqlite3_strnicmp(argv[0], "fts3", 4)==0 && !isFts4)
+ );
+
+ nDb = (int)strlen(argv[1]) + 1;
+ nName = (int)strlen(argv[2]) + 1;
+
+ nByte = sizeof(const char *) * (argc-2);
+ aCol = (const char **)sqlite3_malloc64(nByte);
+ if( aCol ){
+ memset((void*)aCol, 0, nByte);
+ azNotindexed = (char **)sqlite3_malloc64(nByte);
+ }
+ if( azNotindexed ){
+ memset(azNotindexed, 0, nByte);
+ }
+ if( !aCol || !azNotindexed ){
+ rc = SQLITE_NOMEM;
+ goto fts3_init_out;
+ }
+
+ /* Loop through all of the arguments passed by the user to the FTS3/4
+ ** module (i.e. all the column names and special arguments). This loop
+ ** does the following:
+ **
+ ** + Figures out the number of columns the FTSX table will have, and
+ ** the number of bytes of space that must be allocated to store copies
+ ** of the column names.
+ **
+ ** + If there is a tokenizer specification included in the arguments,
+ ** initializes the tokenizer pTokenizer.
+ */
+ for(i=3; rc==SQLITE_OK && i8
+ && 0==sqlite3_strnicmp(z, "tokenize", 8)
+ && 0==sqlite3Fts3IsIdChar(z[8])
+ ){
+ rc = sqlite3Fts3InitTokenizer(pHash, &z[9], &pTokenizer, pzErr);
+ }
+
+ /* Check if it is an FTS4 special argument. */
+ else if( isFts4 && fts3IsSpecialColumn(z, &nKey, &zVal) ){
+ struct Fts4Option {
+ const char *zOpt;
+ int nOpt;
+ } aFts4Opt[] = {
+ { "matchinfo", 9 }, /* 0 -> MATCHINFO */
+ { "prefix", 6 }, /* 1 -> PREFIX */
+ { "compress", 8 }, /* 2 -> COMPRESS */
+ { "uncompress", 10 }, /* 3 -> UNCOMPRESS */
+ { "order", 5 }, /* 4 -> ORDER */
+ { "content", 7 }, /* 5 -> CONTENT */
+ { "languageid", 10 }, /* 6 -> LANGUAGEID */
+ { "notindexed", 10 } /* 7 -> NOTINDEXED */
+ };
+
+ int iOpt;
+ if( !zVal ){
+ rc = SQLITE_NOMEM;
+ }else{
+ for(iOpt=0; iOptnOpt && !sqlite3_strnicmp(z, pOp->zOpt, pOp->nOpt) ){
+ break;
+ }
+ }
+ switch( iOpt ){
+ case 0: /* MATCHINFO */
+ if( strlen(zVal)!=4 || sqlite3_strnicmp(zVal, "fts3", 4) ){
+ sqlite3Fts3ErrMsg(pzErr, "unrecognized matchinfo: %s", zVal);
+ rc = SQLITE_ERROR;
+ }
+ bNoDocsize = 1;
+ break;
+
+ case 1: /* PREFIX */
+ sqlite3_free(zPrefix);
+ zPrefix = zVal;
+ zVal = 0;
+ break;
+
+ case 2: /* COMPRESS */
+ sqlite3_free(zCompress);
+ zCompress = zVal;
+ zVal = 0;
+ break;
+
+ case 3: /* UNCOMPRESS */
+ sqlite3_free(zUncompress);
+ zUncompress = zVal;
+ zVal = 0;
+ break;
+
+ case 4: /* ORDER */
+ if( (strlen(zVal)!=3 || sqlite3_strnicmp(zVal, "asc", 3))
+ && (strlen(zVal)!=4 || sqlite3_strnicmp(zVal, "desc", 4))
+ ){
+ sqlite3Fts3ErrMsg(pzErr, "unrecognized order: %s", zVal);
+ rc = SQLITE_ERROR;
+ }
+ bDescIdx = (zVal[0]=='d' || zVal[0]=='D');
+ break;
+
+ case 5: /* CONTENT */
+ sqlite3_free(zContent);
+ zContent = zVal;
+ zVal = 0;
+ break;
+
+ case 6: /* LANGUAGEID */
+ assert( iOpt==6 );
+ sqlite3_free(zLanguageid);
+ zLanguageid = zVal;
+ zVal = 0;
+ break;
+
+ case 7: /* NOTINDEXED */
+ azNotindexed[nNotindexed++] = zVal;
+ zVal = 0;
+ break;
+
+ default:
+ assert( iOpt==SizeofArray(aFts4Opt) );
+ sqlite3Fts3ErrMsg(pzErr, "unrecognized parameter: %s", z);
+ rc = SQLITE_ERROR;
+ break;
+ }
+ sqlite3_free(zVal);
+ }
+ }
+
+ /* Otherwise, the argument is a column name. */
+ else {
+ nString += (int)(strlen(z) + 1);
+ aCol[nCol++] = z;
+ }
+ }
+
+ /* If a content=xxx option was specified, the following:
+ **
+ ** 1. Ignore any compress= and uncompress= options.
+ **
+ ** 2. If no column names were specified as part of the CREATE VIRTUAL
+ ** TABLE statement, use all columns from the content table.
+ */
+ if( rc==SQLITE_OK && zContent ){
+ sqlite3_free(zCompress);
+ sqlite3_free(zUncompress);
+ zCompress = 0;
+ zUncompress = 0;
+ if( nCol==0 ){
+ sqlite3_free((void*)aCol);
+ aCol = 0;
+ rc = fts3ContentColumns(db, argv[1], zContent,&aCol,&nCol,&nString,pzErr);
+
+ /* If a languageid= option was specified, remove the language id
+ ** column from the aCol[] array. */
+ if( rc==SQLITE_OK && zLanguageid ){
+ int j;
+ for(j=0; jdb = db;
+ p->nColumn = nCol;
+ p->nPendingData = 0;
+ p->azColumn = (char **)&p[1];
+ p->pTokenizer = pTokenizer;
+ p->nMaxPendingData = FTS3_MAX_PENDING_DATA;
+ p->bHasDocsize = (isFts4 && bNoDocsize==0);
+ p->bHasStat = (u8)isFts4;
+ p->bFts4 = (u8)isFts4;
+ p->bDescIdx = (u8)bDescIdx;
+ p->nAutoincrmerge = 0xff; /* 0xff means setting unknown */
+ p->zContentTbl = zContent;
+ p->zLanguageid = zLanguageid;
+ zContent = 0;
+ zLanguageid = 0;
+ TESTONLY( p->inTransaction = -1 );
+ TESTONLY( p->mxSavepoint = -1 );
+
+ p->aIndex = (struct Fts3Index *)&p->azColumn[nCol];
+ memcpy(p->aIndex, aIndex, sizeof(struct Fts3Index) * nIndex);
+ p->nIndex = nIndex;
+ for(i=0; iaIndex[i].hPending, FTS3_HASH_STRING, 1);
+ }
+ p->abNotindexed = (u8 *)&p->aIndex[nIndex];
+
+ /* Fill in the zName and zDb fields of the vtab structure. */
+ zCsr = (char *)&p->abNotindexed[nCol];
+ p->zName = zCsr;
+ memcpy(zCsr, argv[2], nName);
+ zCsr += nName;
+ p->zDb = zCsr;
+ memcpy(zCsr, argv[1], nDb);
+ zCsr += nDb;
+
+ /* Fill in the azColumn array */
+ for(iCol=0; iCol0 ){
+ memcpy(zCsr, z, n);
+ }
+ zCsr[n] = '\0';
+ sqlite3Fts3Dequote(zCsr);
+ p->azColumn[iCol] = zCsr;
+ zCsr += n+1;
+ assert( zCsr <= &((char *)p)[nByte] );
+ }
+
+ /* Fill in the abNotindexed array */
+ for(iCol=0; iColazColumn[iCol]);
+ for(i=0; iazColumn[iCol], zNot, n)
+ ){
+ p->abNotindexed[iCol] = 1;
+ sqlite3_free(zNot);
+ azNotindexed[i] = 0;
+ }
+ }
+ }
+ for(i=0; izReadExprlist = fts3ReadExprList(p, zUncompress, &rc);
+ p->zWriteExprlist = fts3WriteExprList(p, zCompress, &rc);
+ if( rc!=SQLITE_OK ) goto fts3_init_out;
+
+ /* If this is an xCreate call, create the underlying tables in the
+ ** database. TODO: For xConnect(), it could verify that said tables exist.
+ */
+ if( isCreate ){
+ rc = fts3CreateTables(p);
+ }
+
+ /* Check to see if a legacy fts3 table has been "upgraded" by the
+ ** addition of a %_stat table so that it can use incremental merge.
+ */
+ if( !isFts4 && !isCreate ){
+ p->bHasStat = 2;
+ }
+
+ /* Figure out the page-size for the database. This is required in order to
+ ** estimate the cost of loading large doclists from the database. */
+ fts3DatabasePageSize(&rc, p);
+ p->nNodeSize = p->nPgsz-35;
+
+#if defined(SQLITE_DEBUG)||defined(SQLITE_TEST)
+ p->nMergeCount = FTS3_MERGE_COUNT;
+#endif
+
+ /* Declare the table schema to SQLite. */
+ fts3DeclareVtab(&rc, p);
+
+fts3_init_out:
+ sqlite3_free(zPrefix);
+ sqlite3_free(aIndex);
+ sqlite3_free(zCompress);
+ sqlite3_free(zUncompress);
+ sqlite3_free(zContent);
+ sqlite3_free(zLanguageid);
+ for(i=0; ipModule->xDestroy(pTokenizer);
+ }
+ }else{
+ assert( p->pSegments==0 );
+ *ppVTab = &p->base;
+ }
+ return rc;
+}
+
+/*
+** The xConnect() and xCreate() methods for the virtual table. All the
+** work is done in function fts3InitVtab().
+*/
+static int fts3ConnectMethod(
+ sqlite3 *db, /* Database connection */
+ void *pAux, /* Pointer to tokenizer hash table */
+ int argc, /* Number of elements in argv array */
+ const char * const *argv, /* xCreate/xConnect argument array */
+ sqlite3_vtab **ppVtab, /* OUT: New sqlite3_vtab object */
+ char **pzErr /* OUT: sqlite3_malloc'd error message */
+){
+ return fts3InitVtab(0, db, pAux, argc, argv, ppVtab, pzErr);
+}
+static int fts3CreateMethod(
+ sqlite3 *db, /* Database connection */
+ void *pAux, /* Pointer to tokenizer hash table */
+ int argc, /* Number of elements in argv array */
+ const char * const *argv, /* xCreate/xConnect argument array */
+ sqlite3_vtab **ppVtab, /* OUT: New sqlite3_vtab object */
+ char **pzErr /* OUT: sqlite3_malloc'd error message */
+){
+ return fts3InitVtab(1, db, pAux, argc, argv, ppVtab, pzErr);
+}
+
+/*
+** Set the pIdxInfo->estimatedRows variable to nRow. Unless this
+** extension is currently being used by a version of SQLite too old to
+** support estimatedRows. In that case this function is a no-op.
+*/
+static void fts3SetEstimatedRows(sqlite3_index_info *pIdxInfo, i64 nRow){
+#if SQLITE_VERSION_NUMBER>=3008002
+ if( sqlite3_libversion_number()>=3008002 ){
+ pIdxInfo->estimatedRows = nRow;
+ }
+#endif
+}
+
+/*
+** Set the SQLITE_INDEX_SCAN_UNIQUE flag in pIdxInfo->flags. Unless this
+** extension is currently being used by a version of SQLite too old to
+** support index-info flags. In that case this function is a no-op.
+*/
+static void fts3SetUniqueFlag(sqlite3_index_info *pIdxInfo){
+#if SQLITE_VERSION_NUMBER>=3008012
+ if( sqlite3_libversion_number()>=3008012 ){
+ pIdxInfo->idxFlags |= SQLITE_INDEX_SCAN_UNIQUE;
+ }
+#endif
+}
+
+/*
+** Implementation of the xBestIndex method for FTS3 tables. There
+** are three possible strategies, in order of preference:
+**
+** 1. Direct lookup by rowid or docid.
+** 2. Full-text search using a MATCH operator on a non-docid column.
+** 3. Linear scan of %_content table.
+*/
+static int fts3BestIndexMethod(sqlite3_vtab *pVTab, sqlite3_index_info *pInfo){
+ Fts3Table *p = (Fts3Table *)pVTab;
+ int i; /* Iterator variable */
+ int iCons = -1; /* Index of constraint to use */
+
+ int iLangidCons = -1; /* Index of langid=x constraint, if present */
+ int iDocidGe = -1; /* Index of docid>=x constraint, if present */
+ int iDocidLe = -1; /* Index of docid<=x constraint, if present */
+ int iIdx;
+
+ if( p->bLock ){
+ return SQLITE_ERROR;
+ }
+
+ /* By default use a full table scan. This is an expensive option,
+ ** so search through the constraints to see if a more efficient
+ ** strategy is possible.
+ */
+ pInfo->idxNum = FTS3_FULLSCAN_SEARCH;
+ pInfo->estimatedCost = 5000000;
+ for(i=0; inConstraint; i++){
+ int bDocid; /* True if this constraint is on docid */
+ struct sqlite3_index_constraint *pCons = &pInfo->aConstraint[i];
+ if( pCons->usable==0 ){
+ if( pCons->op==SQLITE_INDEX_CONSTRAINT_MATCH ){
+ /* There exists an unusable MATCH constraint. This means that if
+ ** the planner does elect to use the results of this call as part
+ ** of the overall query plan the user will see an "unable to use
+ ** function MATCH in the requested context" error. To discourage
+ ** this, return a very high cost here. */
+ pInfo->idxNum = FTS3_FULLSCAN_SEARCH;
+ pInfo->estimatedCost = 1e50;
+ fts3SetEstimatedRows(pInfo, ((sqlite3_int64)1) << 50);
+ return SQLITE_OK;
+ }
+ continue;
+ }
+
+ bDocid = (pCons->iColumn<0 || pCons->iColumn==p->nColumn+1);
+
+ /* A direct lookup on the rowid or docid column. Assign a cost of 1.0. */
+ if( iCons<0 && pCons->op==SQLITE_INDEX_CONSTRAINT_EQ && bDocid ){
+ pInfo->idxNum = FTS3_DOCID_SEARCH;
+ pInfo->estimatedCost = 1.0;
+ iCons = i;
+ }
+
+ /* A MATCH constraint. Use a full-text search.
+ **
+ ** If there is more than one MATCH constraint available, use the first
+ ** one encountered. If there is both a MATCH constraint and a direct
+ ** rowid/docid lookup, prefer the MATCH strategy. This is done even
+ ** though the rowid/docid lookup is faster than a MATCH query, selecting
+ ** it would lead to an "unable to use function MATCH in the requested
+ ** context" error.
+ */
+ if( pCons->op==SQLITE_INDEX_CONSTRAINT_MATCH
+ && pCons->iColumn>=0 && pCons->iColumn<=p->nColumn
+ ){
+ pInfo->idxNum = FTS3_FULLTEXT_SEARCH + pCons->iColumn;
+ pInfo->estimatedCost = 2.0;
+ iCons = i;
+ }
+
+ /* Equality constraint on the langid column */
+ if( pCons->op==SQLITE_INDEX_CONSTRAINT_EQ
+ && pCons->iColumn==p->nColumn + 2
+ ){
+ iLangidCons = i;
+ }
+
+ if( bDocid ){
+ switch( pCons->op ){
+ case SQLITE_INDEX_CONSTRAINT_GE:
+ case SQLITE_INDEX_CONSTRAINT_GT:
+ iDocidGe = i;
+ break;
+
+ case SQLITE_INDEX_CONSTRAINT_LE:
+ case SQLITE_INDEX_CONSTRAINT_LT:
+ iDocidLe = i;
+ break;
+ }
+ }
+ }
+
+ /* If using a docid=? or rowid=? strategy, set the UNIQUE flag. */
+ if( pInfo->idxNum==FTS3_DOCID_SEARCH ) fts3SetUniqueFlag(pInfo);
+
+ iIdx = 1;
+ if( iCons>=0 ){
+ pInfo->aConstraintUsage[iCons].argvIndex = iIdx++;
+ pInfo->aConstraintUsage[iCons].omit = 1;
+ }
+ if( iLangidCons>=0 ){
+ pInfo->idxNum |= FTS3_HAVE_LANGID;
+ pInfo->aConstraintUsage[iLangidCons].argvIndex = iIdx++;
+ }
+ if( iDocidGe>=0 ){
+ pInfo->idxNum |= FTS3_HAVE_DOCID_GE;
+ pInfo->aConstraintUsage[iDocidGe].argvIndex = iIdx++;
+ }
+ if( iDocidLe>=0 ){
+ pInfo->idxNum |= FTS3_HAVE_DOCID_LE;
+ pInfo->aConstraintUsage[iDocidLe].argvIndex = iIdx++;
+ }
+
+ /* Regardless of the strategy selected, FTS can deliver rows in rowid (or
+ ** docid) order. Both ascending and descending are possible.
+ */
+ if( pInfo->nOrderBy==1 ){
+ struct sqlite3_index_orderby *pOrder = &pInfo->aOrderBy[0];
+ if( pOrder->iColumn<0 || pOrder->iColumn==p->nColumn+1 ){
+ if( pOrder->desc ){
+ pInfo->idxStr = "DESC";
+ }else{
+ pInfo->idxStr = "ASC";
+ }
+ pInfo->orderByConsumed = 1;
+ }
+ }
+
+ assert( p->pSegments==0 );
+ return SQLITE_OK;
+}
+
+/*
+** Implementation of xOpen method.
+*/
+static int fts3OpenMethod(sqlite3_vtab *pVTab, sqlite3_vtab_cursor **ppCsr){
+ sqlite3_vtab_cursor *pCsr; /* Allocated cursor */
+
+ UNUSED_PARAMETER(pVTab);
+
+ /* Allocate a buffer large enough for an Fts3Cursor structure. If the
+ ** allocation succeeds, zero it and return SQLITE_OK. Otherwise,
+ ** if the allocation fails, return SQLITE_NOMEM.
+ */
+ *ppCsr = pCsr = (sqlite3_vtab_cursor *)sqlite3_malloc(sizeof(Fts3Cursor));
+ if( !pCsr ){
+ return SQLITE_NOMEM;
+ }
+ memset(pCsr, 0, sizeof(Fts3Cursor));
+ return SQLITE_OK;
+}
+
+/*
+** Finalize the statement handle at pCsr->pStmt.
+**
+** Or, if that statement handle is one created by fts3CursorSeekStmt(),
+** and the Fts3Table.pSeekStmt slot is currently NULL, save the statement
+** pointer there instead of finalizing it.
+*/
+static void fts3CursorFinalizeStmt(Fts3Cursor *pCsr){
+ if( pCsr->bSeekStmt ){
+ Fts3Table *p = (Fts3Table *)pCsr->base.pVtab;
+ if( p->pSeekStmt==0 ){
+ p->pSeekStmt = pCsr->pStmt;
+ sqlite3_reset(pCsr->pStmt);
+ pCsr->pStmt = 0;
+ }
+ pCsr->bSeekStmt = 0;
+ }
+ sqlite3_finalize(pCsr->pStmt);
+}
+
+/*
+** Free all resources currently held by the cursor passed as the only
+** argument.
+*/
+static void fts3ClearCursor(Fts3Cursor *pCsr){
+ fts3CursorFinalizeStmt(pCsr);
+ sqlite3Fts3FreeDeferredTokens(pCsr);
+ sqlite3_free(pCsr->aDoclist);
+ sqlite3Fts3MIBufferFree(pCsr->pMIBuffer);
+ sqlite3Fts3ExprFree(pCsr->pExpr);
+ memset(&(&pCsr->base)[1], 0, sizeof(Fts3Cursor)-sizeof(sqlite3_vtab_cursor));
+}
+
+/*
+** Close the cursor. For additional information see the documentation
+** on the xClose method of the virtual table interface.
+*/
+static int fts3CloseMethod(sqlite3_vtab_cursor *pCursor){
+ Fts3Cursor *pCsr = (Fts3Cursor *)pCursor;
+ assert( ((Fts3Table *)pCsr->base.pVtab)->pSegments==0 );
+ fts3ClearCursor(pCsr);
+ assert( ((Fts3Table *)pCsr->base.pVtab)->pSegments==0 );
+ sqlite3_free(pCsr);
+ return SQLITE_OK;
+}
+
+/*
+** If pCsr->pStmt has not been prepared (i.e. if pCsr->pStmt==0), then
+** compose and prepare an SQL statement of the form:
+**
+** "SELECT FROM %_content WHERE rowid = ?"
+**
+** (or the equivalent for a content=xxx table) and set pCsr->pStmt to
+** it. If an error occurs, return an SQLite error code.
+*/
+static int fts3CursorSeekStmt(Fts3Cursor *pCsr){
+ int rc = SQLITE_OK;
+ if( pCsr->pStmt==0 ){
+ Fts3Table *p = (Fts3Table *)pCsr->base.pVtab;
+ char *zSql;
+ if( p->pSeekStmt ){
+ pCsr->pStmt = p->pSeekStmt;
+ p->pSeekStmt = 0;
+ }else{
+ zSql = sqlite3_mprintf("SELECT %s WHERE rowid = ?", p->zReadExprlist);
+ if( !zSql ) return SQLITE_NOMEM;
+ p->bLock++;
+ rc = sqlite3_prepare_v3(
+ p->db, zSql,-1,SQLITE_PREPARE_PERSISTENT,&pCsr->pStmt,0
+ );
+ p->bLock--;
+ sqlite3_free(zSql);
+ }
+ if( rc==SQLITE_OK ) pCsr->bSeekStmt = 1;
+ }
+ return rc;
+}
+
+/*
+** Position the pCsr->pStmt statement so that it is on the row
+** of the %_content table that contains the last match. Return
+** SQLITE_OK on success.
+*/
+static int fts3CursorSeek(sqlite3_context *pContext, Fts3Cursor *pCsr){
+ int rc = SQLITE_OK;
+ if( pCsr->isRequireSeek ){
+ rc = fts3CursorSeekStmt(pCsr);
+ if( rc==SQLITE_OK ){
+ Fts3Table *pTab = (Fts3Table*)pCsr->base.pVtab;
+ pTab->bLock++;
+ sqlite3_bind_int64(pCsr->pStmt, 1, pCsr->iPrevId);
+ pCsr->isRequireSeek = 0;
+ if( SQLITE_ROW==sqlite3_step(pCsr->pStmt) ){
+ pTab->bLock--;
+ return SQLITE_OK;
+ }else{
+ pTab->bLock--;
+ rc = sqlite3_reset(pCsr->pStmt);
+ if( rc==SQLITE_OK && ((Fts3Table *)pCsr->base.pVtab)->zContentTbl==0 ){
+ /* If no row was found and no error has occurred, then the %_content
+ ** table is missing a row that is present in the full-text index.
+ ** The data structures are corrupt. */
+ rc = FTS_CORRUPT_VTAB;
+ pCsr->isEof = 1;
+ }
+ }
+ }
+ }
+
+ if( rc!=SQLITE_OK && pContext ){
+ sqlite3_result_error_code(pContext, rc);
+ }
+ return rc;
+}
+
+/*
+** This function is used to process a single interior node when searching
+** a b-tree for a term or term prefix. The node data is passed to this
+** function via the zNode/nNode parameters. The term to search for is
+** passed in zTerm/nTerm.
+**
+** If piFirst is not NULL, then this function sets *piFirst to the blockid
+** of the child node that heads the sub-tree that may contain the term.
+**
+** If piLast is not NULL, then *piLast is set to the right-most child node
+** that heads a sub-tree that may contain a term for which zTerm/nTerm is
+** a prefix.
+**
+** If an OOM error occurs, SQLITE_NOMEM is returned. Otherwise, SQLITE_OK.
+*/
+static int fts3ScanInteriorNode(
+ const char *zTerm, /* Term to select leaves for */
+ int nTerm, /* Size of term zTerm in bytes */
+ const char *zNode, /* Buffer containing segment interior node */
+ int nNode, /* Size of buffer at zNode */
+ sqlite3_int64 *piFirst, /* OUT: Selected child node */
+ sqlite3_int64 *piLast /* OUT: Selected child node */
+){
+ int rc = SQLITE_OK; /* Return code */
+ const char *zCsr = zNode; /* Cursor to iterate through node */
+ const char *zEnd = &zCsr[nNode];/* End of interior node buffer */
+ char *zBuffer = 0; /* Buffer to load terms into */
+ i64 nAlloc = 0; /* Size of allocated buffer */
+ int isFirstTerm = 1; /* True when processing first term on page */
+ u64 iChild; /* Block id of child node to descend to */
+ int nBuffer = 0; /* Total term size */
+
+ /* Skip over the 'height' varint that occurs at the start of every
+ ** interior node. Then load the blockid of the left-child of the b-tree
+ ** node into variable iChild.
+ **
+ ** Even if the data structure on disk is corrupted, this (reading two
+ ** varints from the buffer) does not risk an overread. If zNode is a
+ ** root node, then the buffer comes from a SELECT statement. SQLite does
+ ** not make this guarantee explicitly, but in practice there are always
+ ** either more than 20 bytes of allocated space following the nNode bytes of
+ ** contents, or two zero bytes. Or, if the node is read from the %_segments
+ ** table, then there are always 20 bytes of zeroed padding following the
+ ** nNode bytes of content (see sqlite3Fts3ReadBlock() for details).
+ */
+ zCsr += sqlite3Fts3GetVarintU(zCsr, &iChild);
+ zCsr += sqlite3Fts3GetVarintU(zCsr, &iChild);
+ if( zCsr>zEnd ){
+ return FTS_CORRUPT_VTAB;
+ }
+
+ while( zCsrnBuffer ){
+ rc = FTS_CORRUPT_VTAB;
+ goto finish_scan;
+ }
+ }
+ isFirstTerm = 0;
+ zCsr += fts3GetVarint32(zCsr, &nSuffix);
+
+ assert( nPrefix>=0 && nSuffix>=0 );
+ if( nPrefix>zCsr-zNode || nSuffix>zEnd-zCsr || nSuffix==0 ){
+ rc = FTS_CORRUPT_VTAB;
+ goto finish_scan;
+ }
+ if( (i64)nPrefix+nSuffix>nAlloc ){
+ char *zNew;
+ nAlloc = ((i64)nPrefix+nSuffix) * 2;
+ zNew = (char *)sqlite3_realloc64(zBuffer, nAlloc);
+ if( !zNew ){
+ rc = SQLITE_NOMEM;
+ goto finish_scan;
+ }
+ zBuffer = zNew;
+ }
+ assert( zBuffer );
+ memcpy(&zBuffer[nPrefix], zCsr, nSuffix);
+ nBuffer = nPrefix + nSuffix;
+ zCsr += nSuffix;
+
+ /* Compare the term we are searching for with the term just loaded from
+ ** the interior node. If the specified term is greater than or equal
+ ** to the term from the interior node, then all terms on the sub-tree
+ ** headed by node iChild are smaller than zTerm. No need to search
+ ** iChild.
+ **
+ ** If the interior node term is larger than the specified term, then
+ ** the tree headed by iChild may contain the specified term.
+ */
+ cmp = memcmp(zTerm, zBuffer, (nBuffer>nTerm ? nTerm : nBuffer));
+ if( piFirst && (cmp<0 || (cmp==0 && nBuffer>nTerm)) ){
+ *piFirst = (i64)iChild;
+ piFirst = 0;
+ }
+
+ if( piLast && cmp<0 ){
+ *piLast = (i64)iChild;
+ piLast = 0;
+ }
+
+ iChild++;
+ };
+
+ if( piFirst ) *piFirst = (i64)iChild;
+ if( piLast ) *piLast = (i64)iChild;
+
+ finish_scan:
+ sqlite3_free(zBuffer);
+ return rc;
+}
+
+
+/*
+** The buffer pointed to by argument zNode (size nNode bytes) contains an
+** interior node of a b-tree segment. The zTerm buffer (size nTerm bytes)
+** contains a term. This function searches the sub-tree headed by the zNode
+** node for the range of leaf nodes that may contain the specified term
+** or terms for which the specified term is a prefix.
+**
+** If piLeaf is not NULL, then *piLeaf is set to the blockid of the
+** left-most leaf node in the tree that may contain the specified term.
+** If piLeaf2 is not NULL, then *piLeaf2 is set to the blockid of the
+** right-most leaf node that may contain a term for which the specified
+** term is a prefix.
+**
+** It is possible that the range of returned leaf nodes does not contain
+** the specified term or any terms for which it is a prefix. However, if the
+** segment does contain any such terms, they are stored within the identified
+** range. Because this function only inspects interior segment nodes (and
+** never loads leaf nodes into memory), it is not possible to be sure.
+**
+** If an error occurs, an error code other than SQLITE_OK is returned.
+*/
+static int fts3SelectLeaf(
+ Fts3Table *p, /* Virtual table handle */
+ const char *zTerm, /* Term to select leaves for */
+ int nTerm, /* Size of term zTerm in bytes */
+ const char *zNode, /* Buffer containing segment interior node */
+ int nNode, /* Size of buffer at zNode */
+ sqlite3_int64 *piLeaf, /* Selected leaf node */
+ sqlite3_int64 *piLeaf2 /* Selected leaf node */
+){
+ int rc = SQLITE_OK; /* Return code */
+ int iHeight; /* Height of this node in tree */
+
+ assert( piLeaf || piLeaf2 );
+
+ fts3GetVarint32(zNode, &iHeight);
+ rc = fts3ScanInteriorNode(zTerm, nTerm, zNode, nNode, piLeaf, piLeaf2);
+ assert_fts3_nc( !piLeaf2 || !piLeaf || rc!=SQLITE_OK || (*piLeaf<=*piLeaf2) );
+
+ if( rc==SQLITE_OK && iHeight>1 ){
+ char *zBlob = 0; /* Blob read from %_segments table */
+ int nBlob = 0; /* Size of zBlob in bytes */
+
+ if( piLeaf && piLeaf2 && (*piLeaf!=*piLeaf2) ){
+ rc = sqlite3Fts3ReadBlock(p, *piLeaf, &zBlob, &nBlob, 0);
+ if( rc==SQLITE_OK ){
+ rc = fts3SelectLeaf(p, zTerm, nTerm, zBlob, nBlob, piLeaf, 0);
+ }
+ sqlite3_free(zBlob);
+ piLeaf = 0;
+ zBlob = 0;
+ }
+
+ if( rc==SQLITE_OK ){
+ rc = sqlite3Fts3ReadBlock(p, piLeaf?*piLeaf:*piLeaf2, &zBlob, &nBlob, 0);
+ }
+ if( rc==SQLITE_OK ){
+ int iNewHeight = 0;
+ fts3GetVarint32(zBlob, &iNewHeight);
+ if( iNewHeight>=iHeight ){
+ rc = FTS_CORRUPT_VTAB;
+ }else{
+ rc = fts3SelectLeaf(p, zTerm, nTerm, zBlob, nBlob, piLeaf, piLeaf2);
+ }
+ }
+ sqlite3_free(zBlob);
+ }
+
+ return rc;
+}
+
+/*
+** This function is used to create delta-encoded serialized lists of FTS3
+** varints. Each call to this function appends a single varint to a list.
+*/
+static void fts3PutDeltaVarint(
+ char **pp, /* IN/OUT: Output pointer */
+ sqlite3_int64 *piPrev, /* IN/OUT: Previous value written to list */
+ sqlite3_int64 iVal /* Write this value to the list */
+){
+ assert_fts3_nc( iVal-*piPrev > 0 || (*piPrev==0 && iVal==0) );
+ *pp += sqlite3Fts3PutVarint(*pp, iVal-*piPrev);
+ *piPrev = iVal;
+}
+
+/*
+** When this function is called, *ppPoslist is assumed to point to the
+** start of a position-list. After it returns, *ppPoslist points to the
+** first byte after the position-list.
+**
+** A position list is list of positions (delta encoded) and columns for
+** a single document record of a doclist. So, in other words, this
+** routine advances *ppPoslist so that it points to the next docid in
+** the doclist, or to the first byte past the end of the doclist.
+**
+** If pp is not NULL, then the contents of the position list are copied
+** to *pp. *pp is set to point to the first byte past the last byte copied
+** before this function returns.
+*/
+static void fts3PoslistCopy(char **pp, char **ppPoslist){
+ char *pEnd = *ppPoslist;
+ char c = 0;
+
+ /* The end of a position list is marked by a zero encoded as an FTS3
+ ** varint. A single POS_END (0) byte. Except, if the 0 byte is preceded by
+ ** a byte with the 0x80 bit set, then it is not a varint 0, but the tail
+ ** of some other, multi-byte, value.
+ **
+ ** The following while-loop moves pEnd to point to the first byte that is not
+ ** immediately preceded by a byte with the 0x80 bit set. Then increments
+ ** pEnd once more so that it points to the byte immediately following the
+ ** last byte in the position-list.
+ */
+ while( *pEnd | c ){
+ c = *pEnd++ & 0x80;
+ testcase( c!=0 && (*pEnd)==0 );
+ }
+ pEnd++; /* Advance past the POS_END terminator byte */
+
+ if( pp ){
+ int n = (int)(pEnd - *ppPoslist);
+ char *p = *pp;
+ memcpy(p, *ppPoslist, n);
+ p += n;
+ *pp = p;
+ }
+ *ppPoslist = pEnd;
+}
+
+/*
+** When this function is called, *ppPoslist is assumed to point to the
+** start of a column-list. After it returns, *ppPoslist points to the
+** to the terminator (POS_COLUMN or POS_END) byte of the column-list.
+**
+** A column-list is list of delta-encoded positions for a single column
+** within a single document within a doclist.
+**
+** The column-list is terminated either by a POS_COLUMN varint (1) or
+** a POS_END varint (0). This routine leaves *ppPoslist pointing to
+** the POS_COLUMN or POS_END that terminates the column-list.
+**
+** If pp is not NULL, then the contents of the column-list are copied
+** to *pp. *pp is set to point to the first byte past the last byte copied
+** before this function returns. The POS_COLUMN or POS_END terminator
+** is not copied into *pp.
+*/
+static void fts3ColumnlistCopy(char **pp, char **ppPoslist){
+ char *pEnd = *ppPoslist;
+ char c = 0;
+
+ /* A column-list is terminated by either a 0x01 or 0x00 byte that is
+ ** not part of a multi-byte varint.
+ */
+ while( 0xFE & (*pEnd | c) ){
+ c = *pEnd++ & 0x80;
+ testcase( c!=0 && ((*pEnd)&0xfe)==0 );
+ }
+ if( pp ){
+ int n = (int)(pEnd - *ppPoslist);
+ char *p = *pp;
+ memcpy(p, *ppPoslist, n);
+ p += n;
+ *pp = p;
+ }
+ *ppPoslist = pEnd;
+}
+
+/*
+** Value used to signify the end of an position-list. This must be
+** as large or larger than any value that might appear on the
+** position-list, even a position list that has been corrupted.
+*/
+#define POSITION_LIST_END LARGEST_INT64
+
+/*
+** This function is used to help parse position-lists. When this function is
+** called, *pp may point to the start of the next varint in the position-list
+** being parsed, or it may point to 1 byte past the end of the position-list
+** (in which case **pp will be a terminator bytes POS_END (0) or
+** (1)).
+**
+** If *pp points past the end of the current position-list, set *pi to
+** POSITION_LIST_END and return. Otherwise, read the next varint from *pp,
+** increment the current value of *pi by the value read, and set *pp to
+** point to the next value before returning.
+**
+** Before calling this routine *pi must be initialized to the value of
+** the previous position, or zero if we are reading the first position
+** in the position-list. Because positions are delta-encoded, the value
+** of the previous position is needed in order to compute the value of
+** the next position.
+*/
+static void fts3ReadNextPos(
+ char **pp, /* IN/OUT: Pointer into position-list buffer */
+ sqlite3_int64 *pi /* IN/OUT: Value read from position-list */
+){
+ if( (**pp)&0xFE ){
+ int iVal;
+ *pp += fts3GetVarint32((*pp), &iVal);
+ *pi += iVal;
+ *pi -= 2;
+ }else{
+ *pi = POSITION_LIST_END;
+ }
+}
+
+/*
+** If parameter iCol is not 0, write an POS_COLUMN (1) byte followed by
+** the value of iCol encoded as a varint to *pp. This will start a new
+** column list.
+**
+** Set *pp to point to the byte just after the last byte written before
+** returning (do not modify it if iCol==0). Return the total number of bytes
+** written (0 if iCol==0).
+*/
+static int fts3PutColNumber(char **pp, int iCol){
+ int n = 0; /* Number of bytes written */
+ if( iCol ){
+ char *p = *pp; /* Output pointer */
+ n = 1 + sqlite3Fts3PutVarint(&p[1], iCol);
+ *p = 0x01;
+ *pp = &p[n];
+ }
+ return n;
+}
+
+/*
+** Compute the union of two position lists. The output written
+** into *pp contains all positions of both *pp1 and *pp2 in sorted
+** order and with any duplicates removed. All pointers are
+** updated appropriately. The caller is responsible for insuring
+** that there is enough space in *pp to hold the complete output.
+*/
+static int fts3PoslistMerge(
+ char **pp, /* Output buffer */
+ char **pp1, /* Left input list */
+ char **pp2 /* Right input list */
+){
+ char *p = *pp;
+ char *p1 = *pp1;
+ char *p2 = *pp2;
+
+ while( *p1 || *p2 ){
+ int iCol1; /* The current column index in pp1 */
+ int iCol2; /* The current column index in pp2 */
+
+ if( *p1==POS_COLUMN ){
+ fts3GetVarint32(&p1[1], &iCol1);
+ if( iCol1==0 ) return FTS_CORRUPT_VTAB;
+ }
+ else if( *p1==POS_END ) iCol1 = 0x7fffffff;
+ else iCol1 = 0;
+
+ if( *p2==POS_COLUMN ){
+ fts3GetVarint32(&p2[1], &iCol2);
+ if( iCol2==0 ) return FTS_CORRUPT_VTAB;
+ }
+ else if( *p2==POS_END ) iCol2 = 0x7fffffff;
+ else iCol2 = 0;
+
+ if( iCol1==iCol2 ){
+ sqlite3_int64 i1 = 0; /* Last position from pp1 */
+ sqlite3_int64 i2 = 0; /* Last position from pp2 */
+ sqlite3_int64 iPrev = 0;
+ int n = fts3PutColNumber(&p, iCol1);
+ p1 += n;
+ p2 += n;
+
+ /* At this point, both p1 and p2 point to the start of column-lists
+ ** for the same column (the column with index iCol1 and iCol2).
+ ** A column-list is a list of non-negative delta-encoded varints, each
+ ** incremented by 2 before being stored. Each list is terminated by a
+ ** POS_END (0) or POS_COLUMN (1). The following block merges the two lists
+ ** and writes the results to buffer p. p is left pointing to the byte
+ ** after the list written. No terminator (POS_END or POS_COLUMN) is
+ ** written to the output.
+ */
+ fts3GetDeltaVarint(&p1, &i1);
+ fts3GetDeltaVarint(&p2, &i2);
+ if( i1<2 || i2<2 ){
+ break;
+ }
+ do {
+ fts3PutDeltaVarint(&p, &iPrev, (i1pos(*pp1) && pos(*pp2)-pos(*pp1)<=nToken). i.e.
+** when the *pp1 token appears before the *pp2 token, but not more than nToken
+** slots before it.
+**
+** e.g. nToken==1 searches for adjacent positions.
+*/
+static int fts3PoslistPhraseMerge(
+ char **pp, /* IN/OUT: Preallocated output buffer */
+ int nToken, /* Maximum difference in token positions */
+ int isSaveLeft, /* Save the left position */
+ int isExact, /* If *pp1 is exactly nTokens before *pp2 */
+ char **pp1, /* IN/OUT: Left input list */
+ char **pp2 /* IN/OUT: Right input list */
+){
+ char *p = *pp;
+ char *p1 = *pp1;
+ char *p2 = *pp2;
+ int iCol1 = 0;
+ int iCol2 = 0;
+
+ /* Never set both isSaveLeft and isExact for the same invocation. */
+ assert( isSaveLeft==0 || isExact==0 );
+
+ assert_fts3_nc( p!=0 && *p1!=0 && *p2!=0 );
+ if( *p1==POS_COLUMN ){
+ p1++;
+ p1 += fts3GetVarint32(p1, &iCol1);
+ /* iCol1==0 indicates corruption. Column 0 does not have a POS_COLUMN
+ ** entry, so this is actually end-of-doclist. */
+ if( iCol1==0 ) return 0;
+ }
+ if( *p2==POS_COLUMN ){
+ p2++;
+ p2 += fts3GetVarint32(p2, &iCol2);
+ /* As above, iCol2==0 indicates corruption. */
+ if( iCol2==0 ) return 0;
+ }
+
+ while( 1 ){
+ if( iCol1==iCol2 ){
+ char *pSave = p;
+ sqlite3_int64 iPrev = 0;
+ sqlite3_int64 iPos1 = 0;
+ sqlite3_int64 iPos2 = 0;
+
+ if( iCol1 ){
+ *p++ = POS_COLUMN;
+ p += sqlite3Fts3PutVarint(p, iCol1);
+ }
+
+ fts3GetDeltaVarint(&p1, &iPos1); iPos1 -= 2;
+ fts3GetDeltaVarint(&p2, &iPos2); iPos2 -= 2;
+ if( iPos1<0 || iPos2<0 ) break;
+
+ while( 1 ){
+ if( iPos2==iPos1+nToken
+ || (isExact==0 && iPos2>iPos1 && iPos2<=iPos1+nToken)
+ ){
+ sqlite3_int64 iSave;
+ iSave = isSaveLeft ? iPos1 : iPos2;
+ fts3PutDeltaVarint(&p, &iPrev, iSave+2); iPrev -= 2;
+ pSave = 0;
+ assert( p );
+ }
+ if( (!isSaveLeft && iPos2<=(iPos1+nToken)) || iPos2<=iPos1 ){
+ if( (*p2&0xFE)==0 ) break;
+ fts3GetDeltaVarint(&p2, &iPos2); iPos2 -= 2;
+ }else{
+ if( (*p1&0xFE)==0 ) break;
+ fts3GetDeltaVarint(&p1, &iPos1); iPos1 -= 2;
+ }
+ }
+
+ if( pSave ){
+ assert( pp && p );
+ p = pSave;
+ }
+
+ fts3ColumnlistCopy(0, &p1);
+ fts3ColumnlistCopy(0, &p2);
+ assert( (*p1&0xFE)==0 && (*p2&0xFE)==0 );
+ if( 0==*p1 || 0==*p2 ) break;
+
+ p1++;
+ p1 += fts3GetVarint32(p1, &iCol1);
+ p2++;
+ p2 += fts3GetVarint32(p2, &iCol2);
+ }
+
+ /* Advance pointer p1 or p2 (whichever corresponds to the smaller of
+ ** iCol1 and iCol2) so that it points to either the 0x00 that marks the
+ ** end of the position list, or the 0x01 that precedes the next
+ ** column-number in the position list.
+ */
+ else if( iCol1=pEnd ){
+ *pp = 0;
+ }else{
+ u64 iVal;
+ *pp += sqlite3Fts3GetVarintU(*pp, &iVal);
+ if( bDescIdx ){
+ *pVal = (i64)((u64)*pVal - iVal);
+ }else{
+ *pVal = (i64)((u64)*pVal + iVal);
+ }
+ }
+}
+
+/*
+** This function is used to write a single varint to a buffer. The varint
+** is written to *pp. Before returning, *pp is set to point 1 byte past the
+** end of the value written.
+**
+** If *pbFirst is zero when this function is called, the value written to
+** the buffer is that of parameter iVal.
+**
+** If *pbFirst is non-zero when this function is called, then the value
+** written is either (iVal-*piPrev) (if bDescIdx is zero) or (*piPrev-iVal)
+** (if bDescIdx is non-zero).
+**
+** Before returning, this function always sets *pbFirst to 1 and *piPrev
+** to the value of parameter iVal.
+*/
+static void fts3PutDeltaVarint3(
+ char **pp, /* IN/OUT: Output pointer */
+ int bDescIdx, /* True for descending docids */
+ sqlite3_int64 *piPrev, /* IN/OUT: Previous value written to list */
+ int *pbFirst, /* IN/OUT: True after first int written */
+ sqlite3_int64 iVal /* Write this value to the list */
+){
+ sqlite3_uint64 iWrite;
+ if( bDescIdx==0 || *pbFirst==0 ){
+ assert_fts3_nc( *pbFirst==0 || iVal>=*piPrev );
+ iWrite = (u64)iVal - (u64)*piPrev;
+ }else{
+ assert_fts3_nc( *piPrev>=iVal );
+ iWrite = (u64)*piPrev - (u64)iVal;
+ }
+ assert( *pbFirst || *piPrev==0 );
+ assert_fts3_nc( *pbFirst==0 || iWrite>0 );
+ *pp += sqlite3Fts3PutVarint(*pp, iWrite);
+ *piPrev = iVal;
+ *pbFirst = 1;
+}
+
+
+/*
+** This macro is used by various functions that merge doclists. The two
+** arguments are 64-bit docid values. If the value of the stack variable
+** bDescDoclist is 0 when this macro is invoked, then it returns (i1-i2).
+** Otherwise, (i2-i1).
+**
+** Using this makes it easier to write code that can merge doclists that are
+** sorted in either ascending or descending order.
+*/
+/* #define DOCID_CMP(i1, i2) ((bDescDoclist?-1:1) * (i64)((u64)i1-i2)) */
+#define DOCID_CMP(i1, i2) ((bDescDoclist?-1:1) * (i1>i2?1:((i1==i2)?0:-1)))
+
+/*
+** This function does an "OR" merge of two doclists (output contains all
+** positions contained in either argument doclist). If the docids in the
+** input doclists are sorted in ascending order, parameter bDescDoclist
+** should be false. If they are sorted in ascending order, it should be
+** passed a non-zero value.
+**
+** If no error occurs, *paOut is set to point at an sqlite3_malloc'd buffer
+** containing the output doclist and SQLITE_OK is returned. In this case
+** *pnOut is set to the number of bytes in the output doclist.
+**
+** If an error occurs, an SQLite error code is returned. The output values
+** are undefined in this case.
+*/
+static int fts3DoclistOrMerge(
+ int bDescDoclist, /* True if arguments are desc */
+ char *a1, int n1, /* First doclist */
+ char *a2, int n2, /* Second doclist */
+ char **paOut, int *pnOut /* OUT: Malloc'd doclist */
+){
+ int rc = SQLITE_OK;
+ sqlite3_int64 i1 = 0;
+ sqlite3_int64 i2 = 0;
+ sqlite3_int64 iPrev = 0;
+ char *pEnd1 = &a1[n1];
+ char *pEnd2 = &a2[n2];
+ char *p1 = a1;
+ char *p2 = a2;
+ char *p;
+ char *aOut;
+ int bFirstOut = 0;
+
+ *paOut = 0;
+ *pnOut = 0;
+
+ /* Allocate space for the output. Both the input and output doclists
+ ** are delta encoded. If they are in ascending order (bDescDoclist==0),
+ ** then the first docid in each list is simply encoded as a varint. For
+ ** each subsequent docid, the varint stored is the difference between the
+ ** current and previous docid (a positive number - since the list is in
+ ** ascending order).
+ **
+ ** The first docid written to the output is therefore encoded using the
+ ** same number of bytes as it is in whichever of the input lists it is
+ ** read from. And each subsequent docid read from the same input list
+ ** consumes either the same or less bytes as it did in the input (since
+ ** the difference between it and the previous value in the output must
+ ** be a positive value less than or equal to the delta value read from
+ ** the input list). The same argument applies to all but the first docid
+ ** read from the 'other' list. And to the contents of all position lists
+ ** that will be copied and merged from the input to the output.
+ **
+ ** However, if the first docid copied to the output is a negative number,
+ ** then the encoding of the first docid from the 'other' input list may
+ ** be larger in the output than it was in the input (since the delta value
+ ** may be a larger positive integer than the actual docid).
+ **
+ ** The space required to store the output is therefore the sum of the
+ ** sizes of the two inputs, plus enough space for exactly one of the input
+ ** docids to grow.
+ **
+ ** A symetric argument may be made if the doclists are in descending
+ ** order.
+ */
+ aOut = sqlite3_malloc64((i64)n1+n2+FTS3_VARINT_MAX-1+FTS3_BUFFER_PADDING);
+ if( !aOut ) return SQLITE_NOMEM;
+
+ p = aOut;
+ fts3GetDeltaVarint3(&p1, pEnd1, 0, &i1);
+ fts3GetDeltaVarint3(&p2, pEnd2, 0, &i2);
+ while( p1 || p2 ){
+ sqlite3_int64 iDiff = DOCID_CMP(i1, i2);
+
+ if( p2 && p1 && iDiff==0 ){
+ fts3PutDeltaVarint3(&p, bDescDoclist, &iPrev, &bFirstOut, i1);
+ rc = fts3PoslistMerge(&p, &p1, &p2);
+ if( rc ) break;
+ fts3GetDeltaVarint3(&p1, pEnd1, bDescDoclist, &i1);
+ fts3GetDeltaVarint3(&p2, pEnd2, bDescDoclist, &i2);
+ }else if( !p2 || (p1 && iDiff<0) ){
+ fts3PutDeltaVarint3(&p, bDescDoclist, &iPrev, &bFirstOut, i1);
+ fts3PoslistCopy(&p, &p1);
+ fts3GetDeltaVarint3(&p1, pEnd1, bDescDoclist, &i1);
+ }else{
+ fts3PutDeltaVarint3(&p, bDescDoclist, &iPrev, &bFirstOut, i2);
+ fts3PoslistCopy(&p, &p2);
+ fts3GetDeltaVarint3(&p2, pEnd2, bDescDoclist, &i2);
+ }
+
+ assert( (p-aOut)<=((p1?(p1-a1):n1)+(p2?(p2-a2):n2)+FTS3_VARINT_MAX-1) );
+ }
+
+ if( rc!=SQLITE_OK ){
+ sqlite3_free(aOut);
+ p = aOut = 0;
+ }else{
+ assert( (p-aOut)<=n1+n2+FTS3_VARINT_MAX-1 );
+ memset(&aOut[(p-aOut)], 0, FTS3_BUFFER_PADDING);
+ }
+ *paOut = aOut;
+ *pnOut = (int)(p-aOut);
+ return rc;
+}
+
+/*
+** This function does a "phrase" merge of two doclists. In a phrase merge,
+** the output contains a copy of each position from the right-hand input
+** doclist for which there is a position in the left-hand input doclist
+** exactly nDist tokens before it.
+**
+** If the docids in the input doclists are sorted in ascending order,
+** parameter bDescDoclist should be false. If they are sorted in ascending
+** order, it should be passed a non-zero value.
+**
+** The right-hand input doclist is overwritten by this function.
+*/
+static int fts3DoclistPhraseMerge(
+ int bDescDoclist, /* True if arguments are desc */
+ int nDist, /* Distance from left to right (1=adjacent) */
+ char *aLeft, int nLeft, /* Left doclist */
+ char **paRight, int *pnRight /* IN/OUT: Right/output doclist */
+){
+ sqlite3_int64 i1 = 0;
+ sqlite3_int64 i2 = 0;
+ sqlite3_int64 iPrev = 0;
+ char *aRight = *paRight;
+ char *pEnd1 = &aLeft[nLeft];
+ char *pEnd2 = &aRight[*pnRight];
+ char *p1 = aLeft;
+ char *p2 = aRight;
+ char *p;
+ int bFirstOut = 0;
+ char *aOut;
+
+ assert( nDist>0 );
+ if( bDescDoclist ){
+ aOut = sqlite3_malloc64((sqlite3_int64)*pnRight + FTS3_VARINT_MAX);
+ if( aOut==0 ) return SQLITE_NOMEM;
+ }else{
+ aOut = aRight;
+ }
+ p = aOut;
+
+ fts3GetDeltaVarint3(&p1, pEnd1, 0, &i1);
+ fts3GetDeltaVarint3(&p2, pEnd2, 0, &i2);
+
+ while( p1 && p2 ){
+ sqlite3_int64 iDiff = DOCID_CMP(i1, i2);
+ if( iDiff==0 ){
+ char *pSave = p;
+ sqlite3_int64 iPrevSave = iPrev;
+ int bFirstOutSave = bFirstOut;
+
+ fts3PutDeltaVarint3(&p, bDescDoclist, &iPrev, &bFirstOut, i1);
+ if( 0==fts3PoslistPhraseMerge(&p, nDist, 0, 1, &p1, &p2) ){
+ p = pSave;
+ iPrev = iPrevSave;
+ bFirstOut = bFirstOutSave;
+ }
+ fts3GetDeltaVarint3(&p1, pEnd1, bDescDoclist, &i1);
+ fts3GetDeltaVarint3(&p2, pEnd2, bDescDoclist, &i2);
+ }else if( iDiff<0 ){
+ fts3PoslistCopy(0, &p1);
+ fts3GetDeltaVarint3(&p1, pEnd1, bDescDoclist, &i1);
+ }else{
+ fts3PoslistCopy(0, &p2);
+ fts3GetDeltaVarint3(&p2, pEnd2, bDescDoclist, &i2);
+ }
+ }
+
+ *pnRight = (int)(p - aOut);
+ if( bDescDoclist ){
+ sqlite3_free(aRight);
+ *paRight = aOut;
+ }
+
+ return SQLITE_OK;
+}
+
+/*
+** Argument pList points to a position list nList bytes in size. This
+** function checks to see if the position list contains any entries for
+** a token in position 0 (of any column). If so, it writes argument iDelta
+** to the output buffer pOut, followed by a position list consisting only
+** of the entries from pList at position 0, and terminated by an 0x00 byte.
+** The value returned is the number of bytes written to pOut (if any).
+*/
+int sqlite3Fts3FirstFilter(
+ sqlite3_int64 iDelta, /* Varint that may be written to pOut */
+ char *pList, /* Position list (no 0x00 term) */
+ int nList, /* Size of pList in bytes */
+ char *pOut /* Write output here */
+){
+ int nOut = 0;
+ int bWritten = 0; /* True once iDelta has been written */
+ char *p = pList;
+ char *pEnd = &pList[nList];
+
+ if( *p!=0x01 ){
+ if( *p==0x02 ){
+ nOut += sqlite3Fts3PutVarint(&pOut[nOut], iDelta);
+ pOut[nOut++] = 0x02;
+ bWritten = 1;
+ }
+ fts3ColumnlistCopy(0, &p);
+ }
+
+ while( paaOutput); i++){
+ if( pTS->aaOutput[i] ){
+ if( !aOut ){
+ aOut = pTS->aaOutput[i];
+ nOut = pTS->anOutput[i];
+ pTS->aaOutput[i] = 0;
+ }else{
+ int nNew;
+ char *aNew;
+
+ int rc = fts3DoclistOrMerge(p->bDescIdx,
+ pTS->aaOutput[i], pTS->anOutput[i], aOut, nOut, &aNew, &nNew
+ );
+ if( rc!=SQLITE_OK ){
+ sqlite3_free(aOut);
+ return rc;
+ }
+
+ sqlite3_free(pTS->aaOutput[i]);
+ sqlite3_free(aOut);
+ pTS->aaOutput[i] = 0;
+ aOut = aNew;
+ nOut = nNew;
+ }
+ }
+ }
+
+ pTS->aaOutput[0] = aOut;
+ pTS->anOutput[0] = nOut;
+ return SQLITE_OK;
+}
+
+/*
+** Merge the doclist aDoclist/nDoclist into the TermSelect object passed
+** as the first argument. The merge is an "OR" merge (see function
+** fts3DoclistOrMerge() for details).
+**
+** This function is called with the doclist for each term that matches
+** a queried prefix. It merges all these doclists into one, the doclist
+** for the specified prefix. Since there can be a very large number of
+** doclists to merge, the merging is done pair-wise using the TermSelect
+** object.
+**
+** This function returns SQLITE_OK if the merge is successful, or an
+** SQLite error code (SQLITE_NOMEM) if an error occurs.
+*/
+static int fts3TermSelectMerge(
+ Fts3Table *p, /* FTS table handle */
+ TermSelect *pTS, /* TermSelect object to merge into */
+ char *aDoclist, /* Pointer to doclist */
+ int nDoclist /* Size of aDoclist in bytes */
+){
+ if( pTS->aaOutput[0]==0 ){
+ /* If this is the first term selected, copy the doclist to the output
+ ** buffer using memcpy().
+ **
+ ** Add FTS3_VARINT_MAX bytes of unused space to the end of the
+ ** allocation. This is so as to ensure that the buffer is big enough
+ ** to hold the current doclist AND'd with any other doclist. If the
+ ** doclists are stored in order=ASC order, this padding would not be
+ ** required (since the size of [doclistA AND doclistB] is always less
+ ** than or equal to the size of [doclistA] in that case). But this is
+ ** not true for order=DESC. For example, a doclist containing (1, -1)
+ ** may be smaller than (-1), as in the first example the -1 may be stored
+ ** as a single-byte delta, whereas in the second it must be stored as a
+ ** FTS3_VARINT_MAX byte varint.
+ **
+ ** Similar padding is added in the fts3DoclistOrMerge() function.
+ */
+ pTS->aaOutput[0] = sqlite3_malloc64((i64)nDoclist + FTS3_VARINT_MAX + 1);
+ pTS->anOutput[0] = nDoclist;
+ if( pTS->aaOutput[0] ){
+ memcpy(pTS->aaOutput[0], aDoclist, nDoclist);
+ memset(&pTS->aaOutput[0][nDoclist], 0, FTS3_VARINT_MAX);
+ }else{
+ return SQLITE_NOMEM;
+ }
+ }else{
+ char *aMerge = aDoclist;
+ int nMerge = nDoclist;
+ int iOut;
+
+ for(iOut=0; iOutaaOutput); iOut++){
+ if( pTS->aaOutput[iOut]==0 ){
+ assert( iOut>0 );
+ pTS->aaOutput[iOut] = aMerge;
+ pTS->anOutput[iOut] = nMerge;
+ break;
+ }else{
+ char *aNew;
+ int nNew;
+
+ int rc = fts3DoclistOrMerge(p->bDescIdx, aMerge, nMerge,
+ pTS->aaOutput[iOut], pTS->anOutput[iOut], &aNew, &nNew
+ );
+ if( rc!=SQLITE_OK ){
+ if( aMerge!=aDoclist ) sqlite3_free(aMerge);
+ return rc;
+ }
+
+ if( aMerge!=aDoclist ) sqlite3_free(aMerge);
+ sqlite3_free(pTS->aaOutput[iOut]);
+ pTS->aaOutput[iOut] = 0;
+
+ aMerge = aNew;
+ nMerge = nNew;
+ if( (iOut+1)==SizeofArray(pTS->aaOutput) ){
+ pTS->aaOutput[iOut] = aMerge;
+ pTS->anOutput[iOut] = nMerge;
+ }
+ }
+ }
+ }
+ return SQLITE_OK;
+}
+
+/*
+** Append SegReader object pNew to the end of the pCsr->apSegment[] array.
+*/
+static int fts3SegReaderCursorAppend(
+ Fts3MultiSegReader *pCsr,
+ Fts3SegReader *pNew
+){
+ if( (pCsr->nSegment%16)==0 ){
+ Fts3SegReader **apNew;
+ sqlite3_int64 nByte = (pCsr->nSegment + 16)*sizeof(Fts3SegReader*);
+ apNew = (Fts3SegReader **)sqlite3_realloc64(pCsr->apSegment, nByte);
+ if( !apNew ){
+ sqlite3Fts3SegReaderFree(pNew);
+ return SQLITE_NOMEM;
+ }
+ pCsr->apSegment = apNew;
+ }
+ pCsr->apSegment[pCsr->nSegment++] = pNew;
+ return SQLITE_OK;
+}
+
+/*
+** Add seg-reader objects to the Fts3MultiSegReader object passed as the
+** 8th argument.
+**
+** This function returns SQLITE_OK if successful, or an SQLite error code
+** otherwise.
+*/
+static int fts3SegReaderCursor(
+ Fts3Table *p, /* FTS3 table handle */
+ int iLangid, /* Language id */
+ int iIndex, /* Index to search (from 0 to p->nIndex-1) */
+ int iLevel, /* Level of segments to scan */
+ const char *zTerm, /* Term to query for */
+ int nTerm, /* Size of zTerm in bytes */
+ int isPrefix, /* True for a prefix search */
+ int isScan, /* True to scan from zTerm to EOF */
+ Fts3MultiSegReader *pCsr /* Cursor object to populate */
+){
+ int rc = SQLITE_OK; /* Error code */
+ sqlite3_stmt *pStmt = 0; /* Statement to iterate through segments */
+ int rc2; /* Result of sqlite3_reset() */
+
+ /* If iLevel is less than 0 and this is not a scan, include a seg-reader
+ ** for the pending-terms. If this is a scan, then this call must be being
+ ** made by an fts4aux module, not an FTS table. In this case calling
+ ** Fts3SegReaderPending might segfault, as the data structures used by
+ ** fts4aux are not completely populated. So it's easiest to filter these
+ ** calls out here. */
+ if( iLevel<0 && p->aIndex && p->iPrevLangid==iLangid ){
+ Fts3SegReader *pSeg = 0;
+ rc = sqlite3Fts3SegReaderPending(p, iIndex, zTerm, nTerm, isPrefix||isScan, &pSeg);
+ if( rc==SQLITE_OK && pSeg ){
+ rc = fts3SegReaderCursorAppend(pCsr, pSeg);
+ }
+ }
+
+ if( iLevel!=FTS3_SEGCURSOR_PENDING ){
+ if( rc==SQLITE_OK ){
+ rc = sqlite3Fts3AllSegdirs(p, iLangid, iIndex, iLevel, &pStmt);
+ }
+
+ while( rc==SQLITE_OK && SQLITE_ROW==(rc = sqlite3_step(pStmt)) ){
+ Fts3SegReader *pSeg = 0;
+
+ /* Read the values returned by the SELECT into local variables. */
+ sqlite3_int64 iStartBlock = sqlite3_column_int64(pStmt, 1);
+ sqlite3_int64 iLeavesEndBlock = sqlite3_column_int64(pStmt, 2);
+ sqlite3_int64 iEndBlock = sqlite3_column_int64(pStmt, 3);
+ int nRoot = sqlite3_column_bytes(pStmt, 4);
+ char const *zRoot = sqlite3_column_blob(pStmt, 4);
+
+ /* If zTerm is not NULL, and this segment is not stored entirely on its
+ ** root node, the range of leaves scanned can be reduced. Do this. */
+ if( iStartBlock && zTerm && zRoot ){
+ sqlite3_int64 *pi = (isPrefix ? &iLeavesEndBlock : 0);
+ rc = fts3SelectLeaf(p, zTerm, nTerm, zRoot, nRoot, &iStartBlock, pi);
+ if( rc!=SQLITE_OK ) goto finished;
+ if( isPrefix==0 && isScan==0 ) iLeavesEndBlock = iStartBlock;
+ }
+
+ rc = sqlite3Fts3SegReaderNew(pCsr->nSegment+1,
+ (isPrefix==0 && isScan==0),
+ iStartBlock, iLeavesEndBlock,
+ iEndBlock, zRoot, nRoot, &pSeg
+ );
+ if( rc!=SQLITE_OK ) goto finished;
+ rc = fts3SegReaderCursorAppend(pCsr, pSeg);
+ }
+ }
+
+ finished:
+ rc2 = sqlite3_reset(pStmt);
+ if( rc==SQLITE_DONE ) rc = rc2;
+
+ return rc;
+}
+
+/*
+** Set up a cursor object for iterating through a full-text index or a
+** single level therein.
+*/
+int sqlite3Fts3SegReaderCursor(
+ Fts3Table *p, /* FTS3 table handle */
+ int iLangid, /* Language-id to search */
+ int iIndex, /* Index to search (from 0 to p->nIndex-1) */
+ int iLevel, /* Level of segments to scan */
+ const char *zTerm, /* Term to query for */
+ int nTerm, /* Size of zTerm in bytes */
+ int isPrefix, /* True for a prefix search */
+ int isScan, /* True to scan from zTerm to EOF */
+ Fts3MultiSegReader *pCsr /* Cursor object to populate */
+){
+ assert( iIndex>=0 && iIndexnIndex );
+ assert( iLevel==FTS3_SEGCURSOR_ALL
+ || iLevel==FTS3_SEGCURSOR_PENDING
+ || iLevel>=0
+ );
+ assert( iLevelbase.pVtab;
+
+ if( isPrefix ){
+ for(i=1; bFound==0 && inIndex; i++){
+ if( p->aIndex[i].nPrefix==nTerm ){
+ bFound = 1;
+ rc = sqlite3Fts3SegReaderCursor(p, pCsr->iLangid,
+ i, FTS3_SEGCURSOR_ALL, zTerm, nTerm, 0, 0, pSegcsr
+ );
+ pSegcsr->bLookup = 1;
+ }
+ }
+
+ for(i=1; bFound==0 && inIndex; i++){
+ if( p->aIndex[i].nPrefix==nTerm+1 ){
+ bFound = 1;
+ rc = sqlite3Fts3SegReaderCursor(p, pCsr->iLangid,
+ i, FTS3_SEGCURSOR_ALL, zTerm, nTerm, 1, 0, pSegcsr
+ );
+ if( rc==SQLITE_OK ){
+ rc = fts3SegReaderCursorAddZero(
+ p, pCsr->iLangid, zTerm, nTerm, pSegcsr
+ );
+ }
+ }
+ }
+ }
+
+ if( bFound==0 ){
+ rc = sqlite3Fts3SegReaderCursor(p, pCsr->iLangid,
+ 0, FTS3_SEGCURSOR_ALL, zTerm, nTerm, isPrefix, 0, pSegcsr
+ );
+ pSegcsr->bLookup = !isPrefix;
+ }
+ }
+
+ *ppSegcsr = pSegcsr;
+ return rc;
+}
+
+/*
+** Free an Fts3MultiSegReader allocated by fts3TermSegReaderCursor().
+*/
+static void fts3SegReaderCursorFree(Fts3MultiSegReader *pSegcsr){
+ sqlite3Fts3SegReaderFinish(pSegcsr);
+ sqlite3_free(pSegcsr);
+}
+
+/*
+** This function retrieves the doclist for the specified term (or term
+** prefix) from the database.
+*/
+static int fts3TermSelect(
+ Fts3Table *p, /* Virtual table handle */
+ Fts3PhraseToken *pTok, /* Token to query for */
+ int iColumn, /* Column to query (or -ve for all columns) */
+ int *pnOut, /* OUT: Size of buffer at *ppOut */
+ char **ppOut /* OUT: Malloced result buffer */
+){
+ int rc; /* Return code */
+ Fts3MultiSegReader *pSegcsr; /* Seg-reader cursor for this term */
+ TermSelect tsc; /* Object for pair-wise doclist merging */
+ Fts3SegFilter filter; /* Segment term filter configuration */
+
+ pSegcsr = pTok->pSegcsr;
+ memset(&tsc, 0, sizeof(TermSelect));
+
+ filter.flags = FTS3_SEGMENT_IGNORE_EMPTY | FTS3_SEGMENT_REQUIRE_POS
+ | (pTok->isPrefix ? FTS3_SEGMENT_PREFIX : 0)
+ | (pTok->bFirst ? FTS3_SEGMENT_FIRST : 0)
+ | (iColumnnColumn ? FTS3_SEGMENT_COLUMN_FILTER : 0);
+ filter.iCol = iColumn;
+ filter.zTerm = pTok->z;
+ filter.nTerm = pTok->n;
+
+ rc = sqlite3Fts3SegReaderStart(p, pSegcsr, &filter);
+ while( SQLITE_OK==rc
+ && SQLITE_ROW==(rc = sqlite3Fts3SegReaderStep(p, pSegcsr))
+ ){
+ rc = fts3TermSelectMerge(p, &tsc, pSegcsr->aDoclist, pSegcsr->nDoclist);
+ }
+
+ if( rc==SQLITE_OK ){
+ rc = fts3TermSelectFinishMerge(p, &tsc);
+ }
+ if( rc==SQLITE_OK ){
+ *ppOut = tsc.aaOutput[0];
+ *pnOut = tsc.anOutput[0];
+ }else{
+ int i;
+ for(i=0; ipSegcsr = 0;
+ return rc;
+}
+
+/*
+** This function counts the total number of docids in the doclist stored
+** in buffer aList[], size nList bytes.
+**
+** If the isPoslist argument is true, then it is assumed that the doclist
+** contains a position-list following each docid. Otherwise, it is assumed
+** that the doclist is simply a list of docids stored as delta encoded
+** varints.
+*/
+static int fts3DoclistCountDocids(char *aList, int nList){
+ int nDoc = 0; /* Return value */
+ if( aList ){
+ char *aEnd = &aList[nList]; /* Pointer to one byte after EOF */
+ char *p = aList; /* Cursor */
+ while( peSearch==FTS3_DOCID_SEARCH || pCsr->eSearch==FTS3_FULLSCAN_SEARCH ){
+ Fts3Table *pTab = (Fts3Table*)pCursor->pVtab;
+ pTab->bLock++;
+ if( SQLITE_ROW!=sqlite3_step(pCsr->pStmt) ){
+ pCsr->isEof = 1;
+ rc = sqlite3_reset(pCsr->pStmt);
+ }else{
+ pCsr->iPrevId = sqlite3_column_int64(pCsr->pStmt, 0);
+ rc = SQLITE_OK;
+ }
+ pTab->bLock--;
+ }else{
+ rc = fts3EvalNext((Fts3Cursor *)pCursor);
+ }
+ assert( ((Fts3Table *)pCsr->base.pVtab)->pSegments==0 );
+ return rc;
+}
+
+/*
+** If the numeric type of argument pVal is "integer", then return it
+** converted to a 64-bit signed integer. Otherwise, return a copy of
+** the second parameter, iDefault.
+*/
+static sqlite3_int64 fts3DocidRange(sqlite3_value *pVal, i64 iDefault){
+ if( pVal ){
+ int eType = sqlite3_value_numeric_type(pVal);
+ if( eType==SQLITE_INTEGER ){
+ return sqlite3_value_int64(pVal);
+ }
+ }
+ return iDefault;
+}
+
+/*
+** This is the xFilter interface for the virtual table. See
+** the virtual table xFilter method documentation for additional
+** information.
+**
+** If idxNum==FTS3_FULLSCAN_SEARCH then do a full table scan against
+** the %_content table.
+**
+** If idxNum==FTS3_DOCID_SEARCH then do a docid lookup for a single entry
+** in the %_content table.
+**
+** If idxNum>=FTS3_FULLTEXT_SEARCH then use the full text index. The
+** column on the left-hand side of the MATCH operator is column
+** number idxNum-FTS3_FULLTEXT_SEARCH, 0 indexed. argv[0] is the right-hand
+** side of the MATCH operator.
+*/
+static int fts3FilterMethod(
+ sqlite3_vtab_cursor *pCursor, /* The cursor used for this query */
+ int idxNum, /* Strategy index */
+ const char *idxStr, /* Unused */
+ int nVal, /* Number of elements in apVal */
+ sqlite3_value **apVal /* Arguments for the indexing scheme */
+){
+ int rc = SQLITE_OK;
+ char *zSql; /* SQL statement used to access %_content */
+ int eSearch;
+ Fts3Table *p = (Fts3Table *)pCursor->pVtab;
+ Fts3Cursor *pCsr = (Fts3Cursor *)pCursor;
+
+ sqlite3_value *pCons = 0; /* The MATCH or rowid constraint, if any */
+ sqlite3_value *pLangid = 0; /* The "langid = ?" constraint, if any */
+ sqlite3_value *pDocidGe = 0; /* The "docid >= ?" constraint, if any */
+ sqlite3_value *pDocidLe = 0; /* The "docid <= ?" constraint, if any */
+ int iIdx;
+
+ UNUSED_PARAMETER(idxStr);
+ UNUSED_PARAMETER(nVal);
+
+ if( p->bLock ){
+ return SQLITE_ERROR;
+ }
+
+ eSearch = (idxNum & 0x0000FFFF);
+ assert( eSearch>=0 && eSearch<=(FTS3_FULLTEXT_SEARCH+p->nColumn) );
+ assert( p->pSegments==0 );
+
+ /* Collect arguments into local variables */
+ iIdx = 0;
+ if( eSearch!=FTS3_FULLSCAN_SEARCH ) pCons = apVal[iIdx++];
+ if( idxNum & FTS3_HAVE_LANGID ) pLangid = apVal[iIdx++];
+ if( idxNum & FTS3_HAVE_DOCID_GE ) pDocidGe = apVal[iIdx++];
+ if( idxNum & FTS3_HAVE_DOCID_LE ) pDocidLe = apVal[iIdx++];
+ assert( iIdx==nVal );
+
+ /* In case the cursor has been used before, clear it now. */
+ fts3ClearCursor(pCsr);
+
+ /* Set the lower and upper bounds on docids to return */
+ pCsr->iMinDocid = fts3DocidRange(pDocidGe, SMALLEST_INT64);
+ pCsr->iMaxDocid = fts3DocidRange(pDocidLe, LARGEST_INT64);
+
+ if( idxStr ){
+ pCsr->bDesc = (idxStr[0]=='D');
+ }else{
+ pCsr->bDesc = p->bDescIdx;
+ }
+ pCsr->eSearch = (i16)eSearch;
+
+ if( eSearch!=FTS3_DOCID_SEARCH && eSearch!=FTS3_FULLSCAN_SEARCH ){
+ int iCol = eSearch-FTS3_FULLTEXT_SEARCH;
+ const char *zQuery = (const char *)sqlite3_value_text(pCons);
+
+ if( zQuery==0 && sqlite3_value_type(pCons)!=SQLITE_NULL ){
+ return SQLITE_NOMEM;
+ }
+
+ pCsr->iLangid = 0;
+ if( pLangid ) pCsr->iLangid = sqlite3_value_int(pLangid);
+
+ assert( p->base.zErrMsg==0 );
+ rc = sqlite3Fts3ExprParse(p->pTokenizer, pCsr->iLangid,
+ p->azColumn, p->bFts4, p->nColumn, iCol, zQuery, -1, &pCsr->pExpr,
+ &p->base.zErrMsg
+ );
+ if( rc!=SQLITE_OK ){
+ return rc;
+ }
+
+ rc = fts3EvalStart(pCsr);
+ sqlite3Fts3SegmentsClose(p);
+ if( rc!=SQLITE_OK ) return rc;
+ pCsr->pNextId = pCsr->aDoclist;
+ pCsr->iPrevId = 0;
+ }
+
+ /* Compile a SELECT statement for this cursor. For a full-table-scan, the
+ ** statement loops through all rows of the %_content table. For a
+ ** full-text query or docid lookup, the statement retrieves a single
+ ** row by docid.
+ */
+ if( eSearch==FTS3_FULLSCAN_SEARCH ){
+ if( pDocidGe || pDocidLe ){
+ zSql = sqlite3_mprintf(
+ "SELECT %s WHERE rowid BETWEEN %lld AND %lld ORDER BY rowid %s",
+ p->zReadExprlist, pCsr->iMinDocid, pCsr->iMaxDocid,
+ (pCsr->bDesc ? "DESC" : "ASC")
+ );
+ }else{
+ zSql = sqlite3_mprintf("SELECT %s ORDER BY rowid %s",
+ p->zReadExprlist, (pCsr->bDesc ? "DESC" : "ASC")
+ );
+ }
+ if( zSql ){
+ p->bLock++;
+ rc = sqlite3_prepare_v3(
+ p->db,zSql,-1,SQLITE_PREPARE_PERSISTENT,&pCsr->pStmt,0
+ );
+ p->bLock--;
+ sqlite3_free(zSql);
+ }else{
+ rc = SQLITE_NOMEM;
+ }
+ }else if( eSearch==FTS3_DOCID_SEARCH ){
+ rc = fts3CursorSeekStmt(pCsr);
+ if( rc==SQLITE_OK ){
+ rc = sqlite3_bind_value(pCsr->pStmt, 1, pCons);
+ }
+ }
+ if( rc!=SQLITE_OK ) return rc;
+
+ return fts3NextMethod(pCursor);
+}
+
+/*
+** This is the xEof method of the virtual table. SQLite calls this
+** routine to find out if it has reached the end of a result set.
+*/
+static int fts3EofMethod(sqlite3_vtab_cursor *pCursor){
+ Fts3Cursor *pCsr = (Fts3Cursor*)pCursor;
+ if( pCsr->isEof ){
+ fts3ClearCursor(pCsr);
+ pCsr->isEof = 1;
+ }
+ return pCsr->isEof;
+}
+
+/*
+** This is the xRowid method. The SQLite core calls this routine to
+** retrieve the rowid for the current row of the result set. fts3
+** exposes %_content.docid as the rowid for the virtual table. The
+** rowid should be written to *pRowid.
+*/
+static int fts3RowidMethod(sqlite3_vtab_cursor *pCursor, sqlite_int64 *pRowid){
+ Fts3Cursor *pCsr = (Fts3Cursor *) pCursor;
+ *pRowid = pCsr->iPrevId;
+ return SQLITE_OK;
+}
+
+/*
+** This is the xColumn method, called by SQLite to request a value from
+** the row that the supplied cursor currently points to.
+**
+** If:
+**
+** (iCol < p->nColumn) -> The value of the iCol'th user column.
+** (iCol == p->nColumn) -> Magic column with the same name as the table.
+** (iCol == p->nColumn+1) -> Docid column
+** (iCol == p->nColumn+2) -> Langid column
+*/
+static int fts3ColumnMethod(
+ sqlite3_vtab_cursor *pCursor, /* Cursor to retrieve value from */
+ sqlite3_context *pCtx, /* Context for sqlite3_result_xxx() calls */
+ int iCol /* Index of column to read value from */
+){
+ int rc = SQLITE_OK; /* Return Code */
+ Fts3Cursor *pCsr = (Fts3Cursor *) pCursor;
+ Fts3Table *p = (Fts3Table *)pCursor->pVtab;
+
+ /* The column value supplied by SQLite must be in range. */
+ assert( iCol>=0 && iCol<=p->nColumn+2 );
+
+ switch( iCol-p->nColumn ){
+ case 0:
+ /* The special 'table-name' column */
+ sqlite3_result_pointer(pCtx, pCsr, "fts3cursor", 0);
+ break;
+
+ case 1:
+ /* The docid column */
+ sqlite3_result_int64(pCtx, pCsr->iPrevId);
+ break;
+
+ case 2:
+ if( pCsr->pExpr ){
+ sqlite3_result_int64(pCtx, pCsr->iLangid);
+ break;
+ }else if( p->zLanguageid==0 ){
+ sqlite3_result_int(pCtx, 0);
+ break;
+ }else{
+ iCol = p->nColumn;
+ /* no break */ deliberate_fall_through
+ }
+
+ default:
+ /* A user column. Or, if this is a full-table scan, possibly the
+ ** language-id column. Seek the cursor. */
+ rc = fts3CursorSeek(0, pCsr);
+ if( rc==SQLITE_OK && sqlite3_data_count(pCsr->pStmt)-1>iCol ){
+ sqlite3_result_value(pCtx, sqlite3_column_value(pCsr->pStmt, iCol+1));
+ }
+ break;
+ }
+
+ assert( ((Fts3Table *)pCsr->base.pVtab)->pSegments==0 );
+ return rc;
+}
+
+/*
+** This function is the implementation of the xUpdate callback used by
+** FTS3 virtual tables. It is invoked by SQLite each time a row is to be
+** inserted, updated or deleted.
+*/
+static int fts3UpdateMethod(
+ sqlite3_vtab *pVtab, /* Virtual table handle */
+ int nArg, /* Size of argument array */
+ sqlite3_value **apVal, /* Array of arguments */
+ sqlite_int64 *pRowid /* OUT: The affected (or effected) rowid */
+){
+ return sqlite3Fts3UpdateMethod(pVtab, nArg, apVal, pRowid);
+}
+
+/*
+** Implementation of xSync() method. Flush the contents of the pending-terms
+** hash-table to the database.
+*/
+static int fts3SyncMethod(sqlite3_vtab *pVtab){
+
+ /* Following an incremental-merge operation, assuming that the input
+ ** segments are not completely consumed (the usual case), they are updated
+ ** in place to remove the entries that have already been merged. This
+ ** involves updating the leaf block that contains the smallest unmerged
+ ** entry and each block (if any) between the leaf and the root node. So
+ ** if the height of the input segment b-trees is N, and input segments
+ ** are merged eight at a time, updating the input segments at the end
+ ** of an incremental-merge requires writing (8*(1+N)) blocks. N is usually
+ ** small - often between 0 and 2. So the overhead of the incremental
+ ** merge is somewhere between 8 and 24 blocks. To avoid this overhead
+ ** dwarfing the actual productive work accomplished, the incremental merge
+ ** is only attempted if it will write at least 64 leaf blocks. Hence
+ ** nMinMerge.
+ **
+ ** Of course, updating the input segments also involves deleting a bunch
+ ** of blocks from the segments table. But this is not considered overhead
+ ** as it would also be required by a crisis-merge that used the same input
+ ** segments.
+ */
+ const u32 nMinMerge = 64; /* Minimum amount of incr-merge work to do */
+
+ Fts3Table *p = (Fts3Table*)pVtab;
+ int rc;
+ i64 iLastRowid = sqlite3_last_insert_rowid(p->db);
+
+ rc = sqlite3Fts3PendingTermsFlush(p);
+ if( rc==SQLITE_OK
+ && p->nLeafAdd>(nMinMerge/16)
+ && p->nAutoincrmerge && p->nAutoincrmerge!=0xff
+ ){
+ int mxLevel = 0; /* Maximum relative level value in db */
+ int A; /* Incr-merge parameter A */
+
+ rc = sqlite3Fts3MaxLevel(p, &mxLevel);
+ assert( rc==SQLITE_OK || mxLevel==0 );
+ A = p->nLeafAdd * mxLevel;
+ A += (A/2);
+ if( A>(int)nMinMerge ) rc = sqlite3Fts3Incrmerge(p, A, p->nAutoincrmerge);
+ }
+ sqlite3Fts3SegmentsClose(p);
+ sqlite3_set_last_insert_rowid(p->db, iLastRowid);
+ return rc;
+}
+
+/*
+** If it is currently unknown whether or not the FTS table has an %_stat
+** table (if p->bHasStat==2), attempt to determine this (set p->bHasStat
+** to 0 or 1). Return SQLITE_OK if successful, or an SQLite error code
+** if an error occurs.
+*/
+static int fts3SetHasStat(Fts3Table *p){
+ int rc = SQLITE_OK;
+ if( p->bHasStat==2 ){
+ char *zTbl = sqlite3_mprintf("%s_stat", p->zName);
+ if( zTbl ){
+ int res = sqlite3_table_column_metadata(p->db, p->zDb, zTbl, 0,0,0,0,0,0);
+ sqlite3_free(zTbl);
+ p->bHasStat = (res==SQLITE_OK);
+ }else{
+ rc = SQLITE_NOMEM;
+ }
+ }
+ return rc;
+}
+
+/*
+** Implementation of xBegin() method.
+*/
+static int fts3BeginMethod(sqlite3_vtab *pVtab){
+ Fts3Table *p = (Fts3Table*)pVtab;
+ int rc;
+ UNUSED_PARAMETER(pVtab);
+ assert( p->pSegments==0 );
+ assert( p->nPendingData==0 );
+ assert( p->inTransaction!=1 );
+ p->nLeafAdd = 0;
+ rc = fts3SetHasStat(p);
+#ifdef SQLITE_DEBUG
+ if( rc==SQLITE_OK ){
+ p->inTransaction = 1;
+ p->mxSavepoint = -1;
+ }
+#endif
+ return rc;
+}
+
+/*
+** Implementation of xCommit() method. This is a no-op. The contents of
+** the pending-terms hash-table have already been flushed into the database
+** by fts3SyncMethod().
+*/
+static int fts3CommitMethod(sqlite3_vtab *pVtab){
+ TESTONLY( Fts3Table *p = (Fts3Table*)pVtab );
+ UNUSED_PARAMETER(pVtab);
+ assert( p->nPendingData==0 );
+ assert( p->inTransaction!=0 );
+ assert( p->pSegments==0 );
+ TESTONLY( p->inTransaction = 0 );
+ TESTONLY( p->mxSavepoint = -1; );
+ return SQLITE_OK;
+}
+
+/*
+** Implementation of xRollback(). Discard the contents of the pending-terms
+** hash-table. Any changes made to the database are reverted by SQLite.
+*/
+static int fts3RollbackMethod(sqlite3_vtab *pVtab){
+ Fts3Table *p = (Fts3Table*)pVtab;
+ sqlite3Fts3PendingTermsClear(p);
+ assert( p->inTransaction!=0 );
+ TESTONLY( p->inTransaction = 0 );
+ TESTONLY( p->mxSavepoint = -1; );
+ return SQLITE_OK;
+}
+
+/*
+** When called, *ppPoslist must point to the byte immediately following the
+** end of a position-list. i.e. ( (*ppPoslist)[-1]==POS_END ). This function
+** moves *ppPoslist so that it instead points to the first byte of the
+** same position list.
+*/
+static void fts3ReversePoslist(char *pStart, char **ppPoslist){
+ char *p = &(*ppPoslist)[-2];
+ char c = 0;
+
+ /* Skip backwards passed any trailing 0x00 bytes added by NearTrim() */
+ while( p>pStart && (c=*p--)==0 );
+
+ /* Search backwards for a varint with value zero (the end of the previous
+ ** poslist). This is an 0x00 byte preceded by some byte that does not
+ ** have the 0x80 bit set. */
+ while( p>pStart && (*p & 0x80) | c ){
+ c = *p--;
+ }
+ assert( p==pStart || c==0 );
+
+ /* At this point p points to that preceding byte without the 0x80 bit
+ ** set. So to find the start of the poslist, skip forward 2 bytes then
+ ** over a varint.
+ **
+ ** Normally. The other case is that p==pStart and the poslist to return
+ ** is the first in the doclist. In this case do not skip forward 2 bytes.
+ ** The second part of the if condition (c==0 && *ppPoslist>&p[2])
+ ** is required for cases where the first byte of a doclist and the
+ ** doclist is empty. For example, if the first docid is 10, a doclist
+ ** that begins with:
+ **
+ ** 0x0A 0x00
+ */
+ if( p>pStart || (c==0 && *ppPoslist>&p[2]) ){ p = &p[2]; }
+ while( *p++&0x80 );
+ *ppPoslist = p;
+}
+
+/*
+** Helper function used by the implementation of the overloaded snippet(),
+** offsets() and optimize() SQL functions.
+**
+** If the value passed as the third argument is a blob of size
+** sizeof(Fts3Cursor*), then the blob contents are copied to the
+** output variable *ppCsr and SQLITE_OK is returned. Otherwise, an error
+** message is written to context pContext and SQLITE_ERROR returned. The
+** string passed via zFunc is used as part of the error message.
+*/
+static int fts3FunctionArg(
+ sqlite3_context *pContext, /* SQL function call context */
+ const char *zFunc, /* Function name */
+ sqlite3_value *pVal, /* argv[0] passed to function */
+ Fts3Cursor **ppCsr /* OUT: Store cursor handle here */
+){
+ int rc;
+ *ppCsr = (Fts3Cursor*)sqlite3_value_pointer(pVal, "fts3cursor");
+ if( (*ppCsr)!=0 ){
+ rc = SQLITE_OK;
+ }else{
+ char *zErr = sqlite3_mprintf("illegal first argument to %s", zFunc);
+ sqlite3_result_error(pContext, zErr, -1);
+ sqlite3_free(zErr);
+ rc = SQLITE_ERROR;
+ }
+ return rc;
+}
+
+/*
+** Implementation of the snippet() function for FTS3
+*/
+static void fts3SnippetFunc(
+ sqlite3_context *pContext, /* SQLite function call context */
+ int nVal, /* Size of apVal[] array */
+ sqlite3_value **apVal /* Array of arguments */
+){
+ Fts3Cursor *pCsr; /* Cursor handle passed through apVal[0] */
+ const char *zStart = "";
+ const char *zEnd = "";
+ const char *zEllipsis = "...";
+ int iCol = -1;
+ int nToken = 15; /* Default number of tokens in snippet */
+
+ /* There must be at least one argument passed to this function (otherwise
+ ** the non-overloaded version would have been called instead of this one).
+ */
+ assert( nVal>=1 );
+
+ if( nVal>6 ){
+ sqlite3_result_error(pContext,
+ "wrong number of arguments to function snippet()", -1);
+ return;
+ }
+ if( fts3FunctionArg(pContext, "snippet", apVal[0], &pCsr) ) return;
+
+ switch( nVal ){
+ case 6: nToken = sqlite3_value_int(apVal[5]);
+ /* no break */ deliberate_fall_through
+ case 5: iCol = sqlite3_value_int(apVal[4]);
+ /* no break */ deliberate_fall_through
+ case 4: zEllipsis = (const char*)sqlite3_value_text(apVal[3]);
+ /* no break */ deliberate_fall_through
+ case 3: zEnd = (const char*)sqlite3_value_text(apVal[2]);
+ /* no break */ deliberate_fall_through
+ case 2: zStart = (const char*)sqlite3_value_text(apVal[1]);
+ }
+ if( !zEllipsis || !zEnd || !zStart ){
+ sqlite3_result_error_nomem(pContext);
+ }else if( nToken==0 ){
+ sqlite3_result_text(pContext, "", -1, SQLITE_STATIC);
+ }else if( SQLITE_OK==fts3CursorSeek(pContext, pCsr) ){
+ sqlite3Fts3Snippet(pContext, pCsr, zStart, zEnd, zEllipsis, iCol, nToken);
+ }
+}
+
+/*
+** Implementation of the offsets() function for FTS3
+*/
+static void fts3OffsetsFunc(
+ sqlite3_context *pContext, /* SQLite function call context */
+ int nVal, /* Size of argument array */
+ sqlite3_value **apVal /* Array of arguments */
+){
+ Fts3Cursor *pCsr; /* Cursor handle passed through apVal[0] */
+
+ UNUSED_PARAMETER(nVal);
+
+ assert( nVal==1 );
+ if( fts3FunctionArg(pContext, "offsets", apVal[0], &pCsr) ) return;
+ assert( pCsr );
+ if( SQLITE_OK==fts3CursorSeek(pContext, pCsr) ){
+ sqlite3Fts3Offsets(pContext, pCsr);
+ }
+}
+
+/*
+** Implementation of the special optimize() function for FTS3. This
+** function merges all segments in the database to a single segment.
+** Example usage is:
+**
+** SELECT optimize(t) FROM t LIMIT 1;
+**
+** where 't' is the name of an FTS3 table.
+*/
+static void fts3OptimizeFunc(
+ sqlite3_context *pContext, /* SQLite function call context */
+ int nVal, /* Size of argument array */
+ sqlite3_value **apVal /* Array of arguments */
+){
+ int rc; /* Return code */
+ Fts3Table *p; /* Virtual table handle */
+ Fts3Cursor *pCursor; /* Cursor handle passed through apVal[0] */
+
+ UNUSED_PARAMETER(nVal);
+
+ assert( nVal==1 );
+ if( fts3FunctionArg(pContext, "optimize", apVal[0], &pCursor) ) return;
+ p = (Fts3Table *)pCursor->base.pVtab;
+ assert( p );
+
+ rc = sqlite3Fts3Optimize(p);
+
+ switch( rc ){
+ case SQLITE_OK:
+ sqlite3_result_text(pContext, "Index optimized", -1, SQLITE_STATIC);
+ break;
+ case SQLITE_DONE:
+ sqlite3_result_text(pContext, "Index already optimal", -1, SQLITE_STATIC);
+ break;
+ default:
+ sqlite3_result_error_code(pContext, rc);
+ break;
+ }
+}
+
+/*
+** Implementation of the matchinfo() function for FTS3
+*/
+static void fts3MatchinfoFunc(
+ sqlite3_context *pContext, /* SQLite function call context */
+ int nVal, /* Size of argument array */
+ sqlite3_value **apVal /* Array of arguments */
+){
+ Fts3Cursor *pCsr; /* Cursor handle passed through apVal[0] */
+ assert( nVal==1 || nVal==2 );
+ if( SQLITE_OK==fts3FunctionArg(pContext, "matchinfo", apVal[0], &pCsr) ){
+ const char *zArg = 0;
+ if( nVal>1 ){
+ zArg = (const char *)sqlite3_value_text(apVal[1]);
+ }
+ sqlite3Fts3Matchinfo(pContext, pCsr, zArg);
+ }
+}
+
+/*
+** This routine implements the xFindFunction method for the FTS3
+** virtual table.
+*/
+static int fts3FindFunctionMethod(
+ sqlite3_vtab *pVtab, /* Virtual table handle */
+ int nArg, /* Number of SQL function arguments */
+ const char *zName, /* Name of SQL function */
+ void (**pxFunc)(sqlite3_context*,int,sqlite3_value**), /* OUT: Result */
+ void **ppArg /* Unused */
+){
+ struct Overloaded {
+ const char *zName;
+ void (*xFunc)(sqlite3_context*,int,sqlite3_value**);
+ } aOverload[] = {
+ { "snippet", fts3SnippetFunc },
+ { "offsets", fts3OffsetsFunc },
+ { "optimize", fts3OptimizeFunc },
+ { "matchinfo", fts3MatchinfoFunc },
+ };
+ int i; /* Iterator variable */
+
+ UNUSED_PARAMETER(pVtab);
+ UNUSED_PARAMETER(nArg);
+ UNUSED_PARAMETER(ppArg);
+
+ for(i=0; idb; /* Database connection */
+ int rc; /* Return Code */
+
+ /* At this point it must be known if the %_stat table exists or not.
+ ** So bHasStat may not be 2. */
+ rc = fts3SetHasStat(p);
+
+ /* As it happens, the pending terms table is always empty here. This is
+ ** because an "ALTER TABLE RENAME TABLE" statement inside a transaction
+ ** always opens a savepoint transaction. And the xSavepoint() method
+ ** flushes the pending terms table. But leave the (no-op) call to
+ ** PendingTermsFlush() in in case that changes.
+ */
+ assert( p->nPendingData==0 );
+ if( rc==SQLITE_OK ){
+ rc = sqlite3Fts3PendingTermsFlush(p);
+ }
+
+ p->bIgnoreSavepoint = 1;
+
+ if( p->zContentTbl==0 ){
+ fts3DbExec(&rc, db,
+ "ALTER TABLE %Q.'%q_content' RENAME TO '%q_content';",
+ p->zDb, p->zName, zName
+ );
+ }
+
+ if( p->bHasDocsize ){
+ fts3DbExec(&rc, db,
+ "ALTER TABLE %Q.'%q_docsize' RENAME TO '%q_docsize';",
+ p->zDb, p->zName, zName
+ );
+ }
+ if( p->bHasStat ){
+ fts3DbExec(&rc, db,
+ "ALTER TABLE %Q.'%q_stat' RENAME TO '%q_stat';",
+ p->zDb, p->zName, zName
+ );
+ }
+ fts3DbExec(&rc, db,
+ "ALTER TABLE %Q.'%q_segments' RENAME TO '%q_segments';",
+ p->zDb, p->zName, zName
+ );
+ fts3DbExec(&rc, db,
+ "ALTER TABLE %Q.'%q_segdir' RENAME TO '%q_segdir';",
+ p->zDb, p->zName, zName
+ );
+
+ p->bIgnoreSavepoint = 0;
+ return rc;
+}
+
+/*
+** The xSavepoint() method.
+**
+** Flush the contents of the pending-terms table to disk.
+*/
+static int fts3SavepointMethod(sqlite3_vtab *pVtab, int iSavepoint){
+ int rc = SQLITE_OK;
+ Fts3Table *pTab = (Fts3Table*)pVtab;
+ assert( pTab->inTransaction );
+ assert( pTab->mxSavepoint<=iSavepoint );
+ TESTONLY( pTab->mxSavepoint = iSavepoint );
+
+ if( pTab->bIgnoreSavepoint==0 ){
+ if( fts3HashCount(&pTab->aIndex[0].hPending)>0 ){
+ char *zSql = sqlite3_mprintf("INSERT INTO %Q.%Q(%Q) VALUES('flush')",
+ pTab->zDb, pTab->zName, pTab->zName
+ );
+ if( zSql ){
+ pTab->bIgnoreSavepoint = 1;
+ rc = sqlite3_exec(pTab->db, zSql, 0, 0, 0);
+ pTab->bIgnoreSavepoint = 0;
+ sqlite3_free(zSql);
+ }else{
+ rc = SQLITE_NOMEM;
+ }
+ }
+ if( rc==SQLITE_OK ){
+ pTab->iSavepoint = iSavepoint+1;
+ }
+ }
+ return rc;
+}
+
+/*
+** The xRelease() method.
+**
+** This is a no-op.
+*/
+static int fts3ReleaseMethod(sqlite3_vtab *pVtab, int iSavepoint){
+ Fts3Table *pTab = (Fts3Table*)pVtab;
+ assert( pTab->inTransaction );
+ assert( pTab->mxSavepoint >= iSavepoint );
+ TESTONLY( pTab->mxSavepoint = iSavepoint-1 );
+ pTab->iSavepoint = iSavepoint;
+ return SQLITE_OK;
+}
+
+/*
+** The xRollbackTo() method.
+**
+** Discard the contents of the pending terms table.
+*/
+static int fts3RollbackToMethod(sqlite3_vtab *pVtab, int iSavepoint){
+ Fts3Table *pTab = (Fts3Table*)pVtab;
+ UNUSED_PARAMETER(iSavepoint);
+ assert( pTab->inTransaction );
+ TESTONLY( pTab->mxSavepoint = iSavepoint );
+ if( (iSavepoint+1)<=pTab->iSavepoint ){
+ sqlite3Fts3PendingTermsClear(pTab);
+ }
+ return SQLITE_OK;
+}
+
+/*
+** Return true if zName is the extension on one of the shadow tables used
+** by this module.
+*/
+static int fts3ShadowName(const char *zName){
+ static const char *azName[] = {
+ "content", "docsize", "segdir", "segments", "stat",
+ };
+ unsigned int i;
+ for(i=0; ibFts4 ? 4 : 3, zSchema, zTabname, sqlite3_errstr(rc));
+ if( *pzErr ) rc = SQLITE_OK;
+ }else if( rc==SQLITE_OK && bOk==0 ){
+ *pzErr = sqlite3_mprintf("malformed inverted index for FTS%d table %s.%s",
+ p->bFts4 ? 4 : 3, zSchema, zTabname);
+ if( *pzErr==0 ) rc = SQLITE_NOMEM;
+ }
+ sqlite3Fts3SegmentsClose(p);
+ return rc;
+}
+
+
+
+static const sqlite3_module fts3Module = {
+ /* iVersion */ 4,
+ /* xCreate */ fts3CreateMethod,
+ /* xConnect */ fts3ConnectMethod,
+ /* xBestIndex */ fts3BestIndexMethod,
+ /* xDisconnect */ fts3DisconnectMethod,
+ /* xDestroy */ fts3DestroyMethod,
+ /* xOpen */ fts3OpenMethod,
+ /* xClose */ fts3CloseMethod,
+ /* xFilter */ fts3FilterMethod,
+ /* xNext */ fts3NextMethod,
+ /* xEof */ fts3EofMethod,
+ /* xColumn */ fts3ColumnMethod,
+ /* xRowid */ fts3RowidMethod,
+ /* xUpdate */ fts3UpdateMethod,
+ /* xBegin */ fts3BeginMethod,
+ /* xSync */ fts3SyncMethod,
+ /* xCommit */ fts3CommitMethod,
+ /* xRollback */ fts3RollbackMethod,
+ /* xFindFunction */ fts3FindFunctionMethod,
+ /* xRename */ fts3RenameMethod,
+ /* xSavepoint */ fts3SavepointMethod,
+ /* xRelease */ fts3ReleaseMethod,
+ /* xRollbackTo */ fts3RollbackToMethod,
+ /* xShadowName */ fts3ShadowName,
+ /* xIntegrity */ fts3IntegrityMethod,
+};
+
+/*
+** This function is registered as the module destructor (called when an
+** FTS3 enabled database connection is closed). It frees the memory
+** allocated for the tokenizer hash table.
+*/
+static void hashDestroy(void *p){
+ Fts3HashWrapper *pHash = (Fts3HashWrapper *)p;
+ pHash->nRef--;
+ if( pHash->nRef<=0 ){
+ sqlite3Fts3HashClear(&pHash->hash);
+ sqlite3_free(pHash);
+ }
+}
+
+/*
+** The fts3 built-in tokenizers - "simple", "porter" and "icu"- are
+** implemented in files fts3_tokenizer1.c, fts3_porter.c and fts3_icu.c
+** respectively. The following three forward declarations are for functions
+** declared in these files used to retrieve the respective implementations.
+**
+** Calling sqlite3Fts3SimpleTokenizerModule() sets the value pointed
+** to by the argument to point to the "simple" tokenizer implementation.
+** And so on.
+*/
+void sqlite3Fts3SimpleTokenizerModule(sqlite3_tokenizer_module const**ppModule);
+void sqlite3Fts3PorterTokenizerModule(sqlite3_tokenizer_module const**ppModule);
+#ifndef SQLITE_DISABLE_FTS3_UNICODE
+void sqlite3Fts3UnicodeTokenizer(sqlite3_tokenizer_module const**ppModule);
+#endif
+#ifdef SQLITE_ENABLE_ICU
+void sqlite3Fts3IcuTokenizerModule(sqlite3_tokenizer_module const**ppModule);
+#endif
+
+/*
+** Initialize the fts3 extension. If this extension is built as part
+** of the sqlite library, then this function is called directly by
+** SQLite. If fts3 is built as a dynamically loadable extension, this
+** function is called by the sqlite3_extension_init() entry point.
+*/
+int sqlite3Fts3Init(sqlite3 *db){
+ int rc = SQLITE_OK;
+ Fts3HashWrapper *pHash = 0;
+ const sqlite3_tokenizer_module *pSimple = 0;
+ const sqlite3_tokenizer_module *pPorter = 0;
+#ifndef SQLITE_DISABLE_FTS3_UNICODE
+ const sqlite3_tokenizer_module *pUnicode = 0;
+#endif
+
+#ifdef SQLITE_ENABLE_ICU
+ const sqlite3_tokenizer_module *pIcu = 0;
+ sqlite3Fts3IcuTokenizerModule(&pIcu);
+#endif
+
+#ifndef SQLITE_DISABLE_FTS3_UNICODE
+ sqlite3Fts3UnicodeTokenizer(&pUnicode);
+#endif
+
+#ifdef SQLITE_TEST
+ rc = sqlite3Fts3InitTerm(db);
+ if( rc!=SQLITE_OK ) return rc;
+#endif
+
+ rc = sqlite3Fts3InitAux(db);
+ if( rc!=SQLITE_OK ) return rc;
+
+ sqlite3Fts3SimpleTokenizerModule(&pSimple);
+ sqlite3Fts3PorterTokenizerModule(&pPorter);
+
+ /* Allocate and initialize the hash-table used to store tokenizers. */
+ pHash = sqlite3_malloc(sizeof(Fts3HashWrapper));
+ if( !pHash ){
+ rc = SQLITE_NOMEM;
+ }else{
+ sqlite3Fts3HashInit(&pHash->hash, FTS3_HASH_STRING, 1);
+ pHash->nRef = 0;
+ }
+
+ /* Load the built-in tokenizers into the hash table */
+ if( rc==SQLITE_OK ){
+ if( sqlite3Fts3HashInsert(&pHash->hash, "simple", 7, (void *)pSimple)
+ || sqlite3Fts3HashInsert(&pHash->hash, "porter", 7, (void *)pPorter)
+
+#ifndef SQLITE_DISABLE_FTS3_UNICODE
+ || sqlite3Fts3HashInsert(&pHash->hash, "unicode61", 10, (void *)pUnicode)
+#endif
+#ifdef SQLITE_ENABLE_ICU
+ || (pIcu && sqlite3Fts3HashInsert(&pHash->hash, "icu", 4, (void *)pIcu))
+#endif
+ ){
+ rc = SQLITE_NOMEM;
+ }
+ }
+
+#ifdef SQLITE_TEST
+ if( rc==SQLITE_OK ){
+ rc = sqlite3Fts3ExprInitTestInterface(db, &pHash->hash);
+ }
+#endif
+
+ /* Create the virtual table wrapper around the hash-table and overload
+ ** the four scalar functions. If this is successful, register the
+ ** module with sqlite.
+ */
+ if( SQLITE_OK==rc
+ && SQLITE_OK==(rc=sqlite3Fts3InitHashTable(db,&pHash->hash,"fts3_tokenizer"))
+ && SQLITE_OK==(rc = sqlite3_overload_function(db, "snippet", -1))
+ && SQLITE_OK==(rc = sqlite3_overload_function(db, "offsets", 1))
+ && SQLITE_OK==(rc = sqlite3_overload_function(db, "matchinfo", 1))
+ && SQLITE_OK==(rc = sqlite3_overload_function(db, "matchinfo", 2))
+ && SQLITE_OK==(rc = sqlite3_overload_function(db, "optimize", 1))
+ ){
+ pHash->nRef++;
+ rc = sqlite3_create_module_v2(
+ db, "fts3", &fts3Module, (void *)pHash, hashDestroy
+ );
+ if( rc==SQLITE_OK ){
+ pHash->nRef++;
+ rc = sqlite3_create_module_v2(
+ db, "fts4", &fts3Module, (void *)pHash, hashDestroy
+ );
+ }
+ if( rc==SQLITE_OK ){
+ pHash->nRef++;
+ rc = sqlite3Fts3InitTok(db, (void *)pHash, hashDestroy);
+ }
+ return rc;
+ }
+
+
+ /* An error has occurred. Delete the hash table and return the error code. */
+ assert( rc!=SQLITE_OK );
+ if( pHash ){
+ sqlite3Fts3HashClear(&pHash->hash);
+ sqlite3_free(pHash);
+ }
+ return rc;
+}
+
+/*
+** Allocate an Fts3MultiSegReader for each token in the expression headed
+** by pExpr.
+**
+** An Fts3SegReader object is a cursor that can seek or scan a range of
+** entries within a single segment b-tree. An Fts3MultiSegReader uses multiple
+** Fts3SegReader objects internally to provide an interface to seek or scan
+** within the union of all segments of a b-tree. Hence the name.
+**
+** If the allocated Fts3MultiSegReader just seeks to a single entry in a
+** segment b-tree (if the term is not a prefix or it is a prefix for which
+** there exists prefix b-tree of the right length) then it may be traversed
+** and merged incrementally. Otherwise, it has to be merged into an in-memory
+** doclist and then traversed.
+*/
+static void fts3EvalAllocateReaders(
+ Fts3Cursor *pCsr, /* FTS cursor handle */
+ Fts3Expr *pExpr, /* Allocate readers for this expression */
+ int *pnToken, /* OUT: Total number of tokens in phrase. */
+ int *pnOr, /* OUT: Total number of OR nodes in expr. */
+ int *pRc /* IN/OUT: Error code */
+){
+ if( pExpr && SQLITE_OK==*pRc ){
+ if( pExpr->eType==FTSQUERY_PHRASE ){
+ int i;
+ int nToken = pExpr->pPhrase->nToken;
+ *pnToken += nToken;
+ for(i=0; ipPhrase->aToken[i];
+ int rc = fts3TermSegReaderCursor(pCsr,
+ pToken->z, pToken->n, pToken->isPrefix, &pToken->pSegcsr
+ );
+ if( rc!=SQLITE_OK ){
+ *pRc = rc;
+ return;
+ }
+ }
+ assert( pExpr->pPhrase->iDoclistToken==0 );
+ pExpr->pPhrase->iDoclistToken = -1;
+ }else{
+ *pnOr += (pExpr->eType==FTSQUERY_OR);
+ fts3EvalAllocateReaders(pCsr, pExpr->pLeft, pnToken, pnOr, pRc);
+ fts3EvalAllocateReaders(pCsr, pExpr->pRight, pnToken, pnOr, pRc);
+ }
+ }
+}
+
+/*
+** Arguments pList/nList contain the doclist for token iToken of phrase p.
+** It is merged into the main doclist stored in p->doclist.aAll/nAll.
+**
+** This function assumes that pList points to a buffer allocated using
+** sqlite3_malloc(). This function takes responsibility for eventually
+** freeing the buffer.
+**
+** SQLITE_OK is returned if successful, or SQLITE_NOMEM if an error occurs.
+*/
+static int fts3EvalPhraseMergeToken(
+ Fts3Table *pTab, /* FTS Table pointer */
+ Fts3Phrase *p, /* Phrase to merge pList/nList into */
+ int iToken, /* Token pList/nList corresponds to */
+ char *pList, /* Pointer to doclist */
+ int nList /* Number of bytes in pList */
+){
+ int rc = SQLITE_OK;
+ assert( iToken!=p->iDoclistToken );
+
+ if( pList==0 ){
+ sqlite3_free(p->doclist.aAll);
+ p->doclist.aAll = 0;
+ p->doclist.nAll = 0;
+ }
+
+ else if( p->iDoclistToken<0 ){
+ p->doclist.aAll = pList;
+ p->doclist.nAll = nList;
+ }
+
+ else if( p->doclist.aAll==0 ){
+ sqlite3_free(pList);
+ }
+
+ else {
+ char *pLeft;
+ char *pRight;
+ int nLeft;
+ int nRight;
+ int nDiff;
+
+ if( p->iDoclistTokendoclist.aAll;
+ nLeft = p->doclist.nAll;
+ pRight = pList;
+ nRight = nList;
+ nDiff = iToken - p->iDoclistToken;
+ }else{
+ pRight = p->doclist.aAll;
+ nRight = p->doclist.nAll;
+ pLeft = pList;
+ nLeft = nList;
+ nDiff = p->iDoclistToken - iToken;
+ }
+
+ rc = fts3DoclistPhraseMerge(
+ pTab->bDescIdx, nDiff, pLeft, nLeft, &pRight, &nRight
+ );
+ sqlite3_free(pLeft);
+ p->doclist.aAll = pRight;
+ p->doclist.nAll = nRight;
+ }
+
+ if( iToken>p->iDoclistToken ) p->iDoclistToken = iToken;
+ return rc;
+}
+
+/*
+** Load the doclist for phrase p into p->doclist.aAll/nAll. The loaded doclist
+** does not take deferred tokens into account.
+**
+** SQLITE_OK is returned if no error occurs, otherwise an SQLite error code.
+*/
+static int fts3EvalPhraseLoad(
+ Fts3Cursor *pCsr, /* FTS Cursor handle */
+ Fts3Phrase *p /* Phrase object */
+){
+ Fts3Table *pTab = (Fts3Table *)pCsr->base.pVtab;
+ int iToken;
+ int rc = SQLITE_OK;
+
+ for(iToken=0; rc==SQLITE_OK && iTokennToken; iToken++){
+ Fts3PhraseToken *pToken = &p->aToken[iToken];
+ assert( pToken->pDeferred==0 || pToken->pSegcsr==0 );
+
+ if( pToken->pSegcsr ){
+ int nThis = 0;
+ char *pThis = 0;
+ rc = fts3TermSelect(pTab, pToken, p->iColumn, &nThis, &pThis);
+ if( rc==SQLITE_OK ){
+ rc = fts3EvalPhraseMergeToken(pTab, p, iToken, pThis, nThis);
+ }
+ }
+ assert( pToken->pSegcsr==0 );
+ }
+
+ return rc;
+}
+
+#ifndef SQLITE_DISABLE_FTS4_DEFERRED
+/*
+** This function is called on each phrase after the position lists for
+** any deferred tokens have been loaded into memory. It updates the phrases
+** current position list to include only those positions that are really
+** instances of the phrase (after considering deferred tokens). If this
+** means that the phrase does not appear in the current row, doclist.pList
+** and doclist.nList are both zeroed.
+**
+** SQLITE_OK is returned if no error occurs, otherwise an SQLite error code.
+*/
+static int fts3EvalDeferredPhrase(Fts3Cursor *pCsr, Fts3Phrase *pPhrase){
+ int iToken; /* Used to iterate through phrase tokens */
+ char *aPoslist = 0; /* Position list for deferred tokens */
+ int nPoslist = 0; /* Number of bytes in aPoslist */
+ int iPrev = -1; /* Token number of previous deferred token */
+ char *aFree = (pPhrase->doclist.bFreeList ? pPhrase->doclist.pList : 0);
+
+ for(iToken=0; iTokennToken; iToken++){
+ Fts3PhraseToken *pToken = &pPhrase->aToken[iToken];
+ Fts3DeferredToken *pDeferred = pToken->pDeferred;
+
+ if( pDeferred ){
+ char *pList;
+ int nList;
+ int rc = sqlite3Fts3DeferredTokenList(pDeferred, &pList, &nList);
+ if( rc!=SQLITE_OK ) return rc;
+
+ if( pList==0 ){
+ sqlite3_free(aPoslist);
+ sqlite3_free(aFree);
+ pPhrase->doclist.pList = 0;
+ pPhrase->doclist.nList = 0;
+ return SQLITE_OK;
+
+ }else if( aPoslist==0 ){
+ aPoslist = pList;
+ nPoslist = nList;
+
+ }else{
+ char *aOut = pList;
+ char *p1 = aPoslist;
+ char *p2 = aOut;
+
+ assert( iPrev>=0 );
+ fts3PoslistPhraseMerge(&aOut, iToken-iPrev, 0, 1, &p1, &p2);
+ sqlite3_free(aPoslist);
+ aPoslist = pList;
+ nPoslist = (int)(aOut - aPoslist);
+ if( nPoslist==0 ){
+ sqlite3_free(aPoslist);
+ sqlite3_free(aFree);
+ pPhrase->doclist.pList = 0;
+ pPhrase->doclist.nList = 0;
+ return SQLITE_OK;
+ }
+ }
+ iPrev = iToken;
+ }
+ }
+
+ if( iPrev>=0 ){
+ int nMaxUndeferred = pPhrase->iDoclistToken;
+ if( nMaxUndeferred<0 ){
+ pPhrase->doclist.pList = aPoslist;
+ pPhrase->doclist.nList = nPoslist;
+ pPhrase->doclist.iDocid = pCsr->iPrevId;
+ pPhrase->doclist.bFreeList = 1;
+ }else{
+ int nDistance;
+ char *p1;
+ char *p2;
+ char *aOut;
+
+ if( nMaxUndeferred>iPrev ){
+ p1 = aPoslist;
+ p2 = pPhrase->doclist.pList;
+ nDistance = nMaxUndeferred - iPrev;
+ }else{
+ p1 = pPhrase->doclist.pList;
+ p2 = aPoslist;
+ nDistance = iPrev - nMaxUndeferred;
+ }
+
+ aOut = (char *)sqlite3Fts3MallocZero(nPoslist+FTS3_BUFFER_PADDING);
+ if( !aOut ){
+ sqlite3_free(aPoslist);
+ return SQLITE_NOMEM;
+ }
+
+ pPhrase->doclist.pList = aOut;
+ assert( p1 && p2 );
+ if( fts3PoslistPhraseMerge(&aOut, nDistance, 0, 1, &p1, &p2) ){
+ pPhrase->doclist.bFreeList = 1;
+ pPhrase->doclist.nList = (int)(aOut - pPhrase->doclist.pList);
+ }else{
+ sqlite3_free(aOut);
+ pPhrase->doclist.pList = 0;
+ pPhrase->doclist.nList = 0;
+ }
+ sqlite3_free(aPoslist);
+ }
+ }
+
+ if( pPhrase->doclist.pList!=aFree ) sqlite3_free(aFree);
+ return SQLITE_OK;
+}
+#endif /* SQLITE_DISABLE_FTS4_DEFERRED */
+
+/*
+** Maximum number of tokens a phrase may have to be considered for the
+** incremental doclists strategy.
+*/
+#define MAX_INCR_PHRASE_TOKENS 4
+
+/*
+** This function is called for each Fts3Phrase in a full-text query
+** expression to initialize the mechanism for returning rows. Once this
+** function has been called successfully on an Fts3Phrase, it may be
+** used with fts3EvalPhraseNext() to iterate through the matching docids.
+**
+** If parameter bOptOk is true, then the phrase may (or may not) use the
+** incremental loading strategy. Otherwise, the entire doclist is loaded into
+** memory within this call.
+**
+** SQLITE_OK is returned if no error occurs, otherwise an SQLite error code.
+*/
+static int fts3EvalPhraseStart(Fts3Cursor *pCsr, int bOptOk, Fts3Phrase *p){
+ Fts3Table *pTab = (Fts3Table *)pCsr->base.pVtab;
+ int rc = SQLITE_OK; /* Error code */
+ int i;
+
+ /* Determine if doclists may be loaded from disk incrementally. This is
+ ** possible if the bOptOk argument is true, the FTS doclists will be
+ ** scanned in forward order, and the phrase consists of
+ ** MAX_INCR_PHRASE_TOKENS or fewer tokens, none of which are are "^first"
+ ** tokens or prefix tokens that cannot use a prefix-index. */
+ int bHaveIncr = 0;
+ int bIncrOk = (bOptOk
+ && pCsr->bDesc==pTab->bDescIdx
+ && p->nToken<=MAX_INCR_PHRASE_TOKENS && p->nToken>0
+#if defined(SQLITE_DEBUG) || defined(SQLITE_TEST)
+ && pTab->bNoIncrDoclist==0
+#endif
+ );
+ for(i=0; bIncrOk==1 && inToken; i++){
+ Fts3PhraseToken *pToken = &p->aToken[i];
+ if( pToken->bFirst || (pToken->pSegcsr!=0 && !pToken->pSegcsr->bLookup) ){
+ bIncrOk = 0;
+ }
+ if( pToken->pSegcsr ) bHaveIncr = 1;
+ }
+
+ if( bIncrOk && bHaveIncr ){
+ /* Use the incremental approach. */
+ int iCol = (p->iColumn >= pTab->nColumn ? -1 : p->iColumn);
+ for(i=0; rc==SQLITE_OK && inToken; i++){
+ Fts3PhraseToken *pToken = &p->aToken[i];
+ Fts3MultiSegReader *pSegcsr = pToken->pSegcsr;
+ if( pSegcsr ){
+ rc = sqlite3Fts3MsrIncrStart(pTab, pSegcsr, iCol, pToken->z, pToken->n);
+ }
+ }
+ p->bIncr = 1;
+ }else{
+ /* Load the full doclist for the phrase into memory. */
+ rc = fts3EvalPhraseLoad(pCsr, p);
+ p->bIncr = 0;
+ }
+
+ assert( rc!=SQLITE_OK || p->nToken<1 || p->aToken[0].pSegcsr==0 || p->bIncr );
+ return rc;
+}
+
+/*
+** This function is used to iterate backwards (from the end to start)
+** through doclists. It is used by this module to iterate through phrase
+** doclists in reverse and by the fts3_write.c module to iterate through
+** pending-terms lists when writing to databases with "order=desc".
+**
+** The doclist may be sorted in ascending (parameter bDescIdx==0) or
+** descending (parameter bDescIdx==1) order of docid. Regardless, this
+** function iterates from the end of the doclist to the beginning.
+*/
+void sqlite3Fts3DoclistPrev(
+ int bDescIdx, /* True if the doclist is desc */
+ char *aDoclist, /* Pointer to entire doclist */
+ int nDoclist, /* Length of aDoclist in bytes */
+ char **ppIter, /* IN/OUT: Iterator pointer */
+ sqlite3_int64 *piDocid, /* IN/OUT: Docid pointer */
+ int *pnList, /* OUT: List length pointer */
+ u8 *pbEof /* OUT: End-of-file flag */
+){
+ char *p = *ppIter;
+
+ assert( nDoclist>0 );
+ assert( *pbEof==0 );
+ assert_fts3_nc( p || *piDocid==0 );
+ assert( !p || (p>aDoclist && p<&aDoclist[nDoclist]) );
+
+ if( p==0 ){
+ sqlite3_int64 iDocid = 0;
+ char *pNext = 0;
+ char *pDocid = aDoclist;
+ char *pEnd = &aDoclist[nDoclist];
+ int iMul = 1;
+
+ while( pDocid0 );
+ assert( *pbEof==0 );
+ assert_fts3_nc( p || *piDocid==0 );
+ assert( !p || (p>=aDoclist && p<=&aDoclist[nDoclist]) );
+
+ if( p==0 ){
+ p = aDoclist;
+ p += sqlite3Fts3GetVarint(p, piDocid);
+ }else{
+ fts3PoslistCopy(0, &p);
+ while( p<&aDoclist[nDoclist] && *p==0 ) p++;
+ if( p>=&aDoclist[nDoclist] ){
+ *pbEof = 1;
+ }else{
+ sqlite3_int64 iVar;
+ p += sqlite3Fts3GetVarint(p, &iVar);
+ *piDocid += ((bDescIdx ? -1 : 1) * iVar);
+ }
+ }
+
+ *ppIter = p;
+}
+
+/*
+** Advance the iterator pDL to the next entry in pDL->aAll/nAll. Set *pbEof
+** to true if EOF is reached.
+*/
+static void fts3EvalDlPhraseNext(
+ Fts3Table *pTab,
+ Fts3Doclist *pDL,
+ u8 *pbEof
+){
+ char *pIter; /* Used to iterate through aAll */
+ char *pEnd; /* 1 byte past end of aAll */
+
+ if( pDL->pNextDocid ){
+ pIter = pDL->pNextDocid;
+ assert( pDL->aAll!=0 || pIter==0 );
+ }else{
+ pIter = pDL->aAll;
+ }
+
+ if( pIter==0 || pIter>=(pEnd = pDL->aAll + pDL->nAll) ){
+ /* We have already reached the end of this doclist. EOF. */
+ *pbEof = 1;
+ }else{
+ sqlite3_int64 iDelta;
+ pIter += sqlite3Fts3GetVarint(pIter, &iDelta);
+ if( pTab->bDescIdx==0 || pDL->pNextDocid==0 ){
+ pDL->iDocid += iDelta;
+ }else{
+ pDL->iDocid -= iDelta;
+ }
+ pDL->pList = pIter;
+ fts3PoslistCopy(0, &pIter);
+ pDL->nList = (int)(pIter - pDL->pList);
+
+ /* pIter now points just past the 0x00 that terminates the position-
+ ** list for document pDL->iDocid. However, if this position-list was
+ ** edited in place by fts3EvalNearTrim(), then pIter may not actually
+ ** point to the start of the next docid value. The following line deals
+ ** with this case by advancing pIter past the zero-padding added by
+ ** fts3EvalNearTrim(). */
+ while( pIterpNextDocid = pIter;
+ assert( pIter>=&pDL->aAll[pDL->nAll] || *pIter );
+ *pbEof = 0;
+ }
+}
+
+/*
+** Helper type used by fts3EvalIncrPhraseNext() and incrPhraseTokenNext().
+*/
+typedef struct TokenDoclist TokenDoclist;
+struct TokenDoclist {
+ int bIgnore;
+ sqlite3_int64 iDocid;
+ char *pList;
+ int nList;
+};
+
+/*
+** Token pToken is an incrementally loaded token that is part of a
+** multi-token phrase. Advance it to the next matching document in the
+** database and populate output variable *p with the details of the new
+** entry. Or, if the iterator has reached EOF, set *pbEof to true.
+**
+** If an error occurs, return an SQLite error code. Otherwise, return
+** SQLITE_OK.
+*/
+static int incrPhraseTokenNext(
+ Fts3Table *pTab, /* Virtual table handle */
+ Fts3Phrase *pPhrase, /* Phrase to advance token of */
+ int iToken, /* Specific token to advance */
+ TokenDoclist *p, /* OUT: Docid and doclist for new entry */
+ u8 *pbEof /* OUT: True if iterator is at EOF */
+){
+ int rc = SQLITE_OK;
+
+ if( pPhrase->iDoclistToken==iToken ){
+ assert( p->bIgnore==0 );
+ assert( pPhrase->aToken[iToken].pSegcsr==0 );
+ fts3EvalDlPhraseNext(pTab, &pPhrase->doclist, pbEof);
+ p->pList = pPhrase->doclist.pList;
+ p->nList = pPhrase->doclist.nList;
+ p->iDocid = pPhrase->doclist.iDocid;
+ }else{
+ Fts3PhraseToken *pToken = &pPhrase->aToken[iToken];
+ assert( pToken->pDeferred==0 );
+ assert( pToken->pSegcsr || pPhrase->iDoclistToken>=0 );
+ if( pToken->pSegcsr ){
+ assert( p->bIgnore==0 );
+ rc = sqlite3Fts3MsrIncrNext(
+ pTab, pToken->pSegcsr, &p->iDocid, &p->pList, &p->nList
+ );
+ if( p->pList==0 ) *pbEof = 1;
+ }else{
+ p->bIgnore = 1;
+ }
+ }
+
+ return rc;
+}
+
+
+/*
+** The phrase iterator passed as the second argument:
+**
+** * features at least one token that uses an incremental doclist, and
+**
+** * does not contain any deferred tokens.
+**
+** Advance it to the next matching documnent in the database and populate
+** the Fts3Doclist.pList and nList fields.
+**
+** If there is no "next" entry and no error occurs, then *pbEof is set to
+** 1 before returning. Otherwise, if no error occurs and the iterator is
+** successfully advanced, *pbEof is set to 0.
+**
+** If an error occurs, return an SQLite error code. Otherwise, return
+** SQLITE_OK.
+*/
+static int fts3EvalIncrPhraseNext(
+ Fts3Cursor *pCsr, /* FTS Cursor handle */
+ Fts3Phrase *p, /* Phrase object to advance to next docid */
+ u8 *pbEof /* OUT: Set to 1 if EOF */
+){
+ int rc = SQLITE_OK;
+ Fts3Doclist *pDL = &p->doclist;
+ Fts3Table *pTab = (Fts3Table *)pCsr->base.pVtab;
+ u8 bEof = 0;
+
+ /* This is only called if it is guaranteed that the phrase has at least
+ ** one incremental token. In which case the bIncr flag is set. */
+ assert( p->bIncr==1 );
+
+ if( p->nToken==1 ){
+ rc = sqlite3Fts3MsrIncrNext(pTab, p->aToken[0].pSegcsr,
+ &pDL->iDocid, &pDL->pList, &pDL->nList
+ );
+ if( pDL->pList==0 ) bEof = 1;
+ }else{
+ int bDescDoclist = pCsr->bDesc;
+ struct TokenDoclist a[MAX_INCR_PHRASE_TOKENS];
+
+ memset(a, 0, sizeof(a));
+ assert( p->nToken<=MAX_INCR_PHRASE_TOKENS );
+ assert( p->iDoclistTokennToken && bEof==0; i++){
+ rc = incrPhraseTokenNext(pTab, p, i, &a[i], &bEof);
+ if( a[i].bIgnore==0 && (bMaxSet==0 || DOCID_CMP(iMax, a[i].iDocid)<0) ){
+ iMax = a[i].iDocid;
+ bMaxSet = 1;
+ }
+ }
+ assert( rc!=SQLITE_OK || (p->nToken>=1 && a[p->nToken-1].bIgnore==0) );
+ assert( rc!=SQLITE_OK || bMaxSet );
+
+ /* Keep advancing iterators until they all point to the same document */
+ for(i=0; inToken; i++){
+ while( rc==SQLITE_OK && bEof==0
+ && a[i].bIgnore==0 && DOCID_CMP(a[i].iDocid, iMax)<0
+ ){
+ rc = incrPhraseTokenNext(pTab, p, i, &a[i], &bEof);
+ if( DOCID_CMP(a[i].iDocid, iMax)>0 ){
+ iMax = a[i].iDocid;
+ i = 0;
+ }
+ }
+ }
+
+ /* Check if the current entries really are a phrase match */
+ if( bEof==0 ){
+ int nList = 0;
+ int nByte = a[p->nToken-1].nList;
+ char *aDoclist = sqlite3_malloc64((i64)nByte+FTS3_BUFFER_PADDING);
+ if( !aDoclist ) return SQLITE_NOMEM;
+ memcpy(aDoclist, a[p->nToken-1].pList, nByte+1);
+ memset(&aDoclist[nByte], 0, FTS3_BUFFER_PADDING);
+
+ for(i=0; i<(p->nToken-1); i++){
+ if( a[i].bIgnore==0 ){
+ char *pL = a[i].pList;
+ char *pR = aDoclist;
+ char *pOut = aDoclist;
+ int nDist = p->nToken-1-i;
+ int res = fts3PoslistPhraseMerge(&pOut, nDist, 0, 1, &pL, &pR);
+ if( res==0 ) break;
+ nList = (int)(pOut - aDoclist);
+ }
+ }
+ if( i==(p->nToken-1) ){
+ pDL->iDocid = iMax;
+ pDL->pList = aDoclist;
+ pDL->nList = nList;
+ pDL->bFreeList = 1;
+ break;
+ }
+ sqlite3_free(aDoclist);
+ }
+ }
+ }
+
+ *pbEof = bEof;
+ return rc;
+}
+
+/*
+** Attempt to move the phrase iterator to point to the next matching docid.
+** If an error occurs, return an SQLite error code. Otherwise, return
+** SQLITE_OK.
+**
+** If there is no "next" entry and no error occurs, then *pbEof is set to
+** 1 before returning. Otherwise, if no error occurs and the iterator is
+** successfully advanced, *pbEof is set to 0.
+*/
+static int fts3EvalPhraseNext(
+ Fts3Cursor *pCsr, /* FTS Cursor handle */
+ Fts3Phrase *p, /* Phrase object to advance to next docid */
+ u8 *pbEof /* OUT: Set to 1 if EOF */
+){
+ int rc = SQLITE_OK;
+ Fts3Doclist *pDL = &p->doclist;
+ Fts3Table *pTab = (Fts3Table *)pCsr->base.pVtab;
+
+ if( p->bIncr ){
+ rc = fts3EvalIncrPhraseNext(pCsr, p, pbEof);
+ }else if( pCsr->bDesc!=pTab->bDescIdx && pDL->nAll ){
+ sqlite3Fts3DoclistPrev(pTab->bDescIdx, pDL->aAll, pDL->nAll,
+ &pDL->pNextDocid, &pDL->iDocid, &pDL->nList, pbEof
+ );
+ pDL->pList = pDL->pNextDocid;
+ }else{
+ fts3EvalDlPhraseNext(pTab, pDL, pbEof);
+ }
+
+ return rc;
+}
+
+/*
+**
+** If *pRc is not SQLITE_OK when this function is called, it is a no-op.
+** Otherwise, fts3EvalPhraseStart() is called on all phrases within the
+** expression. Also the Fts3Expr.bDeferred variable is set to true for any
+** expressions for which all descendent tokens are deferred.
+**
+** If parameter bOptOk is zero, then it is guaranteed that the
+** Fts3Phrase.doclist.aAll/nAll variables contain the entire doclist for
+** each phrase in the expression (subject to deferred token processing).
+** Or, if bOptOk is non-zero, then one or more tokens within the expression
+** may be loaded incrementally, meaning doclist.aAll/nAll is not available.
+**
+** If an error occurs within this function, *pRc is set to an SQLite error
+** code before returning.
+*/
+static void fts3EvalStartReaders(
+ Fts3Cursor *pCsr, /* FTS Cursor handle */
+ Fts3Expr *pExpr, /* Expression to initialize phrases in */
+ int *pRc /* IN/OUT: Error code */
+){
+ if( pExpr && SQLITE_OK==*pRc ){
+ if( pExpr->eType==FTSQUERY_PHRASE ){
+ int nToken = pExpr->pPhrase->nToken;
+ if( nToken ){
+ int i;
+ for(i=0; ipPhrase->aToken[i].pDeferred==0 ) break;
+ }
+ pExpr->bDeferred = (i==nToken);
+ }
+ *pRc = fts3EvalPhraseStart(pCsr, 1, pExpr->pPhrase);
+ }else{
+ fts3EvalStartReaders(pCsr, pExpr->pLeft, pRc);
+ fts3EvalStartReaders(pCsr, pExpr->pRight, pRc);
+ pExpr->bDeferred = (pExpr->pLeft->bDeferred && pExpr->pRight->bDeferred);
+ }
+ }
+}
+
+/*
+** An array of the following structures is assembled as part of the process
+** of selecting tokens to defer before the query starts executing (as part
+** of the xFilter() method). There is one element in the array for each
+** token in the FTS expression.
+**
+** Tokens are divided into AND/NEAR clusters. All tokens in a cluster belong
+** to phrases that are connected only by AND and NEAR operators (not OR or
+** NOT). When determining tokens to defer, each AND/NEAR cluster is considered
+** separately. The root of a tokens AND/NEAR cluster is stored in
+** Fts3TokenAndCost.pRoot.
+*/
+typedef struct Fts3TokenAndCost Fts3TokenAndCost;
+struct Fts3TokenAndCost {
+ Fts3Phrase *pPhrase; /* The phrase the token belongs to */
+ int iToken; /* Position of token in phrase */
+ Fts3PhraseToken *pToken; /* The token itself */
+ Fts3Expr *pRoot; /* Root of NEAR/AND cluster */
+ int nOvfl; /* Number of overflow pages to load doclist */
+ int iCol; /* The column the token must match */
+};
+
+/*
+** This function is used to populate an allocated Fts3TokenAndCost array.
+**
+** If *pRc is not SQLITE_OK when this function is called, it is a no-op.
+** Otherwise, if an error occurs during execution, *pRc is set to an
+** SQLite error code.
+*/
+static void fts3EvalTokenCosts(
+ Fts3Cursor *pCsr, /* FTS Cursor handle */
+ Fts3Expr *pRoot, /* Root of current AND/NEAR cluster */
+ Fts3Expr *pExpr, /* Expression to consider */
+ Fts3TokenAndCost **ppTC, /* Write new entries to *(*ppTC)++ */
+ Fts3Expr ***ppOr, /* Write new OR root to *(*ppOr)++ */
+ int *pRc /* IN/OUT: Error code */
+){
+ if( *pRc==SQLITE_OK ){
+ if( pExpr->eType==FTSQUERY_PHRASE ){
+ Fts3Phrase *pPhrase = pExpr->pPhrase;
+ int i;
+ for(i=0; *pRc==SQLITE_OK && inToken; i++){
+ Fts3TokenAndCost *pTC = (*ppTC)++;
+ pTC->pPhrase = pPhrase;
+ pTC->iToken = i;
+ pTC->pRoot = pRoot;
+ pTC->pToken = &pPhrase->aToken[i];
+ pTC->iCol = pPhrase->iColumn;
+ *pRc = sqlite3Fts3MsrOvfl(pCsr, pTC->pToken->pSegcsr, &pTC->nOvfl);
+ }
+ }else if( pExpr->eType!=FTSQUERY_NOT ){
+ assert( pExpr->eType==FTSQUERY_OR
+ || pExpr->eType==FTSQUERY_AND
+ || pExpr->eType==FTSQUERY_NEAR
+ );
+ assert( pExpr->pLeft && pExpr->pRight );
+ if( pExpr->eType==FTSQUERY_OR ){
+ pRoot = pExpr->pLeft;
+ **ppOr = pRoot;
+ (*ppOr)++;
+ }
+ fts3EvalTokenCosts(pCsr, pRoot, pExpr->pLeft, ppTC, ppOr, pRc);
+ if( pExpr->eType==FTSQUERY_OR ){
+ pRoot = pExpr->pRight;
+ **ppOr = pRoot;
+ (*ppOr)++;
+ }
+ fts3EvalTokenCosts(pCsr, pRoot, pExpr->pRight, ppTC, ppOr, pRc);
+ }
+ }
+}
+
+/*
+** Determine the average document (row) size in pages. If successful,
+** write this value to *pnPage and return SQLITE_OK. Otherwise, return
+** an SQLite error code.
+**
+** The average document size in pages is calculated by first calculating
+** determining the average size in bytes, B. If B is less than the amount
+** of data that will fit on a single leaf page of an intkey table in
+** this database, then the average docsize is 1. Otherwise, it is 1 plus
+** the number of overflow pages consumed by a record B bytes in size.
+*/
+static int fts3EvalAverageDocsize(Fts3Cursor *pCsr, int *pnPage){
+ int rc = SQLITE_OK;
+ if( pCsr->nRowAvg==0 ){
+ /* The average document size, which is required to calculate the cost
+ ** of each doclist, has not yet been determined. Read the required
+ ** data from the %_stat table to calculate it.
+ **
+ ** Entry 0 of the %_stat table is a blob containing (nCol+1) FTS3
+ ** varints, where nCol is the number of columns in the FTS3 table.
+ ** The first varint is the number of documents currently stored in
+ ** the table. The following nCol varints contain the total amount of
+ ** data stored in all rows of each column of the table, from left
+ ** to right.
+ */
+ Fts3Table *p = (Fts3Table*)pCsr->base.pVtab;
+ sqlite3_stmt *pStmt;
+ sqlite3_int64 nDoc = 0;
+ sqlite3_int64 nByte = 0;
+ const char *pEnd;
+ const char *a;
+
+ rc = sqlite3Fts3SelectDoctotal(p, &pStmt);
+ if( rc!=SQLITE_OK ) return rc;
+ a = sqlite3_column_blob(pStmt, 0);
+ testcase( a==0 ); /* If %_stat.value set to X'' */
+ if( a ){
+ pEnd = &a[sqlite3_column_bytes(pStmt, 0)];
+ a += sqlite3Fts3GetVarintBounded(a, pEnd, &nDoc);
+ while( anDoc = nDoc;
+ pCsr->nRowAvg = (int)(((nByte / nDoc) + p->nPgsz) / p->nPgsz);
+ assert( pCsr->nRowAvg>0 );
+ rc = sqlite3_reset(pStmt);
+ }
+
+ *pnPage = pCsr->nRowAvg;
+ return rc;
+}
+
+/*
+** This function is called to select the tokens (if any) that will be
+** deferred. The array aTC[] has already been populated when this is
+** called.
+**
+** This function is called once for each AND/NEAR cluster in the
+** expression. Each invocation determines which tokens to defer within
+** the cluster with root node pRoot. See comments above the definition
+** of struct Fts3TokenAndCost for more details.
+**
+** If no error occurs, SQLITE_OK is returned and sqlite3Fts3DeferToken()
+** called on each token to defer. Otherwise, an SQLite error code is
+** returned.
+*/
+static int fts3EvalSelectDeferred(
+ Fts3Cursor *pCsr, /* FTS Cursor handle */
+ Fts3Expr *pRoot, /* Consider tokens with this root node */
+ Fts3TokenAndCost *aTC, /* Array of expression tokens and costs */
+ int nTC /* Number of entries in aTC[] */
+){
+ Fts3Table *pTab = (Fts3Table *)pCsr->base.pVtab;
+ int nDocSize = 0; /* Number of pages per doc loaded */
+ int rc = SQLITE_OK; /* Return code */
+ int ii; /* Iterator variable for various purposes */
+ int nOvfl = 0; /* Total overflow pages used by doclists */
+ int nToken = 0; /* Total number of tokens in cluster */
+
+ int nMinEst = 0; /* The minimum count for any phrase so far. */
+ int nLoad4 = 1; /* (Phrases that will be loaded)^4. */
+
+ /* Tokens are never deferred for FTS tables created using the content=xxx
+ ** option. The reason being that it is not guaranteed that the content
+ ** table actually contains the same data as the index. To prevent this from
+ ** causing any problems, the deferred token optimization is completely
+ ** disabled for content=xxx tables. */
+ if( pTab->zContentTbl ){
+ return SQLITE_OK;
+ }
+
+ /* Count the tokens in this AND/NEAR cluster. If none of the doclists
+ ** associated with the tokens spill onto overflow pages, or if there is
+ ** only 1 token, exit early. No tokens to defer in this case. */
+ for(ii=0; ii0 );
+
+
+ /* Iterate through all tokens in this AND/NEAR cluster, in ascending order
+ ** of the number of overflow pages that will be loaded by the pager layer
+ ** to retrieve the entire doclist for the token from the full-text index.
+ ** Load the doclists for tokens that are either:
+ **
+ ** a. The cheapest token in the entire query (i.e. the one visited by the
+ ** first iteration of this loop), or
+ **
+ ** b. Part of a multi-token phrase.
+ **
+ ** After each token doclist is loaded, merge it with the others from the
+ ** same phrase and count the number of documents that the merged doclist
+ ** contains. Set variable "nMinEst" to the smallest number of documents in
+ ** any phrase doclist for which 1 or more token doclists have been loaded.
+ ** Let nOther be the number of other phrases for which it is certain that
+ ** one or more tokens will not be deferred.
+ **
+ ** Then, for each token, defer it if loading the doclist would result in
+ ** loading N or more overflow pages into memory, where N is computed as:
+ **
+ ** (nMinEst + 4^nOther - 1) / (4^nOther)
+ */
+ for(ii=0; iinOvfl)
+ ){
+ pTC = &aTC[iTC];
+ }
+ }
+ assert( pTC );
+
+ if( ii && pTC->nOvfl>=((nMinEst+(nLoad4/4)-1)/(nLoad4/4))*nDocSize ){
+ /* The number of overflow pages to load for this (and therefore all
+ ** subsequent) tokens is greater than the estimated number of pages
+ ** that will be loaded if all subsequent tokens are deferred.
+ */
+ Fts3PhraseToken *pToken = pTC->pToken;
+ rc = sqlite3Fts3DeferToken(pCsr, pToken, pTC->iCol);
+ fts3SegReaderCursorFree(pToken->pSegcsr);
+ pToken->pSegcsr = 0;
+ }else{
+ /* Set nLoad4 to the value of (4^nOther) for the next iteration of the
+ ** for-loop. Except, limit the value to 2^24 to prevent it from
+ ** overflowing the 32-bit integer it is stored in. */
+ if( ii<12 ) nLoad4 = nLoad4*4;
+
+ if( ii==0 || (pTC->pPhrase->nToken>1 && ii!=nToken-1) ){
+ /* Either this is the cheapest token in the entire query, or it is
+ ** part of a multi-token phrase. Either way, the entire doclist will
+ ** (eventually) be loaded into memory. It may as well be now. */
+ Fts3PhraseToken *pToken = pTC->pToken;
+ int nList = 0;
+ char *pList = 0;
+ rc = fts3TermSelect(pTab, pToken, pTC->iCol, &nList, &pList);
+ assert( rc==SQLITE_OK || pList==0 );
+ if( rc==SQLITE_OK ){
+ rc = fts3EvalPhraseMergeToken(
+ pTab, pTC->pPhrase, pTC->iToken,pList,nList
+ );
+ }
+ if( rc==SQLITE_OK ){
+ int nCount;
+ nCount = fts3DoclistCountDocids(
+ pTC->pPhrase->doclist.aAll, pTC->pPhrase->doclist.nAll
+ );
+ if( ii==0 || nCountpToken = 0;
+ }
+
+ return rc;
+}
+
+/*
+** This function is called from within the xFilter method. It initializes
+** the full-text query currently stored in pCsr->pExpr. To iterate through
+** the results of a query, the caller does:
+**
+** fts3EvalStart(pCsr);
+** while( 1 ){
+** fts3EvalNext(pCsr);
+** if( pCsr->bEof ) break;
+** ... return row pCsr->iPrevId to the caller ...
+** }
+*/
+static int fts3EvalStart(Fts3Cursor *pCsr){
+ Fts3Table *pTab = (Fts3Table *)pCsr->base.pVtab;
+ int rc = SQLITE_OK;
+ int nToken = 0;
+ int nOr = 0;
+
+ /* Allocate a MultiSegReader for each token in the expression. */
+ fts3EvalAllocateReaders(pCsr, pCsr->pExpr, &nToken, &nOr, &rc);
+
+ /* Determine which, if any, tokens in the expression should be deferred. */
+#ifndef SQLITE_DISABLE_FTS4_DEFERRED
+ if( rc==SQLITE_OK && nToken>1 && pTab->bFts4 ){
+ Fts3TokenAndCost *aTC;
+ aTC = (Fts3TokenAndCost *)sqlite3_malloc64(
+ sizeof(Fts3TokenAndCost) * nToken
+ + sizeof(Fts3Expr *) * nOr * 2
+ );
+
+ if( !aTC ){
+ rc = SQLITE_NOMEM;
+ }else{
+ Fts3Expr **apOr = (Fts3Expr **)&aTC[nToken];
+ int ii;
+ Fts3TokenAndCost *pTC = aTC;
+ Fts3Expr **ppOr = apOr;
+
+ fts3EvalTokenCosts(pCsr, 0, pCsr->pExpr, &pTC, &ppOr, &rc);
+ nToken = (int)(pTC-aTC);
+ nOr = (int)(ppOr-apOr);
+
+ if( rc==SQLITE_OK ){
+ rc = fts3EvalSelectDeferred(pCsr, 0, aTC, nToken);
+ for(ii=0; rc==SQLITE_OK && iipExpr, &rc);
+ return rc;
+}
+
+/*
+** Invalidate the current position list for phrase pPhrase.
+*/
+static void fts3EvalInvalidatePoslist(Fts3Phrase *pPhrase){
+ if( pPhrase->doclist.bFreeList ){
+ sqlite3_free(pPhrase->doclist.pList);
+ }
+ pPhrase->doclist.pList = 0;
+ pPhrase->doclist.nList = 0;
+ pPhrase->doclist.bFreeList = 0;
+}
+
+/*
+** This function is called to edit the position list associated with
+** the phrase object passed as the fifth argument according to a NEAR
+** condition. For example:
+**
+** abc NEAR/5 "def ghi"
+**
+** Parameter nNear is passed the NEAR distance of the expression (5 in
+** the example above). When this function is called, *paPoslist points to
+** the position list, and *pnToken is the number of phrase tokens in the
+** phrase on the other side of the NEAR operator to pPhrase. For example,
+** if pPhrase refers to the "def ghi" phrase, then *paPoslist points to
+** the position list associated with phrase "abc".
+**
+** All positions in the pPhrase position list that are not sufficiently
+** close to a position in the *paPoslist position list are removed. If this
+** leaves 0 positions, zero is returned. Otherwise, non-zero.
+**
+** Before returning, *paPoslist is set to point to the position lsit
+** associated with pPhrase. And *pnToken is set to the number of tokens in
+** pPhrase.
+*/
+static int fts3EvalNearTrim(
+ int nNear, /* NEAR distance. As in "NEAR/nNear". */
+ char *aTmp, /* Temporary space to use */
+ char **paPoslist, /* IN/OUT: Position list */
+ int *pnToken, /* IN/OUT: Tokens in phrase of *paPoslist */
+ Fts3Phrase *pPhrase /* The phrase object to trim the doclist of */
+){
+ int nParam1 = nNear + pPhrase->nToken;
+ int nParam2 = nNear + *pnToken;
+ int nNew;
+ char *p2;
+ char *pOut;
+ int res;
+
+ assert( pPhrase->doclist.pList );
+
+ p2 = pOut = pPhrase->doclist.pList;
+ res = fts3PoslistNearMerge(
+ &pOut, aTmp, nParam1, nParam2, paPoslist, &p2
+ );
+ if( res ){
+ nNew = (int)(pOut - pPhrase->doclist.pList) - 1;
+ assert_fts3_nc( nNew<=pPhrase->doclist.nList && nNew>0 );
+ if( nNew>=0 && nNew<=pPhrase->doclist.nList ){
+ assert( pPhrase->doclist.pList[nNew]=='\0' );
+ memset(&pPhrase->doclist.pList[nNew], 0, pPhrase->doclist.nList - nNew);
+ pPhrase->doclist.nList = nNew;
+ }
+ *paPoslist = pPhrase->doclist.pList;
+ *pnToken = pPhrase->nToken;
+ }
+
+ return res;
+}
+
+/*
+** This function is a no-op if *pRc is other than SQLITE_OK when it is called.
+** Otherwise, it advances the expression passed as the second argument to
+** point to the next matching row in the database. Expressions iterate through
+** matching rows in docid order. Ascending order if Fts3Cursor.bDesc is zero,
+** or descending if it is non-zero.
+**
+** If an error occurs, *pRc is set to an SQLite error code. Otherwise, if
+** successful, the following variables in pExpr are set:
+**
+** Fts3Expr.bEof (non-zero if EOF - there is no next row)
+** Fts3Expr.iDocid (valid if bEof==0. The docid of the next row)
+**
+** If the expression is of type FTSQUERY_PHRASE, and the expression is not
+** at EOF, then the following variables are populated with the position list
+** for the phrase for the visited row:
+**
+** FTs3Expr.pPhrase->doclist.nList (length of pList in bytes)
+** FTs3Expr.pPhrase->doclist.pList (pointer to position list)
+**
+** It says above that this function advances the expression to the next
+** matching row. This is usually true, but there are the following exceptions:
+**
+** 1. Deferred tokens are not taken into account. If a phrase consists
+** entirely of deferred tokens, it is assumed to match every row in
+** the db. In this case the position-list is not populated at all.
+**
+** Or, if a phrase contains one or more deferred tokens and one or
+** more non-deferred tokens, then the expression is advanced to the
+** next possible match, considering only non-deferred tokens. In other
+** words, if the phrase is "A B C", and "B" is deferred, the expression
+** is advanced to the next row that contains an instance of "A * C",
+** where "*" may match any single token. The position list in this case
+** is populated as for "A * C" before returning.
+**
+** 2. NEAR is treated as AND. If the expression is "x NEAR y", it is
+** advanced to point to the next row that matches "x AND y".
+**
+** See sqlite3Fts3EvalTestDeferred() for details on testing if a row is
+** really a match, taking into account deferred tokens and NEAR operators.
+*/
+static void fts3EvalNextRow(
+ Fts3Cursor *pCsr, /* FTS Cursor handle */
+ Fts3Expr *pExpr, /* Expr. to advance to next matching row */
+ int *pRc /* IN/OUT: Error code */
+){
+ if( *pRc==SQLITE_OK && pExpr->bEof==0 ){
+ int bDescDoclist = pCsr->bDesc; /* Used by DOCID_CMP() macro */
+ pExpr->bStart = 1;
+
+ switch( pExpr->eType ){
+ case FTSQUERY_NEAR:
+ case FTSQUERY_AND: {
+ Fts3Expr *pLeft = pExpr->pLeft;
+ Fts3Expr *pRight = pExpr->pRight;
+ assert( !pLeft->bDeferred || !pRight->bDeferred );
+
+ if( pLeft->bDeferred ){
+ /* LHS is entirely deferred. So we assume it matches every row.
+ ** Advance the RHS iterator to find the next row visited. */
+ fts3EvalNextRow(pCsr, pRight, pRc);
+ pExpr->iDocid = pRight->iDocid;
+ pExpr->bEof = pRight->bEof;
+ }else if( pRight->bDeferred ){
+ /* RHS is entirely deferred. So we assume it matches every row.
+ ** Advance the LHS iterator to find the next row visited. */
+ fts3EvalNextRow(pCsr, pLeft, pRc);
+ pExpr->iDocid = pLeft->iDocid;
+ pExpr->bEof = pLeft->bEof;
+ }else{
+ /* Neither the RHS or LHS are deferred. */
+ fts3EvalNextRow(pCsr, pLeft, pRc);
+ fts3EvalNextRow(pCsr, pRight, pRc);
+ while( !pLeft->bEof && !pRight->bEof && *pRc==SQLITE_OK ){
+ sqlite3_int64 iDiff = DOCID_CMP(pLeft->iDocid, pRight->iDocid);
+ if( iDiff==0 ) break;
+ if( iDiff<0 ){
+ fts3EvalNextRow(pCsr, pLeft, pRc);
+ }else{
+ fts3EvalNextRow(pCsr, pRight, pRc);
+ }
+ }
+ pExpr->iDocid = pLeft->iDocid;
+ pExpr->bEof = (pLeft->bEof || pRight->bEof);
+ if( pExpr->eType==FTSQUERY_NEAR && pExpr->bEof ){
+ assert( pRight->eType==FTSQUERY_PHRASE );
+ if( pRight->pPhrase->doclist.aAll ){
+ Fts3Doclist *pDl = &pRight->pPhrase->doclist;
+ while( *pRc==SQLITE_OK && pRight->bEof==0 ){
+ memset(pDl->pList, 0, pDl->nList);
+ fts3EvalNextRow(pCsr, pRight, pRc);
+ }
+ }
+ if( pLeft->pPhrase && pLeft->pPhrase->doclist.aAll ){
+ Fts3Doclist *pDl = &pLeft->pPhrase->doclist;
+ while( *pRc==SQLITE_OK && pLeft->bEof==0 ){
+ memset(pDl->pList, 0, pDl->nList);
+ fts3EvalNextRow(pCsr, pLeft, pRc);
+ }
+ }
+ pRight->bEof = pLeft->bEof = 1;
+ }
+ }
+ break;
+ }
+
+ case FTSQUERY_OR: {
+ Fts3Expr *pLeft = pExpr->pLeft;
+ Fts3Expr *pRight = pExpr->pRight;
+ sqlite3_int64 iCmp = DOCID_CMP(pLeft->iDocid, pRight->iDocid);
+
+ assert_fts3_nc( pLeft->bStart || pLeft->iDocid==pRight->iDocid );
+ assert_fts3_nc( pRight->bStart || pLeft->iDocid==pRight->iDocid );
+
+ if( pRight->bEof || (pLeft->bEof==0 && iCmp<0) ){
+ fts3EvalNextRow(pCsr, pLeft, pRc);
+ }else if( pLeft->bEof || iCmp>0 ){
+ fts3EvalNextRow(pCsr, pRight, pRc);
+ }else{
+ fts3EvalNextRow(pCsr, pLeft, pRc);
+ fts3EvalNextRow(pCsr, pRight, pRc);
+ }
+
+ pExpr->bEof = (pLeft->bEof && pRight->bEof);
+ iCmp = DOCID_CMP(pLeft->iDocid, pRight->iDocid);
+ if( pRight->bEof || (pLeft->bEof==0 && iCmp<0) ){
+ pExpr->iDocid = pLeft->iDocid;
+ }else{
+ pExpr->iDocid = pRight->iDocid;
+ }
+
+ break;
+ }
+
+ case FTSQUERY_NOT: {
+ Fts3Expr *pLeft = pExpr->pLeft;
+ Fts3Expr *pRight = pExpr->pRight;
+
+ if( pRight->bStart==0 ){
+ fts3EvalNextRow(pCsr, pRight, pRc);
+ assert( *pRc!=SQLITE_OK || pRight->bStart );
+ }
+
+ fts3EvalNextRow(pCsr, pLeft, pRc);
+ if( pLeft->bEof==0 ){
+ while( !*pRc
+ && !pRight->bEof
+ && DOCID_CMP(pLeft->iDocid, pRight->iDocid)>0
+ ){
+ fts3EvalNextRow(pCsr, pRight, pRc);
+ }
+ }
+ pExpr->iDocid = pLeft->iDocid;
+ pExpr->bEof = pLeft->bEof;
+ break;
+ }
+
+ default: {
+ Fts3Phrase *pPhrase = pExpr->pPhrase;
+ fts3EvalInvalidatePoslist(pPhrase);
+ *pRc = fts3EvalPhraseNext(pCsr, pPhrase, &pExpr->bEof);
+ pExpr->iDocid = pPhrase->doclist.iDocid;
+ break;
+ }
+ }
+ }
+}
+
+/*
+** If *pRc is not SQLITE_OK, or if pExpr is not the root node of a NEAR
+** cluster, then this function returns 1 immediately.
+**
+** Otherwise, it checks if the current row really does match the NEAR
+** expression, using the data currently stored in the position lists
+** (Fts3Expr->pPhrase.doclist.pList/nList) for each phrase in the expression.
+**
+** If the current row is a match, the position list associated with each
+** phrase in the NEAR expression is edited in place to contain only those
+** phrase instances sufficiently close to their peers to satisfy all NEAR
+** constraints. In this case it returns 1. If the NEAR expression does not
+** match the current row, 0 is returned. The position lists may or may not
+** be edited if 0 is returned.
+*/
+static int fts3EvalNearTest(Fts3Expr *pExpr, int *pRc){
+ int res = 1;
+
+ /* The following block runs if pExpr is the root of a NEAR query.
+ ** For example, the query:
+ **
+ ** "w" NEAR "x" NEAR "y" NEAR "z"
+ **
+ ** which is represented in tree form as:
+ **
+ ** |
+ ** +--NEAR--+ <-- root of NEAR query
+ ** | |
+ ** +--NEAR--+ "z"
+ ** | |
+ ** +--NEAR--+ "y"
+ ** | |
+ ** "w" "x"
+ **
+ ** The right-hand child of a NEAR node is always a phrase. The
+ ** left-hand child may be either a phrase or a NEAR node. There are
+ ** no exceptions to this - it's the way the parser in fts3_expr.c works.
+ */
+ if( *pRc==SQLITE_OK
+ && pExpr->eType==FTSQUERY_NEAR
+ && (pExpr->pParent==0 || pExpr->pParent->eType!=FTSQUERY_NEAR)
+ ){
+ Fts3Expr *p;
+ sqlite3_int64 nTmp = 0; /* Bytes of temp space */
+ char *aTmp; /* Temp space for PoslistNearMerge() */
+
+ /* Allocate temporary working space. */
+ for(p=pExpr; p->pLeft; p=p->pLeft){
+ assert( p->pRight->pPhrase->doclist.nList>0 );
+ nTmp += p->pRight->pPhrase->doclist.nList;
+ }
+ nTmp += p->pPhrase->doclist.nList;
+ aTmp = sqlite3_malloc64(nTmp*2 + FTS3_VARINT_MAX);
+ if( !aTmp ){
+ *pRc = SQLITE_NOMEM;
+ res = 0;
+ }else{
+ char *aPoslist = p->pPhrase->doclist.pList;
+ int nToken = p->pPhrase->nToken;
+
+ for(p=p->pParent;res && p && p->eType==FTSQUERY_NEAR; p=p->pParent){
+ Fts3Phrase *pPhrase = p->pRight->pPhrase;
+ int nNear = p->nNear;
+ res = fts3EvalNearTrim(nNear, aTmp, &aPoslist, &nToken, pPhrase);
+ }
+
+ aPoslist = pExpr->pRight->pPhrase->doclist.pList;
+ nToken = pExpr->pRight->pPhrase->nToken;
+ for(p=pExpr->pLeft; p && res; p=p->pLeft){
+ int nNear;
+ Fts3Phrase *pPhrase;
+ assert( p->pParent && p->pParent->pLeft==p );
+ nNear = p->pParent->nNear;
+ pPhrase = (
+ p->eType==FTSQUERY_NEAR ? p->pRight->pPhrase : p->pPhrase
+ );
+ res = fts3EvalNearTrim(nNear, aTmp, &aPoslist, &nToken, pPhrase);
+ }
+ }
+
+ sqlite3_free(aTmp);
+ }
+
+ return res;
+}
+
+/*
+** This function is a helper function for sqlite3Fts3EvalTestDeferred().
+** Assuming no error occurs or has occurred, It returns non-zero if the
+** expression passed as the second argument matches the row that pCsr
+** currently points to, or zero if it does not.
+**
+** If *pRc is not SQLITE_OK when this function is called, it is a no-op.
+** If an error occurs during execution of this function, *pRc is set to
+** the appropriate SQLite error code. In this case the returned value is
+** undefined.
+*/
+static int fts3EvalTestExpr(
+ Fts3Cursor *pCsr, /* FTS cursor handle */
+ Fts3Expr *pExpr, /* Expr to test. May or may not be root. */
+ int *pRc /* IN/OUT: Error code */
+){
+ int bHit = 1; /* Return value */
+ if( *pRc==SQLITE_OK ){
+ switch( pExpr->eType ){
+ case FTSQUERY_NEAR:
+ case FTSQUERY_AND:
+ bHit = (
+ fts3EvalTestExpr(pCsr, pExpr->pLeft, pRc)
+ && fts3EvalTestExpr(pCsr, pExpr->pRight, pRc)
+ && fts3EvalNearTest(pExpr, pRc)
+ );
+
+ /* If the NEAR expression does not match any rows, zero the doclist for
+ ** all phrases involved in the NEAR. This is because the snippet(),
+ ** offsets() and matchinfo() functions are not supposed to recognize
+ ** any instances of phrases that are part of unmatched NEAR queries.
+ ** For example if this expression:
+ **
+ ** ... MATCH 'a OR (b NEAR c)'
+ **
+ ** is matched against a row containing:
+ **
+ ** 'a b d e'
+ **
+ ** then any snippet() should ony highlight the "a" term, not the "b"
+ ** (as "b" is part of a non-matching NEAR clause).
+ */
+ if( bHit==0
+ && pExpr->eType==FTSQUERY_NEAR
+ && (pExpr->pParent==0 || pExpr->pParent->eType!=FTSQUERY_NEAR)
+ ){
+ Fts3Expr *p;
+ for(p=pExpr; p->pPhrase==0; p=p->pLeft){
+ if( p->pRight->iDocid==pCsr->iPrevId ){
+ fts3EvalInvalidatePoslist(p->pRight->pPhrase);
+ }
+ }
+ if( p->iDocid==pCsr->iPrevId ){
+ fts3EvalInvalidatePoslist(p->pPhrase);
+ }
+ }
+
+ break;
+
+ case FTSQUERY_OR: {
+ int bHit1 = fts3EvalTestExpr(pCsr, pExpr->pLeft, pRc);
+ int bHit2 = fts3EvalTestExpr(pCsr, pExpr->pRight, pRc);
+ bHit = bHit1 || bHit2;
+ break;
+ }
+
+ case FTSQUERY_NOT:
+ bHit = (
+ fts3EvalTestExpr(pCsr, pExpr->pLeft, pRc)
+ && !fts3EvalTestExpr(pCsr, pExpr->pRight, pRc)
+ );
+ break;
+
+ default: {
+#ifndef SQLITE_DISABLE_FTS4_DEFERRED
+ if( pCsr->pDeferred && (pExpr->bDeferred || (
+ pExpr->iDocid==pCsr->iPrevId && pExpr->pPhrase->doclist.pList
+ ))){
+ Fts3Phrase *pPhrase = pExpr->pPhrase;
+ if( pExpr->bDeferred ){
+ fts3EvalInvalidatePoslist(pPhrase);
+ }
+ *pRc = fts3EvalDeferredPhrase(pCsr, pPhrase);
+ bHit = (pPhrase->doclist.pList!=0);
+ pExpr->iDocid = pCsr->iPrevId;
+ }else
+#endif
+ {
+ bHit = (
+ pExpr->bEof==0 && pExpr->iDocid==pCsr->iPrevId
+ && pExpr->pPhrase->doclist.nList>0
+ );
+ }
+ break;
+ }
+ }
+ }
+ return bHit;
+}
+
+/*
+** This function is called as the second part of each xNext operation when
+** iterating through the results of a full-text query. At this point the
+** cursor points to a row that matches the query expression, with the
+** following caveats:
+**
+** * Up until this point, "NEAR" operators in the expression have been
+** treated as "AND".
+**
+** * Deferred tokens have not yet been considered.
+**
+** If *pRc is not SQLITE_OK when this function is called, it immediately
+** returns 0. Otherwise, it tests whether or not after considering NEAR
+** operators and deferred tokens the current row is still a match for the
+** expression. It returns 1 if both of the following are true:
+**
+** 1. *pRc is SQLITE_OK when this function returns, and
+**
+** 2. After scanning the current FTS table row for the deferred tokens,
+** it is determined that the row does *not* match the query.
+**
+** Or, if no error occurs and it seems the current row does match the FTS
+** query, return 0.
+*/
+int sqlite3Fts3EvalTestDeferred(Fts3Cursor *pCsr, int *pRc){
+ int rc = *pRc;
+ int bMiss = 0;
+ if( rc==SQLITE_OK ){
+
+ /* If there are one or more deferred tokens, load the current row into
+ ** memory and scan it to determine the position list for each deferred
+ ** token. Then, see if this row is really a match, considering deferred
+ ** tokens and NEAR operators (neither of which were taken into account
+ ** earlier, by fts3EvalNextRow()).
+ */
+ if( pCsr->pDeferred ){
+ rc = fts3CursorSeek(0, pCsr);
+ if( rc==SQLITE_OK ){
+ rc = sqlite3Fts3CacheDeferredDoclists(pCsr);
+ }
+ }
+ bMiss = (0==fts3EvalTestExpr(pCsr, pCsr->pExpr, &rc));
+
+ /* Free the position-lists accumulated for each deferred token above. */
+ sqlite3Fts3FreeDeferredDoclists(pCsr);
+ *pRc = rc;
+ }
+ return (rc==SQLITE_OK && bMiss);
+}
+
+/*
+** Advance to the next document that matches the FTS expression in
+** Fts3Cursor.pExpr.
+*/
+static int fts3EvalNext(Fts3Cursor *pCsr){
+ int rc = SQLITE_OK; /* Return Code */
+ Fts3Expr *pExpr = pCsr->pExpr;
+ assert( pCsr->isEof==0 );
+ if( pExpr==0 ){
+ pCsr->isEof = 1;
+ }else{
+ do {
+ if( pCsr->isRequireSeek==0 ){
+ sqlite3_reset(pCsr->pStmt);
+ }
+ assert( sqlite3_data_count(pCsr->pStmt)==0 );
+ fts3EvalNextRow(pCsr, pExpr, &rc);
+ pCsr->isEof = pExpr->bEof;
+ pCsr->isRequireSeek = 1;
+ pCsr->isMatchinfoNeeded = 1;
+ pCsr->iPrevId = pExpr->iDocid;
+ }while( pCsr->isEof==0 && sqlite3Fts3EvalTestDeferred(pCsr, &rc) );
+ }
+
+ /* Check if the cursor is past the end of the docid range specified
+ ** by Fts3Cursor.iMinDocid/iMaxDocid. If so, set the EOF flag. */
+ if( rc==SQLITE_OK && (
+ (pCsr->bDesc==0 && pCsr->iPrevId>pCsr->iMaxDocid)
+ || (pCsr->bDesc!=0 && pCsr->iPrevIdiMinDocid)
+ )){
+ pCsr->isEof = 1;
+ }
+
+ return rc;
+}
+
+/*
+** Restart interation for expression pExpr so that the next call to
+** fts3EvalNext() visits the first row. Do not allow incremental
+** loading or merging of phrase doclists for this iteration.
+**
+** If *pRc is other than SQLITE_OK when this function is called, it is
+** a no-op. If an error occurs within this function, *pRc is set to an
+** SQLite error code before returning.
+*/
+static void fts3EvalRestart(
+ Fts3Cursor *pCsr,
+ Fts3Expr *pExpr,
+ int *pRc
+){
+ if( pExpr && *pRc==SQLITE_OK ){
+ Fts3Phrase *pPhrase = pExpr->pPhrase;
+
+ if( pPhrase ){
+ fts3EvalInvalidatePoslist(pPhrase);
+ if( pPhrase->bIncr ){
+ int i;
+ for(i=0; inToken; i++){
+ Fts3PhraseToken *pToken = &pPhrase->aToken[i];
+ assert( pToken->pDeferred==0 );
+ if( pToken->pSegcsr ){
+ sqlite3Fts3MsrIncrRestart(pToken->pSegcsr);
+ }
+ }
+ *pRc = fts3EvalPhraseStart(pCsr, 0, pPhrase);
+ }
+ pPhrase->doclist.pNextDocid = 0;
+ pPhrase->doclist.iDocid = 0;
+ pPhrase->pOrPoslist = 0;
+ }
+
+ pExpr->iDocid = 0;
+ pExpr->bEof = 0;
+ pExpr->bStart = 0;
+
+ fts3EvalRestart(pCsr, pExpr->pLeft, pRc);
+ fts3EvalRestart(pCsr, pExpr->pRight, pRc);
+ }
+}
+
+/*
+** After allocating the Fts3Expr.aMI[] array for each phrase in the
+** expression rooted at pExpr, the cursor iterates through all rows matched
+** by pExpr, calling this function for each row. This function increments
+** the values in Fts3Expr.aMI[] according to the position-list currently
+** found in Fts3Expr.pPhrase->doclist.pList for each of the phrase
+** expression nodes.
+*/
+static void fts3EvalUpdateCounts(Fts3Expr *pExpr, int nCol){
+ if( pExpr ){
+ Fts3Phrase *pPhrase = pExpr->pPhrase;
+ if( pPhrase && pPhrase->doclist.pList ){
+ int iCol = 0;
+ char *p = pPhrase->doclist.pList;
+
+ do{
+ u8 c = 0;
+ int iCnt = 0;
+ while( 0xFE & (*p | c) ){
+ if( (c&0x80)==0 ) iCnt++;
+ c = *p++ & 0x80;
+ }
+
+ /* aMI[iCol*3 + 1] = Number of occurrences
+ ** aMI[iCol*3 + 2] = Number of rows containing at least one instance
+ */
+ pExpr->aMI[iCol*3 + 1] += iCnt;
+ pExpr->aMI[iCol*3 + 2] += (iCnt>0);
+ if( *p==0x00 ) break;
+ p++;
+ p += fts3GetVarint32(p, &iCol);
+ }while( iColpLeft, nCol);
+ fts3EvalUpdateCounts(pExpr->pRight, nCol);
+ }
+}
+
+/*
+** This is an sqlite3Fts3ExprIterate() callback. If the Fts3Expr.aMI[] array
+** has not yet been allocated, allocate and zero it. Otherwise, just zero
+** it.
+*/
+static int fts3AllocateMSI(Fts3Expr *pExpr, int iPhrase, void *pCtx){
+ Fts3Table *pTab = (Fts3Table*)pCtx;
+ UNUSED_PARAMETER(iPhrase);
+ if( pExpr->aMI==0 ){
+ pExpr->aMI = (u32 *)sqlite3_malloc64(pTab->nColumn * 3 * sizeof(u32));
+ if( pExpr->aMI==0 ) return SQLITE_NOMEM;
+ }
+ memset(pExpr->aMI, 0, pTab->nColumn * 3 * sizeof(u32));
+ return SQLITE_OK;
+}
+
+/*
+** Expression pExpr must be of type FTSQUERY_PHRASE.
+**
+** If it is not already allocated and populated, this function allocates and
+** populates the Fts3Expr.aMI[] array for expression pExpr. If pExpr is part
+** of a NEAR expression, then it also allocates and populates the same array
+** for all other phrases that are part of the NEAR expression.
+**
+** SQLITE_OK is returned if the aMI[] array is successfully allocated and
+** populated. Otherwise, if an error occurs, an SQLite error code is returned.
+*/
+static int fts3EvalGatherStats(
+ Fts3Cursor *pCsr, /* Cursor object */
+ Fts3Expr *pExpr /* FTSQUERY_PHRASE expression */
+){
+ int rc = SQLITE_OK; /* Return code */
+
+ assert( pExpr->eType==FTSQUERY_PHRASE );
+ if( pExpr->aMI==0 ){
+ Fts3Table *pTab = (Fts3Table *)pCsr->base.pVtab;
+ Fts3Expr *pRoot; /* Root of NEAR expression */
+
+ sqlite3_int64 iPrevId = pCsr->iPrevId;
+ sqlite3_int64 iDocid;
+ u8 bEof;
+
+ /* Find the root of the NEAR expression */
+ pRoot = pExpr;
+ while( pRoot->pParent
+ && (pRoot->pParent->eType==FTSQUERY_NEAR || pRoot->bDeferred)
+ ){
+ pRoot = pRoot->pParent;
+ }
+ iDocid = pRoot->iDocid;
+ bEof = pRoot->bEof;
+ assert( pRoot->bStart );
+
+ /* Allocate space for the aMSI[] array of each FTSQUERY_PHRASE node */
+ rc = sqlite3Fts3ExprIterate(pRoot, fts3AllocateMSI, (void*)pTab);
+ if( rc!=SQLITE_OK ) return rc;
+ fts3EvalRestart(pCsr, pRoot, &rc);
+
+ while( pCsr->isEof==0 && rc==SQLITE_OK ){
+
+ do {
+ /* Ensure the %_content statement is reset. */
+ if( pCsr->isRequireSeek==0 ) sqlite3_reset(pCsr->pStmt);
+ assert( sqlite3_data_count(pCsr->pStmt)==0 );
+
+ /* Advance to the next document */
+ fts3EvalNextRow(pCsr, pRoot, &rc);
+ pCsr->isEof = pRoot->bEof;
+ pCsr->isRequireSeek = 1;
+ pCsr->isMatchinfoNeeded = 1;
+ pCsr->iPrevId = pRoot->iDocid;
+ }while( pCsr->isEof==0
+ && pRoot->eType==FTSQUERY_NEAR
+ && sqlite3Fts3EvalTestDeferred(pCsr, &rc)
+ );
+
+ if( rc==SQLITE_OK && pCsr->isEof==0 ){
+ fts3EvalUpdateCounts(pRoot, pTab->nColumn);
+ }
+ }
+
+ pCsr->isEof = 0;
+ pCsr->iPrevId = iPrevId;
+
+ if( bEof ){
+ pRoot->bEof = bEof;
+ }else{
+ /* Caution: pRoot may iterate through docids in ascending or descending
+ ** order. For this reason, even though it seems more defensive, the
+ ** do loop can not be written:
+ **
+ ** do {...} while( pRoot->iDocidbEof==0 );
+ if( pRoot->bEof ) rc = FTS_CORRUPT_VTAB;
+ }while( pRoot->iDocid!=iDocid && rc==SQLITE_OK );
+ }
+ }
+ return rc;
+}
+
+/*
+** This function is used by the matchinfo() module to query a phrase
+** expression node for the following information:
+**
+** 1. The total number of occurrences of the phrase in each column of
+** the FTS table (considering all rows), and
+**
+** 2. For each column, the number of rows in the table for which the
+** column contains at least one instance of the phrase.
+**
+** If no error occurs, SQLITE_OK is returned and the values for each column
+** written into the array aiOut as follows:
+**
+** aiOut[iCol*3 + 1] = Number of occurrences
+** aiOut[iCol*3 + 2] = Number of rows containing at least one instance
+**
+** Caveats:
+**
+** * If a phrase consists entirely of deferred tokens, then all output
+** values are set to the number of documents in the table. In other
+** words we assume that very common tokens occur exactly once in each
+** column of each row of the table.
+**
+** * If a phrase contains some deferred tokens (and some non-deferred
+** tokens), count the potential occurrence identified by considering
+** the non-deferred tokens instead of actual phrase occurrences.
+**
+** * If the phrase is part of a NEAR expression, then only phrase instances
+** that meet the NEAR constraint are included in the counts.
+*/
+int sqlite3Fts3EvalPhraseStats(
+ Fts3Cursor *pCsr, /* FTS cursor handle */
+ Fts3Expr *pExpr, /* Phrase expression */
+ u32 *aiOut /* Array to write results into (see above) */
+){
+ Fts3Table *pTab = (Fts3Table *)pCsr->base.pVtab;
+ int rc = SQLITE_OK;
+ int iCol;
+
+ if( pExpr->bDeferred && pExpr->pParent->eType!=FTSQUERY_NEAR ){
+ assert( pCsr->nDoc>0 );
+ for(iCol=0; iColnColumn; iCol++){
+ aiOut[iCol*3 + 1] = (u32)pCsr->nDoc;
+ aiOut[iCol*3 + 2] = (u32)pCsr->nDoc;
+ }
+ }else{
+ rc = fts3EvalGatherStats(pCsr, pExpr);
+ if( rc==SQLITE_OK ){
+ assert( pExpr->aMI );
+ for(iCol=0; iColnColumn; iCol++){
+ aiOut[iCol*3 + 1] = pExpr->aMI[iCol*3 + 1];
+ aiOut[iCol*3 + 2] = pExpr->aMI[iCol*3 + 2];
+ }
+ }
+ }
+
+ return rc;
+}
+
+/*
+** The expression pExpr passed as the second argument to this function
+** must be of type FTSQUERY_PHRASE.
+**
+** The returned value is either NULL or a pointer to a buffer containing
+** a position-list indicating the occurrences of the phrase in column iCol
+** of the current row.
+**
+** More specifically, the returned buffer contains 1 varint for each
+** occurrence of the phrase in the column, stored using the normal (delta+2)
+** compression and is terminated by either an 0x01 or 0x00 byte. For example,
+** if the requested column contains "a b X c d X X" and the position-list
+** for 'X' is requested, the buffer returned may contain:
+**
+** 0x04 0x05 0x03 0x01 or 0x04 0x05 0x03 0x00
+**
+** This function works regardless of whether or not the phrase is deferred,
+** incremental, or neither.
+*/
+int sqlite3Fts3EvalPhrasePoslist(
+ Fts3Cursor *pCsr, /* FTS3 cursor object */
+ Fts3Expr *pExpr, /* Phrase to return doclist for */
+ int iCol, /* Column to return position list for */
+ char **ppOut /* OUT: Pointer to position list */
+){
+ Fts3Phrase *pPhrase = pExpr->pPhrase;
+ Fts3Table *pTab = (Fts3Table *)pCsr->base.pVtab;
+ char *pIter;
+ int iThis;
+ sqlite3_int64 iDocid;
+
+ /* If this phrase is applies specifically to some column other than
+ ** column iCol, return a NULL pointer. */
+ *ppOut = 0;
+ assert( iCol>=0 && iColnColumn );
+ if( (pPhrase->iColumnnColumn && pPhrase->iColumn!=iCol) ){
+ return SQLITE_OK;
+ }
+
+ iDocid = pExpr->iDocid;
+ pIter = pPhrase->doclist.pList;
+ if( iDocid!=pCsr->iPrevId || pExpr->bEof ){
+ int rc = SQLITE_OK;
+ int bDescDoclist = pTab->bDescIdx; /* For DOCID_CMP macro */
+ int bOr = 0;
+ u8 bTreeEof = 0;
+ Fts3Expr *p; /* Used to iterate from pExpr to root */
+ Fts3Expr *pNear; /* Most senior NEAR ancestor (or pExpr) */
+ Fts3Expr *pRun; /* Closest non-deferred ancestor of pNear */
+ int bMatch;
+
+ /* Check if this phrase descends from an OR expression node. If not,
+ ** return NULL. Otherwise, the entry that corresponds to docid
+ ** pCsr->iPrevId may lie earlier in the doclist buffer. Or, if the
+ ** tree that the node is part of has been marked as EOF, but the node
+ ** itself is not EOF, then it may point to an earlier entry. */
+ pNear = pExpr;
+ for(p=pExpr->pParent; p; p=p->pParent){
+ if( p->eType==FTSQUERY_OR ) bOr = 1;
+ if( p->eType==FTSQUERY_NEAR ) pNear = p;
+ if( p->bEof ) bTreeEof = 1;
+ }
+ if( bOr==0 ) return SQLITE_OK;
+ pRun = pNear;
+ while( pRun->bDeferred ){
+ assert( pRun->pParent );
+ pRun = pRun->pParent;
+ }
+
+ /* This is the descendent of an OR node. In this case we cannot use
+ ** an incremental phrase. Load the entire doclist for the phrase
+ ** into memory in this case. */
+ if( pPhrase->bIncr ){
+ int bEofSave = pRun->bEof;
+ fts3EvalRestart(pCsr, pRun, &rc);
+ while( rc==SQLITE_OK && !pRun->bEof ){
+ fts3EvalNextRow(pCsr, pRun, &rc);
+ if( bEofSave==0 && pRun->iDocid==iDocid ) break;
+ }
+ assert( rc!=SQLITE_OK || pPhrase->bIncr==0 );
+ if( rc==SQLITE_OK && pRun->bEof!=bEofSave ){
+ rc = FTS_CORRUPT_VTAB;
+ }
+ }
+ if( bTreeEof ){
+ while( rc==SQLITE_OK && !pRun->bEof ){
+ fts3EvalNextRow(pCsr, pRun, &rc);
+ }
+ }
+ if( rc!=SQLITE_OK ) return rc;
+
+ bMatch = 1;
+ for(p=pNear; p; p=p->pLeft){
+ u8 bEof = 0;
+ Fts3Expr *pTest = p;
+ Fts3Phrase *pPh;
+ assert( pTest->eType==FTSQUERY_NEAR || pTest->eType==FTSQUERY_PHRASE );
+ if( pTest->eType==FTSQUERY_NEAR ) pTest = pTest->pRight;
+ assert( pTest->eType==FTSQUERY_PHRASE );
+ pPh = pTest->pPhrase;
+
+ pIter = pPh->pOrPoslist;
+ iDocid = pPh->iOrDocid;
+ if( pCsr->bDesc==bDescDoclist ){
+ bEof = !pPh->doclist.nAll ||
+ (pIter >= (pPh->doclist.aAll + pPh->doclist.nAll));
+ while( (pIter==0 || DOCID_CMP(iDocid, pCsr->iPrevId)<0 ) && bEof==0 ){
+ sqlite3Fts3DoclistNext(
+ bDescDoclist, pPh->doclist.aAll, pPh->doclist.nAll,
+ &pIter, &iDocid, &bEof
+ );
+ }
+ }else{
+ bEof = !pPh->doclist.nAll || (pIter && pIter<=pPh->doclist.aAll);
+ while( (pIter==0 || DOCID_CMP(iDocid, pCsr->iPrevId)>0 ) && bEof==0 ){
+ int dummy;
+ sqlite3Fts3DoclistPrev(
+ bDescDoclist, pPh->doclist.aAll, pPh->doclist.nAll,
+ &pIter, &iDocid, &dummy, &bEof
+ );
+ }
+ }
+ pPh->pOrPoslist = pIter;
+ pPh->iOrDocid = iDocid;
+ if( bEof || iDocid!=pCsr->iPrevId ) bMatch = 0;
+ }
+
+ if( bMatch ){
+ pIter = pPhrase->pOrPoslist;
+ }else{
+ pIter = 0;
+ }
+ }
+ if( pIter==0 ) return SQLITE_OK;
+
+ if( *pIter==0x01 ){
+ pIter++;
+ pIter += fts3GetVarint32(pIter, &iThis);
+ }else{
+ iThis = 0;
+ }
+ while( iThisdoclist, and
+** * any Fts3MultiSegReader objects held by phrase tokens.
+*/
+void sqlite3Fts3EvalPhraseCleanup(Fts3Phrase *pPhrase){
+ if( pPhrase ){
+ int i;
+ sqlite3_free(pPhrase->doclist.aAll);
+ fts3EvalInvalidatePoslist(pPhrase);
+ memset(&pPhrase->doclist, 0, sizeof(Fts3Doclist));
+ for(i=0; inToken; i++){
+ fts3SegReaderCursorFree(pPhrase->aToken[i].pSegcsr);
+ pPhrase->aToken[i].pSegcsr = 0;
+ }
+ }
+}
+
+
+/*
+** Return SQLITE_CORRUPT_VTAB.
+*/
+#ifdef SQLITE_DEBUG
+int sqlite3Fts3Corrupt(){
+ return SQLITE_CORRUPT_VTAB;
+}
+#endif
+
+#if !defined(SQLITE_CORE)
+/*
+** Initialize API pointer table, if required.
+*/
+#ifdef _WIN32
+__declspec(dllexport)
+#endif
+int sqlite3_fts3_init(
+ sqlite3 *db,
+ char **pzErrMsg,
+ const sqlite3_api_routines *pApi
+){
+ SQLITE_EXTENSION_INIT2(pApi)
+ return sqlite3Fts3Init(db);
+}
+#endif
+
+#endif
diff --git a/ext/fts3/fts3.h b/ext/fts3/fts3.h
new file mode 100644
index 0000000000000000000000000000000000000000..c1aa8caf0927920d6bb78d578d9c111d4161a34a
--- /dev/null
+++ b/ext/fts3/fts3.h
@@ -0,0 +1,26 @@
+/*
+** 2006 Oct 10
+**
+** The author disclaims copyright to this source code. In place of
+** a legal notice, here is a blessing:
+**
+** May you do good and not evil.
+** May you find forgiveness for yourself and forgive others.
+** May you share freely, never taking more than you give.
+**
+******************************************************************************
+**
+** This header file is used by programs that want to link against the
+** FTS3 library. All it does is declare the sqlite3Fts3Init() interface.
+*/
+#include "sqlite3.h"
+
+#ifdef __cplusplus
+extern "C" {
+#endif /* __cplusplus */
+
+int sqlite3Fts3Init(sqlite3 *db);
+
+#ifdef __cplusplus
+} /* extern "C" */
+#endif /* __cplusplus */
diff --git a/ext/fts3/fts3Int.h b/ext/fts3/fts3Int.h
new file mode 100644
index 0000000000000000000000000000000000000000..3b236faf495e758417116d7c7fb1201e8d7d2fcb
--- /dev/null
+++ b/ext/fts3/fts3Int.h
@@ -0,0 +1,659 @@
+/*
+** 2009 Nov 12
+**
+** The author disclaims copyright to this source code. In place of
+** a legal notice, here is a blessing:
+**
+** May you do good and not evil.
+** May you find forgiveness for yourself and forgive others.
+** May you share freely, never taking more than you give.
+**
+******************************************************************************
+**
+*/
+#ifndef _FTSINT_H
+#define _FTSINT_H
+
+#if !defined(NDEBUG) && !defined(SQLITE_DEBUG)
+# define NDEBUG 1
+#endif
+
+/* FTS3/FTS4 require virtual tables */
+#ifdef SQLITE_OMIT_VIRTUALTABLE
+# undef SQLITE_ENABLE_FTS3
+# undef SQLITE_ENABLE_FTS4
+#endif
+
+/*
+** FTS4 is really an extension for FTS3. It is enabled using the
+** SQLITE_ENABLE_FTS3 macro. But to avoid confusion we also all
+** the SQLITE_ENABLE_FTS4 macro to serve as an alisse for SQLITE_ENABLE_FTS3.
+*/
+#if defined(SQLITE_ENABLE_FTS4) && !defined(SQLITE_ENABLE_FTS3)
+# define SQLITE_ENABLE_FTS3
+#endif
+
+#if !defined(SQLITE_CORE) || defined(SQLITE_ENABLE_FTS3)
+
+/* If not building as part of the core, include sqlite3ext.h. */
+#ifndef SQLITE_CORE
+# include "sqlite3ext.h"
+SQLITE_EXTENSION_INIT3
+#endif
+
+#include "sqlite3.h"
+#include "fts3_tokenizer.h"
+#include "fts3_hash.h"
+
+/*
+** This constant determines the maximum depth of an FTS expression tree
+** that the library will create and use. FTS uses recursion to perform
+** various operations on the query tree, so the disadvantage of a large
+** limit is that it may allow very large queries to use large amounts
+** of stack space (perhaps causing a stack overflow).
+*/
+#ifndef SQLITE_FTS3_MAX_EXPR_DEPTH
+# define SQLITE_FTS3_MAX_EXPR_DEPTH 12
+#endif
+
+
+/*
+** This constant controls how often segments are merged. Once there are
+** FTS3_MERGE_COUNT segments of level N, they are merged into a single
+** segment of level N+1.
+*/
+#define FTS3_MERGE_COUNT 16
+
+/*
+** This is the maximum amount of data (in bytes) to store in the
+** Fts3Table.pendingTerms hash table. Normally, the hash table is
+** populated as documents are inserted/updated/deleted in a transaction
+** and used to create a new segment when the transaction is committed.
+** However if this limit is reached midway through a transaction, a new
+** segment is created and the hash table cleared immediately.
+*/
+#define FTS3_MAX_PENDING_DATA (1*1024*1024)
+
+/*
+** Macro to return the number of elements in an array. SQLite has a
+** similar macro called ArraySize(). Use a different name to avoid
+** a collision when building an amalgamation with built-in FTS3.
+*/
+#define SizeofArray(X) ((int)(sizeof(X)/sizeof(X[0])))
+
+
+#ifndef MIN
+# define MIN(x,y) ((x)<(y)?(x):(y))
+#endif
+#ifndef MAX
+# define MAX(x,y) ((x)>(y)?(x):(y))
+#endif
+
+/*
+** Maximum length of a varint encoded integer. The varint format is different
+** from that used by SQLite, so the maximum length is 10, not 9.
+*/
+#define FTS3_VARINT_MAX 10
+
+#define FTS3_BUFFER_PADDING 8
+
+/*
+** FTS4 virtual tables may maintain multiple indexes - one index of all terms
+** in the document set and zero or more prefix indexes. All indexes are stored
+** as one or more b+-trees in the %_segments and %_segdir tables.
+**
+** It is possible to determine which index a b+-tree belongs to based on the
+** value stored in the "%_segdir.level" column. Given this value L, the index
+** that the b+-tree belongs to is (L<<10). In other words, all b+-trees with
+** level values between 0 and 1023 (inclusive) belong to index 0, all levels
+** between 1024 and 2047 to index 1, and so on.
+**
+** It is considered impossible for an index to use more than 1024 levels. In
+** theory though this may happen, but only after at least
+** (FTS3_MERGE_COUNT^1024) separate flushes of the pending-terms tables.
+*/
+#define FTS3_SEGDIR_MAXLEVEL 1024
+#define FTS3_SEGDIR_MAXLEVEL_STR "1024"
+
+/*
+** The testcase() macro is only used by the amalgamation. If undefined,
+** make it a no-op.
+*/
+#ifndef testcase
+# define testcase(X)
+#endif
+
+/*
+** Terminator values for position-lists and column-lists.
+*/
+#define POS_COLUMN (1) /* Column-list terminator */
+#define POS_END (0) /* Position-list terminator */
+
+/*
+** The assert_fts3_nc() macro is similar to the assert() macro, except that it
+** is used for assert() conditions that are true only if it can be
+** guranteed that the database is not corrupt.
+*/
+#ifdef SQLITE_DEBUG
+extern int sqlite3_fts3_may_be_corrupt;
+# define assert_fts3_nc(x) assert(sqlite3_fts3_may_be_corrupt || (x))
+#else
+# define assert_fts3_nc(x) assert(x)
+#endif
+
+/*
+** This section provides definitions to allow the
+** FTS3 extension to be compiled outside of the
+** amalgamation.
+*/
+#ifndef SQLITE_AMALGAMATION
+/*
+** Macros indicating that conditional expressions are always true or
+** false.
+*/
+#if defined(SQLITE_COVERAGE_TEST) || defined(SQLITE_MUTATION_TEST)
+# define SQLITE_OMIT_AUXILIARY_SAFETY_CHECKS 1
+#endif
+#if defined(SQLITE_OMIT_AUXILIARY_SAFETY_CHECKS)
+# define ALWAYS(X) (1)
+# define NEVER(X) (0)
+#elif !defined(NDEBUG)
+# define ALWAYS(X) ((X)?1:(assert(0),0))
+# define NEVER(X) ((X)?(assert(0),1):0)
+#else
+# define ALWAYS(X) (X)
+# define NEVER(X) (X)
+#endif
+
+/*
+** Internal types used by SQLite.
+*/
+typedef unsigned char u8; /* 1-byte (or larger) unsigned integer */
+typedef short int i16; /* 2-byte (or larger) signed integer */
+typedef unsigned int u32; /* 4-byte unsigned integer */
+typedef sqlite3_uint64 u64; /* 8-byte unsigned integer */
+typedef sqlite3_int64 i64; /* 8-byte signed integer */
+
+/*
+** Macro used to suppress compiler warnings for unused parameters.
+*/
+#define UNUSED_PARAMETER(x) (void)(x)
+
+/*
+** Activate assert() only if SQLITE_TEST is enabled.
+*/
+#if !defined(NDEBUG) && !defined(SQLITE_DEBUG)
+# define NDEBUG 1
+#endif
+
+/*
+** The TESTONLY macro is used to enclose variable declarations or
+** other bits of code that are needed to support the arguments
+** within testcase() and assert() macros.
+*/
+#if defined(SQLITE_DEBUG) || defined(SQLITE_COVERAGE_TEST)
+# define TESTONLY(X) X
+#else
+# define TESTONLY(X)
+#endif
+
+#define LARGEST_INT64 (0xffffffff|(((i64)0x7fffffff)<<32))
+#define SMALLEST_INT64 (((i64)-1) - LARGEST_INT64)
+
+#define deliberate_fall_through
+
+#endif /* SQLITE_AMALGAMATION */
+
+#ifdef SQLITE_DEBUG
+int sqlite3Fts3Corrupt(void);
+# define FTS_CORRUPT_VTAB sqlite3Fts3Corrupt()
+#else
+# define FTS_CORRUPT_VTAB SQLITE_CORRUPT_VTAB
+#endif
+
+typedef struct Fts3Table Fts3Table;
+typedef struct Fts3Cursor Fts3Cursor;
+typedef struct Fts3Expr Fts3Expr;
+typedef struct Fts3Phrase Fts3Phrase;
+typedef struct Fts3PhraseToken Fts3PhraseToken;
+
+typedef struct Fts3Doclist Fts3Doclist;
+typedef struct Fts3SegFilter Fts3SegFilter;
+typedef struct Fts3DeferredToken Fts3DeferredToken;
+typedef struct Fts3SegReader Fts3SegReader;
+typedef struct Fts3MultiSegReader Fts3MultiSegReader;
+
+typedef struct MatchinfoBuffer MatchinfoBuffer;
+
+/*
+** A connection to a fulltext index is an instance of the following
+** structure. The xCreate and xConnect methods create an instance
+** of this structure and xDestroy and xDisconnect free that instance.
+** All other methods receive a pointer to the structure as one of their
+** arguments.
+*/
+struct Fts3Table {
+ sqlite3_vtab base; /* Base class used by SQLite core */
+ sqlite3 *db; /* The database connection */
+ const char *zDb; /* logical database name */
+ const char *zName; /* virtual table name */
+ int nColumn; /* number of named columns in virtual table */
+ char **azColumn; /* column names. malloced */
+ u8 *abNotindexed; /* True for 'notindexed' columns */
+ sqlite3_tokenizer *pTokenizer; /* tokenizer for inserts and queries */
+ char *zContentTbl; /* content=xxx option, or NULL */
+ char *zLanguageid; /* languageid=xxx option, or NULL */
+ int nAutoincrmerge; /* Value configured by 'automerge' */
+ u32 nLeafAdd; /* Number of leaf blocks added this trans */
+ int bLock; /* Used to prevent recursive content= tbls */
+
+ /* Precompiled statements used by the implementation. Each of these
+ ** statements is run and reset within a single virtual table API call.
+ */
+ sqlite3_stmt *aStmt[40];
+ sqlite3_stmt *pSeekStmt; /* Cache for fts3CursorSeekStmt() */
+
+ char *zReadExprlist;
+ char *zWriteExprlist;
+
+ int nNodeSize; /* Soft limit for node size */
+ u8 bFts4; /* True for FTS4, false for FTS3 */
+ u8 bHasStat; /* True if %_stat table exists (2==unknown) */
+ u8 bHasDocsize; /* True if %_docsize table exists */
+ u8 bDescIdx; /* True if doclists are in reverse order */
+ u8 bIgnoreSavepoint; /* True to ignore xSavepoint invocations */
+ int nPgsz; /* Page size for host database */
+ char *zSegmentsTbl; /* Name of %_segments table */
+ sqlite3_blob *pSegments; /* Blob handle open on %_segments table */
+ int iSavepoint;
+
+ /*
+ ** The following array of hash tables is used to buffer pending index
+ ** updates during transactions. All pending updates buffered at any one
+ ** time must share a common language-id (see the FTS4 langid= feature).
+ ** The current language id is stored in variable iPrevLangid.
+ **
+ ** A single FTS4 table may have multiple full-text indexes. For each index
+ ** there is an entry in the aIndex[] array. Index 0 is an index of all the
+ ** terms that appear in the document set. Each subsequent index in aIndex[]
+ ** is an index of prefixes of a specific length.
+ **
+ ** Variable nPendingData contains an estimate the memory consumed by the
+ ** pending data structures, including hash table overhead, but not including
+ ** malloc overhead. When nPendingData exceeds nMaxPendingData, all hash
+ ** tables are flushed to disk. Variable iPrevDocid is the docid of the most
+ ** recently inserted record.
+ */
+ int nIndex; /* Size of aIndex[] */
+ struct Fts3Index {
+ int nPrefix; /* Prefix length (0 for main terms index) */
+ Fts3Hash hPending; /* Pending terms table for this index */
+ } *aIndex;
+ int nMaxPendingData; /* Max pending data before flush to disk */
+ int nPendingData; /* Current bytes of pending data */
+ sqlite_int64 iPrevDocid; /* Docid of most recently inserted document */
+ int iPrevLangid; /* Langid of recently inserted document */
+ int bPrevDelete; /* True if last operation was a delete */
+
+#if defined(SQLITE_DEBUG) || defined(SQLITE_COVERAGE_TEST)
+ /* State variables used for validating that the transaction control
+ ** methods of the virtual table are called at appropriate times. These
+ ** values do not contribute to FTS functionality; they are used for
+ ** verifying the operation of the SQLite core.
+ */
+ int inTransaction; /* True after xBegin but before xCommit/xRollback */
+ int mxSavepoint; /* Largest valid xSavepoint integer */
+#endif
+
+#if defined(SQLITE_DEBUG) || defined(SQLITE_TEST)
+ /* True to disable the incremental doclist optimization. This is controled
+ ** by special insert command 'test-no-incr-doclist'. */
+ int bNoIncrDoclist;
+
+ /* Number of segments in a level */
+ int nMergeCount;
+#endif
+};
+
+/* Macro to find the number of segments to merge */
+#if defined(SQLITE_DEBUG) || defined(SQLITE_TEST)
+# define MergeCount(P) ((P)->nMergeCount)
+#else
+# define MergeCount(P) FTS3_MERGE_COUNT
+#endif
+
+/*
+** When the core wants to read from the virtual table, it creates a
+** virtual table cursor (an instance of the following structure) using
+** the xOpen method. Cursors are destroyed using the xClose method.
+*/
+struct Fts3Cursor {
+ sqlite3_vtab_cursor base; /* Base class used by SQLite core */
+ i16 eSearch; /* Search strategy (see below) */
+ u8 isEof; /* True if at End Of Results */
+ u8 isRequireSeek; /* True if must seek pStmt to %_content row */
+ u8 bSeekStmt; /* True if pStmt is a seek */
+ sqlite3_stmt *pStmt; /* Prepared statement in use by the cursor */
+ Fts3Expr *pExpr; /* Parsed MATCH query string */
+ int iLangid; /* Language being queried for */
+ int nPhrase; /* Number of matchable phrases in query */
+ Fts3DeferredToken *pDeferred; /* Deferred search tokens, if any */
+ sqlite3_int64 iPrevId; /* Previous id read from aDoclist */
+ char *pNextId; /* Pointer into the body of aDoclist */
+ char *aDoclist; /* List of docids for full-text queries */
+ int nDoclist; /* Size of buffer at aDoclist */
+ u8 bDesc; /* True to sort in descending order */
+ int eEvalmode; /* An FTS3_EVAL_XX constant */
+ int nRowAvg; /* Average size of database rows, in pages */
+ sqlite3_int64 nDoc; /* Documents in table */
+ i64 iMinDocid; /* Minimum docid to return */
+ i64 iMaxDocid; /* Maximum docid to return */
+ int isMatchinfoNeeded; /* True when aMatchinfo[] needs filling in */
+ MatchinfoBuffer *pMIBuffer; /* Buffer for matchinfo data */
+};
+
+#define FTS3_EVAL_FILTER 0
+#define FTS3_EVAL_NEXT 1
+#define FTS3_EVAL_MATCHINFO 2
+
+/*
+** The Fts3Cursor.eSearch member is always set to one of the following.
+** Actualy, Fts3Cursor.eSearch can be greater than or equal to
+** FTS3_FULLTEXT_SEARCH. If so, then Fts3Cursor.eSearch - 2 is the index
+** of the column to be searched. For example, in
+**
+** CREATE VIRTUAL TABLE ex1 USING fts3(a,b,c,d);
+** SELECT docid FROM ex1 WHERE b MATCH 'one two three';
+**
+** Because the LHS of the MATCH operator is 2nd column "b",
+** Fts3Cursor.eSearch will be set to FTS3_FULLTEXT_SEARCH+1. (+0 for a,
+** +1 for b, +2 for c, +3 for d.) If the LHS of MATCH were "ex1"
+** indicating that all columns should be searched,
+** then eSearch would be set to FTS3_FULLTEXT_SEARCH+4.
+*/
+#define FTS3_FULLSCAN_SEARCH 0 /* Linear scan of %_content table */
+#define FTS3_DOCID_SEARCH 1 /* Lookup by rowid on %_content table */
+#define FTS3_FULLTEXT_SEARCH 2 /* Full-text index search */
+
+/*
+** The lower 16-bits of the sqlite3_index_info.idxNum value set by
+** the xBestIndex() method contains the Fts3Cursor.eSearch value described
+** above. The upper 16-bits contain a combination of the following
+** bits, used to describe extra constraints on full-text searches.
+*/
+#define FTS3_HAVE_LANGID 0x00010000 /* languageid=? */
+#define FTS3_HAVE_DOCID_GE 0x00020000 /* docid>=? */
+#define FTS3_HAVE_DOCID_LE 0x00040000 /* docid<=? */
+
+struct Fts3Doclist {
+ char *aAll; /* Array containing doclist (or NULL) */
+ int nAll; /* Size of a[] in bytes */
+ char *pNextDocid; /* Pointer to next docid */
+
+ sqlite3_int64 iDocid; /* Current docid (if pList!=0) */
+ int bFreeList; /* True if pList should be sqlite3_free()d */
+ char *pList; /* Pointer to position list following iDocid */
+ int nList; /* Length of position list */
+};
+
+/*
+** A "phrase" is a sequence of one or more tokens that must match in
+** sequence. A single token is the base case and the most common case.
+** For a sequence of tokens contained in double-quotes (i.e. "one two three")
+** nToken will be the number of tokens in the string.
+*/
+struct Fts3PhraseToken {
+ char *z; /* Text of the token */
+ int n; /* Number of bytes in buffer z */
+ int isPrefix; /* True if token ends with a "*" character */
+ int bFirst; /* True if token must appear at position 0 */
+
+ /* Variables above this point are populated when the expression is
+ ** parsed (by code in fts3_expr.c). Below this point the variables are
+ ** used when evaluating the expression. */
+ Fts3DeferredToken *pDeferred; /* Deferred token object for this token */
+ Fts3MultiSegReader *pSegcsr; /* Segment-reader for this token */
+};
+
+struct Fts3Phrase {
+ /* Cache of doclist for this phrase. */
+ Fts3Doclist doclist;
+ int bIncr; /* True if doclist is loaded incrementally */
+ int iDoclistToken;
+
+ /* Used by sqlite3Fts3EvalPhrasePoslist() if this is a descendent of an
+ ** OR condition. */
+ char *pOrPoslist;
+ i64 iOrDocid;
+
+ /* Variables below this point are populated by fts3_expr.c when parsing
+ ** a MATCH expression. Everything above is part of the evaluation phase.
+ */
+ int nToken; /* Number of tokens in the phrase */
+ int iColumn; /* Index of column this phrase must match */
+ Fts3PhraseToken aToken[1]; /* One entry for each token in the phrase */
+};
+
+/*
+** A tree of these objects forms the RHS of a MATCH operator.
+**
+** If Fts3Expr.eType is FTSQUERY_PHRASE and isLoaded is true, then aDoclist
+** points to a malloced buffer, size nDoclist bytes, containing the results
+** of this phrase query in FTS3 doclist format. As usual, the initial
+** "Length" field found in doclists stored on disk is omitted from this
+** buffer.
+**
+** Variable aMI is used only for FTSQUERY_NEAR nodes to store the global
+** matchinfo data. If it is not NULL, it points to an array of size nCol*3,
+** where nCol is the number of columns in the queried FTS table. The array
+** is populated as follows:
+**
+** aMI[iCol*3 + 0] = Undefined
+** aMI[iCol*3 + 1] = Number of occurrences
+** aMI[iCol*3 + 2] = Number of rows containing at least one instance
+**
+** The aMI array is allocated using sqlite3_malloc(). It should be freed
+** when the expression node is.
+*/
+struct Fts3Expr {
+ int eType; /* One of the FTSQUERY_XXX values defined below */
+ int nNear; /* Valid if eType==FTSQUERY_NEAR */
+ Fts3Expr *pParent; /* pParent->pLeft==this or pParent->pRight==this */
+ Fts3Expr *pLeft; /* Left operand */
+ Fts3Expr *pRight; /* Right operand */
+ Fts3Phrase *pPhrase; /* Valid if eType==FTSQUERY_PHRASE */
+
+ /* The following are used by the fts3_eval.c module. */
+ sqlite3_int64 iDocid; /* Current docid */
+ u8 bEof; /* True this expression is at EOF already */
+ u8 bStart; /* True if iDocid is valid */
+ u8 bDeferred; /* True if this expression is entirely deferred */
+
+ /* The following are used by the fts3_snippet.c module. */
+ int iPhrase; /* Index of this phrase in matchinfo() results */
+ u32 *aMI; /* See above */
+};
+
+/*
+** Candidate values for Fts3Query.eType. Note that the order of the first
+** four values is in order of precedence when parsing expressions. For
+** example, the following:
+**
+** "a OR b AND c NOT d NEAR e"
+**
+** is equivalent to:
+**
+** "a OR (b AND (c NOT (d NEAR e)))"
+*/
+#define FTSQUERY_NEAR 1
+#define FTSQUERY_NOT 2
+#define FTSQUERY_AND 3
+#define FTSQUERY_OR 4
+#define FTSQUERY_PHRASE 5
+
+
+/* fts3_write.c */
+int sqlite3Fts3UpdateMethod(sqlite3_vtab*,int,sqlite3_value**,sqlite3_int64*);
+int sqlite3Fts3PendingTermsFlush(Fts3Table *);
+void sqlite3Fts3PendingTermsClear(Fts3Table *);
+int sqlite3Fts3Optimize(Fts3Table *);
+int sqlite3Fts3SegReaderNew(int, int, sqlite3_int64,
+ sqlite3_int64, sqlite3_int64, const char *, int, Fts3SegReader**);
+int sqlite3Fts3SegReaderPending(
+ Fts3Table*,int,const char*,int,int,Fts3SegReader**);
+void sqlite3Fts3SegReaderFree(Fts3SegReader *);
+int sqlite3Fts3AllSegdirs(Fts3Table*, int, int, int, sqlite3_stmt **);
+int sqlite3Fts3ReadBlock(Fts3Table*, sqlite3_int64, char **, int*, int*);
+
+int sqlite3Fts3SelectDoctotal(Fts3Table *, sqlite3_stmt **);
+int sqlite3Fts3SelectDocsize(Fts3Table *, sqlite3_int64, sqlite3_stmt **);
+
+#ifndef SQLITE_DISABLE_FTS4_DEFERRED
+void sqlite3Fts3FreeDeferredTokens(Fts3Cursor *);
+int sqlite3Fts3DeferToken(Fts3Cursor *, Fts3PhraseToken *, int);
+int sqlite3Fts3CacheDeferredDoclists(Fts3Cursor *);
+void sqlite3Fts3FreeDeferredDoclists(Fts3Cursor *);
+int sqlite3Fts3DeferredTokenList(Fts3DeferredToken *, char **, int *);
+#else
+# define sqlite3Fts3FreeDeferredTokens(x)
+# define sqlite3Fts3DeferToken(x,y,z) SQLITE_OK
+# define sqlite3Fts3CacheDeferredDoclists(x) SQLITE_OK
+# define sqlite3Fts3FreeDeferredDoclists(x)
+# define sqlite3Fts3DeferredTokenList(x,y,z) SQLITE_OK
+#endif
+
+void sqlite3Fts3SegmentsClose(Fts3Table *);
+int sqlite3Fts3MaxLevel(Fts3Table *, int *);
+
+/* Special values interpreted by sqlite3SegReaderCursor() */
+#define FTS3_SEGCURSOR_PENDING -1
+#define FTS3_SEGCURSOR_ALL -2
+
+int sqlite3Fts3SegReaderStart(Fts3Table*, Fts3MultiSegReader*, Fts3SegFilter*);
+int sqlite3Fts3SegReaderStep(Fts3Table *, Fts3MultiSegReader *);
+void sqlite3Fts3SegReaderFinish(Fts3MultiSegReader *);
+
+int sqlite3Fts3SegReaderCursor(Fts3Table *,
+ int, int, int, const char *, int, int, int, Fts3MultiSegReader *);
+
+/* Flags allowed as part of the 4th argument to SegmentReaderIterate() */
+#define FTS3_SEGMENT_REQUIRE_POS 0x00000001
+#define FTS3_SEGMENT_IGNORE_EMPTY 0x00000002
+#define FTS3_SEGMENT_COLUMN_FILTER 0x00000004
+#define FTS3_SEGMENT_PREFIX 0x00000008
+#define FTS3_SEGMENT_SCAN 0x00000010
+#define FTS3_SEGMENT_FIRST 0x00000020
+
+/* Type passed as 4th argument to SegmentReaderIterate() */
+struct Fts3SegFilter {
+ const char *zTerm;
+ int nTerm;
+ int iCol;
+ int flags;
+};
+
+struct Fts3MultiSegReader {
+ /* Used internally by sqlite3Fts3SegReaderXXX() calls */
+ Fts3SegReader **apSegment; /* Array of Fts3SegReader objects */
+ int nSegment; /* Size of apSegment array */
+ int nAdvance; /* How many seg-readers to advance */
+ Fts3SegFilter *pFilter; /* Pointer to filter object */
+ char *aBuffer; /* Buffer to merge doclists in */
+ i64 nBuffer; /* Allocated size of aBuffer[] in bytes */
+
+ int iColFilter; /* If >=0, filter for this column */
+ int bRestart;
+
+ /* Used by fts3.c only. */
+ int nCost; /* Cost of running iterator */
+ int bLookup; /* True if a lookup of a single entry. */
+
+ /* Output values. Valid only after Fts3SegReaderStep() returns SQLITE_ROW. */
+ char *zTerm; /* Pointer to term buffer */
+ int nTerm; /* Size of zTerm in bytes */
+ char *aDoclist; /* Pointer to doclist buffer */
+ int nDoclist; /* Size of aDoclist[] in bytes */
+};
+
+int sqlite3Fts3Incrmerge(Fts3Table*,int,int);
+
+#define fts3GetVarint32(p, piVal) ( \
+ (*(u8*)(p)&0x80) ? sqlite3Fts3GetVarint32(p, piVal) : (*piVal=*(u8*)(p), 1) \
+)
+
+/* fts3.c */
+void sqlite3Fts3ErrMsg(char**,const char*,...);
+int sqlite3Fts3PutVarint(char *, sqlite3_int64);
+int sqlite3Fts3GetVarint(const char *, sqlite_int64 *);
+int sqlite3Fts3GetVarintU(const char *, sqlite_uint64 *);
+int sqlite3Fts3GetVarintBounded(const char*,const char*,sqlite3_int64*);
+int sqlite3Fts3GetVarint32(const char *, int *);
+int sqlite3Fts3VarintLen(sqlite3_uint64);
+void sqlite3Fts3Dequote(char *);
+void sqlite3Fts3DoclistPrev(int,char*,int,char**,sqlite3_int64*,int*,u8*);
+int sqlite3Fts3EvalPhraseStats(Fts3Cursor *, Fts3Expr *, u32 *);
+int sqlite3Fts3FirstFilter(sqlite3_int64, char *, int, char *);
+void sqlite3Fts3CreateStatTable(int*, Fts3Table*);
+int sqlite3Fts3EvalTestDeferred(Fts3Cursor *pCsr, int *pRc);
+int sqlite3Fts3ReadInt(const char *z, int *pnOut);
+
+/* fts3_tokenizer.c */
+const char *sqlite3Fts3NextToken(const char *, int *);
+int sqlite3Fts3InitHashTable(sqlite3 *, Fts3Hash *, const char *);
+int sqlite3Fts3InitTokenizer(Fts3Hash *pHash, const char *,
+ sqlite3_tokenizer **, char **
+);
+int sqlite3Fts3IsIdChar(char);
+
+/* fts3_snippet.c */
+void sqlite3Fts3Offsets(sqlite3_context*, Fts3Cursor*);
+void sqlite3Fts3Snippet(sqlite3_context *, Fts3Cursor *, const char *,
+ const char *, const char *, int, int
+);
+void sqlite3Fts3Matchinfo(sqlite3_context *, Fts3Cursor *, const char *);
+void sqlite3Fts3MIBufferFree(MatchinfoBuffer *p);
+
+/* fts3_expr.c */
+int sqlite3Fts3ExprParse(sqlite3_tokenizer *, int,
+ char **, int, int, int, const char *, int, Fts3Expr **, char **
+);
+void sqlite3Fts3ExprFree(Fts3Expr *);
+#ifdef SQLITE_TEST
+int sqlite3Fts3ExprInitTestInterface(sqlite3 *db, Fts3Hash*);
+int sqlite3Fts3InitTerm(sqlite3 *db);
+#endif
+void *sqlite3Fts3MallocZero(i64 nByte);
+
+int sqlite3Fts3OpenTokenizer(sqlite3_tokenizer *, int, const char *, int,
+ sqlite3_tokenizer_cursor **
+);
+
+/* fts3_aux.c */
+int sqlite3Fts3InitAux(sqlite3 *db);
+
+void sqlite3Fts3EvalPhraseCleanup(Fts3Phrase *);
+
+int sqlite3Fts3MsrIncrStart(
+ Fts3Table*, Fts3MultiSegReader*, int, const char*, int);
+int sqlite3Fts3MsrIncrNext(
+ Fts3Table *, Fts3MultiSegReader *, sqlite3_int64 *, char **, int *);
+int sqlite3Fts3EvalPhrasePoslist(Fts3Cursor *, Fts3Expr *, int iCol, char **);
+int sqlite3Fts3MsrOvfl(Fts3Cursor *, Fts3MultiSegReader *, int *);
+int sqlite3Fts3MsrIncrRestart(Fts3MultiSegReader *pCsr);
+
+/* fts3_tokenize_vtab.c */
+int sqlite3Fts3InitTok(sqlite3*, Fts3Hash *, void(*xDestroy)(void*));
+
+/* fts3_unicode2.c (functions generated by parsing unicode text files) */
+#ifndef SQLITE_DISABLE_FTS3_UNICODE
+int sqlite3FtsUnicodeFold(int, int);
+int sqlite3FtsUnicodeIsalnum(int);
+int sqlite3FtsUnicodeIsdiacritic(int);
+#endif
+
+int sqlite3Fts3ExprIterate(Fts3Expr*, int (*x)(Fts3Expr*,int,void*), void*);
+
+int sqlite3Fts3IntegrityCheck(Fts3Table *p, int *pbOk);
+
+#endif /* !SQLITE_CORE || SQLITE_ENABLE_FTS3 */
+#endif /* _FTSINT_H */
diff --git a/ext/fts3/fts3_aux.c b/ext/fts3/fts3_aux.c
new file mode 100644
index 0000000000000000000000000000000000000000..439d57936620734f2e741738ed26968dfff09ba3
--- /dev/null
+++ b/ext/fts3/fts3_aux.c
@@ -0,0 +1,557 @@
+/*
+** 2011 Jan 27
+**
+** The author disclaims copyright to this source code. In place of
+** a legal notice, here is a blessing:
+**
+** May you do good and not evil.
+** May you find forgiveness for yourself and forgive others.
+** May you share freely, never taking more than you give.
+**
+******************************************************************************
+**
+*/
+#include "fts3Int.h"
+#if !defined(SQLITE_CORE) || defined(SQLITE_ENABLE_FTS3)
+
+#include
+#include
+
+typedef struct Fts3auxTable Fts3auxTable;
+typedef struct Fts3auxCursor Fts3auxCursor;
+
+struct Fts3auxTable {
+ sqlite3_vtab base; /* Base class used by SQLite core */
+ Fts3Table *pFts3Tab;
+};
+
+struct Fts3auxCursor {
+ sqlite3_vtab_cursor base; /* Base class used by SQLite core */
+ Fts3MultiSegReader csr; /* Must be right after "base" */
+ Fts3SegFilter filter;
+ char *zStop;
+ int nStop; /* Byte-length of string zStop */
+ int iLangid; /* Language id to query */
+ int isEof; /* True if cursor is at EOF */
+ sqlite3_int64 iRowid; /* Current rowid */
+
+ int iCol; /* Current value of 'col' column */
+ int nStat; /* Size of aStat[] array */
+ struct Fts3auxColstats {
+ sqlite3_int64 nDoc; /* 'documents' values for current csr row */
+ sqlite3_int64 nOcc; /* 'occurrences' values for current csr row */
+ } *aStat;
+};
+
+/*
+** Schema of the terms table.
+*/
+#define FTS3_AUX_SCHEMA \
+ "CREATE TABLE x(term, col, documents, occurrences, languageid HIDDEN)"
+
+/*
+** This function does all the work for both the xConnect and xCreate methods.
+** These tables have no persistent representation of their own, so xConnect
+** and xCreate are identical operations.
+*/
+static int fts3auxConnectMethod(
+ sqlite3 *db, /* Database connection */
+ void *pUnused, /* Unused */
+ int argc, /* Number of elements in argv array */
+ const char * const *argv, /* xCreate/xConnect argument array */
+ sqlite3_vtab **ppVtab, /* OUT: New sqlite3_vtab object */
+ char **pzErr /* OUT: sqlite3_malloc'd error message */
+){
+ char const *zDb; /* Name of database (e.g. "main") */
+ char const *zFts3; /* Name of fts3 table */
+ int nDb; /* Result of strlen(zDb) */
+ int nFts3; /* Result of strlen(zFts3) */
+ sqlite3_int64 nByte; /* Bytes of space to allocate here */
+ int rc; /* value returned by declare_vtab() */
+ Fts3auxTable *p; /* Virtual table object to return */
+
+ UNUSED_PARAMETER(pUnused);
+
+ /* The user should invoke this in one of two forms:
+ **
+ ** CREATE VIRTUAL TABLE xxx USING fts4aux(fts4-table);
+ ** CREATE VIRTUAL TABLE xxx USING fts4aux(fts4-table-db, fts4-table);
+ */
+ if( argc!=4 && argc!=5 ) goto bad_args;
+
+ zDb = argv[1];
+ nDb = (int)strlen(zDb);
+ if( argc==5 ){
+ if( nDb==4 && 0==sqlite3_strnicmp("temp", zDb, 4) ){
+ zDb = argv[3];
+ nDb = (int)strlen(zDb);
+ zFts3 = argv[4];
+ }else{
+ goto bad_args;
+ }
+ }else{
+ zFts3 = argv[3];
+ }
+ nFts3 = (int)strlen(zFts3);
+
+ rc = sqlite3_declare_vtab(db, FTS3_AUX_SCHEMA);
+ if( rc!=SQLITE_OK ) return rc;
+
+ nByte = sizeof(Fts3auxTable) + sizeof(Fts3Table) + nDb + nFts3 + 2;
+ p = (Fts3auxTable *)sqlite3_malloc64(nByte);
+ if( !p ) return SQLITE_NOMEM;
+ memset(p, 0, nByte);
+
+ p->pFts3Tab = (Fts3Table *)&p[1];
+ p->pFts3Tab->zDb = (char *)&p->pFts3Tab[1];
+ p->pFts3Tab->zName = &p->pFts3Tab->zDb[nDb+1];
+ p->pFts3Tab->db = db;
+ p->pFts3Tab->nIndex = 1;
+
+ memcpy((char *)p->pFts3Tab->zDb, zDb, nDb);
+ memcpy((char *)p->pFts3Tab->zName, zFts3, nFts3);
+ sqlite3Fts3Dequote((char *)p->pFts3Tab->zName);
+
+ *ppVtab = (sqlite3_vtab *)p;
+ return SQLITE_OK;
+
+ bad_args:
+ sqlite3Fts3ErrMsg(pzErr, "invalid arguments to fts4aux constructor");
+ return SQLITE_ERROR;
+}
+
+/*
+** This function does the work for both the xDisconnect and xDestroy methods.
+** These tables have no persistent representation of their own, so xDisconnect
+** and xDestroy are identical operations.
+*/
+static int fts3auxDisconnectMethod(sqlite3_vtab *pVtab){
+ Fts3auxTable *p = (Fts3auxTable *)pVtab;
+ Fts3Table *pFts3 = p->pFts3Tab;
+ int i;
+
+ /* Free any prepared statements held */
+ for(i=0; iaStmt); i++){
+ sqlite3_finalize(pFts3->aStmt[i]);
+ }
+ sqlite3_free(pFts3->zSegmentsTbl);
+ sqlite3_free(p);
+ return SQLITE_OK;
+}
+
+#define FTS4AUX_EQ_CONSTRAINT 1
+#define FTS4AUX_GE_CONSTRAINT 2
+#define FTS4AUX_LE_CONSTRAINT 4
+
+/*
+** xBestIndex - Analyze a WHERE and ORDER BY clause.
+*/
+static int fts3auxBestIndexMethod(
+ sqlite3_vtab *pVTab,
+ sqlite3_index_info *pInfo
+){
+ int i;
+ int iEq = -1;
+ int iGe = -1;
+ int iLe = -1;
+ int iLangid = -1;
+ int iNext = 1; /* Next free argvIndex value */
+
+ UNUSED_PARAMETER(pVTab);
+
+ /* This vtab delivers always results in "ORDER BY term ASC" order. */
+ if( pInfo->nOrderBy==1
+ && pInfo->aOrderBy[0].iColumn==0
+ && pInfo->aOrderBy[0].desc==0
+ ){
+ pInfo->orderByConsumed = 1;
+ }
+
+ /* Search for equality and range constraints on the "term" column.
+ ** And equality constraints on the hidden "languageid" column. */
+ for(i=0; inConstraint; i++){
+ if( pInfo->aConstraint[i].usable ){
+ int op = pInfo->aConstraint[i].op;
+ int iCol = pInfo->aConstraint[i].iColumn;
+
+ if( iCol==0 ){
+ if( op==SQLITE_INDEX_CONSTRAINT_EQ ) iEq = i;
+ if( op==SQLITE_INDEX_CONSTRAINT_LT ) iLe = i;
+ if( op==SQLITE_INDEX_CONSTRAINT_LE ) iLe = i;
+ if( op==SQLITE_INDEX_CONSTRAINT_GT ) iGe = i;
+ if( op==SQLITE_INDEX_CONSTRAINT_GE ) iGe = i;
+ }
+ if( iCol==4 ){
+ if( op==SQLITE_INDEX_CONSTRAINT_EQ ) iLangid = i;
+ }
+ }
+ }
+
+ if( iEq>=0 ){
+ pInfo->idxNum = FTS4AUX_EQ_CONSTRAINT;
+ pInfo->aConstraintUsage[iEq].argvIndex = iNext++;
+ pInfo->estimatedCost = 5;
+ }else{
+ pInfo->idxNum = 0;
+ pInfo->estimatedCost = 20000;
+ if( iGe>=0 ){
+ pInfo->idxNum += FTS4AUX_GE_CONSTRAINT;
+ pInfo->aConstraintUsage[iGe].argvIndex = iNext++;
+ pInfo->estimatedCost /= 2;
+ }
+ if( iLe>=0 ){
+ pInfo->idxNum += FTS4AUX_LE_CONSTRAINT;
+ pInfo->aConstraintUsage[iLe].argvIndex = iNext++;
+ pInfo->estimatedCost /= 2;
+ }
+ }
+ if( iLangid>=0 ){
+ pInfo->aConstraintUsage[iLangid].argvIndex = iNext++;
+ pInfo->estimatedCost--;
+ }
+
+ return SQLITE_OK;
+}
+
+/*
+** xOpen - Open a cursor.
+*/
+static int fts3auxOpenMethod(sqlite3_vtab *pVTab, sqlite3_vtab_cursor **ppCsr){
+ Fts3auxCursor *pCsr; /* Pointer to cursor object to return */
+
+ UNUSED_PARAMETER(pVTab);
+
+ pCsr = (Fts3auxCursor *)sqlite3_malloc(sizeof(Fts3auxCursor));
+ if( !pCsr ) return SQLITE_NOMEM;
+ memset(pCsr, 0, sizeof(Fts3auxCursor));
+
+ *ppCsr = (sqlite3_vtab_cursor *)pCsr;
+ return SQLITE_OK;
+}
+
+/*
+** xClose - Close a cursor.
+*/
+static int fts3auxCloseMethod(sqlite3_vtab_cursor *pCursor){
+ Fts3Table *pFts3 = ((Fts3auxTable *)pCursor->pVtab)->pFts3Tab;
+ Fts3auxCursor *pCsr = (Fts3auxCursor *)pCursor;
+
+ sqlite3Fts3SegmentsClose(pFts3);
+ sqlite3Fts3SegReaderFinish(&pCsr->csr);
+ sqlite3_free((void *)pCsr->filter.zTerm);
+ sqlite3_free(pCsr->zStop);
+ sqlite3_free(pCsr->aStat);
+ sqlite3_free(pCsr);
+ return SQLITE_OK;
+}
+
+static int fts3auxGrowStatArray(Fts3auxCursor *pCsr, int nSize){
+ if( nSize>pCsr->nStat ){
+ struct Fts3auxColstats *aNew;
+ aNew = (struct Fts3auxColstats *)sqlite3_realloc64(pCsr->aStat,
+ sizeof(struct Fts3auxColstats) * nSize
+ );
+ if( aNew==0 ) return SQLITE_NOMEM;
+ memset(&aNew[pCsr->nStat], 0,
+ sizeof(struct Fts3auxColstats) * (nSize - pCsr->nStat)
+ );
+ pCsr->aStat = aNew;
+ pCsr->nStat = nSize;
+ }
+ return SQLITE_OK;
+}
+
+/*
+** xNext - Advance the cursor to the next row, if any.
+*/
+static int fts3auxNextMethod(sqlite3_vtab_cursor *pCursor){
+ Fts3auxCursor *pCsr = (Fts3auxCursor *)pCursor;
+ Fts3Table *pFts3 = ((Fts3auxTable *)pCursor->pVtab)->pFts3Tab;
+ int rc;
+
+ /* Increment our pretend rowid value. */
+ pCsr->iRowid++;
+
+ for(pCsr->iCol++; pCsr->iColnStat; pCsr->iCol++){
+ if( pCsr->aStat[pCsr->iCol].nDoc>0 ) return SQLITE_OK;
+ }
+
+ rc = sqlite3Fts3SegReaderStep(pFts3, &pCsr->csr);
+ if( rc==SQLITE_ROW ){
+ int i = 0;
+ int nDoclist = pCsr->csr.nDoclist;
+ char *aDoclist = pCsr->csr.aDoclist;
+ int iCol;
+
+ int eState = 0;
+
+ if( pCsr->zStop ){
+ int n = (pCsr->nStopcsr.nTerm) ? pCsr->nStop : pCsr->csr.nTerm;
+ int mc = memcmp(pCsr->zStop, pCsr->csr.zTerm, n);
+ if( mc<0 || (mc==0 && pCsr->csr.nTerm>pCsr->nStop) ){
+ pCsr->isEof = 1;
+ return SQLITE_OK;
+ }
+ }
+
+ if( fts3auxGrowStatArray(pCsr, 2) ) return SQLITE_NOMEM;
+ memset(pCsr->aStat, 0, sizeof(struct Fts3auxColstats) * pCsr->nStat);
+ iCol = 0;
+ rc = SQLITE_OK;
+
+ while( iaStat[0].nDoc++;
+ eState = 1;
+ iCol = 0;
+ break;
+
+ /* State 1. In this state we are expecting either a 1, indicating
+ ** that the following integer will be a column number, or the
+ ** start of a position list for column 0.
+ **
+ ** The only difference between state 1 and state 2 is that if the
+ ** integer encountered in state 1 is not 0 or 1, then we need to
+ ** increment the column 0 "nDoc" count for this term.
+ */
+ case 1:
+ assert( iCol==0 );
+ if( v>1 ){
+ pCsr->aStat[1].nDoc++;
+ }
+ eState = 2;
+ /* fall through */
+
+ case 2:
+ if( v==0 ){ /* 0x00. Next integer will be a docid. */
+ eState = 0;
+ }else if( v==1 ){ /* 0x01. Next integer will be a column number. */
+ eState = 3;
+ }else{ /* 2 or greater. A position. */
+ pCsr->aStat[iCol+1].nOcc++;
+ pCsr->aStat[0].nOcc++;
+ }
+ break;
+
+ /* State 3. The integer just read is a column number. */
+ default: assert( eState==3 );
+ iCol = (int)v;
+ if( iCol<1 ){
+ rc = SQLITE_CORRUPT_VTAB;
+ break;
+ }
+ if( fts3auxGrowStatArray(pCsr, iCol+2) ) return SQLITE_NOMEM;
+ pCsr->aStat[iCol+1].nDoc++;
+ eState = 2;
+ break;
+ }
+ }
+
+ pCsr->iCol = 0;
+ }else{
+ pCsr->isEof = 1;
+ }
+ return rc;
+}
+
+/*
+** xFilter - Initialize a cursor to point at the start of its data.
+*/
+static int fts3auxFilterMethod(
+ sqlite3_vtab_cursor *pCursor, /* The cursor used for this query */
+ int idxNum, /* Strategy index */
+ const char *idxStr, /* Unused */
+ int nVal, /* Number of elements in apVal */
+ sqlite3_value **apVal /* Arguments for the indexing scheme */
+){
+ Fts3auxCursor *pCsr = (Fts3auxCursor *)pCursor;
+ Fts3Table *pFts3 = ((Fts3auxTable *)pCursor->pVtab)->pFts3Tab;
+ int rc;
+ int isScan = 0;
+ int iLangVal = 0; /* Language id to query */
+
+ int iEq = -1; /* Index of term=? value in apVal */
+ int iGe = -1; /* Index of term>=? value in apVal */
+ int iLe = -1; /* Index of term<=? value in apVal */
+ int iLangid = -1; /* Index of languageid=? value in apVal */
+ int iNext = 0;
+
+ UNUSED_PARAMETER(nVal);
+ UNUSED_PARAMETER(idxStr);
+
+ assert( idxStr==0 );
+ assert( idxNum==FTS4AUX_EQ_CONSTRAINT || idxNum==0
+ || idxNum==FTS4AUX_LE_CONSTRAINT || idxNum==FTS4AUX_GE_CONSTRAINT
+ || idxNum==(FTS4AUX_LE_CONSTRAINT|FTS4AUX_GE_CONSTRAINT)
+ );
+
+ if( idxNum==FTS4AUX_EQ_CONSTRAINT ){
+ iEq = iNext++;
+ }else{
+ isScan = 1;
+ if( idxNum & FTS4AUX_GE_CONSTRAINT ){
+ iGe = iNext++;
+ }
+ if( idxNum & FTS4AUX_LE_CONSTRAINT ){
+ iLe = iNext++;
+ }
+ }
+ if( iNextfilter.zTerm);
+ sqlite3Fts3SegReaderFinish(&pCsr->csr);
+ sqlite3_free((void *)pCsr->filter.zTerm);
+ sqlite3_free(pCsr->aStat);
+ sqlite3_free(pCsr->zStop);
+ memset(&pCsr->csr, 0, ((u8*)&pCsr[1]) - (u8*)&pCsr->csr);
+
+ pCsr->filter.flags = FTS3_SEGMENT_REQUIRE_POS|FTS3_SEGMENT_IGNORE_EMPTY;
+ if( isScan ) pCsr->filter.flags |= FTS3_SEGMENT_SCAN;
+
+ if( iEq>=0 || iGe>=0 ){
+ const unsigned char *zStr = sqlite3_value_text(apVal[0]);
+ assert( (iEq==0 && iGe==-1) || (iEq==-1 && iGe==0) );
+ if( zStr ){
+ pCsr->filter.zTerm = sqlite3_mprintf("%s", zStr);
+ if( pCsr->filter.zTerm==0 ) return SQLITE_NOMEM;
+ pCsr->filter.nTerm = (int)strlen(pCsr->filter.zTerm);
+ }
+ }
+
+ if( iLe>=0 ){
+ pCsr->zStop = sqlite3_mprintf("%s", sqlite3_value_text(apVal[iLe]));
+ if( pCsr->zStop==0 ) return SQLITE_NOMEM;
+ pCsr->nStop = (int)strlen(pCsr->zStop);
+ }
+
+ if( iLangid>=0 ){
+ iLangVal = sqlite3_value_int(apVal[iLangid]);
+
+ /* If the user specified a negative value for the languageid, use zero
+ ** instead. This works, as the "languageid=?" constraint will also
+ ** be tested by the VDBE layer. The test will always be false (since
+ ** this module will not return a row with a negative languageid), and
+ ** so the overall query will return zero rows. */
+ if( iLangVal<0 ) iLangVal = 0;
+ }
+ pCsr->iLangid = iLangVal;
+
+ rc = sqlite3Fts3SegReaderCursor(pFts3, iLangVal, 0, FTS3_SEGCURSOR_ALL,
+ pCsr->filter.zTerm, pCsr->filter.nTerm, 0, isScan, &pCsr->csr
+ );
+ if( rc==SQLITE_OK ){
+ rc = sqlite3Fts3SegReaderStart(pFts3, &pCsr->csr, &pCsr->filter);
+ }
+
+ if( rc==SQLITE_OK ) rc = fts3auxNextMethod(pCursor);
+ return rc;
+}
+
+/*
+** xEof - Return true if the cursor is at EOF, or false otherwise.
+*/
+static int fts3auxEofMethod(sqlite3_vtab_cursor *pCursor){
+ Fts3auxCursor *pCsr = (Fts3auxCursor *)pCursor;
+ return pCsr->isEof;
+}
+
+/*
+** xColumn - Return a column value.
+*/
+static int fts3auxColumnMethod(
+ sqlite3_vtab_cursor *pCursor, /* Cursor to retrieve value from */
+ sqlite3_context *pCtx, /* Context for sqlite3_result_xxx() calls */
+ int iCol /* Index of column to read value from */
+){
+ Fts3auxCursor *p = (Fts3auxCursor *)pCursor;
+
+ assert( p->isEof==0 );
+ switch( iCol ){
+ case 0: /* term */
+ sqlite3_result_text(pCtx, p->csr.zTerm, p->csr.nTerm, SQLITE_TRANSIENT);
+ break;
+
+ case 1: /* col */
+ if( p->iCol ){
+ sqlite3_result_int(pCtx, p->iCol-1);
+ }else{
+ sqlite3_result_text(pCtx, "*", -1, SQLITE_STATIC);
+ }
+ break;
+
+ case 2: /* documents */
+ sqlite3_result_int64(pCtx, p->aStat[p->iCol].nDoc);
+ break;
+
+ case 3: /* occurrences */
+ sqlite3_result_int64(pCtx, p->aStat[p->iCol].nOcc);
+ break;
+
+ default: /* languageid */
+ assert( iCol==4 );
+ sqlite3_result_int(pCtx, p->iLangid);
+ break;
+ }
+
+ return SQLITE_OK;
+}
+
+/*
+** xRowid - Return the current rowid for the cursor.
+*/
+static int fts3auxRowidMethod(
+ sqlite3_vtab_cursor *pCursor, /* Cursor to retrieve value from */
+ sqlite_int64 *pRowid /* OUT: Rowid value */
+){
+ Fts3auxCursor *pCsr = (Fts3auxCursor *)pCursor;
+ *pRowid = pCsr->iRowid;
+ return SQLITE_OK;
+}
+
+/*
+** Register the fts3aux module with database connection db. Return SQLITE_OK
+** if successful or an error code if sqlite3_create_module() fails.
+*/
+int sqlite3Fts3InitAux(sqlite3 *db){
+ static const sqlite3_module fts3aux_module = {
+ 0, /* iVersion */
+ fts3auxConnectMethod, /* xCreate */
+ fts3auxConnectMethod, /* xConnect */
+ fts3auxBestIndexMethod, /* xBestIndex */
+ fts3auxDisconnectMethod, /* xDisconnect */
+ fts3auxDisconnectMethod, /* xDestroy */
+ fts3auxOpenMethod, /* xOpen */
+ fts3auxCloseMethod, /* xClose */
+ fts3auxFilterMethod, /* xFilter */
+ fts3auxNextMethod, /* xNext */
+ fts3auxEofMethod, /* xEof */
+ fts3auxColumnMethod, /* xColumn */
+ fts3auxRowidMethod, /* xRowid */
+ 0, /* xUpdate */
+ 0, /* xBegin */
+ 0, /* xSync */
+ 0, /* xCommit */
+ 0, /* xRollback */
+ 0, /* xFindFunction */
+ 0, /* xRename */
+ 0, /* xSavepoint */
+ 0, /* xRelease */
+ 0, /* xRollbackTo */
+ 0, /* xShadowName */
+ 0 /* xIntegrity */
+ };
+ int rc; /* Return code */
+
+ rc = sqlite3_create_module(db, "fts4aux", &fts3aux_module, 0);
+ return rc;
+}
+
+#endif /* !defined(SQLITE_CORE) || defined(SQLITE_ENABLE_FTS3) */
diff --git a/ext/fts3/fts3_expr.c b/ext/fts3/fts3_expr.c
new file mode 100644
index 0000000000000000000000000000000000000000..9e201b168415c81ababc35826ba583e81325ba70
--- /dev/null
+++ b/ext/fts3/fts3_expr.c
@@ -0,0 +1,1292 @@
+/*
+** 2008 Nov 28
+**
+** The author disclaims copyright to this source code. In place of
+** a legal notice, here is a blessing:
+**
+** May you do good and not evil.
+** May you find forgiveness for yourself and forgive others.
+** May you share freely, never taking more than you give.
+**
+******************************************************************************
+**
+** This module contains code that implements a parser for fts3 query strings
+** (the right-hand argument to the MATCH operator). Because the supported
+** syntax is relatively simple, the whole tokenizer/parser system is
+** hand-coded.
+*/
+#include "fts3Int.h"
+#if !defined(SQLITE_CORE) || defined(SQLITE_ENABLE_FTS3)
+
+/*
+** By default, this module parses the legacy syntax that has been
+** traditionally used by fts3. Or, if SQLITE_ENABLE_FTS3_PARENTHESIS
+** is defined, then it uses the new syntax. The differences between
+** the new and the old syntaxes are:
+**
+** a) The new syntax supports parenthesis. The old does not.
+**
+** b) The new syntax supports the AND and NOT operators. The old does not.
+**
+** c) The old syntax supports the "-" token qualifier. This is not
+** supported by the new syntax (it is replaced by the NOT operator).
+**
+** d) When using the old syntax, the OR operator has a greater precedence
+** than an implicit AND. When using the new, both implicity and explicit
+** AND operators have a higher precedence than OR.
+**
+** If compiled with SQLITE_TEST defined, then this module exports the
+** symbol "int sqlite3_fts3_enable_parentheses". Setting this variable
+** to zero causes the module to use the old syntax. If it is set to
+** non-zero the new syntax is activated. This is so both syntaxes can
+** be tested using a single build of testfixture.
+**
+** The following describes the syntax supported by the fts3 MATCH
+** operator in a similar format to that used by the lemon parser
+** generator. This module does not use actually lemon, it uses a
+** custom parser.
+**
+** query ::= andexpr (OR andexpr)*.
+**
+** andexpr ::= notexpr (AND? notexpr)*.
+**
+** notexpr ::= nearexpr (NOT nearexpr|-TOKEN)*.
+** notexpr ::= LP query RP.
+**
+** nearexpr ::= phrase (NEAR distance_opt nearexpr)*.
+**
+** distance_opt ::= .
+** distance_opt ::= / INTEGER.
+**
+** phrase ::= TOKEN.
+** phrase ::= COLUMN:TOKEN.
+** phrase ::= "TOKEN TOKEN TOKEN...".
+*/
+
+#ifdef SQLITE_TEST
+int sqlite3_fts3_enable_parentheses = 0;
+#else
+# ifdef SQLITE_ENABLE_FTS3_PARENTHESIS
+# define sqlite3_fts3_enable_parentheses 1
+# else
+# define sqlite3_fts3_enable_parentheses 0
+# endif
+#endif
+
+/*
+** Default span for NEAR operators.
+*/
+#define SQLITE_FTS3_DEFAULT_NEAR_PARAM 10
+
+#include
+#include
+
+/*
+** isNot:
+** This variable is used by function getNextNode(). When getNextNode() is
+** called, it sets ParseContext.isNot to true if the 'next node' is a
+** FTSQUERY_PHRASE with a unary "-" attached to it. i.e. "mysql" in the
+** FTS3 query "sqlite -mysql". Otherwise, ParseContext.isNot is set to
+** zero.
+*/
+typedef struct ParseContext ParseContext;
+struct ParseContext {
+ sqlite3_tokenizer *pTokenizer; /* Tokenizer module */
+ int iLangid; /* Language id used with tokenizer */
+ const char **azCol; /* Array of column names for fts3 table */
+ int bFts4; /* True to allow FTS4-only syntax */
+ int nCol; /* Number of entries in azCol[] */
+ int iDefaultCol; /* Default column to query */
+ int isNot; /* True if getNextNode() sees a unary - */
+ sqlite3_context *pCtx; /* Write error message here */
+ int nNest; /* Number of nested brackets */
+};
+
+/*
+** This function is equivalent to the standard isspace() function.
+**
+** The standard isspace() can be awkward to use safely, because although it
+** is defined to accept an argument of type int, its behavior when passed
+** an integer that falls outside of the range of the unsigned char type
+** is undefined (and sometimes, "undefined" means segfault). This wrapper
+** is defined to accept an argument of type char, and always returns 0 for
+** any values that fall outside of the range of the unsigned char type (i.e.
+** negative values).
+*/
+static int fts3isspace(char c){
+ return c==' ' || c=='\t' || c=='\n' || c=='\r' || c=='\v' || c=='\f';
+}
+
+/*
+** Allocate nByte bytes of memory using sqlite3_malloc(). If successful,
+** zero the memory before returning a pointer to it. If unsuccessful,
+** return NULL.
+*/
+void *sqlite3Fts3MallocZero(sqlite3_int64 nByte){
+ void *pRet = sqlite3_malloc64(nByte);
+ if( pRet ) memset(pRet, 0, nByte);
+ return pRet;
+}
+
+int sqlite3Fts3OpenTokenizer(
+ sqlite3_tokenizer *pTokenizer,
+ int iLangid,
+ const char *z,
+ int n,
+ sqlite3_tokenizer_cursor **ppCsr
+){
+ sqlite3_tokenizer_module const *pModule = pTokenizer->pModule;
+ sqlite3_tokenizer_cursor *pCsr = 0;
+ int rc;
+
+ rc = pModule->xOpen(pTokenizer, z, n, &pCsr);
+ assert( rc==SQLITE_OK || pCsr==0 );
+ if( rc==SQLITE_OK ){
+ pCsr->pTokenizer = pTokenizer;
+ if( pModule->iVersion>=1 ){
+ rc = pModule->xLanguageid(pCsr, iLangid);
+ if( rc!=SQLITE_OK ){
+ pModule->xClose(pCsr);
+ pCsr = 0;
+ }
+ }
+ }
+ *ppCsr = pCsr;
+ return rc;
+}
+
+/*
+** Function getNextNode(), which is called by fts3ExprParse(), may itself
+** call fts3ExprParse(). So this forward declaration is required.
+*/
+static int fts3ExprParse(ParseContext *, const char *, int, Fts3Expr **, int *);
+
+/*
+** Extract the next token from buffer z (length n) using the tokenizer
+** and other information (column names etc.) in pParse. Create an Fts3Expr
+** structure of type FTSQUERY_PHRASE containing a phrase consisting of this
+** single token and set *ppExpr to point to it. If the end of the buffer is
+** reached before a token is found, set *ppExpr to zero. It is the
+** responsibility of the caller to eventually deallocate the allocated
+** Fts3Expr structure (if any) by passing it to sqlite3_free().
+**
+** Return SQLITE_OK if successful, or SQLITE_NOMEM if a memory allocation
+** fails.
+*/
+static int getNextToken(
+ ParseContext *pParse, /* fts3 query parse context */
+ int iCol, /* Value for Fts3Phrase.iColumn */
+ const char *z, int n, /* Input string */
+ Fts3Expr **ppExpr, /* OUT: expression */
+ int *pnConsumed /* OUT: Number of bytes consumed */
+){
+ sqlite3_tokenizer *pTokenizer = pParse->pTokenizer;
+ sqlite3_tokenizer_module const *pModule = pTokenizer->pModule;
+ int rc;
+ sqlite3_tokenizer_cursor *pCursor;
+ Fts3Expr *pRet = 0;
+ int i = 0;
+
+ /* Set variable i to the maximum number of bytes of input to tokenize. */
+ for(i=0; iiLangid, z, i, &pCursor);
+ if( rc==SQLITE_OK ){
+ const char *zToken;
+ int nToken = 0, iStart = 0, iEnd = 0, iPosition = 0;
+ sqlite3_int64 nByte; /* total space to allocate */
+
+ rc = pModule->xNext(pCursor, &zToken, &nToken, &iStart, &iEnd, &iPosition);
+ if( rc==SQLITE_OK ){
+ nByte = sizeof(Fts3Expr) + sizeof(Fts3Phrase) + nToken;
+ pRet = (Fts3Expr *)sqlite3Fts3MallocZero(nByte);
+ if( !pRet ){
+ rc = SQLITE_NOMEM;
+ }else{
+ pRet->eType = FTSQUERY_PHRASE;
+ pRet->pPhrase = (Fts3Phrase *)&pRet[1];
+ pRet->pPhrase->nToken = 1;
+ pRet->pPhrase->iColumn = iCol;
+ pRet->pPhrase->aToken[0].n = nToken;
+ pRet->pPhrase->aToken[0].z = (char *)&pRet->pPhrase[1];
+ memcpy(pRet->pPhrase->aToken[0].z, zToken, nToken);
+
+ if( iEndpPhrase->aToken[0].isPrefix = 1;
+ iEnd++;
+ }
+
+ while( 1 ){
+ if( !sqlite3_fts3_enable_parentheses
+ && iStart>0 && z[iStart-1]=='-'
+ ){
+ pParse->isNot = 1;
+ iStart--;
+ }else if( pParse->bFts4 && iStart>0 && z[iStart-1]=='^' ){
+ pRet->pPhrase->aToken[0].bFirst = 1;
+ iStart--;
+ }else{
+ break;
+ }
+ }
+
+ }
+ *pnConsumed = iEnd;
+ }else if( i && rc==SQLITE_DONE ){
+ rc = SQLITE_OK;
+ }
+
+ pModule->xClose(pCursor);
+ }
+
+ *ppExpr = pRet;
+ return rc;
+}
+
+
+/*
+** Enlarge a memory allocation. If an out-of-memory allocation occurs,
+** then free the old allocation.
+*/
+static void *fts3ReallocOrFree(void *pOrig, sqlite3_int64 nNew){
+ void *pRet = sqlite3_realloc64(pOrig, nNew);
+ if( !pRet ){
+ sqlite3_free(pOrig);
+ }
+ return pRet;
+}
+
+/*
+** Buffer zInput, length nInput, contains the contents of a quoted string
+** that appeared as part of an fts3 query expression. Neither quote character
+** is included in the buffer. This function attempts to tokenize the entire
+** input buffer and create an Fts3Expr structure of type FTSQUERY_PHRASE
+** containing the results.
+**
+** If successful, SQLITE_OK is returned and *ppExpr set to point at the
+** allocated Fts3Expr structure. Otherwise, either SQLITE_NOMEM (out of memory
+** error) or SQLITE_ERROR (tokenization error) is returned and *ppExpr set
+** to 0.
+*/
+static int getNextString(
+ ParseContext *pParse, /* fts3 query parse context */
+ const char *zInput, int nInput, /* Input string */
+ Fts3Expr **ppExpr /* OUT: expression */
+){
+ sqlite3_tokenizer *pTokenizer = pParse->pTokenizer;
+ sqlite3_tokenizer_module const *pModule = pTokenizer->pModule;
+ int rc;
+ Fts3Expr *p = 0;
+ sqlite3_tokenizer_cursor *pCursor = 0;
+ char *zTemp = 0;
+ int nTemp = 0;
+
+ const int nSpace = sizeof(Fts3Expr) + sizeof(Fts3Phrase);
+ int nToken = 0;
+
+ /* The final Fts3Expr data structure, including the Fts3Phrase,
+ ** Fts3PhraseToken structures token buffers are all stored as a single
+ ** allocation so that the expression can be freed with a single call to
+ ** sqlite3_free(). Setting this up requires a two pass approach.
+ **
+ ** The first pass, in the block below, uses a tokenizer cursor to iterate
+ ** through the tokens in the expression. This pass uses fts3ReallocOrFree()
+ ** to assemble data in two dynamic buffers:
+ **
+ ** Buffer p: Points to the Fts3Expr structure, followed by the Fts3Phrase
+ ** structure, followed by the array of Fts3PhraseToken
+ ** structures. This pass only populates the Fts3PhraseToken array.
+ **
+ ** Buffer zTemp: Contains copies of all tokens.
+ **
+ ** The second pass, in the block that begins "if( rc==SQLITE_DONE )" below,
+ ** appends buffer zTemp to buffer p, and fills in the Fts3Expr and Fts3Phrase
+ ** structures.
+ */
+ rc = sqlite3Fts3OpenTokenizer(
+ pTokenizer, pParse->iLangid, zInput, nInput, &pCursor);
+ if( rc==SQLITE_OK ){
+ int ii;
+ for(ii=0; rc==SQLITE_OK; ii++){
+ const char *zByte;
+ int nByte = 0, iBegin = 0, iEnd = 0, iPos = 0;
+ rc = pModule->xNext(pCursor, &zByte, &nByte, &iBegin, &iEnd, &iPos);
+ if( rc==SQLITE_OK ){
+ Fts3PhraseToken *pToken;
+
+ p = fts3ReallocOrFree(p, nSpace + ii*sizeof(Fts3PhraseToken));
+ zTemp = fts3ReallocOrFree(zTemp, nTemp + nByte);
+ if( !zTemp || !p ){
+ rc = SQLITE_NOMEM;
+ goto getnextstring_out;
+ }
+
+ assert( nToken==ii );
+ pToken = &((Fts3Phrase *)(&p[1]))->aToken[ii];
+ memset(pToken, 0, sizeof(Fts3PhraseToken));
+
+ memcpy(&zTemp[nTemp], zByte, nByte);
+ nTemp += nByte;
+
+ pToken->n = nByte;
+ pToken->isPrefix = (iEndbFirst = (iBegin>0 && zInput[iBegin-1]=='^');
+ nToken = ii+1;
+ }
+ }
+ }
+
+ if( rc==SQLITE_DONE ){
+ int jj;
+ char *zBuf = 0;
+
+ p = fts3ReallocOrFree(p, nSpace + nToken*sizeof(Fts3PhraseToken) + nTemp);
+ if( !p ){
+ rc = SQLITE_NOMEM;
+ goto getnextstring_out;
+ }
+ memset(p, 0, (char *)&(((Fts3Phrase *)&p[1])->aToken[0])-(char *)p);
+ p->eType = FTSQUERY_PHRASE;
+ p->pPhrase = (Fts3Phrase *)&p[1];
+ p->pPhrase->iColumn = pParse->iDefaultCol;
+ p->pPhrase->nToken = nToken;
+
+ zBuf = (char *)&p->pPhrase->aToken[nToken];
+ assert( nTemp==0 || zTemp );
+ if( zTemp ){
+ memcpy(zBuf, zTemp, nTemp);
+ }
+
+ for(jj=0; jjpPhrase->nToken; jj++){
+ p->pPhrase->aToken[jj].z = zBuf;
+ zBuf += p->pPhrase->aToken[jj].n;
+ }
+ rc = SQLITE_OK;
+ }
+
+ getnextstring_out:
+ if( pCursor ){
+ pModule->xClose(pCursor);
+ }
+ sqlite3_free(zTemp);
+ if( rc!=SQLITE_OK ){
+ sqlite3_free(p);
+ p = 0;
+ }
+ *ppExpr = p;
+ return rc;
+}
+
+/*
+** The output variable *ppExpr is populated with an allocated Fts3Expr
+** structure, or set to 0 if the end of the input buffer is reached.
+**
+** Returns an SQLite error code. SQLITE_OK if everything works, SQLITE_NOMEM
+** if a malloc failure occurs, or SQLITE_ERROR if a parse error is encountered.
+** If SQLITE_ERROR is returned, pContext is populated with an error message.
+*/
+static int getNextNode(
+ ParseContext *pParse, /* fts3 query parse context */
+ const char *z, int n, /* Input string */
+ Fts3Expr **ppExpr, /* OUT: expression */
+ int *pnConsumed /* OUT: Number of bytes consumed */
+){
+ static const struct Fts3Keyword {
+ char *z; /* Keyword text */
+ unsigned char n; /* Length of the keyword */
+ unsigned char parenOnly; /* Only valid in paren mode */
+ unsigned char eType; /* Keyword code */
+ } aKeyword[] = {
+ { "OR" , 2, 0, FTSQUERY_OR },
+ { "AND", 3, 1, FTSQUERY_AND },
+ { "NOT", 3, 1, FTSQUERY_NOT },
+ { "NEAR", 4, 0, FTSQUERY_NEAR }
+ };
+ int ii;
+ int iCol;
+ int iColLen;
+ int rc;
+ Fts3Expr *pRet = 0;
+
+ const char *zInput = z;
+ int nInput = n;
+
+ pParse->isNot = 0;
+
+ /* Skip over any whitespace before checking for a keyword, an open or
+ ** close bracket, or a quoted string.
+ */
+ while( nInput>0 && fts3isspace(*zInput) ){
+ nInput--;
+ zInput++;
+ }
+ if( nInput==0 ){
+ return SQLITE_DONE;
+ }
+
+ /* See if we are dealing with a keyword. */
+ for(ii=0; ii<(int)(sizeof(aKeyword)/sizeof(struct Fts3Keyword)); ii++){
+ const struct Fts3Keyword *pKey = &aKeyword[ii];
+
+ if( (pKey->parenOnly & ~sqlite3_fts3_enable_parentheses)!=0 ){
+ continue;
+ }
+
+ if( nInput>=pKey->n && 0==memcmp(zInput, pKey->z, pKey->n) ){
+ int nNear = SQLITE_FTS3_DEFAULT_NEAR_PARAM;
+ int nKey = pKey->n;
+ char cNext;
+
+ /* If this is a "NEAR" keyword, check for an explicit nearness. */
+ if( pKey->eType==FTSQUERY_NEAR ){
+ assert( nKey==4 );
+ if( zInput[4]=='/' && zInput[5]>='0' && zInput[5]<='9' ){
+ nKey += 1+sqlite3Fts3ReadInt(&zInput[nKey+1], &nNear);
+ }
+ }
+
+ /* At this point this is probably a keyword. But for that to be true,
+ ** the next byte must contain either whitespace, an open or close
+ ** parenthesis, a quote character, or EOF.
+ */
+ cNext = zInput[nKey];
+ if( fts3isspace(cNext)
+ || cNext=='"' || cNext=='(' || cNext==')' || cNext==0
+ ){
+ pRet = (Fts3Expr *)sqlite3Fts3MallocZero(sizeof(Fts3Expr));
+ if( !pRet ){
+ return SQLITE_NOMEM;
+ }
+ pRet->eType = pKey->eType;
+ pRet->nNear = nNear;
+ *ppExpr = pRet;
+ *pnConsumed = (int)((zInput - z) + nKey);
+ return SQLITE_OK;
+ }
+
+ /* Turns out that wasn't a keyword after all. This happens if the
+ ** user has supplied a token such as "ORacle". Continue.
+ */
+ }
+ }
+
+ /* See if we are dealing with a quoted phrase. If this is the case, then
+ ** search for the closing quote and pass the whole string to getNextString()
+ ** for processing. This is easy to do, as fts3 has no syntax for escaping
+ ** a quote character embedded in a string.
+ */
+ if( *zInput=='"' ){
+ for(ii=1; iinNest++;
+#if !defined(SQLITE_MAX_EXPR_DEPTH)
+ if( pParse->nNest>1000 ) return SQLITE_ERROR;
+#elif SQLITE_MAX_EXPR_DEPTH>0
+ if( pParse->nNest>SQLITE_MAX_EXPR_DEPTH ) return SQLITE_ERROR;
+#endif
+ rc = fts3ExprParse(pParse, zInput+1, nInput-1, ppExpr, &nConsumed);
+ *pnConsumed = (int)(zInput - z) + 1 + nConsumed;
+ return rc;
+ }else if( *zInput==')' ){
+ pParse->nNest--;
+ *pnConsumed = (int)((zInput - z) + 1);
+ *ppExpr = 0;
+ return SQLITE_DONE;
+ }
+ }
+
+ /* If control flows to this point, this must be a regular token, or
+ ** the end of the input. Read a regular token using the sqlite3_tokenizer
+ ** interface. Before doing so, figure out if there is an explicit
+ ** column specifier for the token.
+ **
+ ** TODO: Strangely, it is not possible to associate a column specifier
+ ** with a quoted phrase, only with a single token. Not sure if this was
+ ** an implementation artifact or an intentional decision when fts3 was
+ ** first implemented. Whichever it was, this module duplicates the
+ ** limitation.
+ */
+ iCol = pParse->iDefaultCol;
+ iColLen = 0;
+ for(ii=0; iinCol; ii++){
+ const char *zStr = pParse->azCol[ii];
+ int nStr = (int)strlen(zStr);
+ if( nInput>nStr && zInput[nStr]==':'
+ && sqlite3_strnicmp(zStr, zInput, nStr)==0
+ ){
+ iCol = ii;
+ iColLen = (int)((zInput - z) + nStr + 1);
+ break;
+ }
+ }
+ rc = getNextToken(pParse, iCol, &z[iColLen], n-iColLen, ppExpr, pnConsumed);
+ *pnConsumed += iColLen;
+ return rc;
+}
+
+/*
+** The argument is an Fts3Expr structure for a binary operator (any type
+** except an FTSQUERY_PHRASE). Return an integer value representing the
+** precedence of the operator. Lower values have a higher precedence (i.e.
+** group more tightly). For example, in the C language, the == operator
+** groups more tightly than ||, and would therefore have a higher precedence.
+**
+** When using the new fts3 query syntax (when SQLITE_ENABLE_FTS3_PARENTHESIS
+** is defined), the order of the operators in precedence from highest to
+** lowest is:
+**
+** NEAR
+** NOT
+** AND (including implicit ANDs)
+** OR
+**
+** Note that when using the old query syntax, the OR operator has a higher
+** precedence than the AND operator.
+*/
+static int opPrecedence(Fts3Expr *p){
+ assert( p->eType!=FTSQUERY_PHRASE );
+ if( sqlite3_fts3_enable_parentheses ){
+ return p->eType;
+ }else if( p->eType==FTSQUERY_NEAR ){
+ return 1;
+ }else if( p->eType==FTSQUERY_OR ){
+ return 2;
+ }
+ assert( p->eType==FTSQUERY_AND );
+ return 3;
+}
+
+/*
+** Argument ppHead contains a pointer to the current head of a query
+** expression tree being parsed. pPrev is the expression node most recently
+** inserted into the tree. This function adds pNew, which is always a binary
+** operator node, into the expression tree based on the relative precedence
+** of pNew and the existing nodes of the tree. This may result in the head
+** of the tree changing, in which case *ppHead is set to the new root node.
+*/
+static void insertBinaryOperator(
+ Fts3Expr **ppHead, /* Pointer to the root node of a tree */
+ Fts3Expr *pPrev, /* Node most recently inserted into the tree */
+ Fts3Expr *pNew /* New binary node to insert into expression tree */
+){
+ Fts3Expr *pSplit = pPrev;
+ while( pSplit->pParent && opPrecedence(pSplit->pParent)<=opPrecedence(pNew) ){
+ pSplit = pSplit->pParent;
+ }
+
+ if( pSplit->pParent ){
+ assert( pSplit->pParent->pRight==pSplit );
+ pSplit->pParent->pRight = pNew;
+ pNew->pParent = pSplit->pParent;
+ }else{
+ *ppHead = pNew;
+ }
+ pNew->pLeft = pSplit;
+ pSplit->pParent = pNew;
+}
+
+/*
+** Parse the fts3 query expression found in buffer z, length n. This function
+** returns either when the end of the buffer is reached or an unmatched
+** closing bracket - ')' - is encountered.
+**
+** If successful, SQLITE_OK is returned, *ppExpr is set to point to the
+** parsed form of the expression and *pnConsumed is set to the number of
+** bytes read from buffer z. Otherwise, *ppExpr is set to 0 and SQLITE_NOMEM
+** (out of memory error) or SQLITE_ERROR (parse error) is returned.
+*/
+static int fts3ExprParse(
+ ParseContext *pParse, /* fts3 query parse context */
+ const char *z, int n, /* Text of MATCH query */
+ Fts3Expr **ppExpr, /* OUT: Parsed query structure */
+ int *pnConsumed /* OUT: Number of bytes consumed */
+){
+ Fts3Expr *pRet = 0;
+ Fts3Expr *pPrev = 0;
+ Fts3Expr *pNotBranch = 0; /* Only used in legacy parse mode */
+ int nIn = n;
+ const char *zIn = z;
+ int rc = SQLITE_OK;
+ int isRequirePhrase = 1;
+
+ while( rc==SQLITE_OK ){
+ Fts3Expr *p = 0;
+ int nByte = 0;
+
+ rc = getNextNode(pParse, zIn, nIn, &p, &nByte);
+ assert( nByte>0 || (rc!=SQLITE_OK && p==0) );
+ if( rc==SQLITE_OK ){
+ if( p ){
+ int isPhrase;
+
+ if( !sqlite3_fts3_enable_parentheses
+ && p->eType==FTSQUERY_PHRASE && pParse->isNot
+ ){
+ /* Create an implicit NOT operator. */
+ Fts3Expr *pNot = sqlite3Fts3MallocZero(sizeof(Fts3Expr));
+ if( !pNot ){
+ sqlite3Fts3ExprFree(p);
+ rc = SQLITE_NOMEM;
+ goto exprparse_out;
+ }
+ pNot->eType = FTSQUERY_NOT;
+ pNot->pRight = p;
+ p->pParent = pNot;
+ if( pNotBranch ){
+ pNot->pLeft = pNotBranch;
+ pNotBranch->pParent = pNot;
+ }
+ pNotBranch = pNot;
+ p = pPrev;
+ }else{
+ int eType = p->eType;
+ isPhrase = (eType==FTSQUERY_PHRASE || p->pLeft);
+
+ /* The isRequirePhrase variable is set to true if a phrase or
+ ** an expression contained in parenthesis is required. If a
+ ** binary operator (AND, OR, NOT or NEAR) is encounted when
+ ** isRequirePhrase is set, this is a syntax error.
+ */
+ if( !isPhrase && isRequirePhrase ){
+ sqlite3Fts3ExprFree(p);
+ rc = SQLITE_ERROR;
+ goto exprparse_out;
+ }
+
+ if( isPhrase && !isRequirePhrase ){
+ /* Insert an implicit AND operator. */
+ Fts3Expr *pAnd;
+ assert( pRet && pPrev );
+ pAnd = sqlite3Fts3MallocZero(sizeof(Fts3Expr));
+ if( !pAnd ){
+ sqlite3Fts3ExprFree(p);
+ rc = SQLITE_NOMEM;
+ goto exprparse_out;
+ }
+ pAnd->eType = FTSQUERY_AND;
+ insertBinaryOperator(&pRet, pPrev, pAnd);
+ pPrev = pAnd;
+ }
+
+ /* This test catches attempts to make either operand of a NEAR
+ ** operator something other than a phrase. For example, either of
+ ** the following:
+ **
+ ** (bracketed expression) NEAR phrase
+ ** phrase NEAR (bracketed expression)
+ **
+ ** Return an error in either case.
+ */
+ if( pPrev && (
+ (eType==FTSQUERY_NEAR && !isPhrase && pPrev->eType!=FTSQUERY_PHRASE)
+ || (eType!=FTSQUERY_PHRASE && isPhrase && pPrev->eType==FTSQUERY_NEAR)
+ )){
+ sqlite3Fts3ExprFree(p);
+ rc = SQLITE_ERROR;
+ goto exprparse_out;
+ }
+
+ if( isPhrase ){
+ if( pRet ){
+ assert( pPrev && pPrev->pLeft && pPrev->pRight==0 );
+ pPrev->pRight = p;
+ p->pParent = pPrev;
+ }else{
+ pRet = p;
+ }
+ }else{
+ insertBinaryOperator(&pRet, pPrev, p);
+ }
+ isRequirePhrase = !isPhrase;
+ }
+ pPrev = p;
+ }
+ assert( nByte>0 );
+ }
+ assert( rc!=SQLITE_OK || (nByte>0 && nByte<=nIn) );
+ nIn -= nByte;
+ zIn += nByte;
+ }
+
+ if( rc==SQLITE_DONE && pRet && isRequirePhrase ){
+ rc = SQLITE_ERROR;
+ }
+
+ if( rc==SQLITE_DONE ){
+ rc = SQLITE_OK;
+ if( !sqlite3_fts3_enable_parentheses && pNotBranch ){
+ if( !pRet ){
+ rc = SQLITE_ERROR;
+ }else{
+ Fts3Expr *pIter = pNotBranch;
+ while( pIter->pLeft ){
+ pIter = pIter->pLeft;
+ }
+ pIter->pLeft = pRet;
+ pRet->pParent = pIter;
+ pRet = pNotBranch;
+ }
+ }
+ }
+ *pnConsumed = n - nIn;
+
+exprparse_out:
+ if( rc!=SQLITE_OK ){
+ sqlite3Fts3ExprFree(pRet);
+ sqlite3Fts3ExprFree(pNotBranch);
+ pRet = 0;
+ }
+ *ppExpr = pRet;
+ return rc;
+}
+
+/*
+** Return SQLITE_ERROR if the maximum depth of the expression tree passed
+** as the only argument is more than nMaxDepth.
+*/
+static int fts3ExprCheckDepth(Fts3Expr *p, int nMaxDepth){
+ int rc = SQLITE_OK;
+ if( p ){
+ if( nMaxDepth<0 ){
+ rc = SQLITE_TOOBIG;
+ }else{
+ rc = fts3ExprCheckDepth(p->pLeft, nMaxDepth-1);
+ if( rc==SQLITE_OK ){
+ rc = fts3ExprCheckDepth(p->pRight, nMaxDepth-1);
+ }
+ }
+ }
+ return rc;
+}
+
+/*
+** This function attempts to transform the expression tree at (*pp) to
+** an equivalent but more balanced form. The tree is modified in place.
+** If successful, SQLITE_OK is returned and (*pp) set to point to the
+** new root expression node.
+**
+** nMaxDepth is the maximum allowable depth of the balanced sub-tree.
+**
+** Otherwise, if an error occurs, an SQLite error code is returned and
+** expression (*pp) freed.
+*/
+static int fts3ExprBalance(Fts3Expr **pp, int nMaxDepth){
+ int rc = SQLITE_OK; /* Return code */
+ Fts3Expr *pRoot = *pp; /* Initial root node */
+ Fts3Expr *pFree = 0; /* List of free nodes. Linked by pParent. */
+ int eType = pRoot->eType; /* Type of node in this tree */
+
+ if( nMaxDepth==0 ){
+ rc = SQLITE_ERROR;
+ }
+
+ if( rc==SQLITE_OK ){
+ if( (eType==FTSQUERY_AND || eType==FTSQUERY_OR) ){
+ Fts3Expr **apLeaf;
+ apLeaf = (Fts3Expr **)sqlite3_malloc64(sizeof(Fts3Expr *) * nMaxDepth);
+ if( 0==apLeaf ){
+ rc = SQLITE_NOMEM;
+ }else{
+ memset(apLeaf, 0, sizeof(Fts3Expr *) * nMaxDepth);
+ }
+
+ if( rc==SQLITE_OK ){
+ int i;
+ Fts3Expr *p;
+
+ /* Set $p to point to the left-most leaf in the tree of eType nodes. */
+ for(p=pRoot; p->eType==eType; p=p->pLeft){
+ assert( p->pParent==0 || p->pParent->pLeft==p );
+ assert( p->pLeft && p->pRight );
+ }
+
+ /* This loop runs once for each leaf in the tree of eType nodes. */
+ while( 1 ){
+ int iLvl;
+ Fts3Expr *pParent = p->pParent; /* Current parent of p */
+
+ assert( pParent==0 || pParent->pLeft==p );
+ p->pParent = 0;
+ if( pParent ){
+ pParent->pLeft = 0;
+ }else{
+ pRoot = 0;
+ }
+ rc = fts3ExprBalance(&p, nMaxDepth-1);
+ if( rc!=SQLITE_OK ) break;
+
+ for(iLvl=0; p && iLvlpLeft = apLeaf[iLvl];
+ pFree->pRight = p;
+ pFree->pLeft->pParent = pFree;
+ pFree->pRight->pParent = pFree;
+
+ p = pFree;
+ pFree = pFree->pParent;
+ p->pParent = 0;
+ apLeaf[iLvl] = 0;
+ }
+ }
+ if( p ){
+ sqlite3Fts3ExprFree(p);
+ rc = SQLITE_TOOBIG;
+ break;
+ }
+
+ /* If that was the last leaf node, break out of the loop */
+ if( pParent==0 ) break;
+
+ /* Set $p to point to the next leaf in the tree of eType nodes */
+ for(p=pParent->pRight; p->eType==eType; p=p->pLeft);
+
+ /* Remove pParent from the original tree. */
+ assert( pParent->pParent==0 || pParent->pParent->pLeft==pParent );
+ pParent->pRight->pParent = pParent->pParent;
+ if( pParent->pParent ){
+ pParent->pParent->pLeft = pParent->pRight;
+ }else{
+ assert( pParent==pRoot );
+ pRoot = pParent->pRight;
+ }
+
+ /* Link pParent into the free node list. It will be used as an
+ ** internal node of the new tree. */
+ pParent->pParent = pFree;
+ pFree = pParent;
+ }
+
+ if( rc==SQLITE_OK ){
+ p = 0;
+ for(i=0; ipParent = 0;
+ }else{
+ assert( pFree!=0 );
+ pFree->pRight = p;
+ pFree->pLeft = apLeaf[i];
+ pFree->pLeft->pParent = pFree;
+ pFree->pRight->pParent = pFree;
+
+ p = pFree;
+ pFree = pFree->pParent;
+ p->pParent = 0;
+ }
+ }
+ }
+ pRoot = p;
+ }else{
+ /* An error occurred. Delete the contents of the apLeaf[] array
+ ** and pFree list. Everything else is cleaned up by the call to
+ ** sqlite3Fts3ExprFree(pRoot) below. */
+ Fts3Expr *pDel;
+ for(i=0; ipParent;
+ sqlite3_free(pDel);
+ }
+ }
+
+ assert( pFree==0 );
+ sqlite3_free( apLeaf );
+ }
+ }else if( eType==FTSQUERY_NOT ){
+ Fts3Expr *pLeft = pRoot->pLeft;
+ Fts3Expr *pRight = pRoot->pRight;
+
+ pRoot->pLeft = 0;
+ pRoot->pRight = 0;
+ pLeft->pParent = 0;
+ pRight->pParent = 0;
+
+ rc = fts3ExprBalance(&pLeft, nMaxDepth-1);
+ if( rc==SQLITE_OK ){
+ rc = fts3ExprBalance(&pRight, nMaxDepth-1);
+ }
+
+ if( rc!=SQLITE_OK ){
+ sqlite3Fts3ExprFree(pRight);
+ sqlite3Fts3ExprFree(pLeft);
+ }else{
+ assert( pLeft && pRight );
+ pRoot->pLeft = pLeft;
+ pLeft->pParent = pRoot;
+ pRoot->pRight = pRight;
+ pRight->pParent = pRoot;
+ }
+ }
+ }
+
+ if( rc!=SQLITE_OK ){
+ sqlite3Fts3ExprFree(pRoot);
+ pRoot = 0;
+ }
+ *pp = pRoot;
+ return rc;
+}
+
+/*
+** This function is similar to sqlite3Fts3ExprParse(), with the following
+** differences:
+**
+** 1. It does not do expression rebalancing.
+** 2. It does not check that the expression does not exceed the
+** maximum allowable depth.
+** 3. Even if it fails, *ppExpr may still be set to point to an
+** expression tree. It should be deleted using sqlite3Fts3ExprFree()
+** in this case.
+*/
+static int fts3ExprParseUnbalanced(
+ sqlite3_tokenizer *pTokenizer, /* Tokenizer module */
+ int iLangid, /* Language id for tokenizer */
+ char **azCol, /* Array of column names for fts3 table */
+ int bFts4, /* True to allow FTS4-only syntax */
+ int nCol, /* Number of entries in azCol[] */
+ int iDefaultCol, /* Default column to query */
+ const char *z, int n, /* Text of MATCH query */
+ Fts3Expr **ppExpr /* OUT: Parsed query structure */
+){
+ int nParsed;
+ int rc;
+ ParseContext sParse;
+
+ memset(&sParse, 0, sizeof(ParseContext));
+ sParse.pTokenizer = pTokenizer;
+ sParse.iLangid = iLangid;
+ sParse.azCol = (const char **)azCol;
+ sParse.nCol = nCol;
+ sParse.iDefaultCol = iDefaultCol;
+ sParse.bFts4 = bFts4;
+ if( z==0 ){
+ *ppExpr = 0;
+ return SQLITE_OK;
+ }
+ if( n<0 ){
+ n = (int)strlen(z);
+ }
+ rc = fts3ExprParse(&sParse, z, n, ppExpr, &nParsed);
+ assert( rc==SQLITE_OK || *ppExpr==0 );
+
+ /* Check for mismatched parenthesis */
+ if( rc==SQLITE_OK && sParse.nNest ){
+ rc = SQLITE_ERROR;
+ }
+
+ return rc;
+}
+
+/*
+** Parameters z and n contain a pointer to and length of a buffer containing
+** an fts3 query expression, respectively. This function attempts to parse the
+** query expression and create a tree of Fts3Expr structures representing the
+** parsed expression. If successful, *ppExpr is set to point to the head
+** of the parsed expression tree and SQLITE_OK is returned. If an error
+** occurs, either SQLITE_NOMEM (out-of-memory error) or SQLITE_ERROR (parse
+** error) is returned and *ppExpr is set to 0.
+**
+** If parameter n is a negative number, then z is assumed to point to a
+** nul-terminated string and the length is determined using strlen().
+**
+** The first parameter, pTokenizer, is passed the fts3 tokenizer module to
+** use to normalize query tokens while parsing the expression. The azCol[]
+** array, which is assumed to contain nCol entries, should contain the names
+** of each column in the target fts3 table, in order from left to right.
+** Column names must be nul-terminated strings.
+**
+** The iDefaultCol parameter should be passed the index of the table column
+** that appears on the left-hand-side of the MATCH operator (the default
+** column to match against for tokens for which a column name is not explicitly
+** specified as part of the query string), or -1 if tokens may by default
+** match any table column.
+*/
+int sqlite3Fts3ExprParse(
+ sqlite3_tokenizer *pTokenizer, /* Tokenizer module */
+ int iLangid, /* Language id for tokenizer */
+ char **azCol, /* Array of column names for fts3 table */
+ int bFts4, /* True to allow FTS4-only syntax */
+ int nCol, /* Number of entries in azCol[] */
+ int iDefaultCol, /* Default column to query */
+ const char *z, int n, /* Text of MATCH query */
+ Fts3Expr **ppExpr, /* OUT: Parsed query structure */
+ char **pzErr /* OUT: Error message (sqlite3_malloc) */
+){
+ int rc = fts3ExprParseUnbalanced(
+ pTokenizer, iLangid, azCol, bFts4, nCol, iDefaultCol, z, n, ppExpr
+ );
+
+ /* Rebalance the expression. And check that its depth does not exceed
+ ** SQLITE_FTS3_MAX_EXPR_DEPTH. */
+ if( rc==SQLITE_OK && *ppExpr ){
+ rc = fts3ExprBalance(ppExpr, SQLITE_FTS3_MAX_EXPR_DEPTH);
+ if( rc==SQLITE_OK ){
+ rc = fts3ExprCheckDepth(*ppExpr, SQLITE_FTS3_MAX_EXPR_DEPTH);
+ }
+ }
+
+ if( rc!=SQLITE_OK ){
+ sqlite3Fts3ExprFree(*ppExpr);
+ *ppExpr = 0;
+ if( rc==SQLITE_TOOBIG ){
+ sqlite3Fts3ErrMsg(pzErr,
+ "FTS expression tree is too large (maximum depth %d)",
+ SQLITE_FTS3_MAX_EXPR_DEPTH
+ );
+ rc = SQLITE_ERROR;
+ }else if( rc==SQLITE_ERROR ){
+ sqlite3Fts3ErrMsg(pzErr, "malformed MATCH expression: [%s]", z);
+ }
+ }
+
+ return rc;
+}
+
+/*
+** Free a single node of an expression tree.
+*/
+static void fts3FreeExprNode(Fts3Expr *p){
+ assert( p->eType==FTSQUERY_PHRASE || p->pPhrase==0 );
+ sqlite3Fts3EvalPhraseCleanup(p->pPhrase);
+ sqlite3_free(p->aMI);
+ sqlite3_free(p);
+}
+
+/*
+** Free a parsed fts3 query expression allocated by sqlite3Fts3ExprParse().
+**
+** This function would be simpler if it recursively called itself. But
+** that would mean passing a sufficiently large expression to ExprParse()
+** could cause a stack overflow.
+*/
+void sqlite3Fts3ExprFree(Fts3Expr *pDel){
+ Fts3Expr *p;
+ assert( pDel==0 || pDel->pParent==0 );
+ for(p=pDel; p && (p->pLeft||p->pRight); p=(p->pLeft ? p->pLeft : p->pRight)){
+ assert( p->pParent==0 || p==p->pParent->pRight || p==p->pParent->pLeft );
+ }
+ while( p ){
+ Fts3Expr *pParent = p->pParent;
+ fts3FreeExprNode(p);
+ if( pParent && p==pParent->pLeft && pParent->pRight ){
+ p = pParent->pRight;
+ while( p && (p->pLeft || p->pRight) ){
+ assert( p==p->pParent->pRight || p==p->pParent->pLeft );
+ p = (p->pLeft ? p->pLeft : p->pRight);
+ }
+ }else{
+ p = pParent;
+ }
+ }
+}
+
+/****************************************************************************
+*****************************************************************************
+** Everything after this point is just test code.
+*/
+
+#ifdef SQLITE_TEST
+
+#include
+
+/*
+** Return a pointer to a buffer containing a text representation of the
+** expression passed as the first argument. The buffer is obtained from
+** sqlite3_malloc(). It is the responsibility of the caller to use
+** sqlite3_free() to release the memory. If an OOM condition is encountered,
+** NULL is returned.
+**
+** If the second argument is not NULL, then its contents are prepended to
+** the returned expression text and then freed using sqlite3_free().
+*/
+static char *exprToString(Fts3Expr *pExpr, char *zBuf){
+ if( pExpr==0 ){
+ return sqlite3_mprintf("");
+ }
+ switch( pExpr->eType ){
+ case FTSQUERY_PHRASE: {
+ Fts3Phrase *pPhrase = pExpr->pPhrase;
+ int i;
+ zBuf = sqlite3_mprintf(
+ "%zPHRASE %d 0", zBuf, pPhrase->iColumn);
+ for(i=0; zBuf && inToken; i++){
+ zBuf = sqlite3_mprintf("%z %.*s%s", zBuf,
+ pPhrase->aToken[i].n, pPhrase->aToken[i].z,
+ (pPhrase->aToken[i].isPrefix?"+":"")
+ );
+ }
+ return zBuf;
+ }
+
+ case FTSQUERY_NEAR:
+ zBuf = sqlite3_mprintf("%zNEAR/%d ", zBuf, pExpr->nNear);
+ break;
+ case FTSQUERY_NOT:
+ zBuf = sqlite3_mprintf("%zNOT ", zBuf);
+ break;
+ case FTSQUERY_AND:
+ zBuf = sqlite3_mprintf("%zAND ", zBuf);
+ break;
+ case FTSQUERY_OR:
+ zBuf = sqlite3_mprintf("%zOR ", zBuf);
+ break;
+ }
+
+ if( zBuf ) zBuf = sqlite3_mprintf("%z{", zBuf);
+ if( zBuf ) zBuf = exprToString(pExpr->pLeft, zBuf);
+ if( zBuf ) zBuf = sqlite3_mprintf("%z} {", zBuf);
+
+ if( zBuf ) zBuf = exprToString(pExpr->pRight, zBuf);
+ if( zBuf ) zBuf = sqlite3_mprintf("%z}", zBuf);
+
+ return zBuf;
+}
+
+/*
+** This is the implementation of a scalar SQL function used to test the
+** expression parser. It should be called as follows:
+**
+** fts3_exprtest(, , , ...);
+**
+** The first argument, , is the name of the fts3 tokenizer used
+** to parse the query expression (see README.tokenizers). The second argument
+** is the query expression to parse. Each subsequent argument is the name
+** of a column of the fts3 table that the query expression may refer to.
+** For example:
+**
+** SELECT fts3_exprtest('simple', 'Bill col2:Bloggs', 'col1', 'col2');
+*/
+static void fts3ExprTestCommon(
+ int bRebalance,
+ sqlite3_context *context,
+ int argc,
+ sqlite3_value **argv
+){
+ sqlite3_tokenizer *pTokenizer = 0;
+ int rc;
+ char **azCol = 0;
+ const char *zExpr;
+ int nExpr;
+ int nCol;
+ int ii;
+ Fts3Expr *pExpr;
+ char *zBuf = 0;
+ Fts3Hash *pHash = (Fts3Hash*)sqlite3_user_data(context);
+ const char *zTokenizer = 0;
+ char *zErr = 0;
+
+ if( argc<3 ){
+ sqlite3_result_error(context,
+ "Usage: fts3_exprtest(tokenizer, expr, col1, ...", -1
+ );
+ return;
+ }
+
+ zTokenizer = (const char*)sqlite3_value_text(argv[0]);
+ rc = sqlite3Fts3InitTokenizer(pHash, zTokenizer, &pTokenizer, &zErr);
+ if( rc!=SQLITE_OK ){
+ if( rc==SQLITE_NOMEM ){
+ sqlite3_result_error_nomem(context);
+ }else{
+ sqlite3_result_error(context, zErr, -1);
+ }
+ sqlite3_free(zErr);
+ return;
+ }
+
+ zExpr = (const char *)sqlite3_value_text(argv[1]);
+ nExpr = sqlite3_value_bytes(argv[1]);
+ nCol = argc-2;
+ azCol = (char **)sqlite3_malloc64(nCol*sizeof(char *));
+ if( !azCol ){
+ sqlite3_result_error_nomem(context);
+ goto exprtest_out;
+ }
+ for(ii=0; iipModule->xDestroy(pTokenizer);
+ }
+ sqlite3_free(azCol);
+}
+
+static void fts3ExprTest(
+ sqlite3_context *context,
+ int argc,
+ sqlite3_value **argv
+){
+ fts3ExprTestCommon(0, context, argc, argv);
+}
+static void fts3ExprTestRebalance(
+ sqlite3_context *context,
+ int argc,
+ sqlite3_value **argv
+){
+ fts3ExprTestCommon(1, context, argc, argv);
+}
+
+/*
+** Register the query expression parser test function fts3_exprtest()
+** with database connection db.
+*/
+int sqlite3Fts3ExprInitTestInterface(sqlite3 *db, Fts3Hash *pHash){
+ int rc = sqlite3_create_function(
+ db, "fts3_exprtest", -1, SQLITE_UTF8, (void*)pHash, fts3ExprTest, 0, 0
+ );
+ if( rc==SQLITE_OK ){
+ rc = sqlite3_create_function(db, "fts3_exprtest_rebalance",
+ -1, SQLITE_UTF8, (void*)pHash, fts3ExprTestRebalance, 0, 0
+ );
+ }
+ return rc;
+}
+
+#endif
+#endif /* !defined(SQLITE_CORE) || defined(SQLITE_ENABLE_FTS3) */
diff --git a/ext/fts3/fts3_hash.c b/ext/fts3/fts3_hash.c
new file mode 100644
index 0000000000000000000000000000000000000000..63e55b3dc9ac3d152e104af29b95b9e179e1ce48
--- /dev/null
+++ b/ext/fts3/fts3_hash.c
@@ -0,0 +1,383 @@
+/*
+** 2001 September 22
+**
+** The author disclaims copyright to this source code. In place of
+** a legal notice, here is a blessing:
+**
+** May you do good and not evil.
+** May you find forgiveness for yourself and forgive others.
+** May you share freely, never taking more than you give.
+**
+*************************************************************************
+** This is the implementation of generic hash-tables used in SQLite.
+** We've modified it slightly to serve as a standalone hash table
+** implementation for the full-text indexing module.
+*/
+
+/*
+** The code in this file is only compiled if:
+**
+** * The FTS3 module is being built as an extension
+** (in which case SQLITE_CORE is not defined), or
+**
+** * The FTS3 module is being built into the core of
+** SQLite (in which case SQLITE_ENABLE_FTS3 is defined).
+*/
+#include "fts3Int.h"
+#if !defined(SQLITE_CORE) || defined(SQLITE_ENABLE_FTS3)
+
+#include
+#include
+#include
+
+#include "fts3_hash.h"
+
+/*
+** Malloc and Free functions
+*/
+static void *fts3HashMalloc(sqlite3_int64 n){
+ void *p = sqlite3_malloc64(n);
+ if( p ){
+ memset(p, 0, n);
+ }
+ return p;
+}
+static void fts3HashFree(void *p){
+ sqlite3_free(p);
+}
+
+/* Turn bulk memory into a hash table object by initializing the
+** fields of the Hash structure.
+**
+** "pNew" is a pointer to the hash table that is to be initialized.
+** keyClass is one of the constants
+** FTS3_HASH_BINARY or FTS3_HASH_STRING. The value of keyClass
+** determines what kind of key the hash table will use. "copyKey" is
+** true if the hash table should make its own private copy of keys and
+** false if it should just use the supplied pointer.
+*/
+void sqlite3Fts3HashInit(Fts3Hash *pNew, char keyClass, char copyKey){
+ assert( pNew!=0 );
+ assert( keyClass>=FTS3_HASH_STRING && keyClass<=FTS3_HASH_BINARY );
+ pNew->keyClass = keyClass;
+ pNew->copyKey = copyKey;
+ pNew->first = 0;
+ pNew->count = 0;
+ pNew->htsize = 0;
+ pNew->ht = 0;
+}
+
+/* Remove all entries from a hash table. Reclaim all memory.
+** Call this routine to delete a hash table or to reset a hash table
+** to the empty state.
+*/
+void sqlite3Fts3HashClear(Fts3Hash *pH){
+ Fts3HashElem *elem; /* For looping over all elements of the table */
+
+ assert( pH!=0 );
+ elem = pH->first;
+ pH->first = 0;
+ fts3HashFree(pH->ht);
+ pH->ht = 0;
+ pH->htsize = 0;
+ while( elem ){
+ Fts3HashElem *next_elem = elem->next;
+ if( pH->copyKey && elem->pKey ){
+ fts3HashFree(elem->pKey);
+ }
+ fts3HashFree(elem);
+ elem = next_elem;
+ }
+ pH->count = 0;
+}
+
+/*
+** Hash and comparison functions when the mode is FTS3_HASH_STRING
+*/
+static int fts3StrHash(const void *pKey, int nKey){
+ const char *z = (const char *)pKey;
+ unsigned h = 0;
+ if( nKey<=0 ) nKey = (int) strlen(z);
+ while( nKey > 0 ){
+ h = (h<<3) ^ h ^ *z++;
+ nKey--;
+ }
+ return (int)(h & 0x7fffffff);
+}
+static int fts3StrCompare(const void *pKey1, int n1, const void *pKey2, int n2){
+ if( n1!=n2 ) return 1;
+ return strncmp((const char*)pKey1,(const char*)pKey2,n1);
+}
+
+/*
+** Hash and comparison functions when the mode is FTS3_HASH_BINARY
+*/
+static int fts3BinHash(const void *pKey, int nKey){
+ int h = 0;
+ const char *z = (const char *)pKey;
+ while( nKey-- > 0 ){
+ h = (h<<3) ^ h ^ *(z++);
+ }
+ return h & 0x7fffffff;
+}
+static int fts3BinCompare(const void *pKey1, int n1, const void *pKey2, int n2){
+ if( n1!=n2 ) return 1;
+ return memcmp(pKey1,pKey2,n1);
+}
+
+/*
+** Return a pointer to the appropriate hash function given the key class.
+**
+** The C syntax in this function definition may be unfamilar to some
+** programmers, so we provide the following additional explanation:
+**
+** The name of the function is "ftsHashFunction". The function takes a
+** single parameter "keyClass". The return value of ftsHashFunction()
+** is a pointer to another function. Specifically, the return value
+** of ftsHashFunction() is a pointer to a function that takes two parameters
+** with types "const void*" and "int" and returns an "int".
+*/
+static int (*ftsHashFunction(int keyClass))(const void*,int){
+ if( keyClass==FTS3_HASH_STRING ){
+ return &fts3StrHash;
+ }else{
+ assert( keyClass==FTS3_HASH_BINARY );
+ return &fts3BinHash;
+ }
+}
+
+/*
+** Return a pointer to the appropriate hash function given the key class.
+**
+** For help in interpreted the obscure C code in the function definition,
+** see the header comment on the previous function.
+*/
+static int (*ftsCompareFunction(int keyClass))(const void*,int,const void*,int){
+ if( keyClass==FTS3_HASH_STRING ){
+ return &fts3StrCompare;
+ }else{
+ assert( keyClass==FTS3_HASH_BINARY );
+ return &fts3BinCompare;
+ }
+}
+
+/* Link an element into the hash table
+*/
+static void fts3HashInsertElement(
+ Fts3Hash *pH, /* The complete hash table */
+ struct _fts3ht *pEntry, /* The entry into which pNew is inserted */
+ Fts3HashElem *pNew /* The element to be inserted */
+){
+ Fts3HashElem *pHead; /* First element already in pEntry */
+ pHead = pEntry->chain;
+ if( pHead ){
+ pNew->next = pHead;
+ pNew->prev = pHead->prev;
+ if( pHead->prev ){ pHead->prev->next = pNew; }
+ else { pH->first = pNew; }
+ pHead->prev = pNew;
+ }else{
+ pNew->next = pH->first;
+ if( pH->first ){ pH->first->prev = pNew; }
+ pNew->prev = 0;
+ pH->first = pNew;
+ }
+ pEntry->count++;
+ pEntry->chain = pNew;
+}
+
+
+/* Resize the hash table so that it cantains "new_size" buckets.
+** "new_size" must be a power of 2. The hash table might fail
+** to resize if sqliteMalloc() fails.
+**
+** Return non-zero if a memory allocation error occurs.
+*/
+static int fts3Rehash(Fts3Hash *pH, int new_size){
+ struct _fts3ht *new_ht; /* The new hash table */
+ Fts3HashElem *elem, *next_elem; /* For looping over existing elements */
+ int (*xHash)(const void*,int); /* The hash function */
+
+ assert( (new_size & (new_size-1))==0 );
+ new_ht = (struct _fts3ht *)fts3HashMalloc( new_size*sizeof(struct _fts3ht) );
+ if( new_ht==0 ) return 1;
+ fts3HashFree(pH->ht);
+ pH->ht = new_ht;
+ pH->htsize = new_size;
+ xHash = ftsHashFunction(pH->keyClass);
+ for(elem=pH->first, pH->first=0; elem; elem = next_elem){
+ int h = (*xHash)(elem->pKey, elem->nKey) & (new_size-1);
+ next_elem = elem->next;
+ fts3HashInsertElement(pH, &new_ht[h], elem);
+ }
+ return 0;
+}
+
+/* This function (for internal use only) locates an element in an
+** hash table that matches the given key. The hash for this key has
+** already been computed and is passed as the 4th parameter.
+*/
+static Fts3HashElem *fts3FindElementByHash(
+ const Fts3Hash *pH, /* The pH to be searched */
+ const void *pKey, /* The key we are searching for */
+ int nKey,
+ int h /* The hash for this key. */
+){
+ Fts3HashElem *elem; /* Used to loop thru the element list */
+ int count; /* Number of elements left to test */
+ int (*xCompare)(const void*,int,const void*,int); /* comparison function */
+
+ if( pH->ht ){
+ struct _fts3ht *pEntry = &pH->ht[h];
+ elem = pEntry->chain;
+ count = pEntry->count;
+ xCompare = ftsCompareFunction(pH->keyClass);
+ while( count-- && elem ){
+ if( (*xCompare)(elem->pKey,elem->nKey,pKey,nKey)==0 ){
+ return elem;
+ }
+ elem = elem->next;
+ }
+ }
+ return 0;
+}
+
+/* Remove a single entry from the hash table given a pointer to that
+** element and a hash on the element's key.
+*/
+static void fts3RemoveElementByHash(
+ Fts3Hash *pH, /* The pH containing "elem" */
+ Fts3HashElem* elem, /* The element to be removed from the pH */
+ int h /* Hash value for the element */
+){
+ struct _fts3ht *pEntry;
+ if( elem->prev ){
+ elem->prev->next = elem->next;
+ }else{
+ pH->first = elem->next;
+ }
+ if( elem->next ){
+ elem->next->prev = elem->prev;
+ }
+ pEntry = &pH->ht[h];
+ if( pEntry->chain==elem ){
+ pEntry->chain = elem->next;
+ }
+ pEntry->count--;
+ if( pEntry->count<=0 ){
+ pEntry->chain = 0;
+ }
+ if( pH->copyKey && elem->pKey ){
+ fts3HashFree(elem->pKey);
+ }
+ fts3HashFree( elem );
+ pH->count--;
+ if( pH->count<=0 ){
+ assert( pH->first==0 );
+ assert( pH->count==0 );
+ fts3HashClear(pH);
+ }
+}
+
+Fts3HashElem *sqlite3Fts3HashFindElem(
+ const Fts3Hash *pH,
+ const void *pKey,
+ int nKey
+){
+ int h; /* A hash on key */
+ int (*xHash)(const void*,int); /* The hash function */
+
+ if( pH==0 || pH->ht==0 ) return 0;
+ xHash = ftsHashFunction(pH->keyClass);
+ assert( xHash!=0 );
+ h = (*xHash)(pKey,nKey);
+ assert( (pH->htsize & (pH->htsize-1))==0 );
+ return fts3FindElementByHash(pH,pKey,nKey, h & (pH->htsize-1));
+}
+
+/*
+** Attempt to locate an element of the hash table pH with a key
+** that matches pKey,nKey. Return the data for this element if it is
+** found, or NULL if there is no match.
+*/
+void *sqlite3Fts3HashFind(const Fts3Hash *pH, const void *pKey, int nKey){
+ Fts3HashElem *pElem; /* The element that matches key (if any) */
+
+ pElem = sqlite3Fts3HashFindElem(pH, pKey, nKey);
+ return pElem ? pElem->data : 0;
+}
+
+/* Insert an element into the hash table pH. The key is pKey,nKey
+** and the data is "data".
+**
+** If no element exists with a matching key, then a new
+** element is created. A copy of the key is made if the copyKey
+** flag is set. NULL is returned.
+**
+** If another element already exists with the same key, then the
+** new data replaces the old data and the old data is returned.
+** The key is not copied in this instance. If a malloc fails, then
+** the new data is returned and the hash table is unchanged.
+**
+** If the "data" parameter to this function is NULL, then the
+** element corresponding to "key" is removed from the hash table.
+*/
+void *sqlite3Fts3HashInsert(
+ Fts3Hash *pH, /* The hash table to insert into */
+ const void *pKey, /* The key */
+ int nKey, /* Number of bytes in the key */
+ void *data /* The data */
+){
+ int hraw; /* Raw hash value of the key */
+ int h; /* the hash of the key modulo hash table size */
+ Fts3HashElem *elem; /* Used to loop thru the element list */
+ Fts3HashElem *new_elem; /* New element added to the pH */
+ int (*xHash)(const void*,int); /* The hash function */
+
+ assert( pH!=0 );
+ xHash = ftsHashFunction(pH->keyClass);
+ assert( xHash!=0 );
+ hraw = (*xHash)(pKey, nKey);
+ assert( (pH->htsize & (pH->htsize-1))==0 );
+ h = hraw & (pH->htsize-1);
+ elem = fts3FindElementByHash(pH,pKey,nKey,h);
+ if( elem ){
+ void *old_data = elem->data;
+ if( data==0 ){
+ fts3RemoveElementByHash(pH,elem,h);
+ }else{
+ elem->data = data;
+ }
+ return old_data;
+ }
+ if( data==0 ) return 0;
+ if( (pH->htsize==0 && fts3Rehash(pH,8))
+ || (pH->count>=pH->htsize && fts3Rehash(pH, pH->htsize*2))
+ ){
+ pH->count = 0;
+ return data;
+ }
+ assert( pH->htsize>0 );
+ new_elem = (Fts3HashElem*)fts3HashMalloc( sizeof(Fts3HashElem) );
+ if( new_elem==0 ) return data;
+ if( pH->copyKey && pKey!=0 ){
+ new_elem->pKey = fts3HashMalloc( nKey );
+ if( new_elem->pKey==0 ){
+ fts3HashFree(new_elem);
+ return data;
+ }
+ memcpy((void*)new_elem->pKey, pKey, nKey);
+ }else{
+ new_elem->pKey = (void*)pKey;
+ }
+ new_elem->nKey = nKey;
+ pH->count++;
+ assert( pH->htsize>0 );
+ assert( (pH->htsize & (pH->htsize-1))==0 );
+ h = hraw & (pH->htsize-1);
+ fts3HashInsertElement(pH, &pH->ht[h], new_elem);
+ new_elem->data = data;
+ return 0;
+}
+
+#endif /* !defined(SQLITE_CORE) || defined(SQLITE_ENABLE_FTS3) */
diff --git a/ext/fts3/fts3_hash.h b/ext/fts3/fts3_hash.h
new file mode 100644
index 0000000000000000000000000000000000000000..dc3fcf8334fb26962bc887fb1fc30e21c543185c
--- /dev/null
+++ b/ext/fts3/fts3_hash.h
@@ -0,0 +1,112 @@
+/*
+** 2001 September 22
+**
+** The author disclaims copyright to this source code. In place of
+** a legal notice, here is a blessing:
+**
+** May you do good and not evil.
+** May you find forgiveness for yourself and forgive others.
+** May you share freely, never taking more than you give.
+**
+*************************************************************************
+** This is the header file for the generic hash-table implementation
+** used in SQLite. We've modified it slightly to serve as a standalone
+** hash table implementation for the full-text indexing module.
+**
+*/
+#ifndef _FTS3_HASH_H_
+#define _FTS3_HASH_H_
+
+/* Forward declarations of structures. */
+typedef struct Fts3Hash Fts3Hash;
+typedef struct Fts3HashElem Fts3HashElem;
+
+/* A complete hash table is an instance of the following structure.
+** The internals of this structure are intended to be opaque -- client
+** code should not attempt to access or modify the fields of this structure
+** directly. Change this structure only by using the routines below.
+** However, many of the "procedures" and "functions" for modifying and
+** accessing this structure are really macros, so we can't really make
+** this structure opaque.
+*/
+struct Fts3Hash {
+ char keyClass; /* HASH_INT, _POINTER, _STRING, _BINARY */
+ char copyKey; /* True if copy of key made on insert */
+ int count; /* Number of entries in this table */
+ Fts3HashElem *first; /* The first element of the array */
+ int htsize; /* Number of buckets in the hash table */
+ struct _fts3ht { /* the hash table */
+ int count; /* Number of entries with this hash */
+ Fts3HashElem *chain; /* Pointer to first entry with this hash */
+ } *ht;
+};
+
+/* Each element in the hash table is an instance of the following
+** structure. All elements are stored on a single doubly-linked list.
+**
+** Again, this structure is intended to be opaque, but it can't really
+** be opaque because it is used by macros.
+*/
+struct Fts3HashElem {
+ Fts3HashElem *next, *prev; /* Next and previous elements in the table */
+ void *data; /* Data associated with this element */
+ void *pKey; int nKey; /* Key associated with this element */
+};
+
+/*
+** There are 2 different modes of operation for a hash table:
+**
+** FTS3_HASH_STRING pKey points to a string that is nKey bytes long
+** (including the null-terminator, if any). Case
+** is respected in comparisons.
+**
+** FTS3_HASH_BINARY pKey points to binary data nKey bytes long.
+** memcmp() is used to compare keys.
+**
+** A copy of the key is made if the copyKey parameter to fts3HashInit is 1.
+*/
+#define FTS3_HASH_STRING 1
+#define FTS3_HASH_BINARY 2
+
+/*
+** Access routines. To delete, insert a NULL pointer.
+*/
+void sqlite3Fts3HashInit(Fts3Hash *pNew, char keyClass, char copyKey);
+void *sqlite3Fts3HashInsert(Fts3Hash*, const void *pKey, int nKey, void *pData);
+void *sqlite3Fts3HashFind(const Fts3Hash*, const void *pKey, int nKey);
+void sqlite3Fts3HashClear(Fts3Hash*);
+Fts3HashElem *sqlite3Fts3HashFindElem(const Fts3Hash *, const void *, int);
+
+/*
+** Shorthand for the functions above
+*/
+#define fts3HashInit sqlite3Fts3HashInit
+#define fts3HashInsert sqlite3Fts3HashInsert
+#define fts3HashFind sqlite3Fts3HashFind
+#define fts3HashClear sqlite3Fts3HashClear
+#define fts3HashFindElem sqlite3Fts3HashFindElem
+
+/*
+** Macros for looping over all elements of a hash table. The idiom is
+** like this:
+**
+** Fts3Hash h;
+** Fts3HashElem *p;
+** ...
+** for(p=fts3HashFirst(&h); p; p=fts3HashNext(p)){
+** SomeStructure *pData = fts3HashData(p);
+** // do something with pData
+** }
+*/
+#define fts3HashFirst(H) ((H)->first)
+#define fts3HashNext(E) ((E)->next)
+#define fts3HashData(E) ((E)->data)
+#define fts3HashKey(E) ((E)->pKey)
+#define fts3HashKeysize(E) ((E)->nKey)
+
+/*
+** Number of entries in a hash table
+*/
+#define fts3HashCount(H) ((H)->count)
+
+#endif /* _FTS3_HASH_H_ */
diff --git a/ext/fts3/fts3_icu.c b/ext/fts3/fts3_icu.c
new file mode 100644
index 0000000000000000000000000000000000000000..0848a5aabaeb642279d24d8d8dbeb5a38fda131f
--- /dev/null
+++ b/ext/fts3/fts3_icu.c
@@ -0,0 +1,262 @@
+/*
+** 2007 June 22
+**
+** The author disclaims copyright to this source code. In place of
+** a legal notice, here is a blessing:
+**
+** May you do good and not evil.
+** May you find forgiveness for yourself and forgive others.
+** May you share freely, never taking more than you give.
+**
+*************************************************************************
+** This file implements a tokenizer for fts3 based on the ICU library.
+*/
+#include "fts3Int.h"
+#if !defined(SQLITE_CORE) || defined(SQLITE_ENABLE_FTS3)
+#ifdef SQLITE_ENABLE_ICU
+
+#include
+#include
+#include "fts3_tokenizer.h"
+
+#include
+#include
+#include
+#include
+
+typedef struct IcuTokenizer IcuTokenizer;
+typedef struct IcuCursor IcuCursor;
+
+struct IcuTokenizer {
+ sqlite3_tokenizer base;
+ char *zLocale;
+};
+
+struct IcuCursor {
+ sqlite3_tokenizer_cursor base;
+
+ UBreakIterator *pIter; /* ICU break-iterator object */
+ int nChar; /* Number of UChar elements in pInput */
+ UChar *aChar; /* Copy of input using utf-16 encoding */
+ int *aOffset; /* Offsets of each character in utf-8 input */
+
+ int nBuffer;
+ char *zBuffer;
+
+ int iToken;
+};
+
+/*
+** Create a new tokenizer instance.
+*/
+static int icuCreate(
+ int argc, /* Number of entries in argv[] */
+ const char * const *argv, /* Tokenizer creation arguments */
+ sqlite3_tokenizer **ppTokenizer /* OUT: Created tokenizer */
+){
+ IcuTokenizer *p;
+ int n = 0;
+
+ if( argc>0 ){
+ n = strlen(argv[0])+1;
+ }
+ p = (IcuTokenizer *)sqlite3_malloc64(sizeof(IcuTokenizer)+n);
+ if( !p ){
+ return SQLITE_NOMEM;
+ }
+ memset(p, 0, sizeof(IcuTokenizer));
+
+ if( n ){
+ p->zLocale = (char *)&p[1];
+ memcpy(p->zLocale, argv[0], n);
+ }
+
+ *ppTokenizer = (sqlite3_tokenizer *)p;
+
+ return SQLITE_OK;
+}
+
+/*
+** Destroy a tokenizer
+*/
+static int icuDestroy(sqlite3_tokenizer *pTokenizer){
+ IcuTokenizer *p = (IcuTokenizer *)pTokenizer;
+ sqlite3_free(p);
+ return SQLITE_OK;
+}
+
+/*
+** Prepare to begin tokenizing a particular string. The input
+** string to be tokenized is pInput[0..nBytes-1]. A cursor
+** used to incrementally tokenize this string is returned in
+** *ppCursor.
+*/
+static int icuOpen(
+ sqlite3_tokenizer *pTokenizer, /* The tokenizer */
+ const char *zInput, /* Input string */
+ int nInput, /* Length of zInput in bytes */
+ sqlite3_tokenizer_cursor **ppCursor /* OUT: Tokenization cursor */
+){
+ IcuTokenizer *p = (IcuTokenizer *)pTokenizer;
+ IcuCursor *pCsr;
+
+ const int32_t opt = U_FOLD_CASE_DEFAULT;
+ UErrorCode status = U_ZERO_ERROR;
+ int nChar;
+
+ UChar32 c;
+ int iInput = 0;
+ int iOut = 0;
+
+ *ppCursor = 0;
+
+ if( zInput==0 ){
+ nInput = 0;
+ zInput = "";
+ }else if( nInput<0 ){
+ nInput = strlen(zInput);
+ }
+ nChar = nInput+1;
+ pCsr = (IcuCursor *)sqlite3_malloc64(
+ sizeof(IcuCursor) + /* IcuCursor */
+ ((nChar+3)&~3) * sizeof(UChar) + /* IcuCursor.aChar[] */
+ (nChar+1) * sizeof(int) /* IcuCursor.aOffset[] */
+ );
+ if( !pCsr ){
+ return SQLITE_NOMEM;
+ }
+ memset(pCsr, 0, sizeof(IcuCursor));
+ pCsr->aChar = (UChar *)&pCsr[1];
+ pCsr->aOffset = (int *)&pCsr->aChar[(nChar+3)&~3];
+
+ pCsr->aOffset[iOut] = iInput;
+ U8_NEXT(zInput, iInput, nInput, c);
+ while( c>0 ){
+ int isError = 0;
+ c = u_foldCase(c, opt);
+ U16_APPEND(pCsr->aChar, iOut, nChar, c, isError);
+ if( isError ){
+ sqlite3_free(pCsr);
+ return SQLITE_ERROR;
+ }
+ pCsr->aOffset[iOut] = iInput;
+
+ if( iInputpIter = ubrk_open(UBRK_WORD, p->zLocale, pCsr->aChar, iOut, &status);
+ if( !U_SUCCESS(status) ){
+ sqlite3_free(pCsr);
+ return SQLITE_ERROR;
+ }
+ pCsr->nChar = iOut;
+
+ ubrk_first(pCsr->pIter);
+ *ppCursor = (sqlite3_tokenizer_cursor *)pCsr;
+ return SQLITE_OK;
+}
+
+/*
+** Close a tokenization cursor previously opened by a call to icuOpen().
+*/
+static int icuClose(sqlite3_tokenizer_cursor *pCursor){
+ IcuCursor *pCsr = (IcuCursor *)pCursor;
+ ubrk_close(pCsr->pIter);
+ sqlite3_free(pCsr->zBuffer);
+ sqlite3_free(pCsr);
+ return SQLITE_OK;
+}
+
+/*
+** Extract the next token from a tokenization cursor.
+*/
+static int icuNext(
+ sqlite3_tokenizer_cursor *pCursor, /* Cursor returned by simpleOpen */
+ const char **ppToken, /* OUT: *ppToken is the token text */
+ int *pnBytes, /* OUT: Number of bytes in token */
+ int *piStartOffset, /* OUT: Starting offset of token */
+ int *piEndOffset, /* OUT: Ending offset of token */
+ int *piPosition /* OUT: Position integer of token */
+){
+ IcuCursor *pCsr = (IcuCursor *)pCursor;
+
+ int iStart = 0;
+ int iEnd = 0;
+ int nByte = 0;
+
+ while( iStart==iEnd ){
+ UChar32 c;
+
+ iStart = ubrk_current(pCsr->pIter);
+ iEnd = ubrk_next(pCsr->pIter);
+ if( iEnd==UBRK_DONE ){
+ return SQLITE_DONE;
+ }
+
+ while( iStartaChar, iWhite, pCsr->nChar, c);
+ if( u_isspace(c) ){
+ iStart = iWhite;
+ }else{
+ break;
+ }
+ }
+ assert(iStart<=iEnd);
+ }
+
+ do {
+ UErrorCode status = U_ZERO_ERROR;
+ if( nByte ){
+ char *zNew = sqlite3_realloc(pCsr->zBuffer, nByte);
+ if( !zNew ){
+ return SQLITE_NOMEM;
+ }
+ pCsr->zBuffer = zNew;
+ pCsr->nBuffer = nByte;
+ }
+
+ u_strToUTF8(
+ pCsr->zBuffer, pCsr->nBuffer, &nByte, /* Output vars */
+ &pCsr->aChar[iStart], iEnd-iStart, /* Input vars */
+ &status /* Output success/failure */
+ );
+ } while( nByte>pCsr->nBuffer );
+
+ *ppToken = pCsr->zBuffer;
+ *pnBytes = nByte;
+ *piStartOffset = pCsr->aOffset[iStart];
+ *piEndOffset = pCsr->aOffset[iEnd];
+ *piPosition = pCsr->iToken++;
+
+ return SQLITE_OK;
+}
+
+/*
+** The set of routines that implement the simple tokenizer
+*/
+static const sqlite3_tokenizer_module icuTokenizerModule = {
+ 0, /* iVersion */
+ icuCreate, /* xCreate */
+ icuDestroy, /* xCreate */
+ icuOpen, /* xOpen */
+ icuClose, /* xClose */
+ icuNext, /* xNext */
+ 0, /* xLanguageid */
+};
+
+/*
+** Set *ppModule to point at the implementation of the ICU tokenizer.
+*/
+void sqlite3Fts3IcuTokenizerModule(
+ sqlite3_tokenizer_module const**ppModule
+){
+ *ppModule = &icuTokenizerModule;
+}
+
+#endif /* defined(SQLITE_ENABLE_ICU) */
+#endif /* !defined(SQLITE_CORE) || defined(SQLITE_ENABLE_FTS3) */
diff --git a/ext/fts3/fts3_porter.c b/ext/fts3/fts3_porter.c
new file mode 100644
index 0000000000000000000000000000000000000000..fbe7913020a4e975ce997b02a619ca589dc1dc1d
--- /dev/null
+++ b/ext/fts3/fts3_porter.c
@@ -0,0 +1,662 @@
+/*
+** 2006 September 30
+**
+** The author disclaims copyright to this source code. In place of
+** a legal notice, here is a blessing:
+**
+** May you do good and not evil.
+** May you find forgiveness for yourself and forgive others.
+** May you share freely, never taking more than you give.
+**
+*************************************************************************
+** Implementation of the full-text-search tokenizer that implements
+** a Porter stemmer.
+*/
+
+/*
+** The code in this file is only compiled if:
+**
+** * The FTS3 module is being built as an extension
+** (in which case SQLITE_CORE is not defined), or
+**
+** * The FTS3 module is being built into the core of
+** SQLite (in which case SQLITE_ENABLE_FTS3 is defined).
+*/
+#include "fts3Int.h"
+#if !defined(SQLITE_CORE) || defined(SQLITE_ENABLE_FTS3)
+
+#include
+#include
+#include
+#include
+
+#include "fts3_tokenizer.h"
+
+/*
+** Class derived from sqlite3_tokenizer
+*/
+typedef struct porter_tokenizer {
+ sqlite3_tokenizer base; /* Base class */
+} porter_tokenizer;
+
+/*
+** Class derived from sqlite3_tokenizer_cursor
+*/
+typedef struct porter_tokenizer_cursor {
+ sqlite3_tokenizer_cursor base;
+ const char *zInput; /* input we are tokenizing */
+ int nInput; /* size of the input */
+ int iOffset; /* current position in zInput */
+ int iToken; /* index of next token to be returned */
+ char *zToken; /* storage for current token */
+ int nAllocated; /* space allocated to zToken buffer */
+} porter_tokenizer_cursor;
+
+
+/*
+** Create a new tokenizer instance.
+*/
+static int porterCreate(
+ int argc, const char * const *argv,
+ sqlite3_tokenizer **ppTokenizer
+){
+ porter_tokenizer *t;
+
+ UNUSED_PARAMETER(argc);
+ UNUSED_PARAMETER(argv);
+
+ t = (porter_tokenizer *) sqlite3_malloc(sizeof(*t));
+ if( t==NULL ) return SQLITE_NOMEM;
+ memset(t, 0, sizeof(*t));
+ *ppTokenizer = &t->base;
+ return SQLITE_OK;
+}
+
+/*
+** Destroy a tokenizer
+*/
+static int porterDestroy(sqlite3_tokenizer *pTokenizer){
+ sqlite3_free(pTokenizer);
+ return SQLITE_OK;
+}
+
+/*
+** Prepare to begin tokenizing a particular string. The input
+** string to be tokenized is zInput[0..nInput-1]. A cursor
+** used to incrementally tokenize this string is returned in
+** *ppCursor.
+*/
+static int porterOpen(
+ sqlite3_tokenizer *pTokenizer, /* The tokenizer */
+ const char *zInput, int nInput, /* String to be tokenized */
+ sqlite3_tokenizer_cursor **ppCursor /* OUT: Tokenization cursor */
+){
+ porter_tokenizer_cursor *c;
+
+ UNUSED_PARAMETER(pTokenizer);
+
+ c = (porter_tokenizer_cursor *) sqlite3_malloc(sizeof(*c));
+ if( c==NULL ) return SQLITE_NOMEM;
+
+ c->zInput = zInput;
+ if( zInput==0 ){
+ c->nInput = 0;
+ }else if( nInput<0 ){
+ c->nInput = (int)strlen(zInput);
+ }else{
+ c->nInput = nInput;
+ }
+ c->iOffset = 0; /* start tokenizing at the beginning */
+ c->iToken = 0;
+ c->zToken = NULL; /* no space allocated, yet. */
+ c->nAllocated = 0;
+
+ *ppCursor = &c->base;
+ return SQLITE_OK;
+}
+
+/*
+** Close a tokenization cursor previously opened by a call to
+** porterOpen() above.
+*/
+static int porterClose(sqlite3_tokenizer_cursor *pCursor){
+ porter_tokenizer_cursor *c = (porter_tokenizer_cursor *) pCursor;
+ sqlite3_free(c->zToken);
+ sqlite3_free(c);
+ return SQLITE_OK;
+}
+/*
+** Vowel or consonant
+*/
+static const char cType[] = {
+ 0, 1, 1, 1, 0, 1, 1, 1, 0, 1, 1, 1, 1, 1, 0, 1, 1, 1, 1, 1, 0,
+ 1, 1, 1, 2, 1
+};
+
+/*
+** isConsonant() and isVowel() determine if their first character in
+** the string they point to is a consonant or a vowel, according
+** to Porter ruls.
+**
+** A consonate is any letter other than 'a', 'e', 'i', 'o', or 'u'.
+** 'Y' is a consonant unless it follows another consonant,
+** in which case it is a vowel.
+**
+** In these routine, the letters are in reverse order. So the 'y' rule
+** is that 'y' is a consonant unless it is followed by another
+** consonent.
+*/
+static int isVowel(const char*);
+static int isConsonant(const char *z){
+ int j;
+ char x = *z;
+ if( x==0 ) return 0;
+ assert( x>='a' && x<='z' );
+ j = cType[x-'a'];
+ if( j<2 ) return j;
+ return z[1]==0 || isVowel(z + 1);
+}
+static int isVowel(const char *z){
+ int j;
+ char x = *z;
+ if( x==0 ) return 0;
+ assert( x>='a' && x<='z' );
+ j = cType[x-'a'];
+ if( j<2 ) return 1-j;
+ return isConsonant(z + 1);
+}
+
+/*
+** Let any sequence of one or more vowels be represented by V and let
+** C be sequence of one or more consonants. Then every word can be
+** represented as:
+**
+** [C] (VC){m} [V]
+**
+** In prose: A word is an optional consonant followed by zero or
+** vowel-consonant pairs followed by an optional vowel. "m" is the
+** number of vowel consonant pairs. This routine computes the value
+** of m for the first i bytes of a word.
+**
+** Return true if the m-value for z is 1 or more. In other words,
+** return true if z contains at least one vowel that is followed
+** by a consonant.
+**
+** In this routine z[] is in reverse order. So we are really looking
+** for an instance of a consonant followed by a vowel.
+*/
+static int m_gt_0(const char *z){
+ while( isVowel(z) ){ z++; }
+ if( *z==0 ) return 0;
+ while( isConsonant(z) ){ z++; }
+ return *z!=0;
+}
+
+/* Like mgt0 above except we are looking for a value of m which is
+** exactly 1
+*/
+static int m_eq_1(const char *z){
+ while( isVowel(z) ){ z++; }
+ if( *z==0 ) return 0;
+ while( isConsonant(z) ){ z++; }
+ if( *z==0 ) return 0;
+ while( isVowel(z) ){ z++; }
+ if( *z==0 ) return 1;
+ while( isConsonant(z) ){ z++; }
+ return *z==0;
+}
+
+/* Like mgt0 above except we are looking for a value of m>1 instead
+** or m>0
+*/
+static int m_gt_1(const char *z){
+ while( isVowel(z) ){ z++; }
+ if( *z==0 ) return 0;
+ while( isConsonant(z) ){ z++; }
+ if( *z==0 ) return 0;
+ while( isVowel(z) ){ z++; }
+ if( *z==0 ) return 0;
+ while( isConsonant(z) ){ z++; }
+ return *z!=0;
+}
+
+/*
+** Return TRUE if there is a vowel anywhere within z[0..n-1]
+*/
+static int hasVowel(const char *z){
+ while( isConsonant(z) ){ z++; }
+ return *z!=0;
+}
+
+/*
+** Return TRUE if the word ends in a double consonant.
+**
+** The text is reversed here. So we are really looking at
+** the first two characters of z[].
+*/
+static int doubleConsonant(const char *z){
+ return isConsonant(z) && z[0]==z[1];
+}
+
+/*
+** Return TRUE if the word ends with three letters which
+** are consonant-vowel-consonent and where the final consonant
+** is not 'w', 'x', or 'y'.
+**
+** The word is reversed here. So we are really checking the
+** first three letters and the first one cannot be in [wxy].
+*/
+static int star_oh(const char *z){
+ return
+ isConsonant(z) &&
+ z[0]!='w' && z[0]!='x' && z[0]!='y' &&
+ isVowel(z+1) &&
+ isConsonant(z+2);
+}
+
+/*
+** If the word ends with zFrom and xCond() is true for the stem
+** of the word that preceeds the zFrom ending, then change the
+** ending to zTo.
+**
+** The input word *pz and zFrom are both in reverse order. zTo
+** is in normal order.
+**
+** Return TRUE if zFrom matches. Return FALSE if zFrom does not
+** match. Not that TRUE is returned even if xCond() fails and
+** no substitution occurs.
+*/
+static int stem(
+ char **pz, /* The word being stemmed (Reversed) */
+ const char *zFrom, /* If the ending matches this... (Reversed) */
+ const char *zTo, /* ... change the ending to this (not reversed) */
+ int (*xCond)(const char*) /* Condition that must be true */
+){
+ char *z = *pz;
+ while( *zFrom && *zFrom==*z ){ z++; zFrom++; }
+ if( *zFrom!=0 ) return 0;
+ if( xCond && !xCond(z) ) return 1;
+ while( *zTo ){
+ *(--z) = *(zTo++);
+ }
+ *pz = z;
+ return 1;
+}
+
+/*
+** This is the fallback stemmer used when the porter stemmer is
+** inappropriate. The input word is copied into the output with
+** US-ASCII case folding. If the input word is too long (more
+** than 20 bytes if it contains no digits or more than 6 bytes if
+** it contains digits) then word is truncated to 20 or 6 bytes
+** by taking 10 or 3 bytes from the beginning and end.
+*/
+static void copy_stemmer(const char *zIn, int nIn, char *zOut, int *pnOut){
+ int i, mx, j;
+ int hasDigit = 0;
+ for(i=0; i='A' && c<='Z' ){
+ zOut[i] = c - 'A' + 'a';
+ }else{
+ if( c>='0' && c<='9' ) hasDigit = 1;
+ zOut[i] = c;
+ }
+ }
+ mx = hasDigit ? 3 : 10;
+ if( nIn>mx*2 ){
+ for(j=mx, i=nIn-mx; i=(int)sizeof(zReverse)-7 ){
+ /* The word is too big or too small for the porter stemmer.
+ ** Fallback to the copy stemmer */
+ copy_stemmer(zIn, nIn, zOut, pnOut);
+ return;
+ }
+ for(i=0, j=sizeof(zReverse)-6; i='A' && c<='Z' ){
+ zReverse[j] = c + 'a' - 'A';
+ }else if( c>='a' && c<='z' ){
+ zReverse[j] = c;
+ }else{
+ /* The use of a character not in [a-zA-Z] means that we fallback
+ ** to the copy stemmer */
+ copy_stemmer(zIn, nIn, zOut, pnOut);
+ return;
+ }
+ }
+ memset(&zReverse[sizeof(zReverse)-5], 0, 5);
+ z = &zReverse[j+1];
+
+
+ /* Step 1a */
+ if( z[0]=='s' ){
+ if(
+ !stem(&z, "sess", "ss", 0) &&
+ !stem(&z, "sei", "i", 0) &&
+ !stem(&z, "ss", "ss", 0)
+ ){
+ z++;
+ }
+ }
+
+ /* Step 1b */
+ z2 = z;
+ if( stem(&z, "dee", "ee", m_gt_0) ){
+ /* Do nothing. The work was all in the test */
+ }else if(
+ (stem(&z, "gni", "", hasVowel) || stem(&z, "de", "", hasVowel))
+ && z!=z2
+ ){
+ if( stem(&z, "ta", "ate", 0) ||
+ stem(&z, "lb", "ble", 0) ||
+ stem(&z, "zi", "ize", 0) ){
+ /* Do nothing. The work was all in the test */
+ }else if( doubleConsonant(z) && (*z!='l' && *z!='s' && *z!='z') ){
+ z++;
+ }else if( m_eq_1(z) && star_oh(z) ){
+ *(--z) = 'e';
+ }
+ }
+
+ /* Step 1c */
+ if( z[0]=='y' && hasVowel(z+1) ){
+ z[0] = 'i';
+ }
+
+ /* Step 2 */
+ switch( z[1] ){
+ case 'a':
+ if( !stem(&z, "lanoita", "ate", m_gt_0) ){
+ stem(&z, "lanoit", "tion", m_gt_0);
+ }
+ break;
+ case 'c':
+ if( !stem(&z, "icne", "ence", m_gt_0) ){
+ stem(&z, "icna", "ance", m_gt_0);
+ }
+ break;
+ case 'e':
+ stem(&z, "rezi", "ize", m_gt_0);
+ break;
+ case 'g':
+ stem(&z, "igol", "log", m_gt_0);
+ break;
+ case 'l':
+ if( !stem(&z, "ilb", "ble", m_gt_0)
+ && !stem(&z, "illa", "al", m_gt_0)
+ && !stem(&z, "iltne", "ent", m_gt_0)
+ && !stem(&z, "ile", "e", m_gt_0)
+ ){
+ stem(&z, "ilsuo", "ous", m_gt_0);
+ }
+ break;
+ case 'o':
+ if( !stem(&z, "noitazi", "ize", m_gt_0)
+ && !stem(&z, "noita", "ate", m_gt_0)
+ ){
+ stem(&z, "rota", "ate", m_gt_0);
+ }
+ break;
+ case 's':
+ if( !stem(&z, "msila", "al", m_gt_0)
+ && !stem(&z, "ssenevi", "ive", m_gt_0)
+ && !stem(&z, "ssenluf", "ful", m_gt_0)
+ ){
+ stem(&z, "ssensuo", "ous", m_gt_0);
+ }
+ break;
+ case 't':
+ if( !stem(&z, "itila", "al", m_gt_0)
+ && !stem(&z, "itivi", "ive", m_gt_0)
+ ){
+ stem(&z, "itilib", "ble", m_gt_0);
+ }
+ break;
+ }
+
+ /* Step 3 */
+ switch( z[0] ){
+ case 'e':
+ if( !stem(&z, "etaci", "ic", m_gt_0)
+ && !stem(&z, "evita", "", m_gt_0)
+ ){
+ stem(&z, "ezila", "al", m_gt_0);
+ }
+ break;
+ case 'i':
+ stem(&z, "itici", "ic", m_gt_0);
+ break;
+ case 'l':
+ if( !stem(&z, "laci", "ic", m_gt_0) ){
+ stem(&z, "luf", "", m_gt_0);
+ }
+ break;
+ case 's':
+ stem(&z, "ssen", "", m_gt_0);
+ break;
+ }
+
+ /* Step 4 */
+ switch( z[1] ){
+ case 'a':
+ if( z[0]=='l' && m_gt_1(z+2) ){
+ z += 2;
+ }
+ break;
+ case 'c':
+ if( z[0]=='e' && z[2]=='n' && (z[3]=='a' || z[3]=='e') && m_gt_1(z+4) ){
+ z += 4;
+ }
+ break;
+ case 'e':
+ if( z[0]=='r' && m_gt_1(z+2) ){
+ z += 2;
+ }
+ break;
+ case 'i':
+ if( z[0]=='c' && m_gt_1(z+2) ){
+ z += 2;
+ }
+ break;
+ case 'l':
+ if( z[0]=='e' && z[2]=='b' && (z[3]=='a' || z[3]=='i') && m_gt_1(z+4) ){
+ z += 4;
+ }
+ break;
+ case 'n':
+ if( z[0]=='t' ){
+ if( z[2]=='a' ){
+ if( m_gt_1(z+3) ){
+ z += 3;
+ }
+ }else if( z[2]=='e' ){
+ if( !stem(&z, "tneme", "", m_gt_1)
+ && !stem(&z, "tnem", "", m_gt_1)
+ ){
+ stem(&z, "tne", "", m_gt_1);
+ }
+ }
+ }
+ break;
+ case 'o':
+ if( z[0]=='u' ){
+ if( m_gt_1(z+2) ){
+ z += 2;
+ }
+ }else if( z[3]=='s' || z[3]=='t' ){
+ stem(&z, "noi", "", m_gt_1);
+ }
+ break;
+ case 's':
+ if( z[0]=='m' && z[2]=='i' && m_gt_1(z+3) ){
+ z += 3;
+ }
+ break;
+ case 't':
+ if( !stem(&z, "eta", "", m_gt_1) ){
+ stem(&z, "iti", "", m_gt_1);
+ }
+ break;
+ case 'u':
+ if( z[0]=='s' && z[2]=='o' && m_gt_1(z+3) ){
+ z += 3;
+ }
+ break;
+ case 'v':
+ case 'z':
+ if( z[0]=='e' && z[2]=='i' && m_gt_1(z+3) ){
+ z += 3;
+ }
+ break;
+ }
+
+ /* Step 5a */
+ if( z[0]=='e' ){
+ if( m_gt_1(z+1) ){
+ z++;
+ }else if( m_eq_1(z+1) && !star_oh(z+1) ){
+ z++;
+ }
+ }
+
+ /* Step 5b */
+ if( m_gt_1(z) && z[0]=='l' && z[1]=='l' ){
+ z++;
+ }
+
+ /* z[] is now the stemmed word in reverse order. Flip it back
+ ** around into forward order and return.
+ */
+ *pnOut = i = (int)strlen(z);
+ zOut[i] = 0;
+ while( *z ){
+ zOut[--i] = *(z++);
+ }
+}
+
+/*
+** Characters that can be part of a token. We assume any character
+** whose value is greater than 0x80 (any UTF character) can be
+** part of a token. In other words, delimiters all must have
+** values of 0x7f or lower.
+*/
+static const char porterIdChar[] = {
+/* x0 x1 x2 x3 x4 x5 x6 x7 x8 x9 xA xB xC xD xE xF */
+ 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 0, 0, 0, 0, 0, 0, /* 3x */
+ 0, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, /* 4x */
+ 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 0, 0, 0, 0, 1, /* 5x */
+ 0, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, /* 6x */
+ 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 0, 0, 0, 0, 0, /* 7x */
+};
+#define isDelim(C) (((ch=C)&0x80)==0 && (ch<0x30 || !porterIdChar[ch-0x30]))
+
+/*
+** Extract the next token from a tokenization cursor. The cursor must
+** have been opened by a prior call to porterOpen().
+*/
+static int porterNext(
+ sqlite3_tokenizer_cursor *pCursor, /* Cursor returned by porterOpen */
+ const char **pzToken, /* OUT: *pzToken is the token text */
+ int *pnBytes, /* OUT: Number of bytes in token */
+ int *piStartOffset, /* OUT: Starting offset of token */
+ int *piEndOffset, /* OUT: Ending offset of token */
+ int *piPosition /* OUT: Position integer of token */
+){
+ porter_tokenizer_cursor *c = (porter_tokenizer_cursor *) pCursor;
+ const char *z = c->zInput;
+
+ while( c->iOffsetnInput ){
+ int iStartOffset, ch;
+
+ /* Scan past delimiter characters */
+ while( c->iOffsetnInput && isDelim(z[c->iOffset]) ){
+ c->iOffset++;
+ }
+
+ /* Count non-delimiter characters. */
+ iStartOffset = c->iOffset;
+ while( c->iOffsetnInput && !isDelim(z[c->iOffset]) ){
+ c->iOffset++;
+ }
+
+ if( c->iOffset>iStartOffset ){
+ int n = c->iOffset-iStartOffset;
+ if( n>c->nAllocated ){
+ char *pNew;
+ c->nAllocated = n+20;
+ pNew = sqlite3_realloc64(c->zToken, c->nAllocated);
+ if( !pNew ) return SQLITE_NOMEM;
+ c->zToken = pNew;
+ }
+ porter_stemmer(&z[iStartOffset], n, c->zToken, pnBytes);
+ *pzToken = c->zToken;
+ *piStartOffset = iStartOffset;
+ *piEndOffset = c->iOffset;
+ *piPosition = c->iToken++;
+ return SQLITE_OK;
+ }
+ }
+ return SQLITE_DONE;
+}
+
+/*
+** The set of routines that implement the porter-stemmer tokenizer
+*/
+static const sqlite3_tokenizer_module porterTokenizerModule = {
+ 0,
+ porterCreate,
+ porterDestroy,
+ porterOpen,
+ porterClose,
+ porterNext,
+ 0
+};
+
+/*
+** Allocate a new porter tokenizer. Return a pointer to the new
+** tokenizer in *ppModule
+*/
+void sqlite3Fts3PorterTokenizerModule(
+ sqlite3_tokenizer_module const**ppModule
+){
+ *ppModule = &porterTokenizerModule;
+}
+
+#endif /* !defined(SQLITE_CORE) || defined(SQLITE_ENABLE_FTS3) */
diff --git a/ext/fts3/fts3_snippet.c b/ext/fts3/fts3_snippet.c
new file mode 100644
index 0000000000000000000000000000000000000000..80f62eb3bbfcfe55968e0394d00ec3be5cf9ee37
--- /dev/null
+++ b/ext/fts3/fts3_snippet.c
@@ -0,0 +1,1756 @@
+/*
+** 2009 Oct 23
+**
+** The author disclaims copyright to this source code. In place of
+** a legal notice, here is a blessing:
+**
+** May you do good and not evil.
+** May you find forgiveness for yourself and forgive others.
+** May you share freely, never taking more than you give.
+**
+******************************************************************************
+*/
+
+#include "fts3Int.h"
+#if !defined(SQLITE_CORE) || defined(SQLITE_ENABLE_FTS3)
+
+#include
+#include
+
+#ifndef SQLITE_AMALGAMATION
+typedef sqlite3_int64 i64;
+#endif
+
+/*
+** Characters that may appear in the second argument to matchinfo().
+*/
+#define FTS3_MATCHINFO_NPHRASE 'p' /* 1 value */
+#define FTS3_MATCHINFO_NCOL 'c' /* 1 value */
+#define FTS3_MATCHINFO_NDOC 'n' /* 1 value */
+#define FTS3_MATCHINFO_AVGLENGTH 'a' /* nCol values */
+#define FTS3_MATCHINFO_LENGTH 'l' /* nCol values */
+#define FTS3_MATCHINFO_LCS 's' /* nCol values */
+#define FTS3_MATCHINFO_HITS 'x' /* 3*nCol*nPhrase values */
+#define FTS3_MATCHINFO_LHITS 'y' /* nCol*nPhrase values */
+#define FTS3_MATCHINFO_LHITS_BM 'b' /* nCol*nPhrase values */
+
+/*
+** The default value for the second argument to matchinfo().
+*/
+#define FTS3_MATCHINFO_DEFAULT "pcx"
+
+
+/*
+** Used as an sqlite3Fts3ExprIterate() context when loading phrase doclists to
+** Fts3Expr.aDoclist[]/nDoclist.
+*/
+typedef struct LoadDoclistCtx LoadDoclistCtx;
+struct LoadDoclistCtx {
+ Fts3Cursor *pCsr; /* FTS3 Cursor */
+ int nPhrase; /* Number of phrases seen so far */
+ int nToken; /* Number of tokens seen so far */
+};
+
+/*
+** The following types are used as part of the implementation of the
+** fts3BestSnippet() routine.
+*/
+typedef struct SnippetIter SnippetIter;
+typedef struct SnippetPhrase SnippetPhrase;
+typedef struct SnippetFragment SnippetFragment;
+
+struct SnippetIter {
+ Fts3Cursor *pCsr; /* Cursor snippet is being generated from */
+ int iCol; /* Extract snippet from this column */
+ int nSnippet; /* Requested snippet length (in tokens) */
+ int nPhrase; /* Number of phrases in query */
+ SnippetPhrase *aPhrase; /* Array of size nPhrase */
+ int iCurrent; /* First token of current snippet */
+};
+
+struct SnippetPhrase {
+ int nToken; /* Number of tokens in phrase */
+ char *pList; /* Pointer to start of phrase position list */
+ i64 iHead; /* Next value in position list */
+ char *pHead; /* Position list data following iHead */
+ i64 iTail; /* Next value in trailing position list */
+ char *pTail; /* Position list data following iTail */
+};
+
+struct SnippetFragment {
+ int iCol; /* Column snippet is extracted from */
+ int iPos; /* Index of first token in snippet */
+ u64 covered; /* Mask of query phrases covered */
+ u64 hlmask; /* Mask of snippet terms to highlight */
+};
+
+/*
+** This type is used as an sqlite3Fts3ExprIterate() context object while
+** accumulating the data returned by the matchinfo() function.
+*/
+typedef struct MatchInfo MatchInfo;
+struct MatchInfo {
+ Fts3Cursor *pCursor; /* FTS3 Cursor */
+ int nCol; /* Number of columns in table */
+ int nPhrase; /* Number of matchable phrases in query */
+ sqlite3_int64 nDoc; /* Number of docs in database */
+ char flag;
+ u32 *aMatchinfo; /* Pre-allocated buffer */
+};
+
+/*
+** An instance of this structure is used to manage a pair of buffers, each
+** (nElem * sizeof(u32)) bytes in size. See the MatchinfoBuffer code below
+** for details.
+*/
+struct MatchinfoBuffer {
+ u8 aRef[3];
+ int nElem;
+ int bGlobal; /* Set if global data is loaded */
+ char *zMatchinfo;
+ u32 aMatchinfo[1];
+};
+
+
+/*
+** The snippet() and offsets() functions both return text values. An instance
+** of the following structure is used to accumulate those values while the
+** functions are running. See fts3StringAppend() for details.
+*/
+typedef struct StrBuffer StrBuffer;
+struct StrBuffer {
+ char *z; /* Pointer to buffer containing string */
+ int n; /* Length of z in bytes (excl. nul-term) */
+ int nAlloc; /* Allocated size of buffer z in bytes */
+};
+
+
+/*************************************************************************
+** Start of MatchinfoBuffer code.
+*/
+
+/*
+** Allocate a two-slot MatchinfoBuffer object.
+*/
+static MatchinfoBuffer *fts3MIBufferNew(size_t nElem, const char *zMatchinfo){
+ MatchinfoBuffer *pRet;
+ sqlite3_int64 nByte = sizeof(u32) * (2*(sqlite3_int64)nElem + 1)
+ + sizeof(MatchinfoBuffer);
+ sqlite3_int64 nStr = strlen(zMatchinfo);
+
+ pRet = sqlite3Fts3MallocZero(nByte + nStr+1);
+ if( pRet ){
+ pRet->aMatchinfo[0] = (u8*)(&pRet->aMatchinfo[1]) - (u8*)pRet;
+ pRet->aMatchinfo[1+nElem] = pRet->aMatchinfo[0]
+ + sizeof(u32)*((int)nElem+1);
+ pRet->nElem = (int)nElem;
+ pRet->zMatchinfo = ((char*)pRet) + nByte;
+ memcpy(pRet->zMatchinfo, zMatchinfo, nStr+1);
+ pRet->aRef[0] = 1;
+ }
+
+ return pRet;
+}
+
+static void fts3MIBufferFree(void *p){
+ MatchinfoBuffer *pBuf = (MatchinfoBuffer*)((u8*)p - ((u32*)p)[-1]);
+
+ assert( (u32*)p==&pBuf->aMatchinfo[1]
+ || (u32*)p==&pBuf->aMatchinfo[pBuf->nElem+2]
+ );
+ if( (u32*)p==&pBuf->aMatchinfo[1] ){
+ pBuf->aRef[1] = 0;
+ }else{
+ pBuf->aRef[2] = 0;
+ }
+
+ if( pBuf->aRef[0]==0 && pBuf->aRef[1]==0 && pBuf->aRef[2]==0 ){
+ sqlite3_free(pBuf);
+ }
+}
+
+static void (*fts3MIBufferAlloc(MatchinfoBuffer *p, u32 **paOut))(void*){
+ void (*xRet)(void*) = 0;
+ u32 *aOut = 0;
+
+ if( p->aRef[1]==0 ){
+ p->aRef[1] = 1;
+ aOut = &p->aMatchinfo[1];
+ xRet = fts3MIBufferFree;
+ }
+ else if( p->aRef[2]==0 ){
+ p->aRef[2] = 1;
+ aOut = &p->aMatchinfo[p->nElem+2];
+ xRet = fts3MIBufferFree;
+ }else{
+ aOut = (u32*)sqlite3_malloc64(p->nElem * sizeof(u32));
+ if( aOut ){
+ xRet = sqlite3_free;
+ if( p->bGlobal ) memcpy(aOut, &p->aMatchinfo[1], p->nElem*sizeof(u32));
+ }
+ }
+
+ *paOut = aOut;
+ return xRet;
+}
+
+static void fts3MIBufferSetGlobal(MatchinfoBuffer *p){
+ p->bGlobal = 1;
+ memcpy(&p->aMatchinfo[2+p->nElem], &p->aMatchinfo[1], p->nElem*sizeof(u32));
+}
+
+/*
+** Free a MatchinfoBuffer object allocated using fts3MIBufferNew()
+*/
+void sqlite3Fts3MIBufferFree(MatchinfoBuffer *p){
+ if( p ){
+ assert( p->aRef[0]==1 );
+ p->aRef[0] = 0;
+ if( p->aRef[0]==0 && p->aRef[1]==0 && p->aRef[2]==0 ){
+ sqlite3_free(p);
+ }
+ }
+}
+
+/*
+** End of MatchinfoBuffer code.
+*************************************************************************/
+
+
+/*
+** This function is used to help iterate through a position-list. A position
+** list is a list of unique integers, sorted from smallest to largest. Each
+** element of the list is represented by an FTS3 varint that takes the value
+** of the difference between the current element and the previous one plus
+** two. For example, to store the position-list:
+**
+** 4 9 113
+**
+** the three varints:
+**
+** 6 7 106
+**
+** are encoded.
+**
+** When this function is called, *pp points to the start of an element of
+** the list. *piPos contains the value of the previous entry in the list.
+** After it returns, *piPos contains the value of the next element of the
+** list and *pp is advanced to the following varint.
+*/
+static void fts3GetDeltaPosition(char **pp, i64 *piPos){
+ int iVal;
+ *pp += fts3GetVarint32(*pp, &iVal);
+ *piPos += (iVal-2);
+}
+
+/*
+** Helper function for sqlite3Fts3ExprIterate() (see below).
+*/
+static int fts3ExprIterate2(
+ Fts3Expr *pExpr, /* Expression to iterate phrases of */
+ int *piPhrase, /* Pointer to phrase counter */
+ int (*x)(Fts3Expr*,int,void*), /* Callback function to invoke for phrases */
+ void *pCtx /* Second argument to pass to callback */
+){
+ int rc; /* Return code */
+ int eType = pExpr->eType; /* Type of expression node pExpr */
+
+ if( eType!=FTSQUERY_PHRASE ){
+ assert( pExpr->pLeft && pExpr->pRight );
+ rc = fts3ExprIterate2(pExpr->pLeft, piPhrase, x, pCtx);
+ if( rc==SQLITE_OK && eType!=FTSQUERY_NOT ){
+ rc = fts3ExprIterate2(pExpr->pRight, piPhrase, x, pCtx);
+ }
+ }else{
+ rc = x(pExpr, *piPhrase, pCtx);
+ (*piPhrase)++;
+ }
+ return rc;
+}
+
+/*
+** Iterate through all phrase nodes in an FTS3 query, except those that
+** are part of a sub-tree that is the right-hand-side of a NOT operator.
+** For each phrase node found, the supplied callback function is invoked.
+**
+** If the callback function returns anything other than SQLITE_OK,
+** the iteration is abandoned and the error code returned immediately.
+** Otherwise, SQLITE_OK is returned after a callback has been made for
+** all eligible phrase nodes.
+*/
+int sqlite3Fts3ExprIterate(
+ Fts3Expr *pExpr, /* Expression to iterate phrases of */
+ int (*x)(Fts3Expr*,int,void*), /* Callback function to invoke for phrases */
+ void *pCtx /* Second argument to pass to callback */
+){
+ int iPhrase = 0; /* Variable used as the phrase counter */
+ return fts3ExprIterate2(pExpr, &iPhrase, x, pCtx);
+}
+
+/*
+** This is an sqlite3Fts3ExprIterate() callback used while loading the
+** doclists for each phrase into Fts3Expr.aDoclist[]/nDoclist. See also
+** fts3ExprLoadDoclists().
+*/
+static int fts3ExprLoadDoclistsCb(Fts3Expr *pExpr, int iPhrase, void *ctx){
+ int rc = SQLITE_OK;
+ Fts3Phrase *pPhrase = pExpr->pPhrase;
+ LoadDoclistCtx *p = (LoadDoclistCtx *)ctx;
+
+ UNUSED_PARAMETER(iPhrase);
+
+ p->nPhrase++;
+ p->nToken += pPhrase->nToken;
+
+ return rc;
+}
+
+/*
+** Load the doclists for each phrase in the query associated with FTS3 cursor
+** pCsr.
+**
+** If pnPhrase is not NULL, then *pnPhrase is set to the number of matchable
+** phrases in the expression (all phrases except those directly or
+** indirectly descended from the right-hand-side of a NOT operator). If
+** pnToken is not NULL, then it is set to the number of tokens in all
+** matchable phrases of the expression.
+*/
+static int fts3ExprLoadDoclists(
+ Fts3Cursor *pCsr, /* Fts3 cursor for current query */
+ int *pnPhrase, /* OUT: Number of phrases in query */
+ int *pnToken /* OUT: Number of tokens in query */
+){
+ int rc; /* Return Code */
+ LoadDoclistCtx sCtx = {0,0,0}; /* Context for sqlite3Fts3ExprIterate() */
+ sCtx.pCsr = pCsr;
+ rc = sqlite3Fts3ExprIterate(pCsr->pExpr,fts3ExprLoadDoclistsCb,(void*)&sCtx);
+ if( pnPhrase ) *pnPhrase = sCtx.nPhrase;
+ if( pnToken ) *pnToken = sCtx.nToken;
+ return rc;
+}
+
+static int fts3ExprPhraseCountCb(Fts3Expr *pExpr, int iPhrase, void *ctx){
+ (*(int *)ctx)++;
+ pExpr->iPhrase = iPhrase;
+ return SQLITE_OK;
+}
+static int fts3ExprPhraseCount(Fts3Expr *pExpr){
+ int nPhrase = 0;
+ (void)sqlite3Fts3ExprIterate(pExpr, fts3ExprPhraseCountCb, (void *)&nPhrase);
+ return nPhrase;
+}
+
+/*
+** Advance the position list iterator specified by the first two
+** arguments so that it points to the first element with a value greater
+** than or equal to parameter iNext.
+*/
+static void fts3SnippetAdvance(char **ppIter, i64 *piIter, int iNext){
+ char *pIter = *ppIter;
+ if( pIter ){
+ i64 iIter = *piIter;
+
+ while( iIteriCurrent<0 ){
+ /* The SnippetIter object has just been initialized. The first snippet
+ ** candidate always starts at offset 0 (even if this candidate has a
+ ** score of 0.0).
+ */
+ pIter->iCurrent = 0;
+
+ /* Advance the 'head' iterator of each phrase to the first offset that
+ ** is greater than or equal to (iNext+nSnippet).
+ */
+ for(i=0; inPhrase; i++){
+ SnippetPhrase *pPhrase = &pIter->aPhrase[i];
+ fts3SnippetAdvance(&pPhrase->pHead, &pPhrase->iHead, pIter->nSnippet);
+ }
+ }else{
+ int iStart;
+ int iEnd = 0x7FFFFFFF;
+
+ for(i=0; inPhrase; i++){
+ SnippetPhrase *pPhrase = &pIter->aPhrase[i];
+ if( pPhrase->pHead && pPhrase->iHeadiHead;
+ }
+ }
+ if( iEnd==0x7FFFFFFF ){
+ return 1;
+ }
+
+ assert( pIter->nSnippet>=0 );
+ pIter->iCurrent = iStart = iEnd - pIter->nSnippet + 1;
+ for(i=0; inPhrase; i++){
+ SnippetPhrase *pPhrase = &pIter->aPhrase[i];
+ fts3SnippetAdvance(&pPhrase->pHead, &pPhrase->iHead, iEnd+1);
+ fts3SnippetAdvance(&pPhrase->pTail, &pPhrase->iTail, iStart);
+ }
+ }
+
+ return 0;
+}
+
+/*
+** Retrieve information about the current candidate snippet of snippet
+** iterator pIter.
+*/
+static void fts3SnippetDetails(
+ SnippetIter *pIter, /* Snippet iterator */
+ u64 mCovered, /* Bitmask of phrases already covered */
+ int *piToken, /* OUT: First token of proposed snippet */
+ int *piScore, /* OUT: "Score" for this snippet */
+ u64 *pmCover, /* OUT: Bitmask of phrases covered */
+ u64 *pmHighlight /* OUT: Bitmask of terms to highlight */
+){
+ int iStart = pIter->iCurrent; /* First token of snippet */
+ int iScore = 0; /* Score of this snippet */
+ int i; /* Loop counter */
+ u64 mCover = 0; /* Mask of phrases covered by this snippet */
+ u64 mHighlight = 0; /* Mask of tokens to highlight in snippet */
+
+ for(i=0; inPhrase; i++){
+ SnippetPhrase *pPhrase = &pIter->aPhrase[i];
+ if( pPhrase->pTail ){
+ char *pCsr = pPhrase->pTail;
+ i64 iCsr = pPhrase->iTail;
+
+ while( iCsr<(iStart+pIter->nSnippet) && iCsr>=iStart ){
+ int j;
+ u64 mPhrase = (u64)1 << (i%64);
+ u64 mPos = (u64)1 << (iCsr - iStart);
+ assert( iCsr>=iStart && (iCsr - iStart)<=64 );
+ assert( i>=0 );
+ if( (mCover|mCovered)&mPhrase ){
+ iScore++;
+ }else{
+ iScore += 1000;
+ }
+ mCover |= mPhrase;
+
+ for(j=0; jnToken && jnSnippet; j++){
+ mHighlight |= (mPos>>j);
+ }
+
+ if( 0==(*pCsr & 0x0FE) ) break;
+ fts3GetDeltaPosition(&pCsr, &iCsr);
+ }
+ }
+ }
+
+ /* Set the output variables before returning. */
+ *piToken = iStart;
+ *piScore = iScore;
+ *pmCover = mCover;
+ *pmHighlight = mHighlight;
+}
+
+/*
+** This function is an sqlite3Fts3ExprIterate() callback used by
+** fts3BestSnippet(). Each invocation populates an element of the
+** SnippetIter.aPhrase[] array.
+*/
+static int fts3SnippetFindPositions(Fts3Expr *pExpr, int iPhrase, void *ctx){
+ SnippetIter *p = (SnippetIter *)ctx;
+ SnippetPhrase *pPhrase = &p->aPhrase[iPhrase];
+ char *pCsr;
+ int rc;
+
+ pPhrase->nToken = pExpr->pPhrase->nToken;
+ rc = sqlite3Fts3EvalPhrasePoslist(p->pCsr, pExpr, p->iCol, &pCsr);
+ assert( rc==SQLITE_OK || pCsr==0 );
+ if( pCsr ){
+ i64 iFirst = 0;
+ pPhrase->pList = pCsr;
+ fts3GetDeltaPosition(&pCsr, &iFirst);
+ if( iFirst<0 ){
+ rc = FTS_CORRUPT_VTAB;
+ }else{
+ pPhrase->pHead = pCsr;
+ pPhrase->pTail = pCsr;
+ pPhrase->iHead = iFirst;
+ pPhrase->iTail = iFirst;
+ }
+ }else{
+ assert( rc!=SQLITE_OK || (
+ pPhrase->pList==0 && pPhrase->pHead==0 && pPhrase->pTail==0
+ ));
+ }
+
+ return rc;
+}
+
+/*
+** Select the fragment of text consisting of nFragment contiguous tokens
+** from column iCol that represent the "best" snippet. The best snippet
+** is the snippet with the highest score, where scores are calculated
+** by adding:
+**
+** (a) +1 point for each occurrence of a matchable phrase in the snippet.
+**
+** (b) +1000 points for the first occurrence of each matchable phrase in
+** the snippet for which the corresponding mCovered bit is not set.
+**
+** The selected snippet parameters are stored in structure *pFragment before
+** returning. The score of the selected snippet is stored in *piScore
+** before returning.
+*/
+static int fts3BestSnippet(
+ int nSnippet, /* Desired snippet length */
+ Fts3Cursor *pCsr, /* Cursor to create snippet for */
+ int iCol, /* Index of column to create snippet from */
+ u64 mCovered, /* Mask of phrases already covered */
+ u64 *pmSeen, /* IN/OUT: Mask of phrases seen */
+ SnippetFragment *pFragment, /* OUT: Best snippet found */
+ int *piScore /* OUT: Score of snippet pFragment */
+){
+ int rc; /* Return Code */
+ int nList; /* Number of phrases in expression */
+ SnippetIter sIter; /* Iterates through snippet candidates */
+ sqlite3_int64 nByte; /* Number of bytes of space to allocate */
+ int iBestScore = -1; /* Best snippet score found so far */
+ int i; /* Loop counter */
+
+ memset(&sIter, 0, sizeof(sIter));
+
+ /* Iterate through the phrases in the expression to count them. The same
+ ** callback makes sure the doclists are loaded for each phrase.
+ */
+ rc = fts3ExprLoadDoclists(pCsr, &nList, 0);
+ if( rc!=SQLITE_OK ){
+ return rc;
+ }
+
+ /* Now that it is known how many phrases there are, allocate and zero
+ ** the required space using malloc().
+ */
+ nByte = sizeof(SnippetPhrase) * nList;
+ sIter.aPhrase = (SnippetPhrase *)sqlite3Fts3MallocZero(nByte);
+ if( !sIter.aPhrase ){
+ return SQLITE_NOMEM;
+ }
+
+ /* Initialize the contents of the SnippetIter object. Then iterate through
+ ** the set of phrases in the expression to populate the aPhrase[] array.
+ */
+ sIter.pCsr = pCsr;
+ sIter.iCol = iCol;
+ sIter.nSnippet = nSnippet;
+ sIter.nPhrase = nList;
+ sIter.iCurrent = -1;
+ rc = sqlite3Fts3ExprIterate(
+ pCsr->pExpr, fts3SnippetFindPositions, (void*)&sIter
+ );
+ if( rc==SQLITE_OK ){
+
+ /* Set the *pmSeen output variable. */
+ for(i=0; iiCol = iCol;
+ while( !fts3SnippetNextCandidate(&sIter) ){
+ int iPos;
+ int iScore;
+ u64 mCover;
+ u64 mHighlite;
+ fts3SnippetDetails(&sIter, mCovered, &iPos, &iScore, &mCover,&mHighlite);
+ assert( iScore>=0 );
+ if( iScore>iBestScore ){
+ pFragment->iPos = iPos;
+ pFragment->hlmask = mHighlite;
+ pFragment->covered = mCover;
+ iBestScore = iScore;
+ }
+ }
+
+ *piScore = iBestScore;
+ }
+ sqlite3_free(sIter.aPhrase);
+ return rc;
+}
+
+
+/*
+** Append a string to the string-buffer passed as the first argument.
+**
+** If nAppend is negative, then the length of the string zAppend is
+** determined using strlen().
+*/
+static int fts3StringAppend(
+ StrBuffer *pStr, /* Buffer to append to */
+ const char *zAppend, /* Pointer to data to append to buffer */
+ int nAppend /* Size of zAppend in bytes (or -1) */
+){
+ if( nAppend<0 ){
+ nAppend = (int)strlen(zAppend);
+ }
+
+ /* If there is insufficient space allocated at StrBuffer.z, use realloc()
+ ** to grow the buffer until so that it is big enough to accomadate the
+ ** appended data.
+ */
+ if( pStr->n+nAppend+1>=pStr->nAlloc ){
+ sqlite3_int64 nAlloc = pStr->nAlloc+(sqlite3_int64)nAppend+100;
+ char *zNew = sqlite3_realloc64(pStr->z, nAlloc);
+ if( !zNew ){
+ return SQLITE_NOMEM;
+ }
+ pStr->z = zNew;
+ pStr->nAlloc = nAlloc;
+ }
+ assert( pStr->z!=0 && (pStr->nAlloc >= pStr->n+nAppend+1) );
+
+ /* Append the data to the string buffer. */
+ memcpy(&pStr->z[pStr->n], zAppend, nAppend);
+ pStr->n += nAppend;
+ pStr->z[pStr->n] = '\0';
+
+ return SQLITE_OK;
+}
+
+/*
+** The fts3BestSnippet() function often selects snippets that end with a
+** query term. That is, the final term of the snippet is always a term
+** that requires highlighting. For example, if 'X' is a highlighted term
+** and '.' is a non-highlighted term, BestSnippet() may select:
+**
+** ........X.....X
+**
+** This function "shifts" the beginning of the snippet forward in the
+** document so that there are approximately the same number of
+** non-highlighted terms to the right of the final highlighted term as there
+** are to the left of the first highlighted term. For example, to this:
+**
+** ....X.....X....
+**
+** This is done as part of extracting the snippet text, not when selecting
+** the snippet. Snippet selection is done based on doclists only, so there
+** is no way for fts3BestSnippet() to know whether or not the document
+** actually contains terms that follow the final highlighted term.
+*/
+static int fts3SnippetShift(
+ Fts3Table *pTab, /* FTS3 table snippet comes from */
+ int iLangid, /* Language id to use in tokenizing */
+ int nSnippet, /* Number of tokens desired for snippet */
+ const char *zDoc, /* Document text to extract snippet from */
+ int nDoc, /* Size of buffer zDoc in bytes */
+ int *piPos, /* IN/OUT: First token of snippet */
+ u64 *pHlmask /* IN/OUT: Mask of tokens to highlight */
+){
+ u64 hlmask = *pHlmask; /* Local copy of initial highlight-mask */
+
+ if( hlmask ){
+ int nLeft; /* Tokens to the left of first highlight */
+ int nRight; /* Tokens to the right of last highlight */
+ int nDesired; /* Ideal number of tokens to shift forward */
+
+ for(nLeft=0; !(hlmask & ((u64)1 << nLeft)); nLeft++);
+ for(nRight=0; !(hlmask & ((u64)1 << (nSnippet-1-nRight))); nRight++);
+ assert( (nSnippet-1-nRight)<=63 && (nSnippet-1-nRight)>=0 );
+ nDesired = (nLeft-nRight)/2;
+
+ /* Ideally, the start of the snippet should be pushed forward in the
+ ** document nDesired tokens. This block checks if there are actually
+ ** nDesired tokens to the right of the snippet. If so, *piPos and
+ ** *pHlMask are updated to shift the snippet nDesired tokens to the
+ ** right. Otherwise, the snippet is shifted by the number of tokens
+ ** available.
+ */
+ if( nDesired>0 ){
+ int nShift; /* Number of tokens to shift snippet by */
+ int iCurrent = 0; /* Token counter */
+ int rc; /* Return Code */
+ sqlite3_tokenizer_module *pMod;
+ sqlite3_tokenizer_cursor *pC;
+ pMod = (sqlite3_tokenizer_module *)pTab->pTokenizer->pModule;
+
+ /* Open a cursor on zDoc/nDoc. Check if there are (nSnippet+nDesired)
+ ** or more tokens in zDoc/nDoc.
+ */
+ rc = sqlite3Fts3OpenTokenizer(pTab->pTokenizer, iLangid, zDoc, nDoc, &pC);
+ if( rc!=SQLITE_OK ){
+ return rc;
+ }
+ while( rc==SQLITE_OK && iCurrent<(nSnippet+nDesired) ){
+ const char *ZDUMMY; int DUMMY1 = 0, DUMMY2 = 0, DUMMY3 = 0;
+ rc = pMod->xNext(pC, &ZDUMMY, &DUMMY1, &DUMMY2, &DUMMY3, &iCurrent);
+ }
+ pMod->xClose(pC);
+ if( rc!=SQLITE_OK && rc!=SQLITE_DONE ){ return rc; }
+
+ nShift = (rc==SQLITE_DONE)+iCurrent-nSnippet;
+ assert( nShift<=nDesired );
+ if( nShift>0 ){
+ *piPos += nShift;
+ *pHlmask = hlmask >> nShift;
+ }
+ }
+ }
+ return SQLITE_OK;
+}
+
+/*
+** Extract the snippet text for fragment pFragment from cursor pCsr and
+** append it to string buffer pOut.
+*/
+static int fts3SnippetText(
+ Fts3Cursor *pCsr, /* FTS3 Cursor */
+ SnippetFragment *pFragment, /* Snippet to extract */
+ int iFragment, /* Fragment number */
+ int isLast, /* True for final fragment in snippet */
+ int nSnippet, /* Number of tokens in extracted snippet */
+ const char *zOpen, /* String inserted before highlighted term */
+ const char *zClose, /* String inserted after highlighted term */
+ const char *zEllipsis, /* String inserted between snippets */
+ StrBuffer *pOut /* Write output here */
+){
+ Fts3Table *pTab = (Fts3Table *)pCsr->base.pVtab;
+ int rc; /* Return code */
+ const char *zDoc; /* Document text to extract snippet from */
+ int nDoc; /* Size of zDoc in bytes */
+ int iCurrent = 0; /* Current token number of document */
+ int iEnd = 0; /* Byte offset of end of current token */
+ int isShiftDone = 0; /* True after snippet is shifted */
+ int iPos = pFragment->iPos; /* First token of snippet */
+ u64 hlmask = pFragment->hlmask; /* Highlight-mask for snippet */
+ int iCol = pFragment->iCol+1; /* Query column to extract text from */
+ sqlite3_tokenizer_module *pMod; /* Tokenizer module methods object */
+ sqlite3_tokenizer_cursor *pC; /* Tokenizer cursor open on zDoc/nDoc */
+
+ zDoc = (const char *)sqlite3_column_text(pCsr->pStmt, iCol);
+ if( zDoc==0 ){
+ if( sqlite3_column_type(pCsr->pStmt, iCol)!=SQLITE_NULL ){
+ return SQLITE_NOMEM;
+ }
+ return SQLITE_OK;
+ }
+ nDoc = sqlite3_column_bytes(pCsr->pStmt, iCol);
+
+ /* Open a token cursor on the document. */
+ pMod = (sqlite3_tokenizer_module *)pTab->pTokenizer->pModule;
+ rc = sqlite3Fts3OpenTokenizer(pTab->pTokenizer, pCsr->iLangid, zDoc,nDoc,&pC);
+ if( rc!=SQLITE_OK ){
+ return rc;
+ }
+
+ while( rc==SQLITE_OK ){
+ const char *ZDUMMY; /* Dummy argument used with tokenizer */
+ int DUMMY1 = -1; /* Dummy argument used with tokenizer */
+ int iBegin = 0; /* Offset in zDoc of start of token */
+ int iFin = 0; /* Offset in zDoc of end of token */
+ int isHighlight = 0; /* True for highlighted terms */
+
+ /* Variable DUMMY1 is initialized to a negative value above. Elsewhere
+ ** in the FTS code the variable that the third argument to xNext points to
+ ** is initialized to zero before the first (*but not necessarily
+ ** subsequent*) call to xNext(). This is done for a particular application
+ ** that needs to know whether or not the tokenizer is being used for
+ ** snippet generation or for some other purpose.
+ **
+ ** Extreme care is required when writing code to depend on this
+ ** initialization. It is not a documented part of the tokenizer interface.
+ ** If a tokenizer is used directly by any code outside of FTS, this
+ ** convention might not be respected. */
+ rc = pMod->xNext(pC, &ZDUMMY, &DUMMY1, &iBegin, &iFin, &iCurrent);
+ if( rc!=SQLITE_OK ){
+ if( rc==SQLITE_DONE ){
+ /* Special case - the last token of the snippet is also the last token
+ ** of the column. Append any punctuation that occurred between the end
+ ** of the previous token and the end of the document to the output.
+ ** Then break out of the loop. */
+ rc = fts3StringAppend(pOut, &zDoc[iEnd], -1);
+ }
+ break;
+ }
+ if( iCurrentiLangid, nSnippet, &zDoc[iBegin], n, &iPos, &hlmask
+ );
+ isShiftDone = 1;
+
+ /* Now that the shift has been done, check if the initial "..." are
+ ** required. They are required if (a) this is not the first fragment,
+ ** or (b) this fragment does not begin at position 0 of its column.
+ */
+ if( rc==SQLITE_OK ){
+ if( iPos>0 || iFragment>0 ){
+ rc = fts3StringAppend(pOut, zEllipsis, -1);
+ }else if( iBegin ){
+ rc = fts3StringAppend(pOut, zDoc, iBegin);
+ }
+ }
+ if( rc!=SQLITE_OK || iCurrent=(iPos+nSnippet) ){
+ if( isLast ){
+ rc = fts3StringAppend(pOut, zEllipsis, -1);
+ }
+ break;
+ }
+
+ /* Set isHighlight to true if this term should be highlighted. */
+ isHighlight = (hlmask & ((u64)1 << (iCurrent-iPos)))!=0;
+
+ if( iCurrent>iPos ) rc = fts3StringAppend(pOut, &zDoc[iEnd], iBegin-iEnd);
+ if( rc==SQLITE_OK && isHighlight ) rc = fts3StringAppend(pOut, zOpen, -1);
+ if( rc==SQLITE_OK ) rc = fts3StringAppend(pOut, &zDoc[iBegin], iFin-iBegin);
+ if( rc==SQLITE_OK && isHighlight ) rc = fts3StringAppend(pOut, zClose, -1);
+
+ iEnd = iFin;
+ }
+
+ pMod->xClose(pC);
+ return rc;
+}
+
+
+/*
+** This function is used to count the entries in a column-list (a
+** delta-encoded list of term offsets within a single column of a single
+** row). When this function is called, *ppCollist should point to the
+** beginning of the first varint in the column-list (the varint that
+** contains the position of the first matching term in the column data).
+** Before returning, *ppCollist is set to point to the first byte after
+** the last varint in the column-list (either the 0x00 signifying the end
+** of the position-list, or the 0x01 that precedes the column number of
+** the next column in the position-list).
+**
+** The number of elements in the column-list is returned.
+*/
+static int fts3ColumnlistCount(char **ppCollist){
+ char *pEnd = *ppCollist;
+ char c = 0;
+ int nEntry = 0;
+
+ /* A column-list is terminated by either a 0x01 or 0x00. */
+ while( 0xFE & (*pEnd | c) ){
+ c = *pEnd++ & 0x80;
+ if( !c ) nEntry++;
+ }
+
+ *ppCollist = pEnd;
+ return nEntry;
+}
+
+/*
+** This function gathers 'y' or 'b' data for a single phrase.
+*/
+static int fts3ExprLHits(
+ Fts3Expr *pExpr, /* Phrase expression node */
+ MatchInfo *p /* Matchinfo context */
+){
+ Fts3Table *pTab = (Fts3Table *)p->pCursor->base.pVtab;
+ int iStart;
+ Fts3Phrase *pPhrase = pExpr->pPhrase;
+ char *pIter = pPhrase->doclist.pList;
+ int iCol = 0;
+
+ assert( p->flag==FTS3_MATCHINFO_LHITS_BM || p->flag==FTS3_MATCHINFO_LHITS );
+ if( p->flag==FTS3_MATCHINFO_LHITS ){
+ iStart = pExpr->iPhrase * p->nCol;
+ }else{
+ iStart = pExpr->iPhrase * ((p->nCol + 31) / 32);
+ }
+
+ if( pIter ) while( 1 ){
+ int nHit = fts3ColumnlistCount(&pIter);
+ if( (pPhrase->iColumn>=pTab->nColumn || pPhrase->iColumn==iCol) ){
+ if( p->flag==FTS3_MATCHINFO_LHITS ){
+ p->aMatchinfo[iStart + iCol] = (u32)nHit;
+ }else if( nHit ){
+ p->aMatchinfo[iStart + (iCol+1)/32] |= (1 << (iCol&0x1F));
+ }
+ }
+ assert( *pIter==0x00 || *pIter==0x01 );
+ if( *pIter!=0x01 ) break;
+ pIter++;
+ pIter += fts3GetVarint32(pIter, &iCol);
+ if( iCol>=p->nCol ) return FTS_CORRUPT_VTAB;
+ }
+ return SQLITE_OK;
+}
+
+/*
+** Gather the results for matchinfo directives 'y' and 'b'.
+*/
+static int fts3ExprLHitGather(
+ Fts3Expr *pExpr,
+ MatchInfo *p
+){
+ int rc = SQLITE_OK;
+ assert( (pExpr->pLeft==0)==(pExpr->pRight==0) );
+ if( pExpr->bEof==0 && pExpr->iDocid==p->pCursor->iPrevId ){
+ if( pExpr->pLeft ){
+ rc = fts3ExprLHitGather(pExpr->pLeft, p);
+ if( rc==SQLITE_OK ) rc = fts3ExprLHitGather(pExpr->pRight, p);
+ }else{
+ rc = fts3ExprLHits(pExpr, p);
+ }
+ }
+ return rc;
+}
+
+/*
+** sqlite3Fts3ExprIterate() callback used to collect the "global" matchinfo
+** stats for a single query.
+**
+** sqlite3Fts3ExprIterate() callback to load the 'global' elements of a
+** FTS3_MATCHINFO_HITS matchinfo array. The global stats are those elements
+** of the matchinfo array that are constant for all rows returned by the
+** current query.
+**
+** Argument pCtx is actually a pointer to a struct of type MatchInfo. This
+** function populates Matchinfo.aMatchinfo[] as follows:
+**
+** for(iCol=0; iColpCursor, pExpr, &p->aMatchinfo[3*iPhrase*p->nCol]
+ );
+}
+
+/*
+** sqlite3Fts3ExprIterate() callback used to collect the "local" part of the
+** FTS3_MATCHINFO_HITS array. The local stats are those elements of the
+** array that are different for each row returned by the query.
+*/
+static int fts3ExprLocalHitsCb(
+ Fts3Expr *pExpr, /* Phrase expression node */
+ int iPhrase, /* Phrase number */
+ void *pCtx /* Pointer to MatchInfo structure */
+){
+ int rc = SQLITE_OK;
+ MatchInfo *p = (MatchInfo *)pCtx;
+ int iStart = iPhrase * p->nCol * 3;
+ int i;
+
+ for(i=0; inCol && rc==SQLITE_OK; i++){
+ char *pCsr;
+ rc = sqlite3Fts3EvalPhrasePoslist(p->pCursor, pExpr, i, &pCsr);
+ if( pCsr ){
+ p->aMatchinfo[iStart+i*3] = fts3ColumnlistCount(&pCsr);
+ }else{
+ p->aMatchinfo[iStart+i*3] = 0;
+ }
+ }
+
+ return rc;
+}
+
+static int fts3MatchinfoCheck(
+ Fts3Table *pTab,
+ char cArg,
+ char **pzErr
+){
+ if( (cArg==FTS3_MATCHINFO_NPHRASE)
+ || (cArg==FTS3_MATCHINFO_NCOL)
+ || (cArg==FTS3_MATCHINFO_NDOC && pTab->bFts4)
+ || (cArg==FTS3_MATCHINFO_AVGLENGTH && pTab->bFts4)
+ || (cArg==FTS3_MATCHINFO_LENGTH && pTab->bHasDocsize)
+ || (cArg==FTS3_MATCHINFO_LCS)
+ || (cArg==FTS3_MATCHINFO_HITS)
+ || (cArg==FTS3_MATCHINFO_LHITS)
+ || (cArg==FTS3_MATCHINFO_LHITS_BM)
+ ){
+ return SQLITE_OK;
+ }
+ sqlite3Fts3ErrMsg(pzErr, "unrecognized matchinfo request: %c", cArg);
+ return SQLITE_ERROR;
+}
+
+static size_t fts3MatchinfoSize(MatchInfo *pInfo, char cArg){
+ size_t nVal; /* Number of integers output by cArg */
+
+ switch( cArg ){
+ case FTS3_MATCHINFO_NDOC:
+ case FTS3_MATCHINFO_NPHRASE:
+ case FTS3_MATCHINFO_NCOL:
+ nVal = 1;
+ break;
+
+ case FTS3_MATCHINFO_AVGLENGTH:
+ case FTS3_MATCHINFO_LENGTH:
+ case FTS3_MATCHINFO_LCS:
+ nVal = pInfo->nCol;
+ break;
+
+ case FTS3_MATCHINFO_LHITS:
+ nVal = pInfo->nCol * pInfo->nPhrase;
+ break;
+
+ case FTS3_MATCHINFO_LHITS_BM:
+ nVal = pInfo->nPhrase * ((pInfo->nCol + 31) / 32);
+ break;
+
+ default:
+ assert( cArg==FTS3_MATCHINFO_HITS );
+ nVal = pInfo->nCol * pInfo->nPhrase * 3;
+ break;
+ }
+
+ return nVal;
+}
+
+static int fts3MatchinfoSelectDoctotal(
+ Fts3Table *pTab,
+ sqlite3_stmt **ppStmt,
+ sqlite3_int64 *pnDoc,
+ const char **paLen,
+ const char **ppEnd
+){
+ sqlite3_stmt *pStmt;
+ const char *a;
+ const char *pEnd;
+ sqlite3_int64 nDoc;
+ int n;
+
+
+ if( !*ppStmt ){
+ int rc = sqlite3Fts3SelectDoctotal(pTab, ppStmt);
+ if( rc!=SQLITE_OK ) return rc;
+ }
+ pStmt = *ppStmt;
+ assert( sqlite3_data_count(pStmt)==1 );
+
+ n = sqlite3_column_bytes(pStmt, 0);
+ a = sqlite3_column_blob(pStmt, 0);
+ if( a==0 ){
+ return FTS_CORRUPT_VTAB;
+ }
+ pEnd = a + n;
+ a += sqlite3Fts3GetVarintBounded(a, pEnd, &nDoc);
+ if( nDoc<=0 || a>pEnd ){
+ return FTS_CORRUPT_VTAB;
+ }
+ *pnDoc = nDoc;
+
+ if( paLen ) *paLen = a;
+ if( ppEnd ) *ppEnd = pEnd;
+ return SQLITE_OK;
+}
+
+/*
+** An instance of the following structure is used to store state while
+** iterating through a multi-column position-list corresponding to the
+** hits for a single phrase on a single row in order to calculate the
+** values for a matchinfo() FTS3_MATCHINFO_LCS request.
+*/
+typedef struct LcsIterator LcsIterator;
+struct LcsIterator {
+ Fts3Expr *pExpr; /* Pointer to phrase expression */
+ int iPosOffset; /* Tokens count up to end of this phrase */
+ char *pRead; /* Cursor used to iterate through aDoclist */
+ int iPos; /* Current position */
+};
+
+/*
+** If LcsIterator.iCol is set to the following value, the iterator has
+** finished iterating through all offsets for all columns.
+*/
+#define LCS_ITERATOR_FINISHED 0x7FFFFFFF;
+
+static int fts3MatchinfoLcsCb(
+ Fts3Expr *pExpr, /* Phrase expression node */
+ int iPhrase, /* Phrase number (numbered from zero) */
+ void *pCtx /* Pointer to MatchInfo structure */
+){
+ LcsIterator *aIter = (LcsIterator *)pCtx;
+ aIter[iPhrase].pExpr = pExpr;
+ return SQLITE_OK;
+}
+
+/*
+** Advance the iterator passed as an argument to the next position. Return
+** 1 if the iterator is at EOF or if it now points to the start of the
+** position list for the next column.
+*/
+static int fts3LcsIteratorAdvance(LcsIterator *pIter){
+ char *pRead;
+ sqlite3_int64 iRead;
+ int rc = 0;
+
+ if( NEVER(pIter==0) ) return 1;
+ pRead = pIter->pRead;
+ pRead += sqlite3Fts3GetVarint(pRead, &iRead);
+ if( iRead==0 || iRead==1 ){
+ pRead = 0;
+ rc = 1;
+ }else{
+ pIter->iPos += (int)(iRead-2);
+ }
+
+ pIter->pRead = pRead;
+ return rc;
+}
+
+/*
+** This function implements the FTS3_MATCHINFO_LCS matchinfo() flag.
+**
+** If the call is successful, the longest-common-substring lengths for each
+** column are written into the first nCol elements of the pInfo->aMatchinfo[]
+** array before returning. SQLITE_OK is returned in this case.
+**
+** Otherwise, if an error occurs, an SQLite error code is returned and the
+** data written to the first nCol elements of pInfo->aMatchinfo[] is
+** undefined.
+*/
+static int fts3MatchinfoLcs(Fts3Cursor *pCsr, MatchInfo *pInfo){
+ LcsIterator *aIter;
+ int i;
+ int iCol;
+ int nToken = 0;
+ int rc = SQLITE_OK;
+
+ /* Allocate and populate the array of LcsIterator objects. The array
+ ** contains one element for each matchable phrase in the query.
+ **/
+ aIter = sqlite3Fts3MallocZero(sizeof(LcsIterator) * pCsr->nPhrase);
+ if( !aIter ) return SQLITE_NOMEM;
+ (void)sqlite3Fts3ExprIterate(pCsr->pExpr, fts3MatchinfoLcsCb, (void*)aIter);
+
+ for(i=0; inPhrase; i++){
+ LcsIterator *pIter = &aIter[i];
+ nToken -= pIter->pExpr->pPhrase->nToken;
+ pIter->iPosOffset = nToken;
+ }
+
+ for(iCol=0; iColnCol; iCol++){
+ int nLcs = 0; /* LCS value for this column */
+ int nLive = 0; /* Number of iterators in aIter not at EOF */
+
+ for(i=0; inPhrase; i++){
+ LcsIterator *pIt = &aIter[i];
+ rc = sqlite3Fts3EvalPhrasePoslist(pCsr, pIt->pExpr, iCol, &pIt->pRead);
+ if( rc!=SQLITE_OK ) goto matchinfo_lcs_out;
+ if( pIt->pRead ){
+ pIt->iPos = pIt->iPosOffset;
+ fts3LcsIteratorAdvance(pIt);
+ if( pIt->pRead==0 ){
+ rc = FTS_CORRUPT_VTAB;
+ goto matchinfo_lcs_out;
+ }
+ nLive++;
+ }
+ }
+
+ while( nLive>0 ){
+ LcsIterator *pAdv = 0; /* The iterator to advance by one position */
+ int nThisLcs = 0; /* LCS for the current iterator positions */
+
+ for(i=0; inPhrase; i++){
+ LcsIterator *pIter = &aIter[i];
+ if( pIter->pRead==0 ){
+ /* This iterator is already at EOF for this column. */
+ nThisLcs = 0;
+ }else{
+ if( pAdv==0 || pIter->iPosiPos ){
+ pAdv = pIter;
+ }
+ if( nThisLcs==0 || pIter->iPos==pIter[-1].iPos ){
+ nThisLcs++;
+ }else{
+ nThisLcs = 1;
+ }
+ if( nThisLcs>nLcs ) nLcs = nThisLcs;
+ }
+ }
+ if( fts3LcsIteratorAdvance(pAdv) ) nLive--;
+ }
+
+ pInfo->aMatchinfo[iCol] = nLcs;
+ }
+
+ matchinfo_lcs_out:
+ sqlite3_free(aIter);
+ return rc;
+}
+
+/*
+** Populate the buffer pInfo->aMatchinfo[] with an array of integers to
+** be returned by the matchinfo() function. Argument zArg contains the
+** format string passed as the second argument to matchinfo (or the
+** default value "pcx" if no second argument was specified). The format
+** string has already been validated and the pInfo->aMatchinfo[] array
+** is guaranteed to be large enough for the output.
+**
+** If bGlobal is true, then populate all fields of the matchinfo() output.
+** If it is false, then assume that those fields that do not change between
+** rows (i.e. FTS3_MATCHINFO_NPHRASE, NCOL, NDOC, AVGLENGTH and part of HITS)
+** have already been populated.
+**
+** Return SQLITE_OK if successful, or an SQLite error code if an error
+** occurs. If a value other than SQLITE_OK is returned, the state the
+** pInfo->aMatchinfo[] buffer is left in is undefined.
+*/
+static int fts3MatchinfoValues(
+ Fts3Cursor *pCsr, /* FTS3 cursor object */
+ int bGlobal, /* True to grab the global stats */
+ MatchInfo *pInfo, /* Matchinfo context object */
+ const char *zArg /* Matchinfo format string */
+){
+ int rc = SQLITE_OK;
+ int i;
+ Fts3Table *pTab = (Fts3Table *)pCsr->base.pVtab;
+ sqlite3_stmt *pSelect = 0;
+
+ for(i=0; rc==SQLITE_OK && zArg[i]; i++){
+ pInfo->flag = zArg[i];
+ switch( zArg[i] ){
+ case FTS3_MATCHINFO_NPHRASE:
+ if( bGlobal ) pInfo->aMatchinfo[0] = pInfo->nPhrase;
+ break;
+
+ case FTS3_MATCHINFO_NCOL:
+ if( bGlobal ) pInfo->aMatchinfo[0] = pInfo->nCol;
+ break;
+
+ case FTS3_MATCHINFO_NDOC:
+ if( bGlobal ){
+ sqlite3_int64 nDoc = 0;
+ rc = fts3MatchinfoSelectDoctotal(pTab, &pSelect, &nDoc, 0, 0);
+ pInfo->aMatchinfo[0] = (u32)nDoc;
+ }
+ break;
+
+ case FTS3_MATCHINFO_AVGLENGTH:
+ if( bGlobal ){
+ sqlite3_int64 nDoc; /* Number of rows in table */
+ const char *a; /* Aggregate column length array */
+ const char *pEnd; /* First byte past end of length array */
+
+ rc = fts3MatchinfoSelectDoctotal(pTab, &pSelect, &nDoc, &a, &pEnd);
+ if( rc==SQLITE_OK ){
+ int iCol;
+ for(iCol=0; iColnCol; iCol++){
+ u32 iVal;
+ sqlite3_int64 nToken;
+ a += sqlite3Fts3GetVarint(a, &nToken);
+ if( a>pEnd ){
+ rc = SQLITE_CORRUPT_VTAB;
+ break;
+ }
+ iVal = (u32)(((u32)(nToken&0xffffffff)+nDoc/2)/nDoc);
+ pInfo->aMatchinfo[iCol] = iVal;
+ }
+ }
+ }
+ break;
+
+ case FTS3_MATCHINFO_LENGTH: {
+ sqlite3_stmt *pSelectDocsize = 0;
+ rc = sqlite3Fts3SelectDocsize(pTab, pCsr->iPrevId, &pSelectDocsize);
+ if( rc==SQLITE_OK ){
+ int iCol;
+ const char *a = sqlite3_column_blob(pSelectDocsize, 0);
+ const char *pEnd = a + sqlite3_column_bytes(pSelectDocsize, 0);
+ for(iCol=0; iColnCol; iCol++){
+ sqlite3_int64 nToken;
+ a += sqlite3Fts3GetVarintBounded(a, pEnd, &nToken);
+ if( a>pEnd ){
+ rc = SQLITE_CORRUPT_VTAB;
+ break;
+ }
+ pInfo->aMatchinfo[iCol] = (u32)nToken;
+ }
+ }
+ sqlite3_reset(pSelectDocsize);
+ break;
+ }
+
+ case FTS3_MATCHINFO_LCS:
+ rc = fts3ExprLoadDoclists(pCsr, 0, 0);
+ if( rc==SQLITE_OK ){
+ rc = fts3MatchinfoLcs(pCsr, pInfo);
+ }
+ break;
+
+ case FTS3_MATCHINFO_LHITS_BM:
+ case FTS3_MATCHINFO_LHITS: {
+ size_t nZero = fts3MatchinfoSize(pInfo, zArg[i]) * sizeof(u32);
+ memset(pInfo->aMatchinfo, 0, nZero);
+ rc = fts3ExprLHitGather(pCsr->pExpr, pInfo);
+ break;
+ }
+
+ default: {
+ Fts3Expr *pExpr;
+ assert( zArg[i]==FTS3_MATCHINFO_HITS );
+ pExpr = pCsr->pExpr;
+ rc = fts3ExprLoadDoclists(pCsr, 0, 0);
+ if( rc!=SQLITE_OK ) break;
+ if( bGlobal ){
+ if( pCsr->pDeferred ){
+ rc = fts3MatchinfoSelectDoctotal(pTab, &pSelect, &pInfo->nDoc,0,0);
+ if( rc!=SQLITE_OK ) break;
+ }
+ rc = sqlite3Fts3ExprIterate(pExpr, fts3ExprGlobalHitsCb,(void*)pInfo);
+ sqlite3Fts3EvalTestDeferred(pCsr, &rc);
+ if( rc!=SQLITE_OK ) break;
+ }
+ (void)sqlite3Fts3ExprIterate(pExpr, fts3ExprLocalHitsCb,(void*)pInfo);
+ break;
+ }
+ }
+
+ pInfo->aMatchinfo += fts3MatchinfoSize(pInfo, zArg[i]);
+ }
+
+ sqlite3_reset(pSelect);
+ return rc;
+}
+
+
+/*
+** Populate pCsr->aMatchinfo[] with data for the current row. The
+** 'matchinfo' data is an array of 32-bit unsigned integers (C type u32).
+*/
+static void fts3GetMatchinfo(
+ sqlite3_context *pCtx, /* Return results here */
+ Fts3Cursor *pCsr, /* FTS3 Cursor object */
+ const char *zArg /* Second argument to matchinfo() function */
+){
+ MatchInfo sInfo;
+ Fts3Table *pTab = (Fts3Table *)pCsr->base.pVtab;
+ int rc = SQLITE_OK;
+ int bGlobal = 0; /* Collect 'global' stats as well as local */
+
+ u32 *aOut = 0;
+ void (*xDestroyOut)(void*) = 0;
+
+ memset(&sInfo, 0, sizeof(MatchInfo));
+ sInfo.pCursor = pCsr;
+ sInfo.nCol = pTab->nColumn;
+
+ /* If there is cached matchinfo() data, but the format string for the
+ ** cache does not match the format string for this request, discard
+ ** the cached data. */
+ if( pCsr->pMIBuffer && strcmp(pCsr->pMIBuffer->zMatchinfo, zArg) ){
+ sqlite3Fts3MIBufferFree(pCsr->pMIBuffer);
+ pCsr->pMIBuffer = 0;
+ }
+
+ /* If Fts3Cursor.pMIBuffer is NULL, then this is the first time the
+ ** matchinfo function has been called for this query. In this case
+ ** allocate the array used to accumulate the matchinfo data and
+ ** initialize those elements that are constant for every row.
+ */
+ if( pCsr->pMIBuffer==0 ){
+ size_t nMatchinfo = 0; /* Number of u32 elements in match-info */
+ int i; /* Used to iterate through zArg */
+
+ /* Determine the number of phrases in the query */
+ pCsr->nPhrase = fts3ExprPhraseCount(pCsr->pExpr);
+ sInfo.nPhrase = pCsr->nPhrase;
+
+ /* Determine the number of integers in the buffer returned by this call. */
+ for(i=0; zArg[i]; i++){
+ char *zErr = 0;
+ if( fts3MatchinfoCheck(pTab, zArg[i], &zErr) ){
+ sqlite3_result_error(pCtx, zErr, -1);
+ sqlite3_free(zErr);
+ return;
+ }
+ nMatchinfo += fts3MatchinfoSize(&sInfo, zArg[i]);
+ }
+
+ /* Allocate space for Fts3Cursor.aMatchinfo[] and Fts3Cursor.zMatchinfo. */
+ pCsr->pMIBuffer = fts3MIBufferNew(nMatchinfo, zArg);
+ if( !pCsr->pMIBuffer ) rc = SQLITE_NOMEM;
+
+ pCsr->isMatchinfoNeeded = 1;
+ bGlobal = 1;
+ }
+
+ if( rc==SQLITE_OK ){
+ xDestroyOut = fts3MIBufferAlloc(pCsr->pMIBuffer, &aOut);
+ if( xDestroyOut==0 ){
+ rc = SQLITE_NOMEM;
+ }
+ }
+
+ if( rc==SQLITE_OK ){
+ sInfo.aMatchinfo = aOut;
+ sInfo.nPhrase = pCsr->nPhrase;
+ rc = fts3MatchinfoValues(pCsr, bGlobal, &sInfo, zArg);
+ if( bGlobal ){
+ fts3MIBufferSetGlobal(pCsr->pMIBuffer);
+ }
+ }
+
+ if( rc!=SQLITE_OK ){
+ sqlite3_result_error_code(pCtx, rc);
+ if( xDestroyOut ) xDestroyOut(aOut);
+ }else{
+ int n = pCsr->pMIBuffer->nElem * sizeof(u32);
+ sqlite3_result_blob(pCtx, aOut, n, xDestroyOut);
+ }
+}
+
+/*
+** Implementation of snippet() function.
+*/
+void sqlite3Fts3Snippet(
+ sqlite3_context *pCtx, /* SQLite function call context */
+ Fts3Cursor *pCsr, /* Cursor object */
+ const char *zStart, /* Snippet start text - "" */
+ const char *zEnd, /* Snippet end text - "" */
+ const char *zEllipsis, /* Snippet ellipsis text - "..." */
+ int iCol, /* Extract snippet from this column */
+ int nToken /* Approximate number of tokens in snippet */
+){
+ Fts3Table *pTab = (Fts3Table *)pCsr->base.pVtab;
+ int rc = SQLITE_OK;
+ int i;
+ StrBuffer res = {0, 0, 0};
+
+ /* The returned text includes up to four fragments of text extracted from
+ ** the data in the current row. The first iteration of the for(...) loop
+ ** below attempts to locate a single fragment of text nToken tokens in
+ ** size that contains at least one instance of all phrases in the query
+ ** expression that appear in the current row. If such a fragment of text
+ ** cannot be found, the second iteration of the loop attempts to locate
+ ** a pair of fragments, and so on.
+ */
+ int nSnippet = 0; /* Number of fragments in this snippet */
+ SnippetFragment aSnippet[4]; /* Maximum of 4 fragments per snippet */
+ int nFToken = -1; /* Number of tokens in each fragment */
+
+ if( !pCsr->pExpr ){
+ sqlite3_result_text(pCtx, "", 0, SQLITE_STATIC);
+ return;
+ }
+
+ /* Limit the snippet length to 64 tokens. */
+ if( nToken<-64 ) nToken = -64;
+ if( nToken>+64 ) nToken = +64;
+
+ for(nSnippet=1; 1; nSnippet++){
+
+ int iSnip; /* Loop counter 0..nSnippet-1 */
+ u64 mCovered = 0; /* Bitmask of phrases covered by snippet */
+ u64 mSeen = 0; /* Bitmask of phrases seen by BestSnippet() */
+
+ if( nToken>=0 ){
+ nFToken = (nToken+nSnippet-1) / nSnippet;
+ }else{
+ nFToken = -1 * nToken;
+ }
+
+ for(iSnip=0; iSnipnColumn; iRead++){
+ SnippetFragment sF = {0, 0, 0, 0};
+ int iS = 0;
+ if( iCol>=0 && iRead!=iCol ) continue;
+
+ /* Find the best snippet of nFToken tokens in column iRead. */
+ rc = fts3BestSnippet(nFToken, pCsr, iRead, mCovered, &mSeen, &sF, &iS);
+ if( rc!=SQLITE_OK ){
+ goto snippet_out;
+ }
+ if( iS>iBestScore ){
+ *pFragment = sF;
+ iBestScore = iS;
+ }
+ }
+
+ mCovered |= pFragment->covered;
+ }
+
+ /* If all query phrases seen by fts3BestSnippet() are present in at least
+ ** one of the nSnippet snippet fragments, break out of the loop.
+ */
+ assert( (mCovered&mSeen)==mCovered );
+ if( mSeen==mCovered || nSnippet==SizeofArray(aSnippet) ) break;
+ }
+
+ assert( nFToken>0 );
+
+ for(i=0; ipCsr, pExpr, p->iCol, &pList);
+ nTerm = pExpr->pPhrase->nToken;
+ if( pList ){
+ fts3GetDeltaPosition(&pList, &iPos);
+ assert_fts3_nc( iPos>=0 );
+ }
+
+ for(iTerm=0; iTermaTerm[p->iTerm++];
+ pT->iOff = nTerm-iTerm-1;
+ pT->pList = pList;
+ pT->iPos = iPos;
+ }
+
+ return rc;
+}
+
+/*
+** Implementation of offsets() function.
+*/
+void sqlite3Fts3Offsets(
+ sqlite3_context *pCtx, /* SQLite function call context */
+ Fts3Cursor *pCsr /* Cursor object */
+){
+ Fts3Table *pTab = (Fts3Table *)pCsr->base.pVtab;
+ sqlite3_tokenizer_module const *pMod = pTab->pTokenizer->pModule;
+ int rc; /* Return Code */
+ int nToken; /* Number of tokens in query */
+ int iCol; /* Column currently being processed */
+ StrBuffer res = {0, 0, 0}; /* Result string */
+ TermOffsetCtx sCtx; /* Context for fts3ExprTermOffsetInit() */
+
+ if( !pCsr->pExpr ){
+ sqlite3_result_text(pCtx, "", 0, SQLITE_STATIC);
+ return;
+ }
+
+ memset(&sCtx, 0, sizeof(sCtx));
+ assert( pCsr->isRequireSeek==0 );
+
+ /* Count the number of terms in the query */
+ rc = fts3ExprLoadDoclists(pCsr, 0, &nToken);
+ if( rc!=SQLITE_OK ) goto offsets_out;
+
+ /* Allocate the array of TermOffset iterators. */
+ sCtx.aTerm = (TermOffset *)sqlite3Fts3MallocZero(sizeof(TermOffset)*nToken);
+ if( 0==sCtx.aTerm ){
+ rc = SQLITE_NOMEM;
+ goto offsets_out;
+ }
+ sCtx.iDocid = pCsr->iPrevId;
+ sCtx.pCsr = pCsr;
+
+ /* Loop through the table columns, appending offset information to
+ ** string-buffer res for each column.
+ */
+ for(iCol=0; iColnColumn; iCol++){
+ sqlite3_tokenizer_cursor *pC; /* Tokenizer cursor */
+ const char *ZDUMMY; /* Dummy argument used with xNext() */
+ int NDUMMY = 0; /* Dummy argument used with xNext() */
+ int iStart = 0;
+ int iEnd = 0;
+ int iCurrent = 0;
+ const char *zDoc;
+ int nDoc;
+
+ /* Initialize the contents of sCtx.aTerm[] for column iCol. This
+ ** operation may fail if the database contains corrupt records.
+ */
+ sCtx.iCol = iCol;
+ sCtx.iTerm = 0;
+ rc = sqlite3Fts3ExprIterate(
+ pCsr->pExpr, fts3ExprTermOffsetInit, (void*)&sCtx
+ );
+ if( rc!=SQLITE_OK ) goto offsets_out;
+
+ /* Retreive the text stored in column iCol. If an SQL NULL is stored
+ ** in column iCol, jump immediately to the next iteration of the loop.
+ ** If an OOM occurs while retrieving the data (this can happen if SQLite
+ ** needs to transform the data from utf-16 to utf-8), return SQLITE_NOMEM
+ ** to the caller.
+ */
+ zDoc = (const char *)sqlite3_column_text(pCsr->pStmt, iCol+1);
+ nDoc = sqlite3_column_bytes(pCsr->pStmt, iCol+1);
+ if( zDoc==0 ){
+ if( sqlite3_column_type(pCsr->pStmt, iCol+1)==SQLITE_NULL ){
+ continue;
+ }
+ rc = SQLITE_NOMEM;
+ goto offsets_out;
+ }
+
+ /* Initialize a tokenizer iterator to iterate through column iCol. */
+ rc = sqlite3Fts3OpenTokenizer(pTab->pTokenizer, pCsr->iLangid,
+ zDoc, nDoc, &pC
+ );
+ if( rc!=SQLITE_OK ) goto offsets_out;
+
+ rc = pMod->xNext(pC, &ZDUMMY, &NDUMMY, &iStart, &iEnd, &iCurrent);
+ while( rc==SQLITE_OK ){
+ int i; /* Used to loop through terms */
+ int iMinPos = 0x7FFFFFFF; /* Position of next token */
+ TermOffset *pTerm = 0; /* TermOffset associated with next token */
+
+ for(i=0; ipList && (pT->iPos-pT->iOff)iPos-pT->iOff;
+ pTerm = pT;
+ }
+ }
+
+ if( !pTerm ){
+ /* All offsets for this column have been gathered. */
+ rc = SQLITE_DONE;
+ }else{
+ assert_fts3_nc( iCurrent<=iMinPos );
+ if( 0==(0xFE&*pTerm->pList) ){
+ pTerm->pList = 0;
+ }else{
+ fts3GetDeltaPosition(&pTerm->pList, &pTerm->iPos);
+ }
+ while( rc==SQLITE_OK && iCurrentxNext(pC, &ZDUMMY, &NDUMMY, &iStart, &iEnd, &iCurrent);
+ }
+ if( rc==SQLITE_OK ){
+ char aBuffer[64];
+ sqlite3_snprintf(sizeof(aBuffer), aBuffer,
+ "%d %d %d %d ", iCol, pTerm-sCtx.aTerm, iStart, iEnd-iStart
+ );
+ rc = fts3StringAppend(&res, aBuffer, -1);
+ }else if( rc==SQLITE_DONE && pTab->zContentTbl==0 ){
+ rc = FTS_CORRUPT_VTAB;
+ }
+ }
+ }
+ if( rc==SQLITE_DONE ){
+ rc = SQLITE_OK;
+ }
+
+ pMod->xClose(pC);
+ if( rc!=SQLITE_OK ) goto offsets_out;
+ }
+
+ offsets_out:
+ sqlite3_free(sCtx.aTerm);
+ assert( rc!=SQLITE_DONE );
+ sqlite3Fts3SegmentsClose(pTab);
+ if( rc!=SQLITE_OK ){
+ sqlite3_result_error_code(pCtx, rc);
+ sqlite3_free(res.z);
+ }else{
+ sqlite3_result_text(pCtx, res.z, res.n-1, sqlite3_free);
+ }
+ return;
+}
+
+/*
+** Implementation of matchinfo() function.
+*/
+void sqlite3Fts3Matchinfo(
+ sqlite3_context *pContext, /* Function call context */
+ Fts3Cursor *pCsr, /* FTS3 table cursor */
+ const char *zArg /* Second arg to matchinfo() function */
+){
+ Fts3Table *pTab = (Fts3Table *)pCsr->base.pVtab;
+ const char *zFormat;
+
+ if( zArg ){
+ zFormat = zArg;
+ }else{
+ zFormat = FTS3_MATCHINFO_DEFAULT;
+ }
+
+ if( !pCsr->pExpr ){
+ sqlite3_result_blob(pContext, "", 0, SQLITE_STATIC);
+ return;
+ }else{
+ /* Retrieve matchinfo() data. */
+ fts3GetMatchinfo(pContext, pCsr, zFormat);
+ sqlite3Fts3SegmentsClose(pTab);
+ }
+}
+
+#endif
diff --git a/ext/fts3/fts3_term.c b/ext/fts3/fts3_term.c
new file mode 100644
index 0000000000000000000000000000000000000000..655dd9f35a82a0804cb69f54e32e9bce7e7a091c
--- /dev/null
+++ b/ext/fts3/fts3_term.c
@@ -0,0 +1,383 @@
+/*
+** 2011 Jan 27
+**
+** The author disclaims copyright to this source code. In place of
+** a legal notice, here is a blessing:
+**
+** May you do good and not evil.
+** May you find forgiveness for yourself and forgive others.
+** May you share freely, never taking more than you give.
+**
+******************************************************************************
+**
+** This file is not part of the production FTS code. It is only used for
+** testing. It contains a virtual table implementation that provides direct
+** access to the full-text index of an FTS table.
+*/
+
+#include "fts3Int.h"
+#if !defined(SQLITE_CORE) || defined(SQLITE_ENABLE_FTS3)
+#ifdef SQLITE_TEST
+
+#include
+#include
+#include
+
+typedef struct Fts3termTable Fts3termTable;
+typedef struct Fts3termCursor Fts3termCursor;
+
+struct Fts3termTable {
+ sqlite3_vtab base; /* Base class used by SQLite core */
+ int iIndex; /* Index for Fts3Table.aIndex[] */
+ Fts3Table *pFts3Tab;
+};
+
+struct Fts3termCursor {
+ sqlite3_vtab_cursor base; /* Base class used by SQLite core */
+ Fts3MultiSegReader csr; /* Must be right after "base" */
+ Fts3SegFilter filter;
+
+ int isEof; /* True if cursor is at EOF */
+ char *pNext;
+
+ sqlite3_int64 iRowid; /* Current 'rowid' value */
+ sqlite3_int64 iDocid; /* Current 'docid' value */
+ int iCol; /* Current 'col' value */
+ int iPos; /* Current 'pos' value */
+};
+
+/*
+** Schema of the terms table.
+*/
+#define FTS3_TERMS_SCHEMA "CREATE TABLE x(term, docid, col, pos)"
+
+/*
+** This function does all the work for both the xConnect and xCreate methods.
+** These tables have no persistent representation of their own, so xConnect
+** and xCreate are identical operations.
+*/
+static int fts3termConnectMethod(
+ sqlite3 *db, /* Database connection */
+ void *pCtx, /* Non-zero for an fts4prefix table */
+ int argc, /* Number of elements in argv array */
+ const char * const *argv, /* xCreate/xConnect argument array */
+ sqlite3_vtab **ppVtab, /* OUT: New sqlite3_vtab object */
+ char **pzErr /* OUT: sqlite3_malloc'd error message */
+){
+ char const *zDb; /* Name of database (e.g. "main") */
+ char const *zFts3; /* Name of fts3 table */
+ int nDb; /* Result of strlen(zDb) */
+ int nFts3; /* Result of strlen(zFts3) */
+ sqlite3_int64 nByte; /* Bytes of space to allocate here */
+ int rc; /* value returned by declare_vtab() */
+ Fts3termTable *p; /* Virtual table object to return */
+ int iIndex = 0;
+
+ UNUSED_PARAMETER(pCtx);
+ if( argc==5 ){
+ iIndex = atoi(argv[4]);
+ argc--;
+ }
+
+ *ppVtab = 0;
+
+ /* The user should specify a single argument - the name of an fts3 table. */
+ if( argc!=4 ){
+ sqlite3Fts3ErrMsg(pzErr,
+ "wrong number of arguments to fts4term constructor"
+ );
+ return SQLITE_ERROR;
+ }
+
+ zDb = argv[1];
+ nDb = (int)strlen(zDb);
+ zFts3 = argv[3];
+ nFts3 = (int)strlen(zFts3);
+
+ rc = sqlite3_declare_vtab(db, FTS3_TERMS_SCHEMA);
+ if( rc!=SQLITE_OK ) return rc;
+
+ nByte = sizeof(Fts3termTable);
+ p = (Fts3termTable *)sqlite3Fts3MallocZero(nByte);
+ if( !p ) return SQLITE_NOMEM;
+
+ p->pFts3Tab = (Fts3Table*)sqlite3Fts3MallocZero(
+ sizeof(Fts3Table) + nDb + nFts3 + 2
+ );
+ if( p->pFts3Tab==0 ){
+ sqlite3_free(p);
+ return SQLITE_NOMEM;
+ }
+ p->pFts3Tab->zDb = (char *)&p->pFts3Tab[1];
+ p->pFts3Tab->zName = &p->pFts3Tab->zDb[nDb+1];
+ p->pFts3Tab->db = db;
+ p->pFts3Tab->nIndex = iIndex+1;
+ p->iIndex = iIndex;
+
+ memcpy((char *)p->pFts3Tab->zDb, zDb, nDb);
+ memcpy((char *)p->pFts3Tab->zName, zFts3, nFts3);
+ sqlite3Fts3Dequote((char *)p->pFts3Tab->zName);
+
+ *ppVtab = (sqlite3_vtab *)p;
+ return SQLITE_OK;
+}
+
+/*
+** This function does the work for both the xDisconnect and xDestroy methods.
+** These tables have no persistent representation of their own, so xDisconnect
+** and xDestroy are identical operations.
+*/
+static int fts3termDisconnectMethod(sqlite3_vtab *pVtab){
+ Fts3termTable *p = (Fts3termTable *)pVtab;
+ Fts3Table *pFts3 = p->pFts3Tab;
+ int i;
+
+ /* Free any prepared statements held */
+ for(i=0; iaStmt); i++){
+ sqlite3_finalize(pFts3->aStmt[i]);
+ }
+ sqlite3_free(pFts3->zSegmentsTbl);
+ sqlite3_free(pFts3);
+ sqlite3_free(p);
+ return SQLITE_OK;
+}
+
+#define FTS4AUX_EQ_CONSTRAINT 1
+#define FTS4AUX_GE_CONSTRAINT 2
+#define FTS4AUX_LE_CONSTRAINT 4
+
+/*
+** xBestIndex - Analyze a WHERE and ORDER BY clause.
+*/
+static int fts3termBestIndexMethod(
+ sqlite3_vtab *pVTab,
+ sqlite3_index_info *pInfo
+){
+ UNUSED_PARAMETER(pVTab);
+
+ /* This vtab naturally does "ORDER BY term, docid, col, pos". */
+ if( pInfo->nOrderBy ){
+ int i;
+ for(i=0; inOrderBy; i++){
+ if( pInfo->aOrderBy[i].iColumn!=i || pInfo->aOrderBy[i].desc ) break;
+ }
+ if( i==pInfo->nOrderBy ){
+ pInfo->orderByConsumed = 1;
+ }
+ }
+
+ return SQLITE_OK;
+}
+
+/*
+** xOpen - Open a cursor.
+*/
+static int fts3termOpenMethod(sqlite3_vtab *pVTab, sqlite3_vtab_cursor **ppCsr){
+ Fts3termCursor *pCsr; /* Pointer to cursor object to return */
+
+ UNUSED_PARAMETER(pVTab);
+
+ pCsr = (Fts3termCursor *)sqlite3_malloc(sizeof(Fts3termCursor));
+ if( !pCsr ) return SQLITE_NOMEM;
+ memset(pCsr, 0, sizeof(Fts3termCursor));
+
+ *ppCsr = (sqlite3_vtab_cursor *)pCsr;
+ return SQLITE_OK;
+}
+
+/*
+** xClose - Close a cursor.
+*/
+static int fts3termCloseMethod(sqlite3_vtab_cursor *pCursor){
+ Fts3Table *pFts3 = ((Fts3termTable *)pCursor->pVtab)->pFts3Tab;
+ Fts3termCursor *pCsr = (Fts3termCursor *)pCursor;
+
+ sqlite3Fts3SegmentsClose(pFts3);
+ sqlite3Fts3SegReaderFinish(&pCsr->csr);
+ sqlite3_free(pCsr);
+ return SQLITE_OK;
+}
+
+/*
+** xNext - Advance the cursor to the next row, if any.
+*/
+static int fts3termNextMethod(sqlite3_vtab_cursor *pCursor){
+ Fts3termCursor *pCsr = (Fts3termCursor *)pCursor;
+ Fts3Table *pFts3 = ((Fts3termTable *)pCursor->pVtab)->pFts3Tab;
+ int rc;
+ sqlite3_int64 v;
+
+ /* Increment our pretend rowid value. */
+ pCsr->iRowid++;
+
+ /* Advance to the next term in the full-text index. */
+ if( pCsr->csr.aDoclist==0
+ || pCsr->pNext>=&pCsr->csr.aDoclist[pCsr->csr.nDoclist-1]
+ ){
+ rc = sqlite3Fts3SegReaderStep(pFts3, &pCsr->csr);
+ if( rc!=SQLITE_ROW ){
+ pCsr->isEof = 1;
+ return rc;
+ }
+
+ pCsr->iCol = 0;
+ pCsr->iPos = 0;
+ pCsr->iDocid = 0;
+ pCsr->pNext = pCsr->csr.aDoclist;
+
+ /* Read docid */
+ pCsr->pNext += sqlite3Fts3GetVarint(pCsr->pNext, &pCsr->iDocid);
+ }
+
+ pCsr->pNext += sqlite3Fts3GetVarint(pCsr->pNext, &v);
+ if( v==0 ){
+ pCsr->pNext += sqlite3Fts3GetVarint(pCsr->pNext, &v);
+ pCsr->iDocid += v;
+ pCsr->pNext += sqlite3Fts3GetVarint(pCsr->pNext, &v);
+ pCsr->iCol = 0;
+ pCsr->iPos = 0;
+ }
+
+ if( v==1 ){
+ pCsr->pNext += sqlite3Fts3GetVarint(pCsr->pNext, &v);
+ pCsr->iCol += (int)v;
+ pCsr->iPos = 0;
+ pCsr->pNext += sqlite3Fts3GetVarint(pCsr->pNext, &v);
+ }
+
+ pCsr->iPos += (int)(v - 2);
+
+ return SQLITE_OK;
+}
+
+/*
+** xFilter - Initialize a cursor to point at the start of its data.
+*/
+static int fts3termFilterMethod(
+ sqlite3_vtab_cursor *pCursor, /* The cursor used for this query */
+ int idxNum, /* Strategy index */
+ const char *idxStr, /* Unused */
+ int nVal, /* Number of elements in apVal */
+ sqlite3_value **apVal /* Arguments for the indexing scheme */
+){
+ Fts3termCursor *pCsr = (Fts3termCursor *)pCursor;
+ Fts3termTable *p = (Fts3termTable *)pCursor->pVtab;
+ Fts3Table *pFts3 = p->pFts3Tab;
+ int rc;
+
+ UNUSED_PARAMETER(nVal);
+ UNUSED_PARAMETER(idxNum);
+ UNUSED_PARAMETER(idxStr);
+ UNUSED_PARAMETER(apVal);
+
+ assert( idxStr==0 && idxNum==0 );
+
+ /* In case this cursor is being reused, close and zero it. */
+ testcase(pCsr->filter.zTerm);
+ sqlite3Fts3SegReaderFinish(&pCsr->csr);
+ memset(&pCsr->csr, 0, ((u8*)&pCsr[1]) - (u8*)&pCsr->csr);
+
+ pCsr->filter.flags = FTS3_SEGMENT_REQUIRE_POS|FTS3_SEGMENT_IGNORE_EMPTY;
+ pCsr->filter.flags |= FTS3_SEGMENT_SCAN;
+
+ rc = sqlite3Fts3SegReaderCursor(pFts3, 0, p->iIndex, FTS3_SEGCURSOR_ALL,
+ pCsr->filter.zTerm, pCsr->filter.nTerm, 0, 1, &pCsr->csr
+ );
+ if( rc==SQLITE_OK ){
+ rc = sqlite3Fts3SegReaderStart(pFts3, &pCsr->csr, &pCsr->filter);
+ }
+ if( rc==SQLITE_OK ){
+ rc = fts3termNextMethod(pCursor);
+ }
+ return rc;
+}
+
+/*
+** xEof - Return true if the cursor is at EOF, or false otherwise.
+*/
+static int fts3termEofMethod(sqlite3_vtab_cursor *pCursor){
+ Fts3termCursor *pCsr = (Fts3termCursor *)pCursor;
+ return pCsr->isEof;
+}
+
+/*
+** xColumn - Return a column value.
+*/
+static int fts3termColumnMethod(
+ sqlite3_vtab_cursor *pCursor, /* Cursor to retrieve value from */
+ sqlite3_context *pCtx, /* Context for sqlite3_result_xxx() calls */
+ int iCol /* Index of column to read value from */
+){
+ Fts3termCursor *p = (Fts3termCursor *)pCursor;
+
+ assert( iCol>=0 && iCol<=3 );
+ switch( iCol ){
+ case 0:
+ sqlite3_result_text(pCtx, p->csr.zTerm, p->csr.nTerm, SQLITE_TRANSIENT);
+ break;
+ case 1:
+ sqlite3_result_int64(pCtx, p->iDocid);
+ break;
+ case 2:
+ sqlite3_result_int64(pCtx, p->iCol);
+ break;
+ default:
+ sqlite3_result_int64(pCtx, p->iPos);
+ break;
+ }
+
+ return SQLITE_OK;
+}
+
+/*
+** xRowid - Return the current rowid for the cursor.
+*/
+static int fts3termRowidMethod(
+ sqlite3_vtab_cursor *pCursor, /* Cursor to retrieve value from */
+ sqlite_int64 *pRowid /* OUT: Rowid value */
+){
+ Fts3termCursor *pCsr = (Fts3termCursor *)pCursor;
+ *pRowid = pCsr->iRowid;
+ return SQLITE_OK;
+}
+
+/*
+** Register the fts3term module with database connection db. Return SQLITE_OK
+** if successful or an error code if sqlite3_create_module() fails.
+*/
+int sqlite3Fts3InitTerm(sqlite3 *db){
+ static const sqlite3_module fts3term_module = {
+ 0, /* iVersion */
+ fts3termConnectMethod, /* xCreate */
+ fts3termConnectMethod, /* xConnect */
+ fts3termBestIndexMethod, /* xBestIndex */
+ fts3termDisconnectMethod, /* xDisconnect */
+ fts3termDisconnectMethod, /* xDestroy */
+ fts3termOpenMethod, /* xOpen */
+ fts3termCloseMethod, /* xClose */
+ fts3termFilterMethod, /* xFilter */
+ fts3termNextMethod, /* xNext */
+ fts3termEofMethod, /* xEof */
+ fts3termColumnMethod, /* xColumn */
+ fts3termRowidMethod, /* xRowid */
+ 0, /* xUpdate */
+ 0, /* xBegin */
+ 0, /* xSync */
+ 0, /* xCommit */
+ 0, /* xRollback */
+ 0, /* xFindFunction */
+ 0, /* xRename */
+ 0, /* xSavepoint */
+ 0, /* xRelease */
+ 0, /* xRollbackTo */
+ 0, /* xShadowName */
+ 0 /* xIntegrity */
+ };
+ int rc; /* Return code */
+
+ rc = sqlite3_create_module(db, "fts4term", &fts3term_module, 0);
+ return rc;
+}
+
+#endif
+#endif /* !defined(SQLITE_CORE) || defined(SQLITE_ENABLE_FTS3) */
diff --git a/ext/fts3/fts3_test.c b/ext/fts3/fts3_test.c
new file mode 100644
index 0000000000000000000000000000000000000000..3c42a7bf02963c8430d99eae92ad988f073c4b69
--- /dev/null
+++ b/ext/fts3/fts3_test.c
@@ -0,0 +1,620 @@
+/*
+** 2011 Jun 13
+**
+** The author disclaims copyright to this source code. In place of
+** a legal notice, here is a blessing:
+**
+** May you do good and not evil.
+** May you find forgiveness for yourself and forgive others.
+** May you share freely, never taking more than you give.
+**
+******************************************************************************
+**
+** This file is not part of the production FTS code. It is only used for
+** testing. It contains a Tcl command that can be used to test if a document
+** matches an FTS NEAR expression.
+**
+** As of March 2012, it also contains a version 1 tokenizer used for testing
+** that the sqlite3_tokenizer_module.xLanguage() method is invoked correctly.
+*/
+
+#include "tclsqlite.h"
+#include
+#include
+
+#if defined(SQLITE_TEST)
+#if defined(SQLITE_ENABLE_FTS3) || defined(SQLITE_ENABLE_FTS4)
+
+/* Required so that the "ifdef SQLITE_ENABLE_FTS3" below works */
+#include "fts3Int.h"
+
+#define NM_MAX_TOKEN 12
+
+typedef struct NearPhrase NearPhrase;
+typedef struct NearDocument NearDocument;
+typedef struct NearToken NearToken;
+
+struct NearDocument {
+ int nToken; /* Length of token in bytes */
+ NearToken *aToken; /* Token array */
+};
+
+struct NearToken {
+ int n; /* Length of token in bytes */
+ const char *z; /* Pointer to token string */
+};
+
+struct NearPhrase {
+ int nNear; /* Preceding NEAR value */
+ int nToken; /* Number of tokens in this phrase */
+ NearToken aToken[NM_MAX_TOKEN]; /* Array of tokens in this phrase */
+};
+
+static int nm_phrase_match(
+ NearPhrase *p,
+ NearToken *aToken
+){
+ int ii;
+
+ for(ii=0; iinToken; ii++){
+ NearToken *pToken = &p->aToken[ii];
+ if( pToken->n>0 && pToken->z[pToken->n-1]=='*' ){
+ if( aToken[ii].n<(pToken->n-1) ) return 0;
+ if( memcmp(aToken[ii].z, pToken->z, pToken->n-1) ) return 0;
+ }else{
+ if( aToken[ii].n!=pToken->n ) return 0;
+ if( memcmp(aToken[ii].z, pToken->z, pToken->n) ) return 0;
+ }
+ }
+
+ return 1;
+}
+
+static int nm_near_chain(
+ int iDir, /* Direction to iterate through aPhrase[] */
+ NearDocument *pDoc, /* Document to match against */
+ int iPos, /* Position at which iPhrase was found */
+ int nPhrase, /* Size of phrase array */
+ NearPhrase *aPhrase, /* Phrase array */
+ int iPhrase /* Index of phrase found */
+){
+ int iStart;
+ int iStop;
+ int ii;
+ int nNear;
+ int iPhrase2;
+ NearPhrase *p;
+ NearPhrase *pPrev;
+
+ assert( iDir==1 || iDir==-1 );
+
+ if( iDir==1 ){
+ if( (iPhrase+1)==nPhrase ) return 1;
+ nNear = aPhrase[iPhrase+1].nNear;
+ }else{
+ if( iPhrase==0 ) return 1;
+ nNear = aPhrase[iPhrase].nNear;
+ }
+ pPrev = &aPhrase[iPhrase];
+ iPhrase2 = iPhrase+iDir;
+ p = &aPhrase[iPhrase2];
+
+ iStart = iPos - nNear - p->nToken;
+ iStop = iPos + nNear + pPrev->nToken;
+
+ if( iStart<0 ) iStart = 0;
+ if( iStop > pDoc->nToken - p->nToken ) iStop = pDoc->nToken - p->nToken;
+
+ for(ii=iStart; ii<=iStop; ii++){
+ if( nm_phrase_match(p, &pDoc->aToken[ii]) ){
+ if( nm_near_chain(iDir, pDoc, ii, nPhrase, aPhrase, iPhrase2) ) return 1;
+ }
+ }
+
+ return 0;
+}
+
+static int nm_match_count(
+ NearDocument *pDoc, /* Document to match against */
+ int nPhrase, /* Size of phrase array */
+ NearPhrase *aPhrase, /* Phrase array */
+ int iPhrase /* Index of phrase to count matches for */
+){
+ int nOcc = 0;
+ int ii;
+ NearPhrase *p = &aPhrase[iPhrase];
+
+ for(ii=0; ii<(pDoc->nToken + 1 - p->nToken); ii++){
+ if( nm_phrase_match(p, &pDoc->aToken[ii]) ){
+ /* Test forward NEAR chain (i>iPhrase) */
+ if( 0==nm_near_chain(1, pDoc, ii, nPhrase, aPhrase, iPhrase) ) continue;
+
+ /* Test reverse NEAR chain (iNM_MAX_TOKEN ){
+ Tcl_AppendResult(interp, "Too many tokens in phrase", 0);
+ rc = TCL_ERROR;
+ goto near_match_out;
+ }
+ for(jj=0; jj<(int)nToken; jj++){
+ NearToken *pT = &aPhrase[ii].aToken[jj];
+ pT->z = Tcl_GetStringFromObj(apToken[jj], &nn);
+ pT->n = (int)nn;
+ }
+ aPhrase[ii].nToken = (int)nToken;
+ }
+ for(ii=1; ii0));
+
+ near_match_out:
+ ckfree((char *)aPhrase);
+ ckfree((char *)doc.aToken);
+ return rc;
+}
+
+/*
+** Tclcmd: fts3_configure_incr_load ?CHUNKSIZE THRESHOLD?
+**
+** Normally, FTS uses hard-coded values to determine the minimum doclist
+** size eligible for incremental loading, and the size of the chunks loaded
+** when a doclist is incrementally loaded. This command allows the built-in
+** values to be overridden for testing purposes.
+**
+** If present, the first argument is the chunksize in bytes to load doclists
+** in. The second argument is the minimum doclist size in bytes to use
+** incremental loading with.
+**
+** Whether or not the arguments are present, this command returns a list of
+** two integers - the initial chunksize and threshold when the command is
+** invoked. This can be used to restore the default behavior after running
+** tests. For example:
+**
+** # Override incr-load settings for testing:
+** set cfg [fts3_configure_incr_load $new_chunksize $new_threshold]
+**
+** .... run tests ....
+**
+** # Restore initial incr-load settings:
+** eval fts3_configure_incr_load $cfg
+*/
+static int SQLITE_TCLAPI fts3_configure_incr_load_cmd(
+ ClientData clientData,
+ Tcl_Interp *interp,
+ int objc,
+ Tcl_Obj *CONST objv[]
+){
+#ifdef SQLITE_ENABLE_FTS3
+ extern int test_fts3_node_chunksize;
+ extern int test_fts3_node_chunk_threshold;
+ Tcl_Obj *pRet;
+
+ if( objc!=1 && objc!=3 ){
+ Tcl_WrongNumArgs(interp, 1, objv, "?CHUNKSIZE THRESHOLD?");
+ return TCL_ERROR;
+ }
+
+ pRet = Tcl_NewObj();
+ Tcl_IncrRefCount(pRet);
+ Tcl_ListObjAppendElement(
+ interp, pRet, Tcl_NewIntObj(test_fts3_node_chunksize));
+ Tcl_ListObjAppendElement(
+ interp, pRet, Tcl_NewIntObj(test_fts3_node_chunk_threshold));
+
+ if( objc==3 ){
+ int iArg1;
+ int iArg2;
+ if( Tcl_GetIntFromObj(interp, objv[1], &iArg1)
+ || Tcl_GetIntFromObj(interp, objv[2], &iArg2)
+ ){
+ Tcl_DecrRefCount(pRet);
+ return TCL_ERROR;
+ }
+ test_fts3_node_chunksize = iArg1;
+ test_fts3_node_chunk_threshold = iArg2;
+ }
+
+ Tcl_SetObjResult(interp, pRet);
+ Tcl_DecrRefCount(pRet);
+#endif
+ UNUSED_PARAMETER(clientData);
+ return TCL_OK;
+}
+
+#ifdef SQLITE_ENABLE_FTS3
+/**************************************************************************
+** Beginning of test tokenizer code.
+**
+** For language 0, this tokenizer is similar to the default 'simple'
+** tokenizer. For other languages L, the following:
+**
+** * Odd numbered languages are case-sensitive. Even numbered
+** languages are not.
+**
+** * Language ids 100 or greater are considered an error.
+**
+** The implementation assumes that the input contains only ASCII characters
+** (i.e. those that may be encoded in UTF-8 using a single byte).
+*/
+typedef struct test_tokenizer {
+ sqlite3_tokenizer base;
+} test_tokenizer;
+
+typedef struct test_tokenizer_cursor {
+ sqlite3_tokenizer_cursor base;
+ const char *aInput; /* Input being tokenized */
+ int nInput; /* Size of the input in bytes */
+ int iInput; /* Current offset in aInput */
+ int iToken; /* Index of next token to be returned */
+ char *aBuffer; /* Buffer containing current token */
+ int nBuffer; /* Number of bytes allocated at pToken */
+ int iLangid; /* Configured language id */
+} test_tokenizer_cursor;
+
+static int testTokenizerCreate(
+ int argc, const char * const *argv,
+ sqlite3_tokenizer **ppTokenizer
+){
+ test_tokenizer *pNew;
+ UNUSED_PARAMETER(argc);
+ UNUSED_PARAMETER(argv);
+
+ pNew = sqlite3_malloc(sizeof(test_tokenizer));
+ if( !pNew ) return SQLITE_NOMEM;
+ memset(pNew, 0, sizeof(test_tokenizer));
+
+ *ppTokenizer = (sqlite3_tokenizer *)pNew;
+ return SQLITE_OK;
+}
+
+static int testTokenizerDestroy(sqlite3_tokenizer *pTokenizer){
+ test_tokenizer *p = (test_tokenizer *)pTokenizer;
+ sqlite3_free(p);
+ return SQLITE_OK;
+}
+
+static int testTokenizerOpen(
+ sqlite3_tokenizer *pTokenizer, /* The tokenizer */
+ const char *pInput, int nBytes, /* String to be tokenized */
+ sqlite3_tokenizer_cursor **ppCursor /* OUT: Tokenization cursor */
+){
+ int rc = SQLITE_OK; /* Return code */
+ test_tokenizer_cursor *pCsr; /* New cursor object */
+
+ UNUSED_PARAMETER(pTokenizer);
+
+ pCsr = (test_tokenizer_cursor *)sqlite3_malloc(sizeof(test_tokenizer_cursor));
+ if( pCsr==0 ){
+ rc = SQLITE_NOMEM;
+ }else{
+ memset(pCsr, 0, sizeof(test_tokenizer_cursor));
+ pCsr->aInput = pInput;
+ if( nBytes<0 ){
+ pCsr->nInput = (int)strlen(pInput);
+ }else{
+ pCsr->nInput = nBytes;
+ }
+ }
+
+ *ppCursor = (sqlite3_tokenizer_cursor *)pCsr;
+ return rc;
+}
+
+static int testTokenizerClose(sqlite3_tokenizer_cursor *pCursor){
+ test_tokenizer_cursor *pCsr = (test_tokenizer_cursor *)pCursor;
+ sqlite3_free(pCsr->aBuffer);
+ sqlite3_free(pCsr);
+ return SQLITE_OK;
+}
+
+static int testIsTokenChar(char c){
+ return (c>='a' && c<='z') || (c>='A' && c<='Z');
+}
+static int testTolower(char c){
+ char ret = c;
+ if( ret>='A' && ret<='Z') ret = ret - ('A'-'a');
+ return ret;
+}
+
+static int testTokenizerNext(
+ sqlite3_tokenizer_cursor *pCursor, /* Cursor returned by testTokenizerOpen */
+ const char **ppToken, /* OUT: *ppToken is the token text */
+ int *pnBytes, /* OUT: Number of bytes in token */
+ int *piStartOffset, /* OUT: Starting offset of token */
+ int *piEndOffset, /* OUT: Ending offset of token */
+ int *piPosition /* OUT: Position integer of token */
+){
+ test_tokenizer_cursor *pCsr = (test_tokenizer_cursor *)pCursor;
+ int rc = SQLITE_OK;
+ const char *p;
+ const char *pEnd;
+
+ p = &pCsr->aInput[pCsr->iInput];
+ pEnd = &pCsr->aInput[pCsr->nInput];
+
+ /* Skip past any white-space */
+ assert( p<=pEnd );
+ while( ppCsr->nBuffer ){
+ sqlite3_free(pCsr->aBuffer);
+ pCsr->aBuffer = sqlite3_malloc64(nToken);
+ }
+ if( pCsr->aBuffer==0 ){
+ rc = SQLITE_NOMEM;
+ }else{
+ int i;
+
+ if( pCsr->iLangid & 0x00000001 ){
+ for(i=0; iaBuffer[i] = pToken[i];
+ }else{
+ for(i=0; iaBuffer[i] = (char)testTolower(pToken[i]);
+ }
+ pCsr->iToken++;
+ pCsr->iInput = (int)(p - pCsr->aInput);
+
+ *ppToken = pCsr->aBuffer;
+ *pnBytes = (int)nToken;
+ *piStartOffset = (int)(pToken - pCsr->aInput);
+ *piEndOffset = (int)(p - pCsr->aInput);
+ *piPosition = pCsr->iToken;
+ }
+ }
+
+ return rc;
+}
+
+static int testTokenizerLanguage(
+ sqlite3_tokenizer_cursor *pCursor,
+ int iLangid
+){
+ int rc = SQLITE_OK;
+ test_tokenizer_cursor *pCsr = (test_tokenizer_cursor *)pCursor;
+ pCsr->iLangid = iLangid;
+ if( pCsr->iLangid>=100 ){
+ rc = SQLITE_ERROR;
+ }
+ return rc;
+}
+#endif
+
+static int SQLITE_TCLAPI fts3_test_tokenizer_cmd(
+ ClientData clientData,
+ Tcl_Interp *interp,
+ int objc,
+ Tcl_Obj *CONST objv[]
+){
+#ifdef SQLITE_ENABLE_FTS3
+ static const sqlite3_tokenizer_module testTokenizerModule = {
+ 1,
+ testTokenizerCreate,
+ testTokenizerDestroy,
+ testTokenizerOpen,
+ testTokenizerClose,
+ testTokenizerNext,
+ testTokenizerLanguage
+ };
+ const sqlite3_tokenizer_module *pPtr = &testTokenizerModule;
+ if( objc!=1 ){
+ Tcl_WrongNumArgs(interp, 1, objv, "");
+ return TCL_ERROR;
+ }
+ Tcl_SetObjResult(interp, Tcl_NewByteArrayObj(
+ (const unsigned char *)&pPtr, sizeof(sqlite3_tokenizer_module *)
+ ));
+#endif
+ UNUSED_PARAMETER(clientData);
+ return TCL_OK;
+}
+
+static int SQLITE_TCLAPI fts3_test_varint_cmd(
+ ClientData clientData,
+ Tcl_Interp *interp,
+ int objc,
+ Tcl_Obj *CONST objv[]
+){
+#ifdef SQLITE_ENABLE_FTS3
+ char aBuf[24];
+ int rc;
+ Tcl_WideInt w;
+ sqlite3_int64 w2;
+ int nByte, nByte2;
+
+ if( objc!=2 ){
+ Tcl_WrongNumArgs(interp, 1, objv, "INTEGER");
+ return TCL_ERROR;
+ }
+
+ rc = Tcl_GetWideIntFromObj(interp, objv[1], &w);
+ if( rc!=TCL_OK ) return rc;
+
+ nByte = sqlite3Fts3PutVarint(aBuf, w);
+ nByte2 = sqlite3Fts3GetVarint(aBuf, &w2);
+ if( w!=w2 || nByte!=nByte2 ){
+ char *zErr = sqlite3_mprintf("error testing %lld", w);
+ Tcl_ResetResult(interp);
+ Tcl_AppendResult(interp, zErr, 0);
+ return TCL_ERROR;
+ }
+
+ if( w<=2147483647 && w>=0 ){
+ int i;
+ nByte2 = fts3GetVarint32(aBuf, &i);
+ if( (int)w!=i || nByte!=nByte2 ){
+ char *zErr = sqlite3_mprintf("error testing %lld (32-bit)", w);
+ Tcl_ResetResult(interp);
+ Tcl_AppendResult(interp, zErr, 0);
+ return TCL_ERROR;
+ }
+ }
+
+#endif
+ UNUSED_PARAMETER(clientData);
+ return TCL_OK;
+}
+
+/*
+** End of tokenizer code.
+**************************************************************************/
+
+/*
+** sqlite3_fts3_may_be_corrupt BOOLEAN
+**
+** Set or clear the global "may-be-corrupt" flag. Return the old value.
+*/
+static int SQLITE_TCLAPI fts3_may_be_corrupt(
+ void * clientData,
+ Tcl_Interp *interp,
+ int objc,
+ Tcl_Obj *CONST objv[]
+){
+#ifdef SQLITE_DEBUG
+ int bOld = sqlite3_fts3_may_be_corrupt;
+
+ if( objc!=2 && objc!=1 ){
+ Tcl_WrongNumArgs(interp, 1, objv, "?BOOLEAN?");
+ return TCL_ERROR;
+ }
+ if( objc==2 ){
+ int bNew;
+ if( Tcl_GetBooleanFromObj(interp, objv[1], &bNew) ) return TCL_ERROR;
+ sqlite3_fts3_may_be_corrupt = bNew;
+ }
+
+ Tcl_SetObjResult(interp, Tcl_NewIntObj(bOld));
+#endif
+ return TCL_OK;
+}
+
+int Sqlitetestfts3_Init(Tcl_Interp *interp){
+ Tcl_CreateObjCommand(interp, "fts3_near_match", fts3_near_match_cmd, 0, 0);
+ Tcl_CreateObjCommand(interp,
+ "fts3_configure_incr_load", fts3_configure_incr_load_cmd, 0, 0
+ );
+ Tcl_CreateObjCommand(
+ interp, "fts3_test_tokenizer", fts3_test_tokenizer_cmd, 0, 0
+ );
+ Tcl_CreateObjCommand(
+ interp, "fts3_test_varint", fts3_test_varint_cmd, 0, 0
+ );
+ Tcl_CreateObjCommand(
+ interp, "sqlite3_fts3_may_be_corrupt", fts3_may_be_corrupt, 0, 0
+ );
+ return TCL_OK;
+}
+#endif /* SQLITE_ENABLE_FTS3 || SQLITE_ENABLE_FTS4 */
+#endif /* ifdef SQLITE_TEST */
diff --git a/ext/fts3/fts3_tokenize_vtab.c b/ext/fts3/fts3_tokenize_vtab.c
new file mode 100644
index 0000000000000000000000000000000000000000..7e8d09bd4834e73cb13f8bdf808b1d55b98ee808
--- /dev/null
+++ b/ext/fts3/fts3_tokenize_vtab.c
@@ -0,0 +1,459 @@
+/*
+** 2013 Apr 22
+**
+** The author disclaims copyright to this source code. In place of
+** a legal notice, here is a blessing:
+**
+** May you do good and not evil.
+** May you find forgiveness for yourself and forgive others.
+** May you share freely, never taking more than you give.
+**
+******************************************************************************
+**
+** This file contains code for the "fts3tokenize" virtual table module.
+** An fts3tokenize virtual table is created as follows:
+**
+** CREATE VIRTUAL TABLE USING fts3tokenize(
+** , , ...
+** );
+**
+** The table created has the following schema:
+**
+** CREATE TABLE (input, token, start, end, position)
+**
+** When queried, the query must include a WHERE clause of type:
+**
+** input =
+**
+** The virtual table module tokenizes this , using the FTS3
+** tokenizer specified by the arguments to the CREATE VIRTUAL TABLE
+** statement and returns one row for each token in the result. With
+** fields set as follows:
+**
+** input: Always set to a copy of
+** token: A token from the input.
+** start: Byte offset of the token within the input .
+** end: Byte offset of the byte immediately following the end of the
+** token within the input string.
+** pos: Token offset of token within input.
+**
+*/
+#include "fts3Int.h"
+#if !defined(SQLITE_CORE) || defined(SQLITE_ENABLE_FTS3)
+
+#include
+#include
+
+typedef struct Fts3tokTable Fts3tokTable;
+typedef struct Fts3tokCursor Fts3tokCursor;
+
+/*
+** Virtual table structure.
+*/
+struct Fts3tokTable {
+ sqlite3_vtab base; /* Base class used by SQLite core */
+ const sqlite3_tokenizer_module *pMod;
+ sqlite3_tokenizer *pTok;
+};
+
+/*
+** Virtual table cursor structure.
+*/
+struct Fts3tokCursor {
+ sqlite3_vtab_cursor base; /* Base class used by SQLite core */
+ char *zInput; /* Input string */
+ sqlite3_tokenizer_cursor *pCsr; /* Cursor to iterate through zInput */
+ int iRowid; /* Current 'rowid' value */
+ const char *zToken; /* Current 'token' value */
+ int nToken; /* Size of zToken in bytes */
+ int iStart; /* Current 'start' value */
+ int iEnd; /* Current 'end' value */
+ int iPos; /* Current 'pos' value */
+};
+
+/*
+** Query FTS for the tokenizer implementation named zName.
+*/
+static int fts3tokQueryTokenizer(
+ Fts3Hash *pHash,
+ const char *zName,
+ const sqlite3_tokenizer_module **pp,
+ char **pzErr
+){
+ sqlite3_tokenizer_module *p;
+ int nName = (int)strlen(zName);
+
+ p = (sqlite3_tokenizer_module *)sqlite3Fts3HashFind(pHash, zName, nName+1);
+ if( !p ){
+ sqlite3Fts3ErrMsg(pzErr, "unknown tokenizer: %s", zName);
+ return SQLITE_ERROR;
+ }
+
+ *pp = p;
+ return SQLITE_OK;
+}
+
+/*
+** The second argument, argv[], is an array of pointers to nul-terminated
+** strings. This function makes a copy of the array and strings into a
+** single block of memory. It then dequotes any of the strings that appear
+** to be quoted.
+**
+** If successful, output parameter *pazDequote is set to point at the
+** array of dequoted strings and SQLITE_OK is returned. The caller is
+** responsible for eventually calling sqlite3_free() to free the array
+** in this case. Or, if an error occurs, an SQLite error code is returned.
+** The final value of *pazDequote is undefined in this case.
+*/
+static int fts3tokDequoteArray(
+ int argc, /* Number of elements in argv[] */
+ const char * const *argv, /* Input array */
+ char ***pazDequote /* Output array */
+){
+ int rc = SQLITE_OK; /* Return code */
+ if( argc==0 ){
+ *pazDequote = 0;
+ }else{
+ int i;
+ int nByte = 0;
+ char **azDequote;
+
+ for(i=0; i1 ) azArg = (const char * const *)&azDequote[1];
+ rc = pMod->xCreate((nDequote>1 ? nDequote-1 : 0), azArg, &pTok);
+ }
+
+ if( rc==SQLITE_OK ){
+ pTab = (Fts3tokTable *)sqlite3_malloc(sizeof(Fts3tokTable));
+ if( pTab==0 ){
+ rc = SQLITE_NOMEM;
+ }
+ }
+
+ if( rc==SQLITE_OK ){
+ memset(pTab, 0, sizeof(Fts3tokTable));
+ pTab->pMod = pMod;
+ pTab->pTok = pTok;
+ *ppVtab = &pTab->base;
+ }else{
+ if( pTok ){
+ pMod->xDestroy(pTok);
+ }
+ }
+
+ sqlite3_free(azDequote);
+ return rc;
+}
+
+/*
+** This function does the work for both the xDisconnect and xDestroy methods.
+** These tables have no persistent representation of their own, so xDisconnect
+** and xDestroy are identical operations.
+*/
+static int fts3tokDisconnectMethod(sqlite3_vtab *pVtab){
+ Fts3tokTable *pTab = (Fts3tokTable *)pVtab;
+
+ pTab->pMod->xDestroy(pTab->pTok);
+ sqlite3_free(pTab);
+ return SQLITE_OK;
+}
+
+/*
+** xBestIndex - Analyze a WHERE and ORDER BY clause.
+*/
+static int fts3tokBestIndexMethod(
+ sqlite3_vtab *pVTab,
+ sqlite3_index_info *pInfo
+){
+ int i;
+ UNUSED_PARAMETER(pVTab);
+
+ for(i=0; inConstraint; i++){
+ if( pInfo->aConstraint[i].usable
+ && pInfo->aConstraint[i].iColumn==0
+ && pInfo->aConstraint[i].op==SQLITE_INDEX_CONSTRAINT_EQ
+ ){
+ pInfo->idxNum = 1;
+ pInfo->aConstraintUsage[i].argvIndex = 1;
+ pInfo->aConstraintUsage[i].omit = 1;
+ pInfo->estimatedCost = 1;
+ return SQLITE_OK;
+ }
+ }
+
+ pInfo->idxNum = 0;
+ assert( pInfo->estimatedCost>1000000.0 );
+
+ return SQLITE_OK;
+}
+
+/*
+** xOpen - Open a cursor.
+*/
+static int fts3tokOpenMethod(sqlite3_vtab *pVTab, sqlite3_vtab_cursor **ppCsr){
+ Fts3tokCursor *pCsr;
+ UNUSED_PARAMETER(pVTab);
+
+ pCsr = (Fts3tokCursor *)sqlite3_malloc(sizeof(Fts3tokCursor));
+ if( pCsr==0 ){
+ return SQLITE_NOMEM;
+ }
+ memset(pCsr, 0, sizeof(Fts3tokCursor));
+
+ *ppCsr = (sqlite3_vtab_cursor *)pCsr;
+ return SQLITE_OK;
+}
+
+/*
+** Reset the tokenizer cursor passed as the only argument. As if it had
+** just been returned by fts3tokOpenMethod().
+*/
+static void fts3tokResetCursor(Fts3tokCursor *pCsr){
+ if( pCsr->pCsr ){
+ Fts3tokTable *pTab = (Fts3tokTable *)(pCsr->base.pVtab);
+ pTab->pMod->xClose(pCsr->pCsr);
+ pCsr->pCsr = 0;
+ }
+ sqlite3_free(pCsr->zInput);
+ pCsr->zInput = 0;
+ pCsr->zToken = 0;
+ pCsr->nToken = 0;
+ pCsr->iStart = 0;
+ pCsr->iEnd = 0;
+ pCsr->iPos = 0;
+ pCsr->iRowid = 0;
+}
+
+/*
+** xClose - Close a cursor.
+*/
+static int fts3tokCloseMethod(sqlite3_vtab_cursor *pCursor){
+ Fts3tokCursor *pCsr = (Fts3tokCursor *)pCursor;
+
+ fts3tokResetCursor(pCsr);
+ sqlite3_free(pCsr);
+ return SQLITE_OK;
+}
+
+/*
+** xNext - Advance the cursor to the next row, if any.
+*/
+static int fts3tokNextMethod(sqlite3_vtab_cursor *pCursor){
+ Fts3tokCursor *pCsr = (Fts3tokCursor *)pCursor;
+ Fts3tokTable *pTab = (Fts3tokTable *)(pCursor->pVtab);
+ int rc; /* Return code */
+
+ pCsr->iRowid++;
+ rc = pTab->pMod->xNext(pCsr->pCsr,
+ &pCsr->zToken, &pCsr->nToken,
+ &pCsr->iStart, &pCsr->iEnd, &pCsr->iPos
+ );
+
+ if( rc!=SQLITE_OK ){
+ fts3tokResetCursor(pCsr);
+ if( rc==SQLITE_DONE ) rc = SQLITE_OK;
+ }
+
+ return rc;
+}
+
+/*
+** xFilter - Initialize a cursor to point at the start of its data.
+*/
+static int fts3tokFilterMethod(
+ sqlite3_vtab_cursor *pCursor, /* The cursor used for this query */
+ int idxNum, /* Strategy index */
+ const char *idxStr, /* Unused */
+ int nVal, /* Number of elements in apVal */
+ sqlite3_value **apVal /* Arguments for the indexing scheme */
+){
+ int rc = SQLITE_ERROR;
+ Fts3tokCursor *pCsr = (Fts3tokCursor *)pCursor;
+ Fts3tokTable *pTab = (Fts3tokTable *)(pCursor->pVtab);
+ UNUSED_PARAMETER(idxStr);
+ UNUSED_PARAMETER(nVal);
+
+ fts3tokResetCursor(pCsr);
+ if( idxNum==1 ){
+ const char *zByte = (const char *)sqlite3_value_text(apVal[0]);
+ int nByte = sqlite3_value_bytes(apVal[0]);
+ pCsr->zInput = sqlite3_malloc64(nByte+1);
+ if( pCsr->zInput==0 ){
+ rc = SQLITE_NOMEM;
+ }else{
+ if( nByte>0 ) memcpy(pCsr->zInput, zByte, nByte);
+ pCsr->zInput[nByte] = 0;
+ rc = pTab->pMod->xOpen(pTab->pTok, pCsr->zInput, nByte, &pCsr->pCsr);
+ if( rc==SQLITE_OK ){
+ pCsr->pCsr->pTokenizer = pTab->pTok;
+ }
+ }
+ }
+
+ if( rc!=SQLITE_OK ) return rc;
+ return fts3tokNextMethod(pCursor);
+}
+
+/*
+** xEof - Return true if the cursor is at EOF, or false otherwise.
+*/
+static int fts3tokEofMethod(sqlite3_vtab_cursor *pCursor){
+ Fts3tokCursor *pCsr = (Fts3tokCursor *)pCursor;
+ return (pCsr->zToken==0);
+}
+
+/*
+** xColumn - Return a column value.
+*/
+static int fts3tokColumnMethod(
+ sqlite3_vtab_cursor *pCursor, /* Cursor to retrieve value from */
+ sqlite3_context *pCtx, /* Context for sqlite3_result_xxx() calls */
+ int iCol /* Index of column to read value from */
+){
+ Fts3tokCursor *pCsr = (Fts3tokCursor *)pCursor;
+
+ /* CREATE TABLE x(input, token, start, end, position) */
+ switch( iCol ){
+ case 0:
+ sqlite3_result_text(pCtx, pCsr->zInput, -1, SQLITE_TRANSIENT);
+ break;
+ case 1:
+ sqlite3_result_text(pCtx, pCsr->zToken, pCsr->nToken, SQLITE_TRANSIENT);
+ break;
+ case 2:
+ sqlite3_result_int(pCtx, pCsr->iStart);
+ break;
+ case 3:
+ sqlite3_result_int(pCtx, pCsr->iEnd);
+ break;
+ default:
+ assert( iCol==4 );
+ sqlite3_result_int(pCtx, pCsr->iPos);
+ break;
+ }
+ return SQLITE_OK;
+}
+
+/*
+** xRowid - Return the current rowid for the cursor.
+*/
+static int fts3tokRowidMethod(
+ sqlite3_vtab_cursor *pCursor, /* Cursor to retrieve value from */
+ sqlite_int64 *pRowid /* OUT: Rowid value */
+){
+ Fts3tokCursor *pCsr = (Fts3tokCursor *)pCursor;
+ *pRowid = (sqlite3_int64)pCsr->iRowid;
+ return SQLITE_OK;
+}
+
+/*
+** Register the fts3tok module with database connection db. Return SQLITE_OK
+** if successful or an error code if sqlite3_create_module() fails.
+*/
+int sqlite3Fts3InitTok(sqlite3 *db, Fts3Hash *pHash, void(*xDestroy)(void*)){
+ static const sqlite3_module fts3tok_module = {
+ 0, /* iVersion */
+ fts3tokConnectMethod, /* xCreate */
+ fts3tokConnectMethod, /* xConnect */
+ fts3tokBestIndexMethod, /* xBestIndex */
+ fts3tokDisconnectMethod, /* xDisconnect */
+ fts3tokDisconnectMethod, /* xDestroy */
+ fts3tokOpenMethod, /* xOpen */
+ fts3tokCloseMethod, /* xClose */
+ fts3tokFilterMethod, /* xFilter */
+ fts3tokNextMethod, /* xNext */
+ fts3tokEofMethod, /* xEof */
+ fts3tokColumnMethod, /* xColumn */
+ fts3tokRowidMethod, /* xRowid */
+ 0, /* xUpdate */
+ 0, /* xBegin */
+ 0, /* xSync */
+ 0, /* xCommit */
+ 0, /* xRollback */
+ 0, /* xFindFunction */
+ 0, /* xRename */
+ 0, /* xSavepoint */
+ 0, /* xRelease */
+ 0, /* xRollbackTo */
+ 0, /* xShadowName */
+ 0 /* xIntegrity */
+ };
+ int rc; /* Return code */
+
+ rc = sqlite3_create_module_v2(
+ db, "fts3tokenize", &fts3tok_module, (void*)pHash, xDestroy
+ );
+ return rc;
+}
+
+#endif /* !defined(SQLITE_CORE) || defined(SQLITE_ENABLE_FTS3) */
diff --git a/ext/fts3/fts3_tokenizer.c b/ext/fts3/fts3_tokenizer.c
new file mode 100644
index 0000000000000000000000000000000000000000..24c237a89d5a79142fb6cf12d00bc7260d0a28c7
--- /dev/null
+++ b/ext/fts3/fts3_tokenizer.c
@@ -0,0 +1,516 @@
+/*
+** 2007 June 22
+**
+** The author disclaims copyright to this source code. In place of
+** a legal notice, here is a blessing:
+**
+** May you do good and not evil.
+** May you find forgiveness for yourself and forgive others.
+** May you share freely, never taking more than you give.
+**
+******************************************************************************
+**
+** This is part of an SQLite module implementing full-text search.
+** This particular file implements the generic tokenizer interface.
+*/
+
+/*
+** The code in this file is only compiled if:
+**
+** * The FTS3 module is being built as an extension
+** (in which case SQLITE_CORE is not defined), or
+**
+** * The FTS3 module is being built into the core of
+** SQLite (in which case SQLITE_ENABLE_FTS3 is defined).
+*/
+#include "fts3Int.h"
+#if !defined(SQLITE_CORE) || defined(SQLITE_ENABLE_FTS3)
+
+#include
+#include
+
+/*
+** Return true if the two-argument version of fts3_tokenizer()
+** has been activated via a prior call to sqlite3_db_config(db,
+** SQLITE_DBCONFIG_ENABLE_FTS3_TOKENIZER, 1, 0);
+*/
+static int fts3TokenizerEnabled(sqlite3_context *context){
+ sqlite3 *db = sqlite3_context_db_handle(context);
+ int isEnabled = 0;
+ sqlite3_db_config(db,SQLITE_DBCONFIG_ENABLE_FTS3_TOKENIZER,-1,&isEnabled);
+ return isEnabled;
+}
+
+/*
+** Implementation of the SQL scalar function for accessing the underlying
+** hash table. This function may be called as follows:
+**
+** SELECT ();
+** SELECT (, );
+**
+** where is the name passed as the second argument
+** to the sqlite3Fts3InitHashTable() function (e.g. 'fts3_tokenizer').
+**
+** If the argument is specified, it must be a blob value
+** containing a pointer to be stored as the hash data corresponding
+** to the string . If is not specified, then
+** the string must already exist in the has table. Otherwise,
+** an error is returned.
+**
+** Whether or not the argument is specified, the value returned
+** is a blob containing the pointer stored as the hash data corresponding
+** to string (after the hash-table is updated, if applicable).
+*/
+static void fts3TokenizerFunc(
+ sqlite3_context *context,
+ int argc,
+ sqlite3_value **argv
+){
+ Fts3Hash *pHash;
+ void *pPtr = 0;
+ const unsigned char *zName;
+ int nName;
+
+ assert( argc==1 || argc==2 );
+
+ pHash = (Fts3Hash *)sqlite3_user_data(context);
+
+ zName = sqlite3_value_text(argv[0]);
+ nName = sqlite3_value_bytes(argv[0])+1;
+
+ if( argc==2 ){
+ if( fts3TokenizerEnabled(context) || sqlite3_value_frombind(argv[1]) ){
+ void *pOld;
+ int n = sqlite3_value_bytes(argv[1]);
+ if( zName==0 || n!=sizeof(pPtr) ){
+ sqlite3_result_error(context, "argument type mismatch", -1);
+ return;
+ }
+ pPtr = *(void **)sqlite3_value_blob(argv[1]);
+ pOld = sqlite3Fts3HashInsert(pHash, (void *)zName, nName, pPtr);
+ if( pOld==pPtr ){
+ sqlite3_result_error(context, "out of memory", -1);
+ }
+ }else{
+ sqlite3_result_error(context, "fts3tokenize disabled", -1);
+ return;
+ }
+ }else{
+ if( zName ){
+ pPtr = sqlite3Fts3HashFind(pHash, zName, nName);
+ }
+ if( !pPtr ){
+ char *zErr = sqlite3_mprintf("unknown tokenizer: %s", zName);
+ sqlite3_result_error(context, zErr, -1);
+ sqlite3_free(zErr);
+ return;
+ }
+ }
+ if( fts3TokenizerEnabled(context) || sqlite3_value_frombind(argv[0]) ){
+ sqlite3_result_blob(context, (void *)&pPtr, sizeof(pPtr), SQLITE_TRANSIENT);
+ }
+}
+
+int sqlite3Fts3IsIdChar(char c){
+ static const char isFtsIdChar[] = {
+ 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, /* 0x */
+ 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, /* 1x */
+ 0, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, /* 2x */
+ 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 0, 0, 0, 0, 0, 0, /* 3x */
+ 0, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, /* 4x */
+ 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 0, 0, 0, 0, 1, /* 5x */
+ 0, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, /* 6x */
+ 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 0, 0, 0, 0, 0, /* 7x */
+ };
+ return (c&0x80 || isFtsIdChar[(int)(c)]);
+}
+
+const char *sqlite3Fts3NextToken(const char *zStr, int *pn){
+ const char *z1;
+ const char *z2 = 0;
+
+ /* Find the start of the next token. */
+ z1 = zStr;
+ while( z2==0 ){
+ char c = *z1;
+ switch( c ){
+ case '\0': return 0; /* No more tokens here */
+ case '\'':
+ case '"':
+ case '`': {
+ z2 = z1;
+ while( *++z2 && (*z2!=c || *++z2==c) );
+ break;
+ }
+ case '[':
+ z2 = &z1[1];
+ while( *z2 && z2[0]!=']' ) z2++;
+ if( *z2 ) z2++;
+ break;
+
+ default:
+ if( sqlite3Fts3IsIdChar(*z1) ){
+ z2 = &z1[1];
+ while( sqlite3Fts3IsIdChar(*z2) ) z2++;
+ }else{
+ z1++;
+ }
+ }
+ }
+
+ *pn = (int)(z2-z1);
+ return z1;
+}
+
+int sqlite3Fts3InitTokenizer(
+ Fts3Hash *pHash, /* Tokenizer hash table */
+ const char *zArg, /* Tokenizer name */
+ sqlite3_tokenizer **ppTok, /* OUT: Tokenizer (if applicable) */
+ char **pzErr /* OUT: Set to malloced error message */
+){
+ int rc;
+ char *z = (char *)zArg;
+ int n = 0;
+ char *zCopy;
+ char *zEnd; /* Pointer to nul-term of zCopy */
+ sqlite3_tokenizer_module *m;
+
+ zCopy = sqlite3_mprintf("%s", zArg);
+ if( !zCopy ) return SQLITE_NOMEM;
+ zEnd = &zCopy[strlen(zCopy)];
+
+ z = (char *)sqlite3Fts3NextToken(zCopy, &n);
+ if( z==0 ){
+ assert( n==0 );
+ z = zCopy;
+ }
+ z[n] = '\0';
+ sqlite3Fts3Dequote(z);
+
+ m = (sqlite3_tokenizer_module *)sqlite3Fts3HashFind(pHash,z,(int)strlen(z)+1);
+ if( !m ){
+ sqlite3Fts3ErrMsg(pzErr, "unknown tokenizer: %s", z);
+ rc = SQLITE_ERROR;
+ }else{
+ char const **aArg = 0;
+ int iArg = 0;
+ z = &z[n+1];
+ while( zxCreate(iArg, aArg, ppTok);
+ assert( rc!=SQLITE_OK || *ppTok );
+ if( rc!=SQLITE_OK ){
+ sqlite3Fts3ErrMsg(pzErr, "unknown tokenizer");
+ }else{
+ (*ppTok)->pModule = m;
+ }
+ sqlite3_free((void *)aArg);
+ }
+
+ sqlite3_free(zCopy);
+ return rc;
+}
+
+
+#ifdef SQLITE_TEST
+
+#include "tclsqlite.h"
+#include
+
+/*
+** Implementation of a special SQL scalar function for testing tokenizers
+** designed to be used in concert with the Tcl testing framework. This
+** function must be called with two or more arguments:
+**
+** SELECT (, ..., );
+**
+** where is the name passed as the second argument
+** to the sqlite3Fts3InitHashTable() function (e.g. 'fts3_tokenizer')
+** concatenated with the string '_test' (e.g. 'fts3_tokenizer_test').
+**
+** The return value is a string that may be interpreted as a Tcl
+** list. For each token in the , three elements are
+** added to the returned list. The first is the token position, the
+** second is the token text (folded, stemmed, etc.) and the third is the
+** substring of associated with the token. For example,
+** using the built-in "simple" tokenizer:
+**
+** SELECT fts_tokenizer_test('simple', 'I don't see how');
+**
+** will return the string:
+**
+** "{0 i I 1 dont don't 2 see see 3 how how}"
+**
+*/
+static void testFunc(
+ sqlite3_context *context,
+ int argc,
+ sqlite3_value **argv
+){
+ Fts3Hash *pHash;
+ sqlite3_tokenizer_module *p;
+ sqlite3_tokenizer *pTokenizer = 0;
+ sqlite3_tokenizer_cursor *pCsr = 0;
+
+ const char *zErr = 0;
+
+ const char *zName;
+ int nName;
+ const char *zInput;
+ int nInput;
+
+ const char *azArg[64];
+
+ const char *zToken;
+ int nToken = 0;
+ int iStart = 0;
+ int iEnd = 0;
+ int iPos = 0;
+ int i;
+
+ Tcl_Obj *pRet;
+
+ if( argc<2 ){
+ sqlite3_result_error(context, "insufficient arguments", -1);
+ return;
+ }
+
+ nName = sqlite3_value_bytes(argv[0]);
+ zName = (const char *)sqlite3_value_text(argv[0]);
+ nInput = sqlite3_value_bytes(argv[argc-1]);
+ zInput = (const char *)sqlite3_value_text(argv[argc-1]);
+
+ pHash = (Fts3Hash *)sqlite3_user_data(context);
+ p = (sqlite3_tokenizer_module *)sqlite3Fts3HashFind(pHash, zName, nName+1);
+
+ if( !p ){
+ char *zErr2 = sqlite3_mprintf("unknown tokenizer: %s", zName);
+ sqlite3_result_error(context, zErr2, -1);
+ sqlite3_free(zErr2);
+ return;
+ }
+
+ pRet = Tcl_NewObj();
+ Tcl_IncrRefCount(pRet);
+
+ for(i=1; ixCreate(argc-2, azArg, &pTokenizer) ){
+ zErr = "error in xCreate()";
+ goto finish;
+ }
+ pTokenizer->pModule = p;
+ if( sqlite3Fts3OpenTokenizer(pTokenizer, 0, zInput, nInput, &pCsr) ){
+ zErr = "error in xOpen()";
+ goto finish;
+ }
+
+ while( SQLITE_OK==p->xNext(pCsr, &zToken, &nToken, &iStart, &iEnd, &iPos) ){
+ Tcl_ListObjAppendElement(0, pRet, Tcl_NewIntObj(iPos));
+ Tcl_ListObjAppendElement(0, pRet, Tcl_NewStringObj(zToken, nToken));
+ zToken = &zInput[iStart];
+ nToken = iEnd-iStart;
+ Tcl_ListObjAppendElement(0, pRet, Tcl_NewStringObj(zToken, nToken));
+ }
+
+ if( SQLITE_OK!=p->xClose(pCsr) ){
+ zErr = "error in xClose()";
+ goto finish;
+ }
+ if( SQLITE_OK!=p->xDestroy(pTokenizer) ){
+ zErr = "error in xDestroy()";
+ goto finish;
+ }
+
+finish:
+ if( zErr ){
+ sqlite3_result_error(context, zErr, -1);
+ }else{
+ sqlite3_result_text(context, Tcl_GetString(pRet), -1, SQLITE_TRANSIENT);
+ }
+ Tcl_DecrRefCount(pRet);
+}
+
+static
+int registerTokenizer(
+ sqlite3 *db,
+ char *zName,
+ const sqlite3_tokenizer_module *p
+){
+ int rc;
+ sqlite3_stmt *pStmt;
+ const char zSql[] = "SELECT fts3_tokenizer(?, ?)";
+
+ rc = sqlite3_prepare_v2(db, zSql, -1, &pStmt, 0);
+ if( rc!=SQLITE_OK ){
+ return rc;
+ }
+
+ sqlite3_bind_text(pStmt, 1, zName, -1, SQLITE_STATIC);
+ sqlite3_bind_blob(pStmt, 2, &p, sizeof(p), SQLITE_STATIC);
+ sqlite3_step(pStmt);
+
+ return sqlite3_finalize(pStmt);
+}
+
+
+static
+int queryTokenizer(
+ sqlite3 *db,
+ char *zName,
+ const sqlite3_tokenizer_module **pp
+){
+ int rc;
+ sqlite3_stmt *pStmt;
+ const char zSql[] = "SELECT fts3_tokenizer(?)";
+
+ *pp = 0;
+ rc = sqlite3_prepare_v2(db, zSql, -1, &pStmt, 0);
+ if( rc!=SQLITE_OK ){
+ return rc;
+ }
+
+ sqlite3_bind_text(pStmt, 1, zName, -1, SQLITE_STATIC);
+ if( SQLITE_ROW==sqlite3_step(pStmt) ){
+ if( sqlite3_column_type(pStmt, 0)==SQLITE_BLOB
+ && sqlite3_column_bytes(pStmt, 0)==sizeof(*pp)
+ ){
+ memcpy((void *)pp, sqlite3_column_blob(pStmt, 0), sizeof(*pp));
+ }
+ }
+
+ return sqlite3_finalize(pStmt);
+}
+
+void sqlite3Fts3SimpleTokenizerModule(sqlite3_tokenizer_module const**ppModule);
+
+/*
+** Implementation of the scalar function fts3_tokenizer_internal_test().
+** This function is used for testing only, it is not included in the
+** build unless SQLITE_TEST is defined.
+**
+** The purpose of this is to test that the fts3_tokenizer() function
+** can be used as designed by the C-code in the queryTokenizer and
+** registerTokenizer() functions above. These two functions are repeated
+** in the README.tokenizer file as an example, so it is important to
+** test them.
+**
+** To run the tests, evaluate the fts3_tokenizer_internal_test() scalar
+** function with no arguments. An assert() will fail if a problem is
+** detected. i.e.:
+**
+** SELECT fts3_tokenizer_internal_test();
+**
+*/
+static void intTestFunc(
+ sqlite3_context *context,
+ int argc,
+ sqlite3_value **argv
+){
+ int rc;
+ const sqlite3_tokenizer_module *p1;
+ const sqlite3_tokenizer_module *p2;
+ sqlite3 *db = (sqlite3 *)sqlite3_user_data(context);
+
+ UNUSED_PARAMETER(argc);
+ UNUSED_PARAMETER(argv);
+
+ /* Test the query function */
+ sqlite3Fts3SimpleTokenizerModule(&p1);
+ rc = queryTokenizer(db, "simple", &p2);
+ assert( rc==SQLITE_OK );
+ assert( p1==p2 );
+ rc = queryTokenizer(db, "nosuchtokenizer", &p2);
+ assert( rc==SQLITE_ERROR );
+ assert( p2==0 );
+ assert( 0==strcmp(sqlite3_errmsg(db), "unknown tokenizer: nosuchtokenizer") );
+
+ /* Test the storage function */
+ if( fts3TokenizerEnabled(context) ){
+ rc = registerTokenizer(db, "nosuchtokenizer", p1);
+ assert( rc==SQLITE_OK );
+ rc = queryTokenizer(db, "nosuchtokenizer", &p2);
+ assert( rc==SQLITE_OK );
+ assert( p2==p1 );
+ }
+
+ sqlite3_result_text(context, "ok", -1, SQLITE_STATIC);
+}
+
+#endif
+
+/*
+** Set up SQL objects in database db used to access the contents of
+** the hash table pointed to by argument pHash. The hash table must
+** been initialized to use string keys, and to take a private copy
+** of the key when a value is inserted. i.e. by a call similar to:
+**
+** sqlite3Fts3HashInit(pHash, FTS3_HASH_STRING, 1);
+**
+** This function adds a scalar function (see header comment above
+** fts3TokenizerFunc() in this file for details) and, if ENABLE_TABLE is
+** defined at compilation time, a temporary virtual table (see header
+** comment above struct HashTableVtab) to the database schema. Both
+** provide read/write access to the contents of *pHash.
+**
+** The third argument to this function, zName, is used as the name
+** of both the scalar and, if created, the virtual table.
+*/
+int sqlite3Fts3InitHashTable(
+ sqlite3 *db,
+ Fts3Hash *pHash,
+ const char *zName
+){
+ int rc = SQLITE_OK;
+ void *p = (void *)pHash;
+ const int any = SQLITE_UTF8|SQLITE_DIRECTONLY;
+
+#ifdef SQLITE_TEST
+ char *zTest = 0;
+ char *zTest2 = 0;
+ void *pdb = (void *)db;
+ zTest = sqlite3_mprintf("%s_test", zName);
+ zTest2 = sqlite3_mprintf("%s_internal_test", zName);
+ if( !zTest || !zTest2 ){
+ rc = SQLITE_NOMEM;
+ }
+#endif
+
+ if( SQLITE_OK==rc ){
+ rc = sqlite3_create_function(db, zName, 1, any, p, fts3TokenizerFunc, 0, 0);
+ }
+ if( SQLITE_OK==rc ){
+ rc = sqlite3_create_function(db, zName, 2, any, p, fts3TokenizerFunc, 0, 0);
+ }
+#ifdef SQLITE_TEST
+ if( SQLITE_OK==rc ){
+ rc = sqlite3_create_function(db, zTest, -1, any, p, testFunc, 0, 0);
+ }
+ if( SQLITE_OK==rc ){
+ rc = sqlite3_create_function(db, zTest2, 0, any, pdb, intTestFunc, 0, 0);
+ }
+#endif
+
+#ifdef SQLITE_TEST
+ sqlite3_free(zTest);
+ sqlite3_free(zTest2);
+#endif
+
+ return rc;
+}
+
+#endif /* !defined(SQLITE_CORE) || defined(SQLITE_ENABLE_FTS3) */
diff --git a/ext/fts3/fts3_tokenizer.h b/ext/fts3/fts3_tokenizer.h
new file mode 100644
index 0000000000000000000000000000000000000000..4a40b2b38503833bf17ded9ee2e2742e86267fc6
--- /dev/null
+++ b/ext/fts3/fts3_tokenizer.h
@@ -0,0 +1,161 @@
+/*
+** 2006 July 10
+**
+** The author disclaims copyright to this source code.
+**
+*************************************************************************
+** Defines the interface to tokenizers used by fulltext-search. There
+** are three basic components:
+**
+** sqlite3_tokenizer_module is a singleton defining the tokenizer
+** interface functions. This is essentially the class structure for
+** tokenizers.
+**
+** sqlite3_tokenizer is used to define a particular tokenizer, perhaps
+** including customization information defined at creation time.
+**
+** sqlite3_tokenizer_cursor is generated by a tokenizer to generate
+** tokens from a particular input.
+*/
+#ifndef _FTS3_TOKENIZER_H_
+#define _FTS3_TOKENIZER_H_
+
+/* TODO(shess) Only used for SQLITE_OK and SQLITE_DONE at this time.
+** If tokenizers are to be allowed to call sqlite3_*() functions, then
+** we will need a way to register the API consistently.
+*/
+#include "sqlite3.h"
+
+/*
+** Structures used by the tokenizer interface. When a new tokenizer
+** implementation is registered, the caller provides a pointer to
+** an sqlite3_tokenizer_module containing pointers to the callback
+** functions that make up an implementation.
+**
+** When an fts3 table is created, it passes any arguments passed to
+** the tokenizer clause of the CREATE VIRTUAL TABLE statement to the
+** sqlite3_tokenizer_module.xCreate() function of the requested tokenizer
+** implementation. The xCreate() function in turn returns an
+** sqlite3_tokenizer structure representing the specific tokenizer to
+** be used for the fts3 table (customized by the tokenizer clause arguments).
+**
+** To tokenize an input buffer, the sqlite3_tokenizer_module.xOpen()
+** method is called. It returns an sqlite3_tokenizer_cursor object
+** that may be used to tokenize a specific input buffer based on
+** the tokenization rules supplied by a specific sqlite3_tokenizer
+** object.
+*/
+typedef struct sqlite3_tokenizer_module sqlite3_tokenizer_module;
+typedef struct sqlite3_tokenizer sqlite3_tokenizer;
+typedef struct sqlite3_tokenizer_cursor sqlite3_tokenizer_cursor;
+
+struct sqlite3_tokenizer_module {
+
+ /*
+ ** Structure version. Should always be set to 0 or 1.
+ */
+ int iVersion;
+
+ /*
+ ** Create a new tokenizer. The values in the argv[] array are the
+ ** arguments passed to the "tokenizer" clause of the CREATE VIRTUAL
+ ** TABLE statement that created the fts3 table. For example, if
+ ** the following SQL is executed:
+ **
+ ** CREATE .. USING fts3( ... , tokenizer arg1 arg2)
+ **
+ ** then argc is set to 2, and the argv[] array contains pointers
+ ** to the strings "arg1" and "arg2".
+ **
+ ** This method should return either SQLITE_OK (0), or an SQLite error
+ ** code. If SQLITE_OK is returned, then *ppTokenizer should be set
+ ** to point at the newly created tokenizer structure. The generic
+ ** sqlite3_tokenizer.pModule variable should not be initialized by
+ ** this callback. The caller will do so.
+ */
+ int (*xCreate)(
+ int argc, /* Size of argv array */
+ const char *const*argv, /* Tokenizer argument strings */
+ sqlite3_tokenizer **ppTokenizer /* OUT: Created tokenizer */
+ );
+
+ /*
+ ** Destroy an existing tokenizer. The fts3 module calls this method
+ ** exactly once for each successful call to xCreate().
+ */
+ int (*xDestroy)(sqlite3_tokenizer *pTokenizer);
+
+ /*
+ ** Create a tokenizer cursor to tokenize an input buffer. The caller
+ ** is responsible for ensuring that the input buffer remains valid
+ ** until the cursor is closed (using the xClose() method).
+ */
+ int (*xOpen)(
+ sqlite3_tokenizer *pTokenizer, /* Tokenizer object */
+ const char *pInput, int nBytes, /* Input buffer */
+ sqlite3_tokenizer_cursor **ppCursor /* OUT: Created tokenizer cursor */
+ );
+
+ /*
+ ** Destroy an existing tokenizer cursor. The fts3 module calls this
+ ** method exactly once for each successful call to xOpen().
+ */
+ int (*xClose)(sqlite3_tokenizer_cursor *pCursor);
+
+ /*
+ ** Retrieve the next token from the tokenizer cursor pCursor. This
+ ** method should either return SQLITE_OK and set the values of the
+ ** "OUT" variables identified below, or SQLITE_DONE to indicate that
+ ** the end of the buffer has been reached, or an SQLite error code.
+ **
+ ** *ppToken should be set to point at a buffer containing the
+ ** normalized version of the token (i.e. after any case-folding and/or
+ ** stemming has been performed). *pnBytes should be set to the length
+ ** of this buffer in bytes. The input text that generated the token is
+ ** identified by the byte offsets returned in *piStartOffset and
+ ** *piEndOffset. *piStartOffset should be set to the index of the first
+ ** byte of the token in the input buffer. *piEndOffset should be set
+ ** to the index of the first byte just past the end of the token in
+ ** the input buffer.
+ **
+ ** The buffer *ppToken is set to point at is managed by the tokenizer
+ ** implementation. It is only required to be valid until the next call
+ ** to xNext() or xClose().
+ */
+ /* TODO(shess) current implementation requires pInput to be
+ ** nul-terminated. This should either be fixed, or pInput/nBytes
+ ** should be converted to zInput.
+ */
+ int (*xNext)(
+ sqlite3_tokenizer_cursor *pCursor, /* Tokenizer cursor */
+ const char **ppToken, int *pnBytes, /* OUT: Normalized text for token */
+ int *piStartOffset, /* OUT: Byte offset of token in input buffer */
+ int *piEndOffset, /* OUT: Byte offset of end of token in input buffer */
+ int *piPosition /* OUT: Number of tokens returned before this one */
+ );
+
+ /***********************************************************************
+ ** Methods below this point are only available if iVersion>=1.
+ */
+
+ /*
+ ** Configure the language id of a tokenizer cursor.
+ */
+ int (*xLanguageid)(sqlite3_tokenizer_cursor *pCsr, int iLangid);
+};
+
+struct sqlite3_tokenizer {
+ const sqlite3_tokenizer_module *pModule; /* The module for this tokenizer */
+ /* Tokenizer implementations will typically add additional fields */
+};
+
+struct sqlite3_tokenizer_cursor {
+ sqlite3_tokenizer *pTokenizer; /* Tokenizer for this cursor. */
+ /* Tokenizer implementations will typically add additional fields */
+};
+
+int fts3_global_term_cnt(int iTerm, int iCol);
+int fts3_term_cnt(int iTerm, int iCol);
+
+
+#endif /* _FTS3_TOKENIZER_H_ */
diff --git a/ext/fts3/fts3_tokenizer1.c b/ext/fts3/fts3_tokenizer1.c
new file mode 100644
index 0000000000000000000000000000000000000000..78e5889da5250184db89879b059a66eb5637aa93
--- /dev/null
+++ b/ext/fts3/fts3_tokenizer1.c
@@ -0,0 +1,234 @@
+/*
+** 2006 Oct 10
+**
+** The author disclaims copyright to this source code. In place of
+** a legal notice, here is a blessing:
+**
+** May you do good and not evil.
+** May you find forgiveness for yourself and forgive others.
+** May you share freely, never taking more than you give.
+**
+******************************************************************************
+**
+** Implementation of the "simple" full-text-search tokenizer.
+*/
+
+/*
+** The code in this file is only compiled if:
+**
+** * The FTS3 module is being built as an extension
+** (in which case SQLITE_CORE is not defined), or
+**
+** * The FTS3 module is being built into the core of
+** SQLite (in which case SQLITE_ENABLE_FTS3 is defined).
+*/
+#include "fts3Int.h"
+#if !defined(SQLITE_CORE) || defined(SQLITE_ENABLE_FTS3)
+
+#include
+#include
+#include
+#include
+
+#include "fts3_tokenizer.h"
+
+typedef struct simple_tokenizer {
+ sqlite3_tokenizer base;
+ char delim[128]; /* flag ASCII delimiters */
+} simple_tokenizer;
+
+typedef struct simple_tokenizer_cursor {
+ sqlite3_tokenizer_cursor base;
+ const char *pInput; /* input we are tokenizing */
+ int nBytes; /* size of the input */
+ int iOffset; /* current position in pInput */
+ int iToken; /* index of next token to be returned */
+ char *pToken; /* storage for current token */
+ int nTokenAllocated; /* space allocated to zToken buffer */
+} simple_tokenizer_cursor;
+
+
+static int simpleDelim(simple_tokenizer *t, unsigned char c){
+ return c<0x80 && t->delim[c];
+}
+static int fts3_isalnum(int x){
+ return (x>='0' && x<='9') || (x>='A' && x<='Z') || (x>='a' && x<='z');
+}
+
+/*
+** Create a new tokenizer instance.
+*/
+static int simpleCreate(
+ int argc, const char * const *argv,
+ sqlite3_tokenizer **ppTokenizer
+){
+ simple_tokenizer *t;
+
+ t = (simple_tokenizer *) sqlite3_malloc(sizeof(*t));
+ if( t==NULL ) return SQLITE_NOMEM;
+ memset(t, 0, sizeof(*t));
+
+ /* TODO(shess) Delimiters need to remain the same from run to run,
+ ** else we need to reindex. One solution would be a meta-table to
+ ** track such information in the database, then we'd only want this
+ ** information on the initial create.
+ */
+ if( argc>1 ){
+ int i, n = (int)strlen(argv[1]);
+ for(i=0; i=0x80 ){
+ sqlite3_free(t);
+ return SQLITE_ERROR;
+ }
+ t->delim[ch] = 1;
+ }
+ } else {
+ /* Mark non-alphanumeric ASCII characters as delimiters */
+ int i;
+ for(i=1; i<0x80; i++){
+ t->delim[i] = !fts3_isalnum(i) ? -1 : 0;
+ }
+ }
+
+ *ppTokenizer = &t->base;
+ return SQLITE_OK;
+}
+
+/*
+** Destroy a tokenizer
+*/
+static int simpleDestroy(sqlite3_tokenizer *pTokenizer){
+ sqlite3_free(pTokenizer);
+ return SQLITE_OK;
+}
+
+/*
+** Prepare to begin tokenizing a particular string. The input
+** string to be tokenized is pInput[0..nBytes-1]. A cursor
+** used to incrementally tokenize this string is returned in
+** *ppCursor.
+*/
+static int simpleOpen(
+ sqlite3_tokenizer *pTokenizer, /* The tokenizer */
+ const char *pInput, int nBytes, /* String to be tokenized */
+ sqlite3_tokenizer_cursor **ppCursor /* OUT: Tokenization cursor */
+){
+ simple_tokenizer_cursor *c;
+
+ UNUSED_PARAMETER(pTokenizer);
+
+ c = (simple_tokenizer_cursor *) sqlite3_malloc(sizeof(*c));
+ if( c==NULL ) return SQLITE_NOMEM;
+
+ c->pInput = pInput;
+ if( pInput==0 ){
+ c->nBytes = 0;
+ }else if( nBytes<0 ){
+ c->nBytes = (int)strlen(pInput);
+ }else{
+ c->nBytes = nBytes;
+ }
+ c->iOffset = 0; /* start tokenizing at the beginning */
+ c->iToken = 0;
+ c->pToken = NULL; /* no space allocated, yet. */
+ c->nTokenAllocated = 0;
+
+ *ppCursor = &c->base;
+ return SQLITE_OK;
+}
+
+/*
+** Close a tokenization cursor previously opened by a call to
+** simpleOpen() above.
+*/
+static int simpleClose(sqlite3_tokenizer_cursor *pCursor){
+ simple_tokenizer_cursor *c = (simple_tokenizer_cursor *) pCursor;
+ sqlite3_free(c->pToken);
+ sqlite3_free(c);
+ return SQLITE_OK;
+}
+
+/*
+** Extract the next token from a tokenization cursor. The cursor must
+** have been opened by a prior call to simpleOpen().
+*/
+static int simpleNext(
+ sqlite3_tokenizer_cursor *pCursor, /* Cursor returned by simpleOpen */
+ const char **ppToken, /* OUT: *ppToken is the token text */
+ int *pnBytes, /* OUT: Number of bytes in token */
+ int *piStartOffset, /* OUT: Starting offset of token */
+ int *piEndOffset, /* OUT: Ending offset of token */
+ int *piPosition /* OUT: Position integer of token */
+){
+ simple_tokenizer_cursor *c = (simple_tokenizer_cursor *) pCursor;
+ simple_tokenizer *t = (simple_tokenizer *) pCursor->pTokenizer;
+ unsigned char *p = (unsigned char *)c->pInput;
+
+ while( c->iOffsetnBytes ){
+ int iStartOffset;
+
+ /* Scan past delimiter characters */
+ while( c->iOffsetnBytes && simpleDelim(t, p[c->iOffset]) ){
+ c->iOffset++;
+ }
+
+ /* Count non-delimiter characters. */
+ iStartOffset = c->iOffset;
+ while( c->iOffsetnBytes && !simpleDelim(t, p[c->iOffset]) ){
+ c->iOffset++;
+ }
+
+ if( c->iOffset>iStartOffset ){
+ int i, n = c->iOffset-iStartOffset;
+ if( n>c->nTokenAllocated ){
+ char *pNew;
+ c->nTokenAllocated = n+20;
+ pNew = sqlite3_realloc64(c->pToken, c->nTokenAllocated);
+ if( !pNew ) return SQLITE_NOMEM;
+ c->pToken = pNew;
+ }
+ for(i=0; ipToken[i] = (char)((ch>='A' && ch<='Z') ? ch-'A'+'a' : ch);
+ }
+ *ppToken = c->pToken;
+ *pnBytes = n;
+ *piStartOffset = iStartOffset;
+ *piEndOffset = c->iOffset;
+ *piPosition = c->iToken++;
+
+ return SQLITE_OK;
+ }
+ }
+ return SQLITE_DONE;
+}
+
+/*
+** The set of routines that implement the simple tokenizer
+*/
+static const sqlite3_tokenizer_module simpleTokenizerModule = {
+ 0,
+ simpleCreate,
+ simpleDestroy,
+ simpleOpen,
+ simpleClose,
+ simpleNext,
+ 0,
+};
+
+/*
+** Allocate a new simple tokenizer. Return a pointer to the new
+** tokenizer in *ppModule
+*/
+void sqlite3Fts3SimpleTokenizerModule(
+ sqlite3_tokenizer_module const**ppModule
+){
+ *ppModule = &simpleTokenizerModule;
+}
+
+#endif /* !defined(SQLITE_CORE) || defined(SQLITE_ENABLE_FTS3) */
diff --git a/ext/fts3/fts3_unicode.c b/ext/fts3/fts3_unicode.c
new file mode 100644
index 0000000000000000000000000000000000000000..2fd4b39fb7721a8df03477546875b23a0e6ba43f
--- /dev/null
+++ b/ext/fts3/fts3_unicode.c
@@ -0,0 +1,397 @@
+/*
+** 2012 May 24
+**
+** The author disclaims copyright to this source code. In place of
+** a legal notice, here is a blessing:
+**
+** May you do good and not evil.
+** May you find forgiveness for yourself and forgive others.
+** May you share freely, never taking more than you give.
+**
+******************************************************************************
+**
+** Implementation of the "unicode" full-text-search tokenizer.
+*/
+
+#ifndef SQLITE_DISABLE_FTS3_UNICODE
+
+#include "fts3Int.h"
+#if !defined(SQLITE_CORE) || defined(SQLITE_ENABLE_FTS3)
+
+#include
+#include
+#include
+#include
+
+#include "fts3_tokenizer.h"
+
+/*
+** The following two macros - READ_UTF8 and WRITE_UTF8 - have been copied
+** from the sqlite3 source file utf.c. If this file is compiled as part
+** of the amalgamation, they are not required.
+*/
+#ifndef SQLITE_AMALGAMATION
+
+static const unsigned char sqlite3Utf8Trans1[] = {
+ 0x00, 0x01, 0x02, 0x03, 0x04, 0x05, 0x06, 0x07,
+ 0x08, 0x09, 0x0a, 0x0b, 0x0c, 0x0d, 0x0e, 0x0f,
+ 0x10, 0x11, 0x12, 0x13, 0x14, 0x15, 0x16, 0x17,
+ 0x18, 0x19, 0x1a, 0x1b, 0x1c, 0x1d, 0x1e, 0x1f,
+ 0x00, 0x01, 0x02, 0x03, 0x04, 0x05, 0x06, 0x07,
+ 0x08, 0x09, 0x0a, 0x0b, 0x0c, 0x0d, 0x0e, 0x0f,
+ 0x00, 0x01, 0x02, 0x03, 0x04, 0x05, 0x06, 0x07,
+ 0x00, 0x01, 0x02, 0x03, 0x00, 0x01, 0x00, 0x00,
+};
+
+#define READ_UTF8(zIn, zTerm, c) \
+ c = *(zIn++); \
+ if( c>=0xc0 ){ \
+ c = sqlite3Utf8Trans1[c-0xc0]; \
+ while( zIn!=zTerm && (*zIn & 0xc0)==0x80 ){ \
+ c = (c<<6) + (0x3f & *(zIn++)); \
+ } \
+ if( c<0x80 \
+ || (c&0xFFFFF800)==0xD800 \
+ || (c&0xFFFFFFFE)==0xFFFE ){ c = 0xFFFD; } \
+ }
+
+#define WRITE_UTF8(zOut, c) { \
+ if( c<0x00080 ){ \
+ *zOut++ = (u8)(c&0xFF); \
+ } \
+ else if( c<0x00800 ){ \
+ *zOut++ = 0xC0 + (u8)((c>>6)&0x1F); \
+ *zOut++ = 0x80 + (u8)(c & 0x3F); \
+ } \
+ else if( c<0x10000 ){ \
+ *zOut++ = 0xE0 + (u8)((c>>12)&0x0F); \
+ *zOut++ = 0x80 + (u8)((c>>6) & 0x3F); \
+ *zOut++ = 0x80 + (u8)(c & 0x3F); \
+ }else{ \
+ *zOut++ = 0xF0 + (u8)((c>>18) & 0x07); \
+ *zOut++ = 0x80 + (u8)((c>>12) & 0x3F); \
+ *zOut++ = 0x80 + (u8)((c>>6) & 0x3F); \
+ *zOut++ = 0x80 + (u8)(c & 0x3F); \
+ } \
+}
+
+#endif /* ifndef SQLITE_AMALGAMATION */
+
+typedef struct unicode_tokenizer unicode_tokenizer;
+typedef struct unicode_cursor unicode_cursor;
+
+struct unicode_tokenizer {
+ sqlite3_tokenizer base;
+ int eRemoveDiacritic;
+ int nException;
+ int *aiException;
+};
+
+struct unicode_cursor {
+ sqlite3_tokenizer_cursor base;
+ const unsigned char *aInput; /* Input text being tokenized */
+ int nInput; /* Size of aInput[] in bytes */
+ int iOff; /* Current offset within aInput[] */
+ int iToken; /* Index of next token to be returned */
+ char *zToken; /* storage for current token */
+ int nAlloc; /* space allocated at zToken */
+};
+
+
+/*
+** Destroy a tokenizer allocated by unicodeCreate().
+*/
+static int unicodeDestroy(sqlite3_tokenizer *pTokenizer){
+ if( pTokenizer ){
+ unicode_tokenizer *p = (unicode_tokenizer *)pTokenizer;
+ sqlite3_free(p->aiException);
+ sqlite3_free(p);
+ }
+ return SQLITE_OK;
+}
+
+/*
+** As part of a tokenchars= or separators= option, the CREATE VIRTUAL TABLE
+** statement has specified that the tokenizer for this table shall consider
+** all characters in string zIn/nIn to be separators (if bAlnum==0) or
+** token characters (if bAlnum==1).
+**
+** For each codepoint in the zIn/nIn string, this function checks if the
+** sqlite3FtsUnicodeIsalnum() function already returns the desired result.
+** If so, no action is taken. Otherwise, the codepoint is added to the
+** unicode_tokenizer.aiException[] array. For the purposes of tokenization,
+** the return value of sqlite3FtsUnicodeIsalnum() is inverted for all
+** codepoints in the aiException[] array.
+**
+** If a standalone diacritic mark (one that sqlite3FtsUnicodeIsdiacritic()
+** identifies as a diacritic) occurs in the zIn/nIn string it is ignored.
+** It is not possible to change the behavior of the tokenizer with respect
+** to these codepoints.
+*/
+static int unicodeAddExceptions(
+ unicode_tokenizer *p, /* Tokenizer to add exceptions to */
+ int bAlnum, /* Replace Isalnum() return value with this */
+ const char *zIn, /* Array of characters to make exceptions */
+ int nIn /* Length of z in bytes */
+){
+ const unsigned char *z = (const unsigned char *)zIn;
+ const unsigned char *zTerm = &z[nIn];
+ unsigned int iCode;
+ int nEntry = 0;
+
+ assert( bAlnum==0 || bAlnum==1 );
+
+ while( zaiException,(p->nException+nEntry)*sizeof(int));
+ if( aNew==0 ) return SQLITE_NOMEM;
+ nNew = p->nException;
+
+ z = (const unsigned char *)zIn;
+ while( zi; j--) aNew[j] = aNew[j-1];
+ aNew[i] = (int)iCode;
+ nNew++;
+ }
+ }
+ p->aiException = aNew;
+ p->nException = nNew;
+ }
+
+ return SQLITE_OK;
+}
+
+/*
+** Return true if the p->aiException[] array contains the value iCode.
+*/
+static int unicodeIsException(unicode_tokenizer *p, int iCode){
+ if( p->nException>0 ){
+ int *a = p->aiException;
+ int iLo = 0;
+ int iHi = p->nException-1;
+
+ while( iHi>=iLo ){
+ int iTest = (iHi + iLo) / 2;
+ if( iCode==a[iTest] ){
+ return 1;
+ }else if( iCode>a[iTest] ){
+ iLo = iTest+1;
+ }else{
+ iHi = iTest-1;
+ }
+ }
+ }
+
+ return 0;
+}
+
+/*
+** Return true if, for the purposes of tokenization, codepoint iCode is
+** considered a token character (not a separator).
+*/
+static int unicodeIsAlnum(unicode_tokenizer *p, int iCode){
+ assert( (sqlite3FtsUnicodeIsalnum(iCode) & 0xFFFFFFFE)==0 );
+ return sqlite3FtsUnicodeIsalnum(iCode) ^ unicodeIsException(p, iCode);
+}
+
+/*
+** Create a new tokenizer instance.
+*/
+static int unicodeCreate(
+ int nArg, /* Size of array argv[] */
+ const char * const *azArg, /* Tokenizer creation arguments */
+ sqlite3_tokenizer **pp /* OUT: New tokenizer handle */
+){
+ unicode_tokenizer *pNew; /* New tokenizer object */
+ int i;
+ int rc = SQLITE_OK;
+
+ pNew = (unicode_tokenizer *) sqlite3_malloc(sizeof(unicode_tokenizer));
+ if( pNew==NULL ) return SQLITE_NOMEM;
+ memset(pNew, 0, sizeof(unicode_tokenizer));
+ pNew->eRemoveDiacritic = 1;
+
+ for(i=0; rc==SQLITE_OK && ieRemoveDiacritic = 1;
+ }
+ else if( n==19 && memcmp("remove_diacritics=0", z, 19)==0 ){
+ pNew->eRemoveDiacritic = 0;
+ }
+ else if( n==19 && memcmp("remove_diacritics=2", z, 19)==0 ){
+ pNew->eRemoveDiacritic = 2;
+ }
+ else if( n>=11 && memcmp("tokenchars=", z, 11)==0 ){
+ rc = unicodeAddExceptions(pNew, 1, &z[11], n-11);
+ }
+ else if( n>=11 && memcmp("separators=", z, 11)==0 ){
+ rc = unicodeAddExceptions(pNew, 0, &z[11], n-11);
+ }
+ else{
+ /* Unrecognized argument */
+ rc = SQLITE_ERROR;
+ }
+ }
+
+ if( rc!=SQLITE_OK ){
+ unicodeDestroy((sqlite3_tokenizer *)pNew);
+ pNew = 0;
+ }
+ *pp = (sqlite3_tokenizer *)pNew;
+ return rc;
+}
+
+/*
+** Prepare to begin tokenizing a particular string. The input
+** string to be tokenized is pInput[0..nBytes-1]. A cursor
+** used to incrementally tokenize this string is returned in
+** *ppCursor.
+*/
+static int unicodeOpen(
+ sqlite3_tokenizer *p, /* The tokenizer */
+ const char *aInput, /* Input string */
+ int nInput, /* Size of string aInput in bytes */
+ sqlite3_tokenizer_cursor **pp /* OUT: New cursor object */
+){
+ unicode_cursor *pCsr;
+
+ pCsr = (unicode_cursor *)sqlite3_malloc(sizeof(unicode_cursor));
+ if( pCsr==0 ){
+ return SQLITE_NOMEM;
+ }
+ memset(pCsr, 0, sizeof(unicode_cursor));
+
+ pCsr->aInput = (const unsigned char *)aInput;
+ if( aInput==0 ){
+ pCsr->nInput = 0;
+ pCsr->aInput = (const unsigned char*)"";
+ }else if( nInput<0 ){
+ pCsr->nInput = (int)strlen(aInput);
+ }else{
+ pCsr->nInput = nInput;
+ }
+
+ *pp = &pCsr->base;
+ UNUSED_PARAMETER(p);
+ return SQLITE_OK;
+}
+
+/*
+** Close a tokenization cursor previously opened by a call to
+** simpleOpen() above.
+*/
+static int unicodeClose(sqlite3_tokenizer_cursor *pCursor){
+ unicode_cursor *pCsr = (unicode_cursor *) pCursor;
+ sqlite3_free(pCsr->zToken);
+ sqlite3_free(pCsr);
+ return SQLITE_OK;
+}
+
+/*
+** Extract the next token from a tokenization cursor. The cursor must
+** have been opened by a prior call to simpleOpen().
+*/
+static int unicodeNext(
+ sqlite3_tokenizer_cursor *pC, /* Cursor returned by simpleOpen */
+ const char **paToken, /* OUT: Token text */
+ int *pnToken, /* OUT: Number of bytes at *paToken */
+ int *piStart, /* OUT: Starting offset of token */
+ int *piEnd, /* OUT: Ending offset of token */
+ int *piPos /* OUT: Position integer of token */
+){
+ unicode_cursor *pCsr = (unicode_cursor *)pC;
+ unicode_tokenizer *p = ((unicode_tokenizer *)pCsr->base.pTokenizer);
+ unsigned int iCode = 0;
+ char *zOut;
+ const unsigned char *z = &pCsr->aInput[pCsr->iOff];
+ const unsigned char *zStart = z;
+ const unsigned char *zEnd;
+ const unsigned char *zTerm = &pCsr->aInput[pCsr->nInput];
+
+ /* Scan past any delimiter characters before the start of the next token.
+ ** Return SQLITE_DONE early if this takes us all the way to the end of
+ ** the input. */
+ while( z=zTerm ) return SQLITE_DONE;
+
+ zOut = pCsr->zToken;
+ do {
+ int iOut;
+
+ /* Grow the output buffer if required. */
+ if( (zOut-pCsr->zToken)>=(pCsr->nAlloc-4) ){
+ char *zNew = sqlite3_realloc64(pCsr->zToken, pCsr->nAlloc+64);
+ if( !zNew ) return SQLITE_NOMEM;
+ zOut = &zNew[zOut - pCsr->zToken];
+ pCsr->zToken = zNew;
+ pCsr->nAlloc += 64;
+ }
+
+ /* Write the folded case of the last character read to the output */
+ zEnd = z;
+ iOut = sqlite3FtsUnicodeFold((int)iCode, p->eRemoveDiacritic);
+ if( iOut ){
+ WRITE_UTF8(zOut, iOut);
+ }
+
+ /* If the cursor is not at EOF, read the next character */
+ if( z>=zTerm ) break;
+ READ_UTF8(z, zTerm, iCode);
+ }while( unicodeIsAlnum(p, (int)iCode)
+ || sqlite3FtsUnicodeIsdiacritic((int)iCode)
+ );
+
+ /* Set the output variables and return. */
+ pCsr->iOff = (int)(z - pCsr->aInput);
+ *paToken = pCsr->zToken;
+ *pnToken = (int)(zOut - pCsr->zToken);
+ *piStart = (int)(zStart - pCsr->aInput);
+ *piEnd = (int)(zEnd - pCsr->aInput);
+ *piPos = pCsr->iToken++;
+ return SQLITE_OK;
+}
+
+/*
+** Set *ppModule to a pointer to the sqlite3_tokenizer_module
+** structure for the unicode tokenizer.
+*/
+void sqlite3Fts3UnicodeTokenizer(sqlite3_tokenizer_module const **ppModule){
+ static const sqlite3_tokenizer_module module = {
+ 0,
+ unicodeCreate,
+ unicodeDestroy,
+ unicodeOpen,
+ unicodeClose,
+ unicodeNext,
+ 0,
+ };
+ *ppModule = &module;
+}
+
+#endif /* !defined(SQLITE_CORE) || defined(SQLITE_ENABLE_FTS3) */
+#endif /* ifndef SQLITE_DISABLE_FTS3_UNICODE */
diff --git a/ext/fts3/fts3_unicode2.c b/ext/fts3/fts3_unicode2.c
new file mode 100644
index 0000000000000000000000000000000000000000..c510162496dd95c8d21c0b588b8f247114503aca
--- /dev/null
+++ b/ext/fts3/fts3_unicode2.c
@@ -0,0 +1,383 @@
+/*
+** 2012-05-25
+**
+** The author disclaims copyright to this source code. In place of
+** a legal notice, here is a blessing:
+**
+** May you do good and not evil.
+** May you find forgiveness for yourself and forgive others.
+** May you share freely, never taking more than you give.
+**
+******************************************************************************
+*/
+
+/*
+** DO NOT EDIT THIS MACHINE GENERATED FILE.
+*/
+
+#ifndef SQLITE_DISABLE_FTS3_UNICODE
+#if defined(SQLITE_ENABLE_FTS3) || defined(SQLITE_ENABLE_FTS4)
+
+#include