/*
** Execute zSql on database db.
**
** If zSql returns rows, then each row will have exactly one
** column. (This will only happen if zSql begins with "SELECT".)
** Take each row of result and call execSql() again recursively.
**
** The execSqlF() routine does the same thing, except it accepts
** a format string as its third argument
*/
static int execSql(sqlite3 *db, char **pzErrMsg, const char *zSql){
sqlite3_stmt *pStmt;
int rc;
/* printf("SQL: [%s]\n", zSql); fflush(stdout); */
rc = sqlite3_prepare_v2(db, zSql, -1, &pStmt, 0);
if( rc!=SQLITE_OK ) return rc;
while( SQLITE_ROW==(rc = sqlite3_step(pStmt)) ){
const char *zSubSql = (const char*)sqlite3_column_text(pStmt,0);
assert( sqlite3_strnicmp(zSql,"SELECT",6)==0 );
/* The secondary SQL must be one of CREATE TABLE, CREATE INDEX,
** or INSERT. Historically there have been attacks that first
** corrupt the sqlite_master.sql field with other kinds of statements
** then run VACUUM to get those statements to execute at inappropriate
** times. */
if( zSubSql
&& (strncmp(zSubSql,"CRE",3)==0 || strncmp(zSubSql,"INS",3)==0)
){
rc = execSql(db, pzErrMsg, zSubSql);
if( rc!=SQLITE_OK ) break;
}
}
assert( rc!=SQLITE_ROW );
if( rc==SQLITE_DONE ) rc = SQLITE_OK;
if( rc ){
sqlite3SetString(pzErrMsg, db, sqlite3_errmsg(db));
}
(void)sqlite3_finalize(pStmt);
return rc;
}
/*
** Case insensitive string comparison.
*/
int fossil_strnicmp(const char *zA, const char *zB, int nByte){
if( zA==0 ){
if( zB==0 ) return 0;
return -1;
}else if( zB==0 ){
return +1;
}
if( nByte<0 ) nByte = strlen(zB);
return sqlite3_strnicmp(zA, zB, nByte);
}
/*
** This is the callback routine for the code that initializes the
** database. See sqlite3Init() below for additional information.
** This routine is also called from the OP_ParseSchema opcode of the VDBE.
**
** Each callback contains the following information:
**
** argv[0] = name of thing being created
** argv[1] = root page number for table or index. 0 for trigger or view.
** argv[2] = SQL text for the CREATE statement.
**
*/
int sqlite3InitCallback(void *pInit, int argc, char **argv, char **NotUsed){
InitData *pData = (InitData*)pInit;
sqlite3 *db = pData->db;
int iDb = pData->iDb;
assert( argc==3 );
UNUSED_PARAMETER2(NotUsed, argc);
assert( sqlite3_mutex_held(db->mutex) );
DbClearProperty(db, iDb, DB_Empty);
if( db->mallocFailed ){
corruptSchema(pData, argv[0], 0);
return 1;
}
assert( iDb>=0 && iDb<db->nDb );
if( argv==0 ) return 0; /* Might happen if EMPTY_RESULT_CALLBACKS are on */
if( argv[1]==0 ){
corruptSchema(pData, argv[0], 0);
}else if( sqlite3_strnicmp(argv[2],"create ",7)==0 ){
/* Call the parser to process a CREATE TABLE, INDEX or VIEW.
** But because db->init.busy is set to 1, no VDBE code is generated
** or executed. All the parser does is build the internal data
** structures that describe the table, index, or view.
*/
int rc;
u8 saved_iDb = db->init.iDb;
sqlite3_stmt *pStmt;
TESTONLY(int rcp); /* Return code from sqlite3_prepare() */
assert( db->init.busy );
db->init.iDb = iDb;
db->init.newTnum = sqlite3Atoi(argv[1]);
db->init.orphanTrigger = 0;
TESTONLY(rcp = ) sqlite3_prepare(db, argv[2], -1, &pStmt, 0);
rc = db->errCode;
assert( (rc&0xFF)==(rcp&0xFF) );
db->init.iDb = saved_iDb;
assert( saved_iDb==0 || (db->flags & SQLITE_Vacuum)!=0 );
if( SQLITE_OK!=rc ){
if( db->init.orphanTrigger ){
assert( iDb==1 );
}else{
pData->rc = rc;
if( rc==SQLITE_NOMEM ){
sqlite3OomFault(db);
}else if( rc!=SQLITE_INTERRUPT && (rc&0xFF)!=SQLITE_LOCKED ){
corruptSchema(pData, argv[0], sqlite3_errmsg(db));
}
}
}
sqlite3_finalize(pStmt);
}else if( argv[0]==0 || (argv[2]!=0 && argv[2][0]!=0) ){
int fossil_stricmp(const char *zA, const char *zB){
int nByte;
int rc;
if( zA==0 ){
if( zB==0 ) return 0;
return -1;
}else if( zB==0 ){
return +1;
}
nByte = strlen(zB);
rc = sqlite3_strnicmp(zA, zB, nByte);
if( rc==0 && zA[nByte] ) rc = 1;
return rc;
}
/*
** This function is used to determine if the main database file for
** connection db is open in WAL mode or not. If no error occurs and the
** database file is in WAL mode, set *pbWal to true and return SQLITE_OK.
** If it is not in WAL mode, set *pbWal to false.
**
** If an error occurs, return an SQLite error code. The value of *pbWal
** is undefined in this case.
*/
static int superlockIsWal(Superlock *pLock){
int rc; /* Return Code */
sqlite3_stmt *pStmt; /* Compiled PRAGMA journal_mode statement */
rc = sqlite3_prepare(pLock->db, "PRAGMA main.journal_mode", -1, &pStmt, 0);
if( rc!=SQLITE_OK ) return rc;
pLock->bWal = 0;
if( SQLITE_ROW==sqlite3_step(pStmt) ){
const char *zMode = (const char *)sqlite3_column_text(pStmt, 0);
if( zMode && strlen(zMode)==3 && sqlite3_strnicmp("wal", zMode, 3)==0 ){
pLock->bWal = 1;
}
}
return sqlite3_finalize(pStmt);
}
static int fts5ConfigSetEnum(
const Fts5Enum *aEnum,
const char *zEnum,
int *peVal
){
int nEnum = strlen(zEnum);
int i;
int iVal = -1;
for(i=0; aEnum[i].zName; i++){
if( sqlite3_strnicmp(aEnum[i].zName, zEnum, nEnum)==0 ){
if( iVal>=0 ) return SQLITE_ERROR;
iVal = aEnum[i].eVal;
}
}
*peVal = iVal;
return iVal<0 ? SQLITE_ERROR : SQLITE_OK;
}
/*
** Convert a symbolic name into a RID. Acceptable forms:
**
** * SHA1 hash
** * SHA1 hash prefix of at least 4 characters
** * Symbolic Name
** * "tag:" + symbolic name
** * Date or date-time
** * "date:" + Date or date-time
** * symbolic-name ":" date-time
** * "tip"
**
** The following additional forms are available in local checkouts:
**
** * "current"
** * "prev" or "previous"
** * "next"
**
** Return the RID of the matching artifact. Or return 0 if the name does not
** match any known object. Or return -1 if the name is ambiguous.
**
** The zType parameter specifies the type of artifact: ci, t, w, e, g.
** If zType is NULL or "" or "*" then any type of artifact will serve.
** If zType is "br" then find the first check-in of the named branch
** rather than the last.
** zType is "ci" in most use cases since we are usually searching for
** a check-in.
**
** Note that the input zTag for types "t" and "e" is the SHA1 hash of
** the ticket-change or event-change artifact, not the randomly generated
** hexadecimal identifier assigned to tickets and events. Those identifiers
** live in a separate namespace.
*/
int symbolic_name_to_rid(const char *zTag, const char *zType){
int vid;
int rid = 0;
int nTag;
int i;
int startOfBranch = 0;
if( zType==0 || zType[0]==0 ){
zType = "*";
}else if( zType[0]=='b' ){
zType = "ci";
startOfBranch = 1;
}
if( zTag==0 || zTag[0]==0 ) return 0;
/* special keyword: "tip" */
if( fossil_strcmp(zTag, "tip")==0 && (zType[0]=='*' || zType[0]=='c') ){
rid = db_int(0,
"SELECT objid"
" FROM event"
" WHERE type='ci'"
" ORDER BY event.mtime DESC"
);
if( rid ) return rid;
}
/* special keywords: "prev", "previous", "current", and "next" */
if( g.localOpen && (vid=db_lget_int("checkout",0))!=0 ){
if( fossil_strcmp(zTag, "current")==0 ){
rid = vid;
}else if( fossil_strcmp(zTag, "prev")==0
|| fossil_strcmp(zTag, "previous")==0 ){
rid = db_int(0, "SELECT pid FROM plink WHERE cid=%d AND isprim", vid);
}else if( fossil_strcmp(zTag, "next")==0 ){
rid = db_int(0, "SELECT cid FROM plink WHERE pid=%d"
" ORDER BY isprim DESC, mtime DESC", vid);
}
if( rid ) return rid;
}
/* Date and times */
if( memcmp(zTag, "date:", 5)==0 ){
rid = db_int(0,
"SELECT objid FROM event"
" WHERE mtime<=julianday(%Q,'utc') AND type GLOB '%q'"
" ORDER BY mtime DESC LIMIT 1",
&zTag[5], zType);
return rid;
}
if( fossil_isdate(zTag) ){
rid = db_int(0,
"SELECT objid FROM event"
" WHERE mtime<=julianday(%Q,'utc') AND type GLOB '%q'"
" ORDER BY mtime DESC LIMIT 1",
zTag, zType);
if( rid) return rid;
}
/* Deprecated date & time formats: "local:" + date-time and
** "utc:" + date-time */
if( memcmp(zTag, "local:", 6)==0 ){
rid = db_int(0,
"SELECT objid FROM event"
" WHERE mtime<=julianday(%Q) AND type GLOB '%q'"
" ORDER BY mtime DESC LIMIT 1",
&zTag[6], zType);
return rid;
}
if( memcmp(zTag, "utc:", 4)==0 ){
rid = db_int(0,
"SELECT objid FROM event"
" WHERE mtime<=julianday('%qz') AND type GLOB '%q'"
" ORDER BY mtime DESC LIMIT 1",
&zTag[4], zType);
return rid;
}
/* "tag:" + symbolic-name */
if( memcmp(zTag, "tag:", 4)==0 ){
rid = db_int(0,
"SELECT event.objid, max(event.mtime)"
" FROM tag, tagxref, event"
" WHERE tag.tagname='sym-%q' "
" AND tagxref.tagid=tag.tagid AND tagxref.tagtype>0 "
" AND event.objid=tagxref.rid "
" AND event.type GLOB '%q'",
&zTag[4], zType
);
if( startOfBranch ) rid = start_of_branch(rid,1);
return rid;
}
/* root:TAG -> The origin of the branch */
if( memcmp(zTag, "root:", 5)==0 ){
rid = symbolic_name_to_rid(zTag+5, zType);
return start_of_branch(rid, 0);
}
/* symbolic-name ":" date-time */
nTag = strlen(zTag);
for(i=0; i<nTag-10 && zTag[i]!=':'; i++){}
if( zTag[i]==':' && fossil_isdate(&zTag[i+1]) ){
char *zDate = mprintf("%s", &zTag[i+1]);
char *zTagBase = mprintf("%.*s", i, zTag);
int nDate = strlen(zDate);
if( sqlite3_strnicmp(&zDate[nDate-3],"utc",3)==0 ){
zDate[nDate-3] = 'z';
zDate[nDate-2] = 0;
}
rid = db_int(0,
"SELECT event.objid, max(event.mtime)"
" FROM tag, tagxref, event"
" WHERE tag.tagname='sym-%q' "
" AND tagxref.tagid=tag.tagid AND tagxref.tagtype>0 "
" AND event.objid=tagxref.rid "
" AND event.mtime<=julianday(%Q)"
" AND event.type GLOB '%q'",
zTagBase, zDate, zType
);
return rid;
}
/* SHA1 hash or prefix */
if( nTag>=4 && nTag<=UUID_SIZE && validate16(zTag, nTag) ){
Stmt q;
char zUuid[UUID_SIZE+1];
memcpy(zUuid, zTag, nTag+1);
canonical16(zUuid, nTag);
rid = 0;
if( zType[0]=='*' ){
db_prepare(&q, "SELECT rid FROM blob WHERE uuid GLOB '%q*'", zUuid);
}else{
db_prepare(&q,
"SELECT blob.rid"
" FROM blob, event"
" WHERE blob.uuid GLOB '%q*'"
" AND event.objid=blob.rid"
" AND event.type GLOB '%q'",
zUuid, zType
);
}
if( db_step(&q)==SQLITE_ROW ){
rid = db_column_int(&q, 0);
if( db_step(&q)==SQLITE_ROW ) rid = -1;
}
db_finalize(&q);
if( rid ) return rid;
}
/* Symbolic name */
rid = db_int(0,
"SELECT event.objid, max(event.mtime)"
" FROM tag, tagxref, event"
" WHERE tag.tagname='sym-%q' "
" AND tagxref.tagid=tag.tagid AND tagxref.tagtype>0 "
" AND event.objid=tagxref.rid "
" AND event.type GLOB '%q'",
zTag, zType
);
if( rid>0 ){
if( startOfBranch ) rid = start_of_branch(rid,1);
return rid;
}
/* Undocumented: numeric tags get translated directly into the RID */
if( memcmp(zTag, "rid:", 4)==0 ){
zTag += 4;
for(i=0; fossil_isdigit(zTag[i]); i++){}
if( zTag[i]==0 ){
if( strcmp(zType,"*")==0 ){
rid = atoi(zTag);
}else{
rid = db_int(0,
"SELECT event.objid"
" FROM event"
" WHERE event.objid=%s"
" AND event.type GLOB '%q'", zTag /*safe-for-%s*/, zType);
}
}
}
return rid;
}
/*
** Attempt to load an SQLite extension library contained in the file
** zFile. The entry point is zProc. zProc may be 0 in which case a
** default entry point name (sqlite3_extension_init) is used. Use
** of the default name is recommended.
**
** Return SQLITE_OK on success and SQLITE_ERROR if something goes wrong.
**
** If an error occurs and pzErrMsg is not 0, then fill *pzErrMsg with
** error message text. The calling function should free this memory
** by calling sqlite3DbFree(db, ).
*/
static int sqlite3LoadExtension(
sqlite3 *db, /* Load the extension into this database connection */
const char *zFile, /* Name of the shared library containing extension */
const char *zProc, /* Entry point. Use "sqlite3_extension_init" if 0 */
char **pzErrMsg /* Put error message here if not 0 */
){
sqlite3_vfs *pVfs = db->pVfs;
void *handle;
int (*xInit)(sqlite3*,char**,const sqlite3_api_routines*);
char *zErrmsg = 0;
const char *zEntry;
char *zAltEntry = 0;
void **aHandle;
int nMsg = 300 + sqlite3Strlen30(zFile);
int ii;
/* Shared library endings to try if zFile cannot be loaded as written */
static const char *azEndings[] = {
#if SQLITE_OS_WIN
"dll"
#elif defined(__APPLE__)
"dylib"
#else
"so"
#endif
};
if( pzErrMsg ) *pzErrMsg = 0;
/* Ticket #1863. To avoid a creating security problems for older
** applications that relink against newer versions of SQLite, the
** ability to run load_extension is turned off by default. One
** must call sqlite3_enable_load_extension() to turn on extension
** loading. Otherwise you get the following error.
*/
if( (db->flags & SQLITE_LoadExtension)==0 ){
if( pzErrMsg ){
*pzErrMsg = sqlite3_mprintf("not authorized");
}
return SQLITE_ERROR;
}
zEntry = zProc ? zProc : "sqlite3_extension_init";
handle = sqlite3OsDlOpen(pVfs, zFile);
#if SQLITE_OS_UNIX || SQLITE_OS_WIN
for(ii=0; ii<ArraySize(azEndings) && handle==0; ii++){
char *zAltFile = sqlite3_mprintf("%s.%s", zFile, azEndings[ii]);
if( zAltFile==0 ) return SQLITE_NOMEM;
handle = sqlite3OsDlOpen(pVfs, zAltFile);
sqlite3_free(zAltFile);
}
#endif
if( handle==0 ){
if( pzErrMsg ){
*pzErrMsg = zErrmsg = sqlite3_malloc(nMsg);
if( zErrmsg ){
sqlite3_snprintf(nMsg, zErrmsg,
"unable to open shared library [%s]", zFile);
sqlite3OsDlError(pVfs, nMsg-1, zErrmsg);
}
}
return SQLITE_ERROR;
}
xInit = (int(*)(sqlite3*,char**,const sqlite3_api_routines*))
sqlite3OsDlSym(pVfs, handle, zEntry);
/* If no entry point was specified and the default legacy
** entry point name "sqlite3_extension_init" was not found, then
** construct an entry point name "sqlite3_X_init" where the X is
** replaced by the lowercase value of every ASCII alphabetic
** character in the filename after the last "/" upto the first ".",
** and eliding the first three characters if they are "lib".
** Examples:
**
** /usr/local/lib/libExample5.4.3.so ==> sqlite3_example_init
** C:/lib/mathfuncs.dll ==> sqlite3_mathfuncs_init
*/
if( xInit==0 && zProc==0 ){
int iFile, iEntry, c;
int ncFile = sqlite3Strlen30(zFile);
zAltEntry = sqlite3_malloc(ncFile+30);
if( zAltEntry==0 ){
sqlite3OsDlClose(pVfs, handle);
return SQLITE_NOMEM;
}
memcpy(zAltEntry, "sqlite3_", 8);
for(iFile=ncFile-1; iFile>=0 && zFile[iFile]!='/'; iFile--){}
iFile++;
if( sqlite3_strnicmp(zFile+iFile, "lib", 3)==0 ) iFile += 3;
for(iEntry=8; (c = zFile[iFile])!=0 && c!='.'; iFile++){
if( sqlite3Isalpha(c) ){
zAltEntry[iEntry++] = (char)sqlite3UpperToLower[(unsigned)c];
}
}
memcpy(zAltEntry+iEntry, "_init", 6);
zEntry = zAltEntry;
xInit = (int(*)(sqlite3*,char**,const sqlite3_api_routines*))
sqlite3OsDlSym(pVfs, handle, zEntry);
}
if( xInit==0 ){
if( pzErrMsg ){
nMsg += sqlite3Strlen30(zEntry);
*pzErrMsg = zErrmsg = sqlite3_malloc(nMsg);
if( zErrmsg ){
sqlite3_snprintf(nMsg, zErrmsg,
"no entry point [%s] in shared library [%s]", zEntry, zFile);
sqlite3OsDlError(pVfs, nMsg-1, zErrmsg);
}
}
sqlite3OsDlClose(pVfs, handle);
sqlite3_free(zAltEntry);
return SQLITE_ERROR;
}
sqlite3_free(zAltEntry);
if( xInit(db, &zErrmsg, &sqlite3Apis) ){
if( pzErrMsg ){
*pzErrMsg = sqlite3_mprintf("error during initialization: %s", zErrmsg);
}
sqlite3_free(zErrmsg);
sqlite3OsDlClose(pVfs, handle);
return SQLITE_ERROR;
}
/* Append the new shared library handle to the db->aExtension array. */
aHandle = sqlite3DbMallocZero(db, sizeof(handle)*(db->nExtension+1));
if( aHandle==0 ){
return SQLITE_NOMEM;
}
if( db->nExtension>0 ){
memcpy(aHandle, db->aExtension, sizeof(handle)*db->nExtension);
}
sqlite3DbFree(db, db->aExtension);
db->aExtension = aHandle;
db->aExtension[db->nExtension++] = handle;
return SQLITE_OK;
}
static const char *fts5ConfigSkipLiteral(const char *pIn){
const char *p = pIn;
switch( *p ){
case 'n': case 'N':
if( sqlite3_strnicmp("null", p, 4)==0 ){
p = &p[4];
}else{
p = 0;
}
break;
case 'x': case 'X':
p++;
if( *p=='\'' ){
p++;
while( (*p>='a' && *p<='f')
|| (*p>='A' && *p<='F')
|| (*p>='0' && *p<='9')
){
p++;
}
if( *p=='\'' && 0==((p-pIn)%2) ){
p++;
}else{
p = 0;
}
}else{
p = 0;
}
break;
case '\'':
p++;
while( p ){
if( *p=='\'' ){
p++;
if( *p!='\'' ) break;
}
p++;
if( *p==0 ) p = 0;
}
break;
default:
/* maybe a number */
if( *p=='+' || *p=='-' ) p++;
while( fts5_isdigit(*p) ) p++;
/* At this point, if the literal was an integer, the parse is
** finished. Or, if it is a floating point value, it may continue
** with either a decimal point or an 'E' character. */
if( *p=='.' && fts5_isdigit(p[1]) ){
p += 2;
while( fts5_isdigit(*p) ) p++;
}
if( p==pIn ) p = 0;
break;
}
return p;
}
/*
** Parse a "special" CREATE VIRTUAL TABLE directive and update
** configuration object pConfig as appropriate.
**
** If successful, object pConfig is updated and SQLITE_OK returned. If
** an error occurs, an SQLite error code is returned and an error message
** may be left in *pzErr. It is the responsibility of the caller to
** eventually free any such error message using sqlite3_free().
*/
static int fts5ConfigParseSpecial(
Fts5Global *pGlobal,
Fts5Config *pConfig, /* Configuration object to update */
const char *zCmd, /* Special command to parse */
const char *zArg, /* Argument to parse */
char **pzErr /* OUT: Error message */
){
int rc = SQLITE_OK;
int nCmd = (int)strlen(zCmd);
if( sqlite3_strnicmp("prefix", zCmd, nCmd)==0 ){
const int nByte = sizeof(int) * FTS5_MAX_PREFIX_INDEXES;
const char *p;
int bFirst = 1;
if( pConfig->aPrefix==0 ){
pConfig->aPrefix = sqlite3Fts5MallocZero(&rc, nByte);
if( rc ) return rc;
}
p = zArg;
while( 1 ){
int nPre = 0;
while( p[0]==' ' ) p++;
if( bFirst==0 && p[0]==',' ){
p++;
while( p[0]==' ' ) p++;
}else if( p[0]=='\0' ){
break;
}
if( p[0]<'0' || p[0]>'9' ){
*pzErr = sqlite3_mprintf("malformed prefix=... directive");
rc = SQLITE_ERROR;
break;
}
if( pConfig->nPrefix==FTS5_MAX_PREFIX_INDEXES ){
*pzErr = sqlite3_mprintf(
"too many prefix indexes (max %d)", FTS5_MAX_PREFIX_INDEXES
);
rc = SQLITE_ERROR;
break;
}
while( p[0]>='0' && p[0]<='9' && nPre<1000 ){
nPre = nPre*10 + (p[0] - '0');
p++;
}
if( rc==SQLITE_OK && (nPre<=0 || nPre>=1000) ){
*pzErr = sqlite3_mprintf("prefix length out of range (max 999)");
rc = SQLITE_ERROR;
break;
}
pConfig->aPrefix[pConfig->nPrefix] = nPre;
pConfig->nPrefix++;
bFirst = 0;
}
assert( pConfig->nPrefix<=FTS5_MAX_PREFIX_INDEXES );
return rc;
}
if( sqlite3_strnicmp("tokenize", zCmd, nCmd)==0 ){
const char *p = (const char*)zArg;
int nArg = (int)strlen(zArg) + 1;
char **azArg = sqlite3Fts5MallocZero(&rc, sizeof(char*) * nArg);
char *pDel = sqlite3Fts5MallocZero(&rc, nArg * 2);
char *pSpace = pDel;
if( azArg && pSpace ){
if( pConfig->pTok ){
*pzErr = sqlite3_mprintf("multiple tokenize=... directives");
rc = SQLITE_ERROR;
}else{
for(nArg=0; p && *p; nArg++){
const char *p2 = fts5ConfigSkipWhitespace(p);
if( *p2=='\'' ){
p = fts5ConfigSkipLiteral(p2);
}else{
p = fts5ConfigSkipBareword(p2);
}
if( p ){
memcpy(pSpace, p2, p-p2);
azArg[nArg] = pSpace;
sqlite3Fts5Dequote(pSpace);
pSpace += (p - p2) + 1;
p = fts5ConfigSkipWhitespace(p);
}
}
if( p==0 ){
*pzErr = sqlite3_mprintf("parse error in tokenize directive");
rc = SQLITE_ERROR;
}else{
rc = sqlite3Fts5GetTokenizer(pGlobal,
(const char**)azArg, nArg, &pConfig->pTok, &pConfig->pTokApi,
pzErr
);
}
}
}
sqlite3_free(azArg);
sqlite3_free(pDel);
return rc;
}
if( sqlite3_strnicmp("content", zCmd, nCmd)==0 ){
if( pConfig->eContent!=FTS5_CONTENT_NORMAL ){
*pzErr = sqlite3_mprintf("multiple content=... directives");
rc = SQLITE_ERROR;
}else{
if( zArg[0] ){
pConfig->eContent = FTS5_CONTENT_EXTERNAL;
pConfig->zContent = sqlite3Fts5Mprintf(&rc, "%Q.%Q", pConfig->zDb,zArg);
}else{
pConfig->eContent = FTS5_CONTENT_NONE;
}
}
return rc;
}
if( sqlite3_strnicmp("content_rowid", zCmd, nCmd)==0 ){
if( pConfig->zContentRowid ){
*pzErr = sqlite3_mprintf("multiple content_rowid=... directives");
rc = SQLITE_ERROR;
}else{
pConfig->zContentRowid = sqlite3Fts5Strndup(&rc, zArg, -1);
}
return rc;
}
if( sqlite3_strnicmp("columnsize", zCmd, nCmd)==0 ){
if( (zArg[0]!='0' && zArg[0]!='1') || zArg[1]!='\0' ){
*pzErr = sqlite3_mprintf("malformed columnsize=... directive");
rc = SQLITE_ERROR;
}else{
pConfig->bColumnsize = (zArg[0]=='1');
}
return rc;
}
if( sqlite3_strnicmp("detail", zCmd, nCmd)==0 ){
const Fts5Enum aDetail[] = {
{ "none", FTS5_DETAIL_NONE },
{ "full", FTS5_DETAIL_FULL },
{ "columns", FTS5_DETAIL_COLUMNS },
{ 0, 0 }
};
if( (rc = fts5ConfigSetEnum(aDetail, zArg, &pConfig->eDetail)) ){
*pzErr = sqlite3_mprintf("malformed detail=... directive");
}
return rc;
}
*pzErr = sqlite3_mprintf("unrecognized option: \"%.*s\"", nCmd, zCmd);
return SQLITE_ERROR;
}
int sqlite3Fts3InitTokenizer(
Fts3Hash *pHash, /* Tokenizer hash table */
const char *zArg, /* Possible tokenizer specification */
sqlite3_tokenizer **ppTok, /* OUT: Tokenizer (if applicable) */
const char **pzTokenizer, /* OUT: Set to zArg if is tokenizer */
char **pzErr /* OUT: Set to malloced error message */
){
int rc;
char *z = (char *)zArg;
int n;
char *zCopy;
char *zEnd; /* Pointer to nul-term of zCopy */
sqlite3_tokenizer_module *m;
if( !z ){
zCopy = sqlite3_mprintf("simple");
}else{
if( sqlite3_strnicmp(z, "tokenize", 8) || fts3IsIdChar(z[8])){
return SQLITE_OK;
}
zCopy = sqlite3_mprintf("%s", &z[8]);
*pzTokenizer = zArg;
}
if( !zCopy ){
return SQLITE_NOMEM;
}
zEnd = &zCopy[strlen(zCopy)];
z = (char *)sqlite3Fts3NextToken(zCopy, &n);
z[n] = '\0';
sqlite3Fts3Dequote(z);
m = (sqlite3_tokenizer_module *)sqlite3Fts3HashFind(pHash, z, (int)strlen(z)+1);
if( !m ){
*pzErr = sqlite3_mprintf("unknown tokenizer: %s", z);
rc = SQLITE_ERROR;
}else{
char const **aArg = 0;
int iArg = 0;
z = &z[n+1];
while( z<zEnd && (NULL!=(z = (char *)sqlite3Fts3NextToken(z, &n))) ){
int nNew = sizeof(char *)*(iArg+1);
char const **aNew = (const char **)sqlite3_realloc((void *)aArg, nNew);
if( !aNew ){
sqlite3_free(zCopy);
sqlite3_free((void *)aArg);
return SQLITE_NOMEM;
}
aArg = aNew;
aArg[iArg++] = z;
z[n] = '\0';
sqlite3Fts3Dequote(z);
z = &z[n+1];
}
rc = m->xCreate(iArg, aArg, ppTok);
assert( rc!=SQLITE_OK || *ppTok );
if( rc!=SQLITE_OK ){
*pzErr = sqlite3_mprintf("unknown tokenizer");
}else{
(*ppTok)->pModule = m;
}
sqlite3_free((void *)aArg);
}
sqlite3_free(zCopy);
return rc;
}
/*
** The output variable *ppExpr is populated with an allocated Fts3Expr
** structure, or set to 0 if the end of the input buffer is reached.
**
** Returns an SQLite error code. SQLITE_OK if everything works, SQLITE_NOMEM
** if a malloc failure occurs, or SQLITE_ERROR if a parse error is encountered.
** If SQLITE_ERROR is returned, pContext is populated with an error message.
*/
static int getNextNode(
ParseContext *pParse, /* fts3 query parse context */
const char *z, int n, /* Input string */
Fts3Expr **ppExpr, /* OUT: expression */
int *pnConsumed /* OUT: Number of bytes consumed */
){
static const struct Fts3Keyword {
char *z; /* Keyword text */
unsigned char n; /* Length of the keyword */
unsigned char parenOnly; /* Only valid in paren mode */
unsigned char eType; /* Keyword code */
} aKeyword[] = {
{ "OR" , 2, 0, FTSQUERY_OR },
{ "AND", 3, 1, FTSQUERY_AND },
{ "NOT", 3, 1, FTSQUERY_NOT },
{ "NEAR", 4, 0, FTSQUERY_NEAR }
};
int ii;
int iCol;
int iColLen;
int rc;
Fts3Expr *pRet = 0;
const char *zInput = z;
int nInput = n;
/* Skip over any whitespace before checking for a keyword, an open or
** close bracket, or a quoted string.
*/
while( nInput>0 && fts3isspace(*zInput) ){
nInput--;
zInput++;
}
if( nInput==0 ){
return SQLITE_DONE;
}
/* See if we are dealing with a keyword. */
for(ii=0; ii<(int)(sizeof(aKeyword)/sizeof(struct Fts3Keyword)); ii++){
const struct Fts3Keyword *pKey = &aKeyword[ii];
if( (pKey->parenOnly & ~sqlite3_fts3_enable_parentheses)!=0 ){
continue;
}
if( nInput>=pKey->n && 0==memcmp(zInput, pKey->z, pKey->n) ){
int nNear = SQLITE_FTS3_DEFAULT_NEAR_PARAM;
int nKey = pKey->n;
char cNext;
/* If this is a "NEAR" keyword, check for an explicit nearness. */
if( pKey->eType==FTSQUERY_NEAR ){
assert( nKey==4 );
if( zInput[4]=='/' && zInput[5]>='0' && zInput[5]<='9' ){
nNear = 0;
for(nKey=5; zInput[nKey]>='0' && zInput[nKey]<='9'; nKey++){
nNear = nNear * 10 + (zInput[nKey] - '0');
}
}
}
/* At this point this is probably a keyword. But for that to be true,
** the next byte must contain either whitespace, an open or close
** parenthesis, a quote character, or EOF.
*/
cNext = zInput[nKey];
if( fts3isspace(cNext)
|| cNext=='"' || cNext=='(' || cNext==')' || cNext==0
){
pRet = (Fts3Expr *)fts3MallocZero(sizeof(Fts3Expr));
if( !pRet ){
return SQLITE_NOMEM;
}
pRet->eType = pKey->eType;
pRet->nNear = nNear;
*ppExpr = pRet;
*pnConsumed = (int)((zInput - z) + nKey);
return SQLITE_OK;
}
/* Turns out that wasn't a keyword after all. This happens if the
** user has supplied a token such as "ORacle". Continue.
*/
}
}
/* Check for an open bracket. */
if( sqlite3_fts3_enable_parentheses ){
if( *zInput=='(' ){
int nConsumed;
pParse->nNest++;
rc = fts3ExprParse(pParse, &zInput[1], nInput-1, ppExpr, &nConsumed);
if( rc==SQLITE_OK && !*ppExpr ){
rc = SQLITE_DONE;
}
*pnConsumed = (int)((zInput - z) + 1 + nConsumed);
return rc;
}
/* Check for a close bracket. */
if( *zInput==')' ){
pParse->nNest--;
*pnConsumed = (int)((zInput - z) + 1);
return SQLITE_DONE;
}
}
/* See if we are dealing with a quoted phrase. If this is the case, then
** search for the closing quote and pass the whole string to getNextString()
** for processing. This is easy to do, as fts3 has no syntax for escaping
** a quote character embedded in a string.
*/
if( *zInput=='"' ){
for(ii=1; ii<nInput && zInput[ii]!='"'; ii++);
*pnConsumed = (int)((zInput - z) + ii + 1);
if( ii==nInput ){
return SQLITE_ERROR;
}
return getNextString(pParse, &zInput[1], ii-1, ppExpr);
}
/* If control flows to this point, this must be a regular token, or
** the end of the input. Read a regular token using the sqlite3_tokenizer
** interface. Before doing so, figure out if there is an explicit
** column specifier for the token.
**
** TODO: Strangely, it is not possible to associate a column specifier
** with a quoted phrase, only with a single token. Not sure if this was
** an implementation artifact or an intentional decision when fts3 was
** first implemented. Whichever it was, this module duplicates the
** limitation.
*/
iCol = pParse->iDefaultCol;
iColLen = 0;
for(ii=0; ii<pParse->nCol; ii++){
const char *zStr = pParse->azCol[ii];
int nStr = (int)strlen(zStr);
if( nInput>nStr && zInput[nStr]==':'
&& sqlite3_strnicmp(zStr, zInput, nStr)==0
){
iCol = ii;
iColLen = (int)((zInput - z) + nStr + 1);
break;
}
}
rc = getNextToken(pParse, iCol, &z[iColLen], n-iColLen, ppExpr, pnConsumed);
*pnConsumed += iColLen;
return rc;
}
static void parse_command_line(
int argc,
char **argv,
int iStart,
CmdLineOption *aOpt,
void *pStruct,
const char *zHelp
){
char *pOut = (char*)pStruct;
int bVerbose = 0;
int iArg;
for(iArg=iStart; iArg<argc; iArg++){
const char *zArg = argv[iArg];
int nArg = strlen(zArg);
int nMatch = 0;
int iOpt;
for(iOpt=0; aOpt[iOpt].zText; iOpt++){
CmdLineOption *pOpt = &aOpt[iOpt];
if( 0==sqlite3_strnicmp(pOpt->zText, zArg, nArg) ){
if( nMatch ){
ambiguous_option_error(zArg);
}
nMatch++;
if( pOpt->eType==CMDLINE_BARE ){
*(int*)(&pOut[pOpt->iOff]) = 1;
}else{
iArg++;
if( iArg==argc ){
option_requires_argument_error(pOpt);
}
switch( pOpt->eType ){
case CMDLINE_INTEGER:
*(int*)(&pOut[pOpt->iOff]) = get_integer_option(pOpt, argv[iArg]);
break;
case CMDLINE_STRING:
*(const char**)(&pOut[pOpt->iOff]) = argv[iArg];
break;
case CMDLINE_BOOLEAN:
*(int*)(&pOut[pOpt->iOff]) = get_boolean_option(pOpt, argv[iArg]);
break;
}
}
}
}
if( nMatch==0 && 0==sqlite3_strnicmp("-cmdline:verbose", zArg, nArg) ){
bVerbose = 1;
nMatch = 1;
}
if( nMatch==0 ){
unknown_option_error(zArg, aOpt, zHelp);
}
}
if( bVerbose ){
int iOpt;
fprintf(stdout, "Options are: ");
for(iOpt=0; aOpt[iOpt].zText; iOpt++){
CmdLineOption *pOpt = &aOpt[iOpt];
if( pOpt->eType!=CMDLINE_BARE || *(int*)(&pOut[pOpt->iOff]) ){
fprintf(stdout, "%s ", pOpt->zText);
}
switch( pOpt->eType ){
case CMDLINE_INTEGER:
fprintf(stdout, "%d ", *(int*)(&pOut[pOpt->iOff]));
break;
case CMDLINE_BOOLEAN:
fprintf(stdout, "%d ", *(int*)(&pOut[pOpt->iOff]));
break;
case CMDLINE_STRING:
fprintf(stdout, "%s ", *(const char**)(&pOut[pOpt->iOff]));
break;
}
}
fprintf(stdout, "\n");
}
}
int main(int argc, char **argv){
const char *zDb;
int rc = 0;
char *zErr = 0;
int i;
int iVerbose = 1; /* -verbose option */
sqlite3 *db = 0;
sqlite3expert *p = 0;
if( argc<2 ) usage(argv);
zDb = argv[argc-1];
if( zDb[0]=='-' ) usage(argv);
rc = sqlite3_open(zDb, &db);
if( rc!=SQLITE_OK ){
fprintf(stderr, "Cannot open db file: %s - %s\n", zDb, sqlite3_errmsg(db));
exit(-2);
}
p = sqlite3_expert_new(db, &zErr);
if( p==0 ){
fprintf(stderr, "Cannot run analysis: %s\n", zErr);
rc = 1;
}else{
for(i=1; i<(argc-1); i++){
char *zArg = argv[i];
int nArg;
if( zArg[0]=='-' && zArg[1]=='-' && zArg[2]!=0 ) zArg++;
nArg = (int)strlen(zArg);
if( nArg>=2 && 0==sqlite3_strnicmp(zArg, "-file", nArg) ){
if( ++i==(argc-1) ) option_requires_argument("-file");
rc = readSqlFromFile(p, argv[i], &zErr);
}
else if( nArg>=3 && 0==sqlite3_strnicmp(zArg, "-sql", nArg) ){
if( ++i==(argc-1) ) option_requires_argument("-sql");
rc = sqlite3_expert_sql(p, argv[i], &zErr);
}
else if( nArg>=3 && 0==sqlite3_strnicmp(zArg, "-sample", nArg) ){
int iSample;
if( ++i==(argc-1) ) option_requires_argument("-sample");
iSample = option_integer_arg(argv[i]);
sqlite3_expert_config(p, EXPERT_CONFIG_SAMPLE, iSample);
}
else if( nArg>=2 && 0==sqlite3_strnicmp(zArg, "-verbose", nArg) ){
if( ++i==(argc-1) ) option_requires_argument("-verbose");
iVerbose = option_integer_arg(argv[i]);
}
else{
usage(argv);
}
}
}
if( rc==SQLITE_OK ){
rc = sqlite3_expert_analyze(p, &zErr);
}
if( rc==SQLITE_OK ){
int nQuery = sqlite3_expert_count(p);
if( iVerbose>0 ){
const char *zCand = sqlite3_expert_report(p,0,EXPERT_REPORT_CANDIDATES);
fprintf(stdout, "-- Candidates -------------------------------\n");
fprintf(stdout, "%s\n", zCand);
}
for(i=0; i<nQuery; i++){
const char *zSql = sqlite3_expert_report(p, i, EXPERT_REPORT_SQL);
const char *zIdx = sqlite3_expert_report(p, i, EXPERT_REPORT_INDEXES);
const char *zEQP = sqlite3_expert_report(p, i, EXPERT_REPORT_PLAN);
if( zIdx==0 ) zIdx = "(no new indexes)\n";
if( iVerbose>0 ){
fprintf(stdout, "-- Query %d ----------------------------------\n",i+1);
fprintf(stdout, "%s\n\n", zSql);
}
fprintf(stdout, "%s\n%s\n", zIdx, zEQP);
}
}else{
fprintf(stderr, "Error: %s\n", zErr ? zErr : "?");
}
sqlite3_expert_destroy(p);
sqlite3_free(zErr);
return rc;
}
