diff options
Diffstat (limited to 'src')
| -rw-r--r-- | src/btree.h | 1 | ||||
| -rw-r--r-- | src/expr.c | 3 | ||||
| -rw-r--r-- | src/os_unix.c | 112 | ||||
| -rw-r--r-- | src/pager.c | 10 | ||||
| -rw-r--r-- | src/parse.y | 2 | ||||
| -rw-r--r-- | src/pragma.c | 9 | ||||
| -rw-r--r-- | src/resolve.c | 2 | ||||
| -rw-r--r-- | src/select.c | 5 | ||||
| -rw-r--r-- | src/shell.c | 12 | ||||
| -rw-r--r-- | src/sqlite.h.in | 15 | ||||
| -rw-r--r-- | src/sqliteInt.h | 3 | ||||
| -rw-r--r-- | src/tclsqlite.c | 6 | ||||
| -rw-r--r-- | src/test1.c | 64 | ||||
| -rw-r--r-- | src/test8.c | 24 | ||||
| -rw-r--r-- | src/test_fuzzer.c | 1215 | ||||
| -rw-r--r-- | src/test_regexp.c | 770 | ||||
| -rw-r--r-- | src/test_spellfix.c | 2847 | ||||
| -rw-r--r-- | src/test_wholenumber.c | 311 | ||||
| -rw-r--r-- | src/vacuum.c | 1 | ||||
| -rw-r--r-- | src/wal.h | 1 | ||||
| -rw-r--r-- | src/walker.c | 17 | ||||
| -rw-r--r-- | src/where.c | 3 |
22 files changed, 191 insertions, 5242 deletions
diff --git a/src/btree.h b/src/btree.h index 23f6ad706..ace0f8cd2 100644 --- a/src/btree.h +++ b/src/btree.h @@ -140,6 +140,7 @@ int sqlite3BtreeNewDb(Btree *p); #define BTREE_TEXT_ENCODING 5 #define BTREE_USER_VERSION 6 #define BTREE_INCR_VACUUM 7 +#define BTREE_APPLICATION_ID 8 /* ** Values that may be OR'd together to form the second argument of an diff --git a/src/expr.c b/src/expr.c index ae6a1dec1..a974c5a61 100644 --- a/src/expr.c +++ b/src/expr.c @@ -1214,6 +1214,7 @@ static int selectNodeIsConstant(Walker *pWalker, Select *NotUsed){ } static int exprIsConst(Expr *p, int initFlag){ Walker w; + memset(&w, 0, sizeof(w)); w.u.i = initFlag; w.xExprCallback = exprNodeIsConstant; w.xSelectCallback = selectNodeIsConstant; @@ -3428,8 +3429,8 @@ void sqlite3ExprCodeConstants(Parse *pParse, Expr *pExpr){ Walker w; if( pParse->cookieGoto ) return; if( OptimizationDisabled(pParse->db, SQLITE_FactorOutConst) ) return; + memset(&w, 0, sizeof(w)); w.xExprCallback = evalConstExpr; - w.xSelectCallback = 0; w.pParse = pParse; sqlite3WalkExpr(&w, pExpr); } diff --git a/src/os_unix.c b/src/os_unix.c index 7448afe4c..6eed67407 100644 --- a/src/os_unix.c +++ b/src/os_unix.c @@ -126,7 +126,7 @@ #include <time.h> #include <sys/time.h> #include <errno.h> -#ifndef SQLITE_OMIT_WAL +#if !defined(SQLITE_OMIT_WAL) || SQLITE_MAX_MMAP_SIZE>0 #include <sys/mman.h> #endif @@ -3194,46 +3194,59 @@ static int unixRead( } /* -** Seek to the offset in id->offset then read cnt bytes into pBuf. -** Return the number of bytes actually read. Update the offset. -** -** To avoid stomping the errno value on a failed write the lastErrno value -** is set before returning. +** Attempt to seek the file-descriptor passed as the first argument to +** absolute offset iOff, then attempt to write nBuf bytes of data from +** pBuf to it. If an error occurs, return -1 and set *piErrno. Otherwise, +** return the actual number of bytes written (which may be less than +** nBuf). */ -static int seekAndWrite(unixFile *id, i64 offset, const void *pBuf, int cnt){ - int got; -#if (!defined(USE_PREAD) && !defined(USE_PREAD64)) - i64 newOffset; -#endif - assert( cnt==(cnt&0x1ffff) ); - cnt &= 0x1ffff; +static int seekAndWriteFd( + int fd, /* File descriptor to write to */ + i64 iOff, /* File offset to begin writing at */ + const void *pBuf, /* Copy data from this buffer to the file */ + int nBuf, /* Size of buffer pBuf in bytes */ + int *piErrno /* OUT: Error number if error occurs */ +){ + int rc = 0; /* Value returned by system call */ + + assert( nBuf==(nBuf&0x1ffff) ); + nBuf &= 0x1ffff; TIMER_START; + #if defined(USE_PREAD) - do{ got = osPwrite(id->h, pBuf, cnt, offset); }while( got<0 && errno==EINTR ); + do{ rc = osPwrite(fd, pBuf, nBuf, iOff); }while( rc<0 && errno==EINTR ); #elif defined(USE_PREAD64) - do{ got = osPwrite64(id->h, pBuf, cnt, offset);}while( got<0 && errno==EINTR); + do{ rc = osPwrite64(fd, pBuf, nBuf, iOff);}while( rc<0 && errno==EINTR); #else do{ - newOffset = lseek(id->h, offset, SEEK_SET); - SimulateIOError( newOffset-- ); - if( newOffset!=offset ){ - if( newOffset == -1 ){ - ((unixFile*)id)->lastErrno = errno; - }else{ - ((unixFile*)id)->lastErrno = 0; - } + i64 iSeek = lseek(fd, iOff, SEEK_SET); + SimulateIOError( iSeek-- ); + + if( iSeek!=iOff ){ + if( piErrno ) *piErrno = (iSeek==-1 ? errno : 0); return -1; } - got = osWrite(id->h, pBuf, cnt); - }while( got<0 && errno==EINTR ); + rc = osWrite(fd, pBuf, nBuf); + }while( rc<0 && errno==EINTR ); #endif + TIMER_END; - if( got<0 ){ - ((unixFile*)id)->lastErrno = errno; - } + OSTRACE(("WRITE %-3d %5d %7lld %llu\n", fd, rc, iOff, TIMER_ELAPSED)); + + if( rc<0 && piErrno ) *piErrno = errno; + return rc; +} + - OSTRACE(("WRITE %-3d %5d %7lld %llu\n", id->h, got, offset, TIMER_ELAPSED)); - return got; +/* +** Seek to the offset in id->offset then read cnt bytes into pBuf. +** Return the number of bytes actually read. Update the offset. +** +** To avoid stomping the errno value on a failed write the lastErrno value +** is set before returning. +*/ +static int seekAndWrite(unixFile *id, i64 offset, const void *pBuf, int cnt){ + return seekAndWriteFd(id->h, offset, pBuf, cnt, &id->lastErrno); } @@ -4322,24 +4335,32 @@ static int unixShmMap( if( sStat.st_size<nByte ){ /* The requested memory region does not exist. If bExtend is set to ** false, exit early. *pp will be set to NULL and SQLITE_OK returned. - ** - ** Alternatively, if bExtend is true, use ftruncate() to allocate - ** the requested memory region. */ - if( !bExtend ) goto shmpage_out; -#if defined(HAVE_POSIX_FALLOCATE) && HAVE_POSIX_FALLOCATE - if( osFallocate(pShmNode->h, sStat.st_size, nByte)!=0 ){ - rc = unixLogError(SQLITE_IOERR_SHMSIZE, "fallocate", - pShmNode->zFilename); + if( !bExtend ){ goto shmpage_out; } -#else - if( robust_ftruncate(pShmNode->h, nByte) ){ - rc = unixLogError(SQLITE_IOERR_SHMSIZE, "ftruncate", - pShmNode->zFilename); - goto shmpage_out; + + /* Alternatively, if bExtend is true, extend the file. Do this by + ** writing a single byte to the end of each (OS) page being + ** allocated or extended. Technically, we need only write to the + ** last page in order to extend the file. But writing to all new + ** pages forces the OS to allocate them immediately, which reduces + ** the chances of SIGBUS while accessing the mapped region later on. + */ + else{ + static const int pgsz = 4096; + int iPg; + + /* Write to the last byte of each newly allocated or extended page */ + assert( (nByte % pgsz)==0 ); + for(iPg=(sStat.st_size/pgsz); iPg<(nByte/pgsz); iPg++){ + if( seekAndWriteFd(pShmNode->h, iPg*pgsz + pgsz-1, "", 1, 0)!=1 ){ + const char *zFile = pShmNode->zFilename; + rc = unixLogError(SQLITE_IOERR_SHMSIZE, "write", zFile); + goto shmpage_out; + } + } } -#endif } } @@ -5266,9 +5287,8 @@ static int fillInUnixFile( if( h>=0 ) robust_close(pNew, h, __LINE__); h = -1; osUnlink(zFilename); - isDelete = 0; + pNew->ctrlFlags |= UNIXFILE_DELETE; } - if( isDelete ) pNew->ctrlFlags |= UNIXFILE_DELETE; #endif if( rc!=SQLITE_OK ){ if( h>=0 ) robust_close(pNew, h, __LINE__); diff --git a/src/pager.c b/src/pager.c index 1c8b5df65..1c6a84fea 100644 --- a/src/pager.c +++ b/src/pager.c @@ -2871,10 +2871,13 @@ static int readDbPage(PgHdr *pPg, u32 iFrame){ return SQLITE_OK; } +#ifndef SQLITE_OMIT_WAL if( iFrame ){ /* Try to pull the page from the write-ahead log. */ rc = sqlite3WalReadFrame(pPager->pWal, iFrame, pgsz, pPg->pData); - }else{ + }else +#endif + { i64 iOffset = (pgno-1)*(i64)pPager->pageSize; rc = sqlite3OsRead(pPager->fd, pPg->pData, pgsz, iOffset); if( rc==SQLITE_IOERR_SHORT_READ ){ @@ -5927,6 +5930,11 @@ static int pager_incr_changecounter(Pager *pPager, int isDirectMode){ pPager->aStat[PAGER_STAT_WRITE]++; } if( rc==SQLITE_OK ){ + /* Update the pager's copy of the change-counter. Otherwise, the + ** next time a read transaction is opened the cache will be + ** flushed (as the change-counter values will not match). */ + const void *pCopy = (const void *)&((const char *)zBuf)[24]; + memcpy(&pPager->dbFileVers, pCopy, sizeof(pPager->dbFileVers)); pPager->changeCountDone = 1; } }else{ diff --git a/src/parse.y b/src/parse.y index abc0e7dc0..8310b2698 100644 --- a/src/parse.y +++ b/src/parse.y @@ -520,7 +520,9 @@ seltablist(A) ::= stl_prefix(X) nm(Y) dbnm(D) as(Z) indexed_opt(I) struct SrcList_item *pOld = F->a; pNew->zName = pOld->zName; pNew->zDatabase = pOld->zDatabase; + pNew->pSelect = pOld->pSelect; pOld->zName = pOld->zDatabase = 0; + pOld->pSelect = 0; } sqlite3SrcListDelete(pParse->db, F); }else{ diff --git a/src/pragma.c b/src/pragma.c index 65efbd8b2..3056a7d8e 100644 --- a/src/pragma.c +++ b/src/pragma.c @@ -1567,6 +1567,11 @@ void sqlite3Pragma( ** PRAGMA [database.]user_version ** PRAGMA [database.]user_version = <integer> ** + ** PRAGMA [database.]freelist_count = <integer> + ** + ** PRAGMA [database.]application_id + ** PRAGMA [database.]application_id = <integer> + ** ** The pragma's schema_version and user_version are used to set or get ** the value of the schema-version and user-version, respectively. Both ** the schema-version and the user-version are 32-bit signed integers @@ -1588,10 +1593,14 @@ void sqlite3Pragma( if( sqlite3StrICmp(zLeft, "schema_version")==0 || sqlite3StrICmp(zLeft, "user_version")==0 || sqlite3StrICmp(zLeft, "freelist_count")==0 + || sqlite3StrICmp(zLeft, "application_id")==0 ){ int iCookie; /* Cookie index. 1 for schema-cookie, 6 for user-cookie. */ sqlite3VdbeUsesBtree(v, iDb); switch( zLeft[0] ){ + case 'a': case 'A': + iCookie = BTREE_APPLICATION_ID; + break; case 'f': case 'F': iCookie = BTREE_FREE_PAGE_COUNT; break; diff --git a/src/resolve.c b/src/resolve.c index 9b350caf8..a8e196926 100644 --- a/src/resolve.c +++ b/src/resolve.c @@ -1283,6 +1283,7 @@ int sqlite3ResolveExprNames( #endif savedHasAgg = pNC->ncFlags & NC_HasAgg; pNC->ncFlags &= ~NC_HasAgg; + memset(&w, 0, sizeof(w)); w.xExprCallback = resolveExprStep; w.xSelectCallback = resolveSelectStep; w.pParse = pNC->pParse; @@ -1323,6 +1324,7 @@ void sqlite3ResolveSelectNames( Walker w; assert( p!=0 ); + memset(&w, 0, sizeof(w)); w.xExprCallback = resolveExprStep; w.xSelectCallback = resolveSelectStep; w.pParse = pParse; diff --git a/src/select.c b/src/select.c index c7948096c..a745dc370 100644 --- a/src/select.c +++ b/src/select.c @@ -3576,6 +3576,7 @@ static int exprWalkNoop(Walker *NotUsed, Expr *NotUsed2){ */ static void sqlite3SelectExpand(Parse *pParse, Select *pSelect){ Walker w; + memset(&w, 0, sizeof(w)); w.xSelectCallback = selectExpander; w.xExprCallback = exprWalkNoop; w.pParse = pParse; @@ -3634,9 +3635,11 @@ static int selectAddSubqueryTypeInfo(Walker *pWalker, Select *p){ static void sqlite3SelectAddTypeInfo(Parse *pParse, Select *pSelect){ #ifndef SQLITE_OMIT_SUBQUERY Walker w; + memset(&w, 0, sizeof(w)); w.xSelectCallback = selectAddSubqueryTypeInfo; w.xExprCallback = exprWalkNoop; w.pParse = pParse; + w.bSelectDepthFirst = 1; sqlite3WalkSelect(&w, pSelect); #endif } @@ -4047,7 +4050,7 @@ int sqlite3Select( pItem->addrFillSub = topAddr+1; VdbeNoopComment((v, "materialize %s", pItem->pTab->zName)); if( pItem->isCorrelated==0 ){ - /* If the subquery is no correlated and if we are not inside of + /* If the subquery is not correlated and if we are not inside of ** a trigger, then we only need to compute the value of the subquery ** once. */ onceAddr = sqlite3CodeOnce(pParse); diff --git a/src/shell.c b/src/shell.c index 13a2f70cb..09c3b810c 100644 --- a/src/shell.c +++ b/src/shell.c @@ -1481,18 +1481,6 @@ static void open_db(struct callback_data *p){ #ifndef SQLITE_OMIT_LOAD_EXTENSION sqlite3_enable_load_extension(p->db, 1); #endif -#ifdef SQLITE_ENABLE_REGEXP - { - extern int sqlite3_add_regexp_func(sqlite3*); - sqlite3_add_regexp_func(db); - } -#endif -#ifdef SQLITE_ENABLE_SPELLFIX - { - extern int sqlite3_spellfix1_register(sqlite3*); - sqlite3_spellfix1_register(db); - } -#endif } } diff --git a/src/sqlite.h.in b/src/sqlite.h.in index 548a8ffd4..e277185ff 100644 --- a/src/sqlite.h.in +++ b/src/sqlite.h.in @@ -1587,7 +1587,9 @@ struct sqlite3_mem_methods { ** page cache implementation into that object.)^ </dd> ** ** [[SQLITE_CONFIG_LOG]] <dt>SQLITE_CONFIG_LOG</dt> -** <dd> ^The SQLITE_CONFIG_LOG option takes two arguments: a pointer to a +** <dd> The SQLITE_CONFIG_LOG option is used to configure the SQLite +** global [error log]. +** (^The SQLITE_CONFIG_LOG option takes two arguments: a pointer to a ** function with a call signature of void(*)(void*,int,const char*), ** and a pointer to void. ^If the function pointer is not NULL, it is ** invoked by [sqlite3_log()] to process each logging event. ^If the @@ -2522,6 +2524,9 @@ int sqlite3_set_authorizer( ** as each triggered subprogram is entered. The callbacks for triggers ** contain a UTF-8 SQL comment that identifies the trigger.)^ ** +** The [SQLITE_TRACE_SIZE_LIMIT] compile-time option can be used to limit +** the length of [bound parameter] expansion in the output of sqlite3_trace(). +** ** ^The callback function registered by sqlite3_profile() is invoked ** as each SQL statement finishes. ^The profile callback contains ** the original statement text and an estimate of wall-clock time @@ -3060,7 +3065,8 @@ int sqlite3_limit(sqlite3*, int id, int newVal); ** <li> ** ^If the database schema changes, instead of returning [SQLITE_SCHEMA] as it ** always used to do, [sqlite3_step()] will automatically recompile the SQL -** statement and try to run it again. +** statement and try to run it again. As many as [SQLITE_MAX_SCHEMA_RETRY] +** retries will occur before sqlite3_step() gives up and returns an error. ** </li> ** ** <li> @@ -3264,6 +3270,9 @@ typedef struct sqlite3_context sqlite3_context; ** parameter [SQLITE_LIMIT_VARIABLE_NUMBER] (default value: 999). ** ** ^The third argument is the value to bind to the parameter. +** ^If the third parameter to sqlite3_bind_text() or sqlite3_bind_text16() +** or sqlite3_bind_blob() is a NULL pointer then the fourth parameter +** is ignored and the end result is the same as sqlite3_bind_null(). ** ** ^(In those routines that have a fourth argument, its value is the ** number of bytes in the parameter. To be clear: the value is the @@ -6872,7 +6881,7 @@ int sqlite3_strglob(const char *zGlob, const char *zStr); /* ** CAPI3REF: Error Logging Interface ** -** ^The [sqlite3_log()] interface writes a message into the error log +** ^The [sqlite3_log()] interface writes a message into the [error log] ** established by the [SQLITE_CONFIG_LOG] option to [sqlite3_config()]. ** ^If logging is enabled, the zFormat string and subsequent arguments are ** used with [sqlite3_snprintf()] to generate the final output string. diff --git a/src/sqliteInt.h b/src/sqliteInt.h index 7cde2bcaf..611f7a556 100644 --- a/src/sqliteInt.h +++ b/src/sqliteInt.h @@ -563,7 +563,7 @@ extern const int sqlite3one; || defined(_WIN32) \ || (defined(__APPLE__) && defined(__MACH__)) \ || defined(__sun) -# define SQLITE_MAX_MMAP_SIZE 2147483648 +# define SQLITE_MAX_MMAP_SIZE 0x7fff0000 /* 2147418112 */ # else # define SQLITE_MAX_MMAP_SIZE 0 # endif @@ -2600,6 +2600,7 @@ struct Walker { int (*xSelectCallback)(Walker*,Select*); /* Callback for SELECTs */ Parse *pParse; /* Parser context. */ int walkerDepth; /* Number of subqueries */ + u8 bSelectDepthFirst; /* Do subqueries first */ union { /* Extra data for callback */ NameContext *pNC; /* Naming context */ int i; /* Integer value */ diff --git a/src/tclsqlite.c b/src/tclsqlite.c index 3b175ce2a..9bd111d94 100644 --- a/src/tclsqlite.c +++ b/src/tclsqlite.c @@ -3829,12 +3829,9 @@ static void init_all(Tcl_Interp *interp){ extern int Sqlitemultiplex_Init(Tcl_Interp*); extern int SqliteSuperlock_Init(Tcl_Interp*); extern int SqlitetestSyscall_Init(Tcl_Interp*); - extern int Sqlitetestfuzzer_Init(Tcl_Interp*); - extern int Sqlitetestwholenumber_Init(Tcl_Interp*); #if defined(SQLITE_ENABLE_SESSION) && defined(SQLITE_ENABLE_PREUPDATE_HOOK) extern int TestSession_Init(Tcl_Interp*); #endif - extern int Sqlitetestregexp_Init(Tcl_Interp*); #if defined(SQLITE_ENABLE_FTS3) || defined(SQLITE_ENABLE_FTS4) extern int Sqlitetestfts3_Init(Tcl_Interp *interp); @@ -3877,12 +3874,9 @@ static void init_all(Tcl_Interp *interp){ Sqlitemultiplex_Init(interp); SqliteSuperlock_Init(interp); SqlitetestSyscall_Init(interp); - Sqlitetestfuzzer_Init(interp); - Sqlitetestwholenumber_Init(interp); #if defined(SQLITE_ENABLE_SESSION) && defined(SQLITE_ENABLE_PREUPDATE_HOOK) TestSession_Init(interp); #endif - Sqlitetestregexp_Init(interp); #if defined(SQLITE_ENABLE_FTS3) || defined(SQLITE_ENABLE_FTS4) Sqlitetestfts3_Init(interp); diff --git a/src/test1.c b/src/test1.c index 63d014cc1..34370d48e 100644 --- a/src/test1.c +++ b/src/test1.c @@ -6045,6 +6045,69 @@ static int optimization_control( return TCL_OK; } +typedef struct sqlite3_api_routines sqlite3_api_routines; +/* +** load_static_extension DB NAME ... +** +** Load one or more statically linked extensions. +*/ +static int tclLoadStaticExtensionCmd( + void * clientData, + Tcl_Interp *interp, + int objc, + Tcl_Obj *CONST objv[] +){ + extern int sqlite3_amatch_init(sqlite3*,char**,const sqlite3_api_routines*); + extern int sqlite3_closure_init(sqlite3*,char**,const sqlite3_api_routines*); + extern int sqlite3_fuzzer_init(sqlite3*,char**,const sqlite3_api_routines*); + extern int sqlite3_ieee_init(sqlite3*,char**,const sqlite3_api_routines*); + extern int sqlite3_nextchar_init(sqlite3*,char**,const sqlite3_api_routines*); + extern int sqlite3_regexp_init(sqlite3*,char**,const sqlite3_api_routines*); + extern int sqlite3_spellfix_init(sqlite3*,char**,const sqlite3_api_routines*); + extern int sqlite3_wholenumber_init(sqlite3*,char**,const sqlite3_api_routines*); + static const struct { + const char *zExtName; + int (*pInit)(sqlite3*,char**,const sqlite3_api_routines*); + } aExtension[] = { + { "amatch", sqlite3_amatch_init }, + { "closure", sqlite3_closure_init }, + { "fuzzer", sqlite3_fuzzer_init }, + { "ieee754", sqlite3_ieee_init }, + { "nextchar", sqlite3_nextchar_init }, + { "regexp", sqlite3_regexp_init }, + { "spellfix", sqlite3_spellfix_init }, + { "wholenumber", sqlite3_wholenumber_init }, + }; + sqlite3 *db; + const char *zName; + int i, j, rc; + char *zErrMsg = 0; + if( objc<3 ){ + Tcl_WrongNumArgs(interp, 1, objv, "DB NAME ..."); + return TCL_ERROR; + } + if( getDbPointer(interp, Tcl_GetString(objv[1]), &db) ) return TCL_ERROR; + for(j=2; j<objc; j++){ + zName = Tcl_GetString(objv[j]); + for(i=0; i<ArraySize(aExtension); i++){ + if( strcmp(zName, aExtension[i].zExtName)==0 ) break; + } + if( i>=ArraySize(aExtension) ){ + Tcl_AppendResult(interp, "no such extension: ", zName, (char*)0); + return TCL_ERROR; + } + rc = aExtension[i].pInit(db, &zErrMsg, 0); + if( rc!=SQLITE_OK || zErrMsg ){ + Tcl_AppendResult(interp, "initialization of ", zName, " failed: ", zErrMsg, + (char*)0); + sqlite3_free(zErrMsg); + return TCL_ERROR; + } + } + return TCL_OK; +} + + /* ** Register commands with the TCL interpreter. */ @@ -6266,6 +6329,7 @@ int Sqlitetest1_Init(Tcl_Interp *interp){ #if SQLITE_OS_UNIX { "getrusage", test_getrusage }, #endif + { "load_static_extension", tclLoadStaticExtensionCmd }, }; static int bitmask_size = sizeof(Bitmask)*8; int i; diff --git a/src/test8.c b/src/test8.c index 6bd4e396b..f7b2e6813 100644 --- a/src/test8.c +++ b/src/test8.c @@ -1370,29 +1370,6 @@ static int declare_vtab( return TCL_OK; } -#include "test_spellfix.c" - -/* -** Register the spellfix virtual table module. -*/ -static int register_spellfix_module( - ClientData clientData, - Tcl_Interp *interp, - int objc, - Tcl_Obj *CONST objv[] -){ - sqlite3 *db; - - if( objc!=2 ){ - Tcl_WrongNumArgs(interp, 1, objv, "DB"); - return TCL_ERROR; - } - if( getDbPointer(interp, Tcl_GetString(objv[1]), &db) ) return TCL_ERROR; - - sqlite3_spellfix1_register(db); - return TCL_OK; -} - #endif /* ifndef SQLITE_OMIT_VIRTUALTABLE */ /* @@ -1406,7 +1383,6 @@ int Sqlitetest8_Init(Tcl_Interp *interp){ void *clientData; } aObjCmd[] = { { "register_echo_module", register_echo_module, 0 }, - { "register_spellfix_module", register_spellfix_module, 0 }, { "sqlite3_declare_vtab", declare_vtab, 0 }, }; int i; diff --git a/src/test_fuzzer.c b/src/test_fuzzer.c deleted file mode 100644 index 10496f2ea..000000000 --- a/src/test_fuzzer.c +++ /dev/null @@ -1,1215 +0,0 @@ -/* -** 2011 March 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. -** -************************************************************************* -** -** Code for a demonstration virtual table that generates variations -** on an input word at increasing edit distances from the original. -** -** A fuzzer virtual table is created like this: -** -** CREATE VIRTUAL TABLE f USING fuzzer(<fuzzer-data-table>); -** -** When it is created, the new fuzzer table must be supplied with the -** name of a "fuzzer data table", which must reside in the same database -** file as the new fuzzer table. The fuzzer data table contains the various -** transformations and their costs that the fuzzer logic uses to generate -** variations. -** -** The fuzzer data table must contain exactly four columns (more precisely, -** the statement "SELECT * FROM <fuzzer_data_table>" must return records -** that consist of four columns). It does not matter what the columns are -** named. -** -** Each row in the fuzzer data table represents a single character -** transformation. The left most column of the row (column 0) contains an -** integer value - the identifier of the ruleset to which the transformation -** rule belongs (see "MULTIPLE RULE SETS" below). The second column of the -** row (column 0) contains the input character or characters. The third -** column contains the output character or characters. And the fourth column -** contains the integer cost of making the transformation. For example: -** -** CREATE TABLE f_data(ruleset, cFrom, cTo, Cost); -** INSERT INTO f_data(ruleset, cFrom, cTo, Cost) VALUES(0, '', 'a', 100); -** INSERT INTO f_data(ruleset, cFrom, cTo, Cost) VALUES(0, 'b', '', 87); -** INSERT INTO f_data(ruleset, cFrom, cTo, Cost) VALUES(0, 'o', 'oe', 38); -** INSERT INTO f_data(ruleset, cFrom, cTo, Cost) VALUES(0, 'oe', 'o', 40); -** -** The first row inserted into the fuzzer data table by the SQL script -** above indicates that the cost of inserting a letter 'a' is 100. (All -** costs are integers. We recommend that costs be scaled so that the -** average cost is around 100.) The second INSERT statement creates a rule -** saying that the cost of deleting a single letter 'b' is 87. The third -** and fourth INSERT statements mean that the cost of transforming a -** single letter "o" into the two-letter sequence "oe" is 38 and that the -** cost of transforming "oe" back into "o" is 40. -** -** The contents of the fuzzer data table are loaded into main memory when -** a fuzzer table is first created, and may be internally reloaded by the -** system at any subsequent time. Therefore, the fuzzer data table should be -** populated before the fuzzer table is created and not modified thereafter. -** If you do need to modify the contents of the fuzzer data table, it is -** recommended that the associated fuzzer table be dropped, the fuzzer data -** table edited, and the fuzzer table recreated within a single transaction. -** Alternatively, the fuzzer data table can be edited then the database -** connection can be closed and reopened. -** -** Once it has been created, the fuzzer table can be queried as follows: -** -** SELECT word, distance FROM f -** WHERE word MATCH 'abcdefg' -** AND distance<200; -** -** This first query outputs the string "abcdefg" and all strings that -** can be derived from that string by appling the specified transformations. -** The strings are output together with their total transformation cost -** (called "distance") and appear in order of increasing cost. No string -** is output more than once. If there are multiple ways to transform the -** target string into the output string then the lowest cost transform is -** the one that is returned. In the example, the search is limited to -** strings with a total distance of less than 200. -** -** The fuzzer is a read-only table. Any attempt to DELETE, INSERT, or -** UPDATE on a fuzzer table will throw an error. -** -** It is important to put some kind of a limit on the fuzzer output. This -** can be either in the form of a LIMIT clause at the end of the query, -** or better, a "distance<NNN" constraint where NNN is some number. The -** running time and memory requirement is exponential in the value of NNN -** so you want to make sure that NNN is not too big. A value of NNN that -** is about twice the average transformation cost seems to give good results. -** -** The fuzzer table can be useful for tasks such as spelling correction. -** Suppose there is a second table vocabulary(w) where the w column contains -** all correctly spelled words. Let $word be a word you want to look up. -** -** SELECT vocabulary.w FROM f, vocabulary -** WHERE f.word MATCH $word -** AND f.distance<=200 -** AND f.word=vocabulary.w -** LIMIT 20 -** -** The query above gives the 20 closest words to the $word being tested. -** (Note that for good performance, the vocubulary.w column should be -** indexed.) -** -** A similar query can be used to find all words in the dictionary that -** begin with some prefix $prefix: -** -** SELECT vocabulary.w FROM f, vocabulary -** WHERE f.word MATCH $prefix -** AND f.distance<=200 -** AND vocabulary.w BETWEEN f.word AND (f.word || x'F7BFBFBF') -** LIMIT 50 -** -** This last query will show up to 50 words out of the vocabulary that -** match or nearly match the $prefix. -** -** MULTIPLE RULE SETS -** -** Normally, the "ruleset" value associated with all character transformations -** in the fuzzer data table is zero. However, if required, the fuzzer table -** allows multiple rulesets to be defined. Each query uses only a single -** ruleset. This allows, for example, a single fuzzer table to support -** multiple languages. -** -** By default, only the rules from ruleset 0 are used. To specify an -** alternative ruleset, a "ruleset = ?" expression must be added to the -** WHERE clause of a SELECT, where ? is the identifier of the desired -** ruleset. For example: -** -** SELECT vocabulary.w FROM f, vocabulary -** WHERE f.word MATCH $word -** AND f.distance<=200 -** AND f.word=vocabulary.w -** AND f.ruleset=1 -- Specify the ruleset to use here -** LIMIT 20 -** -** If no "ruleset = ?" constraint is specified in the WHERE clause, ruleset -** 0 is used. -** -** LIMITS -** -** The maximum ruleset number is 2147483647. The maximum length of either -** of the strings in the second or third column of the fuzzer data table -** is 50 bytes. The maximum cost on a rule is 1000. -*/ - -/* If SQLITE_DEBUG is not defined, disable assert statements. */ -#if !defined(NDEBUG) && !defined(SQLITE_DEBUG) -# define NDEBUG -#endif - -#include "sqlite3.h" -#include <stdlib.h> -#include <string.h> -#include <assert.h> -#include <stdio.h> - -#ifndef SQLITE_OMIT_VIRTUALTABLE - -/* -** Forward declaration of objects used by this implementation -*/ -typedef struct fuzzer_vtab fuzzer_vtab; -typedef struct fuzzer_cursor fuzzer_cursor; -typedef struct fuzzer_rule fuzzer_rule; -typedef struct fuzzer_seen fuzzer_seen; -typedef struct fuzzer_stem fuzzer_stem; - -/* -** Various types. -** -** fuzzer_cost is the "cost" of an edit operation. -** -** fuzzer_len is the length of a matching string. -** -** fuzzer_ruleid is an ruleset identifier. -*/ -typedef int fuzzer_cost; -typedef signed char fuzzer_len; -typedef int fuzzer_ruleid; - -/* -** Limits -*/ -#define FUZZER_MX_LENGTH 50 /* Maximum length of a rule string */ -#define FUZZER_MX_RULEID 2147483647 /* Maximum rule ID */ -#define FUZZER_MX_COST 1000 /* Maximum single-rule cost */ -#define FUZZER_MX_OUTPUT_LENGTH 100 /* Maximum length of an output string */ - - -/* -** Each transformation rule is stored as an instance of this object. -** All rules are kept on a linked list sorted by rCost. -*/ -struct fuzzer_rule { - fuzzer_rule *pNext; /* Next rule in order of increasing rCost */ - char *zFrom; /* Transform from */ - fuzzer_cost rCost; /* Cost of this transformation */ - fuzzer_len nFrom, nTo; /* Length of the zFrom and zTo strings */ - fuzzer_ruleid iRuleset; /* The rule set to which this rule belongs */ - char zTo[4]; /* Transform to (extra space appended) */ -}; - -/* -** A stem object is used to generate variants. It is also used to record -** previously generated outputs. -** -** Every stem is added to a hash table as it is output. Generation of -** duplicate stems is suppressed. -** -** Active stems (those that might generate new outputs) are kepts on a linked -** list sorted by increasing cost. The cost is the sum of rBaseCost and -** pRule->rCost. -*/ -struct fuzzer_stem { - char *zBasis; /* Word being fuzzed */ - const fuzzer_rule *pRule; /* Current rule to apply */ - fuzzer_stem *pNext; /* Next stem in rCost order */ - fuzzer_stem *pHash; /* Next stem with same hash on zBasis */ - fuzzer_cost rBaseCost; /* Base cost of getting to zBasis */ - fuzzer_cost rCostX; /* Precomputed rBaseCost + pRule->rCost */ - fuzzer_len nBasis; /* Length of the zBasis string */ - fuzzer_len n; /* Apply pRule at this character offset */ -}; - -/* -** A fuzzer virtual-table object -*/ -struct fuzzer_vtab { - sqlite3_vtab base; /* Base class - must be first */ - char *zClassName; /* Name of this class. Default: "fuzzer" */ - fuzzer_rule *pRule; /* All active rules in this fuzzer */ - int nCursor; /* Number of active cursors */ -}; - -#define FUZZER_HASH 4001 /* Hash table size */ -#define FUZZER_NQUEUE 20 /* Number of slots on the stem queue */ - -/* A fuzzer cursor object */ -struct fuzzer_cursor { - sqlite3_vtab_cursor base; /* Base class - must be first */ - sqlite3_int64 iRowid; /* The rowid of the current word */ - fuzzer_vtab *pVtab; /* The virtual table this cursor belongs to */ - fuzzer_cost rLimit; /* Maximum cost of any term */ - fuzzer_stem *pStem; /* Stem with smallest rCostX */ - fuzzer_stem *pDone; /* Stems already processed to completion */ - fuzzer_stem *aQueue[FUZZER_NQUEUE]; /* Queue of stems with higher rCostX */ - int mxQueue; /* Largest used index in aQueue[] */ - char *zBuf; /* Temporary use buffer */ - int nBuf; /* Bytes allocated for zBuf */ - int nStem; /* Number of stems allocated */ - int iRuleset; /* Only process rules from this ruleset */ - fuzzer_rule nullRule; /* Null rule used first */ - fuzzer_stem *apHash[FUZZER_HASH]; /* Hash of previously generated terms */ -}; - -/* -** The two input rule lists are both sorted in order of increasing -** cost. Merge them together into a single list, sorted by cost, and -** return a pointer to the head of that list. -*/ -static fuzzer_rule *fuzzerMergeRules(fuzzer_rule *pA, fuzzer_rule *pB){ - fuzzer_rule head; - fuzzer_rule *pTail; - - pTail = &head; - while( pA && pB ){ - if( pA->rCost<=pB->rCost ){ - pTail->pNext = pA; - pTail = pA; - pA = pA->pNext; - }else{ - pTail->pNext = pB; - pTail = pB; - pB = pB->pNext; - } - } - if( pA==0 ){ - pTail->pNext = pB; - }else{ - pTail->pNext = pA; - } - return head.pNext; -} - -/* -** Statement pStmt currently points to a row in the fuzzer data table. This -** function allocates and populates a fuzzer_rule structure according to -** the content of the row. -** -** If successful, *ppRule is set to point to the new object and SQLITE_OK -** is returned. Otherwise, *ppRule is zeroed, *pzErr may be set to point -** to an error message and an SQLite error code returned. -*/ -static int fuzzerLoadOneRule( - fuzzer_vtab *p, /* Fuzzer virtual table handle */ - sqlite3_stmt *pStmt, /* Base rule on statements current row */ - fuzzer_rule **ppRule, /* OUT: New rule object */ - char **pzErr /* OUT: Error message */ -){ - sqlite3_int64 iRuleset = sqlite3_column_int64(pStmt, 0); - const char *zFrom = (const char *)sqlite3_column_text(pStmt, 1); - const char *zTo = (const char *)sqlite3_column_text(pStmt, 2); - int nCost = sqlite3_column_int(pStmt, 3); - - int rc = SQLITE_OK; /* Return code */ - int nFrom; /* Size of string zFrom, in bytes */ - int nTo; /* Size of string zTo, in bytes */ - fuzzer_rule *pRule = 0; /* New rule object to return */ - - if( zFrom==0 ) zFrom = ""; - if( zTo==0 ) zTo = ""; - nFrom = (int)strlen(zFrom); - nTo = (int)strlen(zTo); - - /* Silently ignore null transformations */ - if( strcmp(zFrom, zTo)==0 ){ - *ppRule = 0; - return SQLITE_OK; - } - - if( nCost<=0 || nCost>FUZZER_MX_COST ){ - *pzErr = sqlite3_mprintf("%s: cost must be between 1 and %d", - p->zClassName, FUZZER_MX_COST - ); - rc = SQLITE_ERROR; - }else - if( nFrom>FUZZER_MX_LENGTH || nTo>FUZZER_MX_LENGTH ){ - *pzErr = sqlite3_mprintf("%s: maximum string length is %d", - p->zClassName, FUZZER_MX_LENGTH - ); - rc = SQLITE_ERROR; - }else - if( iRuleset<0 || iRuleset>FUZZER_MX_RULEID ){ - *pzErr = sqlite3_mprintf("%s: ruleset must be between 0 and %d", - p->zClassName, FUZZER_MX_RULEID - ); - rc = SQLITE_ERROR; - }else{ - - pRule = sqlite3_malloc( sizeof(*pRule) + nFrom + nTo ); - if( pRule==0 ){ - rc = SQLITE_NOMEM; - }else{ - memset(pRule, 0, sizeof(*pRule)); - pRule->zFrom = &pRule->zTo[nTo+1]; - pRule->nFrom = nFrom; - memcpy(pRule->zFrom, zFrom, nFrom+1); - memcpy(pRule->zTo, zTo, nTo+1); - pRule->nTo = nTo; - pRule->rCost = nCost; - pRule->iRuleset = (int)iRuleset; - } - } - - *ppRule = pRule; - return rc; -} - -/* -** Load the content of the fuzzer data table into memory. -*/ -static int fuzzerLoadRules( - sqlite3 *db, /* Database handle */ - fuzzer_vtab *p, /* Virtual fuzzer table to configure */ - const char *zDb, /* Database containing rules data */ - const char *zData, /* Table containing rules data */ - char **pzErr /* OUT: Error message */ -){ - int rc = SQLITE_OK; /* Return code */ - char *zSql; /* SELECT used to read from rules table */ - fuzzer_rule *pHead = 0; - - zSql = sqlite3_mprintf("SELECT * FROM %Q.%Q", zDb, zData); - if( zSql==0 ){ - rc = SQLITE_NOMEM; - }else{ - int rc2; /* finalize() return code */ - sqlite3_stmt *pStmt = 0; - rc = sqlite3_prepare_v2(db, zSql, -1, &pStmt, 0); - if( rc!=SQLITE_OK ){ - *pzErr = sqlite3_mprintf("%s: %s", p->zClassName, sqlite3_errmsg(db)); - }else if( sqlite3_column_count(pStmt)!=4 ){ - *pzErr = sqlite3_mprintf("%s: %s has %d columns, expected 4", - p->zClassName, zData, sqlite3_column_count(pStmt) - ); - rc = SQLITE_ERROR; - }else{ - while( rc==SQLITE_OK && SQLITE_ROW==sqlite3_step(pStmt) ){ - fuzzer_rule *pRule = 0; - rc = fuzzerLoadOneRule(p, pStmt, &pRule, pzErr); - if( pRule ){ - pRule->pNext = pHead; - pHead = pRule; - } - } - } - rc2 = sqlite3_finalize(pStmt); - if( rc==SQLITE_OK ) rc = rc2; - } - sqlite3_free(zSql); - - /* All rules are now in a singly linked list starting at pHead. This - ** block sorts them by cost and then sets fuzzer_vtab.pRule to point to - ** point to the head of the sorted list. - */ - if( rc==SQLITE_OK ){ - unsigned int i; - fuzzer_rule *pX; - fuzzer_rule *a[15]; - for(i=0; i<sizeof(a)/sizeof(a[0]); i++) a[i] = 0; - while( (pX = pHead)!=0 ){ - pHead = pX->pNext; - pX->pNext = 0; - for(i=0; a[i] && i<sizeof(a)/sizeof(a[0])-1; i++){ - pX = fuzzerMergeRules(a[i], pX); - a[i] = 0; - } - a[i] = fuzzerMergeRules(a[i], pX); - } - for(pX=a[0], i=1; i<sizeof(a)/sizeof(a[0]); i++){ - pX = fuzzerMergeRules(a[i], pX); - } - p->pRule = fuzzerMergeRules(p->pRule, pX); - }else{ - /* An error has occurred. Setting p->pRule to point to the head of the - ** allocated list ensures that the list will be cleaned up in this case. - */ - assert( p->pRule==0 ); - p->pRule = pHead; - } - - return rc; -} - -/* -** This function converts an SQL quoted string into an unquoted string -** and returns a pointer to a buffer allocated using sqlite3_malloc() -** containing the result. The caller should eventually free this buffer -** using sqlite3_free. -** -** Examples: -** -** "abc" becomes abc -** 'xyz' becomes xyz -** [pqr] becomes pqr -** `mno` becomes mno -*/ -static char *fuzzerDequote(const char *zIn){ - int nIn; /* Size of input string, in bytes */ - char *zOut; /* Output (dequoted) string */ - - nIn = (int)strlen(zIn); - zOut = sqlite3_malloc(nIn+1); - if( zOut ){ - char q = zIn[0]; /* Quote character (if any ) */ - - if( q!='[' && q!= '\'' && q!='"' && q!='`' ){ - memcpy(zOut, zIn, nIn+1); - }else{ - int iOut = 0; /* Index of next byte to write to output */ - int iIn; /* Index of next byte to read from input */ - - if( q=='[' ) q = ']'; - for(iIn=1; iIn<nIn; iIn++){ - if( zIn[iIn]==q ) iIn++; - zOut[iOut++] = zIn[iIn]; - } - } - assert( (int)strlen(zOut)<=nIn ); - } - return zOut; -} - -/* -** xDisconnect/xDestroy method for the fuzzer module. -*/ -static int fuzzerDisconnect(sqlite3_vtab *pVtab){ - fuzzer_vtab *p = (fuzzer_vtab*)pVtab; - assert( p->nCursor==0 ); - while( p->pRule ){ - fuzzer_rule *pRule = p->pRule; - p->pRule = pRule->pNext; - sqlite3_free(pRule); - } - sqlite3_free(p); - return SQLITE_OK; -} - -/* -** xConnect/xCreate method for the fuzzer module. Arguments are: -** -** argv[0] -> module name ("fuzzer") -** argv[1] -> database name -** argv[2] -> table name -** argv[3] -> fuzzer rule table name -*/ -static int fuzzerConnect( - sqlite3 *db, - void *pAux, - int argc, const char *const*argv, - sqlite3_vtab **ppVtab, - char **pzErr -){ - int rc = SQLITE_OK; /* Return code */ - fuzzer_vtab *pNew = 0; /* New virtual table */ - const char *zModule = argv[0]; - const char *zDb = argv[1]; - - if( argc!=4 ){ - *pzErr = sqlite3_mprintf( - "%s: wrong number of CREATE VIRTUAL TABLE arguments", zModule - ); - rc = SQLITE_ERROR; - }else{ - int nModule; /* Length of zModule, in bytes */ - - nModule = (int)strlen(zModule); - pNew = sqlite3_malloc( sizeof(*pNew) + nModule + 1); - if( pNew==0 ){ - rc = SQLITE_NOMEM; - }else{ - char *zTab; /* Dequoted name of fuzzer data table */ - - memset(pNew, 0, sizeof(*pNew)); - pNew->zClassName = (char*)&pNew[1]; - memcpy(pNew->zClassName, zModule, nModule+1); - - zTab = fuzzerDequote(argv[3]); - if( zTab==0 ){ - rc = SQLITE_NOMEM; - }else{ - rc = fuzzerLoadRules(db, pNew, zDb, zTab, pzErr); - sqlite3_free(zTab); - } - - if( rc==SQLITE_OK ){ - rc = sqlite3_declare_vtab(db, "CREATE TABLE x(word,distance,ruleset)"); - } - if( rc!=SQLITE_OK ){ - fuzzerDisconnect((sqlite3_vtab *)pNew); - pNew = 0; - } - } - } - - *ppVtab = (sqlite3_vtab *)pNew; - return rc; -} - -/* -** Open a new fuzzer cursor. -*/ -static int fuzzerOpen(sqlite3_vtab *pVTab, sqlite3_vtab_cursor **ppCursor){ - fuzzer_vtab *p = (fuzzer_vtab*)pVTab; - fuzzer_cursor *pCur; - pCur = sqlite3_malloc( sizeof(*pCur) ); - if( pCur==0 ) return SQLITE_NOMEM; - memset(pCur, 0, sizeof(*pCur)); - pCur->pVtab = p; - *ppCursor = &pCur->base; - p->nCursor++; - return SQLITE_OK; -} - -/* -** Free all stems in a list. -*/ -static void fuzzerClearStemList(fuzzer_stem *pStem){ - while( pStem ){ - fuzzer_stem *pNext = pStem->pNext; - sqlite3_free(pStem); - pStem = pNext; - } -} - -/* -** Free up all the memory allocated by a cursor. Set it rLimit to 0 -** to indicate that it is at EOF. -*/ -static void fuzzerClearCursor(fuzzer_cursor *pCur, int clearHash){ - int i; - fuzzerClearStemList(pCur->pStem); - fuzzerClearStemList(pCur->pDone); - for(i=0; i<FUZZER_NQUEUE; i++) fuzzerClearStemList(pCur->aQueue[i]); - pCur->rLimit = (fuzzer_cost)0; - if( clearHash && pCur->nStem ){ - pCur->mxQueue = 0; - pCur->pStem = 0; - pCur->pDone = 0; - memset(pCur->aQueue, 0, sizeof(pCur->aQueue)); - memset(pCur->apHash, 0, sizeof(pCur->apHash)); - } - pCur->nStem = 0; -} - -/* -** Close a fuzzer cursor. -*/ -static int fuzzerClose(sqlite3_vtab_cursor *cur){ - fuzzer_cursor *pCur = (fuzzer_cursor *)cur; - fuzzerClearCursor(pCur, 0); - sqlite3_free(pCur->zBuf); - pCur->pVtab->nCursor--; - sqlite3_free(pCur); - return SQLITE_OK; -} - -/* -** Compute the current output term for a fuzzer_stem. -*/ -static int fuzzerRender( - fuzzer_stem *pStem, /* The stem to be rendered */ - char **pzBuf, /* Write results into this buffer. realloc if needed */ - int *pnBuf /* Size of the buffer */ -){ - const fuzzer_rule *pRule = pStem->pRule; - int n; /* Size of output term without nul-term */ - char *z; /* Buffer to assemble output term in */ - - n = pStem->nBasis + pRule->nTo - pRule->nFrom; - if( (*pnBuf)<n+1 ){ - (*pzBuf) = sqlite3_realloc((*pzBuf), n+100); - if( (*pzBuf)==0 ) return SQLITE_NOMEM; - (*pnBuf) = n+100; - } - n = pStem->n; - z = *pzBuf; - if( n<0 ){ - memcpy(z, pStem->zBasis, pStem->nBasis+1); - }else{ - memcpy(z, pStem->zBasis, n); - memcpy(&z[n], pRule->zTo, pRule->nTo); - memcpy(&z[n+pRule->nTo], &pStem->zBasis[n+pRule->nFrom], - pStem->nBasis-n-pRule->nFrom+1); - } - - assert( z[pStem->nBasis + pRule->nTo - pRule->nFrom]==0 ); - return SQLITE_OK; -} - -/* -** Compute a hash on zBasis. -*/ -static unsigned int fuzzerHash(const char *z){ - unsigned int h = 0; - while( *z ){ h = (h<<3) ^ (h>>29) ^ *(z++); } - return h % FUZZER_HASH; -} - -/* -** Current cost of a stem -*/ -static fuzzer_cost fuzzerCost(fuzzer_stem *pStem){ - return pStem->rCostX = pStem->rBaseCost + pStem->pRule->rCost; -} - -#if 0 -/* -** Print a description of a fuzzer_stem on stderr. -*/ -static void fuzzerStemPrint( - const char *zPrefix, - fuzzer_stem *pStem, - const char *zSuffix -){ - if( pStem->n<0 ){ - fprintf(stderr, "%s[%s](%d)-->self%s", - zPrefix, - pStem->zBasis, pStem->rBaseCost, - zSuffix - ); - }else{ - char *zBuf = 0; - int nBuf = 0; - if( fuzzerRender(pStem, &zBuf, &nBuf)!=SQLITE_OK ) return; - fprintf(stderr, "%s[%s](%d)-->{%s}(%d)%s", - zPrefix, - pStem->zBasis, pStem->rBaseCost, zBuf, pStem->, - zSuffix - ); - sqlite3_free(zBuf); - } -} -#endif - -/* -** Return 1 if the string to which the cursor is point has already -** been emitted. Return 0 if not. Return -1 on a memory allocation -** failures. -*/ -static int fuzzerSeen(fuzzer_cursor *pCur, fuzzer_stem *pStem){ - unsigned int h; - fuzzer_stem *pLookup; - - if( fuzzerRender(pStem, &pCur->zBuf, &pCur->nBuf)==SQLITE_NOMEM ){ - return -1; - } - h = fuzzerHash(pCur->zBuf); - pLookup = pCur->apHash[h]; - while( pLookup && strcmp(pLookup->zBasis, pCur->zBuf)!=0 ){ - pLookup = pLookup->pHash; - } - return pLookup!=0; -} - -/* -** If argument pRule is NULL, this function returns false. -** -** Otherwise, it returns true if rule pRule should be skipped. A rule -** should be skipped if it does not belong to rule-set iRuleset, or if -** applying it to stem pStem would create a string longer than -** FUZZER_MX_OUTPUT_LENGTH bytes. -*/ -static int fuzzerSkipRule( - const fuzzer_rule *pRule, /* Determine whether or not to skip this */ - fuzzer_stem *pStem, /* Stem rule may be applied to */ - int iRuleset /* Rule-set used by the current query */ -){ - return pRule && ( - (pRule->iRuleset!=iRuleset) - || (pStem->nBasis + pRule->nTo - pRule->nFrom)>FUZZER_MX_OUTPUT_LENGTH - ); -} - -/* -** Advance a fuzzer_stem to its next value. Return 0 if there are -** no more values that can be generated by this fuzzer_stem. Return -** -1 on a memory allocation failure. -*/ -static int fuzzerAdvance(fuzzer_cursor *pCur, fuzzer_stem *pStem){ - const fuzzer_rule *pRule; - while( (pRule = pStem->pRule)!=0 ){ - assert( pRule==&pCur->nullRule || pRule->iRuleset==pCur->iRuleset ); - while( pStem->n < pStem->nBasis - pRule->nFrom ){ - pStem->n++; - if( pRule->nFrom==0 - || memcmp(&pStem->zBasis[pStem->n], pRule->zFrom, pRule->nFrom)==0 - ){ - /* Found a rewrite case. Make sure it is not a duplicate */ - int rc = fuzzerSeen(pCur, pStem); - if( rc<0 ) return -1; - if( rc==0 ){ - fuzzerCost(pStem); - return 1; - } - } - } - pStem->n = -1; - do{ - pRule = pRule->pNext; - }while( fuzzerSkipRule(pRule, pStem, pCur->iRuleset) ); - pStem->pRule = pRule; - if( pRule && fuzzerCost(pStem)>pCur->rLimit ) pStem->pRule = 0; - } - return 0; -} - -/* -** The two input stem lists are both sorted in order of increasing -** rCostX. Merge them together into a single list, sorted by rCostX, and -** return a pointer to the head of that new list. -*/ -static fuzzer_stem *fuzzerMergeStems(fuzzer_stem *pA, fuzzer_stem *pB){ - fuzzer_stem head; - fuzzer_stem *pTail; - - pTail = &head; - while( pA && pB ){ - if( pA->rCostX<=pB->rCostX ){ - pTail->pNext = pA; - pTail = pA; - pA = pA->pNext; - }else{ - pTail->pNext = pB; - pTail = pB; - pB = pB->pNext; - } - } - if( pA==0 ){ - pTail->pNext = pB; - }else{ - pTail->pNext = pA; - } - return head.pNext; -} - -/* -** Load pCur->pStem with the lowest-cost stem. Return a pointer -** to the lowest-cost stem. -*/ -static fuzzer_stem *fuzzerLowestCostStem(fuzzer_cursor *pCur){ - fuzzer_stem *pBest, *pX; - int iBest; - int i; - - if( pCur->pStem==0 ){ - iBest = -1; - pBest = 0; - for(i=0; i<=pCur->mxQueue; i++){ - pX = pCur->aQueue[i]; - if( pX==0 ) continue; - if( pBest==0 || pBest->rCostX>pX->rCostX ){ - pBest = pX; - iBest = i; - } - } - if( pBest ){ - pCur->aQueue[iBest] = pBest->pNext; - pBest->pNext = 0; - pCur->pStem = pBest; - } - } - return pCur->pStem; -} - -/* -** Insert pNew into queue of pending stems. Then find the stem -** with the lowest rCostX and move it into pCur->pStem. -** list. The insert is done such the pNew is in the correct order -** according to fuzzer_stem.zBaseCost+fuzzer_stem.pRule->rCost. -*/ -static fuzzer_stem *fuzzerInsert(fuzzer_cursor *pCur, fuzzer_stem *pNew){ - fuzzer_stem *pX; - int i; - - /* If pCur->pStem exists and is greater than pNew, then make pNew - ** the new pCur->pStem and insert the old pCur->pStem instead. - */ - if( (pX = pCur->pStem)!=0 && pX->rCostX>pNew->rCostX ){ - pNew->pNext = 0; - pCur->pStem = pNew; - pNew = pX; - } - - /* Insert the new value */ - pNew->pNext = 0; - pX = pNew; - for(i=0; i<=pCur->mxQueue; i++){ - if( pCur->aQueue[i] ){ - pX = fuzzerMergeStems(pX, pCur->aQueue[i]); - pCur->aQueue[i] = 0; - }else{ - pCur->aQueue[i] = pX; - break; - } - } - if( i>pCur->mxQueue ){ - if( i<FUZZER_NQUEUE ){ - pCur->mxQueue = i; - pCur->aQueue[i] = pX; - }else{ - assert( pCur->mxQueue==FUZZER_NQUEUE-1 ); - pX = fuzzerMergeStems(pX, pCur->aQueue[FUZZER_NQUEUE-1]); - pCur->aQueue[FUZZER_NQUEUE-1] = pX; - } - } - - return fuzzerLowestCostStem(pCur); -} - -/* -** Allocate a new fuzzer_stem. Add it to the hash table but do not -** link it into either the pCur->pStem or pCur->pDone lists. -*/ -static fuzzer_stem *fuzzerNewStem( - fuzzer_cursor *pCur, - const char *zWord, - fuzzer_cost rBaseCost -){ - fuzzer_stem *pNew; - fuzzer_rule *pRule; - unsigned int h; - - pNew = sqlite3_malloc( sizeof(*pNew) + (int)strlen(zWord) + 1 ); - if( pNew==0 ) return 0; - memset(pNew, 0, sizeof(*pNew)); - pNew->zBasis = (char*)&pNew[1]; - pNew->nBasis = (int)strlen(zWord); - memcpy(pNew->zBasis, zWord, pNew->nBasis+1); - pRule = pCur->pVtab->pRule; - while( fuzzerSkipRule(pRule, pNew, pCur->iRuleset) ){ - pRule = pRule->pNext; - } - pNew->pRule = pRule; - pNew->n = -1; - pNew->rBaseCost = pNew->rCostX = rBaseCost; - h = fuzzerHash(pNew->zBasis); - pNew->pHash = pCur->apHash[h]; - pCur->apHash[h] = pNew; - pCur->nStem++; - return pNew; -} - - -/* -** Advance a cursor to its next row of output -*/ -static int fuzzerNext(sqlite3_vtab_cursor *cur){ - fuzzer_cursor *pCur = (fuzzer_cursor*)cur; - int rc; - fuzzer_stem *pStem, *pNew; - - pCur->iRowid++; - - /* Use the element the cursor is currently point to to create - ** a new stem and insert the new stem into the priority queue. - */ - pStem = pCur->pStem; - if( pStem->rCostX>0 ){ - rc = fuzzerRender(pStem, &pCur->zBuf, &pCur->nBuf); - if( rc==SQLITE_NOMEM ) return SQLITE_NOMEM; - pNew = fuzzerNewStem(pCur, pCur->zBuf, pStem->rCostX); - if( pNew ){ - if( fuzzerAdvance(pCur, pNew)==0 ){ - pNew->pNext = pCur->pDone; - pCur->pDone = pNew; - }else{ - if( fuzzerInsert(pCur, pNew)==pNew ){ - return SQLITE_OK; - } - } - }else{ - return SQLITE_NOMEM; - } - } - - /* Adjust the priority queue so that the first element of the - ** stem list is the next lowest cost word. - */ - while( (pStem = pCur->pStem)!=0 ){ - int res = fuzzerAdvance(pCur, pStem); - if( res<0 ){ - return SQLITE_NOMEM; - }else if( res>0 ){ - pCur->pStem = 0; - pStem = fuzzerInsert(pCur, pStem); - if( (rc = fuzzerSeen(pCur, pStem))!=0 ){ - if( rc<0 ) return SQLITE_NOMEM; - continue; - } - return SQLITE_OK; /* New word found */ - } - pCur->pStem = 0; - pStem->pNext = pCur->pDone; - pCur->pDone = pStem; - if( fuzzerLowestCostStem(pCur) ){ - rc = fuzzerSeen(pCur, pCur->pStem); - if( rc<0 ) return SQLITE_NOMEM; - if( rc==0 ){ - return SQLITE_OK; - } - } - } - - /* Reach this point only if queue has been exhausted and there is - ** nothing left to be output. */ - pCur->rLimit = (fuzzer_cost)0; - return SQLITE_OK; -} - -/* -** Called to "rewind" a cursor back to the beginning so that -** it starts its output over again. Always called at least once -** prior to any fuzzerColumn, fuzzerRowid, or fuzzerEof call. -*/ -static int fuzzerFilter( - sqlite3_vtab_cursor *pVtabCursor, - int idxNum, const char *idxStr, - int argc, sqlite3_value **argv -){ - fuzzer_cursor *pCur = (fuzzer_cursor *)pVtabCursor; - const char *zWord = ""; - fuzzer_stem *pStem; - int idx; - - fuzzerClearCursor(pCur, 1); - pCur->rLimit = 2147483647; - idx = 0; - if( idxNum & 1 ){ - zWord = (const char*)sqlite3_value_text(argv[0]); - idx++; - } - if( idxNum & 2 ){ - pCur->rLimit = (fuzzer_cost)sqlite3_value_int(argv[idx]); - idx++; - } - if( idxNum & 4 ){ - pCur->iRuleset = (fuzzer_cost)sqlite3_value_int(argv[idx]); - idx++; - } - pCur->nullRule.pNext = pCur->pVtab->pRule; - pCur->nullRule.rCost = 0; - pCur->nullRule.nFrom = 0; - pCur->nullRule.nTo = 0; - pCur->nullRule.zFrom = ""; - pCur->iRowid = 1; - assert( pCur->pStem==0 ); - - /* If the query term is longer than FUZZER_MX_OUTPUT_LENGTH bytes, this - ** query will return zero rows. */ - if( (int)strlen(zWord)<FUZZER_MX_OUTPUT_LENGTH ){ - pCur->pStem = pStem = fuzzerNewStem(pCur, zWord, (fuzzer_cost)0); - if( pStem==0 ) return SQLITE_NOMEM; - pStem->pRule = &pCur->nullRule; - pStem->n = pStem->nBasis; - }else{ - pCur->rLimit = 0; - } - - return SQLITE_OK; -} - -/* -** Only the word and distance columns have values. All other columns -** return NULL -*/ -static int fuzzerColumn(sqlite3_vtab_cursor *cur, sqlite3_context *ctx, int i){ - fuzzer_cursor *pCur = (fuzzer_cursor*)cur; - if( i==0 ){ - /* the "word" column */ - if( fuzzerRender(pCur->pStem, &pCur->zBuf, &pCur->nBuf)==SQLITE_NOMEM ){ - return SQLITE_NOMEM; - } - sqlite3_result_text(ctx, pCur->zBuf, -1, SQLITE_TRANSIENT); - }else if( i==1 ){ - /* the "distance" column */ - sqlite3_result_int(ctx, pCur->pStem->rCostX); - }else{ - /* All other columns are NULL */ - sqlite3_result_null(ctx); - } - return SQLITE_OK; -} - -/* -** The rowid. -*/ -static int fuzzerRowid(sqlite3_vtab_cursor *cur, sqlite_int64 *pRowid){ - fuzzer_cursor *pCur = (fuzzer_cursor*)cur; - *pRowid = pCur->iRowid; - return SQLITE_OK; -} - -/* -** When the fuzzer_cursor.rLimit value is 0 or less, that is a signal -** that the cursor has nothing more to output. -*/ -static int fuzzerEof(sqlite3_vtab_cursor *cur){ - fuzzer_cursor *pCur = (fuzzer_cursor*)cur; - return pCur->rLimit<=(fuzzer_cost)0; -} - -/* -** Search for terms of these forms: -** -** (A) word MATCH $str -** (B1) distance < $value -** (B2) distance <= $value -** (C) ruleid == $ruleid -** -** The distance< and distance<= are both treated as distance<=. -** The query plan number is a bit vector: -** -** bit 1: Term of the form (A) found -** bit 2: Term like (B1) or (B2) found -** bit 3: Term like (C) found -** -** If bit-1 is set, $str is always in filter.argv[0]. If bit-2 is set -** then $value is in filter.argv[0] if bit-1 is clear and is in -** filter.argv[1] if bit-1 is set. If bit-3 is set, then $ruleid is -** in filter.argv[0] if bit-1 and bit-2 are both zero, is in -** filter.argv[1] if exactly one of bit-1 and bit-2 are set, and is in -** filter.argv[2] if both bit-1 and bit-2 are set. -*/ -static int fuzzerBestIndex(sqlite3_vtab *tab, sqlite3_index_info *pIdxInfo){ - int iPlan = 0; - int iDistTerm = -1; - int iRulesetTerm = -1; - int i; - const struct sqlite3_index_constraint *pConstraint; - pConstraint = pIdxInfo->aConstraint; - for(i=0; i<pIdxInfo->nConstraint; i++, pConstraint++){ - if( pConstraint->usable==0 ) continue; - if( (iPlan & 1)==0 - && pConstraint->iColumn==0 - && pConstraint->op==SQLITE_INDEX_CONSTRAINT_MATCH - ){ - iPlan |= 1; - pIdxInfo->aConstraintUsage[i].argvIndex = 1; - pIdxInfo->aConstraintUsage[i].omit = 1; - } - if( (iPlan & 2)==0 - && pConstraint->iColumn==1 - && (pConstraint->op==SQLITE_INDEX_CONSTRAINT_LT - || pConstraint->op==SQLITE_INDEX_CONSTRAINT_LE) - ){ - iPlan |= 2; - iDistTerm = i; - } - if( (iPlan & 4)==0 - && pConstraint->iColumn==2 - && pConstraint->op==SQLITE_INDEX_CONSTRAINT_EQ - ){ - iPlan |= 4; - pIdxInfo->aConstraintUsage[i].omit = 1; - iRulesetTerm = i; - } - } - if( iPlan & 2 ){ - pIdxInfo->aConstraintUsage[iDistTerm].argvIndex = 1+((iPlan&1)!=0); - } - if( iPlan & 4 ){ - int idx = 1; - if( iPlan & 1 ) idx++; - if( iPlan & 2 ) idx++; - pIdxInfo->aConstraintUsage[iRulesetTerm].argvIndex = idx; - } - pIdxInfo->idxNum = iPlan; - if( pIdxInfo->nOrderBy==1 - && pIdxInfo->aOrderBy[0].iColumn==1 - && pIdxInfo->aOrderBy[0].desc==0 - ){ - pIdxInfo->orderByConsumed = 1; - } - pIdxInfo->estimatedCost = (double)10000; - - return SQLITE_OK; -} - -/* -** A virtual table module that implements the "fuzzer". -*/ -static sqlite3_module fuzzerModule = { - 0, /* iVersion */ - fuzzerConnect, - fuzzerConnect, - fuzzerBestIndex, - fuzzerDisconnect, - fuzzerDisconnect, - fuzzerOpen, /* xOpen - open a cursor */ - fuzzerClose, /* xClose - close a cursor */ - fuzzerFilter, /* xFilter - configure scan constraints */ - fuzzerNext, /* xNext - advance a cursor */ - fuzzerEof, /* xEof - check for end of scan */ - fuzzerColumn, /* xColumn - read data */ - fuzzerRowid, /* xRowid - read data */ - 0, /* xUpdate */ - 0, /* xBegin */ - 0, /* xSync */ - 0, /* xCommit */ - 0, /* xRollback */ - 0, /* xFindMethod */ - 0, /* xRename */ -}; - -#endif /* SQLITE_OMIT_VIRTUALTABLE */ - - -/* -** Register the fuzzer virtual table -*/ -int fuzzer_register(sqlite3 *db){ - int rc = SQLITE_OK; -#ifndef SQLITE_OMIT_VIRTUALTABLE - rc = sqlite3_create_module(db, "fuzzer", &fuzzerModule, 0); -#endif - return rc; -} - -#ifdef SQLITE_TEST -#include <tcl.h> -/* -** Decode a pointer to an sqlite3 object. -*/ -extern int getDbPointer(Tcl_Interp *interp, const char *zA, sqlite3 **ppDb); - -/* -** Register the echo virtual table module. -*/ -static int register_fuzzer_module( - ClientData clientData, /* Pointer to sqlite3_enable_XXX function */ - Tcl_Interp *interp, /* The TCL interpreter that invoked this command */ - int objc, /* Number of arguments */ - Tcl_Obj *CONST objv[] /* Command arguments */ -){ - sqlite3 *db; - if( objc!=2 ){ - Tcl_WrongNumArgs(interp, 1, objv, "DB"); - return TCL_ERROR; - } - getDbPointer(interp, Tcl_GetString(objv[1]), &db); - fuzzer_register(db); - return TCL_OK; -} - - -/* -** Register commands with the TCL interpreter. -*/ -int Sqlitetestfuzzer_Init(Tcl_Interp *interp){ - static struct { - char *zName; - Tcl_ObjCmdProc *xProc; - void *clientData; - } aObjCmd[] = { - { "register_fuzzer_module", register_fuzzer_module, 0 }, - }; - int i; - for(i=0; i<sizeof(aObjCmd)/sizeof(aObjCmd[0]); i++){ - Tcl_CreateObjCommand(interp, aObjCmd[i].zName, - aObjCmd[i].xProc, aObjCmd[i].clientData, 0); - } - return TCL_OK; -} - -#endif /* SQLITE_TEST */ diff --git a/src/test_regexp.c b/src/test_regexp.c deleted file mode 100644 index 829d22ad2..000000000 --- a/src/test_regexp.c +++ /dev/null @@ -1,770 +0,0 @@ -/* -** 2012-11-13 -** -** 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. -** -****************************************************************************** -** -** The code in this file implements a compact but reasonably -** efficient regular-expression matcher for posix extended regular -** expressions against UTF8 text. The following syntax is supported: -** -** X* zero or more occurrences of X -** X+ one or more occurrences of X -** X? zero or one occurrences of X -** X{p,q} between p and q occurrences of X -** (X) match X -** X|Y X or Y -** ^X X occurring at the beginning of the string -** X$ X occurring at the end of the string -** . Match any single character -** \c Character c where c is one of \{}()[]|*+?. -** \c C-language escapes for c in afnrtv. ex: \t or \n -** \uXXXX Where XXXX is exactly 4 hex digits, unicode value XXXX -** \xXX Where XX is exactly 2 hex digits, unicode value XX -** [abc] Any single character from the set abc -** [^abc] Any single character not in the set abc -** [a-z] Any single character in the range a-z -** [^a-z] Any single character not in the range a-z -** \b Word boundary -** \w Word character. [A-Za-z0-9_] -** \W Non-word character -** \d Digit -** \D Non-digit -** \s Whitespace character -** \S Non-whitespace character -** -** A nondeterministic finite automaton (NFA) is used for matching, so the -** performance is bounded by O(N*M) where N is the size of the regular -** expression and M is the size of the input string. The matcher never -** exhibits exponential behavior. Note that the X{p,q} operator expands -** to p copies of X following by q-p copies of X? and that the size of the -** regular expression in the O(N*M) performance bound is computed after -** this expansion. -*/ -#include <string.h> -#include <stdlib.h> -#include "sqlite3.h" - -/* The end-of-input character */ -#define RE_EOF 0 /* End of input */ - -/* The NFA is implemented as sequence of opcodes taken from the following -** set. Each opcode has a single integer argument. -*/ -#define RE_OP_MATCH 1 /* Match the one character in the argument */ -#define RE_OP_ANY 2 /* Match any one character. (Implements ".") */ -#define RE_OP_ANYSTAR 3 /* Special optimized version of .* */ -#define RE_OP_FORK 4 /* Continue to both next and opcode at iArg */ -#define RE_OP_GOTO 5 /* Jump to opcode at iArg */ -#define RE_OP_ACCEPT 6 /* Halt and indicate a successful match */ -#define RE_OP_CC_INC 7 /* Beginning of a [...] character class */ -#define RE_OP_CC_EXC 8 /* Beginning of a [^...] character class */ -#define RE_OP_CC_VALUE 9 /* Single value in a character class */ -#define RE_OP_CC_RANGE 10 /* Range of values in a character class */ -#define RE_OP_WORD 11 /* Perl word character [A-Za-z0-9_] */ -#define RE_OP_NOTWORD 12 /* Not a perl word character */ -#define RE_OP_DIGIT 13 /* digit: [0-9] */ -#define RE_OP_NOTDIGIT 14 /* Not a digit */ -#define RE_OP_SPACE 15 /* space: [ \t\n\r\v\f] */ -#define RE_OP_NOTSPACE 16 /* Not a digit */ -#define RE_OP_BOUNDARY 17 /* Boundary between word and non-word */ - -/* Each opcode is a "state" in the NFA */ -typedef unsigned short ReStateNumber; - -/* Because this is an NFA and not a DFA, multiple states can be active at -** once. An instance of the following object records all active states in -** the NFA. The implementation is optimized for the common case where the -** number of actives states is small. -*/ -typedef struct ReStateSet { - unsigned nState; /* Number of current states */ - ReStateNumber *aState; /* Current states */ -} ReStateSet; - -/* An input string read one character at a time. -*/ -typedef struct ReInput ReInput; -struct ReInput { - const unsigned char *z; /* All text */ - int i; /* Next byte to read */ - int mx; /* EOF when i>=mx */ -}; - -/* A compiled NFA (or an NFA that is in the process of being compiled) is -** an instance of the following object. -*/ -typedef struct ReCompiled ReCompiled; -struct ReCompiled { - ReInput sIn; /* Regular expression text */ - const char *zErr; /* Error message to return */ - char *aOp; /* Operators for the virtual machine */ - int *aArg; /* Arguments to each operator */ - unsigned (*xNextChar)(ReInput*); /* Next character function */ - unsigned char zInit[12]; /* Initial text to match */ - int nInit; /* Number of characters in zInit */ - unsigned nState; /* Number of entries in aOp[] and aArg[] */ - unsigned nAlloc; /* Slots allocated for aOp[] and aArg[] */ -}; - -/* Add a state to the given state set if it is not already there */ -static void re_add_state(ReStateSet *pSet, int newState){ - unsigned i; - for(i=0; i<pSet->nState; i++) if( pSet->aState[i]==newState ) return; - pSet->aState[pSet->nState++] = newState; -} - -/* Extract the next unicode character from *pzIn and return it. Advance -** *pzIn to the first byte past the end of the character returned. To -** be clear: this routine converts utf8 to unicode. This routine is -** optimized for the common case where the next character is a single byte. -*/ -static unsigned re_next_char(ReInput *p){ - unsigned c; - if( p->i>=p->mx ) return 0; - c = p->z[p->i++]; - if( c>=0x80 ){ - if( (c&0xe0)==0xc0 && p->i<p->mx && (p->z[p->i]&0xc0)==0x80 ){ - c = (c&0x1f)<<6 | (p->z[p->i++]&0x3f); - if( c<0x80 ) c = 0xfffd; - }else if( (c&0xf0)==0xe0 && p->i+1<p->mx && (p->z[p->i]&0xc0)==0x80 - && (p->z[p->i+1]&0xc0)==0x80 ){ - c = (c&0x0f)<<12 | ((p->z[p->i]&0x3f)<<6) | (p->z[p->i+1]&0x3f); - p->i += 2; - if( c<=0x3ff || (c>=0xd800 && c<=0xdfff) ) c = 0xfffd; - }else if( (c&0xf8)==0xf0 && p->i+3<p->mx && (p->z[p->i]&0xc0)==0x80 - && (p->z[p->i+1]&0xc0)==0x80 && (p->z[p->i+2]&0xc0)==0x80 ){ - c = (c&0x07)<<18 | ((p->z[p->i]&0x3f)<<12) | ((p->z[p->i+1]&0x3f)<<6) - | (p->z[p->i+2]&0x3f); - p->i += 3; - if( c<=0xffff || c>0x10ffff ) c = 0xfffd; - }else{ - c = 0xfffd; - } - } - return c; -} -static unsigned re_next_char_nocase(ReInput *p){ - unsigned c = re_next_char(p); - if( c>='A' && c<='Z' ) c += 'a' - 'A'; - return c; -} - -/* Return true if c is a perl "word" character: [A-Za-z0-9_] */ -static int re_word_char(int c){ - return (c>='0' && c<='9') || (c>='a' && c<='z') - || (c>='A' && c<='Z') || c=='_'; -} - -/* Return true if c is a "digit" character: [0-9] */ -static int re_digit_char(int c){ - return (c>='0' && c<='9'); -} - -/* Return true if c is a perl "space" character: [ \t\r\n\v\f] */ -static int re_space_char(int c){ - return c==' ' || c=='\t' || c=='\n' || c=='\r' || c=='\v' || c=='\f'; -} - -/* Run a compiled regular expression on the zero-terminated input -** string zIn[]. Return true on a match and false if there is no match. -*/ -int re_match(ReCompiled *pRe, const unsigned char *zIn, int nIn){ - ReStateSet aStateSet[2], *pThis, *pNext; - ReStateNumber aSpace[100]; - ReStateNumber *pToFree; - unsigned int i = 0; - unsigned int iSwap = 0; - int c = RE_EOF+1; - int cPrev = 0; - int rc = 0; - ReInput in; - - in.z = zIn; - in.i = 0; - in.mx = nIn>=0 ? nIn : (int)strlen((char const*)zIn); - - /* Look for the initial prefix match, if there is one. */ - if( pRe->nInit ){ - unsigned char x = pRe->zInit[0]; - while( in.i+pRe->nInit<=in.mx - && (zIn[in.i]!=x || - strncmp((const char*)zIn+in.i, (const char*)pRe->zInit, pRe->nInit)!=0) - ){ - in.i++; - } - if( in.i+pRe->nInit>in.mx ) return 0; - } - - if( pRe->nState<=(sizeof(aSpace)/(sizeof(aSpace[0])*2)) ){ - pToFree = 0; - aStateSet[0].aState = aSpace; - }else{ - pToFree = sqlite3_malloc( sizeof(ReStateNumber)*2*pRe->nState ); - if( pToFree==0 ) return -1; - aStateSet[0].aState = pToFree; - } - aStateSet[1].aState = &aStateSet[0].aState[pRe->nState]; - pNext = &aStateSet[1]; - pNext->nState = 0; - re_add_state(pNext, 0); - while( c!=RE_EOF && pNext->nState>0 ){ - cPrev = c; - c = pRe->xNextChar(&in); - pThis = pNext; - pNext = &aStateSet[iSwap]; - iSwap = 1 - iSwap; - pNext->nState = 0; - for(i=0; i<pThis->nState; i++){ - int x = pThis->aState[i]; - switch( pRe->aOp[x] ){ - case RE_OP_MATCH: { - if( pRe->aArg[x]==c ) re_add_state(pNext, x+1); - break; - } - case RE_OP_ANY: { - re_add_state(pNext, x+1); - break; - } - case RE_OP_WORD: { - if( re_word_char(c) ) re_add_state(pNext, x+1); - break; - } - case RE_OP_NOTWORD: { - if( !re_word_char(c) ) re_add_state(pNext, x+1); - break; - } - case RE_OP_DIGIT: { - if( re_digit_char(c) ) re_add_state(pNext, x+1); - break; - } - case RE_OP_NOTDIGIT: { - if( !re_digit_char(c) ) re_add_state(pNext, x+1); - break; - } - case RE_OP_SPACE: { - if( re_space_char(c) ) re_add_state(pNext, x+1); - break; - } - case RE_OP_NOTSPACE: { - if( !re_space_char(c) ) re_add_state(pNext, x+1); - break; - } - case RE_OP_BOUNDARY: { - if( re_word_char(c)!=re_word_char(cPrev) ) re_add_state(pThis, x+1); - break; - } - case RE_OP_ANYSTAR: { - re_add_state(pNext, x); - re_add_state(pThis, x+1); - break; - } - case RE_OP_FORK: { - re_add_state(pThis, x+pRe->aArg[x]); - re_add_state(pThis, x+1); - break; - } - case RE_OP_GOTO: { - re_add_state(pThis, x+pRe->aArg[x]); - break; - } - case RE_OP_ACCEPT: { - rc = 1; - goto re_match_end; - } - case RE_OP_CC_INC: - case RE_OP_CC_EXC: { - int j = 1; - int n = pRe->aArg[x]; - int hit = 0; - for(j=1; j>0 && j<n; j++){ - if( pRe->aOp[x+j]==RE_OP_CC_VALUE ){ - if( pRe->aArg[x+j]==c ){ - hit = 1; - j = -1; - } - }else{ - if( pRe->aArg[x+j]<=c && pRe->aArg[x+j+1]>=c ){ - hit = 1; - j = -1; - }else{ - j++; - } - } - } - if( pRe->aOp[x]==RE_OP_CC_EXC ) hit = !hit; - if( hit ) re_add_state(pNext, x+n); - break; - } - } - } - } - for(i=0; i<pNext->nState; i++){ - if( pRe->aOp[pNext->aState[i]]==RE_OP_ACCEPT ){ rc = 1; break; } - } -re_match_end: - sqlite3_free(pToFree); - return rc; -} - -/* Resize the opcode and argument arrays for an RE under construction. -*/ -static int re_resize(ReCompiled *p, int N){ - char *aOp; - int *aArg; - aOp = sqlite3_realloc(p->aOp, N*sizeof(p->aOp[0])); - if( aOp==0 ) return 1; - p->aOp = aOp; - aArg = sqlite3_realloc(p->aArg, N*sizeof(p->aArg[0])); - if( aArg==0 ) return 1; - p->aArg = aArg; - p->nAlloc = N; - return 0; -} - -/* Insert a new opcode and argument into an RE under construction. The -** insertion point is just prior to existing opcode iBefore. -*/ -static int re_insert(ReCompiled *p, int iBefore, int op, int arg){ - int i; - if( p->nAlloc<=p->nState && re_resize(p, p->nAlloc*2) ) return 0; - for(i=p->nState; i>iBefore; i--){ - p->aOp[i] = p->aOp[i-1]; - p->aArg[i] = p->aArg[i-1]; - } - p->nState++; - p->aOp[iBefore] = op; - p->aArg[iBefore] = arg; - return iBefore; -} - -/* Append a new opcode and argument to the end of the RE under construction. -*/ -static int re_append(ReCompiled *p, int op, int arg){ - return re_insert(p, p->nState, op, arg); -} - -/* Make a copy of N opcodes starting at iStart onto the end of the RE -** under construction. -*/ -static void re_copy(ReCompiled *p, int iStart, int N){ - if( p->nState+N>=p->nAlloc && re_resize(p, p->nAlloc*2+N) ) return; - memcpy(&p->aOp[p->nState], &p->aOp[iStart], N*sizeof(p->aOp[0])); - memcpy(&p->aArg[p->nState], &p->aArg[iStart], N*sizeof(p->aArg[0])); - p->nState += N; -} - -/* Return true if c is a hexadecimal digit character: [0-9a-fA-F] -** If c is a hex digit, also set *pV = (*pV)*16 + valueof(c). If -** c is not a hex digit *pV is unchanged. -*/ -static int re_hex(int c, int *pV){ - if( c>='0' && c<='9' ){ - c -= '0'; - }else if( c>='a' && c<='f' ){ - c -= 'a' - 10; - }else if( c>='A' && c<='F' ){ - c -= 'A' - 10; - }else{ - return 0; - } - *pV = (*pV)*16 + (c & 0xff); - return 1; -} - -/* A backslash character has been seen, read the next character and -** return its interpretation. -*/ -static unsigned re_esc_char(ReCompiled *p){ - static const char zEsc[] = "afnrtv\\()*.+?[$^{|}]"; - static const char zTrans[] = "\a\f\n\r\t\v"; - int i, v = 0; - char c; - if( p->sIn.i>=p->sIn.mx ) return 0; - c = p->sIn.z[p->sIn.i]; - if( c=='u' && p->sIn.i+4<p->sIn.mx ){ - const unsigned char *zIn = p->sIn.z + p->sIn.i; - if( re_hex(zIn[1],&v) - && re_hex(zIn[2],&v) - && re_hex(zIn[3],&v) - && re_hex(zIn[4],&v) - ){ - p->sIn.i += 5; - return v; - } - } - if( c=='x' && p->sIn.i+2<p->sIn.mx ){ - const unsigned char *zIn = p->sIn.z + p->sIn.i; - if( re_hex(zIn[1],&v) - && re_hex(zIn[2],&v) - ){ - p->sIn.i += 3; - return v; - } - } - for(i=0; zEsc[i] && zEsc[i]!=c; i++){} - if( zEsc[i] ){ - if( i<6 ) c = zTrans[i]; - p->sIn.i++; - }else{ - p->zErr = "unknown \\ escape"; - } - return c; -} - -/* Forward declaration */ -static const char *re_subcompile_string(ReCompiled*); - -/* Peek at the next byte of input */ -static unsigned char rePeek(ReCompiled *p){ - return p->sIn.i<p->sIn.mx ? p->sIn.z[p->sIn.i] : 0; -} - -/* Compile RE text into a sequence of opcodes. Continue up to the -** first unmatched ")" character, then return. If an error is found, -** return a pointer to the error message string. -*/ -static const char *re_subcompile_re(ReCompiled *p){ - const char *zErr; - int iStart, iEnd, iGoto; - iStart = p->nState; - zErr = re_subcompile_string(p); - if( zErr ) return zErr; - while( rePeek(p)=='|' ){ - iEnd = p->nState; - re_insert(p, iStart, RE_OP_FORK, iEnd + 2 - iStart); - iGoto = re_append(p, RE_OP_GOTO, 0); - p->sIn.i++; - zErr = re_subcompile_string(p); - if( zErr ) return zErr; - p->aArg[iGoto] = p->nState - iGoto; - } - return 0; -} - -/* Compile an element of regular expression text (anything that can be -** an operand to the "|" operator). Return NULL on success or a pointer -** to the error message if there is a problem. -*/ -static const char *re_subcompile_string(ReCompiled *p){ - int iPrev = -1; - int iStart; - unsigned c; - const char *zErr; - while( (c = p->xNextChar(&p->sIn))!=0 ){ - iStart = p->nState; - switch( c ){ - case '|': - case '$': - case ')': { - p->sIn.i--; - return 0; - } - case '(': { - zErr = re_subcompile_re(p); - if( zErr ) return zErr; - if( rePeek(p)!=')' ) return "unmatched '('"; - p->sIn.i++; - break; - } - case '.': { - if( rePeek(p)=='*' ){ - re_append(p, RE_OP_ANYSTAR, 0); - p->sIn.i++; - }else{ - re_append(p, RE_OP_ANY, 0); - } - break; - } - case '*': { - if( iPrev<0 ) return "'*' without operand"; - re_insert(p, iPrev, RE_OP_GOTO, p->nState - iPrev + 1); - re_append(p, RE_OP_FORK, iPrev - p->nState + 1); - break; - } - case '+': { - if( iPrev<0 ) return "'+' without operand"; - re_append(p, RE_OP_FORK, iPrev - p->nState); - break; - } - case '?': { - if( iPrev<0 ) return "'?' without operand"; - re_insert(p, iPrev, RE_OP_FORK, p->nState - iPrev+1); - break; - } - case '{': { - int m = 0, n = 0; - int sz, j; - if( iPrev<0 ) return "'{m,n}' without operand"; - while( (c=rePeek(p))>='0' && c<='9' ){ m = m*10 + c - '0'; p->sIn.i++; } - n = m; - if( c==',' ){ - p->sIn.i++; - n = 0; - while( (c=rePeek(p))>='0' && c<='9' ){ n = n*10 + c-'0'; p->sIn.i++; } - } - if( c!='}' ) return "unmatched '{'"; - if( n>0 && n<m ) return "n less than m in '{m,n}'"; - p->sIn.i++; - sz = p->nState - iPrev; - if( m==0 ){ - if( n==0 ) return "both m and n are zero in '{m,n}'"; - re_insert(p, iPrev, RE_OP_FORK, sz+1); - n--; - }else{ - for(j=1; j<m; j++) re_copy(p, iPrev, sz); - } - for(j=m; j<n; j++){ - re_append(p, RE_OP_FORK, sz+1); - re_copy(p, iPrev, sz); - } - if( n==0 && m>0 ){ - re_append(p, RE_OP_FORK, -sz); - } - break; - } - case '[': { - int iFirst = p->nState; - if( rePeek(p)=='^' ){ - re_append(p, RE_OP_CC_EXC, 0); - p->sIn.i++; - }else{ - re_append(p, RE_OP_CC_INC, 0); - } - while( (c = p->xNextChar(&p->sIn))!=0 ){ - if( c=='[' && rePeek(p)==':' ){ - return "POSIX character classes not supported"; - } - if( c=='\\' ) c = re_esc_char(p); - if( rePeek(p)=='-' ){ - re_append(p, RE_OP_CC_RANGE, c); - p->sIn.i++; - c = p->xNextChar(&p->sIn); - if( c=='\\' ) c = re_esc_char(p); - re_append(p, RE_OP_CC_RANGE, c); - }else{ - re_append(p, RE_OP_CC_VALUE, c); - } - if( rePeek(p)==']' ){ p->sIn.i++; break; } - } - if( c==0 ) return "unclosed '['"; - p->aArg[iFirst] = p->nState - iFirst; - break; - } - case '\\': { - int specialOp = 0; - switch( rePeek(p) ){ - case 'b': specialOp = RE_OP_BOUNDARY; break; - case 'd': specialOp = RE_OP_DIGIT; break; - case 'D': specialOp = RE_OP_NOTDIGIT; break; - case 's': specialOp = RE_OP_SPACE; break; - case 'S': specialOp = RE_OP_NOTSPACE; break; - case 'w': specialOp = RE_OP_WORD; break; - case 'W': specialOp = RE_OP_NOTWORD; break; - } - if( specialOp ){ - p->sIn.i++; - re_append(p, specialOp, 0); - }else{ - c = re_esc_char(p); - re_append(p, RE_OP_MATCH, c); - } - break; - } - default: { - re_append(p, RE_OP_MATCH, c); - break; - } - } - iPrev = iStart; - } - return 0; -} - -/* Free and reclaim all the memory used by a previously compiled -** regular expression. Applications should invoke this routine once -** for every call to re_compile() to avoid memory leaks. -*/ -void re_free(ReCompiled *pRe){ - if( pRe ){ - sqlite3_free(pRe->aOp); - sqlite3_free(pRe->aArg); - sqlite3_free(pRe); - } -} - -/* -** Compile a textual regular expression in zIn[] into a compiled regular -** expression suitable for us by re_match() and return a pointer to the -** compiled regular expression in *ppRe. Return NULL on success or an -** error message if something goes wrong. -*/ -const char *re_compile(ReCompiled **ppRe, const char *zIn, int noCase){ - ReCompiled *pRe; - const char *zErr; - int i, j; - - *ppRe = 0; - pRe = sqlite3_malloc( sizeof(*pRe) ); - if( pRe==0 ){ - return "out of memory"; - } - memset(pRe, 0, sizeof(*pRe)); - pRe->xNextChar = noCase ? re_next_char_nocase : re_next_char; - if( re_resize(pRe, 30) ){ - re_free(pRe); - return "out of memory"; - } - if( zIn[0]=='^' ){ - zIn++; - }else{ - re_append(pRe, RE_OP_ANYSTAR, 0); - } - pRe->sIn.z = (unsigned char*)zIn; - pRe->sIn.i = 0; - pRe->sIn.mx = (int)strlen(zIn); - zErr = re_subcompile_re(pRe); - if( zErr ){ - re_free(pRe); - return zErr; - } - if( rePeek(pRe)=='$' && pRe->sIn.i+1>=pRe->sIn.mx ){ - re_append(pRe, RE_OP_MATCH, RE_EOF); - re_append(pRe, RE_OP_ACCEPT, 0); - *ppRe = pRe; - }else if( pRe->sIn.i>=pRe->sIn.mx ){ - re_append(pRe, RE_OP_ACCEPT, 0); - *ppRe = pRe; - }else{ - re_free(pRe); - return "unrecognized character"; - } - - /* The following is a performance optimization. If the regex begins with - ** ".*" (if the input regex lacks an initial "^") and afterwards there are - ** one or more matching characters, enter those matching characters into - ** zInit[]. The re_match() routine can then search ahead in the input - ** string looking for the initial match without having to run the whole - ** regex engine over the string. Do not worry able trying to match - ** unicode characters beyond plane 0 - those are very rare and this is - ** just an optimization. */ - if( pRe->aOp[0]==RE_OP_ANYSTAR ){ - for(j=0, i=1; j<sizeof(pRe->zInit)-2 && pRe->aOp[i]==RE_OP_MATCH; i++){ - unsigned x = pRe->aArg[i]; - if( x<=127 ){ - pRe->zInit[j++] = x; - }else if( x<=0xfff ){ - pRe->zInit[j++] = 0xc0 | (x>>6); - pRe->zInit[j++] = 0x80 | (x&0x3f); - }else if( x<=0xffff ){ - pRe->zInit[j++] = 0xd0 | (x>>12); - pRe->zInit[j++] = 0x80 | ((x>>6)&0x3f); - pRe->zInit[j++] = 0x80 | (x&0x3f); - }else{ - break; - } - } - if( j>0 && pRe->zInit[j-1]==0 ) j--; - pRe->nInit = j; - } - return pRe->zErr; -} - -/* -** Implementation of the regexp() SQL function. This function implements -** the build-in REGEXP operator. The first argument to the function is the -** pattern and the second argument is the string. So, the SQL statements: -** -** A REGEXP B -** -** is implemented as regexp(B,A). -*/ -static void re_sql_func( - sqlite3_context *context, - int argc, - sqlite3_value **argv -){ - ReCompiled *pRe; /* Compiled regular expression */ - const char *zPattern; /* The regular expression */ - const unsigned char *zStr;/* String being searched */ - const char *zErr; /* Compile error message */ - - pRe = sqlite3_get_auxdata(context, 0); - if( pRe==0 ){ - zPattern = (const char*)sqlite3_value_text(argv[0]); - if( zPattern==0 ) return; - zErr = re_compile(&pRe, zPattern, 0); - if( zErr ){ - re_free(pRe); - sqlite3_result_error(context, zErr, -1); - return; - } - if( pRe==0 ){ - sqlite3_result_error_nomem(context); - return; - } - sqlite3_set_auxdata(context, 0, pRe, (void(*)(void*))re_free); - } - zStr = (const unsigned char*)sqlite3_value_text(argv[1]); - if( zStr!=0 ){ - sqlite3_result_int(context, re_match(pRe, zStr, -1)); - } -} - -/* -** Invoke this routine in order to install the REGEXP function in an -** SQLite database connection. -** -** Use: -** -** sqlite3_auto_extension(sqlite3_add_regexp_func); -** -** to cause this extension to be automatically loaded into each new -** database connection. -*/ -int sqlite3_add_regexp_func(sqlite3 *db){ - return sqlite3_create_function(db, "regexp", 2, SQLITE_UTF8, 0, - re_sql_func, 0, 0); -} - - -/***************************** Test Code ***********************************/ -#ifdef SQLITE_TEST -#include <tcl.h> -extern int getDbPointer(Tcl_Interp *interp, const char *zA, sqlite3 **ppDb); - -/* Implementation of the TCL command: -** -** sqlite3_add_regexp_func $DB -*/ -static int tclSqlite3AddRegexpFunc( - void * clientData, - Tcl_Interp *interp, - int objc, - Tcl_Obj *CONST objv[] -){ - sqlite3 *db; - if( objc!=2 ){ - Tcl_WrongNumArgs(interp, 1, objv, "DB"); - return TCL_ERROR; - } - if( getDbPointer(interp, Tcl_GetString(objv[1]), &db) ) return TCL_ERROR; - sqlite3_add_regexp_func(db); - return TCL_OK; -} - -/* Register the sqlite3_add_regexp_func TCL command with the TCL interpreter. -*/ -int Sqlitetestregexp_Init(Tcl_Interp *interp){ - Tcl_CreateObjCommand(interp, "sqlite3_add_regexp_func", - tclSqlite3AddRegexpFunc, 0, 0); - return TCL_OK; -} -#endif /* SQLITE_TEST */ -/**************************** End Of Test Code *******************************/ diff --git a/src/test_spellfix.c b/src/test_spellfix.c deleted file mode 100644 index 16376244a..000000000 --- a/src/test_spellfix.c +++ /dev/null @@ -1,2847 +0,0 @@ -/* -** 2012 April 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 module implements the spellfix1 VIRTUAL TABLE that can be used -** to search a large vocabulary for close matches. See separate -** documentation files (spellfix1.wiki and editdist3.wiki) for details. -*/ -#if SQLITE_CORE -# include "sqliteInt.h" -#else -# include <string.h> -# include <stdio.h> -# include <stdlib.h> -# include "sqlite3ext.h" -# include <assert.h> -# define ALWAYS(X) 1 -# define NEVER(X) 0 - typedef unsigned char u8; - typedef unsigned short u16; - SQLITE_EXTENSION_INIT1 -#endif /* !SQLITE_CORE */ -#include <ctype.h> - -/* -** Character classes for ASCII characters: -** -** 0 '' Silent letters: H W -** 1 'A' Any vowel: A E I O U (Y) -** 2 'B' A bilabeal stop or fricative: B F P V W -** 3 'C' Other fricatives or back stops: C G J K Q S X Z -** 4 'D' Alveolar stops: D T -** 5 'H' Letter H at the beginning of a word -** 6 'L' Glide: L -** 7 'R' Semivowel: R -** 8 'M' Nasals: M N -** 9 'Y' Letter Y at the beginning of a word. -** 10 '9' Digits: 0 1 2 3 4 5 6 7 8 9 -** 11 ' ' White space -** 12 '?' Other. -*/ -#define CCLASS_SILENT 0 -#define CCLASS_VOWEL 1 -#define CCLASS_B 2 -#define CCLASS_C 3 -#define CCLASS_D 4 -#define CCLASS_H 5 -#define CCLASS_L 6 -#define CCLASS_R 7 -#define CCLASS_M 8 -#define CCLASS_Y 9 -#define CCLASS_DIGIT 10 -#define CCLASS_SPACE 11 -#define CCLASS_OTHER 12 - -/* -** The following table gives the character class for non-initial ASCII -** characters. -*/ -static const unsigned char midClass[] = { - /* */ CCLASS_OTHER, /* */ CCLASS_OTHER, /* */ CCLASS_OTHER, - /* */ CCLASS_OTHER, /* */ CCLASS_OTHER, /* */ CCLASS_OTHER, - /* */ CCLASS_OTHER, /* */ CCLASS_OTHER, /* */ CCLASS_OTHER, - /* */ CCLASS_SPACE, /* */ CCLASS_OTHER, /* */ CCLASS_OTHER, - /* */ CCLASS_SPACE, /* */ CCLASS_SPACE, /* */ CCLASS_OTHER, - /* */ CCLASS_OTHER, /* */ CCLASS_OTHER, /* */ CCLASS_OTHER, - /* */ CCLASS_OTHER, /* */ CCLASS_OTHER, /* */ CCLASS_OTHER, - /* */ CCLASS_OTHER, /* */ CCLASS_OTHER, /* */ CCLASS_OTHER, - /* */ CCLASS_OTHER, /* */ CCLASS_OTHER, /* */ CCLASS_OTHER, - /* */ CCLASS_OTHER, /* */ CCLASS_OTHER, /* */ CCLASS_OTHER, - /* */ CCLASS_OTHER, /* */ CCLASS_OTHER, /* */ CCLASS_SPACE, - /* ! */ CCLASS_OTHER, /* " */ CCLASS_OTHER, /* # */ CCLASS_OTHER, - /* $ */ CCLASS_OTHER, /* % */ CCLASS_OTHER, /* & */ CCLASS_OTHER, - /* ' */ CCLASS_SILENT, /* ( */ CCLASS_OTHER, /* ) */ CCLASS_OTHER, - /* * */ CCLASS_OTHER, /* + */ CCLASS_OTHER, /* , */ CCLASS_OTHER, - /* - */ CCLASS_OTHER, /* . */ CCLASS_OTHER, /* / */ CCLASS_OTHER, - /* 0 */ CCLASS_DIGIT, /* 1 */ CCLASS_DIGIT, /* 2 */ CCLASS_DIGIT, - /* 3 */ CCLASS_DIGIT, /* 4 */ CCLASS_DIGIT, /* 5 */ CCLASS_DIGIT, - /* 6 */ CCLASS_DIGIT, /* 7 */ CCLASS_DIGIT, /* 8 */ CCLASS_DIGIT, - /* 9 */ CCLASS_DIGIT, /* : */ CCLASS_OTHER, /* ; */ CCLASS_OTHER, - /* < */ CCLASS_OTHER, /* = */ CCLASS_OTHER, /* > */ CCLASS_OTHER, - /* ? */ CCLASS_OTHER, /* @ */ CCLASS_OTHER, /* A */ CCLASS_VOWEL, - /* B */ CCLASS_B, /* C */ CCLASS_C, /* D */ CCLASS_D, - /* E */ CCLASS_VOWEL, /* F */ CCLASS_B, /* G */ CCLASS_C, - /* H */ CCLASS_SILENT, /* I */ CCLASS_VOWEL, /* J */ CCLASS_C, - /* K */ CCLASS_C, /* L */ CCLASS_L, /* M */ CCLASS_M, - /* N */ CCLASS_M, /* O */ CCLASS_VOWEL, /* P */ CCLASS_B, - /* Q */ CCLASS_C, /* R */ CCLASS_R, /* S */ CCLASS_C, - /* T */ CCLASS_D, /* U */ CCLASS_VOWEL, /* V */ CCLASS_B, - /* W */ CCLASS_B, /* X */ CCLASS_C, /* Y */ CCLASS_VOWEL, - /* Z */ CCLASS_C, /* [ */ CCLASS_OTHER, /* \ */ CCLASS_OTHER, - /* ] */ CCLASS_OTHER, /* ^ */ CCLASS_OTHER, /* _ */ CCLASS_OTHER, - /* ` */ CCLASS_OTHER, /* a */ CCLASS_VOWEL, /* b */ CCLASS_B, - /* c */ CCLASS_C, /* d */ CCLASS_D, /* e */ CCLASS_VOWEL, - /* f */ CCLASS_B, /* g */ CCLASS_C, /* h */ CCLASS_SILENT, - /* i */ CCLASS_VOWEL, /* j */ CCLASS_C, /* k */ CCLASS_C, - /* l */ CCLASS_L, /* m */ CCLASS_M, /* n */ CCLASS_M, - /* o */ CCLASS_VOWEL, /* p */ CCLASS_B, /* q */ CCLASS_C, - /* r */ CCLASS_R, /* s */ CCLASS_C, /* t */ CCLASS_D, - /* u */ CCLASS_VOWEL, /* v */ CCLASS_B, /* w */ CCLASS_B, - /* x */ CCLASS_C, /* y */ CCLASS_VOWEL, /* z */ CCLASS_C, - /* { */ CCLASS_OTHER, /* | */ CCLASS_OTHER, /* } */ CCLASS_OTHER, - /* ~ */ CCLASS_OTHER, /* */ CCLASS_OTHER, -}; -/* -** This tables gives the character class for ASCII characters that form the -** initial character of a word. The only difference from midClass is with -** the letters H, W, and Y. -*/ -static const unsigned char initClass[] = { - /* */ CCLASS_OTHER, /* */ CCLASS_OTHER, /* */ CCLASS_OTHER, - /* */ CCLASS_OTHER, /* */ CCLASS_OTHER, /* */ CCLASS_OTHER, - /* */ CCLASS_OTHER, /* */ CCLASS_OTHER, /* */ CCLASS_OTHER, - /* */ CCLASS_SPACE, /* */ CCLASS_OTHER, /* */ CCLASS_OTHER, - /* */ CCLASS_SPACE, /* */ CCLASS_SPACE, /* */ CCLASS_OTHER, - /* */ CCLASS_OTHER, /* */ CCLASS_OTHER, /* */ CCLASS_OTHER, - /* */ CCLASS_OTHER, /* */ CCLASS_OTHER, /* */ CCLASS_OTHER, - /* */ CCLASS_OTHER, /* */ CCLASS_OTHER, /* */ CCLASS_OTHER, - /* */ CCLASS_OTHER, /* */ CCLASS_OTHER, /* */ CCLASS_OTHER, - /* */ CCLASS_OTHER, /* */ CCLASS_OTHER, /* */ CCLASS_OTHER, - /* */ CCLASS_OTHER, /* */ CCLASS_OTHER, /* */ CCLASS_SPACE, - /* ! */ CCLASS_OTHER, /* " */ CCLASS_OTHER, /* # */ CCLASS_OTHER, - /* $ */ CCLASS_OTHER, /* % */ CCLASS_OTHER, /* & */ CCLASS_OTHER, - /* ' */ CCLASS_OTHER, /* ( */ CCLASS_OTHER, /* ) */ CCLASS_OTHER, - /* * */ CCLASS_OTHER, /* + */ CCLASS_OTHER, /* , */ CCLASS_OTHER, - /* - */ CCLASS_OTHER, /* . */ CCLASS_OTHER, /* / */ CCLASS_OTHER, - /* 0 */ CCLASS_DIGIT, /* 1 */ CCLASS_DIGIT, /* 2 */ CCLASS_DIGIT, - /* 3 */ CCLASS_DIGIT, /* 4 */ CCLASS_DIGIT, /* 5 */ CCLASS_DIGIT, - /* 6 */ CCLASS_DIGIT, /* 7 */ CCLASS_DIGIT, /* 8 */ CCLASS_DIGIT, - /* 9 */ CCLASS_DIGIT, /* : */ CCLASS_OTHER, /* ; */ CCLASS_OTHER, - /* < */ CCLASS_OTHER, /* = */ CCLASS_OTHER, /* > */ CCLASS_OTHER, - /* ? */ CCLASS_OTHER, /* @ */ CCLASS_OTHER, /* A */ CCLASS_VOWEL, - /* B */ CCLASS_B, /* C */ CCLASS_C, /* D */ CCLASS_D, - /* E */ CCLASS_VOWEL, /* F */ CCLASS_B, /* G */ CCLASS_C, - /* H */ CCLASS_SILENT, /* I */ CCLASS_VOWEL, /* J */ CCLASS_C, - /* K */ CCLASS_C, /* L */ CCLASS_L, /* M */ CCLASS_M, - /* N */ CCLASS_M, /* O */ CCLASS_VOWEL, /* P */ CCLASS_B, - /* Q */ CCLASS_C, /* R */ CCLASS_R, /* S */ CCLASS_C, - /* T */ CCLASS_D, /* U */ CCLASS_VOWEL, /* V */ CCLASS_B, - /* W */ CCLASS_B, /* X */ CCLASS_C, /* Y */ CCLASS_Y, - /* Z */ CCLASS_C, /* [ */ CCLASS_OTHER, /* \ */ CCLASS_OTHER, - /* ] */ CCLASS_OTHER, /* ^ */ CCLASS_OTHER, /* _ */ CCLASS_OTHER, - /* ` */ CCLASS_OTHER, /* a */ CCLASS_VOWEL, /* b */ CCLASS_B, - /* c */ CCLASS_C, /* d */ CCLASS_D, /* e */ CCLASS_VOWEL, - /* f */ CCLASS_B, /* g */ CCLASS_C, /* h */ CCLASS_SILENT, - /* i */ CCLASS_VOWEL, /* j */ CCLASS_C, /* k */ CCLASS_C, - /* l */ CCLASS_L, /* m */ CCLASS_M, /* n */ CCLASS_M, - /* o */ CCLASS_VOWEL, /* p */ CCLASS_B, /* q */ CCLASS_C, - /* r */ CCLASS_R, /* s */ CCLASS_C, /* t */ CCLASS_D, - /* u */ CCLASS_VOWEL, /* v */ CCLASS_B, /* w */ CCLASS_B, - /* x */ CCLASS_C, /* y */ CCLASS_Y, /* z */ CCLASS_C, - /* { */ CCLASS_OTHER, /* | */ CCLASS_OTHER, /* } */ CCLASS_OTHER, - /* ~ */ CCLASS_OTHER, /* */ CCLASS_OTHER, -}; - -/* -** Mapping from the character class number (0-13) to a symbol for each -** character class. Note that initClass[] can be used to map the class -** symbol back into the class number. -*/ -static const unsigned char className[] = ".ABCDHLRMY9 ?"; - -/* -** Generate a "phonetic hash" from a string of ASCII characters -** in zIn[0..nIn-1]. -** -** * Map characters by character class as defined above. -** * Omit double-letters -** * Omit vowels beside R and L -** * Omit T when followed by CH -** * Omit W when followed by R -** * Omit D when followed by J or G -** * Omit K in KN or G in GN at the beginning of a word -** -** Space to hold the result is obtained from sqlite3_malloc() -** -** Return NULL if memory allocation fails. -*/ -static unsigned char *phoneticHash(const unsigned char *zIn, int nIn){ - unsigned char *zOut = sqlite3_malloc( nIn + 1 ); - int i; - int nOut = 0; - char cPrev = 0x77; - char cPrevX = 0x77; - const unsigned char *aClass = initClass; - - if( zOut==0 ) return 0; - if( nIn>2 ){ - switch( zIn[0] ){ - case 'g': - case 'k': { - if( zIn[1]=='n' ){ zIn++; nIn--; } - break; - } - } - } - for(i=0; i<nIn; i++){ - unsigned char c = zIn[i]; - if( i+1<nIn ){ - if( c=='w' && zIn[i+1]=='r' ) continue; - if( c=='d' && (zIn[i+1]=='j' || zIn[i+1]=='g') ) continue; - if( i+2<nIn ){ - if( c=='t' && zIn[i+1]=='c' && zIn[i+2]=='h' ) continue; - } - } - c = aClass[c&0x7f]; - if( c==CCLASS_SPACE ) continue; - if( c==CCLASS_OTHER && cPrev!=CCLASS_DIGIT ) continue; - aClass = midClass; - if( c==CCLASS_VOWEL && (cPrevX==CCLASS_R || cPrevX==CCLASS_L) ){ - continue; /* No vowels beside L or R */ - } - if( (c==CCLASS_R || c==CCLASS_L) && cPrevX==CCLASS_VOWEL ){ - nOut--; /* No vowels beside L or R */ - } - cPrev = c; - if( c==CCLASS_SILENT ) continue; - cPrevX = c; - c = className[c]; - assert( nOut>=0 ); - if( nOut==0 || c!=zOut[nOut-1] ) zOut[nOut++] = c; - } - zOut[nOut] = 0; - return zOut; -} - -/* -** This is an SQL function wrapper around phoneticHash(). See -** the description of phoneticHash() for additional information. -*/ -static void phoneticHashSqlFunc( - sqlite3_context *context, - int argc, - sqlite3_value **argv -){ - const unsigned char *zIn; - unsigned char *zOut; - - zIn = sqlite3_value_text(argv[0]); - if( zIn==0 ) return; - zOut = phoneticHash(zIn, sqlite3_value_bytes(argv[0])); - if( zOut==0 ){ - sqlite3_result_error_nomem(context); - }else{ - sqlite3_result_text(context, (char*)zOut, -1, sqlite3_free); - } -} - -/* -** Return the character class number for a character given its -** context. -*/ -static char characterClass(char cPrev, char c){ - return cPrev==0 ? initClass[c&0x7f] : midClass[c&0x7f]; -} - -/* -** Return the cost of inserting or deleting character c immediately -** following character cPrev. If cPrev==0, that means c is the first -** character of the word. -*/ -static int insertOrDeleteCost(char cPrev, char c, char cNext){ - char classC = characterClass(cPrev, c); - char classCprev; - - if( classC==CCLASS_SILENT ){ - /* Insert or delete "silent" characters such as H or W */ - return 1; - } - if( cPrev==c ){ - /* Repeated characters, or miss a repeat */ - return 10; - } - if( classC==CCLASS_VOWEL && (cPrev=='r' || cNext=='r') ){ - return 20; /* Insert a vowel before or after 'r' */ - } - classCprev = characterClass(cPrev, cPrev); - if( classC==classCprev ){ - if( classC==CCLASS_VOWEL ){ - /* Remove or add a new vowel to a vowel cluster */ - return 15; - }else{ - /* Remove or add a consonant not in the same class */ - return 50; - } - } - - /* any other character insertion or deletion */ - return 100; -} - -/* -** Divide the insertion cost by this factor when appending to the -** end of the word. -*/ -#define FINAL_INS_COST_DIV 4 - -/* -** Return the cost of substituting cTo in place of cFrom assuming -** the previous character is cPrev. If cPrev==0 then cTo is the first -** character of the word. -*/ -static int substituteCost(char cPrev, char cFrom, char cTo){ - char classFrom, classTo; - if( cFrom==cTo ){ - /* Exact match */ - return 0; - } - if( cFrom==(cTo^0x20) && ((cTo>='A' && cTo<='Z') || (cTo>='a' && cTo<='z')) ){ - /* differ only in case */ - return 0; - } - classFrom = characterClass(cPrev, cFrom); - classTo = characterClass(cPrev, cTo); - if( classFrom==classTo ){ - /* Same character class */ - return 40; - } - if( classFrom>=CCLASS_B && classFrom<=CCLASS_Y - && classTo>=CCLASS_B && classTo<=CCLASS_Y ){ - /* Convert from one consonant to another, but in a different class */ - return 75; - } - /* Any other subsitution */ - return 100; -} - -/* -** Given two strings zA and zB which are pure ASCII, return the cost -** of transforming zA into zB. If zA ends with '*' assume that it is -** a prefix of zB and give only minimal penalty for extra characters -** on the end of zB. -** -** Smaller numbers mean a closer match. -** -** Negative values indicate an error: -** -1 One of the inputs is NULL -** -2 Non-ASCII characters on input -** -3 Unable to allocate memory -** -** If pnMatch is not NULL, then *pnMatch is set to the number of bytes -** of zB that matched the pattern in zA. If zA does not end with a '*', -** then this value is always the number of bytes in zB (i.e. strlen(zB)). -** If zA does end in a '*', then it is the number of bytes in the prefix -** of zB that was deemed to match zA. -*/ -static int editdist1(const char *zA, const char *zB, int *pnMatch){ - int nA, nB; /* Number of characters in zA[] and zB[] */ - int xA, xB; /* Loop counters for zA[] and zB[] */ - char cA, cB; /* Current character of zA and zB */ - char cAprev, cBprev; /* Previous character of zA and zB */ - char cAnext, cBnext; /* Next character in zA and zB */ - int d; /* North-west cost value */ - int dc = 0; /* North-west character value */ - int res; /* Final result */ - int *m; /* The cost matrix */ - char *cx; /* Corresponding character values */ - int *toFree = 0; /* Malloced space */ - int mStack[60+15]; /* Stack space to use if not too much is needed */ - int nMatch = 0; - - /* Early out if either input is NULL */ - if( zA==0 || zB==0 ) return -1; - - /* Skip any common prefix */ - while( zA[0] && zA[0]==zB[0] ){ dc = zA[0]; zA++; zB++; nMatch++; } - if( pnMatch ) *pnMatch = nMatch; - if( zA[0]==0 && zB[0]==0 ) return 0; - -#if 0 - printf("A=\"%s\" B=\"%s\" dc=%c\n", zA, zB, dc?dc:' '); -#endif - - /* Verify input strings and measure their lengths */ - for(nA=0; zA[nA]; nA++){ - if( zA[nA]&0x80 ) return -2; - } - for(nB=0; zB[nB]; nB++){ - if( zB[nB]&0x80 ) return -2; - } - - /* Special processing if either string is empty */ - if( nA==0 ){ - cBprev = dc; - for(xB=res=0; (cB = zB[xB])!=0; xB++){ - res += insertOrDeleteCost(cBprev, cB, zB[xB+1])/FINAL_INS_COST_DIV; - cBprev = cB; - } - return res; - } - if( nB==0 ){ - cAprev = dc; - for(xA=res=0; (cA = zA[xA])!=0; xA++){ - res += insertOrDeleteCost(cAprev, cA, zA[xA+1]); - cAprev = cA; - } - return res; - } - - /* A is a prefix of B */ - if( zA[0]=='*' && zA[1]==0 ) return 0; - - /* Allocate and initialize the Wagner matrix */ - if( nB<(sizeof(mStack)*4)/(sizeof(mStack[0])*5) ){ - m = mStack; - }else{ - m = toFree = sqlite3_malloc( (nB+1)*5*sizeof(m[0])/4 ); - if( m==0 ) return -3; - } - cx = (char*)&m[nB+1]; - - /* Compute the Wagner edit distance */ - m[0] = 0; - cx[0] = dc; - cBprev = dc; - for(xB=1; xB<=nB; xB++){ - cBnext = zB[xB]; - cB = zB[xB-1]; - cx[xB] = cB; - m[xB] = m[xB-1] + insertOrDeleteCost(cBprev, cB, cBnext); - cBprev = cB; - } - cAprev = dc; - for(xA=1; xA<=nA; xA++){ - int lastA = (xA==nA); - cA = zA[xA-1]; - cAnext = zA[xA]; - if( cA=='*' && lastA ) break; - d = m[0]; - dc = cx[0]; - m[0] = d + insertOrDeleteCost(cAprev, cA, cAnext); - cBprev = 0; - for(xB=1; xB<=nB; xB++){ - int totalCost, insCost, delCost, subCost, ncx; - cB = zB[xB-1]; - cBnext = zB[xB]; - - /* Cost to insert cB */ - insCost = insertOrDeleteCost(cx[xB-1], cB, cBnext); - if( lastA ) insCost /= FINAL_INS_COST_DIV; - - /* Cost to delete cA */ - delCost = insertOrDeleteCost(cx[xB], cA, cBnext); - - /* Cost to substitute cA->cB */ - subCost = substituteCost(cx[xB-1], cA, cB); - - /* Best cost */ - totalCost = insCost + m[xB-1]; - ncx = cB; - if( (delCost + m[xB])<totalCost ){ - totalCost = delCost + m[xB]; - ncx = cA; - } - if( (subCost + d)<totalCost ){ - totalCost = subCost + d; - } - -#if 0 - printf("%d,%d d=%4d u=%4d r=%4d dc=%c cA=%c cB=%c" - " ins=%4d del=%4d sub=%4d t=%4d ncx=%c\n", - xA, xB, d, m[xB], m[xB-1], dc?dc:' ', cA, cB, - insCost, delCost, subCost, totalCost, ncx?ncx:' '); -#endif - - /* Update the matrix */ - d = m[xB]; - dc = cx[xB]; - m[xB] = totalCost; - cx[xB] = ncx; - cBprev = cB; - } - cAprev = cA; - } - - /* Free the wagner matrix and return the result */ - if( cA=='*' ){ - res = m[1]; - for(xB=1; xB<=nB; xB++){ - if( m[xB]<res ){ - res = m[xB]; - if( pnMatch ) *pnMatch = xB+nMatch; - } - } - }else{ - res = m[nB]; - /* In the current implementation, pnMatch is always NULL if zA does - ** not end in "*" */ - assert( pnMatch==0 ); - } - sqlite3_free(toFree); - return res; -} - -/* -** Function: editdist(A,B) -** -** Return the cost of transforming string A into string B. Both strings -** must be pure ASCII text. If A ends with '*' then it is assumed to be -** a prefix of B and extra characters on the end of B have minimal additional -** cost. -*/ -static void editdistSqlFunc( - sqlite3_context *context, - int argc, - sqlite3_value **argv -){ - int res = editdist1( - (const char*)sqlite3_value_text(argv[0]), - (const char*)sqlite3_value_text(argv[1]), - 0); - if( res<0 ){ - if( res==(-3) ){ - sqlite3_result_error_nomem(context); - }else if( res==(-2) ){ - sqlite3_result_error(context, "non-ASCII input to editdist()", -1); - }else{ - sqlite3_result_error(context, "NULL input to editdist()", -1); - } - }else{ - sqlite3_result_int(context, res); - } -} - -/* End of the fixed-cost edit distance implementation -****************************************************************************** -***************************************************************************** -** Begin: Configurable cost unicode edit distance routines -*/ -/* Forward declaration of structures */ -typedef struct EditDist3Cost EditDist3Cost; -typedef struct EditDist3Config EditDist3Config; -typedef struct EditDist3Point EditDist3Point; -typedef struct EditDist3From EditDist3From; -typedef struct EditDist3FromString EditDist3FromString; -typedef struct EditDist3To EditDist3To; -typedef struct EditDist3ToString EditDist3ToString; -typedef struct EditDist3Lang EditDist3Lang; - - -/* -** An entry in the edit cost table -*/ -struct EditDist3Cost { - EditDist3Cost *pNext; /* Next cost element */ - u8 nFrom; /* Number of bytes in aFrom */ - u8 nTo; /* Number of bytes in aTo */ - u16 iCost; /* Cost of this transformation */ - char a[4] ; /* FROM string followed by TO string */ - /* Additional TO and FROM string bytes appended as necessary */ -}; - -/* -** Edit costs for a particular language ID -*/ -struct EditDist3Lang { - int iLang; /* Language ID */ - int iInsCost; /* Default insertion cost */ - int iDelCost; /* Default deletion cost */ - int iSubCost; /* Default substitution cost */ - EditDist3Cost *pCost; /* Costs */ -}; - - -/* -** The default EditDist3Lang object, with default costs. -*/ -static const EditDist3Lang editDist3Lang = { 0, 100, 100, 150, 0 }; - -/* -** Complete configuration -*/ -struct EditDist3Config { - int nLang; /* Number of language IDs. Size of a[] */ - EditDist3Lang *a; /* One for each distinct language ID */ -}; - -/* -** Extra information about each character in the FROM string. -*/ -struct EditDist3From { - int nSubst; /* Number of substitution cost entries */ - int nDel; /* Number of deletion cost entries */ - int nByte; /* Number of bytes in this character */ - EditDist3Cost **apSubst; /* Array of substitution costs for this element */ - EditDist3Cost **apDel; /* Array of deletion cost entries */ -}; - -/* -** A precompiled FROM string. -* -** In the common case we expect the FROM string to be reused multiple times. -** In other words, the common case will be to measure the edit distance -** from a single origin string to multiple target strings. -*/ -struct EditDist3FromString { - char *z; /* The complete text of the FROM string */ - int n; /* Number of characters in the FROM string */ - int isPrefix; /* True if ends with '*' character */ - EditDist3From *a; /* Extra info about each char of the FROM string */ -}; - -/* -** Extra information about each character in the TO string. -*/ -struct EditDist3To { - int nIns; /* Number of insertion cost entries */ - int nByte; /* Number of bytes in this character */ - EditDist3Cost **apIns; /* Array of deletion cost entries */ -}; - -/* -** A precompiled FROM string -*/ -struct EditDist3ToString { - char *z; /* The complete text of the TO string */ - int n; /* Number of characters in the TO string */ - EditDist3To *a; /* Extra info about each char of the TO string */ -}; - -/* -** Clear or delete an instance of the object that records all edit-distance -** weights. -*/ -static void editDist3ConfigClear(EditDist3Config *p){ - int i; - if( p==0 ) return; - for(i=0; i<p->nLang; i++){ - EditDist3Cost *pCost, *pNext; - pCost = p->a[i].pCost; - while( pCost ){ - pNext = pCost->pNext; - sqlite3_free(pCost); - pCost = pNext; - } - } - sqlite3_free(p->a); - memset(p, 0, sizeof(*p)); -} -static void editDist3ConfigDelete(void *pIn){ - EditDist3Config *p = (EditDist3Config*)pIn; - editDist3ConfigClear(p); - sqlite3_free(p); -} - -/* -** Load all edit-distance weights from a table. -*/ -static int editDist3ConfigLoad( - EditDist3Config *p, /* The edit distance configuration to load */ - sqlite3 *db, /* Load from this database */ - const char *zTable /* Name of the table from which to load */ -){ - sqlite3_stmt *pStmt; - int rc, rc2; - char *zSql; - int iLangPrev = -9999; - EditDist3Lang *pLang = 0; - - zSql = sqlite3_mprintf("SELECT iLang, cFrom, cTo, iCost" - " FROM \"%w\" WHERE iLang>=0 ORDER BY iLang", zTable); - if( zSql==0 ) return SQLITE_NOMEM; - rc = sqlite3_prepare(db, zSql, -1, &pStmt, 0); - sqlite3_free(zSql); - if( rc ) return rc; - editDist3ConfigClear(p); - while( sqlite3_step(pStmt)==SQLITE_ROW ){ - int iLang = sqlite3_column_int(pStmt, 0); - const char *zFrom = (const char*)sqlite3_column_text(pStmt, 1); - int nFrom = zFrom ? sqlite3_column_bytes(pStmt, 1) : 0; - const char *zTo = (const char*)sqlite3_column_text(pStmt, 2); - int nTo = zTo ? sqlite3_column_bytes(pStmt, 2) : 0; - int iCost = sqlite3_column_int(pStmt, 3); - - assert( zFrom!=0 || nFrom==0 ); - assert( zTo!=0 || nTo==0 ); - if( nFrom>100 || nTo>100 ) continue; - if( iCost<0 ) continue; - if( pLang==0 || iLang!=iLangPrev ){ - EditDist3Lang *pNew; - pNew = sqlite3_realloc(p->a, (p->nLang+1)*sizeof(p->a[0])); - if( pNew==0 ){ rc = SQLITE_NOMEM; break; } - p->a = pNew; - pLang = &p->a[p->nLang]; - p->nLang++; - pLang->iLang = iLang; - pLang->iInsCost = 100; - pLang->iDelCost = 100; - pLang->iSubCost = 150; - pLang->pCost = 0; - iLangPrev = iLang; - } - if( nFrom==1 && zFrom[0]=='?' && nTo==0 ){ - pLang->iDelCost = iCost; - }else if( nFrom==0 && nTo==1 && zTo[0]=='?' ){ - pLang->iInsCost = iCost; - }else if( nFrom==1 && nTo==1 && zFrom[0]=='?' && zTo[0]=='?' ){ - pLang->iSubCost = iCost; - }else{ - EditDist3Cost *pCost; - int nExtra = nFrom + nTo - 4; - if( nExtra<0 ) nExtra = 0; - pCost = sqlite3_malloc( sizeof(*pCost) + nExtra ); - if( pCost==0 ){ rc = SQLITE_NOMEM; break; } - pCost->nFrom = nFrom; - pCost->nTo = nTo; - pCost->iCost = iCost; - memcpy(pCost->a, zFrom, nFrom); - memcpy(pCost->a + nFrom, zTo, nTo); - pCost->pNext = pLang->pCost; - pLang->pCost = pCost; - } - } - rc2 = sqlite3_finalize(pStmt); - if( rc==SQLITE_OK ) rc = rc2; - return rc; -} - -/* -** Return the length (in bytes) of a utf-8 character. Or return a maximum -** of N. -*/ -static int utf8Len(unsigned char c, int N){ - int len = 1; - if( c>0x7f ){ - if( (c&0xe0)==0xc0 ){ - len = 2; - }else if( (c&0xf0)==0xe0 ){ - len = 3; - }else{ - len = 4; - } - } - if( len>N ) len = N; - return len; -} - -/* -** Return TRUE (non-zero) if the To side of the given cost matches -** the given string. -*/ -static int matchTo(EditDist3Cost *p, const char *z, int n){ - if( p->nTo>n ) return 0; - if( strncmp(p->a+p->nFrom, z, p->nTo)!=0 ) return 0; - return 1; -} - -/* -** Return TRUE (non-zero) if the From side of the given cost matches -** the given string. -*/ -static int matchFrom(EditDist3Cost *p, const char *z, int n){ - assert( p->nFrom<=n ); - if( strncmp(p->a, z, p->nFrom)!=0 ) return 0; - return 1; -} - -/* -** Return TRUE (non-zero) of the next FROM character and the next TO -** character are the same. -*/ -static int matchFromTo( - EditDist3FromString *pStr, /* Left hand string */ - int n1, /* Index of comparison character on the left */ - const char *z2, /* Right-handl comparison character */ - int n2 /* Bytes remaining in z2[] */ -){ - int b1 = pStr->a[n1].nByte; - if( b1>n2 ) return 0; - if( memcmp(pStr->z+n1, z2, b1)!=0 ) return 0; - return 1; -} - -/* -** Delete an EditDist3FromString objecct -*/ -static void editDist3FromStringDelete(EditDist3FromString *p){ - int i; - if( p ){ - for(i=0; i<p->n; i++){ - sqlite3_free(p->a[i].apDel); - sqlite3_free(p->a[i].apSubst); - } - sqlite3_free(p); - } -} - -/* -** Create a EditDist3FromString object. -*/ -static EditDist3FromString *editDist3FromStringNew( - const EditDist3Lang *pLang, - const char *z, - int n -){ - EditDist3FromString *pStr; - EditDist3Cost *p; - int i; - - if( z==0 ) return 0; - if( n<0 ) n = (int)strlen(z); - pStr = sqlite3_malloc( sizeof(*pStr) + sizeof(pStr->a[0])*n + n + 1 ); - if( pStr==0 ) return 0; - pStr->a = (EditDist3From*)&pStr[1]; - memset(pStr->a, 0, sizeof(pStr->a[0])*n); - pStr->n = n; - pStr->z = (char*)&pStr->a[n]; - memcpy(pStr->z, z, n+1); - if( n && z[n-1]=='*' ){ - pStr->isPrefix = 1; - n--; - pStr->n--; - pStr->z[n] = 0; - }else{ - pStr->isPrefix = 0; - } - - for(i=0; i<n; i++){ - EditDist3From *pFrom = &pStr->a[i]; - memset(pFrom, 0, sizeof(*pFrom)); - pFrom->nByte = utf8Len((unsigned char)z[i], n-i); - for(p=pLang->pCost; p; p=p->pNext){ - EditDist3Cost **apNew; - if( i+p->nFrom>n ) continue; - if( matchFrom(p, z+i, n-i)==0 ) continue; - if( p->nTo==0 ){ - apNew = sqlite3_realloc(pFrom->apDel, - sizeof(*apNew)*(pFrom->nDel+1)); - if( apNew==0 ) break; - pFrom->apDel = apNew; - apNew[pFrom->nDel++] = p; - }else{ - apNew = sqlite3_realloc(pFrom->apSubst, - sizeof(*apNew)*(pFrom->nSubst+1)); - if( apNew==0 ) break; - pFrom->apSubst = apNew; - apNew[pFrom->nSubst++] = p; - } - } - if( p ){ - editDist3FromStringDelete(pStr); - pStr = 0; - break; - } - } - return pStr; -} - -/* -** Update entry m[i] such that it is the minimum of its current value -** and m[j]+iCost. -** -** If the iCost is 1,000,000 or greater, then consider the cost to be -** infinite and skip the update. -*/ -static void updateCost( - unsigned int *m, - int i, - int j, - int iCost -){ - assert( iCost>=0 ); - if( iCost<10000 ){ - unsigned int b = m[j] + iCost; - if( b<m[i] ) m[i] = b; - } -} - -/* Compute the edit distance between two strings. -** -** If an error occurs, return a negative number which is the error code. -** -** If pnMatch is not NULL, then *pnMatch is set to the number of characters -** (not bytes) in z2 that matched the search pattern in *pFrom. If pFrom does -** not contain the pattern for a prefix-search, then this is always the number -** of characters in z2. If pFrom does contain a prefix search pattern, then -** it is the number of characters in the prefix of z2 that was deemed to -** match pFrom. -*/ -static int editDist3Core( - EditDist3FromString *pFrom, /* The FROM string */ - const char *z2, /* The TO string */ - int n2, /* Length of the TO string */ - const EditDist3Lang *pLang, /* Edit weights for a particular language ID */ - int *pnMatch /* OUT: Characters in matched prefix */ -){ - int k, n; - int i1, b1; - int i2, b2; - EditDist3FromString f = *pFrom; - EditDist3To *a2; - unsigned int *m; - int szRow; - EditDist3Cost *p; - int res; - - /* allocate the Wagner matrix and the aTo[] array for the TO string */ - n = (f.n+1)*(n2+1); - n = (n+1)&~1; - m = sqlite3_malloc( n*sizeof(m[0]) + sizeof(a2[0])*n2 ); - if( m==0 ) return -1; /* Out of memory */ - a2 = (EditDist3To*)&m[n]; - memset(a2, 0, sizeof(a2[0])*n2); - - /* Fill in the a1[] matrix for all characters of the TO string */ - for(i2=0; i2<n2; i2++){ - a2[i2].nByte = utf8Len((unsigned char)z2[i2], n2-i2); - for(p=pLang->pCost; p; p=p->pNext){ - EditDist3Cost **apNew; - if( p->nFrom>0 ) continue; - if( i2+p->nTo>n2 ) continue; - if( matchTo(p, z2+i2, n2-i2)==0 ) continue; - a2[i2].nIns++; - apNew = sqlite3_realloc(a2[i2].apIns, sizeof(*apNew)*a2[i2].nIns); - if( apNew==0 ){ - res = -1; /* Out of memory */ - goto editDist3Abort; - } - a2[i2].apIns = apNew; - a2[i2].apIns[a2[i2].nIns-1] = p; - } - } - - /* Prepare to compute the minimum edit distance */ - szRow = f.n+1; - memset(m, 0x01, (n2+1)*szRow*sizeof(m[0])); - m[0] = 0; - - /* First fill in the top-row of the matrix with FROM deletion costs */ - for(i1=0; i1<f.n; i1 += b1){ - b1 = f.a[i1].nByte; - updateCost(m, i1+b1, i1, pLang->iDelCost); - for(k=0; k<f.a[i1].nDel; k++){ - p = f.a[i1].apDel[k]; - updateCost(m, i1+p->nFrom, i1, p->iCost); - } - } - - /* Fill in all subsequent rows, top-to-bottom, left-to-right */ - for(i2=0; i2<n2; i2 += b2){ - int rx; /* Starting index for current row */ - int rxp; /* Starting index for previous row */ - b2 = a2[i2].nByte; - rx = szRow*(i2+b2); - rxp = szRow*i2; - updateCost(m, rx, rxp, pLang->iInsCost); - for(k=0; k<a2[i2].nIns; k++){ - p = a2[i2].apIns[k]; - updateCost(m, szRow*(i2+p->nTo), rxp, p->iCost); - } - for(i1=0; i1<f.n; i1+=b1){ - int cx; /* Index of current cell */ - int cxp; /* Index of cell immediately to the left */ - int cxd; /* Index of cell to the left and one row above */ - int cxu; /* Index of cell immediately above */ - b1 = f.a[i1].nByte; - cxp = rx + i1; - cx = cxp + b1; - cxd = rxp + i1; - cxu = cxd + b1; - updateCost(m, cx, cxp, pLang->iDelCost); - for(k=0; k<f.a[i1].nDel; k++){ - p = f.a[i1].apDel[k]; - updateCost(m, cxp+p->nFrom, cxp, p->iCost); - } - updateCost(m, cx, cxu, pLang->iInsCost); - if( matchFromTo(&f, i1, z2+i2, n2-i2) ){ - updateCost(m, cx, cxd, 0); - } - updateCost(m, cx, cxd, pLang->iSubCost); - for(k=0; k<f.a[i1].nSubst; k++){ - p = f.a[i1].apSubst[k]; - if( matchTo(p, z2+i2, n2-i2) ){ - updateCost(m, cxd+p->nFrom+szRow*p->nTo, cxd, p->iCost); - } - } - } - } - -#if 0 /* Enable for debugging */ - printf(" ^"); - for(i1=0; i1<f.n; i1++) printf(" %c-%2x", f.z[i1], f.z[i1]&0xff); - printf("\n ^:"); - for(i1=0; i1<szRow; i1++){ - int v = m[i1]; - if( v>9999 ) printf(" ****"); - else printf(" %4d", v); - } - printf("\n"); - for(i2=0; i2<n2; i2++){ - printf("%c-%02x:", z2[i2], z2[i2]&0xff); - for(i1=0; i1<szRow; i1++){ - int v = m[(i2+1)*szRow+i1]; - if( v>9999 ) printf(" ****"); - else printf(" %4d", v); - } - printf("\n"); - } -#endif - - /* Free memory allocations and return the result */ - res = (int)m[szRow*(n2+1)-1]; - n = n2; - if( f.isPrefix ){ - for(i2=1; i2<=n2; i2++){ - int b = m[szRow*i2-1]; - if( b<=res ){ - res = b; - n = i2 - 1; - } - } - } - if( pnMatch ){ - int nExtra = 0; - for(k=0; k<n; k++){ - if( (z2[k] & 0xc0)==0x80 ) nExtra++; - } - *pnMatch = n - nExtra; - } - -editDist3Abort: - for(i2=0; i2<n2; i2++) sqlite3_free(a2[i2].apIns); - sqlite3_free(m); - return res; -} - -/* -** Get an appropriate EditDist3Lang object. -*/ -static const EditDist3Lang *editDist3FindLang( - EditDist3Config *pConfig, - int iLang -){ - int i; - for(i=0; i<pConfig->nLang; i++){ - if( pConfig->a[i].iLang==iLang ) return &pConfig->a[i]; - } - return &editDist3Lang; -} - -/* -** Function: editdist3(A,B,iLang) -** editdist3(tablename) -** -** Return the cost of transforming string A into string B using edit -** weights for iLang. -** -** The second form loads edit weights into memory from a table. -*/ -static void editDist3SqlFunc( - sqlite3_context *context, - int argc, - sqlite3_value **argv -){ - EditDist3Config *pConfig = (EditDist3Config*)sqlite3_user_data(context); - sqlite3 *db = sqlite3_context_db_handle(context); - int rc; - if( argc==1 ){ - const char *zTable = (const char*)sqlite3_value_text(argv[0]); - rc = editDist3ConfigLoad(pConfig, db, zTable); - if( rc ) sqlite3_result_error_code(context, rc); - }else{ - const char *zA = (const char*)sqlite3_value_text(argv[0]); - const char *zB = (const char*)sqlite3_value_text(argv[1]); - int nA = sqlite3_value_bytes(argv[0]); - int nB = sqlite3_value_bytes(argv[1]); - int iLang = argc==3 ? sqlite3_value_int(argv[2]) : 0; - const EditDist3Lang *pLang = editDist3FindLang(pConfig, iLang); - EditDist3FromString *pFrom; - int dist; - - pFrom = editDist3FromStringNew(pLang, zA, nA); - if( pFrom==0 ){ - sqlite3_result_error_nomem(context); - return; - } - dist = editDist3Core(pFrom, zB, nB, pLang, 0); - editDist3FromStringDelete(pFrom); - if( dist==(-1) ){ - sqlite3_result_error_nomem(context); - }else{ - sqlite3_result_int(context, dist); - } - } -} - -/* -** Register the editDist3 function with SQLite -*/ -static int editDist3Install(sqlite3 *db){ - int rc; - EditDist3Config *pConfig = sqlite3_malloc( sizeof(*pConfig) ); - if( pConfig==0 ) return SQLITE_NOMEM; - memset(pConfig, 0, sizeof(*pConfig)); - rc = sqlite3_create_function_v2(db, "editdist3", - 2, SQLITE_UTF8, pConfig, editDist3SqlFunc, 0, 0, 0); - if( rc==SQLITE_OK ){ - rc = sqlite3_create_function_v2(db, "editdist3", - 3, SQLITE_UTF8, pConfig, editDist3SqlFunc, 0, 0, 0); - } - if( rc==SQLITE_OK ){ - rc = sqlite3_create_function_v2(db, "editdist3", - 1, SQLITE_UTF8, pConfig, editDist3SqlFunc, 0, 0, - editDist3ConfigDelete); - }else{ - sqlite3_free(pConfig); - } - return rc; -} -/* End configurable cost unicode edit distance routines -****************************************************************************** -****************************************************************************** -** Begin transliterate unicode-to-ascii implementation -*/ - -#if !SQLITE_AMALGAMATION -/* -** This lookup table is used to help decode the first byte of -** a multi-byte UTF8 character. -*/ -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, -}; -#endif - -/* -** Return the value of the first UTF-8 character in the string. -*/ -static int utf8Read(const unsigned char *z, int n, int *pSize){ - int c, i; - - /* All callers to this routine (in the current implementation) - ** always have n>0. */ - if( NEVER(n==0) ){ - c = i = 0; - }else{ - c = z[0]; - i = 1; - if( c>=0xc0 ){ - c = sqlite3Utf8Trans1[c-0xc0]; - while( i<n && (z[i] & 0xc0)==0x80 ){ - c = (c<<6) + (0x3f & z[i++]); - } - } - } - *pSize = i; - return c; -} - -/* -** Return the number of characters in the utf-8 string in the nIn byte -** buffer pointed to by zIn. -*/ -static int utf8Charlen(const char *zIn, int nIn){ - int i; - int nChar = 0; - for(i=0; i<nIn; nChar++){ - int sz; - utf8Read((const unsigned char *)&zIn[i], nIn-i, &sz); - i += sz; - } - return nChar; -} - -/* -** Table of translations from unicode characters into ASCII. -*/ -static const struct { - unsigned short int cFrom; - unsigned char cTo0, cTo1; -} translit[] = { - { 0x00A0, 0x20, 0x00 }, /* to */ - { 0x00B5, 0x75, 0x00 }, /* µ to u */ - { 0x00C0, 0x41, 0x00 }, /* À to A */ - { 0x00C1, 0x41, 0x00 }, /* Á to A */ - { 0x00C2, 0x41, 0x00 }, /*  to A */ - { 0x00C3, 0x41, 0x00 }, /* à to A */ - { 0x00C4, 0x41, 0x65 }, /* Ä to Ae */ - { 0x00C5, 0x41, 0x61 }, /* Å to Aa */ - { 0x00C6, 0x41, 0x45 }, /* Æ to AE */ - { 0x00C7, 0x43, 0x00 }, /* Ç to C */ - { 0x00C8, 0x45, 0x00 }, /* È to E */ - { 0x00C9, 0x45, 0x00 }, /* É to E */ - { 0x00CA, 0x45, 0x00 }, /* Ê to E */ - { 0x00CB, 0x45, 0x00 }, /* Ë to E */ - { 0x00CC, 0x49, 0x00 }, /* Ì to I */ - { 0x00CD, 0x49, 0x00 }, /* Í to I */ - { 0x00CE, 0x49, 0x00 }, /* Î to I */ - { 0x00CF, 0x49, 0x00 }, /* Ï to I */ - { 0x00D0, 0x44, 0x00 }, /* Ð to D */ - { 0x00D1, 0x4E, 0x00 }, /* Ñ to N */ - { 0x00D2, 0x4F, 0x00 }, /* Ò to O */ - { 0x00D3, 0x4F, 0x00 }, /* Ó to O */ - { 0x00D4, 0x4F, 0x00 }, /* Ô to O */ - { 0x00D5, 0x4F, 0x00 }, /* Õ to O */ - { 0x00D6, 0x4F, 0x65 }, /* Ö to Oe */ - { 0x00D7, 0x78, 0x00 }, /* × to x */ - { 0x00D8, 0x4F, 0x00 }, /* Ø to O */ - { 0x00D9, 0x55, 0x00 }, /* Ù to U */ - { 0x00DA, 0x55, 0x00 }, /* Ú to U */ - { 0x00DB, 0x55, 0x00 }, /* Û to U */ - { 0x00DC, 0x55, 0x65 }, /* Ü to Ue */ - { 0x00DD, 0x59, 0x00 }, /* Ý to Y */ - { 0x00DE, 0x54, 0x68 }, /* Þ to Th */ - { 0x00DF, 0x73, 0x73 }, /* ß to ss */ - { 0x00E0, 0x61, 0x00 }, /* à to a */ - { 0x00E1, 0x61, 0x00 }, /* á to a */ - { 0x00E2, 0x61, 0x00 }, /* â to a */ - { 0x00E3, 0x61, 0x00 }, /* ã to a */ - { 0x00E4, 0x61, 0x65 }, /* ä to ae */ - { 0x00E5, 0x61, 0x61 }, /* å to aa */ - { 0x00E6, 0x61, 0x65 }, /* æ to ae */ - { 0x00E7, 0x63, 0x00 }, /* ç to c */ - { 0x00E8, 0x65, 0x00 }, /* è to e */ - { 0x00E9, 0x65, 0x00 }, /* é to e */ - { 0x00EA, 0x65, 0x00 }, /* ê to e */ - { 0x00EB, 0x65, 0x00 }, /* ë to e */ - { 0x00EC, 0x69, 0x00 }, /* ì to i */ - { 0x00ED, 0x69, 0x00 }, /* í to i */ - { 0x00EE, 0x69, 0x00 }, /* î to i */ - { 0x00EF, 0x69, 0x00 }, /* ï to i */ - { 0x00F0, 0x64, 0x00 }, /* ð to d */ - { 0x00F1, 0x6E, 0x00 }, /* ñ to n */ - { 0x00F2, 0x6F, 0x00 }, /* ò to o */ - { 0x00F3, 0x6F, 0x00 }, /* ó to o */ - { 0x00F4, 0x6F, 0x00 }, /* ô to o */ - { 0x00F5, 0x6F, 0x00 }, /* õ to o */ - { 0x00F6, 0x6F, 0x65 }, /* ö to oe */ - { 0x00F7, 0x3A, 0x00 }, /* ÷ to : */ - { 0x00F8, 0x6F, 0x00 }, /* ø to o */ - { 0x00F9, 0x75, 0x00 }, /* ù to u */ - { 0x00FA, 0x75, 0x00 }, /* ú to u */ - { 0x00FB, 0x75, 0x00 }, /* û to u */ - { 0x00FC, 0x75, 0x65 }, /* ü to ue */ - { 0x00FD, 0x79, 0x00 }, /* ý to y */ - { 0x00FE, 0x74, 0x68 }, /* þ to th */ - { 0x00FF, 0x79, 0x00 }, /* ÿ to y */ - { 0x0100, 0x41, 0x00 }, /* Ā to A */ - { 0x0101, 0x61, 0x00 }, /* ā to a */ - { 0x0102, 0x41, 0x00 }, /* Ă to A */ - { 0x0103, 0x61, 0x00 }, /* ă to a */ - { 0x0104, 0x41, 0x00 }, /* Ą to A */ - { 0x0105, 0x61, 0x00 }, /* ą to a */ - { 0x0106, 0x43, 0x00 }, /* Ć to C */ - { 0x0107, 0x63, 0x00 }, /* ć to c */ - { 0x0108, 0x43, 0x68 }, /* Ĉ to Ch */ - { 0x0109, 0x63, 0x68 }, /* ĉ to ch */ - { 0x010A, 0x43, 0x00 }, /* Ċ to C */ - { 0x010B, 0x63, 0x00 }, /* ċ to c */ - { 0x010C, 0x43, 0x00 }, /* Č to C */ - { 0x010D, 0x63, 0x00 }, /* č to c */ - { 0x010E, 0x44, 0x00 }, /* Ď to D */ - { 0x010F, 0x64, 0x00 }, /* ď to d */ - { 0x0110, 0x44, 0x00 }, /* Đ to D */ - { 0x0111, 0x64, 0x00 }, /* đ to d */ - { 0x0112, 0x45, 0x00 }, /* Ē to E */ - { 0x0113, 0x65, 0x00 }, /* ē to e */ - { 0x0114, 0x45, 0x00 }, /* Ĕ to E */ - { 0x0115, 0x65, 0x00 }, /* ĕ to e */ - { 0x0116, 0x45, 0x00 }, /* Ė to E */ - { 0x0117, 0x65, 0x00 }, /* ė to e */ - { 0x0118, 0x45, 0x00 }, /* Ę to E */ - { 0x0119, 0x65, 0x00 }, /* ę to e */ - { 0x011A, 0x45, 0x00 }, /* Ě to E */ - { 0x011B, 0x65, 0x00 }, /* ě to e */ - { 0x011C, 0x47, 0x68 }, /* Ĝ to Gh */ - { 0x011D, 0x67, 0x68 }, /* ĝ to gh */ - { 0x011E, 0x47, 0x00 }, /* Ğ to G */ - { 0x011F, 0x67, 0x00 }, /* ğ to g */ - { 0x0120, 0x47, 0x00 }, /* Ġ to G */ - { 0x0121, 0x67, 0x00 }, /* ġ to g */ - { 0x0122, 0x47, 0x00 }, /* Ģ to G */ - { 0x0123, 0x67, 0x00 }, /* ģ to g */ - { 0x0124, 0x48, 0x68 }, /* Ĥ to Hh */ - { 0x0125, 0x68, 0x68 }, /* ĥ to hh */ - { 0x0126, 0x48, 0x00 }, /* Ħ to H */ - { 0x0127, 0x68, 0x00 }, /* ħ to h */ - { 0x0128, 0x49, 0x00 }, /* Ĩ to I */ - { 0x0129, 0x69, 0x00 }, /* ĩ to i */ - { 0x012A, 0x49, 0x00 }, /* Ī to I */ - { 0x012B, 0x69, 0x00 }, /* ī to i */ - { 0x012C, 0x49, 0x00 }, /* Ĭ to I */ - { 0x012D, 0x69, 0x00 }, /* ĭ to i */ - { 0x012E, 0x49, 0x00 }, /* Į to I */ - { 0x012F, 0x69, 0x00 }, /* į to i */ - { 0x0130, 0x49, 0x00 }, /* İ to I */ - { 0x0131, 0x69, 0x00 }, /* ı to i */ - { 0x0132, 0x49, 0x4A }, /* IJ to IJ */ - { 0x0133, 0x69, 0x6A }, /* ij to ij */ - { 0x0134, 0x4A, 0x68 }, /* Ĵ to Jh */ - { 0x0135, 0x6A, 0x68 }, /* ĵ to jh */ - { 0x0136, 0x4B, 0x00 }, /* Ķ to K */ - { 0x0137, 0x6B, 0x00 }, /* ķ to k */ - { 0x0138, 0x6B, 0x00 }, /* ĸ to k */ - { 0x0139, 0x4C, 0x00 }, /* Ĺ to L */ - { 0x013A, 0x6C, 0x00 }, /* ĺ to l */ - { 0x013B, 0x4C, 0x00 }, /* Ļ to L */ - { 0x013C, 0x6C, 0x00 }, /* ļ to l */ - { 0x013D, 0x4C, 0x00 }, /* Ľ to L */ - { 0x013E, 0x6C, 0x00 }, /* ľ to l */ - { 0x013F, 0x4C, 0x2E }, /* Ŀ to L. */ - { 0x0140, 0x6C, 0x2E }, /* ŀ to l. */ - { 0x0141, 0x4C, 0x00 }, /* Ł to L */ - { 0x0142, 0x6C, 0x00 }, /* ł to l */ - { 0x0143, 0x4E, 0x00 }, /* Ń to N */ - { 0x0144, 0x6E, 0x00 }, /* ń to n */ - { 0x0145, 0x4E, 0x00 }, /* Ņ to N */ - { 0x0146, 0x6E, 0x00 }, /* ņ to n */ - { 0x0147, 0x4E, 0x00 }, /* Ň to N */ - { 0x0148, 0x6E, 0x00 }, /* ň to n */ - { 0x0149, 0x27, 0x6E }, /* ʼn to 'n */ - { 0x014A, 0x4E, 0x47 }, /* Ŋ to NG */ - { 0x014B, 0x6E, 0x67 }, /* ŋ to ng */ - { 0x014C, 0x4F, 0x00 }, /* Ō to O */ - { 0x014D, 0x6F, 0x00 }, /* ō to o */ - { 0x014E, 0x4F, 0x00 }, /* Ŏ to O */ - { 0x014F, 0x6F, 0x00 }, /* ŏ to o */ - { 0x0150, 0x4F, 0x00 }, /* Ő to O */ - { 0x0151, 0x6F, 0x00 }, /* ő to o */ - { 0x0152, 0x4F, 0x45 }, /* Œ to OE */ - { 0x0153, 0x6F, 0x65 }, /* œ to oe */ - { 0x0154, 0x52, 0x00 }, /* Ŕ to R */ - { 0x0155, 0x72, 0x00 }, /* ŕ to r */ - { 0x0156, 0x52, 0x00 }, /* Ŗ to R */ - { 0x0157, 0x72, 0x00 }, /* ŗ to r */ - { 0x0158, 0x52, 0x00 }, /* Ř to R */ - { 0x0159, 0x72, 0x00 }, /* ř to r */ - { 0x015A, 0x53, 0x00 }, /* Ś to S */ - { 0x015B, 0x73, 0x00 }, /* ś to s */ - { 0x015C, 0x53, 0x68 }, /* Ŝ to Sh */ - { 0x015D, 0x73, 0x68 }, /* ŝ to sh */ - { 0x015E, 0x53, 0x00 }, /* Ş to S */ - { 0x015F, 0x73, 0x00 }, /* ş to s */ - { 0x0160, 0x53, 0x00 }, /* Š to S */ - { 0x0161, 0x73, 0x00 }, /* š to s */ - { 0x0162, 0x54, 0x00 }, /* Ţ to T */ - { 0x0163, 0x74, 0x00 }, /* ţ to t */ - { 0x0164, 0x54, 0x00 }, /* Ť to T */ - { 0x0165, 0x74, 0x00 }, /* ť to t */ - { 0x0166, 0x54, 0x00 }, /* Ŧ to T */ - { 0x0167, 0x74, 0x00 }, /* ŧ to t */ - { 0x0168, 0x55, 0x00 }, /* Ũ to U */ - { 0x0169, 0x75, 0x00 }, /* ũ to u */ - { 0x016A, 0x55, 0x00 }, /* Ū to U */ - { 0x016B, 0x75, 0x00 }, /* ū to u */ - { 0x016C, 0x55, 0x00 }, /* Ŭ to U */ - { 0x016D, 0x75, 0x00 }, /* ŭ to u */ - { 0x016E, 0x55, 0x00 }, /* Ů to U */ - { 0x016F, 0x75, 0x00 }, /* ů to u */ - { 0x0170, 0x55, 0x00 }, /* Ű to U */ - { 0x0171, 0x75, 0x00 }, /* ű to u */ - { 0x0172, 0x55, 0x00 }, /* Ų to U */ - { 0x0173, 0x75, 0x00 }, /* ų to u */ - { 0x0174, 0x57, 0x00 }, /* Ŵ to W */ - { 0x0175, 0x77, 0x00 }, /* ŵ to w */ - { 0x0176, 0x59, 0x00 }, /* Ŷ to Y */ - { 0x0177, 0x79, 0x00 }, /* ŷ to y */ - { 0x0178, 0x59, 0x00 }, /* Ÿ to Y */ - { 0x0179, 0x5A, 0x00 }, /* Ź to Z */ - { 0x017A, 0x7A, 0x00 }, /* ź to z */ - { 0x017B, 0x5A, 0x00 }, /* Ż to Z */ - { 0x017C, 0x7A, 0x00 }, /* ż to z */ - { 0x017D, 0x5A, 0x00 }, /* Ž to Z */ - { 0x017E, 0x7A, 0x00 }, /* ž to z */ - { 0x017F, 0x73, 0x00 }, /* ſ to s */ - { 0x0192, 0x66, 0x00 }, /* ƒ to f */ - { 0x0218, 0x53, 0x00 }, /* Ș to S */ - { 0x0219, 0x73, 0x00 }, /* ș to s */ - { 0x021A, 0x54, 0x00 }, /* Ț to T */ - { 0x021B, 0x74, 0x00 }, /* ț to t */ - { 0x0386, 0x41, 0x00 }, /* Ά to A */ - { 0x0388, 0x45, 0x00 }, /* Έ to E */ - { 0x0389, 0x49, 0x00 }, /* Ή to I */ - { 0x038A, 0x49, 0x00 }, /* Ί to I */ - { 0x038C, 0x4f, 0x00 }, /* Ό to O */ - { 0x038E, 0x59, 0x00 }, /* Ύ to Y */ - { 0x038F, 0x4f, 0x00 }, /* Ώ to O */ - { 0x0390, 0x69, 0x00 }, /* ΐ to i */ - { 0x0391, 0x41, 0x00 }, /* Α to A */ - { 0x0392, 0x42, 0x00 }, /* Β to B */ - { 0x0393, 0x47, 0x00 }, /* Γ to G */ - { 0x0394, 0x44, 0x00 }, /* Δ to D */ - { 0x0395, 0x45, 0x00 }, /* Ε to E */ - { 0x0396, 0x5a, 0x00 }, /* Ζ to Z */ - { 0x0397, 0x49, 0x00 }, /* Η to I */ - { 0x0398, 0x54, 0x68 }, /* Θ to Th */ - { 0x0399, 0x49, 0x00 }, /* Ι to I */ - { 0x039A, 0x4b, 0x00 }, /* Κ to K */ - { 0x039B, 0x4c, 0x00 }, /* Λ to L */ - { 0x039C, 0x4d, 0x00 }, /* Μ to M */ - { 0x039D, 0x4e, 0x00 }, /* Ν to N */ - { 0x039E, 0x58, 0x00 }, /* Ξ to X */ - { 0x039F, 0x4f, 0x00 }, /* Ο to O */ - { 0x03A0, 0x50, 0x00 }, /* Π to P */ - { 0x03A1, 0x52, 0x00 }, /* Ρ to R */ - { 0x03A3, 0x53, 0x00 }, /* Σ to S */ - { 0x03A4, 0x54, 0x00 }, /* Τ to T */ - { 0x03A5, 0x59, 0x00 }, /* Υ to Y */ - { 0x03A6, 0x46, 0x00 }, /* Φ to F */ - { 0x03A7, 0x43, 0x68 }, /* Χ to Ch */ - { 0x03A8, 0x50, 0x73 }, /* Ψ to Ps */ - { 0x03A9, 0x4f, 0x00 }, /* Ω to O */ - { 0x03AA, 0x49, 0x00 }, /* Ϊ to I */ - { 0x03AB, 0x59, 0x00 }, /* Ϋ to Y */ - { 0x03AC, 0x61, 0x00 }, /* ά to a */ - { 0x03AD, 0x65, 0x00 }, /* έ to e */ - { 0x03AE, 0x69, 0x00 }, /* ή to i */ - { 0x03AF, 0x69, 0x00 }, /* ί to i */ - { 0x03B1, 0x61, 0x00 }, /* α to a */ - { 0x03B2, 0x62, 0x00 }, /* β to b */ - { 0x03B3, 0x67, 0x00 }, /* γ to g */ - { 0x03B4, 0x64, 0x00 }, /* δ to d */ - { 0x03B5, 0x65, 0x00 }, /* ε to e */ - { 0x03B6, 0x7a, 0x00 }, /* ζ to z */ - { 0x03B7, 0x69, 0x00 }, /* η to i */ - { 0x03B8, 0x74, 0x68 }, /* θ to th */ - { 0x03B9, 0x69, 0x00 }, /* ι to i */ - { 0x03BA, 0x6b, 0x00 }, /* κ to k */ - { 0x03BB, 0x6c, 0x00 }, /* λ to l */ - { 0x03BC, 0x6d, 0x00 }, /* μ to m */ - { 0x03BD, 0x6e, 0x00 }, /* ν to n */ - { 0x03BE, 0x78, 0x00 }, /* ξ to x */ - { 0x03BF, 0x6f, 0x00 }, /* ο to o */ - { 0x03C0, 0x70, 0x00 }, /* π to p */ - { 0x03C1, 0x72, 0x00 }, /* ρ to r */ - { 0x03C3, 0x73, 0x00 }, /* σ to s */ - { 0x03C4, 0x74, 0x00 }, /* τ to t */ - { 0x03C5, 0x79, 0x00 }, /* υ to y */ - { 0x03C6, 0x66, 0x00 }, /* φ to f */ - { 0x03C7, 0x63, 0x68 }, /* χ to ch */ - { 0x03C8, 0x70, 0x73 }, /* ψ to ps */ - { 0x03C9, 0x6f, 0x00 }, /* ω to o */ - { 0x03CA, 0x69, 0x00 }, /* ϊ to i */ - { 0x03CB, 0x79, 0x00 }, /* ϋ to y */ - { 0x03CC, 0x6f, 0x00 }, /* ό to o */ - { 0x03CD, 0x79, 0x00 }, /* ύ to y */ - { 0x03CE, 0x69, 0x00 }, /* ώ to i */ - { 0x0400, 0x45, 0x00 }, /* Ѐ to E */ - { 0x0401, 0x45, 0x00 }, /* Ё to E */ - { 0x0402, 0x44, 0x00 }, /* Ђ to D */ - { 0x0403, 0x47, 0x00 }, /* Ѓ to G */ - { 0x0404, 0x45, 0x00 }, /* Є to E */ - { 0x0405, 0x5a, 0x00 }, /* Ѕ to Z */ - { 0x0406, 0x49, 0x00 }, /* І to I */ - { 0x0407, 0x49, 0x00 }, /* Ї to I */ - { 0x0408, 0x4a, 0x00 }, /* Ј to J */ - { 0x0409, 0x49, 0x00 }, /* Љ to I */ - { 0x040A, 0x4e, 0x00 }, /* Њ to N */ - { 0x040B, 0x44, 0x00 }, /* Ћ to D */ - { 0x040C, 0x4b, 0x00 }, /* Ќ to K */ - { 0x040D, 0x49, 0x00 }, /* Ѝ to I */ - { 0x040E, 0x55, 0x00 }, /* Ў to U */ - { 0x040F, 0x44, 0x00 }, /* Џ to D */ - { 0x0410, 0x41, 0x00 }, /* А to A */ - { 0x0411, 0x42, 0x00 }, /* Б to B */ - { 0x0412, 0x56, 0x00 }, /* В to V */ - { 0x0413, 0x47, 0x00 }, /* Г to G */ - { 0x0414, 0x44, 0x00 }, /* Д to D */ - { 0x0415, 0x45, 0x00 }, /* Е to E */ - { 0x0416, 0x5a, 0x68 }, /* Ж to Zh */ - { 0x0417, 0x5a, 0x00 }, /* З to Z */ - { 0x0418, 0x49, 0x00 }, /* И to I */ - { 0x0419, 0x49, 0x00 }, /* Й to I */ - { 0x041A, 0x4b, 0x00 }, /* К to K */ - { 0x041B, 0x4c, 0x00 }, /* Л to L */ - { 0x041C, 0x4d, 0x00 }, /* М to M */ - { 0x041D, 0x4e, 0x00 }, /* Н to N */ - { 0x041E, 0x4f, 0x00 }, /* О to O */ - { 0x041F, 0x50, 0x00 }, /* П to P */ - { 0x0420, 0x52, 0x00 }, /* Р to R */ - { 0x0421, 0x53, 0x00 }, /* С to S */ - { 0x0422, 0x54, 0x00 }, /* Т to T */ - { 0x0423, 0x55, 0x00 }, /* У to U */ - { 0x0424, 0x46, 0x00 }, /* Ф to F */ - { 0x0425, 0x4b, 0x68 }, /* Х to Kh */ - { 0x0426, 0x54, 0x63 }, /* Ц to Tc */ - { 0x0427, 0x43, 0x68 }, /* Ч to Ch */ - { 0x0428, 0x53, 0x68 }, /* Ш to Sh */ - { 0x0429, 0x53, 0x68 }, /* Щ to Shch */ - { 0x042A, 0x61, 0x00 }, /* to A */ - { 0x042B, 0x59, 0x00 }, /* Ы to Y */ - { 0x042C, 0x59, 0x00 }, /* to Y */ - { 0x042D, 0x45, 0x00 }, /* Э to E */ - { 0x042E, 0x49, 0x75 }, /* Ю to Iu */ - { 0x042F, 0x49, 0x61 }, /* Я to Ia */ - { 0x0430, 0x61, 0x00 }, /* а to a */ - { 0x0431, 0x62, 0x00 }, /* б to b */ - { 0x0432, 0x76, 0x00 }, /* в to v */ - { 0x0433, 0x67, 0x00 }, /* г to g */ - { 0x0434, 0x64, 0x00 }, /* д to d */ - { 0x0435, 0x65, 0x00 }, /* е to e */ - { 0x0436, 0x7a, 0x68 }, /* ж to zh */ - { 0x0437, 0x7a, 0x00 }, /* з to z */ - { 0x0438, 0x69, 0x00 }, /* и to i */ - { 0x0439, 0x69, 0x00 }, /* й to i */ - { 0x043A, 0x6b, 0x00 }, /* к to k */ - { 0x043B, 0x6c, 0x00 }, /* л to l */ - { 0x043C, 0x6d, 0x00 }, /* м to m */ - { 0x043D, 0x6e, 0x00 }, /* н to n */ - { 0x043E, 0x6f, 0x00 }, /* о to o */ - { 0x043F, 0x70, 0x00 }, /* п to p */ - { 0x0440, 0x72, 0x00 }, /* р to r */ - { 0x0441, 0x73, 0x00 }, /* с to s */ - { 0x0442, 0x74, 0x00 }, /* т to t */ - { 0x0443, 0x75, 0x00 }, /* у to u */ - { 0x0444, 0x66, 0x00 }, /* ф to f */ - { 0x0445, 0x6b, 0x68 }, /* х to kh */ - { 0x0446, 0x74, 0x63 }, /* ц to tc */ - { 0x0447, 0x63, 0x68 }, /* ч to ch */ - { 0x0448, 0x73, 0x68 }, /* ш to sh */ - { 0x0449, 0x73, 0x68 }, /* щ to shch */ - { 0x044A, 0x61, 0x00 }, /* to a */ - { 0x044B, 0x79, 0x00 }, /* ы to y */ - { 0x044C, 0x79, 0x00 }, /* to y */ - { 0x044D, 0x65, 0x00 }, /* э to e */ - { 0x044E, 0x69, 0x75 }, /* ю to iu */ - { 0x044F, 0x69, 0x61 }, /* я to ia */ - { 0x0450, 0x65, 0x00 }, /* ѐ to e */ - { 0x0451, 0x65, 0x00 }, /* ё to e */ - { 0x0452, 0x64, 0x00 }, /* ђ to d */ - { 0x0453, 0x67, 0x00 }, /* ѓ to g */ - { 0x0454, 0x65, 0x00 }, /* є to e */ - { 0x0455, 0x7a, 0x00 }, /* ѕ to z */ - { 0x0456, 0x69, 0x00 }, /* і to i */ - { 0x0457, 0x69, 0x00 }, /* ї to i */ - { 0x0458, 0x6a, 0x00 }, /* ј to j */ - { 0x0459, 0x69, 0x00 }, /* љ to i */ - { 0x045A, 0x6e, 0x00 }, /* њ to n */ - { 0x045B, 0x64, 0x00 }, /* ћ to d */ - { 0x045C, 0x6b, 0x00 }, /* ќ to k */ - { 0x045D, 0x69, 0x00 }, /* ѝ to i */ - { 0x045E, 0x75, 0x00 }, /* ў to u */ - { 0x045F, 0x64, 0x00 }, /* џ to d */ - { 0x1E02, 0x42, 0x00 }, /* Ḃ to B */ - { 0x1E03, 0x62, 0x00 }, /* ḃ to b */ - { 0x1E0A, 0x44, 0x00 }, /* Ḋ to D */ - { 0x1E0B, 0x64, 0x00 }, /* ḋ to d */ - { 0x1E1E, 0x46, 0x00 }, /* Ḟ to F */ - { 0x1E1F, 0x66, 0x00 }, /* ḟ to f */ - { 0x1E40, 0x4D, 0x00 }, /* Ṁ to M */ - { 0x1E41, 0x6D, 0x00 }, /* ṁ to m */ - { 0x1E56, 0x50, 0x00 }, /* Ṗ to P */ - { 0x1E57, 0x70, 0x00 }, /* ṗ to p */ - { 0x1E60, 0x53, 0x00 }, /* Ṡ to S */ - { 0x1E61, 0x73, 0x00 }, /* ṡ to s */ - { 0x1E6A, 0x54, 0x00 }, /* Ṫ to T */ - { 0x1E6B, 0x74, 0x00 }, /* ṫ to t */ - { 0x1E80, 0x57, 0x00 }, /* Ẁ to W */ - { 0x1E81, 0x77, 0x00 }, /* ẁ to w */ - { 0x1E82, 0x57, 0x00 }, /* Ẃ to W */ - { 0x1E83, 0x77, 0x00 }, /* ẃ to w */ - { 0x1E84, 0x57, 0x00 }, /* Ẅ to W */ - { 0x1E85, 0x77, 0x00 }, /* ẅ to w */ - { 0x1EF2, 0x59, 0x00 }, /* Ỳ to Y */ - { 0x1EF3, 0x79, 0x00 }, /* ỳ to y */ - { 0xFB00, 0x66, 0x66 }, /* ff to ff */ - { 0xFB01, 0x66, 0x69 }, /* fi to fi */ - { 0xFB02, 0x66, 0x6C }, /* fl to fl */ - { 0xFB05, 0x73, 0x74 }, /* ſt to st */ - { 0xFB06, 0x73, 0x74 }, /* st to st */ -}; - -/* -** Convert the input string from UTF-8 into pure ASCII by converting -** all non-ASCII characters to some combination of characters in the -** ASCII subset. -** -** The returned string might contain more characters than the input. -** -** Space to hold the returned string comes from sqlite3_malloc() and -** should be freed by the caller. -*/ -static unsigned char *transliterate(const unsigned char *zIn, int nIn){ - unsigned char *zOut = sqlite3_malloc( nIn*4 + 1 ); - int c, sz, nOut; - if( zOut==0 ) return 0; - nOut = 0; - while( nIn>0 ){ - c = utf8Read(zIn, nIn, &sz); - zIn += sz; - nIn -= sz; - if( c<=127 ){ - zOut[nOut++] = c; - }else{ - int xTop, xBtm, x; - xTop = sizeof(translit)/sizeof(translit[0]) - 1; - xBtm = 0; - while( xTop>=xBtm ){ - x = (xTop + xBtm)/2; - if( translit[x].cFrom==c ){ - zOut[nOut++] = translit[x].cTo0; - if( translit[x].cTo1 ){ - zOut[nOut++] = translit[x].cTo1; - /* Add an extra "ch" after the "sh" for Щ and щ */ - if( c==0x0429 || c== 0x0449 ){ - zOut[nOut++] = 'c'; - zOut[nOut++] = 'h'; - } - } - c = 0; - break; - }else if( translit[x].cFrom>c ){ - xTop = x-1; - }else{ - xBtm = x+1; - } - } - if( c ) zOut[nOut++] = '?'; - } - } - zOut[nOut] = 0; - return zOut; -} - -/* -** Return the number of characters in the shortest prefix of the input -** string that transliterates to an ASCII string nTrans bytes or longer. -** Or, if the transliteration of the input string is less than nTrans -** bytes in size, return the number of characters in the input string. -*/ -static int translen_to_charlen(const char *zIn, int nIn, int nTrans){ - int i, c, sz, nOut; - int nChar; - - i = nOut = 0; - for(nChar=0; i<nIn && nOut<nTrans; nChar++){ - c = utf8Read((const unsigned char *)&zIn[i], nIn-i, &sz); - i += sz; - - nOut++; - if( c>=128 ){ - int xTop, xBtm, x; - xTop = sizeof(translit)/sizeof(translit[0]) - 1; - xBtm = 0; - while( xTop>=xBtm ){ - x = (xTop + xBtm)/2; - if( translit[x].cFrom==c ){ - if( translit[x].cTo1 ) nOut++; - if( c==0x0429 || c== 0x0449 ) nOut += 2; - break; - }else if( translit[x].cFrom>c ){ - xTop = x-1; - }else{ - xBtm = x+1; - } - } - } - } - - return nChar; -} - - -/* -** spellfix1_translit(X) -** -** Convert a string that contains non-ASCII Roman characters into -** pure ASCII. -*/ -static void transliterateSqlFunc( - sqlite3_context *context, - int argc, - sqlite3_value **argv -){ - const unsigned char *zIn = sqlite3_value_text(argv[0]); - int nIn = sqlite3_value_bytes(argv[0]); - unsigned char *zOut = transliterate(zIn, nIn); - if( zOut==0 ){ - sqlite3_result_error_nomem(context); - }else{ - sqlite3_result_text(context, (char*)zOut, -1, sqlite3_free); - } -} - -/* -** spellfix1_scriptcode(X) -** -** Try to determine the dominant script used by the word X and return -** its ISO 15924 numeric code. -** -** The current implementation only understands the following scripts: -** -** 215 (Latin) -** 220 (Cyrillic) -** 200 (Greek) -** -** This routine will return 998 if the input X contains characters from -** two or more of the above scripts or 999 if X contains no characters -** from any of the above scripts. -*/ -static void scriptCodeSqlFunc( - sqlite3_context *context, - int argc, - sqlite3_value **argv -){ - const unsigned char *zIn = sqlite3_value_text(argv[0]); - int nIn = sqlite3_value_bytes(argv[0]); - int c, sz; - int scriptMask = 0; - int res; -# define SCRIPT_LATIN 0x0001 -# define SCRIPT_CYRILLIC 0x0002 -# define SCRIPT_GREEK 0x0004 - - while( nIn>0 ){ - c = utf8Read(zIn, nIn, &sz); - zIn += sz; - nIn -= sz; - if( c<0x02af ){ - scriptMask |= SCRIPT_LATIN; - }else if( c>=0x0400 && c<=0x04ff ){ - scriptMask |= SCRIPT_CYRILLIC; - }else if( c>=0x0386 && c<=0x03ce ){ - scriptMask |= SCRIPT_GREEK; - } - } - switch( scriptMask ){ - case 0: res = 999; break; - case SCRIPT_LATIN: res = 215; break; - case SCRIPT_CYRILLIC: res = 220; break; - case SCRIPT_GREEK: res = 200; break; - default: res = 998; break; - } - sqlite3_result_int(context, res); -} - -/* End transliterate -****************************************************************************** -****************************************************************************** -** Begin spellfix1 virtual table. -*/ - -/* Maximum length of a phonehash used for querying the shadow table */ -#define SPELLFIX_MX_HASH 8 - -/* Maximum number of hash strings to examine per query */ -#define SPELLFIX_MX_RUN 1 - -typedef struct spellfix1_vtab spellfix1_vtab; -typedef struct spellfix1_cursor spellfix1_cursor; - -/* Fuzzy-search virtual table object */ -struct spellfix1_vtab { - sqlite3_vtab base; /* Base class - must be first */ - sqlite3 *db; /* Database connection */ - char *zDbName; /* Name of database holding this table */ - char *zTableName; /* Name of the virtual table */ - char *zCostTable; /* Table holding edit-distance cost numbers */ - EditDist3Config *pConfig3; /* Parsed edit distance costs */ -}; - -/* Fuzzy-search cursor object */ -struct spellfix1_cursor { - sqlite3_vtab_cursor base; /* Base class - must be first */ - spellfix1_vtab *pVTab; /* The table to which this cursor belongs */ - char *zPattern; /* rhs of MATCH clause */ - int nRow; /* Number of rows of content */ - int nAlloc; /* Number of allocated rows */ - int iRow; /* Current row of content */ - int iLang; /* Value of the langid= constraint */ - int iTop; /* Value of the top= constraint */ - int iScope; /* Value of the scope= constraint */ - int nSearch; /* Number of vocabulary items checked */ - sqlite3_stmt *pFullScan; /* Shadow query for a full table scan */ - struct spellfix1_row { /* For each row of content */ - sqlite3_int64 iRowid; /* Rowid for this row */ - char *zWord; /* Text for this row */ - int iRank; /* Rank for this row */ - int iDistance; /* Distance from pattern for this row */ - int iScore; /* Score for sorting */ - int iMatchlen; /* Value of matchlen column (or -1) */ - char zHash[SPELLFIX_MX_HASH]; /* the phonehash used for this match */ - } *a; -}; - -/* -** 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 spellfix1DbExec( - 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); - } -} - -/* -** xDisconnect/xDestroy method for the fuzzy-search module. -*/ -static int spellfix1Uninit(int isDestroy, sqlite3_vtab *pVTab){ - spellfix1_vtab *p = (spellfix1_vtab*)pVTab; - int rc = SQLITE_OK; - if( isDestroy ){ - sqlite3 *db = p->db; - spellfix1DbExec(&rc, db, "DROP TABLE IF EXISTS \"%w\".\"%w_vocab\"", - p->zDbName, p->zTableName); - } - if( rc==SQLITE_OK ){ - sqlite3_free(p->zTableName); - editDist3ConfigDelete(p->pConfig3); - sqlite3_free(p->zCostTable); - sqlite3_free(p); - } - return rc; -} -static int spellfix1Disconnect(sqlite3_vtab *pVTab){ - return spellfix1Uninit(0, pVTab); -} -static int spellfix1Destroy(sqlite3_vtab *pVTab){ - return spellfix1Uninit(1, pVTab); -} - -/* -** Make a copy of a string. Remove leading and trailing whitespace -** and dequote it. -*/ -static char *spellfix1Dequote(const char *zIn){ - char *zOut; - int i, j; - char c; - while( isspace(zIn[0]) ) zIn++; - zOut = sqlite3_mprintf("%s", zIn); - if( zOut==0 ) return 0; - i = (int)strlen(zOut); -#if 0 /* The parser will never leave spaces at the end */ - while( i>0 && isspace(zOut[i-1]) ){ i--; } -#endif - zOut[i] = 0; - c = zOut[0]; - if( c=='\'' || c=='"' ){ - for(i=1, j=0; ALWAYS(zOut[i]); i++){ - zOut[j++] = zOut[i]; - if( zOut[i]==c ){ - if( zOut[i+1]==c ){ - i++; - }else{ - zOut[j-1] = 0; - break; - } - } - } - } - return zOut; -} - - -/* -** xConnect/xCreate method for the spellfix1 module. Arguments are: -** -** argv[0] -> module name ("spellfix1") -** argv[1] -> database name -** argv[2] -> table name -** argv[3].. -> optional arguments (i.e. "edit_cost_table" parameter) -*/ -static int spellfix1Init( - int isCreate, - sqlite3 *db, - void *pAux, - int argc, const char *const*argv, - sqlite3_vtab **ppVTab, - char **pzErr -){ - spellfix1_vtab *pNew = 0; - const char *zModule = argv[0]; - const char *zDbName = argv[1]; - const char *zTableName = argv[2]; - int nDbName; - int rc = SQLITE_OK; - int i; - - nDbName = (int)strlen(zDbName); - pNew = sqlite3_malloc( sizeof(*pNew) + nDbName + 1); - if( pNew==0 ){ - rc = SQLITE_NOMEM; - }else{ - memset(pNew, 0, sizeof(*pNew)); - pNew->zDbName = (char*)&pNew[1]; - memcpy(pNew->zDbName, zDbName, nDbName+1); - pNew->zTableName = sqlite3_mprintf("%s", zTableName); - pNew->db = db; - if( pNew->zTableName==0 ){ - rc = SQLITE_NOMEM; - }else{ - rc = sqlite3_declare_vtab(db, - "CREATE TABLE x(word,rank,distance,langid, " - "score, matchlen, phonehash HIDDEN, " - "top HIDDEN, scope HIDDEN, srchcnt HIDDEN, " - "soundslike HIDDEN, command HIDDEN)" - ); -#define SPELLFIX_COL_WORD 0 -#define SPELLFIX_COL_RANK 1 -#define SPELLFIX_COL_DISTANCE 2 -#define SPELLFIX_COL_LANGID 3 -#define SPELLFIX_COL_SCORE 4 -#define SPELLFIX_COL_MATCHLEN 5 -#define SPELLFIX_COL_PHONEHASH 6 -#define SPELLFIX_COL_TOP 7 -#define SPELLFIX_COL_SCOPE 8 -#define SPELLFIX_COL_SRCHCNT 9 -#define SPELLFIX_COL_SOUNDSLIKE 10 -#define SPELLFIX_COL_COMMAND 11 - } - if( rc==SQLITE_OK && isCreate ){ - sqlite3_uint64 r; - spellfix1DbExec(&rc, db, - "CREATE TABLE IF NOT EXISTS \"%w\".\"%w_vocab\"(\n" - " id INTEGER PRIMARY KEY,\n" - " rank INT,\n" - " langid INT,\n" - " word TEXT,\n" - " k1 TEXT,\n" - " k2 TEXT\n" - ");\n", - zDbName, zTableName - ); - sqlite3_randomness(sizeof(r), &r); - spellfix1DbExec(&rc, db, - "CREATE INDEX IF NOT EXISTS \"%w\".\"%w_index_%llx\" " - "ON \"%w_vocab\"(langid,k2);", - zDbName, zModule, r, zTableName - ); - } - for(i=3; rc==SQLITE_OK && i<argc; i++){ - if( strncmp(argv[i],"edit_cost_table=",16)==0 && pNew->zCostTable==0 ){ - pNew->zCostTable = spellfix1Dequote(&argv[i][16]); - if( pNew->zCostTable==0 ) rc = SQLITE_NOMEM; - continue; - } - *pzErr = sqlite3_mprintf("bad argument to spellfix1(): \"%s\"", argv[i]); - rc = SQLITE_ERROR; - } - } - - if( rc && pNew ){ - *ppVTab = 0; - spellfix1Uninit(0, &pNew->base); - }else{ - *ppVTab = (sqlite3_vtab *)pNew; - } - return rc; -} - -/* -** The xConnect and xCreate methods -*/ -static int spellfix1Connect( - sqlite3 *db, - void *pAux, - int argc, const char *const*argv, - sqlite3_vtab **ppVTab, - char **pzErr -){ - return spellfix1Init(0, db, pAux, argc, argv, ppVTab, pzErr); -} -static int spellfix1Create( - sqlite3 *db, - void *pAux, - int argc, const char *const*argv, - sqlite3_vtab **ppVTab, - char **pzErr -){ - return spellfix1Init(1, db, pAux, argc, argv, ppVTab, pzErr); -} - -/* -** Clear all of the content from a cursor. -*/ -static void spellfix1ResetCursor(spellfix1_cursor *pCur){ - int i; - for(i=0; i<pCur->nRow; i++){ - sqlite3_free(pCur->a[i].zWord); - } - pCur->nRow = 0; - pCur->iRow = 0; - pCur->nSearch = 0; - if( pCur->pFullScan ){ - sqlite3_finalize(pCur->pFullScan); - pCur->pFullScan = 0; - } -} - -/* -** Resize the cursor to hold up to N rows of content -*/ -static void spellfix1ResizeCursor(spellfix1_cursor *pCur, int N){ - struct spellfix1_row *aNew; - assert( N>=pCur->nRow ); - aNew = sqlite3_realloc(pCur->a, sizeof(pCur->a[0])*N); - if( aNew==0 && N>0 ){ - spellfix1ResetCursor(pCur); - sqlite3_free(pCur->a); - pCur->nAlloc = 0; - pCur->a = 0; - }else{ - pCur->nAlloc = N; - pCur->a = aNew; - } -} - - -/* -** Close a fuzzy-search cursor. -*/ -static int spellfix1Close(sqlite3_vtab_cursor *cur){ - spellfix1_cursor *pCur = (spellfix1_cursor *)cur; - spellfix1ResetCursor(pCur); - spellfix1ResizeCursor(pCur, 0); - sqlite3_free(pCur->zPattern); - sqlite3_free(pCur); - return SQLITE_OK; -} - -/* -** Search for terms of these forms: -** -** (A) word MATCH $str -** (B) langid == $langid -** (C) top = $top -** (D) scope = $scope -** (E) distance < $distance -** (F) distance <= $distance -** -** The plan number is a bit mask formed with these bits: -** -** 0x01 (A) is found -** 0x02 (B) is found -** 0x04 (C) is found -** 0x08 (D) is found -** 0x10 (E) is found -** 0x20 (F) is found -** -** filter.argv[*] values contains $str, $langid, $top, and $scope, -** if specified and in that order. -*/ -static int spellfix1BestIndex(sqlite3_vtab *tab, sqlite3_index_info *pIdxInfo){ - int iPlan = 0; - int iLangTerm = -1; - int iTopTerm = -1; - int iScopeTerm = -1; - int iDistTerm = -1; - int i; - const struct sqlite3_index_constraint *pConstraint; - pConstraint = pIdxInfo->aConstraint; - for(i=0; i<pIdxInfo->nConstraint; i++, pConstraint++){ - if( pConstraint->usable==0 ) continue; - - /* Terms of the form: word MATCH $str */ - if( (iPlan & 1)==0 - && pConstraint->iColumn==SPELLFIX_COL_WORD - && pConstraint->op==SQLITE_INDEX_CONSTRAINT_MATCH - ){ - iPlan |= 1; - pIdxInfo->aConstraintUsage[i].argvIndex = 1; - pIdxInfo->aConstraintUsage[i].omit = 1; - } - - /* Terms of the form: langid = $langid */ - if( (iPlan & 2)==0 - && pConstraint->iColumn==SPELLFIX_COL_LANGID - && pConstraint->op==SQLITE_INDEX_CONSTRAINT_EQ - ){ - iPlan |= 2; - iLangTerm = i; - } - - /* Terms of the form: top = $top */ - if( (iPlan & 4)==0 - && pConstraint->iColumn==SPELLFIX_COL_TOP - && pConstraint->op==SQLITE_INDEX_CONSTRAINT_EQ - ){ - iPlan |= 4; - iTopTerm = i; - } - - /* Terms of the form: scope = $scope */ - if( (iPlan & 8)==0 - && pConstraint->iColumn==SPELLFIX_COL_SCOPE - && pConstraint->op==SQLITE_INDEX_CONSTRAINT_EQ - ){ - iPlan |= 8; - iScopeTerm = i; - } - - /* Terms of the form: distance < $dist or distance <= $dist */ - if( (iPlan & (16|32))==0 - && pConstraint->iColumn==SPELLFIX_COL_DISTANCE - && (pConstraint->op==SQLITE_INDEX_CONSTRAINT_LT - || pConstraint->op==SQLITE_INDEX_CONSTRAINT_LE) - ){ - iPlan |= pConstraint->op==SQLITE_INDEX_CONSTRAINT_LT ? 16 : 32; - iDistTerm = i; - } - } - if( iPlan&1 ){ - int idx = 2; - pIdxInfo->idxNum = iPlan; - if( pIdxInfo->nOrderBy==1 - && pIdxInfo->aOrderBy[0].iColumn==SPELLFIX_COL_SCORE - && pIdxInfo->aOrderBy[0].desc==0 - ){ - pIdxInfo->orderByConsumed = 1; /* Default order by iScore */ - } - if( iPlan&2 ){ - pIdxInfo->aConstraintUsage[iLangTerm].argvIndex = idx++; - pIdxInfo->aConstraintUsage[iLangTerm].omit = 1; - } - if( iPlan&4 ){ - pIdxInfo->aConstraintUsage[iTopTerm].argvIndex = idx++; - pIdxInfo->aConstraintUsage[iTopTerm].omit = 1; - } - if( iPlan&8 ){ - pIdxInfo->aConstraintUsage[iScopeTerm].argvIndex = idx++; - pIdxInfo->aConstraintUsage[iScopeTerm].omit = 1; - } - if( iPlan&(16|32) ){ - pIdxInfo->aConstraintUsage[iDistTerm].argvIndex = idx++; - pIdxInfo->aConstraintUsage[iDistTerm].omit = 1; - } - pIdxInfo->estimatedCost = (double)10000; - }else{ - pIdxInfo->idxNum = 0; - pIdxInfo->estimatedCost = (double)10000000; - } - return SQLITE_OK; -} - -/* -** Open a new fuzzy-search cursor. -*/ -static int spellfix1Open(sqlite3_vtab *pVTab, sqlite3_vtab_cursor **ppCursor){ - spellfix1_vtab *p = (spellfix1_vtab*)pVTab; - spellfix1_cursor *pCur; - pCur = sqlite3_malloc( sizeof(*pCur) ); - if( pCur==0 ) return SQLITE_NOMEM; - memset(pCur, 0, sizeof(*pCur)); - pCur->pVTab = p; - *ppCursor = &pCur->base; - return SQLITE_OK; -} - -/* -** Adjust a distance measurement by the words rank in order to show -** preference to common words. -*/ -static int spellfix1Score(int iDistance, int iRank){ - int iLog2; - for(iLog2=0; iRank>0; iLog2++, iRank>>=1){} - return iDistance + 32 - iLog2; -} - -/* -** Compare two spellfix1_row objects for sorting purposes in qsort() such -** that they sort in order of increasing distance. -*/ -static int spellfix1RowCompare(const void *A, const void *B){ - const struct spellfix1_row *a = (const struct spellfix1_row*)A; - const struct spellfix1_row *b = (const struct spellfix1_row*)B; - return a->iScore - b->iScore; -} - -/* -** A structure used to pass information from spellfix1FilterForMatch() -** into spellfix1RunQuery(). -*/ -typedef struct MatchQuery { - spellfix1_cursor *pCur; /* The cursor being queried */ - sqlite3_stmt *pStmt; /* shadow table query statment */ - char zHash[SPELLFIX_MX_HASH]; /* The current phonehash for zPattern */ - const char *zPattern; /* Transliterated input string */ - int nPattern; /* Length of zPattern */ - EditDist3FromString *pMatchStr3; /* Original unicode string */ - EditDist3Config *pConfig3; /* Edit-distance cost coefficients */ - const EditDist3Lang *pLang; /* The selected language coefficients */ - int iLang; /* The language id */ - int iScope; /* Default scope */ - int iMaxDist; /* Maximum allowed edit distance, or -1 */ - int rc; /* Error code */ - int nRun; /* Number of prior runs for the same zPattern */ - char azPrior[SPELLFIX_MX_RUN][SPELLFIX_MX_HASH]; /* Prior hashes */ -} MatchQuery; - -/* -** Run a query looking for the best matches against zPattern using -** zHash as the character class seed hash. -*/ -static void spellfix1RunQuery(MatchQuery *p, const char *zQuery, int nQuery){ - const char *zK1; - const char *zWord; - int iDist; - int iRank; - int iScore; - int iWorst = 0; - int idx; - int idxWorst = -1; - int i; - int iScope = p->iScope; - spellfix1_cursor *pCur = p->pCur; - sqlite3_stmt *pStmt = p->pStmt; - char zHash1[SPELLFIX_MX_HASH]; - char zHash2[SPELLFIX_MX_HASH]; - char *zClass; - int nClass; - int rc; - - if( pCur->a==0 || p->rc ) return; /* Prior memory allocation failure */ - zClass = (char*)phoneticHash((unsigned char*)zQuery, nQuery); - if( zClass==0 ){ - p->rc = SQLITE_NOMEM; - return; - } - nClass = (int)strlen(zClass); - if( nClass>SPELLFIX_MX_HASH-2 ){ - nClass = SPELLFIX_MX_HASH-2; - zClass[nClass] = 0; - } - if( nClass<=iScope ){ - if( nClass>2 ){ - iScope = nClass-1; - }else{ - iScope = nClass; - } - } - memcpy(zHash1, zClass, iScope); - sqlite3_free(zClass); - zHash1[iScope] = 0; - memcpy(zHash2, zHash1, iScope); - zHash2[iScope] = 'Z'; - zHash2[iScope+1] = 0; -#if SPELLFIX_MX_RUN>1 - for(i=0; i<p->nRun; i++){ - if( strcmp(p->azPrior[i], zHash1)==0 ) return; - } -#endif - assert( p->nRun<SPELLFIX_MX_RUN ); - memcpy(p->azPrior[p->nRun++], zHash1, iScope+1); - if( sqlite3_bind_text(pStmt, 1, zHash1, -1, SQLITE_STATIC)==SQLITE_NOMEM - || sqlite3_bind_text(pStmt, 2, zHash2, -1, SQLITE_STATIC)==SQLITE_NOMEM - ){ - p->rc = SQLITE_NOMEM; - return; - } -#if SPELLFIX_MX_RUN>1 - for(i=0; i<pCur->nRow; i++){ - if( pCur->a[i].iScore>iWorst ){ - iWorst = pCur->a[i].iScore; - idxWorst = i; - } - } -#endif - while( sqlite3_step(pStmt)==SQLITE_ROW ){ - int iMatchlen = -1; - iRank = sqlite3_column_int(pStmt, 2); - if( p->pMatchStr3 ){ - int nWord = sqlite3_column_bytes(pStmt, 1); - zWord = (const char*)sqlite3_column_text(pStmt, 1); - iDist = editDist3Core(p->pMatchStr3, zWord, nWord, p->pLang, &iMatchlen); - }else{ - zK1 = (const char*)sqlite3_column_text(pStmt, 3); - if( zK1==0 ) continue; - iDist = editdist1(p->zPattern, zK1, 0); - } - if( iDist<0 ){ - p->rc = SQLITE_NOMEM; - break; - } - pCur->nSearch++; - iScore = spellfix1Score(iDist,iRank); - if( p->iMaxDist>=0 ){ - if( iDist>p->iMaxDist ) continue; - if( pCur->nRow>=pCur->nAlloc-1 ){ - spellfix1ResizeCursor(pCur, pCur->nAlloc*2 + 10); - if( pCur->a==0 ) break; - } - idx = pCur->nRow; - }else if( pCur->nRow<pCur->nAlloc ){ - idx = pCur->nRow; - }else if( iScore<iWorst ){ - idx = idxWorst; - sqlite3_free(pCur->a[idx].zWord); - }else{ - continue; - } - pCur->a[idx].zWord = sqlite3_mprintf("%s", sqlite3_column_text(pStmt, 1)); - if( pCur->a[idx].zWord==0 ){ - p->rc = SQLITE_NOMEM; - break; - } - pCur->a[idx].iRowid = sqlite3_column_int64(pStmt, 0); - pCur->a[idx].iRank = iRank; - pCur->a[idx].iDistance = iDist; - pCur->a[idx].iScore = iScore; - pCur->a[idx].iMatchlen = iMatchlen; - memcpy(pCur->a[idx].zHash, zHash1, iScope+1); - if( pCur->nRow<pCur->nAlloc ) pCur->nRow++; - if( pCur->nRow==pCur->nAlloc ){ - iWorst = pCur->a[0].iScore; - idxWorst = 0; - for(i=1; i<pCur->nRow; i++){ - iScore = pCur->a[i].iScore; - if( iWorst<iScore ){ - iWorst = iScore; - idxWorst = i; - } - } - } - } - rc = sqlite3_reset(pStmt); - if( rc ) p->rc = rc; -} - -/* -** This version of the xFilter method work if the MATCH term is present -** and we are doing a scan. -*/ -static int spellfix1FilterForMatch( - spellfix1_cursor *pCur, - int idxNum, - int argc, - sqlite3_value **argv -){ - const unsigned char *zMatchThis; /* RHS of the MATCH operator */ - EditDist3FromString *pMatchStr3 = 0; /* zMatchThis as an editdist string */ - char *zPattern; /* Transliteration of zMatchThis */ - int nPattern; /* Length of zPattern */ - int iLimit = 20; /* Max number of rows of output */ - int iScope = 3; /* Use this many characters of zClass */ - int iLang = 0; /* Language code */ - char *zSql; /* SQL of shadow table query */ - sqlite3_stmt *pStmt = 0; /* Shadow table query */ - int rc; /* Result code */ - int idx = 1; /* Next available filter parameter */ - spellfix1_vtab *p = pCur->pVTab; /* The virtual table that owns pCur */ - MatchQuery x; /* For passing info to RunQuery() */ - - /* Load the cost table if we have not already done so */ - if( p->zCostTable!=0 && p->pConfig3==0 ){ - p->pConfig3 = sqlite3_malloc( sizeof(p->pConfig3[0]) ); - if( p->pConfig3==0 ) return SQLITE_NOMEM; - memset(p->pConfig3, 0, sizeof(p->pConfig3[0])); - rc = editDist3ConfigLoad(p->pConfig3, p->db, p->zCostTable); - if( rc ) return rc; - } - memset(&x, 0, sizeof(x)); - x.iScope = 3; /* Default scope if none specified by "WHERE scope=N" */ - x.iMaxDist = -1; /* Maximum allowed edit distance */ - - if( idxNum&2 ){ - iLang = sqlite3_value_int(argv[idx++]); - } - if( idxNum&4 ){ - iLimit = sqlite3_value_int(argv[idx++]); - if( iLimit<1 ) iLimit = 1; - } - if( idxNum&8 ){ - x.iScope = sqlite3_value_int(argv[idx++]); - if( x.iScope<1 ) x.iScope = 1; - if( x.iScope>SPELLFIX_MX_HASH-2 ) x.iScope = SPELLFIX_MX_HASH-2; - } - if( idxNum&(16|32) ){ - x.iMaxDist = sqlite3_value_int(argv[idx++]); - if( idxNum&16 ) x.iMaxDist--; - if( x.iMaxDist<0 ) x.iMaxDist = 0; - } - spellfix1ResetCursor(pCur); - spellfix1ResizeCursor(pCur, iLimit); - zMatchThis = sqlite3_value_text(argv[0]); - if( zMatchThis==0 ) return SQLITE_OK; - if( p->pConfig3 ){ - x.pLang = editDist3FindLang(p->pConfig3, iLang); - pMatchStr3 = editDist3FromStringNew(x.pLang, (const char*)zMatchThis, -1); - if( pMatchStr3==0 ){ - x.rc = SQLITE_NOMEM; - goto filter_exit; - } - }else{ - x.pLang = 0; - } - zPattern = (char*)transliterate(zMatchThis, sqlite3_value_bytes(argv[0])); - sqlite3_free(pCur->zPattern); - pCur->zPattern = zPattern; - if( zPattern==0 ){ - x.rc = SQLITE_NOMEM; - goto filter_exit; - } - nPattern = (int)strlen(zPattern); - if( zPattern[nPattern-1]=='*' ) nPattern--; - zSql = sqlite3_mprintf( - "SELECT id, word, rank, k1" - " FROM \"%w\".\"%w_vocab\"" - " WHERE langid=%d AND k2>=?1 AND k2<?2", - p->zDbName, p->zTableName, iLang - ); - if( zSql==0 ){ - x.rc = SQLITE_NOMEM; - pStmt = 0; - goto filter_exit; - } - rc = sqlite3_prepare_v2(p->db, zSql, -1, &pStmt, 0); - sqlite3_free(zSql); - pCur->iLang = iLang; - x.pCur = pCur; - x.pStmt = pStmt; - x.zPattern = zPattern; - x.nPattern = nPattern; - x.pMatchStr3 = pMatchStr3; - x.iLang = iLang; - x.rc = rc; - x.pConfig3 = p->pConfig3; - if( x.rc==SQLITE_OK ){ - spellfix1RunQuery(&x, zPattern, nPattern); - } - - if( pCur->a ){ - qsort(pCur->a, pCur->nRow, sizeof(pCur->a[0]), spellfix1RowCompare); - pCur->iTop = iLimit; - pCur->iScope = iScope; - }else{ - x.rc = SQLITE_NOMEM; - } - -filter_exit: - sqlite3_finalize(pStmt); - editDist3FromStringDelete(pMatchStr3); - return x.rc; -} - -/* -** This version of xFilter handles a full-table scan case -*/ -static int spellfix1FilterForFullScan( - spellfix1_cursor *pCur, - int idxNum, - int argc, - sqlite3_value **argv -){ - int rc; - char *zSql; - spellfix1_vtab *pVTab = pCur->pVTab; - spellfix1ResetCursor(pCur); - zSql = sqlite3_mprintf( - "SELECT word, rank, NULL, langid, id FROM \"%w\".\"%w_vocab\"", - pVTab->zDbName, pVTab->zTableName); - if( zSql==0 ) return SQLITE_NOMEM; - rc = sqlite3_prepare_v2(pVTab->db, zSql, -1, &pCur->pFullScan, 0); - sqlite3_free(zSql); - pCur->nRow = pCur->iRow = 0; - if( rc==SQLITE_OK ){ - rc = sqlite3_step(pCur->pFullScan); - if( rc==SQLITE_ROW ){ pCur->iRow = -1; rc = SQLITE_OK; } - if( rc==SQLITE_DONE ){ rc = SQLITE_OK; } - }else{ - pCur->iRow = 0; - } - return rc; -} - - -/* -** Called to "rewind" a cursor back to the beginning so that -** it starts its output over again. Always called at least once -** prior to any spellfix1Column, spellfix1Rowid, or spellfix1Eof call. -*/ -static int spellfix1Filter( - sqlite3_vtab_cursor *cur, - int idxNum, const char *idxStr, - int argc, sqlite3_value **argv -){ - spellfix1_cursor *pCur = (spellfix1_cursor *)cur; - int rc; - if( idxNum & 1 ){ - rc = spellfix1FilterForMatch(pCur, idxNum, argc, argv); - }else{ - rc = spellfix1FilterForFullScan(pCur, idxNum, argc, argv); - } - return rc; -} - - -/* -** Advance a cursor to its next row of output -*/ -static int spellfix1Next(sqlite3_vtab_cursor *cur){ - spellfix1_cursor *pCur = (spellfix1_cursor *)cur; - int rc = SQLITE_OK; - if( pCur->iRow < pCur->nRow ){ - if( pCur->pFullScan ){ - rc = sqlite3_step(pCur->pFullScan); - if( rc!=SQLITE_ROW ) pCur->iRow = pCur->nRow; - if( rc==SQLITE_ROW || rc==SQLITE_DONE ) rc = SQLITE_OK; - }else{ - pCur->iRow++; - } - } - return rc; -} - -/* -** Return TRUE if we are at the end-of-file -*/ -static int spellfix1Eof(sqlite3_vtab_cursor *cur){ - spellfix1_cursor *pCur = (spellfix1_cursor *)cur; - return pCur->iRow>=pCur->nRow; -} - -/* -** Return columns from the current row. -*/ -static int spellfix1Column( - sqlite3_vtab_cursor *cur, - sqlite3_context *ctx, - int i -){ - spellfix1_cursor *pCur = (spellfix1_cursor*)cur; - if( pCur->pFullScan ){ - if( i<=SPELLFIX_COL_LANGID ){ - sqlite3_result_value(ctx, sqlite3_column_value(pCur->pFullScan, i)); - }else{ - sqlite3_result_null(ctx); - } - return SQLITE_OK; - } - switch( i ){ - case SPELLFIX_COL_WORD: { - sqlite3_result_text(ctx, pCur->a[pCur->iRow].zWord, -1, SQLITE_STATIC); - break; - } - case SPELLFIX_COL_RANK: { - sqlite3_result_int(ctx, pCur->a[pCur->iRow].iRank); - break; - } - case SPELLFIX_COL_DISTANCE: { - sqlite3_result_int(ctx, pCur->a[pCur->iRow].iDistance); - break; - } - case SPELLFIX_COL_LANGID: { - sqlite3_result_int(ctx, pCur->iLang); - break; - } - case SPELLFIX_COL_SCORE: { - sqlite3_result_int(ctx, pCur->a[pCur->iRow].iScore); - break; - } - case SPELLFIX_COL_MATCHLEN: { - int iMatchlen = pCur->a[pCur->iRow].iMatchlen; - if( iMatchlen<0 ){ - int nPattern = (int)strlen(pCur->zPattern); - char *zWord = pCur->a[pCur->iRow].zWord; - int nWord = (int)strlen(zWord); - - if( nPattern>0 && pCur->zPattern[nPattern-1]=='*' ){ - char *zTranslit; - int res; - zTranslit = (char *)transliterate((unsigned char *)zWord, nWord); - if( !zTranslit ) return SQLITE_NOMEM; - res = editdist1(pCur->zPattern, zTranslit, &iMatchlen); - sqlite3_free(zTranslit); - if( res<0 ) return SQLITE_NOMEM; - iMatchlen = translen_to_charlen(zWord, nWord, iMatchlen); - }else{ - iMatchlen = utf8Charlen(zWord, nWord); - } - } - - sqlite3_result_int(ctx, iMatchlen); - break; - } - case SPELLFIX_COL_PHONEHASH: { - sqlite3_result_text(ctx, pCur->a[pCur->iRow].zHash, -1, SQLITE_STATIC); - break; - } - case SPELLFIX_COL_TOP: { - sqlite3_result_int(ctx, pCur->iTop); - break; - } - case SPELLFIX_COL_SCOPE: { - sqlite3_result_int(ctx, pCur->iScope); - break; - } - case SPELLFIX_COL_SRCHCNT: { - sqlite3_result_int(ctx, pCur->nSearch); - break; - } - default: { - sqlite3_result_null(ctx); - break; - } - } - return SQLITE_OK; -} - -/* -** The rowid. -*/ -static int spellfix1Rowid(sqlite3_vtab_cursor *cur, sqlite_int64 *pRowid){ - spellfix1_cursor *pCur = (spellfix1_cursor*)cur; - if( pCur->pFullScan ){ - *pRowid = sqlite3_column_int64(pCur->pFullScan, 4); - }else{ - *pRowid = pCur->a[pCur->iRow].iRowid; - } - return SQLITE_OK; -} - -/* -** The xUpdate() method. -*/ -static int spellfix1Update( - sqlite3_vtab *pVTab, - int argc, - sqlite3_value **argv, - sqlite_int64 *pRowid -){ - int rc = SQLITE_OK; - sqlite3_int64 rowid, newRowid; - spellfix1_vtab *p = (spellfix1_vtab*)pVTab; - sqlite3 *db = p->db; - - if( argc==1 ){ - /* A delete operation on the rowid given by argv[0] */ - rowid = *pRowid = sqlite3_value_int64(argv[0]); - spellfix1DbExec(&rc, db, "DELETE FROM \"%w\".\"%w_vocab\" " - " WHERE id=%lld", - p->zDbName, p->zTableName, rowid); - }else{ - const unsigned char *zWord = sqlite3_value_text(argv[SPELLFIX_COL_WORD+2]); - int nWord = sqlite3_value_bytes(argv[SPELLFIX_COL_WORD+2]); - int iLang = sqlite3_value_int(argv[SPELLFIX_COL_LANGID+2]); - int iRank = sqlite3_value_int(argv[SPELLFIX_COL_RANK+2]); - const unsigned char *zSoundslike = - sqlite3_value_text(argv[SPELLFIX_COL_SOUNDSLIKE+2]); - int nSoundslike = sqlite3_value_bytes(argv[SPELLFIX_COL_SOUNDSLIKE+2]); - char *zK1, *zK2; - int i; - char c; - - if( zWord==0 ){ - /* Inserts of the form: INSERT INTO table(command) VALUES('xyzzy'); - ** cause zWord to be NULL, so we look at the "command" column to see - ** what special actions to take */ - const char *zCmd = - (const char*)sqlite3_value_text(argv[SPELLFIX_COL_COMMAND+2]); - if( zCmd==0 ){ - pVTab->zErrMsg = sqlite3_mprintf("%s.word may not be NULL", - p->zTableName); - return SQLITE_CONSTRAINT_NOTNULL; - } - if( strcmp(zCmd,"reset")==0 ){ - /* Reset the edit cost table (if there is one). */ - editDist3ConfigDelete(p->pConfig3); - p->pConfig3 = 0; - return SQLITE_OK; - } - if( strncmp(zCmd,"edit_cost_table=",16)==0 ){ - editDist3ConfigDelete(p->pConfig3); - p->pConfig3 = 0; - sqlite3_free(p->zCostTable); - p->zCostTable = spellfix1Dequote(zCmd+16); - if( p->zCostTable==0 ) return SQLITE_NOMEM; - if( p->zCostTable[0]==0 || sqlite3_stricmp(p->zCostTable,"null")==0 ){ - sqlite3_free(p->zCostTable); - p->zCostTable = 0; - } - return SQLITE_OK; - } - pVTab->zErrMsg = sqlite3_mprintf("unknown value for %s.command: \"%w\"", - p->zTableName, zCmd); - return SQLITE_ERROR; - } - if( iRank<1 ) iRank = 1; - if( zSoundslike ){ - zK1 = (char*)transliterate(zSoundslike, nSoundslike); - }else{ - zK1 = (char*)transliterate(zWord, nWord); - } - if( zK1==0 ) return SQLITE_NOMEM; - for(i=0; (c = zK1[i])!=0; i++){ - if( c>='A' && c<='Z' ) zK1[i] += 'a' - 'A'; - } - zK2 = (char*)phoneticHash((const unsigned char*)zK1, i); - if( zK2==0 ){ - sqlite3_free(zK1); - return SQLITE_NOMEM; - } - if( sqlite3_value_type(argv[0])==SQLITE_NULL ){ - spellfix1DbExec(&rc, db, - "INSERT INTO \"%w\".\"%w_vocab\"(rank,langid,word,k1,k2) " - "VALUES(%d,%d,%Q,%Q,%Q)", - p->zDbName, p->zTableName, - iRank, iLang, zWord, zK1, zK2 - ); - *pRowid = sqlite3_last_insert_rowid(db); - }else{ - rowid = sqlite3_value_int64(argv[0]); - newRowid = *pRowid = sqlite3_value_int64(argv[1]); - spellfix1DbExec(&rc, db, - "UPDATE \"%w\".\"%w_vocab\" SET id=%lld, rank=%d, langid=%d," - " word=%Q, k1=%Q, k2=%Q WHERE id=%lld", - p->zDbName, p->zTableName, newRowid, iRank, iLang, - zWord, zK1, zK2, rowid - ); - } - sqlite3_free(zK1); - sqlite3_free(zK2); - } - return rc; -} - -/* -** Rename the spellfix1 table. -*/ -static int spellfix1Rename(sqlite3_vtab *pVTab, const char *zNew){ - spellfix1_vtab *p = (spellfix1_vtab*)pVTab; - sqlite3 *db = p->db; - int rc = SQLITE_OK; - char *zNewName = sqlite3_mprintf("%s", zNew); - if( zNewName==0 ){ - return SQLITE_NOMEM; - } - spellfix1DbExec(&rc, db, - "ALTER TABLE \"%w\".\"%w_vocab\" RENAME TO \"%w_vocab\"", - p->zDbName, p->zTableName, zNewName - ); - if( rc==SQLITE_OK ){ - sqlite3_free(p->zTableName); - p->zTableName = zNewName; - }else{ - sqlite3_free(zNewName); - } - return rc; -} - - -/* -** A virtual table module that provides fuzzy search. -*/ -static sqlite3_module spellfix1Module = { - 0, /* iVersion */ - spellfix1Create, /* xCreate - handle CREATE VIRTUAL TABLE */ - spellfix1Connect, /* xConnect - reconnected to an existing table */ - spellfix1BestIndex, /* xBestIndex - figure out how to do a query */ - spellfix1Disconnect, /* xDisconnect - close a connection */ - spellfix1Destroy, /* xDestroy - handle DROP TABLE */ - spellfix1Open, /* xOpen - open a cursor */ - spellfix1Close, /* xClose - close a cursor */ - spellfix1Filter, /* xFilter - configure scan constraints */ - spellfix1Next, /* xNext - advance a cursor */ - spellfix1Eof, /* xEof - check for end of scan */ - spellfix1Column, /* xColumn - read data */ - spellfix1Rowid, /* xRowid - read data */ - spellfix1Update, /* xUpdate */ - 0, /* xBegin */ - 0, /* xSync */ - 0, /* xCommit */ - 0, /* xRollback */ - 0, /* xFindMethod */ - spellfix1Rename, /* xRename */ -}; - -/* -** Register the various functions and the virtual table. -*/ -static int spellfix1Register(sqlite3 *db){ - int rc = SQLITE_OK; - int i; - rc = sqlite3_create_function(db, "spellfix1_translit", 1, SQLITE_UTF8, 0, - transliterateSqlFunc, 0, 0); - if( rc==SQLITE_OK ){ - rc = sqlite3_create_function(db, "spellfix1_editdist", 2, SQLITE_UTF8, 0, - editdistSqlFunc, 0, 0); - } - if( rc==SQLITE_OK ){ - rc = sqlite3_create_function(db, "spellfix1_phonehash", 1, SQLITE_UTF8, 0, - phoneticHashSqlFunc, 0, 0); - } - if( rc==SQLITE_OK ){ - rc = sqlite3_create_function(db, "spellfix1_scriptcode", 1, SQLITE_UTF8, 0, - scriptCodeSqlFunc, 0, 0); - } - if( rc==SQLITE_OK ){ - rc = sqlite3_create_module(db, "spellfix1", &spellfix1Module, 0); - } - if( rc==SQLITE_OK ){ - rc = editDist3Install(db); - } - - /* Verify sanity of the translit[] table */ - for(i=0; i<sizeof(translit)/sizeof(translit[0])-1; i++){ - assert( translit[i].cFrom<translit[i+1].cFrom ); - } - - return rc; -} - -#if SQLITE_CORE || defined(SQLITE_TEST) -/* -** Register the spellfix1 virtual table and its associated functions. -*/ -int sqlite3_spellfix1_register(sqlite3 *db){ - return spellfix1Register(db); -} -#endif - - -#if !SQLITE_CORE -/* -** Extension load function. -*/ -int sqlite3_spellfix1_init( - sqlite3 *db, - char **pzErrMsg, - const sqlite3_api_routines *pApi -){ - SQLITE_EXTENSION_INIT2(pApi); - return spellfix1Register(db); -} -#endif /* !SQLITE_CORE */ diff --git a/src/test_wholenumber.c b/src/test_wholenumber.c deleted file mode 100644 index 7c42d0169..000000000 --- a/src/test_wholenumber.c +++ /dev/null @@ -1,311 +0,0 @@ -/* -** 2011 April 02 -** -** 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 virtual table that returns the whole numbers -** between 1 and 4294967295, inclusive. -** -** Example: -** -** CREATE VIRTUAL TABLE nums USING wholenumber; -** SELECT value FROM nums WHERE value<10; -** -** Results in: -** -** 1 2 3 4 5 6 7 8 9 -*/ -#include "sqlite3.h" -#include <assert.h> -#include <string.h> - -#ifndef SQLITE_OMIT_VIRTUALTABLE - - -/* A wholenumber cursor object */ -typedef struct wholenumber_cursor wholenumber_cursor; -struct wholenumber_cursor { - sqlite3_vtab_cursor base; /* Base class - must be first */ - sqlite3_int64 iValue; /* Current value */ - sqlite3_int64 mxValue; /* Maximum value */ -}; - -/* Methods for the wholenumber module */ -static int wholenumberConnect( - sqlite3 *db, - void *pAux, - int argc, const char *const*argv, - sqlite3_vtab **ppVtab, - char **pzErr -){ - sqlite3_vtab *pNew; - pNew = *ppVtab = sqlite3_malloc( sizeof(*pNew) ); - if( pNew==0 ) return SQLITE_NOMEM; - sqlite3_declare_vtab(db, "CREATE TABLE x(value)"); - memset(pNew, 0, sizeof(*pNew)); - return SQLITE_OK; -} -/* Note that for this virtual table, the xCreate and xConnect -** methods are identical. */ - -static int wholenumberDisconnect(sqlite3_vtab *pVtab){ - sqlite3_free(pVtab); - return SQLITE_OK; -} -/* The xDisconnect and xDestroy methods are also the same */ - - -/* -** Open a new wholenumber cursor. -*/ -static int wholenumberOpen(sqlite3_vtab *p, sqlite3_vtab_cursor **ppCursor){ - wholenumber_cursor *pCur; - pCur = sqlite3_malloc( sizeof(*pCur) ); - if( pCur==0 ) return SQLITE_NOMEM; - memset(pCur, 0, sizeof(*pCur)); - *ppCursor = &pCur->base; - return SQLITE_OK; -} - -/* -** Close a wholenumber cursor. -*/ -static int wholenumberClose(sqlite3_vtab_cursor *cur){ - sqlite3_free(cur); - return SQLITE_OK; -} - - -/* -** Advance a cursor to its next row of output -*/ -static int wholenumberNext(sqlite3_vtab_cursor *cur){ - wholenumber_cursor *pCur = (wholenumber_cursor*)cur; - pCur->iValue++; - return SQLITE_OK; -} - -/* -** Return the value associated with a wholenumber. -*/ -static int wholenumberColumn( - sqlite3_vtab_cursor *cur, - sqlite3_context *ctx, - int i -){ - wholenumber_cursor *pCur = (wholenumber_cursor*)cur; - sqlite3_result_int64(ctx, pCur->iValue); - return SQLITE_OK; -} - -/* -** The rowid. -*/ -static int wholenumberRowid(sqlite3_vtab_cursor *cur, sqlite_int64 *pRowid){ - wholenumber_cursor *pCur = (wholenumber_cursor*)cur; - *pRowid = pCur->iValue; - return SQLITE_OK; -} - -/* -** When the wholenumber_cursor.rLimit value is 0 or less, that is a signal -** that the cursor has nothing more to output. -*/ -static int wholenumberEof(sqlite3_vtab_cursor *cur){ - wholenumber_cursor *pCur = (wholenumber_cursor*)cur; - return pCur->iValue>pCur->mxValue || pCur->iValue==0; -} - -/* -** Called to "rewind" a cursor back to the beginning so that -** it starts its output over again. Always called at least once -** prior to any wholenumberColumn, wholenumberRowid, or wholenumberEof call. -** -** idxNum Constraints -** ------ --------------------- -** 0 (none) -** 1 value > $argv0 -** 2 value >= $argv0 -** 4 value < $argv0 -** 8 value <= $argv0 -** -** 5 value > $argv0 AND value < $argv1 -** 6 value >= $argv0 AND value < $argv1 -** 9 value > $argv0 AND value <= $argv1 -** 10 value >= $argv0 AND value <= $argv1 -*/ -static int wholenumberFilter( - sqlite3_vtab_cursor *pVtabCursor, - int idxNum, const char *idxStr, - int argc, sqlite3_value **argv -){ - wholenumber_cursor *pCur = (wholenumber_cursor *)pVtabCursor; - sqlite3_int64 v; - int i = 0; - pCur->iValue = 1; - pCur->mxValue = 0xffffffff; /* 4294967295 */ - if( idxNum & 3 ){ - v = sqlite3_value_int64(argv[0]) + (idxNum&1); - if( v>pCur->iValue && v<=pCur->mxValue ) pCur->iValue = v; - i++; - } - if( idxNum & 12 ){ - v = sqlite3_value_int64(argv[i]) - ((idxNum>>2)&1); - if( v>=pCur->iValue && v<pCur->mxValue ) pCur->mxValue = v; - } - return SQLITE_OK; -} - -/* -** Search for terms of these forms: -** -** (1) value > $value -** (2) value >= $value -** (4) value < $value -** (8) value <= $value -** -** idxNum is an ORed combination of 1 or 2 with 4 or 8. -*/ -static int wholenumberBestIndex( - sqlite3_vtab *tab, - sqlite3_index_info *pIdxInfo -){ - int i; - int idxNum = 0; - int argvIdx = 1; - int ltIdx = -1; - int gtIdx = -1; - const struct sqlite3_index_constraint *pConstraint; - pConstraint = pIdxInfo->aConstraint; - for(i=0; i<pIdxInfo->nConstraint; i++, pConstraint++){ - if( pConstraint->usable==0 ) continue; - if( (idxNum & 3)==0 && pConstraint->op==SQLITE_INDEX_CONSTRAINT_GT ){ - idxNum |= 1; - ltIdx = i; - } - if( (idxNum & 3)==0 && pConstraint->op==SQLITE_INDEX_CONSTRAINT_GE ){ - idxNum |= 2; - ltIdx = i; - } - if( (idxNum & 12)==0 && pConstraint->op==SQLITE_INDEX_CONSTRAINT_LT ){ - idxNum |= 4; - gtIdx = i; - } - if( (idxNum & 12)==0 && pConstraint->op==SQLITE_INDEX_CONSTRAINT_LE ){ - idxNum |= 8; - gtIdx = i; - } - } - pIdxInfo->idxNum = idxNum; - if( ltIdx>=0 ){ - pIdxInfo->aConstraintUsage[ltIdx].argvIndex = argvIdx++; - pIdxInfo->aConstraintUsage[ltIdx].omit = 1; - } - if( gtIdx>=0 ){ - pIdxInfo->aConstraintUsage[gtIdx].argvIndex = argvIdx; - pIdxInfo->aConstraintUsage[gtIdx].omit = 1; - } - if( pIdxInfo->nOrderBy==1 - && pIdxInfo->aOrderBy[0].desc==0 - ){ - pIdxInfo->orderByConsumed = 1; - } - pIdxInfo->estimatedCost = (double)1; - return SQLITE_OK; -} - -/* -** A virtual table module that provides read-only access to a -** Tcl global variable namespace. -*/ -static sqlite3_module wholenumberModule = { - 0, /* iVersion */ - wholenumberConnect, - wholenumberConnect, - wholenumberBestIndex, - wholenumberDisconnect, - wholenumberDisconnect, - wholenumberOpen, /* xOpen - open a cursor */ - wholenumberClose, /* xClose - close a cursor */ - wholenumberFilter, /* xFilter - configure scan constraints */ - wholenumberNext, /* xNext - advance a cursor */ - wholenumberEof, /* xEof - check for end of scan */ - wholenumberColumn, /* xColumn - read data */ - wholenumberRowid, /* xRowid - read data */ - 0, /* xUpdate */ - 0, /* xBegin */ - 0, /* xSync */ - 0, /* xCommit */ - 0, /* xRollback */ - 0, /* xFindMethod */ - 0, /* xRename */ -}; - -#endif /* SQLITE_OMIT_VIRTUALTABLE */ - - -/* -** Register the wholenumber virtual table -*/ -int wholenumber_register(sqlite3 *db){ - int rc = SQLITE_OK; -#ifndef SQLITE_OMIT_VIRTUALTABLE - rc = sqlite3_create_module(db, "wholenumber", &wholenumberModule, 0); -#endif - return rc; -} - -#ifdef SQLITE_TEST -#include <tcl.h> -/* -** Decode a pointer to an sqlite3 object. -*/ -extern int getDbPointer(Tcl_Interp *interp, const char *zA, sqlite3 **ppDb); - -/* -** Register the echo virtual table module. -*/ -static int register_wholenumber_module( - ClientData clientData, /* Pointer to sqlite3_enable_XXX function */ - Tcl_Interp *interp, /* The TCL interpreter that invoked this command */ - int objc, /* Number of arguments */ - Tcl_Obj *CONST objv[] /* Command arguments */ -){ - sqlite3 *db; - if( objc!=2 ){ - Tcl_WrongNumArgs(interp, 1, objv, "DB"); - return TCL_ERROR; - } - if( getDbPointer(interp, Tcl_GetString(objv[1]), &db) ) return TCL_ERROR; - wholenumber_register(db); - return TCL_OK; -} - - -/* -** Register commands with the TCL interpreter. -*/ -int Sqlitetestwholenumber_Init(Tcl_Interp *interp){ - static struct { - char *zName; - Tcl_ObjCmdProc *xProc; - void *clientData; - } aObjCmd[] = { - { "register_wholenumber_module", register_wholenumber_module, 0 }, - }; - int i; - for(i=0; i<sizeof(aObjCmd)/sizeof(aObjCmd[0]); i++){ - Tcl_CreateObjCommand(interp, aObjCmd[i].zName, - aObjCmd[i].xProc, aObjCmd[i].clientData, 0); - } - return TCL_OK; -} - -#endif /* SQLITE_TEST */ diff --git a/src/vacuum.c b/src/vacuum.c index 7ed247832..4afb2cca6 100644 --- a/src/vacuum.c +++ b/src/vacuum.c @@ -289,6 +289,7 @@ int sqlite3RunVacuum(char **pzErrMsg, sqlite3 *db){ BTREE_DEFAULT_CACHE_SIZE, 0, /* Preserve the default page cache size */ BTREE_TEXT_ENCODING, 0, /* Preserve the text encoding */ BTREE_USER_VERSION, 0, /* Preserve the user version */ + BTREE_APPLICATION_ID, 0, /* Preserve the application id */ }; assert( 1==sqlite3BtreeIsInTrans(pTemp) ); @@ -43,6 +43,7 @@ # define sqlite3WalExclusiveMode(y,z) 0 # define sqlite3WalHeapMemory(z) 0 # define sqlite3WalFramesize(z) 0 +# define sqlite3WalFindFrame(x,y,z) 0 #else #define WAL_SAVEPOINT_NDATA 4 diff --git a/src/walker.c b/src/walker.c index eab96ea24..e71ed2ac4 100644 --- a/src/walker.c +++ b/src/walker.c @@ -113,7 +113,9 @@ int sqlite3WalkSelectFrom(Walker *pWalker, Select *p){ /* ** Call sqlite3WalkExpr() for every expression in Select statement p. ** Invoke sqlite3WalkSelect() for subqueries in the FROM clause and -** on the compound select chain, p->pPrior. +** on the compound select chain, p->pPrior. Invoke the xSelectCallback() +** either before or after the walk of expressions and FROM clause, depending +** on whether pWalker->bSelectDepthFirst is false or true, respectively. ** ** Return WRC_Continue under normal conditions. Return WRC_Abort if ** there is an abort request. @@ -127,14 +129,23 @@ int sqlite3WalkSelect(Walker *pWalker, Select *p){ rc = WRC_Continue; pWalker->walkerDepth++; while( p ){ - rc = pWalker->xSelectCallback(pWalker, p); - if( rc ) break; + if( !pWalker->bSelectDepthFirst ){ + rc = pWalker->xSelectCallback(pWalker, p); + if( rc ) break; + } if( sqlite3WalkSelectExpr(pWalker, p) || sqlite3WalkSelectFrom(pWalker, p) ){ pWalker->walkerDepth--; return WRC_Abort; } + if( pWalker->bSelectDepthFirst ){ + rc = pWalker->xSelectCallback(pWalker, p); + /* Depth-first search is currently only used for + ** selectAddSubqueryTypeInfo() and that routine always returns + ** WRC_Continue (0). So the following branch is never taken. */ + if( NEVER(rc) ) break; + } p = p->pPrior; } pWalker->walkerDepth--; diff --git a/src/where.c b/src/where.c index d70205205..2de894c3e 100644 --- a/src/where.c +++ b/src/where.c @@ -705,7 +705,7 @@ static WhereTerm *findTerm( continue; } } - if( pTerm->prereqRight==0 ){ + if( pTerm->prereqRight==0 && (pTerm->eOperator&WO_EQ)!=0 ){ pResult = pTerm; goto findTerm_success; }else if( pResult==0 ){ @@ -4883,6 +4883,7 @@ static Bitmask codeOneLoopStart( assert( (pTerm->prereqRight & newNotReady)!=0 ); pAlt = findTerm(pWC, iCur, pTerm->u.leftColumn, notReady, WO_EQ|WO_IN, 0); if( pAlt==0 ) continue; + if( pAlt->wtFlags & (TERM_CODED) ) continue; VdbeNoopComment((v, "begin transitive constraint")); sEq = *pAlt->pExpr; sEq.pLeft = pE->pLeft; |