diff --git a/README.md b/README.md index ebb917e218..0975a32d9a 100644 --- a/README.md +++ b/README.md @@ -159,7 +159,7 @@ extension and only later escaped to the wild as an independent library.) Test scripts and programs are found in the **test/** subdirectory. Additional test code is found in other source repositories. -See [How SQLite Is Tested](http://www.sqlite.org/testing.html) for +See [How SQLite Is Tested](https://www.sqlite.org/testing.html) for additional information. The **ext/** subdirectory contains code for extensions. The @@ -183,7 +183,7 @@ manually-edited files and automatically-generated files. The SQLite interface is defined by the **sqlite3.h** header file, which is generated from src/sqlite.h.in, ./manifest.uuid, and ./VERSION. The -[Tcl script](http://www.tcl.tk) at tool/mksqlite3h.tcl does the conversion. +[Tcl script](https://www.tcl.tk) at tool/mksqlite3h.tcl does the conversion. The manifest.uuid file contains the SHA3 hash of the particular check-in and is used to generate the SQLITE\_SOURCE\_ID macro. The VERSION file contains the current SQLite version number. The sqlite3.h header is really @@ -250,14 +250,14 @@ individual source file exceeds 32K lines in length. ## How It All Fits Together SQLite is modular in design. -See the [architectural description](http://www.sqlite.org/arch.html) +See the [architectural description](https://www.sqlite.org/arch.html) for details. Other documents that are useful in (helping to understand how SQLite works include the -[file format](http://www.sqlite.org/fileformat2.html) description, -the [virtual machine](http://www.sqlite.org/opcode.html) that runs +[file format](https://www.sqlite.org/fileformat2.html) description, +the [virtual machine](https://www.sqlite.org/opcode.html) that runs prepared statements, the description of -[how transactions work](http://www.sqlite.org/atomiccommit.html), and -the [overview of the query planner](http://www.sqlite.org/optoverview.html). +[how transactions work](https://www.sqlite.org/atomiccommit.html), and +the [overview of the query planner](https://www.sqlite.org/optoverview.html). Years of effort have gone into optimizing SQLite, both for small size and high performance. And optimizations tend to result in @@ -353,7 +353,7 @@ hidden by also modifying the makefiles. ## Contacts -The main SQLite website is [http:/sqlite.org/](http://sqlite.org/) +The main SQLite website is [https://sqlite.org/](https://sqlite.org/) with geographically distributed backups at -[http://www2.sqlite.org/](http://www2.sqlite.org) and -[http://www3.sqlite.org/](http://www3.sqlite.org). +[https://www2.sqlite.org/](https://www2.sqlite.org) and +[https://www3.sqlite.org/](https://www3.sqlite.org). diff --git a/ext/fts5/extract_api_docs.tcl b/ext/fts5/extract_api_docs.tcl index 2320d70b7d..6762a036d5 100644 --- a/ext/fts5/extract_api_docs.tcl +++ b/ext/fts5/extract_api_docs.tcl @@ -223,10 +223,12 @@ proc main {data} { Fts5ExtensionApi { set struct [get_fts5_struct $data "^struct Fts5ExtensionApi" "^.;"] set map [list] + set lKey [list] foreach {k v} [get_struct_members $data] { if {[string match x* $k]==0} continue - lappend map $k "$k" + lappend lKey $k } + foreach k [lsort -decr $lKey] { lappend map $k "$k" } output [string map $map $struct] } diff --git a/ext/fts5/fts5.h b/ext/fts5/fts5.h index 323d73a28f..c7a5e6b7de 100644 --- a/ext/fts5/fts5.h +++ b/ext/fts5/fts5.h @@ -261,9 +261,34 @@ struct Fts5PhraseIter { ** ** xPhraseNextColumn() ** See xPhraseFirstColumn above. +** +** xQueryToken(pFts5, iPhrase, iToken, ppToken, pnToken) +** This is used to access token iToken of phrase iPhrase of the current +** query. Before returning, output parameter *ppToken is set to point +** to a buffer containing the requested token, and *pnToken to the +** size of this buffer in bytes. +** +** The output text is not a copy of the query text that specified the +** token. It is the output of the tokenizer module. For tokendata=1 +** tables, this includes any embedded 0x00 and trailing data. +** +** xInstToken(pFts5, iIdx, iToken, ppToken, pnToken) +** This is used to access token iToken of phrase hit iIdx within the +** current row. Output variable (*ppToken) is set to point to a buffer +** containing the matching document token, and (*pnToken) to the size +** of that buffer in bytes. This API is not available if the specified +** token matches a prefix query term. In that case both output variables +** are always set to 0. +** +** The output text is not a copy of the document text that was tokenized. +** It is the output of the tokenizer module. For tokendata=1 tables, this +** includes any embedded 0x00 and trailing data. +** +** This API can be quite slow if used with an FTS5 table created with the +** "detail=none" or "detail=column" option. */ struct Fts5ExtensionApi { - int iVersion; /* Currently always set to 2 */ + int iVersion; /* Currently always set to 3 */ void *(*xUserData)(Fts5Context*); @@ -298,6 +323,13 @@ struct Fts5ExtensionApi { int (*xPhraseFirstColumn)(Fts5Context*, int iPhrase, Fts5PhraseIter*, int*); void (*xPhraseNextColumn)(Fts5Context*, Fts5PhraseIter*, int *piCol); + + /* Below this point are iVersion>=3 only */ + int (*xQueryToken)(Fts5Context*, + int iPhrase, int iToken, + const char **ppToken, int *pnToken + ); + int (*xInstToken)(Fts5Context*, int iIdx, int iToken, const char**, int*); }; /* diff --git a/ext/fts5/fts5Int.h b/ext/fts5/fts5Int.h index 8bbafbaaf4..9beb26e056 100644 --- a/ext/fts5/fts5Int.h +++ b/ext/fts5/fts5Int.h @@ -196,6 +196,7 @@ struct Fts5Config { char *zContent; /* content table */ char *zContentRowid; /* "content_rowid=" option value */ int bColumnsize; /* "columnsize=" option value (dflt==1) */ + int bTokendata; /* "tokendata=" option value (dflt==0) */ int eDetail; /* FTS5_DETAIL_XXX value */ char *zContentExprlist; Fts5Tokenizer *pTok; @@ -384,17 +385,19 @@ struct Fts5IndexIter { /* ** Values used as part of the flags argument passed to IndexQuery(). */ -#define FTS5INDEX_QUERY_PREFIX 0x0001 /* Prefix query */ -#define FTS5INDEX_QUERY_DESC 0x0002 /* Docs in descending rowid order */ -#define FTS5INDEX_QUERY_TEST_NOIDX 0x0004 /* Do not use prefix index */ -#define FTS5INDEX_QUERY_SCAN 0x0008 /* Scan query (fts5vocab) */ +#define FTS5INDEX_QUERY_PREFIX 0x0001 /* Prefix query */ +#define FTS5INDEX_QUERY_DESC 0x0002 /* Docs in descending rowid order */ +#define FTS5INDEX_QUERY_TEST_NOIDX 0x0004 /* Do not use prefix index */ +#define FTS5INDEX_QUERY_SCAN 0x0008 /* Scan query (fts5vocab) */ /* The following are used internally by the fts5_index.c module. They are ** defined here only to make it easier to avoid clashes with the flags ** above. */ -#define FTS5INDEX_QUERY_SKIPEMPTY 0x0010 -#define FTS5INDEX_QUERY_NOOUTPUT 0x0020 -#define FTS5INDEX_QUERY_SKIPHASH 0x0040 +#define FTS5INDEX_QUERY_SKIPEMPTY 0x0010 +#define FTS5INDEX_QUERY_NOOUTPUT 0x0020 +#define FTS5INDEX_QUERY_SKIPHASH 0x0040 +#define FTS5INDEX_QUERY_NOTOKENDATA 0x0080 +#define FTS5INDEX_QUERY_SCANONETERM 0x0100 /* ** Create/destroy an Fts5Index object. @@ -463,6 +466,10 @@ void *sqlite3Fts5StructureRef(Fts5Index*); void sqlite3Fts5StructureRelease(void*); int sqlite3Fts5StructureTest(Fts5Index*, void*); +/* +** Used by xInstToken(): +*/ +int sqlite3Fts5IterToken(Fts5IndexIter*, i64, int, int, const char**, int*); /* ** Insert or remove data to or from the index. Each time a document is @@ -540,6 +547,13 @@ int sqlite3Fts5IndexLoadConfig(Fts5Index *p); int sqlite3Fts5IndexGetOrigin(Fts5Index *p, i64 *piOrigin); int sqlite3Fts5IndexContentlessDelete(Fts5Index *p, i64 iOrigin, i64 iRowid); +void sqlite3Fts5IndexIterClearTokendata(Fts5IndexIter*); + +/* Used to populate hash tables for xInstToken in detail=none/column mode. */ +int sqlite3Fts5IndexIterWriteTokendata( + Fts5IndexIter*, const char*, int, i64 iRowid, int iCol, int iOff +); + /* ** End of interface to code in fts5_index.c. **************************************************************************/ @@ -645,6 +659,7 @@ void sqlite3Fts5HashScanNext(Fts5Hash*); int sqlite3Fts5HashScanEof(Fts5Hash*); void sqlite3Fts5HashScanEntry(Fts5Hash *, const char **pzTerm, /* OUT: term (nul-terminated) */ + int *pnTerm, /* OUT: Size of term in bytes */ const u8 **ppDoclist, /* OUT: pointer to doclist */ int *pnDoclist /* OUT: size of doclist in bytes */ ); @@ -771,6 +786,10 @@ int sqlite3Fts5ExprClonePhrase(Fts5Expr*, int, Fts5Expr**); int sqlite3Fts5ExprPhraseCollist(Fts5Expr *, int, const u8 **, int *); +int sqlite3Fts5ExprQueryToken(Fts5Expr*, int, int, const char**, int*); +int sqlite3Fts5ExprInstToken(Fts5Expr*, i64, int, int, int, int, const char**, int*); +void sqlite3Fts5ExprClearTokens(Fts5Expr*); + /******************************************* ** The fts5_expr.c API above this point is used by the other hand-written ** C code in this module. The interfaces below this point are called by diff --git a/ext/fts5/fts5_config.c b/ext/fts5/fts5_config.c index 5d0770502e..d2e8309cd2 100644 --- a/ext/fts5/fts5_config.c +++ b/ext/fts5/fts5_config.c @@ -398,6 +398,16 @@ static int fts5ConfigParseSpecial( return rc; } + if( sqlite3_strnicmp("tokendata", zCmd, nCmd)==0 ){ + if( (zArg[0]!='0' && zArg[0]!='1') || zArg[1]!='\0' ){ + *pzErr = sqlite3_mprintf("malformed tokendata=... directive"); + rc = SQLITE_ERROR; + }else{ + pConfig->bTokendata = (zArg[0]=='1'); + } + return rc; + } + *pzErr = sqlite3_mprintf("unrecognized option: \"%.*s\"", nCmd, zCmd); return SQLITE_ERROR; } diff --git a/ext/fts5/fts5_expr.c b/ext/fts5/fts5_expr.c index f5101ba065..bc7c9741ee 100644 --- a/ext/fts5/fts5_expr.c +++ b/ext/fts5/fts5_expr.c @@ -100,7 +100,9 @@ struct Fts5ExprNode { struct Fts5ExprTerm { u8 bPrefix; /* True for a prefix term */ u8 bFirst; /* True if token must be first in column */ - char *zTerm; /* nul-terminated term */ + char *pTerm; /* Term data */ + int nQueryTerm; /* Effective size of term in bytes */ + int nFullTerm; /* Size of term in bytes incl. tokendata */ Fts5IndexIter *pIter; /* Iterator for this term */ Fts5ExprTerm *pSynonym; /* Pointer to first in list of synonyms */ }; @@ -967,7 +969,7 @@ static int fts5ExprNearInitAll( p->pIter = 0; } rc = sqlite3Fts5IndexQuery( - pExpr->pIndex, p->zTerm, (int)strlen(p->zTerm), + pExpr->pIndex, p->pTerm, p->nQueryTerm, (pTerm->bPrefix ? FTS5INDEX_QUERY_PREFIX : 0) | (pExpr->bDesc ? FTS5INDEX_QUERY_DESC : 0), pNear->pColset, @@ -1604,7 +1606,7 @@ static void fts5ExprPhraseFree(Fts5ExprPhrase *pPhrase){ Fts5ExprTerm *pSyn; Fts5ExprTerm *pNext; Fts5ExprTerm *pTerm = &pPhrase->aTerm[i]; - sqlite3_free(pTerm->zTerm); + sqlite3_free(pTerm->pTerm); sqlite3Fts5IterClose(pTerm->pIter); for(pSyn=pTerm->pSynonym; pSyn; pSyn=pNext){ pNext = pSyn->pSynonym; @@ -1702,6 +1704,7 @@ Fts5ExprNearset *sqlite3Fts5ParseNearset( typedef struct TokenCtx TokenCtx; struct TokenCtx { Fts5ExprPhrase *pPhrase; + Fts5Config *pConfig; int rc; }; @@ -1735,8 +1738,10 @@ static int fts5ParseTokenize( rc = SQLITE_NOMEM; }else{ memset(pSyn, 0, (size_t)nByte); - pSyn->zTerm = ((char*)pSyn) + sizeof(Fts5ExprTerm) + sizeof(Fts5Buffer); - memcpy(pSyn->zTerm, pToken, nToken); + pSyn->pTerm = ((char*)pSyn) + sizeof(Fts5ExprTerm) + sizeof(Fts5Buffer); + pSyn->nFullTerm = pSyn->nQueryTerm = nToken; + if( pCtx->pConfig->bTokendata ) pSyn->nQueryTerm = strlen(pSyn->pTerm); + memcpy(pSyn->pTerm, pToken, nToken); pSyn->pSynonym = pPhrase->aTerm[pPhrase->nTerm-1].pSynonym; pPhrase->aTerm[pPhrase->nTerm-1].pSynonym = pSyn; } @@ -1761,7 +1766,11 @@ static int fts5ParseTokenize( if( rc==SQLITE_OK ){ pTerm = &pPhrase->aTerm[pPhrase->nTerm++]; memset(pTerm, 0, sizeof(Fts5ExprTerm)); - pTerm->zTerm = sqlite3Fts5Strndup(&rc, pToken, nToken); + pTerm->pTerm = sqlite3Fts5Strndup(&rc, pToken, nToken); + pTerm->nFullTerm = pTerm->nQueryTerm = nToken; + if( pCtx->pConfig->bTokendata && rc==SQLITE_OK ){ + pTerm->nQueryTerm = strlen(pTerm->pTerm); + } } } @@ -1828,6 +1837,7 @@ Fts5ExprPhrase *sqlite3Fts5ParseTerm( memset(&sCtx, 0, sizeof(TokenCtx)); sCtx.pPhrase = pAppend; + sCtx.pConfig = pConfig; rc = fts5ParseStringFromToken(pToken, &z); if( rc==SQLITE_OK ){ @@ -1877,8 +1887,7 @@ int sqlite3Fts5ExprClonePhrase( int rc = SQLITE_OK; /* Return code */ Fts5ExprPhrase *pOrig; /* The phrase extracted from pExpr */ Fts5Expr *pNew = 0; /* Expression to return via *ppNew */ - TokenCtx sCtx = {0,0}; /* Context object for fts5ParseTokenize */ - + TokenCtx sCtx = {0,0,0}; /* Context object for fts5ParseTokenize */ pOrig = pExpr->apExprPhrase[iPhrase]; pNew = (Fts5Expr*)sqlite3Fts5MallocZero(&rc, sizeof(Fts5Expr)); if( rc==SQLITE_OK ){ @@ -1909,13 +1918,12 @@ int sqlite3Fts5ExprClonePhrase( if( pOrig->nTerm ){ int i; /* Used to iterate through phrase terms */ + sCtx.pConfig = pExpr->pConfig; for(i=0; rc==SQLITE_OK && inTerm; i++){ int tflags = 0; Fts5ExprTerm *p; for(p=&pOrig->aTerm[i]; p && rc==SQLITE_OK; p=p->pSynonym){ - const char *zTerm = p->zTerm; - rc = fts5ParseTokenize((void*)&sCtx, tflags, zTerm, (int)strlen(zTerm), - 0, 0); + rc = fts5ParseTokenize((void*)&sCtx, tflags, p->pTerm,p->nFullTerm,0,0); tflags = FTS5_TOKEN_COLOCATED; } if( rc==SQLITE_OK ){ @@ -2296,11 +2304,13 @@ static Fts5ExprNode *fts5ParsePhraseToAnd( if( parseGrowPhraseArray(pParse) ){ fts5ExprPhraseFree(pPhrase); }else{ + Fts5ExprTerm *p = &pNear->apPhrase[0]->aTerm[ii]; + Fts5ExprTerm *pTo = &pPhrase->aTerm[0]; pParse->apPhrase[pParse->nPhrase++] = pPhrase; pPhrase->nTerm = 1; - pPhrase->aTerm[0].zTerm = sqlite3Fts5Strndup( - &pParse->rc, pNear->apPhrase[0]->aTerm[ii].zTerm, -1 - ); + pTo->pTerm = sqlite3Fts5Strndup(&pParse->rc, p->pTerm, p->nFullTerm); + pTo->nQueryTerm = p->nQueryTerm; + pTo->nFullTerm = p->nFullTerm; pRet->apChild[ii] = sqlite3Fts5ParseNode(pParse, FTS5_STRING, 0, 0, sqlite3Fts5ParseNearset(pParse, 0, pPhrase) ); @@ -2485,16 +2495,17 @@ static char *fts5ExprTermPrint(Fts5ExprTerm *pTerm){ /* Determine the maximum amount of space required. */ for(p=pTerm; p; p=p->pSynonym){ - nByte += (int)strlen(pTerm->zTerm) * 2 + 3 + 2; + nByte += pTerm->nQueryTerm * 2 + 3 + 2; } zQuoted = sqlite3_malloc64(nByte); if( zQuoted ){ int i = 0; for(p=pTerm; p; p=p->pSynonym){ - char *zIn = p->zTerm; + char *zIn = p->pTerm; + char *zEnd = &zIn[p->nQueryTerm]; zQuoted[i++] = '"'; - while( *zIn ){ + while( zInnTerm; iTerm++){ - char *zTerm = pPhrase->aTerm[iTerm].zTerm; - zRet = fts5PrintfAppend(zRet, "%s%s", iTerm==0?"":" ", zTerm); + Fts5ExprTerm *p = &pPhrase->aTerm[iTerm]; + zRet = fts5PrintfAppend(zRet, "%s%.*s", iTerm==0?"":" ", + p->nQueryTerm, p->pTerm + ); if( pPhrase->aTerm[iTerm].bPrefix ){ zRet = fts5PrintfAppend(zRet, "*"); } @@ -2974,6 +2987,17 @@ static int fts5ExprColsetTest(Fts5Colset *pColset, int iCol){ return 0; } +/* +** pToken is a buffer nToken bytes in size that may or may not contain +** an embedded 0x00 byte. If it does, return the number of bytes in +** the buffer before the 0x00. If it does not, return nToken. +*/ +static int fts5QueryTerm(const char *pToken, int nToken){ + int ii; + for(ii=0; iipExpr; int i; + int nQuery = nToken; + i64 iRowid = pExpr->pRoot->iRowid; UNUSED_PARAM2(iUnused1, iUnused2); - if( nToken>FTS5_MAX_TOKEN_SIZE ) nToken = FTS5_MAX_TOKEN_SIZE; + if( nQuery>FTS5_MAX_TOKEN_SIZE ) nQuery = FTS5_MAX_TOKEN_SIZE; + if( pExpr->pConfig->bTokendata ){ + nQuery = fts5QueryTerm(pToken, nQuery); + } if( (tflags & FTS5_TOKEN_COLOCATED)==0 ) p->iOff++; for(i=0; inPhrase; i++){ - Fts5ExprTerm *pTerm; + Fts5ExprTerm *pT; if( p->aPopulator[i].bOk==0 ) continue; - for(pTerm=&pExpr->apExprPhrase[i]->aTerm[0]; pTerm; pTerm=pTerm->pSynonym){ - int nTerm = (int)strlen(pTerm->zTerm); - if( (nTerm==nToken || (nTermbPrefix)) - && memcmp(pTerm->zTerm, pToken, nTerm)==0 + for(pT=&pExpr->apExprPhrase[i]->aTerm[0]; pT; pT=pT->pSynonym){ + if( (pT->nQueryTerm==nQuery || (pT->nQueryTermbPrefix)) + && memcmp(pT->pTerm, pToken, pT->nQueryTerm)==0 ){ int rc = sqlite3Fts5PoslistWriterAppend( &pExpr->apExprPhrase[i]->poslist, &p->aPopulator[i].writer, p->iOff ); + if( rc==SQLITE_OK && pExpr->pConfig->bTokendata && !pT->bPrefix ){ + int iCol = p->iOff>>32; + int iTokOff = p->iOff & 0x7FFFFFFF; + rc = sqlite3Fts5IndexIterWriteTokendata( + pT->pIter, pToken, nToken, iRowid, iCol, iTokOff + ); + } if( rc ) return rc; break; } @@ -3135,3 +3170,81 @@ int sqlite3Fts5ExprPhraseCollist( return rc; } + +/* +** Does the work of the fts5_api.xQueryToken() API method. +*/ +int sqlite3Fts5ExprQueryToken( + Fts5Expr *pExpr, + int iPhrase, + int iToken, + const char **ppOut, + int *pnOut +){ + Fts5ExprPhrase *pPhrase = 0; + + if( iPhrase<0 || iPhrase>=pExpr->nPhrase ){ + return SQLITE_RANGE; + } + pPhrase = pExpr->apExprPhrase[iPhrase]; + if( iToken<0 || iToken>=pPhrase->nTerm ){ + return SQLITE_RANGE; + } + + *ppOut = pPhrase->aTerm[iToken].pTerm; + *pnOut = pPhrase->aTerm[iToken].nFullTerm; + return SQLITE_OK; +} + +/* +** Does the work of the fts5_api.xInstToken() API method. +*/ +int sqlite3Fts5ExprInstToken( + Fts5Expr *pExpr, + i64 iRowid, + int iPhrase, + int iCol, + int iOff, + int iToken, + const char **ppOut, + int *pnOut +){ + Fts5ExprPhrase *pPhrase = 0; + Fts5ExprTerm *pTerm = 0; + int rc = SQLITE_OK; + + if( iPhrase<0 || iPhrase>=pExpr->nPhrase ){ + return SQLITE_RANGE; + } + pPhrase = pExpr->apExprPhrase[iPhrase]; + if( iToken<0 || iToken>=pPhrase->nTerm ){ + return SQLITE_RANGE; + } + pTerm = &pPhrase->aTerm[iToken]; + if( pTerm->bPrefix==0 ){ + if( pExpr->pConfig->bTokendata ){ + rc = sqlite3Fts5IterToken( + pTerm->pIter, iRowid, iCol, iOff+iToken, ppOut, pnOut + ); + }else{ + *ppOut = pTerm->pTerm; + *pnOut = pTerm->nFullTerm; + } + } + return rc; +} + +/* +** Clear the token mappings for all Fts5IndexIter objects mannaged by +** the expression passed as the only argument. +*/ +void sqlite3Fts5ExprClearTokens(Fts5Expr *pExpr){ + int ii; + for(ii=0; iinPhrase; ii++){ + Fts5ExprTerm *pT; + for(pT=&pExpr->apExprPhrase[ii]->aTerm[0]; pT; pT=pT->pSynonym){ + sqlite3Fts5IndexIterClearTokendata(pT->pIter); + } + } +} + diff --git a/ext/fts5/fts5_hash.c b/ext/fts5/fts5_hash.c index 391791c7ac..5e0959aa8e 100644 --- a/ext/fts5/fts5_hash.c +++ b/ext/fts5/fts5_hash.c @@ -36,10 +36,15 @@ struct Fts5Hash { /* ** Each entry in the hash table is represented by an object of the -** following type. Each object, its key (a nul-terminated string) and -** its current data are stored in a single memory allocation. The -** key immediately follows the object in memory. The position list -** data immediately follows the key data in memory. +** following type. Each object, its key, and its current data are stored +** in a single memory allocation. The key immediately follows the object +** in memory. The position list data immediately follows the key data +** in memory. +** +** The key is Fts5HashEntry.nKey bytes in size. It consists of a single +** byte identifying the index (either the main term index or a prefix-index), +** followed by the term data. For example: "0token". There is no +** nul-terminator - in this case nKey=6. ** ** The data that follows the key is in a similar, but not identical format ** to the doclist data stored in the database. It is: @@ -174,8 +179,7 @@ static int fts5HashResize(Fts5Hash *pHash){ unsigned int iHash; Fts5HashEntry *p = apOld[i]; apOld[i] = p->pHashNext; - iHash = fts5HashKey(nNew, (u8*)fts5EntryKey(p), - (int)strlen(fts5EntryKey(p))); + iHash = fts5HashKey(nNew, (u8*)fts5EntryKey(p), p->nKey); p->pHashNext = apNew[iHash]; apNew[iHash] = p; } @@ -259,7 +263,7 @@ int sqlite3Fts5HashWrite( for(p=pHash->aSlot[iHash]; p; p=p->pHashNext){ char *zKey = fts5EntryKey(p); if( zKey[0]==bByte - && p->nKey==nToken + && p->nKey==nToken+1 && memcmp(&zKey[1], pToken, nToken)==0 ){ break; @@ -289,9 +293,9 @@ int sqlite3Fts5HashWrite( zKey[0] = bByte; memcpy(&zKey[1], pToken, nToken); assert( iHash==fts5HashKey(pHash->nSlot, (u8*)zKey, nToken+1) ); - p->nKey = nToken; + p->nKey = nToken+1; zKey[nToken+1] = '\0'; - p->nData = nToken+1 + 1 + sizeof(Fts5HashEntry); + p->nData = nToken+1 + sizeof(Fts5HashEntry); p->pHashNext = pHash->aSlot[iHash]; pHash->aSlot[iHash] = p; pHash->nEntry++; @@ -408,12 +412,17 @@ static Fts5HashEntry *fts5HashEntryMerge( *ppOut = p1; p1 = 0; }else{ - int i = 0; char *zKey1 = fts5EntryKey(p1); char *zKey2 = fts5EntryKey(p2); - while( zKey1[i]==zKey2[i] ) i++; + int nMin = MIN(p1->nKey, p2->nKey); - if( ((u8)zKey1[i])>((u8)zKey2[i]) ){ + int cmp = memcmp(zKey1, zKey2, nMin); + if( cmp==0 ){ + cmp = p1->nKey - p2->nKey; + } + assert( cmp!=0 ); + + if( cmp>0 ){ /* p2 is smaller */ *ppOut = p2; ppOut = &p2->pScanNext; @@ -455,7 +464,7 @@ static int fts5HashEntrySort( Fts5HashEntry *pIter; for(pIter=pHash->aSlot[iSlot]; pIter; pIter=pIter->pHashNext){ if( pTerm==0 - || (pIter->nKey+1>=nTerm && 0==memcmp(fts5EntryKey(pIter), pTerm, nTerm)) + || (pIter->nKey>=nTerm && 0==memcmp(fts5EntryKey(pIter), pTerm, nTerm)) ){ Fts5HashEntry *pEntry = pIter; pEntry->pScanNext = 0; @@ -494,12 +503,11 @@ int sqlite3Fts5HashQuery( for(p=pHash->aSlot[iHash]; p; p=p->pHashNext){ zKey = fts5EntryKey(p); - assert( p->nKey+1==(int)strlen(zKey) ); - if( nTerm==p->nKey+1 && memcmp(zKey, pTerm, nTerm)==0 ) break; + if( nTerm==p->nKey && memcmp(zKey, pTerm, nTerm)==0 ) break; } if( p ){ - int nHashPre = sizeof(Fts5HashEntry) + nTerm + 1; + int nHashPre = sizeof(Fts5HashEntry) + nTerm; int nList = p->nData - nHashPre; u8 *pRet = (u8*)(*ppOut = sqlite3_malloc64(nPre + nList + 10)); if( pRet ){ @@ -560,19 +568,22 @@ int sqlite3Fts5HashScanEof(Fts5Hash *p){ void sqlite3Fts5HashScanEntry( Fts5Hash *pHash, const char **pzTerm, /* OUT: term (nul-terminated) */ + int *pnTerm, /* OUT: Size of term in bytes */ const u8 **ppDoclist, /* OUT: pointer to doclist */ int *pnDoclist /* OUT: size of doclist in bytes */ ){ Fts5HashEntry *p; if( (p = pHash->pScan) ){ char *zKey = fts5EntryKey(p); - int nTerm = (int)strlen(zKey); + int nTerm = p->nKey; fts5HashAddPoslistSize(pHash, p, 0); *pzTerm = zKey; - *ppDoclist = (const u8*)&zKey[nTerm+1]; - *pnDoclist = p->nData - (sizeof(Fts5HashEntry) + nTerm + 1); + *pnTerm = nTerm; + *ppDoclist = (const u8*)&zKey[nTerm]; + *pnDoclist = p->nData - (sizeof(Fts5HashEntry) + nTerm); }else{ *pzTerm = 0; + *pnTerm = 0; *ppDoclist = 0; *pnDoclist = 0; } diff --git a/ext/fts5/fts5_index.c b/ext/fts5/fts5_index.c index c467addb8a..7ec1976d33 100644 --- a/ext/fts5/fts5_index.c +++ b/ext/fts5/fts5_index.c @@ -323,6 +323,9 @@ typedef struct Fts5SegWriter Fts5SegWriter; typedef struct Fts5Structure Fts5Structure; typedef struct Fts5StructureLevel Fts5StructureLevel; typedef struct Fts5StructureSegment Fts5StructureSegment; +typedef struct Fts5TokenDataIter Fts5TokenDataIter; +typedef struct Fts5TokenDataMap Fts5TokenDataMap; +typedef struct Fts5TombstoneArray Fts5TombstoneArray; struct Fts5Data { u8 *p; /* Pointer to buffer containing record */ @@ -365,6 +368,7 @@ struct Fts5Index { sqlite3_stmt *pIdxWriter; /* "INSERT ... %_idx VALUES(?,?,?,?)" */ sqlite3_stmt *pIdxDeleter; /* "DELETE FROM %_idx WHERE segid=?" */ sqlite3_stmt *pIdxSelect; + sqlite3_stmt *pIdxNextSelect; int nRead; /* Total number of blocks read */ sqlite3_stmt *pDeleteFromIdx; @@ -518,8 +522,7 @@ struct Fts5SegIter { Fts5Data *pLeaf; /* Current leaf data */ Fts5Data *pNextLeaf; /* Leaf page (iLeafPgno+1) */ i64 iLeafOffset; /* Byte offset within current leaf */ - Fts5Data **apTombstone; /* Array of tombstone pages */ - int nTombstone; + Fts5TombstoneArray *pTombArray; /* Array of tombstone pages */ /* Next method */ void (*xNext)(Fts5Index*, Fts5SegIter*, int*); @@ -546,6 +549,15 @@ struct Fts5SegIter { u8 bDel; /* True if the delete flag is set */ }; +/* +** Array of tombstone pages. Reference counted. +*/ +struct Fts5TombstoneArray { + int nRef; /* Number of pointers to this object */ + int nTombstone; + Fts5Data *apTombstone[1]; /* Array of tombstone pages */ +}; + /* ** Argument is a pointer to an Fts5Data structure that contains a ** leaf page. @@ -590,9 +602,16 @@ struct Fts5SegIter { ** poslist: ** Used by sqlite3Fts5IterPoslist() when the poslist needs to be buffered. ** There is no way to tell if this is populated or not. +** +** pColset: +** If not NULL, points to an object containing a set of column indices. +** Only matches that occur in one of these columns will be returned. +** The Fts5Iter does not own the Fts5Colset object, and so it is not +** freed when the iterator is closed - it is owned by the upper layer. */ struct Fts5Iter { Fts5IndexIter base; /* Base class containing output vars */ + Fts5TokenDataIter *pTokenDataIter; Fts5Index *pIndex; /* Index that owns this iterator */ Fts5Buffer poslist; /* Buffer containing current poslist */ @@ -610,7 +629,6 @@ struct Fts5Iter { Fts5SegIter aSeg[1]; /* Array of segment iterators */ }; - /* ** An instance of the following type is used to iterate through the contents ** of a doclist-index record. @@ -1909,18 +1927,20 @@ static void fts5SegIterSetNext(Fts5Index *p, Fts5SegIter *pIter){ } /* -** Allocate a tombstone hash page array (pIter->apTombstone) for the -** iterator passed as the second argument. If an OOM error occurs, leave -** an error in the Fts5Index object. +** Allocate a tombstone hash page array object (pIter->pTombArray) for +** the iterator passed as the second argument. If an OOM error occurs, +** leave an error in the Fts5Index object. */ static void fts5SegIterAllocTombstone(Fts5Index *p, Fts5SegIter *pIter){ const int nTomb = pIter->pSeg->nPgTombstone; if( nTomb>0 ){ - Fts5Data **apTomb = 0; - apTomb = (Fts5Data**)sqlite3Fts5MallocZero(&p->rc, sizeof(Fts5Data)*nTomb); - if( apTomb ){ - pIter->apTombstone = apTomb; - pIter->nTombstone = nTomb; + int nByte = nTomb * sizeof(Fts5Data*) + sizeof(Fts5TombstoneArray); + Fts5TombstoneArray *pNew; + pNew = (Fts5TombstoneArray*)sqlite3Fts5MallocZero(&p->rc, nByte); + if( pNew ){ + pNew->nTombstone = nTomb; + pNew->nRef = 1; + pIter->pTombArray = pNew; } } } @@ -2177,15 +2197,16 @@ static void fts5SegIterNext_None( }else{ const u8 *pList = 0; const char *zTerm = 0; + int nTerm = 0; int nList; sqlite3Fts5HashScanNext(p->pHash); - sqlite3Fts5HashScanEntry(p->pHash, &zTerm, &pList, &nList); + sqlite3Fts5HashScanEntry(p->pHash, &zTerm, &nTerm, &pList, &nList); if( pList==0 ) goto next_none_eof; pIter->pLeaf->p = (u8*)pList; pIter->pLeaf->nn = nList; pIter->pLeaf->szLeaf = nList; pIter->iEndofDoclist = nList; - sqlite3Fts5BufferSet(&p->rc,&pIter->term, (int)strlen(zTerm), (u8*)zTerm); + sqlite3Fts5BufferSet(&p->rc,&pIter->term, nTerm, (u8*)zTerm); pIter->iLeafOffset = fts5GetVarint(pList, (u64*)&pIter->iRowid); } @@ -2251,11 +2272,12 @@ static void fts5SegIterNext( }else if( pIter->pSeg==0 ){ const u8 *pList = 0; const char *zTerm = 0; + int nTerm = 0; int nList = 0; assert( (pIter->flags & FTS5_SEGITER_ONETERM) || pbNewTerm ); if( 0==(pIter->flags & FTS5_SEGITER_ONETERM) ){ sqlite3Fts5HashScanNext(p->pHash); - sqlite3Fts5HashScanEntry(p->pHash, &zTerm, &pList, &nList); + sqlite3Fts5HashScanEntry(p->pHash, &zTerm, &nTerm, &pList, &nList); } if( pList==0 ){ fts5DataRelease(pIter->pLeaf); @@ -2265,8 +2287,7 @@ static void fts5SegIterNext( pIter->pLeaf->nn = nList; pIter->pLeaf->szLeaf = nList; pIter->iEndofDoclist = nList+1; - sqlite3Fts5BufferSet(&p->rc, &pIter->term, (int)strlen(zTerm), - (u8*)zTerm); + sqlite3Fts5BufferSet(&p->rc, &pIter->term, nTerm, (u8*)zTerm); pIter->iLeafOffset = fts5GetVarint(pList, (u64*)&pIter->iRowid); *pbNewTerm = 1; } @@ -2652,7 +2673,7 @@ static void fts5SegIterSeekInit( fts5LeafSeek(p, bGe, pIter, pTerm, nTerm); } - if( p->rc==SQLITE_OK && bGe==0 ){ + if( p->rc==SQLITE_OK && (bGe==0 || (flags & FTS5INDEX_QUERY_SCANONETERM)) ){ pIter->flags |= FTS5_SEGITER_ONETERM; if( pIter->pLeaf ){ if( flags & FTS5INDEX_QUERY_DESC ){ @@ -2668,7 +2689,9 @@ static void fts5SegIterSeekInit( } fts5SegIterSetNext(p, pIter); - fts5SegIterAllocTombstone(p, pIter); + if( 0==(flags & FTS5INDEX_QUERY_SCANONETERM) ){ + fts5SegIterAllocTombstone(p, pIter); + } /* Either: ** @@ -2685,6 +2708,79 @@ static void fts5SegIterSeekInit( ); } + +/* +** SQL used by fts5SegIterNextInit() to find the page to open. +*/ +static sqlite3_stmt *fts5IdxNextStmt(Fts5Index *p){ + if( p->pIdxNextSelect==0 ){ + Fts5Config *pConfig = p->pConfig; + fts5IndexPrepareStmt(p, &p->pIdxNextSelect, sqlite3_mprintf( + "SELECT pgno FROM '%q'.'%q_idx' WHERE " + "segid=? AND term>? ORDER BY term ASC LIMIT 1", + pConfig->zDb, pConfig->zName + )); + + } + return p->pIdxNextSelect; +} + +/* +** This is similar to fts5SegIterSeekInit(), except that it initializes +** the segment iterator to point to the first term following the page +** with pToken/nToken on it. +*/ +static void fts5SegIterNextInit( + Fts5Index *p, + const char *pTerm, int nTerm, + Fts5StructureSegment *pSeg, /* Description of segment */ + Fts5SegIter *pIter /* Object to populate */ +){ + int iPg = -1; /* Page of segment to open */ + int bDlidx = 0; + sqlite3_stmt *pSel = 0; /* SELECT to find iPg */ + + pSel = fts5IdxNextStmt(p); + if( pSel ){ + assert( p->rc==SQLITE_OK ); + sqlite3_bind_int(pSel, 1, pSeg->iSegid); + sqlite3_bind_blob(pSel, 2, pTerm, nTerm, SQLITE_STATIC); + + if( sqlite3_step(pSel)==SQLITE_ROW ){ + i64 val = sqlite3_column_int64(pSel, 0); + iPg = (int)(val>>1); + bDlidx = (val & 0x0001); + } + p->rc = sqlite3_reset(pSel); + sqlite3_bind_null(pSel, 2); + if( p->rc ) return; + } + + memset(pIter, 0, sizeof(*pIter)); + pIter->pSeg = pSeg; + pIter->flags |= FTS5_SEGITER_ONETERM; + if( iPg>=0 ){ + pIter->iLeafPgno = iPg - 1; + fts5SegIterNextPage(p, pIter); + fts5SegIterSetNext(p, pIter); + } + if( pIter->pLeaf ){ + const u8 *a = pIter->pLeaf->p; + int iTermOff = 0; + + pIter->iPgidxOff = pIter->pLeaf->szLeaf; + pIter->iPgidxOff += fts5GetVarint32(&a[pIter->iPgidxOff], iTermOff); + pIter->iLeafOffset = iTermOff; + fts5SegIterLoadTerm(p, pIter, 0); + fts5SegIterLoadNPos(p, pIter); + if( bDlidx ) fts5SegIterLoadDlidx(p, pIter); + + assert( p->rc!=SQLITE_OK || + fts5BufferCompareBlob(&pIter->term, (const u8*)pTerm, nTerm)>0 + ); + } +} + /* ** Initialize the object pIter to point to term pTerm/nTerm within the ** in-memory hash table. If there is no such term in the hash-table, the @@ -2711,8 +2807,7 @@ static void fts5SegIterHashInit( const u8 *pList = 0; p->rc = sqlite3Fts5HashScanInit(p->pHash, (const char*)pTerm, nTerm); - sqlite3Fts5HashScanEntry(p->pHash, (const char**)&z, &pList, &nList); - n = (z ? (int)strlen((const char*)z) : 0); + sqlite3Fts5HashScanEntry(p->pHash, (const char**)&z, &n, &pList, &nList); if( pList ){ pLeaf = fts5IdxMalloc(p, sizeof(Fts5Data)); if( pLeaf ){ @@ -2771,6 +2866,23 @@ static void fts5IndexFreeArray(Fts5Data **ap, int n){ } } +/* +** Decrement the ref-count of the object passed as the only argument. If it +** reaches 0, free it and its contents. +*/ +static void fts5TombstoneArrayDelete(Fts5TombstoneArray *p){ + if( p ){ + p->nRef--; + if( p->nRef<=0 ){ + int ii; + for(ii=0; iinTombstone; ii++){ + fts5DataRelease(p->apTombstone[ii]); + } + sqlite3_free(p); + } + } +} + /* ** Zero the iterator passed as the only argument. */ @@ -2778,7 +2890,7 @@ static void fts5SegIterClear(Fts5SegIter *pIter){ fts5BufferFree(&pIter->term); fts5DataRelease(pIter->pLeaf); fts5DataRelease(pIter->pNextLeaf); - fts5IndexFreeArray(pIter->apTombstone, pIter->nTombstone); + fts5TombstoneArrayDelete(pIter->pTombArray); fts5DlidxIterFree(pIter->pDlidx); sqlite3_free(pIter->aRowidOffset); memset(pIter, 0, sizeof(Fts5SegIter)); @@ -3023,7 +3135,6 @@ static void fts5SegIterNextFrom( }while( p->rc==SQLITE_OK ); } - /* ** Free the iterator object passed as the second argument. */ @@ -3168,24 +3279,25 @@ static int fts5IndexTombstoneQuery( static int fts5MultiIterIsDeleted(Fts5Iter *pIter){ int iFirst = pIter->aFirst[1].iFirst; Fts5SegIter *pSeg = &pIter->aSeg[iFirst]; + Fts5TombstoneArray *pArray = pSeg->pTombArray; - if( pSeg->pLeaf && pSeg->nTombstone ){ + if( pSeg->pLeaf && pArray ){ /* Figure out which page the rowid might be present on. */ - int iPg = ((u64)pSeg->iRowid) % pSeg->nTombstone; + int iPg = ((u64)pSeg->iRowid) % pArray->nTombstone; assert( iPg>=0 ); /* If tombstone hash page iPg has not yet been loaded from the ** database, load it now. */ - if( pSeg->apTombstone[iPg]==0 ){ - pSeg->apTombstone[iPg] = fts5DataRead(pIter->pIndex, + if( pArray->apTombstone[iPg]==0 ){ + pArray->apTombstone[iPg] = fts5DataRead(pIter->pIndex, FTS5_TOMBSTONE_ROWID(pSeg->pSeg->iSegid, iPg) ); - if( pSeg->apTombstone[iPg]==0 ) return 0; + if( pArray->apTombstone[iPg]==0 ) return 0; } return fts5IndexTombstoneQuery( - pSeg->apTombstone[iPg], - pSeg->nTombstone, + pArray->apTombstone[iPg], + pArray->nTombstone, pSeg->iRowid ); } @@ -3724,6 +3836,32 @@ static void fts5IterSetOutputCb(int *pRc, Fts5Iter *pIter){ } } +/* +** All the component segment-iterators of pIter have been set up. This +** functions finishes setup for iterator pIter itself. +*/ +static void fts5MultiIterFinishSetup(Fts5Index *p, Fts5Iter *pIter){ + int iIter; + for(iIter=pIter->nSeg-1; iIter>0; iIter--){ + int iEq; + if( (iEq = fts5MultiIterDoCompare(pIter, iIter)) ){ + Fts5SegIter *pSeg = &pIter->aSeg[iEq]; + if( p->rc==SQLITE_OK ) pSeg->xNext(p, pSeg, 0); + fts5MultiIterAdvanced(p, pIter, iEq, iIter); + } + } + fts5MultiIterSetEof(pIter); + fts5AssertMultiIterSetup(p, pIter); + + if( (pIter->bSkipEmpty && fts5MultiIterIsEmpty(p, pIter)) + || fts5MultiIterIsDeleted(pIter) + ){ + fts5MultiIterNext(p, pIter, 0, 0); + }else if( pIter->base.bEof==0 ){ + Fts5SegIter *pSeg = &pIter->aSeg[pIter->aFirst[1].iFirst]; + pIter->xSetOutputs(pIter, pSeg); + } +} /* ** Allocate a new Fts5Iter object. @@ -3805,31 +3943,12 @@ static void fts5MultiIterNew( assert( iIter==nSeg ); } - /* If the above was successful, each component iterators now points + /* If the above was successful, each component iterator now points ** to the first entry in its segment. In this case initialize the ** aFirst[] array. Or, if an error has occurred, free the iterator ** object and set the output variable to NULL. */ if( p->rc==SQLITE_OK ){ - for(iIter=pNew->nSeg-1; iIter>0; iIter--){ - int iEq; - if( (iEq = fts5MultiIterDoCompare(pNew, iIter)) ){ - Fts5SegIter *pSeg = &pNew->aSeg[iEq]; - if( p->rc==SQLITE_OK ) pSeg->xNext(p, pSeg, 0); - fts5MultiIterAdvanced(p, pNew, iEq, iIter); - } - } - fts5MultiIterSetEof(pNew); - fts5AssertMultiIterSetup(p, pNew); - - if( (pNew->bSkipEmpty && fts5MultiIterIsEmpty(p, pNew)) - || fts5MultiIterIsDeleted(pNew) - ){ - fts5MultiIterNext(p, pNew, 0, 0); - }else if( pNew->base.bEof==0 ){ - Fts5SegIter *pSeg = &pNew->aSeg[pNew->aFirst[1].iFirst]; - pNew->xSetOutputs(pNew, pSeg); - } - + fts5MultiIterFinishSetup(p, pNew); }else{ fts5MultiIterFree(pNew); *ppOut = 0; @@ -3854,7 +3973,6 @@ static void fts5MultiIterNew2( pNew = fts5MultiIterAlloc(p, 2); if( pNew ){ Fts5SegIter *pIter = &pNew->aSeg[1]; - pIter->flags = FTS5_SEGITER_ONETERM; if( pData->szLeaf>0 ){ pIter->pLeaf = pData; @@ -4217,7 +4335,7 @@ static void fts5WriteDlidxAppend( } if( pDlidx->bPrevValid ){ - iVal = iRowid - pDlidx->iPrev; + iVal = (u64)iRowid - (u64)pDlidx->iPrev; }else{ i64 iPgno = (i==0 ? pWriter->writer.pgno : pDlidx[-1].pgno); assert( pDlidx->buf.n==0 ); @@ -5340,10 +5458,10 @@ static void fts5FlushSecureDelete( Fts5Index *p, Fts5Structure *pStruct, const char *zTerm, + int nTerm, i64 iRowid ){ const int f = FTS5INDEX_QUERY_SKIPHASH; - int nTerm = (int)strlen(zTerm); Fts5Iter *pIter = 0; /* Used to find term instance */ fts5MultiIterNew(p, pStruct, f, 0, (const u8*)zTerm, nTerm, -1, 0, &pIter); @@ -5417,8 +5535,7 @@ static void fts5FlushOneHash(Fts5Index *p){ int nDoclist; /* Size of doclist in bytes */ /* Get the term and doclist for this entry. */ - sqlite3Fts5HashScanEntry(pHash, &zTerm, &pDoclist, &nDoclist); - nTerm = (int)strlen(zTerm); + sqlite3Fts5HashScanEntry(pHash, &zTerm, &nTerm, &pDoclist, &nDoclist); if( bSecureDelete==0 ){ fts5WriteAppendTerm(p, &writer, nTerm, (const u8*)zTerm); if( p->rc!=SQLITE_OK ) break; @@ -5448,7 +5565,7 @@ static void fts5FlushOneHash(Fts5Index *p){ if( bSecureDelete ){ if( eDetail==FTS5_DETAIL_NONE ){ if( iOffrc!=SQLITE_OK || pDoclist[iOff]==0x01 ){ iOff++; continue; @@ -6078,7 +6195,7 @@ static void fts5SetupPrefixIter( u8 *pToken, /* Buffer containing prefix to match */ int nToken, /* Size of buffer pToken in bytes */ Fts5Colset *pColset, /* Restrict matches to these columns */ - Fts5Iter **ppIter /* OUT: New iterator */ + Fts5Iter **ppIter /* OUT: New iterator */ ){ Fts5Structure *pStruct; Fts5Buffer *aBuf; @@ -6099,8 +6216,9 @@ static void fts5SetupPrefixIter( aBuf = (Fts5Buffer*)fts5IdxMalloc(p, sizeof(Fts5Buffer)*nBuf); pStruct = fts5StructureRead(p); + assert( p->rc!=SQLITE_OK || (aBuf && pStruct) ); - if( aBuf && pStruct ){ + if( p->rc==SQLITE_OK ){ const int flags = FTS5INDEX_QUERY_SCAN | FTS5INDEX_QUERY_SKIPEMPTY | FTS5INDEX_QUERY_NOOUTPUT; @@ -6112,6 +6230,12 @@ static void fts5SetupPrefixIter( int bNewTerm = 1; memset(&doclist, 0, sizeof(doclist)); + + /* If iIdx is non-zero, then it is the number of a prefix-index for + ** prefixes 1 character longer than the prefix being queried for. That + ** index contains all the doclists required, except for the one + ** corresponding to the prefix itself. That one is extracted from the + ** main term index here. */ if( iIdx!=0 ){ int dummy = 0; const int f2 = FTS5INDEX_QUERY_SKIPEMPTY|FTS5INDEX_QUERY_NOOUTPUT; @@ -6135,6 +6259,7 @@ static void fts5SetupPrefixIter( pToken[0] = FTS5_MAIN_PREFIX + iIdx; fts5MultiIterNew(p, pStruct, flags, pColset, pToken, nToken, -1, 0, &p1); fts5IterSetOutputCb(&p->rc, p1); + for( /* no-op */ ; fts5MultiIterEof(p, p1)==0; fts5MultiIterNext2(p, p1, &bNewTerm) @@ -6150,7 +6275,6 @@ static void fts5SetupPrefixIter( } if( p1->base.nData==0 ) continue; - if( p1->base.iRowid<=iLastRowid && doclist.n>0 ){ for(i=0; p->rc==SQLITE_OK && doclist.n; i++){ int i1 = i*nMerge; @@ -6189,7 +6313,7 @@ static void fts5SetupPrefixIter( } fts5MultiIterFree(p1); - pData = fts5IdxMalloc(p, sizeof(Fts5Data)+doclist.n+FTS5_DATA_ZERO_PADDING); + pData = fts5IdxMalloc(p, sizeof(*pData)+doclist.n+FTS5_DATA_ZERO_PADDING); if( pData ){ pData->p = (u8*)&pData[1]; pData->nn = pData->szLeaf = doclist.n; @@ -6332,6 +6456,7 @@ int sqlite3Fts5IndexClose(Fts5Index *p){ sqlite3_finalize(p->pIdxWriter); sqlite3_finalize(p->pIdxDeleter); sqlite3_finalize(p->pIdxSelect); + sqlite3_finalize(p->pIdxNextSelect); sqlite3_finalize(p->pDataVersion); sqlite3_finalize(p->pDeleteFromIdx); sqlite3Fts5HashFree(p->pHash); @@ -6427,6 +6552,451 @@ int sqlite3Fts5IndexWrite( return rc; } +/* +** pToken points to a buffer of size nToken bytes containing a search +** term, including the index number at the start, used on a tokendata=1 +** table. This function returns true if the term in buffer pBuf matches +** token pToken/nToken. +*/ +static int fts5IsTokendataPrefix( + Fts5Buffer *pBuf, + const u8 *pToken, + int nToken +){ + return ( + pBuf->n>=nToken + && 0==memcmp(pBuf->p, pToken, nToken) + && (pBuf->n==nToken || pBuf->p[nToken]==0x00) + ); +} + +/* +** Ensure the segment-iterator passed as the only argument points to EOF. +*/ +static void fts5SegIterSetEOF(Fts5SegIter *pSeg){ + fts5DataRelease(pSeg->pLeaf); + pSeg->pLeaf = 0; +} + +/* +** Usually, a tokendata=1 iterator (struct Fts5TokenDataIter) accumulates an +** array of these for each row it visits. Or, for an iterator used by an +** "ORDER BY rank" query, it accumulates an array of these for the entire +** query. +** +** Each instance in the array indicates the iterator (and therefore term) +** associated with position iPos of rowid iRowid. This is used by the +** xInstToken() API. +*/ +struct Fts5TokenDataMap { + i64 iRowid; /* Row this token is located in */ + i64 iPos; /* Position of token */ + int iIter; /* Iterator token was read from */ +}; + +/* +** An object used to supplement Fts5Iter for tokendata=1 iterators. +*/ +struct Fts5TokenDataIter { + int nIter; + int nIterAlloc; + + int nMap; + int nMapAlloc; + Fts5TokenDataMap *aMap; + + Fts5PoslistReader *aPoslistReader; + int *aPoslistToIter; + Fts5Iter *apIter[1]; +}; + +/* +** This function appends iterator pAppend to Fts5TokenDataIter pIn and +** returns the result. +*/ +static Fts5TokenDataIter *fts5AppendTokendataIter( + Fts5Index *p, /* Index object (for error code) */ + Fts5TokenDataIter *pIn, /* Current Fts5TokenDataIter struct */ + Fts5Iter *pAppend /* Append this iterator */ +){ + Fts5TokenDataIter *pRet = pIn; + + if( p->rc==SQLITE_OK ){ + if( pIn==0 || pIn->nIter==pIn->nIterAlloc ){ + int nAlloc = pIn ? pIn->nIterAlloc*2 : 16; + int nByte = nAlloc * sizeof(Fts5Iter*) + sizeof(Fts5TokenDataIter); + Fts5TokenDataIter *pNew = (Fts5TokenDataIter*)sqlite3_realloc(pIn, nByte); + + if( pNew==0 ){ + p->rc = SQLITE_NOMEM; + }else{ + if( pIn==0 ) memset(pNew, 0, nByte); + pRet = pNew; + pNew->nIterAlloc = nAlloc; + } + } + } + if( p->rc ){ + sqlite3Fts5IterClose((Fts5IndexIter*)pAppend); + }else{ + pRet->apIter[pRet->nIter++] = pAppend; + } + assert( pRet==0 || pRet->nIter<=pRet->nIterAlloc ); + + return pRet; +} + +/* +** Delete an Fts5TokenDataIter structure and its contents. +*/ +static void fts5TokendataIterDelete(Fts5TokenDataIter *pSet){ + if( pSet ){ + int ii; + for(ii=0; iinIter; ii++){ + fts5MultiIterFree(pSet->apIter[ii]); + } + sqlite3_free(pSet->aPoslistReader); + sqlite3_free(pSet->aMap); + sqlite3_free(pSet); + } +} + +/* +** Append a mapping to the token-map belonging to object pT. +*/ +static void fts5TokendataIterAppendMap( + Fts5Index *p, + Fts5TokenDataIter *pT, + int iIter, + i64 iRowid, + i64 iPos +){ + if( p->rc==SQLITE_OK ){ + if( pT->nMap==pT->nMapAlloc ){ + int nNew = pT->nMapAlloc ? pT->nMapAlloc*2 : 64; + int nByte = nNew * sizeof(Fts5TokenDataMap); + Fts5TokenDataMap *aNew; + + aNew = (Fts5TokenDataMap*)sqlite3_realloc(pT->aMap, nByte); + if( aNew==0 ){ + p->rc = SQLITE_NOMEM; + return; + } + + pT->aMap = aNew; + pT->nMapAlloc = nNew; + } + + pT->aMap[pT->nMap].iRowid = iRowid; + pT->aMap[pT->nMap].iPos = iPos; + pT->aMap[pT->nMap].iIter = iIter; + pT->nMap++; + } +} + +/* +** The iterator passed as the only argument must be a tokendata=1 iterator +** (pIter->pTokenDataIter!=0). This function sets the iterator output +** variables (pIter->base.*) according to the contents of the current +** row. +*/ +static void fts5IterSetOutputsTokendata(Fts5Iter *pIter){ + int ii; + int nHit = 0; + i64 iRowid = SMALLEST_INT64; + int iMin = 0; + + Fts5TokenDataIter *pT = pIter->pTokenDataIter; + + pIter->base.nData = 0; + pIter->base.pData = 0; + + for(ii=0; iinIter; ii++){ + Fts5Iter *p = pT->apIter[ii]; + if( p->base.bEof==0 ){ + if( nHit==0 || p->base.iRowidbase.iRowid; + nHit = 1; + pIter->base.pData = p->base.pData; + pIter->base.nData = p->base.nData; + iMin = ii; + }else if( p->base.iRowid==iRowid ){ + nHit++; + } + } + } + + if( nHit==0 ){ + pIter->base.bEof = 1; + }else{ + int eDetail = pIter->pIndex->pConfig->eDetail; + pIter->base.bEof = 0; + pIter->base.iRowid = iRowid; + + if( nHit==1 && eDetail==FTS5_DETAIL_FULL ){ + fts5TokendataIterAppendMap(pIter->pIndex, pT, iMin, iRowid, -1); + }else + if( nHit>1 && eDetail!=FTS5_DETAIL_NONE ){ + int nReader = 0; + int nByte = 0; + i64 iPrev = 0; + + /* Allocate array of iterators if they are not already allocated. */ + if( pT->aPoslistReader==0 ){ + pT->aPoslistReader = (Fts5PoslistReader*)sqlite3Fts5MallocZero( + &pIter->pIndex->rc, + pT->nIter * (sizeof(Fts5PoslistReader) + sizeof(int)) + ); + if( pT->aPoslistReader==0 ) return; + pT->aPoslistToIter = (int*)&pT->aPoslistReader[pT->nIter]; + } + + /* Populate an iterator for each poslist that will be merged */ + for(ii=0; iinIter; ii++){ + Fts5Iter *p = pT->apIter[ii]; + if( iRowid==p->base.iRowid ){ + pT->aPoslistToIter[nReader] = ii; + sqlite3Fts5PoslistReaderInit( + p->base.pData, p->base.nData, &pT->aPoslistReader[nReader++] + ); + nByte += p->base.nData; + } + } + + /* Ensure the output buffer is large enough */ + if( fts5BufferGrow(&pIter->pIndex->rc, &pIter->poslist, nByte+nHit*10) ){ + return; + } + + /* Ensure the token-mapping is large enough */ + if( eDetail==FTS5_DETAIL_FULL && pT->nMapAlloc<(pT->nMap + nByte) ){ + int nNew = (pT->nMapAlloc + nByte) * 2; + Fts5TokenDataMap *aNew = (Fts5TokenDataMap*)sqlite3_realloc( + pT->aMap, nNew*sizeof(Fts5TokenDataMap) + ); + if( aNew==0 ){ + pIter->pIndex->rc = SQLITE_NOMEM; + return; + } + pT->aMap = aNew; + pT->nMapAlloc = nNew; + } + + pIter->poslist.n = 0; + + while( 1 ){ + i64 iMinPos = LARGEST_INT64; + + /* Find smallest position */ + iMin = 0; + for(ii=0; iiaPoslistReader[ii]; + if( pReader->bEof==0 ){ + if( pReader->iPosiPos; + iMin = ii; + } + } + } + + /* If all readers were at EOF, break out of the loop. */ + if( iMinPos==LARGEST_INT64 ) break; + + sqlite3Fts5PoslistSafeAppend(&pIter->poslist, &iPrev, iMinPos); + sqlite3Fts5PoslistReaderNext(&pT->aPoslistReader[iMin]); + + if( eDetail==FTS5_DETAIL_FULL ){ + pT->aMap[pT->nMap].iPos = iMinPos; + pT->aMap[pT->nMap].iIter = pT->aPoslistToIter[iMin]; + pT->aMap[pT->nMap].iRowid = iRowid; + pT->nMap++; + } + } + + pIter->base.pData = pIter->poslist.p; + pIter->base.nData = pIter->poslist.n; + } + } +} + +/* +** The iterator passed as the only argument must be a tokendata=1 iterator +** (pIter->pTokenDataIter!=0). This function advances the iterator. If +** argument bFrom is false, then the iterator is advanced to the next +** entry. Or, if bFrom is true, it is advanced to the first entry with +** a rowid of iFrom or greater. +*/ +static void fts5TokendataIterNext(Fts5Iter *pIter, int bFrom, i64 iFrom){ + int ii; + Fts5TokenDataIter *pT = pIter->pTokenDataIter; + + for(ii=0; iinIter; ii++){ + Fts5Iter *p = pT->apIter[ii]; + if( p->base.bEof==0 + && (p->base.iRowid==pIter->base.iRowid || (bFrom && p->base.iRowidpIndex, p, bFrom, iFrom); + while( bFrom && p->base.bEof==0 + && p->base.iRowidpIndex->rc==SQLITE_OK + ){ + fts5MultiIterNext(p->pIndex, p, 0, 0); + } + } + } + + fts5IterSetOutputsTokendata(pIter); +} + +/* +** If the segment-iterator passed as the first argument is at EOF, then +** set pIter->term to a copy of buffer pTerm. +*/ +static void fts5TokendataSetTermIfEof(Fts5Iter *pIter, Fts5Buffer *pTerm){ + if( pIter && pIter->aSeg[0].pLeaf==0 ){ + fts5BufferSet(&pIter->pIndex->rc, &pIter->aSeg[0].term, pTerm->n, pTerm->p); + } +} + +/* +** This function sets up an iterator to use for a non-prefix query on a +** tokendata=1 table. +*/ +static Fts5Iter *fts5SetupTokendataIter( + Fts5Index *p, /* FTS index to query */ + const u8 *pToken, /* Buffer containing query term */ + int nToken, /* Size of buffer pToken in bytes */ + Fts5Colset *pColset /* Colset to filter on */ +){ + Fts5Iter *pRet = 0; + Fts5TokenDataIter *pSet = 0; + Fts5Structure *pStruct = 0; + const int flags = FTS5INDEX_QUERY_SCANONETERM | FTS5INDEX_QUERY_SCAN; + + Fts5Buffer bSeek = {0, 0, 0}; + Fts5Buffer *pSmall = 0; + + fts5IndexFlush(p); + pStruct = fts5StructureRead(p); + + while( 1 ){ + Fts5Iter *pPrev = pSet ? pSet->apIter[pSet->nIter-1] : 0; + Fts5Iter *pNew = 0; + Fts5SegIter *pNewIter = 0; + Fts5SegIter *pPrevIter = 0; + + int iLvl, iSeg, ii; + + pNew = fts5MultiIterAlloc(p, pStruct->nSegment); + if( pNew==0 ) break; + + if( pSmall ){ + fts5BufferSet(&p->rc, &bSeek, pSmall->n, pSmall->p); + fts5BufferAppendBlob(&p->rc, &bSeek, 1, (const u8*)"\0"); + }else{ + fts5BufferSet(&p->rc, &bSeek, nToken, pToken); + } + + pNewIter = &pNew->aSeg[0]; + pPrevIter = (pPrev ? &pPrev->aSeg[0] : 0); + for(iLvl=0; iLvlnLevel; iLvl++){ + for(iSeg=pStruct->aLevel[iLvl].nSeg-1; iSeg>=0; iSeg--){ + Fts5StructureSegment *pSeg = &pStruct->aLevel[iLvl].aSeg[iSeg]; + int bDone = 0; + + if( pPrevIter ){ + if( fts5BufferCompare(pSmall, &pPrevIter->term) ){ + memcpy(pNewIter, pPrevIter, sizeof(Fts5SegIter)); + memset(pPrevIter, 0, sizeof(Fts5SegIter)); + bDone = 1; + }else if( pPrevIter->iEndofDoclist>pPrevIter->pLeaf->szLeaf ){ + fts5SegIterNextInit(p,(const char*)bSeek.p,bSeek.n-1,pSeg,pNewIter); + bDone = 1; + } + } + + if( bDone==0 ){ + fts5SegIterSeekInit(p, bSeek.p, bSeek.n, flags, pSeg, pNewIter); + } + + if( pPrevIter ){ + if( pPrevIter->pTombArray ){ + pNewIter->pTombArray = pPrevIter->pTombArray; + pNewIter->pTombArray->nRef++; + } + }else{ + fts5SegIterAllocTombstone(p, pNewIter); + } + + pNewIter++; + if( pPrevIter ) pPrevIter++; + } + } + fts5TokendataSetTermIfEof(pPrev, pSmall); + + pNew->bSkipEmpty = 1; + pNew->pColset = pColset; + fts5IterSetOutputCb(&p->rc, pNew); + + /* Loop through all segments in the new iterator. Find the smallest + ** term that any segment-iterator points to. Iterator pNew will be + ** used for this term. Also, set any iterator that points to a term that + ** does not match pToken/nToken to point to EOF */ + pSmall = 0; + for(ii=0; iinSeg; ii++){ + Fts5SegIter *pII = &pNew->aSeg[ii]; + if( 0==fts5IsTokendataPrefix(&pII->term, pToken, nToken) ){ + fts5SegIterSetEOF(pII); + } + if( pII->pLeaf && (!pSmall || fts5BufferCompare(pSmall, &pII->term)>0) ){ + pSmall = &pII->term; + } + } + + /* If pSmall is still NULL at this point, then the new iterator does + ** not point to any terms that match the query. So delete it and break + ** out of the loop - all required iterators have been collected. */ + if( pSmall==0 ){ + sqlite3Fts5IterClose((Fts5IndexIter*)pNew); + break; + } + + /* Append this iterator to the set and continue. */ + pSet = fts5AppendTokendataIter(p, pSet, pNew); + } + + if( p->rc==SQLITE_OK && pSet ){ + int ii; + for(ii=0; iinIter; ii++){ + Fts5Iter *pIter = pSet->apIter[ii]; + int iSeg; + for(iSeg=0; iSegnSeg; iSeg++){ + pIter->aSeg[iSeg].flags |= FTS5_SEGITER_ONETERM; + } + fts5MultiIterFinishSetup(p, pIter); + } + } + + if( p->rc==SQLITE_OK ){ + pRet = fts5MultiIterAlloc(p, 0); + } + if( pRet ){ + pRet->pTokenDataIter = pSet; + if( pSet ){ + fts5IterSetOutputsTokendata(pRet); + }else{ + pRet->base.bEof = 1; + } + }else{ + fts5TokendataIterDelete(pSet); + } + + fts5StructureRelease(pStruct); + fts5BufferFree(&bSeek); + return pRet; +} + + /* ** Open a new iterator to iterate though all rowid that match the ** specified token or token prefix. @@ -6448,6 +7018,7 @@ int sqlite3Fts5IndexQuery( if( sqlite3Fts5BufferSize(&p->rc, &buf, nToken+1)==0 ){ int iIdx = 0; /* Index to search */ int iPrefixIdx = 0; /* +1 prefix index */ + int bTokendata = pConfig->bTokendata; if( nToken>0 ) memcpy(&buf.p[1], pToken, nToken); /* Figure out which index to search and set iIdx accordingly. If this @@ -6461,6 +7032,7 @@ int sqlite3Fts5IndexQuery( ** for internal sanity checking by the integrity-check in debug ** mode only. */ #ifdef SQLITE_DEBUG + if( flags & FTS5INDEX_QUERY_NOTOKENDATA ) bTokendata = 0; if( pConfig->bPrefixIndex==0 || (flags & FTS5INDEX_QUERY_TEST_NOIDX) ){ assert( flags & FTS5INDEX_QUERY_PREFIX ); iIdx = 1+pConfig->nPrefix; @@ -6475,7 +7047,10 @@ int sqlite3Fts5IndexQuery( } } - if( iIdx<=pConfig->nPrefix ){ + if( bTokendata && iIdx==0 ){ + buf.p[0] = '0'; + pRet = fts5SetupTokendataIter(p, buf.p, nToken+1, pColset); + }else if( iIdx<=pConfig->nPrefix ){ /* Straight index lookup */ Fts5Structure *pStruct = fts5StructureRead(p); buf.p[0] = (u8)(FTS5_MAIN_PREFIX + iIdx); @@ -6522,7 +7097,11 @@ int sqlite3Fts5IndexQuery( int sqlite3Fts5IterNext(Fts5IndexIter *pIndexIter){ Fts5Iter *pIter = (Fts5Iter*)pIndexIter; assert( pIter->pIndex->rc==SQLITE_OK ); - fts5MultiIterNext(pIter->pIndex, pIter, 0, 0); + if( pIter->pTokenDataIter ){ + fts5TokendataIterNext(pIter, 0, 0); + }else{ + fts5MultiIterNext(pIter->pIndex, pIter, 0, 0); + } return fts5IndexReturn(pIter->pIndex); } @@ -6555,7 +7134,11 @@ int sqlite3Fts5IterNextScan(Fts5IndexIter *pIndexIter){ */ int sqlite3Fts5IterNextFrom(Fts5IndexIter *pIndexIter, i64 iMatch){ Fts5Iter *pIter = (Fts5Iter*)pIndexIter; - fts5MultiIterNextFrom(pIter->pIndex, pIter, iMatch); + if( pIter->pTokenDataIter ){ + fts5TokendataIterNext(pIter, 1, iMatch); + }else{ + fts5MultiIterNextFrom(pIter->pIndex, pIter, iMatch); + } return fts5IndexReturn(pIter->pIndex); } @@ -6570,6 +7153,99 @@ const char *sqlite3Fts5IterTerm(Fts5IndexIter *pIndexIter, int *pn){ return (z ? &z[1] : 0); } +/* +** This is used by xInstToken() to access the token at offset iOff, column +** iCol of row iRowid. The token is returned via output variables *ppOut +** and *pnOut. The iterator passed as the first argument must be a tokendata=1 +** iterator (pIter->pTokenDataIter!=0). +*/ +int sqlite3Fts5IterToken( + Fts5IndexIter *pIndexIter, + i64 iRowid, + int iCol, + int iOff, + const char **ppOut, int *pnOut +){ + Fts5Iter *pIter = (Fts5Iter*)pIndexIter; + Fts5TokenDataIter *pT = pIter->pTokenDataIter; + Fts5TokenDataMap *aMap = pT->aMap; + i64 iPos = (((i64)iCol)<<32) + iOff; + + int i1 = 0; + int i2 = pT->nMap; + int iTest = 0; + + while( i2>i1 ){ + iTest = (i1 + i2) / 2; + + if( aMap[iTest].iRowidiRowid ){ + i2 = iTest; + }else{ + if( aMap[iTest].iPosiPos ){ + i2 = iTest; + }else{ + break; + } + } + } + + if( i2>i1 ){ + Fts5Iter *pMap = pT->apIter[aMap[iTest].iIter]; + *ppOut = (const char*)pMap->aSeg[0].term.p+1; + *pnOut = pMap->aSeg[0].term.n-1; + } + + return SQLITE_OK; +} + +/* +** Clear any existing entries from the token-map associated with the +** iterator passed as the only argument. +*/ +void sqlite3Fts5IndexIterClearTokendata(Fts5IndexIter *pIndexIter){ + Fts5Iter *pIter = (Fts5Iter*)pIndexIter; + if( pIter->pTokenDataIter ){ + pIter->pTokenDataIter->nMap = 0; + } +} + +/* +** Set a token-mapping for the iterator passed as the first argument. This +** is used in detail=column or detail=none mode when a token is requested +** using the xInstToken() API. In this case the caller tokenizers the +** current row and configures the token-mapping via multiple calls to this +** function. +*/ +int sqlite3Fts5IndexIterWriteTokendata( + Fts5IndexIter *pIndexIter, + const char *pToken, int nToken, + i64 iRowid, int iCol, int iOff +){ + Fts5Iter *pIter = (Fts5Iter*)pIndexIter; + Fts5TokenDataIter *pT = pIter->pTokenDataIter; + Fts5Index *p = pIter->pIndex; + int ii; + + assert( p->pConfig->eDetail!=FTS5_DETAIL_FULL ); + assert( pIter->pTokenDataIter ); + + for(ii=0; iinIter; ii++){ + Fts5Buffer *pTerm = &pT->apIter[ii]->aSeg[0].term; + if( nToken==pTerm->n-1 && memcmp(pToken, pTerm->p+1, nToken)==0 ) break; + } + if( iinIter ){ + fts5TokendataIterAppendMap(p, pT, ii, iRowid, (((i64)iCol)<<32) + iOff); + } + return fts5IndexReturn(p); +} + /* ** Close an iterator opened by an earlier call to sqlite3Fts5IndexQuery(). */ @@ -6577,6 +7253,7 @@ void sqlite3Fts5IterClose(Fts5IndexIter *pIndexIter){ if( pIndexIter ){ Fts5Iter *pIter = (Fts5Iter*)pIndexIter; Fts5Index *pIndex = pIter->pIndex; + fts5TokendataIterDelete(pIter->pTokenDataIter); fts5MultiIterFree(pIter); sqlite3Fts5IndexCloseReader(pIndex); } @@ -7084,7 +7761,9 @@ static int fts5QueryCksum( int eDetail = p->pConfig->eDetail; u64 cksum = *pCksum; Fts5IndexIter *pIter = 0; - int rc = sqlite3Fts5IndexQuery(p, z, n, flags, 0, &pIter); + int rc = sqlite3Fts5IndexQuery( + p, z, n, (flags | FTS5INDEX_QUERY_NOTOKENDATA), 0, &pIter + ); while( rc==SQLITE_OK && ALWAYS(pIter!=0) && 0==sqlite3Fts5IterEof(pIter) ){ i64 rowid = pIter->iRowid; @@ -7782,6 +8461,24 @@ static void fts5DecodeRowidList( } #endif /* SQLITE_TEST || SQLITE_FTS5_DEBUG */ +#if defined(SQLITE_TEST) || defined(SQLITE_FTS5_DEBUG) +static void fts5BufferAppendTerm(int *pRc, Fts5Buffer *pBuf, Fts5Buffer *pTerm){ + int ii; + fts5BufferGrow(pRc, pBuf, pTerm->n*2 + 1); + if( *pRc==SQLITE_OK ){ + for(ii=0; iin; ii++){ + if( pTerm->p[ii]==0x00 ){ + pBuf->p[pBuf->n++] = '\\'; + pBuf->p[pBuf->n++] = '0'; + }else{ + pBuf->p[pBuf->n++] = pTerm->p[ii]; + } + } + pBuf->p[pBuf->n] = 0x00; + } +} +#endif /* SQLITE_TEST || SQLITE_FTS5_DEBUG */ + #if defined(SQLITE_TEST) || defined(SQLITE_FTS5_DEBUG) /* ** The implementation of user-defined scalar function fts5_decode(). @@ -7889,9 +8586,8 @@ static void fts5DecodeFunction( iOff += fts5GetVarint32(&a[iOff], nAppend); term.n = nKeep; fts5BufferAppendBlob(&rc, &term, nAppend, &a[iOff]); - sqlite3Fts5BufferAppendPrintf( - &rc, &s, " term=%.*s", term.n, (const char*)term.p - ); + sqlite3Fts5BufferAppendPrintf(&rc, &s, " term="); + fts5BufferAppendTerm(&rc, &s, &term); iOff += nAppend; /* Figure out where the doclist for this term ends */ @@ -7999,9 +8695,8 @@ static void fts5DecodeFunction( fts5BufferAppendBlob(&rc, &term, nByte, &a[iOff]); iOff += nByte; - sqlite3Fts5BufferAppendPrintf( - &rc, &s, " term=%.*s", term.n, (const char*)term.p - ); + sqlite3Fts5BufferAppendPrintf(&rc, &s, " term="); + fts5BufferAppendTerm(&rc, &s, &term); iOff += fts5DecodeDoclist(&rc, &s, &a[iOff], iEnd-iOff); } diff --git a/ext/fts5/fts5_main.c b/ext/fts5/fts5_main.c index 6e86ca5951..d35e998da0 100644 --- a/ext/fts5/fts5_main.c +++ b/ext/fts5/fts5_main.c @@ -656,12 +656,15 @@ static int fts5BestIndexMethod(sqlite3_vtab *pVTab, sqlite3_index_info *pInfo){ } idxStr[iIdxStr] = '\0'; - /* Set idxFlags flags for the ORDER BY clause */ + /* Set idxFlags flags for the ORDER BY clause + ** + ** Note that tokendata=1 tables cannot currently handle "ORDER BY rowid DESC". + */ if( pInfo->nOrderBy==1 ){ int iSort = pInfo->aOrderBy[0].iColumn; if( iSort==(pConfig->nCol+1) && bSeenMatch ){ idxFlags |= FTS5_BI_ORDER_RANK; - }else if( iSort==-1 ){ + }else if( iSort==-1 && (!pInfo->aOrderBy[0].desc || !pConfig->bTokendata) ){ idxFlags |= FTS5_BI_ORDER_ROWID; } if( BitFlagTest(idxFlags, FTS5_BI_ORDER_RANK|FTS5_BI_ORDER_ROWID) ){ @@ -913,6 +916,16 @@ static int fts5NextMethod(sqlite3_vtab_cursor *pCursor){ ); assert( !CsrFlagTest(pCsr, FTS5CSR_EOF) ); + /* If this cursor uses FTS5_PLAN_MATCH and this is a tokendata=1 table, + ** clear any token mappings accumulated at the fts5_index.c level. In + ** other cases, specifically FTS5_PLAN_SOURCE and FTS5_PLAN_SORTED_MATCH, + ** we need to retain the mappings for the entire query. */ + if( pCsr->ePlan==FTS5_PLAN_MATCH + && ((Fts5Table*)pCursor->pVtab)->pConfig->bTokendata + ){ + sqlite3Fts5ExprClearTokens(pCsr->pExpr); + } + if( pCsr->ePlan<3 ){ int bSkip = 0; if( (rc = fts5CursorReseek(pCsr, &bSkip)) || bSkip ) return rc; @@ -2063,12 +2076,6 @@ static int fts5ApiInst( ){ if( iIdx<0 || iIdx>=pCsr->nInstCount ){ rc = SQLITE_RANGE; -#if 0 - }else if( fts5IsOffsetless((Fts5Table*)pCsr->base.pVtab) ){ - *piPhrase = pCsr->aInst[iIdx*3]; - *piCol = pCsr->aInst[iIdx*3 + 2]; - *piOff = -1; -#endif }else{ *piPhrase = pCsr->aInst[iIdx*3]; *piCol = pCsr->aInst[iIdx*3 + 1]; @@ -2323,13 +2330,56 @@ static int fts5ApiPhraseFirstColumn( return rc; } +/* +** xQueryToken() API implemenetation. +*/ +static int fts5ApiQueryToken( + Fts5Context* pCtx, + int iPhrase, + int iToken, + const char **ppOut, + int *pnOut +){ + Fts5Cursor *pCsr = (Fts5Cursor*)pCtx; + return sqlite3Fts5ExprQueryToken(pCsr->pExpr, iPhrase, iToken, ppOut, pnOut); +} + +/* +** xInstToken() API implemenetation. +*/ +static int fts5ApiInstToken( + Fts5Context *pCtx, + int iIdx, + int iToken, + const char **ppOut, int *pnOut +){ + Fts5Cursor *pCsr = (Fts5Cursor*)pCtx; + int rc = SQLITE_OK; + if( CsrFlagTest(pCsr, FTS5CSR_REQUIRE_INST)==0 + || SQLITE_OK==(rc = fts5CacheInstArray(pCsr)) + ){ + if( iIdx<0 || iIdx>=pCsr->nInstCount ){ + rc = SQLITE_RANGE; + }else{ + int iPhrase = pCsr->aInst[iIdx*3]; + int iCol = pCsr->aInst[iIdx*3 + 1]; + int iOff = pCsr->aInst[iIdx*3 + 2]; + i64 iRowid = fts5CursorRowid(pCsr); + rc = sqlite3Fts5ExprInstToken( + pCsr->pExpr, iRowid, iPhrase, iCol, iOff, iToken, ppOut, pnOut + ); + } + } + return rc; +} + static int fts5ApiQueryPhrase(Fts5Context*, int, void*, int(*)(const Fts5ExtensionApi*, Fts5Context*, void*) ); static const Fts5ExtensionApi sFts5Api = { - 2, /* iVersion */ + 3, /* iVersion */ fts5ApiUserData, fts5ApiColumnCount, fts5ApiRowCount, @@ -2349,6 +2399,8 @@ static const Fts5ExtensionApi sFts5Api = { fts5ApiPhraseNext, fts5ApiPhraseFirstColumn, fts5ApiPhraseNextColumn, + fts5ApiQueryToken, + fts5ApiInstToken }; /* diff --git a/ext/fts5/fts5_tcl.c b/ext/fts5/fts5_tcl.c index 80c600dbb1..853a41865e 100644 --- a/ext/fts5/fts5_tcl.c +++ b/ext/fts5/fts5_tcl.c @@ -244,6 +244,9 @@ static int SQLITE_TCLAPI xF5tApi( { "xGetAuxdataInt", 1, "CLEAR" }, /* 15 */ { "xPhraseForeach", 4, "IPHRASE COLVAR OFFVAR SCRIPT" }, /* 16 */ { "xPhraseColumnForeach", 3, "IPHRASE COLVAR SCRIPT" }, /* 17 */ + + { "xQueryToken", 2, "IPHRASE ITERM" }, /* 18 */ + { "xInstToken", 2, "IDX ITERM" }, /* 19 */ { 0, 0, 0} }; @@ -500,6 +503,38 @@ static int SQLITE_TCLAPI xF5tApi( break; } + CASE(18, "xQueryToken") { + const char *pTerm = 0; + int nTerm = 0; + int iPhrase = 0; + int iTerm = 0; + + if( Tcl_GetIntFromObj(interp, objv[2], &iPhrase) ) return TCL_ERROR; + if( Tcl_GetIntFromObj(interp, objv[3], &iTerm) ) return TCL_ERROR; + rc = p->pApi->xQueryToken(p->pFts, iPhrase, iTerm, &pTerm, &nTerm); + if( rc==SQLITE_OK ){ + Tcl_SetObjResult(interp, Tcl_NewStringObj(pTerm, nTerm)); + } + + break; + } + + CASE(19, "xInstToken") { + const char *pTerm = 0; + int nTerm = 0; + int iIdx = 0; + int iTerm = 0; + + if( Tcl_GetIntFromObj(interp, objv[2], &iIdx) ) return TCL_ERROR; + if( Tcl_GetIntFromObj(interp, objv[3], &iTerm) ) return TCL_ERROR; + rc = p->pApi->xInstToken(p->pFts, iIdx, iTerm, &pTerm, &nTerm); + if( rc==SQLITE_OK ){ + Tcl_SetObjResult(interp, Tcl_NewStringObj(pTerm, nTerm)); + } + + break; + } + default: assert( 0 ); break; @@ -1117,6 +1152,176 @@ static int SQLITE_TCLAPI f5tRegisterTok( return TCL_OK; } +typedef struct OriginTextCtx OriginTextCtx; +struct OriginTextCtx { + sqlite3 *db; + fts5_api *pApi; +}; + +typedef struct OriginTextTokenizer OriginTextTokenizer; +struct OriginTextTokenizer { + Fts5Tokenizer *pTok; /* Underlying tokenizer object */ + fts5_tokenizer tokapi; /* API implementation for pTok */ +}; + +/* +** Delete the OriginTextCtx object indicated by the only argument. +*/ +static void f5tOrigintextTokenizerDelete(void *pCtx){ + OriginTextCtx *p = (OriginTextCtx*)pCtx; + ckfree(p); +} + +static int f5tOrigintextCreate( + void *pCtx, + const char **azArg, + int nArg, + Fts5Tokenizer **ppOut +){ + OriginTextCtx *p = (OriginTextCtx*)pCtx; + OriginTextTokenizer *pTok = 0; + void *pTokCtx = 0; + int rc = SQLITE_OK; + + pTok = (OriginTextTokenizer*)sqlite3_malloc(sizeof(OriginTextTokenizer)); + if( pTok==0 ){ + rc = SQLITE_NOMEM; + }else if( nArg<1 ){ + rc = SQLITE_ERROR; + }else{ + /* Locate the underlying tokenizer */ + rc = p->pApi->xFindTokenizer(p->pApi, azArg[0], &pTokCtx, &pTok->tokapi); + } + + /* Create the new tokenizer instance */ + if( rc==SQLITE_OK ){ + rc = pTok->tokapi.xCreate(pTokCtx, &azArg[1], nArg-1, &pTok->pTok); + } + + if( rc!=SQLITE_OK ){ + sqlite3_free(pTok); + pTok = 0; + } + *ppOut = (Fts5Tokenizer*)pTok; + return rc; +} + +static void f5tOrigintextDelete(Fts5Tokenizer *pTokenizer){ + OriginTextTokenizer *p = (OriginTextTokenizer*)pTokenizer; + if( p->pTok ){ + p->tokapi.xDelete(p->pTok); + } + sqlite3_free(p); +} + +typedef struct OriginTextCb OriginTextCb; +struct OriginTextCb { + void *pCtx; + const char *pText; + int nText; + int (*xToken)(void *, int, const char *, int, int, int); + + char *aBuf; /* Buffer to use */ + int nBuf; /* Allocated size of aBuf[] */ +}; + +static int xOriginToken( + void *pCtx, /* Copy of 2nd argument to xTokenize() */ + int tflags, /* Mask of FTS5_TOKEN_* flags */ + const char *pToken, /* Pointer to buffer containing token */ + int nToken, /* Size of token in bytes */ + int iStart, /* Byte offset of token within input text */ + int iEnd /* Byte offset of end of token within input */ +){ + OriginTextCb *p = (OriginTextCb*)pCtx; + int ret = 0; + + if( nToken==(iEnd-iStart) && 0==memcmp(pToken, &p->pText[iStart], nToken) ){ + /* Token exactly matches document text. Pass it through as is. */ + ret = p->xToken(p->pCtx, tflags, pToken, nToken, iStart, iEnd); + }else{ + int nReq = nToken + 1 + (iEnd-iStart); + if( nReq>p->nBuf ){ + sqlite3_free(p->aBuf); + p->aBuf = sqlite3_malloc(nReq*2); + if( p->aBuf==0 ) return SQLITE_NOMEM; + p->nBuf = nReq*2; + } + + memcpy(p->aBuf, pToken, nToken); + p->aBuf[nToken] = '\0'; + memcpy(&p->aBuf[nToken+1], &p->pText[iStart], iEnd-iStart); + ret = p->xToken(p->pCtx, tflags, p->aBuf, nReq, iStart, iEnd); + } + + return ret; +} + + +static int f5tOrigintextTokenize( + Fts5Tokenizer *pTokenizer, + void *pCtx, + int flags, /* Mask of FTS5_TOKENIZE_* flags */ + const char *pText, int nText, + int (*xToken)(void *, int, const char *, int, int, int) +){ + OriginTextTokenizer *p = (OriginTextTokenizer*)pTokenizer; + OriginTextCb cb; + int ret; + + memset(&cb, 0, sizeof(cb)); + cb.pCtx = pCtx; + cb.pText = pText; + cb.nText = nText; + cb.xToken = xToken; + + ret = p->tokapi.xTokenize(p->pTok,(void*)&cb,flags,pText,nText,xOriginToken); + sqlite3_free(cb.aBuf); + return ret; +} + +/* +** sqlite3_fts5_register_origintext DB +** +** Description... +*/ +static int SQLITE_TCLAPI f5tRegisterOriginText( + void * clientData, + Tcl_Interp *interp, + int objc, + Tcl_Obj *CONST objv[] +){ + sqlite3 *db = 0; + fts5_api *pApi = 0; + int rc; + fts5_tokenizer tok = {0, 0, 0}; + OriginTextCtx *pCtx = 0; + + if( objc!=2 ){ + Tcl_WrongNumArgs(interp, 1, objv, "DB"); + return TCL_ERROR; + } + if( f5tDbAndApi(interp, objv[1], &db, &pApi) ) return TCL_ERROR; + + pCtx = (OriginTextCtx*)ckalloc(sizeof(OriginTextCtx)); + pCtx->db = db; + pCtx->pApi = pApi; + + tok.xCreate = f5tOrigintextCreate; + tok.xDelete = f5tOrigintextDelete; + tok.xTokenize = f5tOrigintextTokenize; + rc = pApi->xCreateTokenizer( + pApi, "origintext", (void*)pCtx, &tok, f5tOrigintextTokenizerDelete + ); + + Tcl_ResetResult(interp); + if( rc!=SQLITE_OK ){ + Tcl_AppendResult(interp, "error: ", sqlite3_errmsg(db), 0); + return TCL_ERROR; + } + return TCL_OK; +} + /* ** Entry point. */ @@ -1133,7 +1338,8 @@ int Fts5tcl_Init(Tcl_Interp *interp){ { "sqlite3_fts5_may_be_corrupt", f5tMayBeCorrupt, 0 }, { "sqlite3_fts5_token_hash", f5tTokenHash, 0 }, { "sqlite3_fts5_register_matchinfo", f5tRegisterMatchinfo, 0 }, - { "sqlite3_fts5_register_fts5tokenize", f5tRegisterTok, 0 } + { "sqlite3_fts5_register_fts5tokenize", f5tRegisterTok, 0 }, + { "sqlite3_fts5_register_origintext",f5tRegisterOriginText, 0 } }; int i; F5tTokenizerContext *pContext; diff --git a/ext/fts5/test/fts5_common.tcl b/ext/fts5/test/fts5_common.tcl index 9c012932da..001cad1de2 100644 --- a/ext/fts5/test/fts5_common.tcl +++ b/ext/fts5/test/fts5_common.tcl @@ -438,6 +438,20 @@ proc detail_is_none {} { detail_check ; expr {$::detail == "none"} } proc detail_is_col {} { detail_check ; expr {$::detail == "col" } } proc detail_is_full {} { detail_check ; expr {$::detail == "full"} } +proc foreach_tokenizer_mode {prefix script} { + set saved $::testprefix + foreach {d mapping} { + "" {} + "-origintext" {, tokenize="origintext unicode61", tokendata=1} + } { + set s [string map [list %TOKENIZER% $mapping] $script] + set ::testprefix "$prefix$d" + reset_db + sqlite3_fts5_register_origintext db + uplevel $s + } + set ::testprefix $saved +} #------------------------------------------------------------------------- # Convert a poslist of the type returned by fts5_test_poslist() to a diff --git a/ext/fts5/test/fts5aa.test b/ext/fts5/test/fts5aa.test index e1551fc516..a80a307a49 100644 --- a/ext/fts5/test/fts5aa.test +++ b/ext/fts5/test/fts5aa.test @@ -22,6 +22,7 @@ ifcapable !fts5 { } foreach_detail_mode $::testprefix { +foreach_tokenizer_mode $::testprefix { do_execsql_test 1.0 { CREATE VIRTUAL TABLE t1 USING fts5(a, b, c); @@ -44,7 +45,7 @@ do_execsql_test 1.1 { # do_execsql_test 2.0 { - CREATE VIRTUAL TABLE t1 USING fts5(x, y, detail=%DETAIL%); + CREATE VIRTUAL TABLE t1 USING fts5(x, y, detail=%DETAIL% %TOKENIZER%); } do_execsql_test 2.1 { INSERT INTO t1 VALUES('a b c', 'd e f'); @@ -73,8 +74,9 @@ do_execsql_test 2.4 { #------------------------------------------------------------------------- # reset_db +sqlite3_fts5_register_origintext db do_execsql_test 3.0 { - CREATE VIRTUAL TABLE t1 USING fts5(x,y, detail=%DETAIL%); + CREATE VIRTUAL TABLE t1 USING fts5(x,y, detail=%DETAIL% %TOKENIZER%); } foreach {i x y} { 1 {g f d b f} {h h e i a} @@ -97,8 +99,9 @@ foreach {i x y} { #------------------------------------------------------------------------- # reset_db +sqlite3_fts5_register_origintext db do_execsql_test 4.0 { - CREATE VIRTUAL TABLE t1 USING fts5(x,y, detail=%DETAIL%); + CREATE VIRTUAL TABLE t1 USING fts5(x,y, detail=%DETAIL% %TOKENIZER%); INSERT INTO t1(t1, rank) VALUES('pgsz', 32); } foreach {i x y} { @@ -121,8 +124,9 @@ foreach {i x y} { #------------------------------------------------------------------------- # reset_db +sqlite3_fts5_register_origintext db do_execsql_test 5.0 { - CREATE VIRTUAL TABLE t1 USING fts5(x,y, detail=%DETAIL%); + CREATE VIRTUAL TABLE t1 USING fts5(x,y, detail=%DETAIL% %TOKENIZER%); INSERT INTO t1(t1, rank) VALUES('pgsz', 32); } foreach {i x y} { @@ -145,8 +149,9 @@ foreach {i x y} { #------------------------------------------------------------------------- # reset_db +sqlite3_fts5_register_origintext db do_execsql_test 6.0 { - CREATE VIRTUAL TABLE t1 USING fts5(x,y, detail=%DETAIL%); + CREATE VIRTUAL TABLE t1 USING fts5(x,y, detail=%DETAIL% %TOKENIZER%); INSERT INTO t1(t1, rank) VALUES('pgsz', 32); } @@ -181,6 +186,7 @@ do_execsql_test 6.6 { #------------------------------------------------------------------------- # reset_db +sqlite3_fts5_register_origintext db expr srand(0) do_execsql_test 7.0 { CREATE VIRTUAL TABLE t1 USING fts5(x,y,z); @@ -222,6 +228,7 @@ for {set i 1} {$i <= 10} {incr i} { #------------------------------------------------------------------------- # reset_db +sqlite3_fts5_register_origintext db do_execsql_test 8.0 { CREATE VIRTUAL TABLE t1 USING fts5(x, prefix="1,2,3"); INSERT INTO t1(t1, rank) VALUES('pgsz', 32); @@ -236,6 +243,7 @@ do_execsql_test 8.1 { #------------------------------------------------------------------------- # reset_db +sqlite3_fts5_register_origintext db expr srand(0) @@ -280,8 +288,9 @@ for {set i 1} {$i <= 10} {incr i} { #------------------------------------------------------------------------- # reset_db +sqlite3_fts5_register_origintext db do_execsql_test 10.0 { - CREATE VIRTUAL TABLE t1 USING fts5(x,y, detail=%DETAIL%); + CREATE VIRTUAL TABLE t1 USING fts5(x,y, detail=%DETAIL% %TOKENIZER%); } set d10 { 1 {g f d b f} {h h e i a} @@ -314,19 +323,19 @@ do_execsql_test 10.4.2 { INSERT INTO t1(t1) VALUES('integrity-check') } #------------------------------------------------------------------------- # do_catchsql_test 11.1 { - CREATE VIRTUAL TABLE t2 USING fts5(a, b, c, rank, detail=%DETAIL%); + CREATE VIRTUAL TABLE t2 USING fts5(a, b, c, rank, detail=%DETAIL% %TOKENIZER%); } {1 {reserved fts5 column name: rank}} do_catchsql_test 11.2 { - CREATE VIRTUAL TABLE rank USING fts5(a, b, c, detail=%DETAIL%); + CREATE VIRTUAL TABLE rank USING fts5(a, b, c, detail=%DETAIL% %TOKENIZER%); } {1 {reserved fts5 table name: rank}} do_catchsql_test 11.3 { - CREATE VIRTUAL TABLE t2 USING fts5(a, b, c, rowid, detail=%DETAIL%); + CREATE VIRTUAL TABLE t2 USING fts5(a, b, c, rowid, detail=%DETAIL% %TOKENIZER%); } {1 {reserved fts5 column name: rowid}} #------------------------------------------------------------------------- # do_execsql_test 12.1 { - CREATE VIRTUAL TABLE t2 USING fts5(x,y, detail=%DETAIL%); + CREATE VIRTUAL TABLE t2 USING fts5(x,y, detail=%DETAIL% %TOKENIZER%); } {} do_catchsql_test 12.2 { @@ -341,8 +350,9 @@ do_test 12.3 { #------------------------------------------------------------------------- # reset_db +sqlite3_fts5_register_origintext db do_execsql_test 13.1 { - CREATE VIRTUAL TABLE t1 USING fts5(x, detail=%DETAIL%); + CREATE VIRTUAL TABLE t1 USING fts5(x, detail=%DETAIL% %TOKENIZER%); INSERT INTO t1(rowid, x) VALUES(1, 'o n e'), (2, 't w o'); } {} @@ -365,8 +375,9 @@ do_execsql_test 13.6 { #------------------------------------------------------------------------- # reset_db +sqlite3_fts5_register_origintext db do_execsql_test 14.1 { - CREATE VIRTUAL TABLE t1 USING fts5(x, y, detail=%DETAIL%); + CREATE VIRTUAL TABLE t1 USING fts5(x, y, detail=%DETAIL% %TOKENIZER%); INSERT INTO t1(t1, rank) VALUES('pgsz', 32); WITH d(x,y) AS ( SELECT NULL, 'xyz xyz xyz xyz xyz xyz' @@ -449,8 +460,9 @@ do_catchsql_test 16.2 { #------------------------------------------------------------------------- # reset_db +sqlite3_fts5_register_origintext db do_execsql_test 17.1 { - CREATE VIRTUAL TABLE b2 USING fts5(x, detail=%DETAIL%); + CREATE VIRTUAL TABLE b2 USING fts5(x, detail=%DETAIL% %TOKENIZER%); INSERT INTO b2 VALUES('a'); INSERT INTO b2 VALUES('b'); INSERT INTO b2 VALUES('c'); @@ -466,8 +478,9 @@ do_test 17.2 { if {[string match n* %DETAIL%]==0} { reset_db + sqlite3_fts5_register_origintext db do_execsql_test 17.3 { - CREATE VIRTUAL TABLE c2 USING fts5(x, y, detail=%DETAIL%); + CREATE VIRTUAL TABLE c2 USING fts5(x, y, detail=%DETAIL% %TOKENIZER%); INSERT INTO c2 VALUES('x x x', 'x x x'); SELECT rowid FROM c2 WHERE c2 MATCH 'y:x'; } {1} @@ -476,8 +489,9 @@ if {[string match n* %DETAIL%]==0} { #------------------------------------------------------------------------- # reset_db +sqlite3_fts5_register_origintext db do_execsql_test 17.1 { - CREATE VIRTUAL TABLE uio USING fts5(ttt, detail=%DETAIL%); + CREATE VIRTUAL TABLE uio USING fts5(ttt, detail=%DETAIL% %TOKENIZER%); INSERT INTO uio VALUES(NULL); INSERT INTO uio SELECT NULL FROM uio; INSERT INTO uio SELECT NULL FROM uio; @@ -524,8 +538,8 @@ do_execsql_test 17.9 { #-------------------------------------------------------------------- # do_execsql_test 18.1 { - CREATE VIRTUAL TABLE t1 USING fts5(a, b, detail=%DETAIL%); - CREATE VIRTUAL TABLE t2 USING fts5(c, d, detail=%DETAIL%); + CREATE VIRTUAL TABLE t1 USING fts5(a, b, detail=%DETAIL% %TOKENIZER%); + CREATE VIRTUAL TABLE t2 USING fts5(c, d, detail=%DETAIL% %TOKENIZER%); INSERT INTO t1 VALUES('abc*', NULL); INSERT INTO t2 VALUES(1, 'abcdefg'); } @@ -540,8 +554,9 @@ do_execsql_test 18.3 { # fts5 table in the temp schema. # reset_db +sqlite3_fts5_register_origintext db do_execsql_test 19.0 { - CREATE VIRTUAL TABLE temp.t1 USING fts5(x, detail=%DETAIL%); + CREATE VIRTUAL TABLE temp.t1 USING fts5(x, detail=%DETAIL% %TOKENIZER%); INSERT INTO t1 VALUES('x y z'); INSERT INTO t1 VALUES('w x 1'); SELECT rowid FROM t1 WHERE t1 MATCH 'x'; @@ -551,8 +566,9 @@ do_execsql_test 19.0 { # Test that 6 and 7 byte varints can be read. # reset_db +sqlite3_fts5_register_origintext db do_execsql_test 20.0 { - CREATE VIRTUAL TABLE temp.tmp USING fts5(x, detail=%DETAIL%); + CREATE VIRTUAL TABLE temp.tmp USING fts5(x, detail=%DETAIL% %TOKENIZER%); } set ::ids [list \ 0 [expr 1<<36] [expr 2<<36] [expr 1<<43] [expr 2<<43] @@ -570,7 +586,7 @@ do_test 20.1 { # do_execsql_test 21.0 { CREATE TEMP TABLE t8(a, b); - CREATE VIRTUAL TABLE ft USING fts5(x, detail=%DETAIL%); + CREATE VIRTUAL TABLE ft USING fts5(x, detail=%DETAIL% %TOKENIZER%); } do_execsql_test 21.1 { @@ -581,7 +597,7 @@ do_execsql_test 21.1 { } do_execsql_test 22.0 { - CREATE VIRTUAL TABLE t9 USING fts5(x, detail=%DETAIL%); + CREATE VIRTUAL TABLE t9 USING fts5(x, detail=%DETAIL% %TOKENIZER%); INSERT INTO t9(rowid, x) VALUES(2, 'bbb'); BEGIN; INSERT INTO t9(rowid, x) VALUES(1, 'aaa'); @@ -596,7 +612,7 @@ do_execsql_test 22.1 { #------------------------------------------------------------------------- do_execsql_test 23.0 { - CREATE VIRTUAL TABLE t10 USING fts5(x, detail=%DETAIL%); + CREATE VIRTUAL TABLE t10 USING fts5(x, detail=%DETAIL% %TOKENIZER%); CREATE TABLE t11(x); } do_execsql_test 23.1 { @@ -608,7 +624,7 @@ do_execsql_test 23.2 { #------------------------------------------------------------------------- do_execsql_test 24.0 { - CREATE VIRTUAL TABLE t12 USING fts5(x, detail=%DETAIL%); + CREATE VIRTUAL TABLE t12 USING fts5(x, detail=%DETAIL% %TOKENIZER%); INSERT INTO t12 VALUES('aaaa'); } do_execsql_test 24.1 { @@ -618,6 +634,9 @@ do_execsql_test 24.1 { INSERT INTO t12 VALUES('aaaa'); END; } +execsql_pp { + SELECT rowid, hex(block) FROM t12_data +} do_execsql_test 24.2 { INSERT INTO t12(t12) VALUES('integrity-check'); } @@ -627,7 +646,7 @@ do_execsql_test 24.3 { #------------------------------------------------------------------------- do_execsql_test 25.0 { - CREATE VIRTUAL TABLE t13 USING fts5(x, detail=%DETAIL%); + CREATE VIRTUAL TABLE t13 USING fts5(x, detail=%DETAIL% %TOKENIZER%); } do_execsql_test 25.1 { BEGIN; @@ -638,6 +657,7 @@ SELECT * FROM t13('BBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBB } +} } expand_all_sql db diff --git a/ext/fts5/test/fts5faultH.test b/ext/fts5/test/fts5faultH.test new file mode 100644 index 0000000000..9dd4cac0d6 --- /dev/null +++ b/ext/fts5/test/fts5faultH.test @@ -0,0 +1,93 @@ +# 2010 June 15 +# +# 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. +# +#*********************************************************************** +# + +source [file join [file dirname [info script]] fts5_common.tcl] +source $testdir/malloc_common.tcl +set testprefix fts5faultG + +# If SQLITE_ENABLE_FTS5 is defined, omit this file. +ifcapable !fts5 { + finish_test + return +} + +set ::testprefix fts5faultH + +sqlite3_fts5_register_origintext db + +do_execsql_test 1.0 { + CREATE VIRTUAL TABLE t1 USING fts5( + x, tokenize="origintext unicode61", tokendata=1 + ); + + BEGIN; + INSERT INTO t1 VALUES('oNe tWo thRee'); + INSERT INTO t1 VALUES('One Two Three'); + INSERT INTO t1 VALUES('onE twO threE'); + COMMIT; + BEGIN; + INSERT INTO t1 VALUES('one two three'); + INSERT INTO t1 VALUES('one two three'); + INSERT INTO t1 VALUES('one two three'); + COMMIT; +} + +do_faultsim_test 1 -faults oom* -prep { +} -body { + execsql { + SELECT rowid FROM t1('three'); + } +} -test { + faultsim_integrity_check + faultsim_test_result {0 {1 2 3 4 5 6}} +} + + +reset_db +sqlite3_fts5_register_origintext db +do_execsql_test 2.0 { + CREATE VIRTUAL TABLE t1 USING fts5( + x, tokenize="origintext unicode61", tokendata=1 + ); + INSERT INTO t1(t1, rank) VALUES('pgsz', 64); + + BEGIN; + INSERT INTO t1(rowid, x) VALUES(10, 'aaa bbb BBB'); + INSERT INTO t1(rowid, x) VALUES(12, 'bbb bbb bbb'); + INSERT INTO t1(rowid, x) VALUES(13, 'bbb bbb bbb'); + INSERT INTO t1(rowid, x) VALUES(14, 'bbb BBB bbb'); + INSERT INTO t1(rowid, x) VALUES(15, 'bbb bbb bbb'); + INSERT INTO t1(rowid, x) VALUES(16, 'bbb bbb bbb'); + INSERT INTO t1(rowid, x) VALUES(17, 'bbb bbb bbb'); + INSERT INTO t1(rowid, x) VALUES(18, 'bbb bbb bbb'); + INSERT INTO t1(rowid, x) VALUES(19, 'bbb bbb bbb'); + INSERT INTO t1(rowid, x) VALUES(20, 'bbb bbb bbb'); + INSERT INTO t1(rowid, x) VALUES(21, 'bbb bbb bbb'); + INSERT INTO t1(rowid, x) VALUES(22, 'bbb bbb bbb'); + INSERT INTO t1(rowid, x) VALUES(23, 'bbb bbb bbb'); + INSERT INTO t1(rowid, x) VALUES(24, 'aaa bbb BBB'); + COMMIT; +} + +do_faultsim_test 2 -faults oom* -prep { +} -body { + execsql { + SELECT rowid FROM t1('BBB AND AAA'); + } +} -test { + faultsim_integrity_check + faultsim_test_result {0 {10 24}} +} + + + +finish_test diff --git a/ext/fts5/test/fts5origintext.test b/ext/fts5/test/fts5origintext.test new file mode 100644 index 0000000000..9752f35d34 --- /dev/null +++ b/ext/fts5/test/fts5origintext.test @@ -0,0 +1,297 @@ +# 2014 Jan 08 +# +# 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. +# +#*********************************************************************** +# +# Tests focused on phrase queries. +# + +source [file join [file dirname [info script]] fts5_common.tcl] +set testprefix fts5origintext + +# If SQLITE_ENABLE_FTS5 is defined, omit this file. +ifcapable !fts5 { + finish_test + return +} + +foreach_detail_mode $testprefix { + +sqlite3_fts5_register_origintext db +do_execsql_test 1.0 { + CREATE VIRTUAL TABLE ft USING fts5( + x, tokenize="origintext unicode61", detail=%DETAIL% + ); + CREATE VIRTUAL TABLE vocab USING fts5vocab(ft, instance); +} + +do_execsql_test 1.1 { + INSERT INTO ft VALUES('Hello world'); +} + +do_execsql_test 1.2 { + INSERT INTO ft(ft) VALUES('integrity-check'); +} + +proc b {x} { string map [list "\0" "."] $x } +db func b b + +do_execsql_test 1.3 { + select b(term) from vocab; +} { + hello.Hello + world +} + +do_execsql_test 1.4 { + SELECT rowid FROM ft('Hello'); +} {1} + +#------------------------------------------------------------------------- +reset_db + +# Return a random integer between 0 and n-1. +# +proc random {n} { + expr {abs(int(rand()*$n))} +} + +proc select_one {list} { + set n [llength $list] + lindex $list [random $n] +} + +proc term {} { + set first_letter { + a b c d e f g h i j k l m n o p q r s t u v w x y z + A B C D E F G H I J K L M N O P Q R S T U V W X Y Z + } + + set term [select_one $first_letter] + append term [random 100] +} + +proc document {} { + set nTerm [expr [random 5] + 5] + set doc "" + for {set ii 0} {$ii < $nTerm} {incr ii} { + lappend doc [term] + } + set doc +} +db func document document + +sqlite3_fts5_register_origintext db +do_execsql_test 2.0 { + CREATE VIRTUAL TABLE ft USING fts5( + x, tokenize="origintext unicode61", detail=%DETAIL% + ); + INSERT INTO ft(ft, rank) VALUES('pgsz', 128); + CREATE VIRTUAL TABLE vocab USING fts5vocab(ft, instance); +} + +do_test 2.1 { + for {set ii 0} {$ii < 500} {incr ii} { + execsql { INSERT INTO ft VALUES( document() ) } + } +} {} + +do_execsql_test 2.2 { + INSERT INTO ft(ft) VALUES('integrity-check'); +} + +do_execsql_test 2.3 { + INSERT INTO ft(ft, rank) VALUES('merge', 16); +} + +do_execsql_test 2.4 { + INSERT INTO ft(ft) VALUES('integrity-check'); +} + +do_execsql_test 2.5 { + INSERT INTO ft(ft) VALUES('optimize'); +} + +#------------------------------------------------------------------------- +reset_db + +sqlite3_fts5_register_origintext db +do_execsql_test 3.0 { + CREATE VIRTUAL TABLE ft USING fts5( + x, tokenize="origintext unicode61", detail=%DETAIL% + ); + CREATE VIRTUAL TABLE vocab USING fts5vocab(ft, instance); + + INSERT INTO ft(rowid, x) VALUES(1, 'hello'); + INSERT INTO ft(rowid, x) VALUES(2, 'Hello'); + INSERT INTO ft(rowid, x) VALUES(3, 'HELLO'); +} + +#proc b {x} { string map [list "\0" "."] $x } +#db func b b +#execsql_pp { SELECT b(term) FROM vocab } + +do_execsql_test 3.1.1 { SELECT rowid FROM ft('hello') } 1 +do_execsql_test 3.1.2 { SELECT rowid FROM ft('Hello') } 2 +do_execsql_test 3.1.3 { SELECT rowid FROM ft('HELLO') } 3 + +do_execsql_test 3.2 { + CREATE VIRTUAL TABLE ft2 USING fts5(x, + tokenize="origintext unicode61", + tokendata=1, + detail=%DETAIL% + ); + CREATE VIRTUAL TABLE vocab2 USING fts5vocab(ft2, instance); + + INSERT INTO ft2(rowid, x) VALUES(1, 'hello'); + INSERT INTO ft2(rowid, x) VALUES(2, 'Hello'); + INSERT INTO ft2(rowid, x) VALUES(3, 'HELLO'); + + INSERT INTO ft2(rowid, x) VALUES(10, 'helloooo'); +} + +#proc b {x} { string map [list "\0" "."] $x } +#db func b b +#execsql_pp { SELECT b(term) FROM vocab } + +do_execsql_test 3.3.1 { SELECT rowid FROM ft2('hello') } {1 2 3} +do_execsql_test 3.3.2 { SELECT rowid FROM ft2('Hello') } {1 2 3} +do_execsql_test 3.3.3 { SELECT rowid FROM ft2('HELLO') } {1 2 3} + +do_execsql_test 3.3.4 { SELECT rowid FROM ft2('hello*') } {1 2 3 10} + +#------------------------------------------------------------------------- +# +reset_db +sqlite3_fts5_register_origintext db +proc querytoken {cmd iPhrase iToken} { + set txt [$cmd xQueryToken $iPhrase $iToken] + string map [list "\0" "."] $txt +} +sqlite3_fts5_create_function db querytoken querytoken + +do_execsql_test 4.0 { + CREATE VIRTUAL TABLE ft USING fts5( + x, tokenize='origintext unicode61', tokendata=1, detail=%DETAIL% + ); + INSERT INTO ft VALUES('one two three four'); +} + +do_execsql_test 4.1 { + SELECT rowid, querytoken(ft, 0, 0) FROM ft('TwO') +} {1 two.TwO} +do_execsql_test 4.2 { + SELECT rowid, querytoken(ft, 0, 0) FROM ft('one TWO ThreE') +} {1 one} +do_execsql_test 4.3 { + SELECT rowid, querytoken(ft, 1, 0) FROM ft('one TWO ThreE') +} {1 two.TWO} + +if {"%DETAIL%"=="full"} { + # Phrase queries are only supported for detail=full. + # + do_execsql_test 4.4 { + SELECT rowid, querytoken(ft, 0, 2) FROM ft('"one TWO ThreE"') + } {1 three.ThreE} + do_catchsql_test 4.5 { + SELECT rowid, querytoken(ft, 0, 3) FROM ft('"one TWO ThreE"') + } {1 SQLITE_RANGE} + do_catchsql_test 4.6 { + SELECT rowid, querytoken(ft, 1, 0) FROM ft('"one TWO ThreE"') + } {1 SQLITE_RANGE} + do_catchsql_test 4.7 { + SELECT rowid, querytoken(ft, -1, 0) FROM ft('"one TWO ThreE"') + } {1 SQLITE_RANGE} +} + +#------------------------------------------------------------------------- +# +reset_db +sqlite3_fts5_register_origintext db +proc insttoken {cmd iIdx iToken} { + set txt [$cmd xInstToken $iIdx $iToken] + string map [list "\0" "."] $txt +} +sqlite3_fts5_create_function db insttoken insttoken +fts5_aux_test_functions db + +do_execsql_test 5.0 { + CREATE VIRTUAL TABLE ft USING fts5( + x, tokenize='origintext unicode61', tokendata=1, detail=%DETAIL% + ); + INSERT INTO ft VALUES('one ONE One oNe oNE one'); +} + +do_execsql_test 5.1 { + SELECT insttoken(ft, 0, 0), + insttoken(ft, 1, 0), + insttoken(ft, 2, 0), + insttoken(ft, 3, 0), + insttoken(ft, 4, 0), + insttoken(ft, 5, 0) + FROM ft('one'); +} { + one one.ONE one.One one.oNe one.oNE one +} + +do_execsql_test 5.2 { + SELECT insttoken(ft, 1, 0) FROM ft('one'); +} { + one.ONE +} + +do_execsql_test 5.3 { + SELECT fts5_test_poslist(ft) FROM ft('one'); +} { + {0.0.0 0.0.1 0.0.2 0.0.3 0.0.4 0.0.5} +} + +#------------------------------------------------------------------------- +# Test the xInstToken() API with: +# +# * a non tokendata=1 table. +# * prefix queries. +# +reset_db +sqlite3_fts5_register_origintext db +do_execsql_test 6.0 { + CREATE VIRTUAL TABLE ft USING fts5( + x, y, tokenize='origintext unicode61', detail=%DETAIL% + ); + + INSERT INTO ft VALUES('One Two', 'Three two'); + INSERT INTO ft VALUES('three Three', 'one One'); +} +proc tokens {cmd} { + set ret [list] + for {set iTok 0} {$iTok < [$cmd xInstCount]} {incr iTok} { + set txt [$cmd xInstToken $iTok 0] + set txt [string map [list "\0" "."] $txt] + lappend ret $txt + } + set ret +} +sqlite3_fts5_create_function db tokens tokens + +do_execsql_test 6.1 { + SELECT rowid, tokens(ft) FROM ft('One'); +} {1 one.One 2 one.One} + +do_execsql_test 6.2 { + SELECT rowid, tokens(ft) FROM ft('on*'); +} {1 {{}} 2 {{} {}}} + +do_execsql_test 6.3 { + SELECT rowid, tokens(ft) FROM ft('Three*'); +} {1 {{}} 2 {{}}} + +} + +finish_test + diff --git a/ext/fts5/test/fts5origintext2.test b/ext/fts5/test/fts5origintext2.test new file mode 100644 index 0000000000..a8c5d4eb50 --- /dev/null +++ b/ext/fts5/test/fts5origintext2.test @@ -0,0 +1,146 @@ +# 2014 Jan 08 +# +# 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. +# +#*********************************************************************** +# +# Tests focused on phrase queries. +# + +source [file join [file dirname [info script]] fts5_common.tcl] +set testprefix fts5origintext2 + +# If SQLITE_ENABLE_FTS5 is defined, omit this file. +ifcapable !fts5 { + finish_test + return +} + +sqlite3_fts5_register_origintext db +do_execsql_test 1.0 { + CREATE VIRTUAL TABLE ft USING fts5( + x, tokenize="origintext unicode61", tokendata=1 + ); +} + +do_execsql_test 1.1 { + BEGIN; + INSERT INTO ft VALUES('Hello'); + INSERT INTO ft VALUES('hello'); + INSERT INTO ft VALUES('HELLO'); + INSERT INTO ft VALUES('today'); + INSERT INTO ft VALUES('today'); + INSERT INTO ft VALUES('today'); + INSERT INTO ft VALUES('World'); + INSERT INTO ft VALUES('world'); + INSERT INTO ft VALUES('WORLD'); + COMMIT; +} + +do_execsql_test 1.2 { SELECT rowid FROM ft('hello'); } {1 2 3} +do_execsql_test 1.3 { SELECT rowid FROM ft('today'); } {4 5 6} +do_execsql_test 1.4 { SELECT rowid FROM ft('world'); } {7 8 9} + +do_execsql_test 1.5 { + SELECT count(*) FROM ft_data +} 3 + +do_execsql_test 1.6 { + DELETE FROM ft; + INSERT INTO ft(ft, rank) VALUES('pgsz', 64); + BEGIN; + WITH s(i) AS ( + SELECT 1 UNION ALL SELECT i+1 FROM s WHERE i<100 + ) + INSERT INTO ft SELECT 'Hello Hello Hello Hello Hello Hello Hello' FROM s; + INSERT INTO ft VALUES ('hELLO hELLO hELLO'); + INSERT INTO ft VALUES('today today today today today today today'); + INSERT INTO ft VALUES('today today today today today today today'); + INSERT INTO ft VALUES('today today today today today today today'); + INSERT INTO ft VALUES('today today today today today today today'); + INSERT INTO ft VALUES('today today today today today today today'); + INSERT INTO ft VALUES('today today today today today today today'); + INSERT INTO ft VALUES('World World World World World World World'); + INSERT INTO ft VALUES('world world world world world world world'); + INSERT INTO ft VALUES('WORLD WORLD WORLD WORLD WORLD WORLD WORLD'); + INSERT INTO ft VALUES('World World World World World World World'); + INSERT INTO ft VALUES('world world world world world world world'); + INSERT INTO ft VALUES('WORLD WORLD WORLD WORLD WORLD WORLD WORLD'); + COMMIT; +} + +do_execsql_test 1.7 { + SELECT count(*) FROM ft_data; +} 23 + +do_execsql_test 1.8 { SELECT rowid FROM ft('hello') WHERE rowid>100; } {101} + +do_execsql_test 1.9 { + DELETE FROM ft; + INSERT INTO ft(ft) VALUES('optimize'); + SELECT count(*) FROM ft_data; +} {2} +do_execsql_test 1.10 { + BEGIN; + INSERT INTO ft VALUES('Hello'); + INSERT INTO ft VALUES('hello'); + INSERT INTO ft VALUES('HELLO'); + INSERT INTO ft VALUES('today'); + INSERT INTO ft VALUES('today'); + INSERT INTO ft VALUES('today'); + INSERT INTO ft VALUES('World'); + INSERT INTO ft VALUES('world'); + INSERT INTO ft VALUES('WORLD'); +} + +do_execsql_test 1.11 { SELECT rowid FROM ft('hello'); } {1 2 3} +do_execsql_test 1.12 { SELECT rowid FROM ft('today'); } {4 5 6} +do_execsql_test 1.13 { SELECT rowid FROM ft('world'); } {7 8 9} +do_execsql_test 1.14 { SELECT rowid FROM ft('hello') ORDER BY rank; } {1 2 3} + +#------------------------------------------------------------------------ +reset_db +sqlite3_fts5_register_origintext db +proc tokens {cmd} { + set ret [list] + for {set iTok 0} {$iTok < [$cmd xInstCount]} {incr iTok} { + set txt [$cmd xInstToken $iTok 0] + set txt [string map [list "\0" "."] $txt] + lappend ret $txt + } + set ret +} +sqlite3_fts5_create_function db tokens tokens + +do_execsql_test 2.0 { + CREATE VIRTUAL TABLE x1 USING fts5( + v, tokenize="origintext unicode61", tokendata=1, detail=none + ); + + INSERT INTO x1 VALUES('xxx Xxx XXX yyy YYY yyy'); + INSERT INTO x1 VALUES('xxx yyy xxx yyy yyy yyy'); +} + +do_execsql_test 2.1 { + SELECT tokens(x1) FROM x1('xxx'); +} { + {xxx xxx.Xxx xxx.XXX} {xxx xxx} +} + +do_execsql_test 2.2 { + UPDATE x1_content SET c0 = 'xxx xxX xxx yyy yyy yyy' WHERE id=1; +} + +do_execsql_test 2.3 { + SELECT tokens(x1) FROM x1('xxx'); +} { + {xxx {} xxx} {xxx xxx} +} + +finish_test + diff --git a/ext/fts5/test/fts5origintext3.test b/ext/fts5/test/fts5origintext3.test new file mode 100644 index 0000000000..834844595d --- /dev/null +++ b/ext/fts5/test/fts5origintext3.test @@ -0,0 +1,101 @@ +# 2023 November 22 +# +# The author disclaims copyright to this source code. In place of +# a legal notice, here is a blessing: +# +# May you do good and not evil. +# May you find forgiveness for yourself and forgive others. +# May you share freely, never taking more than you give. +# +#*********************************************************************** +# +# Tests focused on phrase queries. +# + +source [file join [file dirname [info script]] fts5_common.tcl] +set testprefix fts5origintext3 + +# If SQLITE_ENABLE_FTS5 is defined, omit this file. +ifcapable !fts5 { + finish_test + return +} + +foreach_detail_mode $testprefix { + reset_db + + sqlite3_fts5_register_origintext db + fts5_aux_test_functions db + proc insttoken {cmd iIdx iToken} { + set txt [$cmd xInstToken $iIdx $iToken] + string map [list "\0" "."] $txt + } + sqlite3_fts5_create_function db insttoken insttoken + + do_execsql_test 1.0 { + CREATE VIRTUAL TABLE ft USING fts5( + x, tokenize="origintext unicode61", tokendata=1, detail=%DETAIL% + ); + } + + do_execsql_test 1.1 { + INSERT INTO ft VALUES('Hello world HELLO WORLD hello'); + } + + do_execsql_test 1.2 { + SELECT fts5_test_poslist(ft) FROM ft('hello'); + } {{0.0.0 0.0.2 0.0.4}} + + do_execsql_test 1.3 { + SELECT + insttoken(ft, 0, 0), + insttoken(ft, 1, 0), + insttoken(ft, 2, 0) + FROM ft('hello'); + } {hello.Hello hello.HELLO hello} + + do_execsql_test 1.4 { + SELECT + insttoken(ft, 0, 0), + insttoken(ft, 1, 0), + insttoken(ft, 2, 0) + FROM ft('hello') ORDER BY rank; + } {hello.Hello hello.HELLO hello} + + do_execsql_test 1.5 { + CREATE VIRTUAL TABLE ft2 USING fts5( + x, tokenize="origintext unicode61", tokendata=1, detail=%DETAIL% + ); + INSERT INTO ft2(rowid, x) VALUES(1, 'ONE one two three ONE'); + INSERT INTO ft2(rowid, x) VALUES(2, 'TWO one two three TWO'); + INSERT INTO ft2(rowid, x) VALUES(3, 'THREE one two three THREE'); + } + + do_execsql_test 1.6 { + SELECT insttoken(ft2, 0, 0), rowid FROM ft2('three') ORDER BY rank; + } {three.THREE 3 three 1 three 2} + + do_execsql_test 1.7 { + INSERT INTO ft2(rowid, x) VALUES(10, 'aaa bbb BBB'); + INSERT INTO ft2(rowid, x) VALUES(12, 'bbb bbb bbb'); + INSERT INTO ft2(rowid, x) VALUES(13, 'bbb bbb bbb'); + INSERT INTO ft2(rowid, x) VALUES(14, 'bbb BBB bbb'); + INSERT INTO ft2(rowid, x) VALUES(15, 'bbb bbb bbb'); + INSERT INTO ft2(rowid, x) VALUES(16, 'bbb bbb bbb'); + INSERT INTO ft2(rowid, x) VALUES(17, 'bbb bbb bbb'); + INSERT INTO ft2(rowid, x) VALUES(18, 'bbb bbb bbb'); + INSERT INTO ft2(rowid, x) VALUES(19, 'bbb bbb bbb'); + INSERT INTO ft2(rowid, x) VALUES(20, 'bbb bbb bbb'); + INSERT INTO ft2(rowid, x) VALUES(21, 'bbb bbb bbb'); + INSERT INTO ft2(rowid, x) VALUES(22, 'bbb bbb bbb'); + INSERT INTO ft2(rowid, x) VALUES(23, 'bbb bbb bbb'); + INSERT INTO ft2(rowid, x) VALUES(24, 'aaa bbb BBB'); + } + + do_execsql_test 1.8 { SELECT rowid FROM ft2('aaa AND bbb'); } {10 24} + do_execsql_test 1.9 { SELECT rowid FROM ft2('bbb AND aaa'); } {10 24} + +} + +finish_test + diff --git a/ext/fts5/test/fts5origintext4.test b/ext/fts5/test/fts5origintext4.test new file mode 100644 index 0000000000..8973a24b05 --- /dev/null +++ b/ext/fts5/test/fts5origintext4.test @@ -0,0 +1,66 @@ +# 2023 November 22 +# +# The author disclaims copyright to this source code. In place of +# a legal notice, here is a blessing: +# +# May you do good and not evil. +# May you find forgiveness for yourself and forgive others. +# May you share freely, never taking more than you give. +# +#*********************************************************************** +# +# Tests focused on phrase queries. +# + +source [file join [file dirname [info script]] fts5_common.tcl] +set testprefix fts5origintext4 + +# If SQLITE_ENABLE_FTS5 is defined, omit this file. +ifcapable !fts5 { + finish_test + return +} + +sqlite3_fts5_register_origintext db +do_execsql_test 1.0 { + CREATE VIRTUAL TABLE ft USING fts5( + x, tokenize="origintext unicode61", tokendata=1 + ); +} + +do_execsql_test 1.1 { + BEGIN; + INSERT INTO ft SELECT 'the first thing'; + + WITH s(i) AS ( + SELECT 1 UNION ALL SELECT i+1 FROM s WHERE i<90000 + ) + INSERT INTO ft SELECT 'The second thing' FROM s; + + INSERT INTO ft SELECT 'the first thing'; + COMMIT; + INSERT INTO ft(ft) VALUES('optimize'); +} + +foreach {tn sql expr} { + 1 { SELECT rowid FROM ft('the') } {$mem > 250000} + 2 { SELECT rowid FROM ft('first') } {$mem < 50000} + 3 { SELECT rowid FROM ft('the first') } {$mem < 50000} +} { + db close + sqlite3 db test.db + sqlite3_fts5_register_origintext db + + execsql $sql + do_test 1.2.$tn { + set mem [lindex [sqlite3_db_status db CACHE_USED 0] 1] + expr $expr + } 1 +} + +proc b {x} { string map [list "\0" "."] $x } +db func b b +# execsql_pp { SELECT segid, b(term), pgno from ft_idx } + +finish_test + diff --git a/ext/fts5/test/fts5origintext5.test b/ext/fts5/test/fts5origintext5.test new file mode 100644 index 0000000000..03d5bee215 --- /dev/null +++ b/ext/fts5/test/fts5origintext5.test @@ -0,0 +1,273 @@ +# 2023 Dec 04 +# +# The author disclaims copyright to this source code. In place of +# a legal notice, here is a blessing: +# +# May you do good and not evil. +# May you find forgiveness for yourself and forgive others. +# May you share freely, never taking more than you give. +# +#*********************************************************************** +# +# Tests for tables that use both tokendata=1 and contentless_delete=1. +# + +source [file join [file dirname [info script]] fts5_common.tcl] +set testprefix fts5origintext + +# If SQLITE_ENABLE_FTS5 is defined, omit this file. +ifcapable !fts5 { + finish_test + return +} + +# Return a random integer between 0 and n-1. +# +proc random {n} { expr {abs(int(rand()*$n))} } + +# Select an element of the list passed as the only argument at random and +# return it. +# +proc select_one {list} { + set n [llength $list] + lindex $list [random $n] +} + +# Given a term that consists entirely of alphabet characters, return all +# permutations of the term using upper and lower case characters. e.g. +# +# "abc" -> {CBA cBA CbA cbA CBa cBa Cba cba} +# +proc casify {term {lRet {{}}}} { + if {$term==""} { return $lRet } + set t [string range $term 1 end] + set f1 [string toupper [string range $term 0 0]] + set f2 [string tolower [string range $term 0 0]] + set ret [list] + foreach x $lRet { + lappend ret "$x$f1" + lappend ret "$x$f2" + } + return [casify $t $ret] +} + +proc vocab {} { + list abc def ghi jkl mno pqr stu vwx yza +} + +# Return a random 3 letter term. +# +proc term {} { + if {[info exists ::expanded_vocab]==0} { + foreach v [vocab] { lappend ::expanded_vocab {*}[casify $v] } + } + + select_one $::expanded_vocab +} + +# Return a document - between 3 and 10 terms. +# +proc document {} { + set nTerm [expr [random 3] + 7] + set doc "" + for {set ii 0} {$ii < $nTerm} {incr ii} { + lappend doc [term] + } + set doc +} +db func document document + +#------------------------------------------------------------------------- + +expr srand(6) + +set NDOC 200 +set NLOOP 50 + +sqlite3_fts5_register_origintext db + +proc tokens {cmd} { + set ret [list] + for {set iTok 0} {$iTok < [$cmd xInstCount]} {incr iTok} { + set txt [$cmd xInstToken $iTok 0] + set txt [string map [list "\0" "."] $txt] + lappend ret $txt + } + set ret +} +sqlite3_fts5_create_function db tokens tokens + +proc rankfunc {cmd} { + $cmd xRowid +} +sqlite3_fts5_create_function db rankfunc rankfunc + +proc ctrl_tokens {term args} { + set ret [list] + set term [string tolower $term] + foreach doc $args { + foreach a $doc { + if {[string tolower $a]==$term} { + if {$a==$term} { + lappend ret $a + } else { + lappend ret [string tolower $a].$a + } + } + } + } + set ret +} +db func ctrl_tokens ctrl_tokens + +proc do_all_vocab_test {tn} { + foreach ::v [concat [vocab] nnn] { + set answer [execsql { + SELECT id, ctrl_tokens($::v, x) FROM ctrl WHERE x LIKE '%' || $::v || '%' + }] + do_execsql_test $tn.$::v.1 { + SELECT rowid, tokens(ft) FROM ft($::v) + } $answer + do_execsql_test $tn.$::v.2 { + SELECT rowid, tokens(ft) FROM ft($::v) ORDER BY rank + } $answer + } +} + +do_execsql_test 1.0 { + CREATE VIRTUAL TABLE ft USING fts5( + x, tokenize="origintext unicode61", content=, contentless_delete=1, + tokendata=1 + ); + + CREATE TABLE ctrl(id INTEGER PRIMARY KEY, x TEXT); + INSERT INTO ft(ft, rank) VALUES('pgsz', 64); + INSERT INTO ft(ft, rank) VALUES('rank', 'rankfunc()'); +} +do_test 1.1 { + for {set ii 0} {$ii < $NDOC} {incr ii} { + set doc [document] + execsql { + INSERT INTO ft(rowid, x) VALUES($ii, $doc); + INSERT INTO ctrl(id, x) VALUES($ii, $doc); + } + } +} {} + +#execsql_pp { SELECT * FROM ctrl } +#execsql_pp { SELECT * FROM ft } +#fts5_aux_test_functions db +#execsql_pp { SELECT rowid, tokens(ft), fts5_test_poslist(ft) FROM ft('ghi'); } + +do_all_vocab_test 1.2 + +for {set ii 0} {$ii < $NLOOP} {incr ii} { + set lRowid [execsql { SELECT id FROM ctrl WHERE random() % 2 }] + foreach r $lRowid { + execsql { DELETE FROM ft WHERE rowid = $r } + execsql { DELETE FROM ctrl WHERE rowid = $r } + + set doc [document] + execsql { INSERT INTO ft(rowid, x) VALUES($r, $doc) } + execsql { INSERT INTO ctrl(id, x) VALUES($r, $doc) } + } + do_all_vocab_test 1.3.$ii +} + +#------------------------------------------------------------------------- + +do_execsql_test 2.0 { + CREATE VIRTUAL TABLE ft2 USING fts5( + x, y, tokenize="origintext unicode61", content=, contentless_delete=1, + tokendata=1 + ); + + CREATE TABLE ctrl2(id INTEGER PRIMARY KEY, x TEXT, y TEXT); + INSERT INTO ft2(ft2, rank) VALUES('pgsz', 64); + INSERT INTO ft2(ft2, rank) VALUES('rank', 'rankfunc()'); +} +do_test 2.1 { + for {set ii 0} {$ii < $NDOC} {incr ii} { + set doc1 [document] + set doc2 [document] + execsql { + INSERT INTO ft2(rowid, x, y) VALUES($ii, $doc, $doc2); + INSERT INTO ctrl2(id, x, y) VALUES($ii, $doc, $doc2); + } + } +} {} + +proc do_all_vocab_test2 {tn} { + foreach ::v [vocab] { + set answer [execsql { + SELECT id, ctrl_tokens($::v, x, y) FROM ctrl2 + WHERE x LIKE '%' || $::v || '%' OR y LIKE '%' || $::v || '%'; + }] + do_execsql_test $tn.$::v.1 { + SELECT rowid, tokens(ft2) FROM ft2($::v) + } $answer + do_execsql_test $tn.$::v.2 { + SELECT rowid, tokens(ft2) FROM ft2($::v) ORDER BY rank + } $answer + } +} + +do_all_vocab_test2 2.2 + +for {set ii 0} {$ii < $NLOOP} {incr ii} { + set lRowid [execsql { SELECT id FROM ctrl2 WHERE random() % 2 }] + foreach r $lRowid { + execsql { DELETE FROM ft2 WHERE rowid = $r } + execsql { DELETE FROM ctrl2 WHERE rowid = $r } + + set doc1 [document] + set doc2 [document] + execsql { INSERT INTO ft2(rowid, x, y) VALUES($r, $doc, $doc1) } + execsql { INSERT INTO ctrl2(id, x, y) VALUES($r, $doc, $doc2) } + } + do_all_vocab_test 2.3.$ii +} + +#------------------------------------------------------------------------- + +unset -nocomplain ::expanded_vocab +proc vocab {} { + list abcde fghij klmno +} + +proc do_all_vocab_test3 {tn} { + foreach ::v [concat [vocab] nnn] { + set answer [execsql { + SELECT rowid, ctrl_tokens($::v, w) FROM ctrl3 WHERE w LIKE '%' || $::v || '%' + }] + do_execsql_test $tn.$::v.1 { + SELECT rowid, tokens(ft3) FROM ft3($::v) + } $answer + do_execsql_test $tn.$::v.2 { + SELECT rowid, tokens(ft3) FROM ft3($::v) ORDER BY rank + } $answer + } +} + +do_execsql_test 3.0 { + CREATE VIRTUAL TABLE ft3 USING fts5( + w, tokenize="origintext unicode61", content=, contentless_delete=1, + tokendata=1 + ); + INSERT INTO ft3(ft3, rank) VALUES('rank', 'rankfunc()'); + CREATE TABLE ctrl3(w); +} + +do_execsql_test 3.1 { + WITH s(i) AS ( + SELECT 1 UNION ALL SELECT i+1 FROM s WHERE i<2 + ) + INSERT INTO ctrl3 SELECT document() FROM s; + INSERT INTO ft3(rowid, w) SELECT rowid, w FROM ctrl3; +} + +do_all_vocab_test3 3.2 + + +finish_test + diff --git a/ext/fts5/test/fts5simple2.test b/ext/fts5/test/fts5simple2.test index e57cea70fa..6c0e0e1662 100644 --- a/ext/fts5/test/fts5simple2.test +++ b/ext/fts5/test/fts5simple2.test @@ -343,7 +343,9 @@ do_execsql_test 17.0 { INSERT INTO t2 VALUES('a aa aaa', 'b bb bbb'); COMMIT; } -do_execsql_test 17.1 { SELECT * FROM t2('y:a*') WHERE rowid BETWEEN 10 AND 20 } +do_execsql_test 17.1 { + SELECT * FROM t2('y:a*') WHERE rowid BETWEEN 10 AND 20 +} do_execsql_test 17.2 { BEGIN; INSERT INTO t2 VALUES('a aa aaa', 'b bb bbb'); diff --git a/ext/recover/dbdata.c b/ext/recover/dbdata.c index 25a6e9fd6a..b6cb26ecd3 100644 --- a/ext/recover/dbdata.c +++ b/ext/recover/dbdata.c @@ -582,6 +582,7 @@ static int dbdataNext(sqlite3_vtab_cursor *pCursor){ bNextPage = 1; }else{ iOff += dbdataGetVarintU32(&pCsr->aPage[iOff], &nPayload); + if( nPayload>0x7fffff00 ) nPayload &= 0x3fff; } /* If this is a leaf intkey cell, load the rowid */ diff --git a/ext/wasm/GNUmakefile b/ext/wasm/GNUmakefile index c0cab212d8..3ef478612c 100644 --- a/ext/wasm/GNUmakefile +++ b/ext/wasm/GNUmakefile @@ -854,7 +854,7 @@ dir.sql := sql speedtest1 := ../../speedtest1 speedtest1.c := ../../test/speedtest1.c speedtest1.sql := $(dir.sql)/speedtest1.sql -speedtest1.cliflags := --size 25 --big-transactions +speedtest1.cliflags := --size 10 --big-transactions $(speedtest1): $(MAKE) -C ../.. speedtest1 $(speedtest1.sql): $(speedtest1) $(MAKEFILE) diff --git a/ext/wasm/batch-runner-sahpool.html b/ext/wasm/batch-runner-sahpool.html new file mode 100644 index 0000000000..ad7e7b5408 --- /dev/null +++ b/ext/wasm/batch-runner-sahpool.html @@ -0,0 +1,86 @@ + + + + + + + + sqlite3-api batch SQL runner for the SAHPool VFS + + +
sqlite3-api batch SQL runner for the SAHPool VFS
+
+ + + + +
+
+ + + + diff --git a/ext/wasm/batch-runner-sahpool.js b/ext/wasm/batch-runner-sahpool.js new file mode 100644 index 0000000000..dfa5044a9e --- /dev/null +++ b/ext/wasm/batch-runner-sahpool.js @@ -0,0 +1,341 @@ +/* + 2023-11-30 + + The author disclaims copyright to this source code. In place of a + legal notice, here is a blessing: + + * May you do good and not evil. + * May you find forgiveness for yourself and forgive others. + * May you share freely, never taking more than you give. + + *********************************************************************** + + A basic batch SQL runner for the SAHPool VFS. This file must be run in + a worker thread. This is not a full-featured app, just a way to get some + measurements for batch execution of SQL for the OPFS SAH Pool VFS. +*/ +'use strict'; + +const wMsg = function(msgType,...args){ + postMessage({ + type: msgType, + data: args + }); +}; +const toss = function(...args){throw new Error(args.join(' '))}; +const warn = (...args)=>{ wMsg('warn',...args); }; +const error = (...args)=>{ wMsg('error',...args); }; +const log = (...args)=>{ wMsg('stdout',...args); } +let sqlite3; +const urlParams = new URL(globalThis.location.href).searchParams; +const cacheSize = (()=>{ + if(urlParams.has('cachesize')) return +urlParams.get('cachesize'); + return 200; +})(); + + +/** Throws if the given sqlite3 result code is not 0. */ +const checkSqliteRc = (dbh,rc)=>{ + if(rc) toss("Prepare failed:",sqlite3.capi.sqlite3_errmsg(dbh)); +}; + +const sqlToDrop = [ + "SELECT type,name FROM sqlite_schema ", + "WHERE name NOT LIKE 'sqlite\\_%' escape '\\' ", + "AND name NOT LIKE '\\_%' escape '\\'" +].join(''); + +const clearDbSqlite = function(db){ + // This would be SO much easier with the oo1 API, but we specifically want to + // inject metrics we can't get via that API, and we cannot reliably (OPFS) + // open the same DB twice to clear it using that API, so... + const rc = sqlite3.wasm.exports.sqlite3_wasm_db_reset(db.handle); + log("reset db rc =",rc,db.id, db.filename); +}; + +const App = { + db: undefined, + cache:Object.create(null), + log: log, + warn: warn, + error: error, + metrics: { + fileCount: 0, + runTimeMs: 0, + prepareTimeMs: 0, + stepTimeMs: 0, + stmtCount: 0, + strcpyMs: 0, + sqlBytes: 0 + }, + fileList: undefined, + execSql: async function(name,sql){ + const db = this.db; + const banner = "========================================"; + this.log(banner, + "Running",name,'('+sql.length,'bytes)'); + const capi = this.sqlite3.capi, wasm = this.sqlite3.wasm; + let pStmt = 0, pSqlBegin; + const metrics = db.metrics = Object.create(null); + metrics.prepTotal = metrics.stepTotal = 0; + metrics.stmtCount = 0; + metrics.malloc = 0; + metrics.strcpy = 0; + if(this.gotErr){ + this.error("Cannot run SQL: error cleanup is pending."); + return; + } + // Run this async so that the UI can be updated for the above header... + const endRun = ()=>{ + metrics.evalSqlEnd = performance.now(); + metrics.evalTimeTotal = (metrics.evalSqlEnd - metrics.evalSqlStart); + this.log("metrics:",JSON.stringify(metrics, undefined, ' ')); + this.log("prepare() count:",metrics.stmtCount); + this.log("Time in prepare_v2():",metrics.prepTotal,"ms", + "("+(metrics.prepTotal / metrics.stmtCount),"ms per prepare())"); + this.log("Time in step():",metrics.stepTotal,"ms", + "("+(metrics.stepTotal / metrics.stmtCount),"ms per step())"); + this.log("Total runtime:",metrics.evalTimeTotal,"ms"); + this.log("Overhead (time - prep - step):", + (metrics.evalTimeTotal - metrics.prepTotal - metrics.stepTotal)+"ms"); + this.log(banner,"End of",name); + this.metrics.prepareTimeMs += metrics.prepTotal; + this.metrics.stepTimeMs += metrics.stepTotal; + this.metrics.stmtCount += metrics.stmtCount; + this.metrics.strcpyMs += metrics.strcpy; + this.metrics.sqlBytes += sql.length; + }; + + const runner = function(resolve, reject){ + ++this.metrics.fileCount; + metrics.evalSqlStart = performance.now(); + const stack = wasm.scopedAllocPush(); + try { + let t, rc; + let sqlByteLen = sql.byteLength; + const [ppStmt, pzTail] = wasm.scopedAllocPtr(2); + t = performance.now(); + pSqlBegin = wasm.scopedAlloc( sqlByteLen + 1/*SQL + NUL*/) || toss("alloc(",sqlByteLen,") failed"); + metrics.malloc = performance.now() - t; + metrics.byteLength = sqlByteLen; + let pSql = pSqlBegin; + const pSqlEnd = pSqlBegin + sqlByteLen; + t = performance.now(); + wasm.heap8().set(sql, pSql); + wasm.poke(pSql + sqlByteLen, 0); + //log("SQL:",wasm.cstrToJs(pSql)); + metrics.strcpy = performance.now() - t; + let breaker = 0; + while(pSql && wasm.peek8(pSql)){ + wasm.pokePtr(ppStmt, 0); + wasm.pokePtr(pzTail, 0); + t = performance.now(); + rc = capi.sqlite3_prepare_v2( + db.handle, pSql, sqlByteLen, ppStmt, pzTail + ); + metrics.prepTotal += performance.now() - t; + checkSqliteRc(db.handle, rc); + pStmt = wasm.peekPtr(ppStmt); + pSql = wasm.peekPtr(pzTail); + sqlByteLen = pSqlEnd - pSql; + if(!pStmt) continue/*empty statement*/; + ++metrics.stmtCount; + t = performance.now(); + rc = capi.sqlite3_step(pStmt); + capi.sqlite3_finalize(pStmt); + pStmt = 0; + metrics.stepTotal += performance.now() - t; + switch(rc){ + case capi.SQLITE_ROW: + case capi.SQLITE_DONE: break; + default: checkSqliteRc(db.handle, rc); toss("Not reached."); + } + } + resolve(this); + }catch(e){ + if(pStmt) capi.sqlite3_finalize(pStmt); + this.gotErr = e; + reject(e); + }finally{ + capi.sqlite3_exec(db.handle,"rollback;",0,0,0); + wasm.scopedAllocPop(stack); + } + }.bind(this); + const p = new Promise(runner); + return p.catch( + (e)=>this.error("Error via execSql("+name+",...):",e.message) + ).finally(()=>{ + endRun(); + }); + }, + + /** + Loads batch-runner.list and populates the selection list from + it. Returns a promise which resolves to nothing in particular + when it completes. Only intended to be run once at the start + of the app. + */ + loadSqlList: async function(){ + const infile = 'batch-runner.list'; + this.log("Loading list of SQL files:", infile); + let txt; + try{ + const r = await fetch(infile); + if(404 === r.status){ + toss("Missing file '"+infile+"'."); + } + if(!r.ok) toss("Loading",infile,"failed:",r.statusText); + txt = await r.text(); + }catch(e){ + this.error(e.message); + throw e; + } + App.fileList = txt.split(/\n+/).filter(x=>!!x); + this.log("Loaded",infile); + }, + + /** Fetch ./fn and return its contents as a Uint8Array. */ + fetchFile: async function(fn, cacheIt=false){ + if(cacheIt && this.cache[fn]) return this.cache[fn]; + this.log("Fetching",fn,"..."); + let sql; + try { + const r = await fetch(fn); + if(!r.ok) toss("Fetch failed:",r.statusText); + sql = new Uint8Array(await r.arrayBuffer()); + }catch(e){ + this.error(e.message); + throw e; + } + this.log("Fetched",sql.length,"bytes from",fn); + if(cacheIt) this.cache[fn] = sql; + return sql; + }/*fetchFile()*/, + + /** + Converts this.metrics() to a form which is suitable for easy conversion to + CSV. It returns an array of arrays. The first sub-array is the column names. + The 2nd and subsequent are the values, one per test file (only the most recent + metrics are kept for any given file). + */ + metricsToArrays: function(){ + const rc = []; + Object.keys(this.dbs).sort().forEach((k)=>{ + const d = this.dbs[k]; + const m = d.metrics; + delete m.evalSqlStart; + delete m.evalSqlEnd; + const mk = Object.keys(m).sort(); + if(!rc.length){ + rc.push(['db', ...mk]); + } + const row = [k.split('/').pop()/*remove dir prefix from filename*/]; + rc.push(row); + row.push(...mk.map((kk)=>m[kk])); + }); + return rc; + }, + + metricsToBlob: function(colSeparator='\t'){ + const ar = [], ma = this.metricsToArrays(); + if(!ma.length){ + this.error("Metrics are empty. Run something."); + return; + } + ma.forEach(function(row){ + ar.push(row.join(colSeparator),'\n'); + }); + return new Blob(ar); + }, + + /** + Fetch file fn and eval it as an SQL blob. This is an async + operation and returns a Promise which resolves to this + object on success. + */ + evalFile: async function(fn){ + const sql = await this.fetchFile(fn); + return this.execSql(fn,sql); + }/*evalFile()*/, + + /** + Fetches the handle of the db associated with + this.e.selImpl.value, opening it if needed. + */ + initDb: function(){ + const capi = this.sqlite3.capi, wasm = this.sqlite3.wasm; + const stack = wasm.scopedAllocPush(); + let pDb = 0; + const d = Object.create(null); + d.filename = "/batch.db"; + try{ + const oFlags = capi.SQLITE_OPEN_CREATE | capi.SQLITE_OPEN_READWRITE; + const ppDb = wasm.scopedAllocPtr(); + const rc = capi.sqlite3_open_v2(d.filename, ppDb, oFlags, this.PoolUtil.vfsName); + pDb = wasm.peekPtr(ppDb) + if(rc) toss("sqlite3_open_v2() failed with code",rc); + capi.sqlite3_exec(pDb, "PRAGMA cache_size="+cacheSize, 0, 0, 0); + this.log("cache_size =",cacheSize); + }catch(e){ + if(pDb) capi.sqlite3_close_v2(pDb); + throw e; + }finally{ + wasm.scopedAllocPop(stack); + } + d.handle = pDb; + this.log("Opened db:",d.filename,'@',d.handle); + return d; + }, + + closeDb: function(){ + if(this.db.handle){ + this.sqlite3.capi.sqlite3_close_v2(this.db.handle); + this.db.handle = undefined; + } + }, + + run: async function(sqlite3){ + delete this.run; + this.sqlite3 = sqlite3; + const capi = sqlite3.capi, wasm = sqlite3.wasm; + this.log("Loaded module:",capi.sqlite3_libversion(), capi.sqlite3_sourceid()); + this.log("WASM heap size =",wasm.heap8().length); + let timeStart; + sqlite3.installOpfsSAHPoolVfs({ + clearOnInit: true, initialCapacity: 4, + name: 'batch-sahpool', + verbosity: 2 + }).then(PoolUtil=>{ + App.PoolUtil = PoolUtil; + App.db = App.initDb(); + }) + .then(async ()=>this.loadSqlList()) + .then(async ()=>{ + timeStart = performance.now(); + for(let i = 0; i < App.fileList.length; ++i){ + const fn = App.fileList[i]; + await App.evalFile(fn); + if(App.gotErr) throw App.gotErr; + } + }) + .then(()=>{ + App.metrics.runTimeMs = performance.now() - timeStart; + App.log("total metrics:",JSON.stringify(App.metrics, undefined, ' ')); + App.log("Reload the page to run this again."); + App.closeDb(); + App.PoolUtil.removeVfs(); + }) + .catch(e=>this.error("ERROR:",e)); + }/*run()*/ +}/*App*/; + +let sqlite3Js = 'sqlite3.js'; +if(urlParams.has('sqlite3.dir')){ + sqlite3Js = urlParams.get('sqlite3.dir') + '/' + sqlite3Js; +} +importScripts(sqlite3Js); +globalThis.sqlite3InitModule().then(async function(sqlite3_){ + log("Done initializing. Running batch runner..."); + sqlite3 = sqlite3_; + App.run(sqlite3_); +}); diff --git a/ext/wasm/batch-runner.js b/ext/wasm/batch-runner.js index ff287a66e7..e7a322b7f0 100644 --- a/ext/wasm/batch-runner.js +++ b/ext/wasm/batch-runner.js @@ -72,7 +72,6 @@ App.logHtml("reset db rc =",rc,db.id, db.filename); }; - const E = (s)=>document.querySelector(s); const App = { e: { @@ -91,6 +90,15 @@ db: Object.create(null), dbs: Object.create(null), cache:{}, + metrics: { + fileCount: 0, + runTimeMs: 0, + prepareTimeMs: 0, + stepTimeMs: 0, + stmtCount: 0, + strcpyMs: 0, + sqlBytes: 0 + }, log: console.log.bind(console), warn: console.warn.bind(console), cls: function(){this.e.output.innerHTML = ''}, @@ -117,7 +125,6 @@ "Running",name,'('+sql.length,'bytes) using',db.id); const capi = this.sqlite3.capi, wasm = this.sqlite3.wasm; let pStmt = 0, pSqlBegin; - const stack = wasm.scopedAllocPush(); const metrics = db.metrics = Object.create(null); metrics.prepTotal = metrics.stepTotal = 0; metrics.stmtCount = 0; @@ -142,6 +149,11 @@ this.logHtml("Overhead (time - prep - step):", (metrics.evalTimeTotal - metrics.prepTotal - metrics.stepTotal)+"ms"); this.logHtml(banner,"End of",name); + this.metrics.prepareTimeMs += metrics.prepTotal; + this.metrics.stepTimeMs += metrics.stepTotal; + this.metrics.stmtCount += metrics.stmtCount; + this.metrics.strcpyMs += metrics.strcpy; + this.metrics.sqlBytes += sql.length; }; let runner; @@ -214,7 +226,9 @@ }.bind(this); }else{/*sqlite3 db...*/ runner = function(resolve, reject){ + ++this.metrics.fileCount; metrics.evalSqlStart = performance.now(); + const stack = wasm.scopedAllocPush(); try { let t; let sqlByteLen = sql.byteLength; @@ -269,7 +283,7 @@ let p; if(1){ p = new Promise(function(res,rej){ - setTimeout(()=>runner(res, rej), 50)/*give UI a chance to output the "running" banner*/; + setTimeout(()=>runner(res, rej), 0)/*give UI a chance to output the "running" banner*/; }); }else{ p = new Promise(runner); @@ -401,7 +415,7 @@ }); return new Blob(ar); }, - + downloadMetrics: function(){ const b = this.metricsToBlob(); if(!b) return; @@ -576,6 +590,8 @@ const timeTotal = performance.now() - timeStart; who.logHtml("Run-remaining time:",timeTotal,"ms ("+(timeTotal/1000/60)+" minute(s))"); who.clearStorage(); + App.metrics.runTimeMs = timeTotal; + who.logHtml("Total metrics:",JSON.stringify(App.metrics,undefined,' ')); }, false); }/*run()*/ }/*App*/; diff --git a/ext/wasm/demo-123.js b/ext/wasm/demo-123.js index 8e03ee80fa..9f90ca7568 100644 --- a/ext/wasm/demo-123.js +++ b/ext/wasm/demo-123.js @@ -20,7 +20,7 @@ the main (UI) thread. */ let logHtml; - if(self.window === self /* UI thread */){ + if(globalThis.window === globalThis /* UI thread */){ console.log("Running demo from main UI thread."); logHtml = function(cssClass,...args){ const ln = document.createElement('div'); @@ -250,7 +250,7 @@ }/*demo1()*/; log("Loading and initializing sqlite3 module..."); - if(self.window!==self) /*worker thread*/{ + if(globalThis.window!==globalThis) /*worker thread*/{ /* If sqlite3.js is in a directory other than this script, in order to get sqlite3.js to resolve sqlite3.wasm properly, we have to @@ -262,19 +262,20 @@ that's not needed. URL arguments passed as part of the filename via importScripts() - are simply lost, and such scripts see the self.location of + are simply lost, and such scripts see the globalThis.location of _this_ script. */ let sqlite3Js = 'sqlite3.js'; - const urlParams = new URL(self.location.href).searchParams; + const urlParams = new URL(globalThis.location.href).searchParams; if(urlParams.has('sqlite3.dir')){ sqlite3Js = urlParams.get('sqlite3.dir') + '/' + sqlite3Js; } importScripts(sqlite3Js); } - self.sqlite3InitModule({ - // We can redirect any stdout/stderr from the module - // like so... + globalThis.sqlite3InitModule({ + /* We can redirect any stdout/stderr from the module like so, but + note that doing so makes use of Emscripten-isms, not + well-defined sqlite APIs. */ print: log, printErr: error }).then(function(sqlite3){ diff --git a/ext/wasm/index-dist.html b/ext/wasm/index-dist.html index 36e21117e3..f5bcdc1cb2 100644 --- a/ext/wasm/index-dist.html +++ b/ext/wasm/index-dist.html @@ -46,6 +46,9 @@ file:// URLs.
  • Any OPFS-related pages or tests require:
      +
    • An OPFS-capable browser released after February + 2023. Some tests will work with Chromium-based browsers + going back to around v102.
    • That the web server emit the so-called COOP and @@ -53,12 +56,6 @@ headers. althttpd requires the -enable-sab flag for that.
      • -
    • A very recent version of a Chromium-based browser - (v102 at least, possibly newer). OPFS support in the - other major browsers is pending. Development and testing - is currently done against a dev-channel release of - Chrome (v111 as of 2023-02-10). -
    • diff --git a/ext/wasm/index.html b/ext/wasm/index.html index 70ce0441e0..ebbfd6763d 100644 --- a/ext/wasm/index.html +++ b/ext/wasm/index.html @@ -31,6 +31,9 @@ file:// URLs.
  • Any OPFS-related pages or tests require:
      +
    • An OPFS-capable browser released after February + 2023. Some tests will work with Chromium-based browsers + going back to around v102.
    • That the web server emit the so-called COOP and @@ -38,12 +41,6 @@ headers. althttpd requires the -enable-sab flag for that.
      • -
    • A very recent version of a - Chromium-based browser (v102 at least, possibly newer). OPFS - support in the other major browsers is pending. Development - and testing is currently done against a dev-channel release - of Chrome (v111 as of 2023-02-10). -
    • diff --git a/ext/wasm/speedtest1-worker.html b/ext/wasm/speedtest1-worker.html index 3adf8f2ee9..8c9a77dc5e 100644 --- a/ext/wasm/speedtest1-worker.html +++ b/ext/wasm/speedtest1-worker.html @@ -350,7 +350,7 @@ eControls.classList.remove('hidden'); break; case 'stdout': log(msg.data); break; - case 'stdout': logErr(msg.data); break; + case 'stderr': logErr(msg.data); break; case 'run-start': eControls.disabled = true; log("Running speedtest1 with argv =",msg.data.join(' ')); diff --git a/ext/wasm/speedtest1-worker.js b/ext/wasm/speedtest1-worker.js index dedc320292..3abc589b59 100644 --- a/ext/wasm/speedtest1-worker.js +++ b/ext/wasm/speedtest1-worker.js @@ -7,9 +7,9 @@ } importScripts(speedtestJs); /** - If this environment contains OPFS, this function initializes it and - returns the name of the dir on which OPFS is mounted, else it returns - an empty string. + If this build includes WASMFS, this function initializes it and + returns the name of the dir on which OPFS is mounted, else it + returns an empty string. */ const wasmfsDir = function f(wasmUtil){ if(undefined !== f._) return f._; @@ -47,6 +47,7 @@ }; const log = (...args)=>logMsg('stdout',args); const logErr = (...args)=>logMsg('stderr',args); + const realSahName = 'opfs-sahpool-speedtest1'; const runSpeedtest = async function(cliFlagsArray){ const scope = App.wasm.scopedAllocPush(); @@ -57,7 +58,6 @@ ]; App.logBuffer.length = 0; const ndxSahPool = argv.indexOf('opfs-sahpool'); - const realSahName = 'opfs-sahpool-speedtest1'; if(ndxSahPool>0){ argv[ndxSahPool] = realSahName; log("Updated argv for opfs-sahpool: --vfs",realSahName); @@ -73,7 +73,7 @@ clearOnInit: true, verbosity: 2 }).then(PoolUtil=>{ - log("opfs-sahpool successfully installed as",realSahName); + log("opfs-sahpool successfully installed as",PoolUtil.vfsName); App.sqlite3.$SAHPoolUtil = PoolUtil; //console.log("sqlite3.oo1.OpfsSAHPoolDb =", App.sqlite3.oo1.OpfsSAHPoolDb); }); @@ -102,19 +102,6 @@ } }; - const sahpSanityChecks = function(sqlite3){ - log("Attempting OpfsSAHPoolDb sanity checks..."); - const db = new sqlite3.oo1.OpfsSAHPoolDb('opfs-sahpoool.db'); - const fn = db.filename; - db.exec([ - 'create table t(a);', - 'insert into t(a) values(1),(2),(3);' - ]); - db.close(); - sqlite3.wasm.sqlite3_wasm_vfs_unlink(sqlite3_vfs_find("opfs-sahpool"), fn); - log("SAH sanity checks done."); - }; - const EmscriptenModule = { print: log, printErr: logErr, diff --git a/manifest b/manifest index 77dc87e927..bb07a7e78f 100644 --- a/manifest +++ b/manifest @@ -1,12 +1,12 @@ -C Update\sthis\sbranch\swith\slatest\schanges\sfrom\strunk. -D 2023-11-28T17:32:30.239 +C Merge\sthe\slatest\strunk\senhancements\sinto\sthe\swal2\sbranch. +D 2023-12-06T21:11:11.025 F .fossil-settings/empty-dirs dbb81e8fc0401ac46a1491ab34a7f2c7c0452f2f06b54ebb845d024ca8283ef1 F .fossil-settings/ignore-glob 35175cdfcf539b2318cb04a9901442804be81cd677d8b889fcc9149c21f239ea F LICENSE.md df5091916dbb40e6e9686186587125e1b2ff51f022cc334e886c19a0e9982724 F Makefile.in 98cf0a52b45f1ffba34a663c7ad0e39bc284123023229beefae193d547052126 F Makefile.linux-gcc f3842a0b1efbfbb74ac0ef60e56b301836d05b4d867d014f714fa750048f1ab6 F Makefile.msc e7455bcbcd4853c5c6980d15d5217ff720aeb6f1c0256117929332554c106c19 -F README.md 963d30019abf0cc06b263cd2824bce022893f3f93a531758f6f04ff2194a16a8 +F README.md 6358805260a03ebead84e168bbf3740ddf3f683b477e478567186aa7afb490d3 F VERSION 73573d4545343f001bf5dc5461173a7c78c203dd046cabcf99153878cf25d3a6 F aclocal.m4 a5c22d164aff7ed549d53a90fa56d56955281f50 F art/sqlite370.eps aa97a671332b432a54e1d74ff5e8775be34200c2 @@ -89,18 +89,18 @@ F ext/fts3/unicode/CaseFolding.txt 8c678ca52ecc95e16bc7afc2dbf6fc9ffa05db8c F ext/fts3/unicode/UnicodeData.txt cd07314edb62d49fde34debdaf92fa2aa69011e7 F ext/fts3/unicode/mkunicode.tcl d5aebf022fa4577ee8cdf27468f0d847879993959101f6dbd6348ef0cfc324a7 F ext/fts3/unicode/parseunicode.tcl a981bd6466d12dd17967515801c3ff23f74a281be1a03cf1e6f52a6959fc77eb -F ext/fts5/extract_api_docs.tcl a36e54ec777172ddd3f9a88daf593b00848368e0 -F ext/fts5/fts5.h 05501612cc655504c5dce8ba765ab621d50fc478490089beaa0d75e00b23e520 -F ext/fts5/fts5Int.h 78a63cc0795186cde5384816a9403a68c65774b35d952e05b81a1b4b158e07c8 +F ext/fts5/extract_api_docs.tcl bc3a0ca78be7d3df08e7602c00ca48021ebae40682d75eb001bfdf6e54ffb44e +F ext/fts5/fts5.h 9d7867cc63631296462c90bafe57c4881b56e480fa510ffaee6a77049258c714 +F ext/fts5/fts5Int.h defa43c0932265138ee910ca416e6baccf8b774e0f3d610e74be1ab2880e9834 F ext/fts5/fts5_aux.c ee770eec0af8646db9e18fc01a0dad7345b5f5e8cbba236704cfae2d777022ad F ext/fts5/fts5_buffer.c 3001fbabb585d6de52947b44b455235072b741038391f830d6b729225eeaf6a5 -F ext/fts5/fts5_config.c 054359543566cbff1ba65a188330660a5457299513ac71c53b3a07d934c7b081 -F ext/fts5/fts5_expr.c bd3b81ce669c4104e34ffe66570af1999a317b142c15fccb112de9fb0caa57a6 -F ext/fts5/fts5_hash.c 076058f93327051952a752dc765df1acfe783eb11b419b30652aa1fc1f987902 -F ext/fts5/fts5_index.c 458cbed8a3e17617cbf7e80cdfb7612000b9bb3781f286b345fb9655858658cf -F ext/fts5/fts5_main.c a07ed863b8bd9e6fefb62db2fd40a3518eb30a5f7dcfda5be915dd2db45efa2f +F ext/fts5/fts5_config.c 8072a207034b51ae9b7694121d1b5715c794e94b275e088f70ae532378ca5cdf +F ext/fts5/fts5_expr.c b1ec526371b9ffde82341423a5b9753c42cbea629a41b69f26fa377d13b95a8e +F ext/fts5/fts5_hash.c adda4272be401566a6e0ba1acbe70ee5cb97fce944bc2e04dc707152a0ec91b1 +F ext/fts5/fts5_index.c 111fdb22e4d78a5c3813ac2c782409b3567291a48f19ae834cd50268c8e9fafc +F ext/fts5/fts5_main.c fb7ec495d663f40d18e420e1986316591041a70e1e4b4696ab2a7384e4c7fd7a F ext/fts5/fts5_storage.c 5d10b9bdcce5b90656cad13c7d12ad4148677d4b9e3fca0481fca56d6601426d -F ext/fts5/fts5_tcl.c b1445cbe69908c411df8084a10b2485500ac70a9c747cdc8cda175a3da59d8ae +F ext/fts5/fts5_tcl.c cf0fd0dbe64ec272491b749e0d594f563cda03336aeb60900129e6d18b0aefb8 F ext/fts5/fts5_test_mi.c 08c11ec968148d4cb4119d96d819f8c1f329812c568bac3684f5464be177d3ee F ext/fts5/fts5_test_tok.c 3cb0a9b508b30d17ef025ccddd26ae3dc8ddffbe76c057616e59a9aa85d36f3b F ext/fts5/fts5_tokenize.c 83cfcede3898001cab84432a36ce1503e3080cf9b1c682b022ec82e267ea4c13 @@ -109,8 +109,8 @@ F ext/fts5/fts5_varint.c e64d2113f6e1bfee0032972cffc1207b77af63319746951bf1d0988 F ext/fts5/fts5_vocab.c aed56169ae5c1aa9b8189c779ffeef04ed516d3c712c06914e6d91a6759f4e4a F ext/fts5/fts5parse.y eb526940f892ade5693f22ffd6c4f2702543a9059942772526eac1fde256bb05 F ext/fts5/mkportersteps.tcl 5acf962d2e0074f701620bb5308155fa1e4a63ba -F ext/fts5/test/fts5_common.tcl a9de9c2209cc4e7ae3c753e783504e67206c6c1467d08f209cd0c5923d3e8d8b -F ext/fts5/test/fts5aa.test ba5158eba7d61359becdfca895ef471072c7bf7b20e5e60dcb4d024c8419c926 +F ext/fts5/test/fts5_common.tcl 8b1848ac2baad10e444e4183034a52050b52d20b3796d9d30e78f01ab0d05583 +F ext/fts5/test/fts5aa.test 4db81519863244a3cab35795fe65ab6b592e7970c7409eba098b23ebbfc08d95 F ext/fts5/test/fts5ab.test bd932720c748383277456b81f91bc00453de2174f9762cd05f95d0495dc50390 F ext/fts5/test/fts5ac.test a7aa7e1fefc6e1918aa4d3111d5c44a09177168e962c5fd2cca9620de8a7ed6d F ext/fts5/test/fts5ad.test e8cf959dfcd57c8e46d6f5f25665686f3b6627130a9a981371dafdf6482790de @@ -171,6 +171,7 @@ F ext/fts5/test/fts5faultD.test e7ed7895abfe6bc98a5e853826f6b74956e7ba7f594f1860 F ext/fts5/test/fts5faultE.test 844586ce71dab4be85bb86880e87b624d089f851654cd22e4710c77eb8ce7075 F ext/fts5/test/fts5faultF.test 4abef99f86e99d9f0c6460dd68c586a766b6b9f1f660ada55bf2e8266bd1bbc1 F ext/fts5/test/fts5faultG.test d2e5a4d9a34e08dcaadcaeafef74d10cbc2abdd11aa2659a18af0294bf2812d3 +F ext/fts5/test/fts5faultH.test d845f45dac3e1a3f20c7e0a2be95280c95d3204c06802f86ab2c110e52ed3d14 F ext/fts5/test/fts5first.test 3fcf2365c00a15fc9704233674789a3b95131d12de18a9b996159f6909dc8079 F ext/fts5/test/fts5full.test e1701a112354e0ff9a1fdffb0c940c576530c33732ee20ac5e8361777070d717 F ext/fts5/test/fts5fuzz1.test 238d8c45f3b81342aa384de3e581ff2fa330bf922a7b69e484bbc06051a1080e @@ -191,6 +192,11 @@ F ext/fts5/test/fts5onepass.test f9b7d9b2c334900c6542a869760290e2ab5382af8fbd618 F ext/fts5/test/fts5optimize.test 36a752d24c818792032e4ff502936fc9cc5ef938721696396fdc79214b2717f1 F ext/fts5/test/fts5optimize2.test 93e742c36b487d8874621360af5b1ce4d39b04fb9e71ce9bc34015c5fc811785 F ext/fts5/test/fts5optimize3.test bf9c91bb927d0fb2b9a06318a217a0419183ac5913842e062c7e0b98ea5d0fca +F ext/fts5/test/fts5origintext.test d2796fa08ee7aecfabdc0c45bb8a2fb16a00ea8757e63fbc153b718dbe430a39 +F ext/fts5/test/fts5origintext2.test f3b9436de540828d01f0672df855b09ebc0863e126d5b56234701d71dfa73634 +F ext/fts5/test/fts5origintext3.test 0d25933506600452a5ab3873cbb418ed5f2de2446c3672b9997b1ea104b0e7f0 +F ext/fts5/test/fts5origintext4.test 296b1b1e6630d492b99db0769e8127087548f0e939376047716a68b77ca3c871 +F ext/fts5/test/fts5origintext5.test a037bdf7235a22033c4663837bdb12d9738245464a3ac2f60c71fc40d07ede7d F ext/fts5/test/fts5phrase.test 13e5d8e9083077b3d9c74315b3c92ec723cc6eb37c8155e0bfe1bba00559f07b F ext/fts5/test/fts5plan.test b65cfcca9ddd6fdaa118c61e17aeec8e8433bc5b6bb307abd116514f79c49c5a F ext/fts5/test/fts5porter.test 8d08010c28527db66bc3feebd2b8767504aaeb9b101a986342fa7833d49d0d15 @@ -213,7 +219,7 @@ F ext/fts5/test/fts5secure7.test fd03d0868d64340a1db8615b02e5508fea409de13910114 F ext/fts5/test/fts5secure8.test eb3579e9d58b0acad97e8082dee1f99b2d393198f03500b453c2b25761c0c298 F ext/fts5/test/fts5securefault.test dbca2b6a1c16700017f5051138991b705410889933f2a37c57ae8a23b296b10b F ext/fts5/test/fts5simple.test a298670508c1458b88ce6030440f26a30673931884eb5f4094ac1773b3ba217b -F ext/fts5/test/fts5simple2.test 258a1b0c590409bfa5271e872c79572b319d2a56554d0585f68f146a0da603f0 +F ext/fts5/test/fts5simple2.test 8dd2389ee75e21a1429fe87e5f8c7d9a97ad1470304a8a2d3ba4b8c3c345fecd F ext/fts5/test/fts5simple3.test d5c74a9d3ca71bd5dd5cacb7c55b86ea12cdddfc8b1910e3de2995206898380f F ext/fts5/test/fts5synonym.test 1651815b8008de170e8e600dcacc17521d765482ea8f074ae82cfa870d8bb7fb F ext/fts5/test/fts5synonym2.test 8f891fc49cc1e8daed727051e77e1f42849c784a6a54bef82564761b2cb3e016 @@ -262,7 +268,7 @@ F ext/jni/src/org/sqlite/jni/capi/CollationCallback.java e29bcfc540fdd343e2f5cca F ext/jni/src/org/sqlite/jni/capi/CollationNeededCallback.java 5bfa226a8e7a92e804fd52d6e42b4c7b875fa7a94f8e2c330af8cc244a8920ab F ext/jni/src/org/sqlite/jni/capi/CommitHookCallback.java 482f53dfec9e3ac2a9070d3fceebd56250932aaaf7c4f5bc8de29fc011416e0c F ext/jni/src/org/sqlite/jni/capi/ConfigLogCallback.java b995ca412f59b631803b93aa5b3684fce62e335d1e123207084c054abfd488d4 -F ext/jni/src/org/sqlite/jni/capi/ConfigSqlLogCallback.java e5723900b6458bc6288f52187090a78ebe0a20f403ac7c887ec9061dfe51aba7 w ext/jni/src/org/sqlite/jni/capi/ConfigSqllogCallback.java +F ext/jni/src/org/sqlite/jni/capi/ConfigSqlLogCallback.java e5723900b6458bc6288f52187090a78ebe0a20f403ac7c887ec9061dfe51aba7 F ext/jni/src/org/sqlite/jni/capi/NativePointerHolder.java b7036dcb1ef1b39f1f36ac605dde0ff1a24a9a01ade6aa1a605039443e089a61 F ext/jni/src/org/sqlite/jni/capi/OutputPointer.java 246b0e66c4603f41c567105a21189d138aaf8c58203ecd4928802333da553e7c F ext/jni/src/org/sqlite/jni/capi/PrepareMultiCallback.java 97352091abd7556167f4799076396279a51749fdae2b72a6ba61cd39b3df0359 @@ -462,7 +468,7 @@ F ext/rbu/rbuvacuum4.test ffccd22f67e2d0b380d2889685742159dfe0d19a3880ca3d2d1d69 F ext/rbu/sqlite3rbu.c d4ddf8f0e93772556e452a6c2814063cf47efb760a0834391a9d0cd9859fa4b9 F ext/rbu/sqlite3rbu.h 9d923eb135c5d04aa6afd7c39ca47b0d1d0707c100e02f19fdde6a494e414304 F ext/rbu/test_rbu.c ee6ede75147bc081fe9bc3931e6b206277418d14d3fbceea6fdc6216d9b47055 -F ext/recover/dbdata.c fc7147a68422cbbbaa481ee92ae1752cc25f5a24302bece1c70dcb76345bd736 +F ext/recover/dbdata.c b7746c2ec801b453840831311be4b31f8c8f9dd97791060a69bbf12392c78949 F ext/recover/recover1.test c484d01502239f11b61f23c1cee9f5dd19fa17617f8974e42e74d64639c524cf F ext/recover/recover_common.tcl a61306c1eb45c0c3fc45652c35b2d4ec19729e340bdf65a272ce4c229cefd85a F ext/recover/recoverbuild.test c74170e0f7b02456af41838afeb5353fdb985a48cc2331d661bbabbca7c6b8e3 @@ -570,7 +576,7 @@ F ext/userauth/sqlite3userauth.h 7f3ea8c4686db8e40b0a0e7a8e0b00fac13aa7a3 F ext/userauth/user-auth.txt e6641021a9210364665fe625d067617d03f27b04 F ext/userauth/userauth.c 7f00cded7dcaa5d47f54539b290a43d2e59f4b1eb5f447545fa865f002fc80cb F ext/wasm/EXPORTED_FUNCTIONS.fiddle 7fb73f7150ab79d83bb45a67d257553c905c78cd3d693101699243f36c5ae6c3 -F ext/wasm/GNUmakefile 0e362f3fc04eab6628cbe4f1e35f4ab4a200881f6b5f753b27fb45eabeddd9d2 +F ext/wasm/GNUmakefile 57439eec2b8b4d4074e5d861e93aab3f32bb0f44339a2b472c23f4e638b7e8a3 F ext/wasm/README-dist.txt 6382cb9548076fca472fb3330bbdba3a55c1ea0b180ff9253f084f07ff383576 F ext/wasm/README.md a8a2962c3aebdf8d2104a9102e336c5554e78fc6072746e5daf9c61514e7d193 F ext/wasm/SQLTester/GNUmakefile e0794f676d55819951bbfae45cc5e8d7818dc460492dc317ce7f0d2eca15caff @@ -600,8 +606,10 @@ F ext/wasm/api/sqlite3-vfs-opfs.c-pp.js 46c4afa6c50d7369252c104f274ad977a97e91cc F ext/wasm/api/sqlite3-wasm.c d0e09eb5ed3743c00294e30019e591c3aa150572ae7ffe8a8994568a7377589f F ext/wasm/api/sqlite3-worker1-promiser.c-pp.js f10c3ecd9df06f6320073c2ce230a7ed7c56034d8b88c1e57095f2a97faf423a F ext/wasm/api/sqlite3-worker1.c-pp.js a541112aa51e16705f13a99bb943c64efe178aa28c86704a955f8fd9afe4ba37 +F ext/wasm/batch-runner-sahpool.html e9a38fdeb36a13eac7b50241dfe7ae066fe3f51f5c0b0151e7baee5fce0d07a7 +F ext/wasm/batch-runner-sahpool.js 54a3ac228e6c4703fe72fb65c897e19156263a51fe9b7e21d2834a45e876aabd F ext/wasm/batch-runner.html 4deeed44fe41496dc6898d9fb17938ea3291f40f4bfb977e29d0cef96fbbe4c8 -F ext/wasm/batch-runner.js 0dad6a02ad796f1003d3b7048947d275c4d6277f63767b8e685c27df8fdac93e +F ext/wasm/batch-runner.js 05ec254f5dbfe605146d9640b3db17d6ef8c3fbef6aa8396051ca72bb5884e3f F ext/wasm/c-pp.c 6d80d8569d85713effe8b0818a3cf51dc779e3f0bf8dc88771b8998552ee25b4 F ext/wasm/common/SqliteTestUtil.js 7adaeffef757d8708418dc9190f72df22367b531831775804b31598b44f6aa51 F ext/wasm/common/emscripten.css 11bd104b6c0d597c67d40cc8ecc0a60dae2b965151e3b6a37fa5708bac3acd15 @@ -609,7 +617,7 @@ F ext/wasm/common/testing.css e97549bab24126c24e0daabfe2de9bb478fb0a69fdb2ddd0a7 F ext/wasm/common/whwasmutil.js 4c64594eecc7af4ae64259e95a71ba2a7edf118881aaff0bba86d0c7164e78e4 F ext/wasm/demo-123-worker.html a0b58d9caef098a626a1a1db567076fca4245e8d60ba94557ede8684350a81ed F ext/wasm/demo-123.html 8c70a412ce386bd3796534257935eb1e3ea5c581e5d5aea0490b8232e570a508 -F ext/wasm/demo-123.js 38aa8faec4d0ace1c973bc8a7a1533584463ebeecd4c420daa7d9687beeb9cb5 +F ext/wasm/demo-123.js c7b3cca50c55841c381a9ca4f9396e5bbdc6114273d0b10a43e378e32e7be5bf F ext/wasm/demo-jsstorage.html 409c4be4af5f207fb2877160724b91b33ea36a3cd8c204e8da1acb828ffe588e F ext/wasm/demo-jsstorage.js 44e3ae7ec2483b6c511384c3c290beb6f305c721186bcf5398ca4e00004a06b8 F ext/wasm/demo-worker1-promiser.html 1de7c248c7c2cfd4a5783d2aa154bce62d74c6de98ab22f5786620b3354ed15f @@ -623,8 +631,8 @@ F ext/wasm/fiddle/emscripten.css 3d253a6fdb8983a2ac983855bfbdd4b6fa1ff267c28d695 F ext/wasm/fiddle/fiddle-worker.js e0153f9af6500805c6f09c0b3cfdb7d857e9d6863dbee9d50d1628fccf5f4b4d F ext/wasm/fiddle/fiddle.html 550c5aafce40bd218de9bf26192749f69f9b10bc379423ecd2e162bcef885c08 F ext/wasm/fiddle/fiddle.js 974b995119ac443685d7d94d3b3c58c6a36540e9eb3fed7069d5653284071715 -F ext/wasm/index-dist.html 22379774f0ad4edcaaa8cf9c674c82e794cc557719a8addabed74eb8069d412e -F ext/wasm/index.html 4e7847b909f4ae0da8c829b150b79454050e53b3658431f138636257729cd42b +F ext/wasm/index-dist.html e91d76e4581185238fd3d42ed86ec600f7023ed3e3a944c5c356f25304bf1263 +F ext/wasm/index.html b31ce41c0da476d5ffcef23069b9d3415b419d65af5779096ebcfbcbade453a9 F ext/wasm/jaccwabyt/jaccwabyt.js 1264710db3cfbcb6887d95665b7aeba60c1126eaef789ca4cf1a4a17d5bc7f54 F ext/wasm/jaccwabyt/jaccwabyt.md 37911f00db12cbcca73aa1ed72594430365f30aafae2fa9c886961de74e5e0eb F ext/wasm/module-symbols.html dc476b403369b26a1a23773e13b80f41b9a49f0825e81435fe3600a7cfbbe337 @@ -632,8 +640,8 @@ F ext/wasm/scratchpad-wasmfs.html a3d7388f3c4b263676b58b526846e9d02dfcb4014ff29d F ext/wasm/scratchpad-wasmfs.mjs 66034b9256b218de59248aad796760a1584c1dd842231505895eff00dbd57c63 F ext/wasm/speedtest1-wasmfs.html 0e9d335a9b5b5fafe6e1bc8dc0f0ca7e22e6eb916682a2d7c36218bb7d67379d F ext/wasm/speedtest1-wasmfs.mjs ac5cadbf4ffe69e9eaac8b45e8523f030521e02bb67d654c6eb5236d9c456cbe -F ext/wasm/speedtest1-worker.html e33e2064bda572c0c3ebaec7306c35aa758d9d27e245d67e807f8cc4a9351cc5 -F ext/wasm/speedtest1-worker.js 315d26198c46be7c85e26fda15d80ef882424276abde25ffd8b026fb02a35d8c +F ext/wasm/speedtest1-worker.html 864b65ed78ce24847a348c180e7f267621a02ca027068a1863ec1c90187c1852 +F ext/wasm/speedtest1-worker.js 4d2ea70a3c24e05bdca78025202841f33d298c4fa9541a0070c3228661f89ecd F ext/wasm/speedtest1.html ff048b4a623aa192e83e143e48f1ce2a899846dd42c023fdedc8772b6e3f07da F ext/wasm/split-speedtest1-script.sh a3e271938d4d14ee49105eb05567c6a69ba4c1f1293583ad5af0cd3a3779e205 x F ext/wasm/sql/000-mandelbrot.sql 775337a4b80938ac8146aedf88808282f04d02d983d82675bd63d9c2d97a15f0 @@ -672,7 +680,7 @@ F src/btmutex.c 79a43670447eacc651519a429f6ece9fd638563cf95b469d6891185ddae2b522 F src/btree.c edb79e04dbb4df37a67bcebdde0de4495dfba3c9948a93f6d10f74fd61ab2391 x F src/btree.h 03e3356f5208bcab8eed4e094240fdac4a7f9f5ddf5e91045ce589f67d47c240 F src/btreeInt.h 3e2589726c4f105e653461814f65857465da68be1fac688de340c43b873f4062 -F src/build.c 189e4517d67f09f0a3e0d8e1faa6e2ef0c2e95f6ac82e33c912cb7efa2a359cc +F src/build.c 386eadecabe2e99a3783eb802ca01e665f8e0c2af0e0aab28161fd7def219a9d F src/callback.c db3a45e376deff6a16c0058163fe0ae2b73a2945f3f408ca32cf74960b28d490 F src/complete.c a3634ab1e687055cd002e11b8f43eb75c17da23e F src/ctime.c 047a0613c4c3ff65e05903d5b6931185b3df8f34b5178ad2f8d865ada4e9da44 @@ -680,7 +688,7 @@ F src/date.c 3b8d02977d160e128469de38493b4085f7c5cf4073193459909a6af3cf6d7c91 F src/dbpage.c 80e46e1df623ec40486da7a5086cb723b0275a6e2a7b01d9f9b5da0f04ba2782 F src/dbstat.c 3b677254d512fcafd4d0b341bf267b38b235ccfddbef24f9154e19360fa22e43 F src/delete.c cb766727c78e715f9fb7ec8a7d03658ed2a3016343ca687acfcec9083cdca500 -F src/expr.c e9a491c7f156e5b25641c28af11b735a424e108a21b9f83b6f3e51c99a8141d9 +F src/expr.c f4dbcca060fa95539f6705b5ec4cb5eeaba5c0d84ff64efca8edab7a1d776807 F src/fault.c 460f3e55994363812d9d60844b2a6de88826e007 F src/fkey.c a47610f0a5c6cb0ad79f8fcef039c01833dec0c751bb695f28dc0ec6a4c3ba00 F src/func.c 472f6dcfa39cf54f89a6aec76c79c225fb880a6c14469c15d361331662b9bf43 @@ -690,7 +698,7 @@ F src/hash.h 3340ab6e1d13e725571d7cee6d3e3135f0779a7d8e76a9ce0a85971fa3953c51 F src/hwtime.h f9c2dfb84dce7acf95ce6d289e46f5f9d3d1afd328e53da8f8e9008e3b3caae6 F src/in-operator.md 10cd8f4bcd225a32518407c2fb2484089112fd71 F src/insert.c 3f0a94082d978bbdd33c38fefea15346c6c6bffb70bc645a71dc0f1f87dd3276 -F src/json.c f93bf3df3651b1e01e2b57f7dc56f727e7b0e212d934eacf21c6fc8b31bf685e +F src/json.c c2e0fea06f40fb0319ed132fc181a25623585c943e08c690b522f216886ba316 F src/legacy.c d7874bc885906868cd51e6c2156698f2754f02d9eee1bae2d687323c3ca8e5aa F src/loadext.c 7432c944ff197046d67a1207790a1b13eec4548c85a9457eb0896bb3641dfb36 F src/main.c 1b89f3de98d1b59fec5bac1d66d6ece21f703821b8eaa0d53d9604c35309f6f9 @@ -723,19 +731,19 @@ F src/parse.y 020d80386eb216ec9520549106353c517d2bbc89be28752ffdca649a9eaf56ec F src/pcache.c 040b165f30622a21b7a9a77c6f2e4877a32fb7f22d4c7f0d2a6fa6833a156a75 F src/pcache.h 1497ce1b823cf00094bb0cf3bac37b345937e6f910890c626b16512316d3abf5 F src/pcache1.c 602acb23c471bb8d557a6f0083cc2be641d6cafcafa19e481eba7ef4c9ca0f00 -F src/pragma.c 087234c661e4f09e0cc53ae29ed59ed2e555d49c0dd12d913665a6bcf5ab9e1b +F src/pragma.c 54dd8e18efaa3c4eecce92dc68d1d148409d216367df3a99b93dbad9870fdbf1 F src/pragma.h e690a356c18e98414d2e870ea791c1be1545a714ba623719deb63f7f226d8bb7 F src/prepare.c 371f6115cb69286ebc12c6f2d7511279c2e47d9f54f475d46a554d687a3b312c -F src/printf.c 9da63b9ae1c14789bcae12840f5d800fd9302500cd2d62733fac77f0041b4750 +F src/printf.c 18fbdf028345c8fbe6044f5f5bfda5a10d48d6287afef088cc21b0ca57985640 F src/random.c 606b00941a1d7dd09c381d3279a058d771f406c5213c9932bbd93d5587be4b9c F src/resolve.c d017bad7ba8e778617701a0e986fdeb393d67d6afa84fb28ef4e8b8ad2acf916 F src/rowset.c 8432130e6c344b3401a8874c3cb49fefe6873fec593294de077afea2dce5ec97 -F src/select.c 85857bedd2913d888aa571755b48c54cd2e6e7fcb0087e19b226ee0368cfda1e -F src/shell.c.in 7bb83293775e1a5586d65212997442bc7acc70a2f1b781745da64ec3c2e4ea97 +F src/select.c a5058ae54211dc49faade7b78f21c69a3fa3ee652eb270fba6881e28e30e0497 +F src/shell.c.in 9b6c3e641de45651ad0b5e9c26cd2f72efabee28179a5315d15c54239515ee3a F src/sqlite.h.in d93a4821d2f792467a60f7dc81268d1bb8634f40c31694ef254cab4f9921f96a F src/sqlite3.rc 5121c9e10c3964d5755191c80dd1180c122fc3a8 F src/sqlite3ext.h 3f046c04ea3595d6bfda99b781926b17e672fd6d27da2ba6d8d8fc39981dcb54 -F src/sqliteInt.h cd171cba32c7a553e7623fbd82b68b36a1b6c81079ab963260777ea9b3abe4d9 +F src/sqliteInt.h 39b6fe0652c763d8ff6bd0a2427c009e5abe05fb70f5a0cb145a34bcc614fb90 F src/sqliteLimit.h 33b1c9baba578d34efe7dfdb43193b366111cdf41476b1e82699e14c11ee1fb6 F src/status.c 160c445d7d28c984a0eae38c144f6419311ed3eace59b44ac6dafc20db4af749 F src/table.c 0f141b58a16de7e2fbe81c308379e7279f4c6b50eb08efeec5892794a0ba30d1 @@ -800,11 +808,11 @@ F src/upsert.c fa125a8d3410ce9a97b02cb50f7ae68a2476c405c76aa692d3acf6b8586e9242 F src/utf.c ee39565f0843775cc2c81135751ddd93eceb91a673ea2c57f61c76f288b041a0 F src/util.c b22cc9f203a8c0b9ee5338a67f8860347d14845864c10248bebe84518a781677 F src/vacuum.c 604fcdaebe76f3497c855afcbf91b8fa5046b32de3045bab89cc008d68e40104 -F src/vdbe.c f8689104bd2ab4dc9a560b3b040d210c8e2a81627bd06b0b29189459d0b67eba -F src/vdbe.h 41485521f68e9437fdb7ec4a90f9d86ab294e9bb8281e33b235915e29122cfc0 +F src/vdbe.c 308ea5118a4422be0dd54bcdc988b2d12d88e1da21afaa08c56f7eeeae9dcf36 +F src/vdbe.h 88e19a982df9027ec1c177c793d1a5d34dc23d8f06e3b2d997f43688b05ee0eb F src/vdbeInt.h 949669dfd8a41550d27dcb905b494f2ccde9a2e6c1b0b04daa1227e2e74c2b2c -F src/vdbeapi.c b07df805110dc6e81f2a3f9cd4e83f56ea523277a59bcec489a12b740c1079e7 -F src/vdbeaux.c ac5a0786cdeff1a2d0f82261a17806016c1e460ce6d222f86ebc7e9d9f18d0d1 +F src/vdbeapi.c 8f57d60c89da0b60e6d4e272358c511f6bae4e24330bdb11f8b42f986d1bf21b +F src/vdbeaux.c 60216d2e30675e1e5f7e3964eac403c3d0e57c2c7b98ab45171928b2d55bfb80 F src/vdbeblob.c 13f9287b55b6356b4b1845410382d6bede203ceb29ef69388a4a3d007ffacbe5 F src/vdbemem.c 0012d5f01cc866833847c2f3ae4c318ac53a1cb3d28acad9c35e688039464cf0 F src/vdbesort.c 237840ca1947511fa59bd4e18b9eeae93f2af2468c34d2427b059f896230a547 @@ -1324,15 +1332,17 @@ F test/jrnlmode2.test 8759a1d4657c064637f8b079592651530db738419e1d649c6df7048cd7 F test/jrnlmode3.test 556b447a05be0e0963f4311e95ab1632b11c9eaa F test/json/README.md 63e3e589e1df8fd3cc1588ba1faaff659214003f8b77a15af5c6452b35e30ee2 F test/json/json-generator.tcl dc0dd0f393800c98658fc4c47eaa6af29d4e17527380cd28656fb261bddc8a3f +F test/json/json-q1-b.txt 1e180fe6491efab307e318b22879e3a736ac9a96539bbde7911a13ee5b33abc7 F test/json/json-q1.txt 65f9d1cdcc4cffa9823fb73ed936aae5658700cd001fde448f68bfb91c807307 -F test/json/json-speed-check.sh 8b7babf530faa58bd59d6d362cec8e9036a68c5457ff46f3b1f1511d21af6737 x -F test/json101.test bc05d2476fd6f7ead31ec05b43d1b24b2b193ae112fd8f0d2ed56d9a904f9fa5 -F test/json102.test 4c69694773a470f1fda34e5f4ba24920b35184fb66050b450fc2ef9ab5ad310b +F test/json/json-speed-check.sh b060a9a6c696c0a807d8929400fa11bd7113edc58b0d66b9795f424f8d0db326 x +F test/json101.test 70587d7d35ef9e2126364ba70f0c951f70827cfbd28649d779ff3df7e8f87547 +F test/json102.test 557a46e16df1aa9bdbc4076a71a45814ea0e7503d6621d87d42a8c04cbc2b0ef F test/json103.test 53df87f83a4e5fa0c0a56eb29ff6c94055c6eb919f33316d62161a8880112dbe F test/json104.test 1b844a70cddcfa2e4cd81a5db0657b2e61e7f00868310f24f56a9ba0114348c1 -F test/json105.test 11670a4387f4308ae0318cadcbd6a918ea7edcd19fbafde020720a073952675d -F test/json501.test c419deb835b70c1a2c8532936927bcc1146730328edd2052276715bfd209724d -F test/json502.test 98c38e3c4573841028a1381dfb81d4c3f9b105d39668167da10d055e503f6d0b +F test/json105.test e64a8d73677fbae67886642cd5076e2ef3efe89f8483b87595cf9c030216c9bd +F test/json501.test ab168a12eb6eb14d479f8c1cdae3ac062fd5a4679f17f976e96f1af518408330 +F test/json502.test 3c697e506fc38ccb455b49660b21b6e62e08ede0f2d0c869a7d171e17809093c +F test/jsonb01.test cace70765b36a36aec9a85a41ea65667d3bbf647d4400ddc3ac76f8fe7d94f90 F test/keyword1.test 37ef6bba5d2ed5b07ecdd6810571de2956599dff F test/kvtest.c 6e0228409ea7ca0497dad503fbd109badb5e59545d131014b6aaac68b56f484a F test/lastinsert.test 42e948fd6442f07d60acbd15d33fb86473e0ef63 @@ -2157,8 +2167,8 @@ F vsixtest/vsixtest.tcl 6a9a6ab600c25a91a7acc6293828957a386a8a93 F vsixtest/vsixtest.vcxproj.data 2ed517e100c66dc455b492e1a33350c1b20fbcdc F vsixtest/vsixtest.vcxproj.filters 37e51ffedcdb064aad6ff33b6148725226cd608e F vsixtest/vsixtest_TemporaryKey.pfx e5b1b036facdb453873e7084e1cae9102ccc67a0 -P 470152fd2847296e0fc325fdf098e49c888e99a2332fb45533f976ad85a67d8d 4c055b7a6e4533e1e571773456226ca7038ce372df3eedbbbcd9a81e8652a6cf -R 2f203e372f781f50a5b21c9b02ebd41a -U dan -Z 210541985fea5715c3a4748323eb934d +P 7f1b61a3ea2cc5132abf91b0f635be4fcaa082de7b33ca131874a75ae11ee576 ee70e4c1c9c41617850228e48d8df44f105cf2fbbe789340ceca6f27ad6ce5eb +R 4f15adc284ca13aa28f0a56a9198a0ad +U drh +Z 829e78e82ea416b7b0a8ac81648de04f # Remove this line to create a well-formed Fossil manifest. diff --git a/manifest.uuid b/manifest.uuid index 4d40b14004..1c12ec5ca8 100644 --- a/manifest.uuid +++ b/manifest.uuid @@ -1 +1 @@ -7f1b61a3ea2cc5132abf91b0f635be4fcaa082de7b33ca131874a75ae11ee576 \ No newline at end of file +457724e7380a8e768d945afb37fb634a6ec83ddebbc4ad144229c0dd498a5b2e \ No newline at end of file diff --git a/src/build.c b/src/build.c index 3d06792358..50ce97b51a 100644 --- a/src/build.c +++ b/src/build.c @@ -720,7 +720,7 @@ void sqlite3ColumnSetExpr( */ Expr *sqlite3ColumnExpr(Table *pTab, Column *pCol){ if( pCol->iDflt==0 ) return 0; - if( NEVER(!IsOrdinaryTable(pTab)) ) return 0; + if( !IsOrdinaryTable(pTab) ) return 0; if( NEVER(pTab->u.tab.pDfltList==0) ) return 0; if( NEVER(pTab->u.tab.pDfltList->nExpriDflt) ) return 0; return pTab->u.tab.pDfltList->a[pCol->iDflt-1].pExpr; @@ -873,6 +873,9 @@ void sqlite3DeleteTable(sqlite3 *db, Table *pTable){ if( db->pnBytesFreed==0 && (--pTable->nTabRef)>0 ) return; deleteTable(db, pTable); } +void sqlite3DeleteTableGeneric(sqlite3 *db, void *pTable){ + sqlite3DeleteTable(db, (Table*)pTable); +} /* @@ -1410,7 +1413,8 @@ void sqlite3ColumnPropertiesFromName(Table *pTab, Column *pCol){ /* ** Clean up the data structures associated with the RETURNING clause. */ -static void sqlite3DeleteReturning(sqlite3 *db, Returning *pRet){ +static void sqlite3DeleteReturning(sqlite3 *db, void *pArg){ + Returning *pRet = (Returning*)pArg; Hash *pHash; pHash = &(db->aDb[1].pSchema->trigHash); sqlite3HashInsert(pHash, pRet->zName, 0); @@ -1452,8 +1456,7 @@ void sqlite3AddReturning(Parse *pParse, ExprList *pList){ pParse->u1.pReturning = pRet; pRet->pParse = pParse; pRet->pReturnEL = pList; - sqlite3ParserAddCleanup(pParse, - (void(*)(sqlite3*,void*))sqlite3DeleteReturning, pRet); + sqlite3ParserAddCleanup(pParse, sqlite3DeleteReturning, pRet); testcase( pParse->earlyCleanup ); if( db->mallocFailed ) return; sqlite3_snprintf(sizeof(pRet->zName), pRet->zName, @@ -5692,4 +5695,7 @@ void sqlite3WithDelete(sqlite3 *db, With *pWith){ sqlite3DbFree(db, pWith); } } +void sqlite3WithDeleteGeneric(sqlite3 *db, void *pWith){ + sqlite3WithDelete(db, (With*)pWith); +} #endif /* !defined(SQLITE_OMIT_CTE) */ diff --git a/src/expr.c b/src/expr.c index 8a664ffb9c..047d2b6b39 100644 --- a/src/expr.c +++ b/src/expr.c @@ -1223,9 +1223,7 @@ void sqlite3ExprAddFunctionOrderBy( assert( ExprUseXList(pExpr) ); if( pExpr->x.pList==0 || NEVER(pExpr->x.pList->nExpr==0) ){ /* Ignore ORDER BY on zero-argument aggregates */ - sqlite3ParserAddCleanup(pParse, - (void(*)(sqlite3*,void*))sqlite3ExprListDelete, - pOrderBy); + sqlite3ParserAddCleanup(pParse, sqlite3ExprListDeleteGeneric, pOrderBy); return; } if( IsWindowFunc(pExpr) ){ @@ -1406,6 +1404,9 @@ static SQLITE_NOINLINE void sqlite3ExprDeleteNN(sqlite3 *db, Expr *p){ void sqlite3ExprDelete(sqlite3 *db, Expr *p){ if( p ) sqlite3ExprDeleteNN(db, p); } +void sqlite3ExprDeleteGeneric(sqlite3 *db, void *p){ + if( p ) sqlite3ExprDeleteNN(db, (Expr*)p); +} /* ** Clear both elements of an OnOrUsing object @@ -1431,9 +1432,7 @@ void sqlite3ClearOnOrUsing(sqlite3 *db, OnOrUsing *p){ ** pExpr to the pParse->pConstExpr list with a register number of 0. */ void sqlite3ExprDeferredDelete(Parse *pParse, Expr *pExpr){ - sqlite3ParserAddCleanup(pParse, - (void(*)(sqlite3*,void*))sqlite3ExprDelete, - pExpr); + sqlite3ParserAddCleanup(pParse, sqlite3ExprDeleteGeneric, pExpr); } /* Invoke sqlite3RenameExprUnmap() and sqlite3ExprDelete() on the @@ -2239,6 +2238,9 @@ static SQLITE_NOINLINE void exprListDeleteNN(sqlite3 *db, ExprList *pList){ void sqlite3ExprListDelete(sqlite3 *db, ExprList *pList){ if( pList ) exprListDeleteNN(db, pList); } +void sqlite3ExprListDeleteGeneric(sqlite3 *db, void *pList){ + if( pList ) exprListDeleteNN(db, (ExprList*)pList); +} /* ** Return the bitwise-OR of all Expr.flags fields in the given diff --git a/src/json.c b/src/json.c index 0f11ace02b..46d7a3fa8c 100644 --- a/src/json.c +++ b/src/json.c @@ -10,24 +10,142 @@ ** ****************************************************************************** ** -** This SQLite JSON functions. +** SQLite JSON functions. ** ** This file began as an extension in ext/misc/json1.c in 2015. That ** extension proved so useful that it has now been moved into the core. ** -** For the time being, all JSON is stored as pure text. (We might add -** a JSONB type in the future which stores a binary encoding of JSON in -** a BLOB, but there is no support for JSONB in the current implementation. -** This implementation parses JSON text at 250 MB/s, so it is hard to see -** how JSONB might improve on that.) +** The original design stored all JSON as pure text, canonical RFC-8259. +** Support for JSON-5 extensions was added with version 3.42.0 (2023-05-16). +** All generated JSON text still conforms strictly to RFC-8259, but text +** with JSON-5 extensions is accepted as input. +** +** Beginning with version 3.45.0 (pending), these routines also accept +** BLOB values that have JSON encoded using a binary representation we +** call JSONB. The name JSONB comes from PostgreSQL, however the on-disk +** format SQLite JSONB is completely different and incompatible with +** PostgreSQL JSONB. +** +** Decoding and interpreting JSONB is still O(N) where N is the size of +** the input, the same as text JSON. However, the constant of proportionality +** for JSONB is much smaller due to faster parsing. The size of each +** element in JSONB is encoded in its header, so there is no need to search +** for delimiters using persnickety syntax rules. JSONB seems to be about +** 3x faster than text JSON as a result. JSONB is also tends to be slightly +** smaller than text JSON, by 5% or 10%, but there are corner cases where +** JSONB can be slightly larger. So you are not far mistaken to say that +** a JSONB blob is the same size as the equivalent RFC-8259 text. +** +** +** THE JSONB ENCODING: +** +** Every JSON element is encoded in JSONB as a header and a payload. +** The header is between 1 and 9 bytes in size. The payload is zero +** or more bytes. +** +** The lower 4 bits of the first byte of the header determines the +** element type: +** +** 0: NULL +** 1: TRUE +** 2: FALSE +** 3: INT -- RFC-8259 integer literal +** 4: INT5 -- JSON5 integer literal +** 5: FLOAT -- RFC-8259 floating point literal +** 6: FLOAT5 -- JSON5 floating point literal +** 7: TEXT -- Text literal acceptable to both SQL and JSON +** 8: TEXTJ -- Text containing RFC-8259 escapes +** 9: TEXT5 -- Text containing JSON5 and/or RFC-8259 escapes +** 10: TEXTRAW -- Text containing unescaped syntax characters +** 11: ARRAY +** 12: OBJECT +** +** The other three possible values (13-15) are reserved for future +** enhancements. +** +** The upper 4 bits of the first byte determine the size of the header +** and sometimes also the size of the payload. If X is the first byte +** of the element and if X>>4 is between 0 and 11, then the payload +** will be that many bytes in size and the header is exactly one byte +** in size. Other four values for X>>4 (12-15) indicate that the header +** is more than one byte in size and that the payload size is determined +** by the remainder of the header, interpreted as a unsigned big-endian +** integer. +** +** Value of X>>4 Size integer Total header size +** ------------- -------------------- ----------------- +** 12 1 byte (0-255) 2 +** 13 2 byte (0-65535) 3 +** 14 4 byte (0-4294967295) 5 +** 15 8 byte (0-1.8e19) 9 +** +** The payload size need not be expressed in its minimal form. For example, +** if the payload size is 10, the size can be expressed in any of 5 different +** ways: (1) (X>>4)==10, (2) (X>>4)==12 following by on 0x0a byte, +** (3) (X>>4)==13 followed by 0x00 and 0x0a, (4) (X>>4)==14 followed by +** 0x00 0x00 0x00 0x0a, or (5) (X>>4)==15 followed by 7 bytes of 0x00 and +** a single byte of 0x0a. The shorter forms are preferred, of course, but +** sometimes when generating JSONB, the payload size is not known in advance +** and it is convenient to reserve sufficient header space to cover the +** largest possible payload size and then come back later and patch up +** the size when it becomes known, resulting in a non-minimal encoding. +** +** The value (X>>4)==15 is not actually used in the current implementation +** (as SQLite is currently unable handle BLOBs larger than about 2GB) +** but is included in the design to allow for future enhancements. +** +** The payload follows the header. NULL, TRUE, and FALSE have no payload and +** their payload size must always be zero. The payload for INT, INT5, +** FLOAT, FLOAT5, TEXT, TEXTJ, TEXT5, and TEXTROW is text. Note that the +** "..." or '...' delimiters are omitted from the various text encodings. +** The payload for ARRAY and OBJECT is a list of additional elements that +** are the content for the array or object. The payload for an OBJECT +** must be an even number of elements. The first element of each pair is +** the label and must be of type TEXT, TEXTJ, TEXT5, or TEXTRAW. +** +** A valid JSONB blob consists of a single element, as described above. +** Usually this will be an ARRAY or OBJECT element which has many more +** elements as its content. But the overall blob is just a single element. +** +** Input validation for JSONB blobs simply checks that the element type +** code is between 0 and 12 and that the total size of the element +** (header plus payload) is the same as the size of the BLOB. If those +** checks are true, the BLOB is assumed to be JSONB and processing continues. +** Errors are only raised if some other miscoding is discovered during +** processing. */ #ifndef SQLITE_OMIT_JSON #include "sqliteInt.h" +/* JSONB element types +*/ +#define JSONB_NULL 0 /* "null" */ +#define JSONB_TRUE 1 /* "true" */ +#define JSONB_FALSE 2 /* "false" */ +#define JSONB_INT 3 /* integer acceptable to JSON and SQL */ +#define JSONB_INT5 4 /* integer in 0x000 notation */ +#define JSONB_FLOAT 5 /* float acceptable to JSON and SQL */ +#define JSONB_FLOAT5 6 /* float with JSON5 extensions */ +#define JSONB_TEXT 7 /* Text compatible with both JSON and SQL */ +#define JSONB_TEXTJ 8 /* Text with JSON escapes */ +#define JSONB_TEXT5 9 /* Text with JSON-5 escape */ +#define JSONB_TEXTRAW 10 /* SQL text that needs escaping for JSON */ +#define JSONB_ARRAY 11 /* An array */ +#define JSONB_OBJECT 12 /* An object */ + +/* Human-readable names for the JSONB values. The index for each +** string must correspond to the JSONB_* integer above. +*/ +static const char * const jsonbType[] = { + "null", "true", "false", "integer", "integer", + "real", "real", "text", "text", "text", + "text", "array", "object", "", "", "", "" +}; + /* ** Growing our own isspace() routine this way is twice as fast as ** the library isspace() function, resulting in a 7% overall performance -** increase for the parser. (Ubuntu14.10 gcc 4.8.4 x64 with -Os). +** increase for the text-JSON parser. (Ubuntu14.10 gcc 4.8.4 x64 with -Os). */ static const char jsonIsSpace[] = { 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 1, 0, 0, 1, 0, 0, @@ -48,10 +166,16 @@ static const char jsonIsSpace[] = { 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, }; -#define fast_isspace(x) (jsonIsSpace[(unsigned char)x]) +#define jsonIsspace(x) (jsonIsSpace[(unsigned char)x]) /* -** Characters that are special to JSON. Control charaters, +** The set of all space characters recognized by jsonIsspace(). +** Useful as the second argument to strspn(). +*/ +static const char jsonSpaces[] = "\011\012\015\040"; + +/* +** Characters that are special to JSON. Control characters, ** '"' and '\\'. */ static const char jsonIsOk[256] = { @@ -74,22 +198,41 @@ static const char jsonIsOk[256] = { 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1 }; - -#if !defined(SQLITE_DEBUG) && !defined(SQLITE_COVERAGE_TEST) -# define VVA(X) -#else -# define VVA(X) X -#endif - /* Objects */ +typedef struct JsonCache JsonCache; typedef struct JsonString JsonString; -typedef struct JsonNode JsonNode; typedef struct JsonParse JsonParse; -typedef struct JsonCleanup JsonCleanup; + +/* +** Magic number used for the JSON parse cache in sqlite3_get_auxdata() +*/ +#define JSON_CACHE_ID (-429938) /* Cache entry */ +#define JSON_CACHE_SIZE 4 /* Max number of cache entries */ + +/* A cache mapping JSON text into JSONB blobs. +** +** Each cache entry is a JsonParse object with the following restrictions: +** +** * The bReadOnly flag must be set +** +** * The aBlob[] array must be owned by the JsonParse object. In other +** words, nBlobAlloc must be non-zero. +** +** * zJson must be an RCStr. In other words bJsonIsRCStr must be true. +*/ +struct JsonCache { + sqlite3 *db; /* Database connection */ + int nUsed; /* Number of active entries in the cache */ + JsonParse *a[JSON_CACHE_SIZE]; /* One line for each cache entry */ +}; /* An instance of this object represents a JSON string ** under construction. Really, this is a generic string accumulator ** that can be and is used to create strings other than JSON. +** +** If the generated string is longer than will fit into the zSpace[] buffer, +** then it will be an RCStr string. This aids with caching of large +** JSON strings. */ struct JsonString { sqlite3_context *pCtx; /* Function context - put error messages here */ @@ -97,121 +240,74 @@ struct JsonString { u64 nAlloc; /* Bytes of storage available in zBuf[] */ u64 nUsed; /* Bytes of zBuf[] currently used */ u8 bStatic; /* True if zBuf is static space */ - u8 bErr; /* True if an error has been encountered */ + u8 eErr; /* True if an error has been encountered */ char zSpace[100]; /* Initial static space */ }; -/* A deferred cleanup task. A list of JsonCleanup objects might be -** run when the JsonParse object is destroyed. -*/ -struct JsonCleanup { - JsonCleanup *pJCNext; /* Next in a list */ - void (*xOp)(void*); /* Routine to run */ - void *pArg; /* Argument to xOp() */ -}; +/* Allowed values for JsonString.eErr */ +#define JSTRING_OOM 0x01 /* Out of memory */ +#define JSTRING_MALFORMED 0x02 /* Malformed JSONB */ +#define JSTRING_ERR 0x04 /* Error already sent to sqlite3_result */ -/* JSON type values +/* The "subtype" set for text JSON values passed through using +** sqlite3_result_subtype() and sqlite3_value_subtype(). */ -#define JSON_SUBST 0 /* Special edit node. Uses u.iPrev */ -#define JSON_NULL 1 -#define JSON_TRUE 2 -#define JSON_FALSE 3 -#define JSON_INT 4 -#define JSON_REAL 5 -#define JSON_STRING 6 -#define JSON_ARRAY 7 -#define JSON_OBJECT 8 - -/* The "subtype" set for JSON values */ #define JSON_SUBTYPE 74 /* Ascii for "J" */ /* -** Names of the various JSON types: +** Bit values for the flags passed into various SQL function implementations +** via the sqlite3_user_data() value. */ -static const char * const jsonType[] = { - "subst", - "null", "true", "false", "integer", "real", "text", "array", "object" -}; - -/* Bit values for the JsonNode.jnFlag field -*/ -#define JNODE_RAW 0x01 /* Content is raw, not JSON encoded */ -#define JNODE_ESCAPE 0x02 /* Content is text with \ escapes */ -#define JNODE_REMOVE 0x04 /* Do not output */ -#define JNODE_REPLACE 0x08 /* Target of a JSON_SUBST node */ -#define JNODE_APPEND 0x10 /* More ARRAY/OBJECT entries at u.iAppend */ -#define JNODE_LABEL 0x20 /* Is a label of an object */ -#define JNODE_JSON5 0x40 /* Node contains JSON5 enhancements */ +#define JSON_JSON 0x01 /* Result is always JSON */ +#define JSON_SQL 0x02 /* Result is always SQL */ +#define JSON_ABPATH 0x03 /* Allow abbreviated JSON path specs */ +#define JSON_ISSET 0x04 /* json_set(), not json_insert() */ +#define JSON_BLOB 0x08 /* Use the BLOB output format */ -/* A single node of parsed JSON. An array of these nodes describes -** a parse of JSON + edits. +/* A parsed JSON value. Lifecycle: ** -** Use the json_parse() SQL function (available when compiled with -** -DSQLITE_DEBUG) to see a dump of complete JsonParse objects, including -** a complete listing and decoding of the array of JsonNodes. -*/ -struct JsonNode { - u8 eType; /* One of the JSON_ type values */ - u8 jnFlags; /* JNODE flags */ - u8 eU; /* Which union element to use */ - u32 n; /* Bytes of content for INT, REAL or STRING - ** Number of sub-nodes for ARRAY and OBJECT - ** Node that SUBST applies to */ - union { - const char *zJContent; /* 1: Content for INT, REAL, and STRING */ - u32 iAppend; /* 2: More terms for ARRAY and OBJECT */ - u32 iKey; /* 3: Key for ARRAY objects in json_tree() */ - u32 iPrev; /* 4: Previous SUBST node, or 0 */ - } u; -}; - - -/* A parsed and possibly edited JSON string. Lifecycle: +** 1. JSON comes in and is parsed into a JSONB value in aBlob. The +** original text is stored in zJson. This step is skipped if the +** input is JSONB instead of text JSON. ** -** 1. JSON comes in and is parsed into an array aNode[]. The original -** JSON text is stored in zJson. +** 2. The aBlob[] array is searched using the JSON path notation, if needed. +** +** 3. Zero or more changes are made to aBlob[] (via json_remove() or +** json_replace() or json_patch() or similar). ** -** 2. Zero or more changes are made (via json_remove() or json_replace() -** or similar) to the aNode[] array. -** -** 3. A new, edited and mimified JSON string is generated from aNode -** and stored in zAlt. The JsonParse object always owns zAlt. -** -** Step 1 always happens. Step 2 and 3 may or may not happen, depending -** on the operation. -** -** aNode[].u.zJContent entries typically point into zJson. Hence zJson -** must remain valid for the lifespan of the parse. For edits, -** aNode[].u.zJContent might point to malloced space other than zJson. -** Entries in pClup are responsible for freeing that extra malloced space. -** -** When walking the parse tree in aNode[], edits are ignored if useMod is -** false. +** 4. New JSON text is generated from the aBlob[] for output. This step +** is skipped the function is one of the jsonb_* functions that returns +** JSONB instead of text JSON. */ struct JsonParse { - u32 nNode; /* Number of slots of aNode[] used */ - u32 nAlloc; /* Number of slots of aNode[] allocated */ - JsonNode *aNode; /* Array of nodes containing the parse */ - char *zJson; /* Original JSON string (before edits) */ - char *zAlt; /* Revised and/or mimified JSON */ - u32 *aUp; /* Index of parent of each node */ - JsonCleanup *pClup;/* Cleanup operations prior to freeing this object */ + u8 *aBlob; /* JSONB representation of JSON value */ + u32 nBlob; /* Bytes of aBlob[] actually used */ + u32 nBlobAlloc; /* Bytes allocated to aBlob[]. 0 if aBlob is external */ + char *zJson; /* Json text used for parsing */ + int nJson; /* Length of the zJson string in bytes */ u16 iDepth; /* Nesting depth */ u8 nErr; /* Number of errors seen */ u8 oom; /* Set to true if out of memory */ u8 bJsonIsRCStr; /* True if zJson is an RCStr */ u8 hasNonstd; /* True if input uses non-standard features like JSON5 */ - u8 useMod; /* Actually use the edits contain inside aNode */ - u8 hasMod; /* aNode contains edits from the original zJson */ + u8 bReadOnly; /* Do not modify. */ u32 nJPRef; /* Number of references to this object */ - int nJson; /* Length of the zJson string in bytes */ - int nAlt; /* Length of alternative JSON string zAlt, in bytes */ u32 iErr; /* Error location in zJson[] */ - u32 iSubst; /* Last JSON_SUBST entry in aNode[] */ - u32 iHold; /* Age of this entry in the cache for LRU replacement */ + /* Search and edit information. See jsonLookupStep() */ + u8 eEdit; /* Edit operation to apply */ + int delta; /* Size change due to the edit */ + u32 nIns; /* Number of bytes to insert */ + u32 iLabel; /* Location of label if search landed on an object value */ + u8 *aIns; /* Content to be inserted */ }; +/* Allowed values for JsonParse.eEdit */ +#define JEDIT_DEL 1 /* Delete if exists */ +#define JEDIT_REPL 2 /* Overwrite if exists */ +#define JEDIT_INS 3 /* Insert if not exists */ +#define JEDIT_SET 4 /* Insert or overwrite */ + /* ** Maximum nesting depth of JSON for this implementation. ** @@ -219,15 +315,147 @@ struct JsonParse { ** descent parser. A depth of 1000 is far deeper than any sane JSON ** should go. Historical note: This limit was 2000 prior to version 3.42.0 */ -#define JSON_MAX_DEPTH 1000 +#ifndef SQLITE_JSON_MAX_DEPTH +# define JSON_MAX_DEPTH 1000 +#else +# define JSON_MAX_DEPTH SQLITE_JSON_MAX_DEPTH +#endif + +/* +** Allowed values for the flgs argument to jsonParseFuncArg(); +*/ +#define JSON_EDITABLE 0x01 /* Generate a writable JsonParse object */ +#define JSON_KEEPERROR 0x02 /* Return non-NULL even if there is an error */ + +/************************************************************************** +** Forward references +**************************************************************************/ +static void jsonReturnStringAsBlob(JsonString*); +static int jsonFuncArgMightBeBinary(sqlite3_value *pJson); +static u32 jsonXlateBlobToText(const JsonParse*,u32,JsonString*); +static void jsonReturnParse(sqlite3_context*,JsonParse*); +static JsonParse *jsonParseFuncArg(sqlite3_context*,sqlite3_value*,u32); +static void jsonParseFree(JsonParse*); + +/************************************************************************** +** Utility routines for dealing with JsonCache objects +**************************************************************************/ + +/* +** Free a JsonCache object. +*/ +static void jsonCacheDelete(JsonCache *p){ + int i; + for(i=0; inUsed; i++){ + jsonParseFree(p->a[i]); + } + sqlite3DbFree(p->db, p); +} +static void jsonCacheDeleteGeneric(void *p){ + jsonCacheDelete((JsonCache*)p); +} + +/* +** Insert a new entry into the cache. If the cache is full, expel +** the least recently used entry. Return SQLITE_OK on success or a +** result code otherwise. +** +** Cache entries are stored in age order, oldest first. +*/ +static int jsonCacheInsert( + sqlite3_context *ctx, /* The SQL statement context holding the cache */ + JsonParse *pParse /* The parse object to be added to the cache */ +){ + JsonCache *p; + + assert( pParse->zJson!=0 ); + assert( pParse->bJsonIsRCStr ); + p = sqlite3_get_auxdata(ctx, JSON_CACHE_ID); + if( p==0 ){ + sqlite3 *db = sqlite3_context_db_handle(ctx); + p = sqlite3DbMallocZero(db, sizeof(*p)); + if( p==0 ) return SQLITE_NOMEM; + p->db = db; + sqlite3_set_auxdata(ctx, JSON_CACHE_ID, p, jsonCacheDeleteGeneric); + p = sqlite3_get_auxdata(ctx, JSON_CACHE_ID); + if( p==0 ) return SQLITE_NOMEM; + } + if( p->nUsed >= JSON_CACHE_SIZE ){ + jsonParseFree(p->a[0]); + memmove(p->a, &p->a[1], (JSON_CACHE_SIZE-1)*sizeof(p->a[0])); + p->nUsed = JSON_CACHE_SIZE-1; + } + assert( pParse->nBlobAlloc>0 ); + pParse->eEdit = 0; + pParse->nJPRef++; + pParse->bReadOnly = 1; + p->a[p->nUsed] = pParse; + p->nUsed++; + return SQLITE_OK; +} + +/* +** Search for a cached translation the json text supplied by pArg. Return +** the JsonParse object if found. Return NULL if not found. +** +** When a match if found, the matching entry is moved to become the +** most-recently used entry if it isn't so already. +** +** The JsonParse object returned still belongs to the Cache and might +** be deleted at any moment. If the caller whants the JsonParse to +** linger, it needs to increment the nPJRef reference counter. +*/ +static JsonParse *jsonCacheSearch( + sqlite3_context *ctx, /* The SQL statement context holding the cache */ + sqlite3_value *pArg /* Function argument containing SQL text */ +){ + JsonCache *p; + int i; + const char *zJson; + int nJson; + + if( sqlite3_value_type(pArg)!=SQLITE_TEXT ){ + return 0; + } + zJson = (const char*)sqlite3_value_text(pArg); + if( zJson==0 ) return 0; + nJson = sqlite3_value_bytes(pArg); + + p = sqlite3_get_auxdata(ctx, JSON_CACHE_ID); + if( p==0 ){ + return 0; + } + for(i=0; inUsed; i++){ + if( p->a[i]->zJson==zJson ) break; + } + if( i>=p->nUsed ){ + for(i=0; inUsed; i++){ + if( p->a[i]->nJson!=nJson ) continue; + if( memcmp(p->a[i]->zJson, zJson, nJson)==0 ) break; + } + } + if( inUsed ){ + if( inUsed-1 ){ + /* Make the matching entry the most recently used entry */ + JsonParse *tmp = p->a[i]; + memmove(&p->a[i], &p->a[i+1], (p->nUsed-i-1)*sizeof(tmp)); + p->a[p->nUsed-1] = tmp; + i = p->nUsed - 1; + } + return p->a[i]; + }else{ + return 0; + } +} /************************************************************************** ** Utility routines for dealing with JsonString objects **************************************************************************/ -/* Set the JsonString object to an empty string +/* Turn uninitialized bulk memory into a valid JsonString object +** holding a zero-length string. */ -static void jsonZero(JsonString *p){ +static void jsonStringZero(JsonString *p){ p->zBuf = p->zSpace; p->nAlloc = sizeof(p->zSpace); p->nUsed = 0; @@ -236,39 +464,39 @@ static void jsonZero(JsonString *p){ /* Initialize the JsonString object */ -static void jsonInit(JsonString *p, sqlite3_context *pCtx){ +static void jsonStringInit(JsonString *p, sqlite3_context *pCtx){ p->pCtx = pCtx; - p->bErr = 0; - jsonZero(p); + p->eErr = 0; + jsonStringZero(p); } /* Free all allocated memory and reset the JsonString object back to its ** initial state. */ -static void jsonReset(JsonString *p){ +static void jsonStringReset(JsonString *p){ if( !p->bStatic ) sqlite3RCStrUnref(p->zBuf); - jsonZero(p); + jsonStringZero(p); } /* Report an out-of-memory (OOM) condition */ -static void jsonOom(JsonString *p){ - p->bErr = 1; - sqlite3_result_error_nomem(p->pCtx); - jsonReset(p); +static void jsonStringOom(JsonString *p){ + p->eErr |= JSTRING_OOM; + if( p->pCtx ) sqlite3_result_error_nomem(p->pCtx); + jsonStringReset(p); } /* Enlarge pJson->zBuf so that it can hold at least N more bytes. ** Return zero on success. Return non-zero on an OOM error */ -static int jsonGrow(JsonString *p, u32 N){ +static int jsonStringGrow(JsonString *p, u32 N){ u64 nTotal = NnAlloc ? p->nAlloc*2 : p->nAlloc+N+10; char *zNew; if( p->bStatic ){ - if( p->bErr ) return 1; + if( p->eErr ) return 1; zNew = sqlite3RCStrNew(nTotal); if( zNew==0 ){ - jsonOom(p); + jsonStringOom(p); return SQLITE_NOMEM; } memcpy(zNew, p->zBuf, (size_t)p->nUsed); @@ -277,8 +505,8 @@ static int jsonGrow(JsonString *p, u32 N){ }else{ p->zBuf = sqlite3RCStrResize(p->zBuf, nTotal); if( p->zBuf==0 ){ - p->bErr = 1; - jsonZero(p); + p->eErr |= JSTRING_OOM; + jsonStringZero(p); return SQLITE_NOMEM; } } @@ -288,20 +516,20 @@ static int jsonGrow(JsonString *p, u32 N){ /* Append N bytes from zIn onto the end of the JsonString string. */ -static SQLITE_NOINLINE void jsonAppendExpand( +static SQLITE_NOINLINE void jsonStringExpandAndAppend( JsonString *p, const char *zIn, u32 N ){ assert( N>0 ); - if( jsonGrow(p,N) ) return; + if( jsonStringGrow(p,N) ) return; memcpy(p->zBuf+p->nUsed, zIn, N); p->nUsed += N; } static void jsonAppendRaw(JsonString *p, const char *zIn, u32 N){ if( N==0 ) return; if( N+p->nUsed >= p->nAlloc ){ - jsonAppendExpand(p,zIn,N); + jsonStringExpandAndAppend(p,zIn,N); }else{ memcpy(p->zBuf+p->nUsed, zIn, N); p->nUsed += N; @@ -310,7 +538,7 @@ static void jsonAppendRaw(JsonString *p, const char *zIn, u32 N){ static void jsonAppendRawNZ(JsonString *p, const char *zIn, u32 N){ assert( N>0 ); if( N+p->nUsed >= p->nAlloc ){ - jsonAppendExpand(p,zIn,N); + jsonStringExpandAndAppend(p,zIn,N); }else{ memcpy(p->zBuf+p->nUsed, zIn, N); p->nUsed += N; @@ -322,7 +550,7 @@ static void jsonAppendRawNZ(JsonString *p, const char *zIn, u32 N){ */ static void jsonPrintf(int N, JsonString *p, const char *zFormat, ...){ va_list ap; - if( (p->nUsed + N >= p->nAlloc) && jsonGrow(p, N) ) return; + if( (p->nUsed + N >= p->nAlloc) && jsonStringGrow(p, N) ) return; va_start(ap, zFormat); sqlite3_vsnprintf(N, p->zBuf+p->nUsed, zFormat, ap); va_end(ap); @@ -332,7 +560,7 @@ static void jsonPrintf(int N, JsonString *p, const char *zFormat, ...){ /* Append a single character */ static SQLITE_NOINLINE void jsonAppendCharExpand(JsonString *p, char c){ - if( jsonGrow(p,1) ) return; + if( jsonStringGrow(p,1) ) return; p->zBuf[p->nUsed++] = c; } static void jsonAppendChar(JsonString *p, char c){ @@ -343,24 +571,17 @@ static void jsonAppendChar(JsonString *p, char c){ } } -/* Try to force the string to be a zero-terminated RCStr string. +/* Make sure there is a zero terminator on p->zBuf[] ** ** Return true on success. Return false if an OOM prevents this ** from happening. */ -static int jsonForceRCStr(JsonString *p){ +static int jsonStringTerminate(JsonString *p){ jsonAppendChar(p, 0); - if( p->bErr ) return 0; p->nUsed--; - if( p->bStatic==0 ) return 1; - p->nAlloc = 0; - p->nUsed++; - jsonGrow(p, p->nUsed); - p->nUsed--; - return p->bStatic==0; + return p->eErr==0; } - /* Append a comma separator to the output buffer, if the previous ** character is not '[' or '{'. */ @@ -373,21 +594,41 @@ static void jsonAppendSeparator(JsonString *p){ } /* Append the N-byte string in zIn to the end of the JsonString string -** under construction. Enclose the string in "..." and escape -** any double-quotes or backslash characters contained within the +** under construction. Enclose the string in double-quotes ("...") and +** escape any double-quotes or backslash characters contained within the ** string. +** +** This routine is a high-runner. There is a measurable performance +** increase associated with unwinding the jsonIsOk[] loop. */ static void jsonAppendString(JsonString *p, const char *zIn, u32 N){ - u32 i; - if( zIn==0 || ((N+p->nUsed+2 >= p->nAlloc) && jsonGrow(p,N+2)!=0) ) return; + u32 k; + u8 c; + const u8 *z = (const u8*)zIn; + if( z==0 ) return; + if( (N+p->nUsed+2 >= p->nAlloc) && jsonStringGrow(p,N+2)!=0 ) return; p->zBuf[p->nUsed++] = '"'; - for(i=0; izBuf[p->nUsed++] = c; - }else if( c=='"' || c=='\\' ){ + while( 1 /*exit-by-break*/ ){ + k = 0; + while( k+1=N ){ + if( k>0 ){ + memcpy(&p->zBuf[p->nUsed], z, k); + p->nUsed += k; + } + break; + } + if( k>0 ){ + memcpy(&p->zBuf[p->nUsed], z, k); + p->nUsed += k; + z += k; + N -= k; + } + c = z[0]; + if( c=='"' || c=='\\' ){ json_simple_escape: - if( (p->nUsed+N+3-i > p->nAlloc) && jsonGrow(p,N+3-i)!=0 ) return; + if( (p->nUsed+N+3 > p->nAlloc) && jsonStringGrow(p,N+3)!=0 ) return; p->zBuf[p->nUsed++] = '\\'; p->zBuf[p->nUsed++] = c; }else if( c=='\'' ){ @@ -408,7 +649,7 @@ static void jsonAppendString(JsonString *p, const char *zIn, u32 N){ c = aSpecial[c]; goto json_simple_escape; } - if( (p->nUsed+N+7+i > p->nAlloc) && jsonGrow(p,N+7-i)!=0 ) return; + if( (p->nUsed+N+7 > p->nAlloc) && jsonStringGrow(p,N+7)!=0 ) return; p->zBuf[p->nUsed++] = '\\'; p->zBuf[p->nUsed++] = 'u'; p->zBuf[p->nUsed++] = '0'; @@ -416,146 +657,18 @@ static void jsonAppendString(JsonString *p, const char *zIn, u32 N){ p->zBuf[p->nUsed++] = "0123456789abcdef"[c>>4]; p->zBuf[p->nUsed++] = "0123456789abcdef"[c&0xf]; } + z++; + N--; } p->zBuf[p->nUsed++] = '"'; assert( p->nUsednAlloc ); } /* -** The zIn[0..N] string is a JSON5 string literal. Append to p a translation -** of the string literal that standard JSON and that omits all JSON5 -** features. +** Append an sqlite3_value (such as a function parameter) to the JSON +** string under construction in p. */ -static void jsonAppendNormalizedString(JsonString *p, const char *zIn, u32 N){ - u32 i; - jsonAppendChar(p, '"'); - zIn++; - N -= 2; - while( N>0 ){ - for(i=0; i0 ){ - jsonAppendRawNZ(p, zIn, i); - zIn += i; - N -= i; - if( N==0 ) break; - } - if( zIn[0]=='"' ){ - jsonAppendRawNZ(p, "\\\"", 2); - zIn++; - N--; - continue; - } - assert( zIn[0]=='\\' ); - switch( (u8)zIn[1] ){ - case '\'': - jsonAppendChar(p, '\''); - break; - case 'v': - jsonAppendRawNZ(p, "\\u0009", 6); - break; - case 'x': - jsonAppendRawNZ(p, "\\u00", 4); - jsonAppendRawNZ(p, &zIn[2], 2); - zIn += 2; - N -= 2; - break; - case '0': - jsonAppendRawNZ(p, "\\u0000", 6); - break; - case '\r': - if( zIn[2]=='\n' ){ - zIn++; - N--; - } - break; - case '\n': - break; - case 0xe2: - assert( N>=4 ); - assert( 0x80==(u8)zIn[2] ); - assert( 0xa8==(u8)zIn[3] || 0xa9==(u8)zIn[3] ); - zIn += 2; - N -= 2; - break; - default: - jsonAppendRawNZ(p, zIn, 2); - break; - } - zIn += 2; - N -= 2; - } - jsonAppendChar(p, '"'); -} - -/* -** The zIn[0..N] string is a JSON5 integer literal. Append to p a translation -** of the string literal that standard JSON and that omits all JSON5 -** features. -*/ -static void jsonAppendNormalizedInt(JsonString *p, const char *zIn, u32 N){ - if( zIn[0]=='+' ){ - zIn++; - N--; - }else if( zIn[0]=='-' ){ - jsonAppendChar(p, '-'); - zIn++; - N--; - } - if( zIn[0]=='0' && (zIn[1]=='x' || zIn[1]=='X') ){ - sqlite3_int64 i = 0; - int rc = sqlite3DecOrHexToI64(zIn, &i); - if( rc<=1 ){ - jsonPrintf(100,p,"%lld",i); - }else{ - assert( rc==2 ); - jsonAppendRawNZ(p, "9.0e999", 7); - } - return; - } - assert( N>0 ); - jsonAppendRawNZ(p, zIn, N); -} - -/* -** The zIn[0..N] string is a JSON5 real literal. Append to p a translation -** of the string literal that standard JSON and that omits all JSON5 -** features. -*/ -static void jsonAppendNormalizedReal(JsonString *p, const char *zIn, u32 N){ - u32 i; - if( zIn[0]=='+' ){ - zIn++; - N--; - }else if( zIn[0]=='-' ){ - jsonAppendChar(p, '-'); - zIn++; - N--; - } - if( zIn[0]=='.' ){ - jsonAppendChar(p, '0'); - } - for(i=0; i0 ){ - jsonAppendRawNZ(p, zIn, N); - } -} - - - -/* -** Append a function parameter value to the JSON string under -** construction. -*/ -static void jsonAppendValue( +static void jsonAppendSqlValue( JsonString *p, /* Append to this JSON string */ sqlite3_value *pValue /* Value to append */ ){ @@ -585,291 +698,127 @@ static void jsonAppendValue( break; } default: { - if( p->bErr==0 ){ + if( jsonFuncArgMightBeBinary(pValue) ){ + JsonParse px; + memset(&px, 0, sizeof(px)); + px.aBlob = (u8*)sqlite3_value_blob(pValue); + px.nBlob = sqlite3_value_bytes(pValue); + jsonXlateBlobToText(&px, 0, p); + }else if( p->eErr==0 ){ sqlite3_result_error(p->pCtx, "JSON cannot hold BLOB values", -1); - p->bErr = 2; - jsonReset(p); + p->eErr = JSTRING_ERR; + jsonStringReset(p); } break; } } } - -/* Make the JSON in p the result of the SQL function. +/* Make the text in p (which is probably a generated JSON text string) +** the result of the SQL function. ** -** The JSON string is reset. +** The JsonString is reset. +** +** If pParse and ctx are both non-NULL, then the SQL string in p is +** loaded into the zJson field of the pParse object as a RCStr and the +** pParse is added to the cache. */ -static void jsonResult(JsonString *p){ - if( p->bErr==0 ){ - if( p->bStatic ){ +static void jsonReturnString( + JsonString *p, /* String to return */ + JsonParse *pParse, /* JSONB source or NULL */ + sqlite3_context *ctx /* Where to cache */ +){ + assert( (pParse!=0)==(ctx!=0) ); + assert( ctx==0 || ctx==p->pCtx ); + if( p->eErr==0 ){ + int flags = SQLITE_PTR_TO_INT(sqlite3_user_data(p->pCtx)); + if( flags & JSON_BLOB ){ + jsonReturnStringAsBlob(p); + }else if( p->bStatic ){ sqlite3_result_text64(p->pCtx, p->zBuf, p->nUsed, SQLITE_TRANSIENT, SQLITE_UTF8); - }else if( jsonForceRCStr(p) ){ - sqlite3RCStrRef(p->zBuf); - sqlite3_result_text64(p->pCtx, p->zBuf, p->nUsed, + }else if( jsonStringTerminate(p) ){ + if( pParse && pParse->bJsonIsRCStr==0 && pParse->nBlobAlloc>0 ){ + int rc; + pParse->zJson = sqlite3RCStrRef(p->zBuf); + pParse->nJson = p->nUsed; + pParse->bJsonIsRCStr = 1; + rc = jsonCacheInsert(ctx, pParse); + if( rc==SQLITE_NOMEM ){ + sqlite3_result_error_nomem(ctx); + jsonStringReset(p); + return; + } + } + sqlite3_result_text64(p->pCtx, sqlite3RCStrRef(p->zBuf), p->nUsed, sqlite3RCStrUnref, SQLITE_UTF8); + }else{ + sqlite3_result_error_nomem(p->pCtx); } - } - if( p->bErr==1 ){ + }else if( p->eErr & JSTRING_OOM ){ sqlite3_result_error_nomem(p->pCtx); + }else if( p->eErr & JSTRING_MALFORMED ){ + sqlite3_result_error(p->pCtx, "malformed JSON", -1); } - jsonReset(p); + jsonStringReset(p); } /************************************************************************** -** Utility routines for dealing with JsonNode and JsonParse objects +** Utility routines for dealing with JsonParse objects **************************************************************************/ -/* -** Return the number of consecutive JsonNode slots need to represent -** the parsed JSON at pNode. The minimum answer is 1. For ARRAY and -** OBJECT types, the number might be larger. -** -** Appended elements are not counted. The value returned is the number -** by which the JsonNode counter should increment in order to go to the -** next peer value. -*/ -static u32 jsonNodeSize(JsonNode *pNode){ - return pNode->eType>=JSON_ARRAY ? pNode->n+1 : 1; -} - /* ** Reclaim all memory allocated by a JsonParse object. But do not ** delete the JsonParse object itself. */ static void jsonParseReset(JsonParse *pParse){ - while( pParse->pClup ){ - JsonCleanup *pTask = pParse->pClup; - pParse->pClup = pTask->pJCNext; - pTask->xOp(pTask->pArg); - sqlite3_free(pTask); - } assert( pParse->nJPRef<=1 ); - if( pParse->aNode ){ - sqlite3_free(pParse->aNode); - pParse->aNode = 0; - } - pParse->nNode = 0; - pParse->nAlloc = 0; - if( pParse->aUp ){ - sqlite3_free(pParse->aUp); - pParse->aUp = 0; - } if( pParse->bJsonIsRCStr ){ sqlite3RCStrUnref(pParse->zJson); pParse->zJson = 0; + pParse->nJson = 0; pParse->bJsonIsRCStr = 0; } - if( pParse->zAlt ){ - sqlite3RCStrUnref(pParse->zAlt); - pParse->zAlt = 0; + if( pParse->nBlobAlloc ){ + sqlite3_free(pParse->aBlob); + pParse->aBlob = 0; + pParse->nBlob = 0; + pParse->nBlobAlloc = 0; } } /* -** Free a JsonParse object that was obtained from sqlite3_malloc(). -** -** Note that destroying JsonParse might call sqlite3RCStrUnref() to -** destroy the zJson value. The RCStr object might recursively invoke -** JsonParse to destroy this pParse object again. Take care to ensure -** that this recursive destructor sequence terminates harmlessly. +** Decrement the reference count on the JsonParse object. When the +** count reaches zero, free the object. */ static void jsonParseFree(JsonParse *pParse){ - if( pParse->nJPRef>1 ){ - pParse->nJPRef--; - }else{ - jsonParseReset(pParse); - sqlite3_free(pParse); - } -} - -/* -** Add a cleanup task to the JsonParse object. -** -** If an OOM occurs, the cleanup operation happens immediately -** and this function returns SQLITE_NOMEM. -*/ -static int jsonParseAddCleanup( - JsonParse *pParse, /* Add the cleanup task to this parser */ - void(*xOp)(void*), /* The cleanup task */ - void *pArg /* Argument to the cleanup */ -){ - JsonCleanup *pTask = sqlite3_malloc64( sizeof(*pTask) ); - if( pTask==0 ){ - pParse->oom = 1; - xOp(pArg); - return SQLITE_ERROR; - } - pTask->pJCNext = pParse->pClup; - pParse->pClup = pTask; - pTask->xOp = xOp; - pTask->pArg = pArg; - return SQLITE_OK; -} - -/* -** Convert the JsonNode pNode into a pure JSON string and -** append to pOut. Subsubstructure is also included. Return -** the number of JsonNode objects that are encoded. -*/ -static void jsonRenderNode( - JsonParse *pParse, /* the complete parse of the JSON */ - JsonNode *pNode, /* The node to render */ - JsonString *pOut /* Write JSON here */ -){ - assert( pNode!=0 ); - while( (pNode->jnFlags & JNODE_REPLACE)!=0 && pParse->useMod ){ - u32 idx = (u32)(pNode - pParse->aNode); - u32 i = pParse->iSubst; - while( 1 /*exit-by-break*/ ){ - assert( inNode ); - assert( pParse->aNode[i].eType==JSON_SUBST ); - assert( pParse->aNode[i].eU==4 ); - assert( pParse->aNode[i].u.iPrevaNode[i].n==idx ){ - pNode = &pParse->aNode[i+1]; - break; - } - i = pParse->aNode[i].u.iPrev; - } - } - switch( pNode->eType ){ - default: { - assert( pNode->eType==JSON_NULL ); - jsonAppendRawNZ(pOut, "null", 4); - break; - } - case JSON_TRUE: { - jsonAppendRawNZ(pOut, "true", 4); - break; - } - case JSON_FALSE: { - jsonAppendRawNZ(pOut, "false", 5); - break; - } - case JSON_STRING: { - assert( pNode->eU==1 ); - if( pNode->jnFlags & JNODE_RAW ){ - if( pNode->jnFlags & JNODE_LABEL ){ - jsonAppendChar(pOut, '"'); - jsonAppendRaw(pOut, pNode->u.zJContent, pNode->n); - jsonAppendChar(pOut, '"'); - }else{ - jsonAppendString(pOut, pNode->u.zJContent, pNode->n); - } - }else if( pNode->jnFlags & JNODE_JSON5 ){ - jsonAppendNormalizedString(pOut, pNode->u.zJContent, pNode->n); - }else{ - assert( pNode->n>0 ); - jsonAppendRawNZ(pOut, pNode->u.zJContent, pNode->n); - } - break; - } - case JSON_REAL: { - assert( pNode->eU==1 ); - if( pNode->jnFlags & JNODE_JSON5 ){ - jsonAppendNormalizedReal(pOut, pNode->u.zJContent, pNode->n); - }else{ - assert( pNode->n>0 ); - jsonAppendRawNZ(pOut, pNode->u.zJContent, pNode->n); - } - break; - } - case JSON_INT: { - assert( pNode->eU==1 ); - if( pNode->jnFlags & JNODE_JSON5 ){ - jsonAppendNormalizedInt(pOut, pNode->u.zJContent, pNode->n); - }else{ - assert( pNode->n>0 ); - jsonAppendRawNZ(pOut, pNode->u.zJContent, pNode->n); - } - break; - } - case JSON_ARRAY: { - u32 j = 1; - jsonAppendChar(pOut, '['); - for(;;){ - while( j<=pNode->n ){ - if( (pNode[j].jnFlags & JNODE_REMOVE)==0 || pParse->useMod==0 ){ - jsonAppendSeparator(pOut); - jsonRenderNode(pParse, &pNode[j], pOut); - } - j += jsonNodeSize(&pNode[j]); - } - if( (pNode->jnFlags & JNODE_APPEND)==0 ) break; - if( pParse->useMod==0 ) break; - assert( pNode->eU==2 ); - pNode = &pParse->aNode[pNode->u.iAppend]; - j = 1; - } - jsonAppendChar(pOut, ']'); - break; - } - case JSON_OBJECT: { - u32 j = 1; - jsonAppendChar(pOut, '{'); - for(;;){ - while( j<=pNode->n ){ - if( (pNode[j+1].jnFlags & JNODE_REMOVE)==0 || pParse->useMod==0 ){ - jsonAppendSeparator(pOut); - jsonRenderNode(pParse, &pNode[j], pOut); - jsonAppendChar(pOut, ':'); - jsonRenderNode(pParse, &pNode[j+1], pOut); - } - j += 1 + jsonNodeSize(&pNode[j+1]); - } - if( (pNode->jnFlags & JNODE_APPEND)==0 ) break; - if( pParse->useMod==0 ) break; - assert( pNode->eU==2 ); - pNode = &pParse->aNode[pNode->u.iAppend]; - j = 1; - } - jsonAppendChar(pOut, '}'); - break; + if( pParse ){ + if( pParse->nJPRef>1 ){ + pParse->nJPRef--; + }else{ + jsonParseReset(pParse); + sqlite3_free(pParse); } } } -/* -** Return a JsonNode and all its descendants as a JSON string. -*/ -static void jsonReturnJson( - JsonParse *pParse, /* The complete JSON */ - JsonNode *pNode, /* Node to return */ - sqlite3_context *pCtx, /* Return value for this function */ - int bGenerateAlt, /* Also store the rendered text in zAlt */ - int omitSubtype /* Do not call sqlite3_result_subtype() */ -){ - JsonString s; - if( pParse->oom ){ - sqlite3_result_error_nomem(pCtx); - return; - } - if( pParse->nErr==0 ){ - jsonInit(&s, pCtx); - jsonRenderNode(pParse, pNode, &s); - if( bGenerateAlt && pParse->zAlt==0 && jsonForceRCStr(&s) ){ - pParse->zAlt = sqlite3RCStrRef(s.zBuf); - pParse->nAlt = s.nUsed; - } - jsonResult(&s); - if( !omitSubtype ) sqlite3_result_subtype(pCtx, JSON_SUBTYPE); - } -} +/************************************************************************** +** Utility routines for the JSON text parser +**************************************************************************/ /* ** Translate a single byte of Hex into an integer. -** This routine only works if h really is a valid hexadecimal -** character: 0..9a..fA..F +** This routine only gives a correct answer if h really is a valid hexadecimal +** character: 0..9a..fA..F. But unlike sqlite3HexToInt(), it does not +** assert() if the digit is not hex. */ static u8 jsonHexToInt(int h){ - assert( (h>='0' && h<='9') || (h>='a' && h<='f') || (h>='A' && h<='F') ); +#ifdef SQLITE_ASCII + h += 9*(1&(h>>6)); +#endif #ifdef SQLITE_EBCDIC h += 9*(1&~(h>>4)); -#else - h += 9*(1&(h>>6)); #endif return (u8)(h & 0xf); } @@ -879,10 +828,6 @@ static u8 jsonHexToInt(int h){ */ static u32 jsonHexToInt4(const char *z){ u32 v; - assert( sqlite3Isxdigit(z[0]) ); - assert( sqlite3Isxdigit(z[1]) ); - assert( sqlite3Isxdigit(z[2]) ); - assert( sqlite3Isxdigit(z[3]) ); v = (jsonHexToInt(z[0])<<12) + (jsonHexToInt(z[1])<<8) + (jsonHexToInt(z[2])<<4) @@ -890,282 +835,6 @@ static u32 jsonHexToInt4(const char *z){ return v; } -/* -** Make the JsonNode the return value of the function. -*/ -static void jsonReturn( - JsonParse *pParse, /* Complete JSON parse tree */ - JsonNode *pNode, /* Node to return */ - sqlite3_context *pCtx, /* Return value for this function */ - int omitSubtype /* Do not call sqlite3_result_subtype() */ -){ - switch( pNode->eType ){ - default: { - assert( pNode->eType==JSON_NULL ); - sqlite3_result_null(pCtx); - break; - } - case JSON_TRUE: { - sqlite3_result_int(pCtx, 1); - break; - } - case JSON_FALSE: { - sqlite3_result_int(pCtx, 0); - break; - } - case JSON_INT: { - sqlite3_int64 i = 0; - int rc; - int bNeg = 0; - const char *z; - - assert( pNode->eU==1 ); - z = pNode->u.zJContent; - if( z[0]=='-' ){ z++; bNeg = 1; } - else if( z[0]=='+' ){ z++; } - rc = sqlite3DecOrHexToI64(z, &i); - if( rc<=1 ){ - sqlite3_result_int64(pCtx, bNeg ? -i : i); - }else if( rc==3 && bNeg ){ - sqlite3_result_int64(pCtx, SMALLEST_INT64); - }else{ - goto to_double; - } - break; - } - case JSON_REAL: { - double r; - const char *z; - assert( pNode->eU==1 ); - to_double: - z = pNode->u.zJContent; - sqlite3AtoF(z, &r, sqlite3Strlen30(z), SQLITE_UTF8); - sqlite3_result_double(pCtx, r); - break; - } - case JSON_STRING: { - if( pNode->jnFlags & JNODE_RAW ){ - assert( pNode->eU==1 ); - sqlite3_result_text(pCtx, pNode->u.zJContent, pNode->n, - SQLITE_TRANSIENT); - }else if( (pNode->jnFlags & JNODE_ESCAPE)==0 ){ - /* JSON formatted without any backslash-escapes */ - assert( pNode->eU==1 ); - sqlite3_result_text(pCtx, pNode->u.zJContent+1, pNode->n-2, - SQLITE_TRANSIENT); - }else{ - /* Translate JSON formatted string into raw text */ - u32 i; - u32 n = pNode->n; - const char *z; - char *zOut; - u32 j; - u32 nOut = n; - assert( pNode->eU==1 ); - z = pNode->u.zJContent; - zOut = sqlite3_malloc( nOut+1 ); - if( zOut==0 ){ - sqlite3_result_error_nomem(pCtx); - break; - } - for(i=1, j=0; i>6)); - zOut[j++] = 0x80 | (v&0x3f); - }else{ - u32 vlo; - if( (v&0xfc00)==0xd800 - && i>18); - zOut[j++] = 0x80 | ((v>>12)&0x3f); - zOut[j++] = 0x80 | ((v>>6)&0x3f); - zOut[j++] = 0x80 | (v&0x3f); - }else{ - zOut[j++] = 0xe0 | (v>>12); - zOut[j++] = 0x80 | ((v>>6)&0x3f); - zOut[j++] = 0x80 | (v&0x3f); - } - } - continue; - }else if( c=='b' ){ - c = '\b'; - }else if( c=='f' ){ - c = '\f'; - }else if( c=='n' ){ - c = '\n'; - }else if( c=='r' ){ - c = '\r'; - }else if( c=='t' ){ - c = '\t'; - }else if( c=='v' ){ - c = '\v'; - }else if( c=='\'' || c=='"' || c=='/' || c=='\\' ){ - /* pass through unchanged */ - }else if( c=='0' ){ - c = 0; - }else if( c=='x' ){ - c = (jsonHexToInt(z[i+1])<<4) | jsonHexToInt(z[i+2]); - i += 2; - }else if( c=='\r' && z[i+1]=='\n' ){ - i++; - continue; - }else if( 0xe2==(u8)c ){ - assert( 0x80==(u8)z[i+1] ); - assert( 0xa8==(u8)z[i+2] || 0xa9==(u8)z[i+2] ); - i += 2; - continue; - }else{ - continue; - } - } /* end if( c=='\\' ) */ - zOut[j++] = c; - } /* end for() */ - zOut[j] = 0; - sqlite3_result_text(pCtx, zOut, j, sqlite3_free); - } - break; - } - case JSON_ARRAY: - case JSON_OBJECT: { - jsonReturnJson(pParse, pNode, pCtx, 0, omitSubtype); - break; - } - } -} - -/* Forward reference */ -static int jsonParseAddNode(JsonParse*,u32,u32,const char*); - -/* -** A macro to hint to the compiler that a function should not be -** inlined. -*/ -#if defined(__GNUC__) -# define JSON_NOINLINE __attribute__((noinline)) -#elif defined(_MSC_VER) && _MSC_VER>=1310 -# define JSON_NOINLINE __declspec(noinline) -#else -# define JSON_NOINLINE -#endif - - -/* -** Add a single node to pParse->aNode after first expanding the -** size of the aNode array. Return the index of the new node. -** -** If an OOM error occurs, set pParse->oom and return -1. -*/ -static JSON_NOINLINE int jsonParseAddNodeExpand( - JsonParse *pParse, /* Append the node to this object */ - u32 eType, /* Node type */ - u32 n, /* Content size or sub-node count */ - const char *zContent /* Content */ -){ - u32 nNew; - JsonNode *pNew; - assert( pParse->nNode>=pParse->nAlloc ); - if( pParse->oom ) return -1; - nNew = pParse->nAlloc*2 + 10; - pNew = sqlite3_realloc64(pParse->aNode, sizeof(JsonNode)*nNew); - if( pNew==0 ){ - pParse->oom = 1; - return -1; - } - pParse->nAlloc = sqlite3_msize(pNew)/sizeof(JsonNode); - pParse->aNode = pNew; - assert( pParse->nNodenAlloc ); - return jsonParseAddNode(pParse, eType, n, zContent); -} - -/* -** Create a new JsonNode instance based on the arguments and append that -** instance to the JsonParse. Return the index in pParse->aNode[] of the -** new node, or -1 if a memory allocation fails. -*/ -static int jsonParseAddNode( - JsonParse *pParse, /* Append the node to this object */ - u32 eType, /* Node type */ - u32 n, /* Content size or sub-node count */ - const char *zContent /* Content */ -){ - JsonNode *p; - assert( pParse->aNode!=0 || pParse->nNode>=pParse->nAlloc ); - if( pParse->nNode>=pParse->nAlloc ){ - return jsonParseAddNodeExpand(pParse, eType, n, zContent); - } - assert( pParse->aNode!=0 ); - p = &pParse->aNode[pParse->nNode]; - assert( p!=0 ); - p->eType = (u8)(eType & 0xff); - p->jnFlags = (u8)(eType >> 8); - VVA( p->eU = zContent ? 1 : 0 ); - p->n = n; - p->u.zJContent = zContent; - return pParse->nNode++; -} - -/* -** Add an array of new nodes to the current pParse->aNode array. -** Return the index of the first node added. -** -** If an OOM error occurs, set pParse->oom. -*/ -static void jsonParseAddNodeArray( - JsonParse *pParse, /* Append the node to this object */ - JsonNode *aNode, /* Array of nodes to add */ - u32 nNode /* Number of elements in aNew */ -){ - assert( aNode!=0 ); - assert( nNode>=1 ); - if( pParse->nNode + nNode > pParse->nAlloc ){ - u32 nNew = pParse->nNode + nNode; - JsonNode *aNew = sqlite3_realloc64(pParse->aNode, nNew*sizeof(JsonNode)); - if( aNew==0 ){ - pParse->oom = 1; - return; - } - pParse->nAlloc = sqlite3_msize(aNew)/sizeof(JsonNode); - pParse->aNode = aNew; - } - memcpy(&pParse->aNode[pParse->nNode], aNode, nNode*sizeof(JsonNode)); - pParse->nNode += nNode; -} - -/* -** Add a new JSON_SUBST node. The node immediately following -** this new node will be the substitute content for iNode. -*/ -static int jsonParseAddSubstNode( - JsonParse *pParse, /* Add the JSON_SUBST here */ - u32 iNode /* References this node */ -){ - int idx = jsonParseAddNode(pParse, JSON_SUBST, iNode, 0); - if( pParse->oom ) return -1; - pParse->aNode[iNode].jnFlags |= JNODE_REPLACE; - pParse->aNode[idx].eU = 4; - pParse->aNode[idx].u.iPrev = pParse->iSubst; - pParse->iSubst = idx; - pParse->hasMod = 1; - pParse->useMod = 1; - return idx; -} - /* ** Return true if z[] begins with 2 (or more) hexadecimal digits */ @@ -1319,63 +988,303 @@ static const struct NanInfName { char *zMatch; char *zRepl; } aNanInfName[] = { - { 'i', 'I', 3, JSON_REAL, 7, "inf", "9.0e999" }, - { 'i', 'I', 8, JSON_REAL, 7, "infinity", "9.0e999" }, - { 'n', 'N', 3, JSON_NULL, 4, "NaN", "null" }, - { 'q', 'Q', 4, JSON_NULL, 4, "QNaN", "null" }, - { 's', 'S', 4, JSON_NULL, 4, "SNaN", "null" }, + { 'i', 'I', 3, JSONB_FLOAT, 7, "inf", "9.0e999" }, + { 'i', 'I', 8, JSONB_FLOAT, 7, "infinity", "9.0e999" }, + { 'n', 'N', 3, JSONB_NULL, 4, "NaN", "null" }, + { 'q', 'Q', 4, JSONB_NULL, 4, "QNaN", "null" }, + { 's', 'S', 4, JSONB_NULL, 4, "SNaN", "null" }, }; + /* -** Parse a single JSON value which begins at pParse->zJson[i]. Return the -** index of the first character past the end of the value parsed. +** Report the wrong number of arguments for json_insert(), json_replace() +** or json_set(). +*/ +static void jsonWrongNumArgs( + sqlite3_context *pCtx, + const char *zFuncName +){ + char *zMsg = sqlite3_mprintf("json_%s() needs an odd number of arguments", + zFuncName); + sqlite3_result_error(pCtx, zMsg, -1); + sqlite3_free(zMsg); +} + +/**************************************************************************** +** Utility routines for dealing with the binary BLOB representation of JSON +****************************************************************************/ + +/* +** Expand pParse->aBlob so that it holds at least N bytes. ** -** Special return values: +** Return the number of errors. +*/ +static int jsonBlobExpand(JsonParse *pParse, u32 N){ + u8 *aNew; + u32 t; + assert( N>pParse->nBlobAlloc ); + if( pParse->nBlobAlloc==0 ){ + t = 100; + }else{ + t = pParse->nBlobAlloc*2; + } + if( taBlob, t ); + if( aNew==0 ){ pParse->oom = 1; return 1; } + pParse->aBlob = aNew; + pParse->nBlobAlloc = t; + return 0; +} + +/* +** If pParse->aBlob is not previously editable (because it is taken +** from sqlite3_value_blob(), as indicated by the fact that +** pParse->nBlobAlloc==0 and pParse->nBlob>0) then make it editable +** by making a copy into space obtained from malloc. +** +** Return true on success. Return false on OOM. +*/ +static int jsonBlobMakeEditable(JsonParse *pParse, u32 nExtra){ + u8 *aOld; + u32 nSize; + assert( !pParse->bReadOnly ); + if( pParse->oom ) return 0; + if( pParse->nBlobAlloc>0 ) return 1; + aOld = pParse->aBlob; + nSize = pParse->nBlob + nExtra; + pParse->aBlob = 0; + if( jsonBlobExpand(pParse, nSize) ){ + return 0; + } + assert( pParse->nBlobAlloc >= pParse->nBlob + nExtra ); + memcpy(pParse->aBlob, aOld, pParse->nBlob); + return 1; +} + +/* Expand pParse->aBlob and append one bytes. +*/ +static SQLITE_NOINLINE void jsonBlobExpandAndAppendOneByte( + JsonParse *pParse, + u8 c +){ + jsonBlobExpand(pParse, pParse->nBlob+1); + if( pParse->oom==0 ){ + assert( pParse->nBlob+1<=pParse->nBlobAlloc ); + pParse->aBlob[pParse->nBlob++] = c; + } +} + +/* Append a single character. +*/ +static void jsonBlobAppendOneByte(JsonParse *pParse, u8 c){ + if( pParse->nBlob >= pParse->nBlobAlloc ){ + jsonBlobExpandAndAppendOneByte(pParse, c); + }else{ + pParse->aBlob[pParse->nBlob++] = c; + } +} + +/* Slow version of jsonBlobAppendNode() that first resizes the +** pParse->aBlob structure. +*/ +static void jsonBlobAppendNode(JsonParse*,u8,u32,const void*); +static SQLITE_NOINLINE void jsonBlobExpandAndAppendNode( + JsonParse *pParse, + u8 eType, + u32 szPayload, + const void *aPayload +){ + if( jsonBlobExpand(pParse, pParse->nBlob+szPayload+9) ) return; + jsonBlobAppendNode(pParse, eType, szPayload, aPayload); +} + + +/* Append an node type byte together with the payload size and +** possibly also the payload. +** +** If aPayload is not NULL, then it is a pointer to the payload which +** is also appended. If aPayload is NULL, the pParse->aBlob[] array +** is resized (if necessary) so that it is big enough to hold the +** payload, but the payload is not appended and pParse->nBlob is left +** pointing to where the first byte of payload will eventually be. +*/ +static void jsonBlobAppendNode( + JsonParse *pParse, /* The JsonParse object under construction */ + u8 eType, /* Node type. One of JSONB_* */ + u32 szPayload, /* Number of bytes of payload */ + const void *aPayload /* The payload. Might be NULL */ +){ + u8 *a; + if( pParse->nBlob+szPayload+9 > pParse->nBlobAlloc ){ + jsonBlobExpandAndAppendNode(pParse,eType,szPayload,aPayload); + return; + } + assert( pParse->aBlob!=0 ); + a = &pParse->aBlob[pParse->nBlob]; + if( szPayload<=11 ){ + a[0] = eType | (szPayload<<4); + pParse->nBlob += 1; + }else if( szPayload<=0xff ){ + a[0] = eType | 0xc0; + a[1] = szPayload & 0xff; + pParse->nBlob += 2; + }else if( szPayload<=0xffff ){ + a[0] = eType | 0xd0; + a[1] = (szPayload >> 8) & 0xff; + a[2] = szPayload & 0xff; + pParse->nBlob += 3; + }else{ + a[0] = eType | 0xe0; + a[1] = (szPayload >> 24) & 0xff; + a[2] = (szPayload >> 16) & 0xff; + a[3] = (szPayload >> 8) & 0xff; + a[4] = szPayload & 0xff; + pParse->nBlob += 5; + } + if( aPayload ){ + pParse->nBlob += szPayload; + memcpy(&pParse->aBlob[pParse->nBlob-szPayload], aPayload, szPayload); + } +} + +/* Change the payload size for the node at index i to be szPayload. +*/ +static int jsonBlobChangePayloadSize( + JsonParse *pParse, + u32 i, + u32 szPayload +){ + u8 *a; + u8 szType; + u8 nExtra; + u8 nNeeded; + int delta; + if( pParse->oom ) return 0; + a = &pParse->aBlob[i]; + szType = a[0]>>4; + if( szType<=11 ){ + nExtra = 0; + }else if( szType==12 ){ + nExtra = 1; + }else if( szType==13 ){ + nExtra = 2; + }else{ + nExtra = 4; + } + if( szPayload<=11 ){ + nNeeded = 0; + }else if( szPayload<=0xff ){ + nNeeded = 1; + }else if( szPayload<=0xffff ){ + nNeeded = 2; + }else{ + nNeeded = 4; + } + delta = nNeeded - nExtra; + if( delta ){ + u32 newSize = pParse->nBlob + delta; + if( delta>0 ){ + if( newSize>pParse->nBlobAlloc && jsonBlobExpand(pParse, newSize) ){ + return 0; /* OOM error. Error state recorded in pParse->oom. */ + } + a = &pParse->aBlob[i]; + memmove(&a[1+delta], &a[1], pParse->nBlob - (i+1)); + }else{ + memmove(&a[1], &a[1-delta], pParse->nBlob - (i+1-delta)); + } + pParse->nBlob = newSize; + } + if( nNeeded==0 ){ + a[0] = (a[0] & 0x0f) | (szPayload<<4); + }else if( nNeeded==1 ){ + a[0] = (a[0] & 0x0f) | 0xc0; + a[1] = szPayload & 0xff; + }else if( nNeeded==2 ){ + a[0] = (a[0] & 0x0f) | 0xd0; + a[1] = (szPayload >> 8) & 0xff; + a[2] = szPayload & 0xff; + }else{ + a[0] = (a[0] & 0x0f) | 0xe0; + a[1] = (szPayload >> 24) & 0xff; + a[2] = (szPayload >> 16) & 0xff; + a[3] = (szPayload >> 8) & 0xff; + a[4] = szPayload & 0xff; + } + return delta; +} + +/* +** If z[0] is 'u' and is followed by exactly 4 hexadecimal character, +** then set *pOp to JSONB_TEXTJ and return true. If not, do not make +** any changes to *pOp and return false. +*/ +static int jsonIs4HexB(const char *z, int *pOp){ + if( z[0]!='u' ) return 0; + if( !sqlite3Isxdigit(z[1]) ) return 0; + if( !sqlite3Isxdigit(z[2]) ) return 0; + if( !sqlite3Isxdigit(z[3]) ) return 0; + if( !sqlite3Isxdigit(z[4]) ) return 0; + *pOp = JSONB_TEXTJ; + return 1; +} + +/* +** Translate a single element of JSON text at pParse->zJson[i] into +** its equivalent binary JSONB representation. Append the translation into +** pParse->aBlob[] beginning at pParse->nBlob. The size of +** pParse->aBlob[] is increased as necessary. +** +** Return the index of the first character past the end of the element parsed, +** or one of the following special result codes: ** ** 0 End of input -** -1 Syntax error -** -2 '}' seen -** -3 ']' seen -** -4 ',' seen -** -5 ':' seen +** -1 Syntax error or OOM +** -2 '}' seen \ +** -3 ']' seen \___ For these returns, pParse->iErr is set to +** -4 ',' seen / the index in zJson[] of the seen character +** -5 ':' seen / */ -static int jsonParseValue(JsonParse *pParse, u32 i){ +static int jsonXlateTextToBlob(JsonParse *pParse, u32 i){ char c; u32 j; - int iThis; + u32 iThis, iStart; int x; - JsonNode *pNode; + u8 t; const char *z = pParse->zJson; json_parse_restart: switch( (u8)z[i] ){ case '{': { /* Parse object */ - iThis = jsonParseAddNode(pParse, JSON_OBJECT, 0, 0); - if( iThis<0 ) return -1; + iThis = pParse->nBlob; + jsonBlobAppendNode(pParse, JSONB_OBJECT, pParse->nJson-i, 0); if( ++pParse->iDepth > JSON_MAX_DEPTH ){ pParse->iErr = i; return -1; } + iStart = pParse->nBlob; for(j=i+1;;j++){ - u32 nNode = pParse->nNode; - x = jsonParseValue(pParse, j); + u32 iBlob = pParse->nBlob; + x = jsonXlateTextToBlob(pParse, j); if( x<=0 ){ + int op; if( x==(-2) ){ j = pParse->iErr; - if( pParse->nNode!=(u32)iThis+1 ) pParse->hasNonstd = 1; + if( pParse->nBlob!=(u32)iStart ) pParse->hasNonstd = 1; break; } j += json5Whitespace(&z[j]); - if( sqlite3JsonId1(z[j]) - || (z[j]=='\\' && z[j+1]=='u' && jsonIs4Hex(&z[j+2])) + op = JSONB_TEXT; + if( sqlite3JsonId1(z[j]) + || (z[j]=='\\' && jsonIs4HexB(&z[j+1], &op)) ){ int k = j+1; while( (sqlite3JsonId2(z[k]) && json5Whitespace(&z[k])==0) - || (z[k]=='\\' && z[k+1]=='u' && jsonIs4Hex(&z[k+2])) + || (z[k]=='\\' && jsonIs4HexB(&z[k+1], &op)) ){ k++; } - jsonParseAddNode(pParse, JSON_STRING | (JNODE_RAW<<8), k-j, &z[j]); + assert( iBlob==pParse->nBlob ); + jsonBlobAppendNode(pParse, op, k-j, &z[j]); pParse->hasNonstd = 1; x = k; }else{ @@ -1384,24 +1293,24 @@ json_parse_restart: } } if( pParse->oom ) return -1; - pNode = &pParse->aNode[nNode]; - if( pNode->eType!=JSON_STRING ){ + t = pParse->aBlob[iBlob] & 0x0f; + if( tJSONB_TEXTRAW ){ pParse->iErr = j; return -1; } - pNode->jnFlags |= JNODE_LABEL; j = x; if( z[j]==':' ){ j++; }else{ - if( fast_isspace(z[j]) ){ - do{ j++; }while( fast_isspace(z[j]) ); + if( jsonIsspace(z[j]) ){ + /* strspn() is not helpful here */ + do{ j++; }while( jsonIsspace(z[j]) ); if( z[j]==':' ){ j++; goto parse_object_value; } } - x = jsonParseValue(pParse, j); + x = jsonXlateTextToBlob(pParse, j); if( x!=(-5) ){ if( x!=(-1) ) pParse->iErr = j; return -1; @@ -1409,7 +1318,7 @@ json_parse_restart: j = pParse->iErr+1; } parse_object_value: - x = jsonParseValue(pParse, j); + x = jsonXlateTextToBlob(pParse, j); if( x<=0 ){ if( x!=(-1) ) pParse->iErr = j; return -1; @@ -1420,15 +1329,15 @@ json_parse_restart: }else if( z[j]=='}' ){ break; }else{ - if( fast_isspace(z[j]) ){ - do{ j++; }while( fast_isspace(z[j]) ); + if( jsonIsspace(z[j]) ){ + j += 1 + strspn(&z[j+1], jsonSpaces); if( z[j]==',' ){ continue; }else if( z[j]=='}' ){ break; } } - x = jsonParseValue(pParse, j); + x = jsonXlateTextToBlob(pParse, j); if( x==(-4) ){ j = pParse->iErr; continue; @@ -1441,25 +1350,26 @@ json_parse_restart: pParse->iErr = j; return -1; } - pParse->aNode[iThis].n = pParse->nNode - (u32)iThis - 1; + jsonBlobChangePayloadSize(pParse, iThis, pParse->nBlob - iStart); pParse->iDepth--; return j+1; } case '[': { /* Parse array */ - iThis = jsonParseAddNode(pParse, JSON_ARRAY, 0, 0); - if( iThis<0 ) return -1; + iThis = pParse->nBlob; + jsonBlobAppendNode(pParse, JSONB_ARRAY, pParse->nJson - i, 0); + iStart = pParse->nBlob; + if( pParse->oom ) return -1; if( ++pParse->iDepth > JSON_MAX_DEPTH ){ pParse->iErr = i; return -1; } - memset(&pParse->aNode[iThis].u, 0, sizeof(pParse->aNode[iThis].u)); for(j=i+1;;j++){ - x = jsonParseValue(pParse, j); + x = jsonXlateTextToBlob(pParse, j); if( x<=0 ){ if( x==(-3) ){ j = pParse->iErr; - if( pParse->nNode!=(u32)iThis+1 ) pParse->hasNonstd = 1; + if( pParse->nBlob!=iStart ) pParse->hasNonstd = 1; break; } if( x!=(-1) ) pParse->iErr = j; @@ -1471,15 +1381,15 @@ json_parse_restart: }else if( z[j]==']' ){ break; }else{ - if( fast_isspace(z[j]) ){ - do{ j++; }while( fast_isspace(z[j]) ); + if( jsonIsspace(z[j]) ){ + j += 1 + strspn(&z[j+1], jsonSpaces); if( z[j]==',' ){ continue; }else if( z[j]==']' ){ break; } } - x = jsonParseValue(pParse, j); + x = jsonXlateTextToBlob(pParse, j); if( x==(-4) ){ j = pParse->iErr; continue; @@ -1492,23 +1402,33 @@ json_parse_restart: pParse->iErr = j; return -1; } - pParse->aNode[iThis].n = pParse->nNode - (u32)iThis - 1; + jsonBlobChangePayloadSize(pParse, iThis, pParse->nBlob - iStart); pParse->iDepth--; return j+1; } case '\'': { - u8 jnFlags; + u8 opcode; char cDelim; pParse->hasNonstd = 1; - jnFlags = JNODE_JSON5; + opcode = JSONB_TEXT; goto parse_string; case '"': /* Parse string */ - jnFlags = 0; + opcode = JSONB_TEXT; parse_string: cDelim = z[i]; - for(j=i+1; 1; j++){ - if( jsonIsOk[(unsigned char)z[j]] ) continue; + j = i+1; + while( 1 /*exit-by-break*/ ){ + if( jsonIsOk[(u8)z[j]] ){ + if( !jsonIsOk[(u8)z[j+1]] ){ + j += 1; + }else if( !jsonIsOk[(u8)z[j+2]] ){ + j += 2; + }else{ + j += 3; + continue; + } + } c = z[j]; if( c==cDelim ){ break; @@ -1517,16 +1437,16 @@ json_parse_restart: if( c=='"' || c=='\\' || c=='/' || c=='b' || c=='f' || c=='n' || c=='r' || c=='t' || (c=='u' && jsonIs4Hex(&z[j+1])) ){ - jnFlags |= JNODE_ESCAPE; + if( opcode==JSONB_TEXT ) opcode = JSONB_TEXTJ; }else if( c=='\'' || c=='0' || c=='v' || c=='\n' || (0xe2==(u8)c && 0x80==(u8)z[j+1] && (0xa8==(u8)z[j+2] || 0xa9==(u8)z[j+2])) || (c=='x' && jsonIs2Hex(&z[j+1])) ){ - jnFlags |= (JNODE_ESCAPE|JNODE_JSON5); + opcode = JSONB_TEXT5; pParse->hasNonstd = 1; }else if( c=='\r' ){ if( z[j+1]=='\n' ) j++; - jnFlags |= (JNODE_ESCAPE|JNODE_JSON5); + opcode = JSONB_TEXT5; pParse->hasNonstd = 1; }else{ pParse->iErr = j; @@ -1537,13 +1457,14 @@ json_parse_restart: pParse->iErr = j; return -1; } + j++; } - jsonParseAddNode(pParse, JSON_STRING | (jnFlags<<8), j+1-i, &z[i]); + jsonBlobAppendNode(pParse, opcode, j-1-i, &z[i+1]); return j+1; } case 't': { if( strncmp(z+i,"true",4)==0 && !sqlite3Isalnum(z[i+4]) ){ - jsonParseAddNode(pParse, JSON_TRUE, 0, 0); + jsonBlobAppendOneByte(pParse, JSONB_TRUE); return i+4; } pParse->iErr = i; @@ -1551,23 +1472,22 @@ json_parse_restart: } case 'f': { if( strncmp(z+i,"false",5)==0 && !sqlite3Isalnum(z[i+5]) ){ - jsonParseAddNode(pParse, JSON_FALSE, 0, 0); + jsonBlobAppendOneByte(pParse, JSONB_FALSE); return i+5; } pParse->iErr = i; return -1; } case '+': { - u8 seenDP, seenE, jnFlags; + u8 seenE; pParse->hasNonstd = 1; - jnFlags = JNODE_JSON5; + t = 0x00; /* Bit 0x01: JSON5. Bit 0x02: FLOAT */ goto parse_number; case '.': if( sqlite3Isdigit(z[i+1]) ){ pParse->hasNonstd = 1; - jnFlags = JNODE_JSON5; + t = 0x03; /* Bit 0x01: JSON5. Bit 0x02: FLOAT */ seenE = 0; - seenDP = JSON_REAL; goto parse_number_2; } pParse->iErr = i; @@ -1584,9 +1504,8 @@ json_parse_restart: case '8': case '9': /* Parse number */ - jnFlags = 0; + t = 0x00; /* Bit 0x01: JSON5. Bit 0x02: FLOAT */ parse_number: - seenDP = JSON_INT; seenE = 0; assert( '-' < '0' ); assert( '+' < '0' ); @@ -1596,9 +1515,9 @@ json_parse_restart: if( c<='0' ){ if( c=='0' ){ if( (z[i+1]=='x' || z[i+1]=='X') && sqlite3Isxdigit(z[i+2]) ){ - assert( seenDP==JSON_INT ); + assert( t==0x00 ); pParse->hasNonstd = 1; - jnFlags |= JNODE_JSON5; + t = 0x01; for(j=i+3; sqlite3Isxdigit(z[j]); j++){} goto parse_number_finish; }else if( sqlite3Isdigit(z[i+1]) ){ @@ -1615,15 +1534,15 @@ json_parse_restart: ){ pParse->hasNonstd = 1; if( z[i]=='-' ){ - jsonParseAddNode(pParse, JSON_REAL, 8, "-9.0e999"); + jsonBlobAppendNode(pParse, JSONB_FLOAT, 6, "-9e999"); }else{ - jsonParseAddNode(pParse, JSON_REAL, 7, "9.0e999"); + jsonBlobAppendNode(pParse, JSONB_FLOAT, 5, "9e999"); } return i + (sqlite3StrNICmp(&z[i+4],"inity",5)==0 ? 9 : 4); } if( z[i+1]=='.' ){ pParse->hasNonstd = 1; - jnFlags |= JNODE_JSON5; + t |= 0x01; goto parse_number_2; } pParse->iErr = i; @@ -1635,30 +1554,31 @@ json_parse_restart: return -1; }else if( (z[i+2]=='x' || z[i+2]=='X') && sqlite3Isxdigit(z[i+3]) ){ pParse->hasNonstd = 1; - jnFlags |= JNODE_JSON5; + t |= 0x01; for(j=i+4; sqlite3Isxdigit(z[j]); j++){} goto parse_number_finish; } } } } + parse_number_2: for(j=i+1;; j++){ c = z[j]; if( sqlite3Isdigit(c) ) continue; if( c=='.' ){ - if( seenDP==JSON_REAL ){ + if( (t & 0x02)!=0 ){ pParse->iErr = j; return -1; } - seenDP = JSON_REAL; + t |= 0x02; continue; } if( c=='e' || c=='E' ){ if( z[j-1]<'0' ){ if( ALWAYS(z[j-1]=='.') && ALWAYS(j-2>=i) && sqlite3Isdigit(z[j-2]) ){ pParse->hasNonstd = 1; - jnFlags |= JNODE_JSON5; + t |= 0x01; }else{ pParse->iErr = j; return -1; @@ -1668,7 +1588,7 @@ json_parse_restart: pParse->iErr = j; return -1; } - seenDP = JSON_REAL; + t |= 0x02; seenE = 1; c = z[j+1]; if( c=='+' || c=='-' ){ @@ -1686,14 +1606,18 @@ json_parse_restart: if( z[j-1]<'0' ){ if( ALWAYS(z[j-1]=='.') && ALWAYS(j-2>=i) && sqlite3Isdigit(z[j-2]) ){ pParse->hasNonstd = 1; - jnFlags |= JNODE_JSON5; + t |= 0x01; }else{ pParse->iErr = j; return -1; } } parse_number_finish: - jsonParseAddNode(pParse, seenDP | (jnFlags<<8), j - i, &z[i]); + assert( JSONB_INT+0x01==JSONB_INT5 ); + assert( JSONB_FLOAT+0x01==JSONB_FLOAT5 ); + assert( JSONB_INT+0x02==JSONB_FLOAT ); + if( z[i]=='+' ) i++; + jsonBlobAppendNode(pParse, JSONB_INT+t, j-i, &z[i]); return j; } case '}': { @@ -1719,9 +1643,7 @@ json_parse_restart: case 0x0a: case 0x0d: case 0x20: { - do{ - i++; - }while( fast_isspace(z[i]) ); + i += 1 + strspn(&z[i+1], jsonSpaces); goto json_parse_restart; } case 0x0b: @@ -1743,7 +1665,7 @@ json_parse_restart: } case 'n': { if( strncmp(z+i,"null",4)==0 && !sqlite3Isalnum(z[i+4]) ){ - jsonParseAddNode(pParse, JSON_NULL, 0, 0); + jsonBlobAppendOneByte(pParse, JSONB_NULL); return i+4; } /* fall-through into the default case that checks for NaN */ @@ -1759,8 +1681,11 @@ json_parse_restart: continue; } if( sqlite3Isalnum(z[i+nn]) ) continue; - jsonParseAddNode(pParse, aNanInfName[k].eType, - aNanInfName[k].nRepl, aNanInfName[k].zRepl); + if( aNanInfName[k].eType==JSONB_FLOAT ){ + jsonBlobAppendNode(pParse, JSONB_FLOAT, 5, "9e999"); + }else{ + jsonBlobAppendOneByte(pParse, JSONB_NULL); + } pParse->hasNonstd = 1; return i + nn; } @@ -1770,6 +1695,7 @@ json_parse_restart: } /* End switch(z[i]) */ } + /* ** Parse a complete JSON string. Return 0 on success or non-zero if there ** are any errors. If an error occurs, free all memory held by pParse, @@ -1778,20 +1704,21 @@ json_parse_restart: ** pParse must be initialized to an empty parse object prior to calling ** this routine. */ -static int jsonParse( +static int jsonConvertTextToBlob( JsonParse *pParse, /* Initialize and fill this JsonParse object */ sqlite3_context *pCtx /* Report errors here */ ){ int i; const char *zJson = pParse->zJson; - i = jsonParseValue(pParse, 0); + i = jsonXlateTextToBlob(pParse, 0); if( pParse->oom ) i = -1; if( i>0 ){ assert( pParse->iDepth==0 ); - while( fast_isspace(zJson[i]) ) i++; + while( jsonIsspace(zJson[i]) ) i++; if( zJson[i] ){ i += json5Whitespace(&zJson[i]); if( zJson[i] ){ + if( pCtx ) sqlite3_result_error(pCtx, "malformed JSON", -1); jsonParseReset(pParse); return 1; } @@ -1812,248 +1739,688 @@ static int jsonParse( return 0; } - -/* Mark node i of pParse as being a child of iParent. Call recursively -** to fill in all the descendants of node i. +/* +** The input string pStr is a well-formed JSON text string. Convert +** this into the JSONB format and make it the return value of the +** SQL function. */ -static void jsonParseFillInParentage(JsonParse *pParse, u32 i, u32 iParent){ - JsonNode *pNode = &pParse->aNode[i]; - u32 j; - pParse->aUp[i] = iParent; - switch( pNode->eType ){ - case JSON_ARRAY: { - for(j=1; j<=pNode->n; j += jsonNodeSize(pNode+j)){ - jsonParseFillInParentage(pParse, i+j, i); - } - break; - } - case JSON_OBJECT: { - for(j=1; j<=pNode->n; j += jsonNodeSize(pNode+j+1)+1){ - pParse->aUp[i+j] = i; - jsonParseFillInParentage(pParse, i+j+1, i); - } - break; - } - default: { - break; - } +static void jsonReturnStringAsBlob(JsonString *pStr){ + JsonParse px; + memset(&px, 0, sizeof(px)); + jsonStringTerminate(pStr); + px.zJson = pStr->zBuf; + px.nJson = pStr->nUsed; + (void)jsonXlateTextToBlob(&px, 0); + if( px.oom ){ + sqlite3_free(px.aBlob); + sqlite3_result_error_nomem(pStr->pCtx); + }else{ + assert( px.nBlobAlloc>0 ); + assert( !px.bReadOnly ); + sqlite3_result_blob(pStr->pCtx, px.aBlob, px.nBlob, sqlite3_free); } } -/* -** Compute the parentage of all nodes in a completed parse. +/* The byte at index i is a node type-code. This routine +** determines the payload size for that node and writes that +** payload size in to *pSz. It returns the offset from i to the +** beginning of the payload. Return 0 on error. */ -static int jsonParseFindParents(JsonParse *pParse){ - u32 *aUp; - assert( pParse->aUp==0 ); - aUp = pParse->aUp = sqlite3_malloc64( sizeof(u32)*pParse->nNode ); - if( aUp==0 ){ - pParse->oom = 1; - return SQLITE_NOMEM; - } - jsonParseFillInParentage(pParse, 0, 0); - return SQLITE_OK; -} - -/* -** Magic number used for the JSON parse cache in sqlite3_get_auxdata() -*/ -#define JSON_CACHE_ID (-429938) /* First cache entry */ -#define JSON_CACHE_SZ 4 /* Max number of cache entries */ - -/* -** Obtain a complete parse of the JSON found in the pJson argument -** -** Use the sqlite3_get_auxdata() cache to find a preexisting parse -** if it is available. If the cache is not available or if it -** is no longer valid, parse the JSON again and return the new parse. -** Also register the new parse so that it will be available for -** future sqlite3_get_auxdata() calls. -** -** If an error occurs and pErrCtx!=0 then report the error on pErrCtx -** and return NULL. -** -** The returned pointer (if it is not NULL) is owned by the cache in -** most cases, not the caller. The caller does NOT need to invoke -** jsonParseFree(), in most cases. -** -** Except, if an error occurs and pErrCtx==0 then return the JsonParse -** object with JsonParse.nErr non-zero and the caller will own the JsonParse -** object. In that case, it will be the responsibility of the caller to -** invoke jsonParseFree(). To summarize: -** -** pErrCtx!=0 || p->nErr==0 ==> Return value p is owned by the -** cache. Call does not need to -** free it. -** -** pErrCtx==0 && p->nErr!=0 ==> Return value is owned by the caller -** and so the caller must free it. -*/ -static JsonParse *jsonParseCached( - sqlite3_context *pCtx, /* Context to use for cache search */ - sqlite3_value *pJson, /* Function param containing JSON text */ - sqlite3_context *pErrCtx, /* Write parse errors here if not NULL */ - int bUnedited /* No prior edits allowed */ -){ - char *zJson = (char*)sqlite3_value_text(pJson); - int nJson = sqlite3_value_bytes(pJson); - JsonParse *p; - JsonParse *pMatch = 0; - int iKey; - int iMinKey = 0; - u32 iMinHold = 0xffffffff; - u32 iMaxHold = 0; - int bJsonRCStr; - - if( zJson==0 ) return 0; - for(iKey=0; iKeynJson==nJson - && (p->hasMod==0 || bUnedited==0) - && (p->zJson==zJson || memcmp(p->zJson,zJson,nJson)==0) - ){ - p->nErr = 0; - p->useMod = 0; - pMatch = p; - }else - if( pMatch==0 - && p->zAlt!=0 - && bUnedited==0 - && p->nAlt==nJson - && memcmp(p->zAlt, zJson, nJson)==0 - ){ - p->nErr = 0; - p->useMod = 1; - pMatch = p; - }else if( p->iHoldiHold; - iMinKey = iKey; - } - if( p->iHold>iMaxHold ){ - iMaxHold = p->iHold; - } - } - if( pMatch ){ - /* The input JSON text was found in the cache. Use the preexisting - ** parse of this JSON */ - pMatch->nErr = 0; - pMatch->iHold = iMaxHold+1; - assert( pMatch->nJPRef>0 ); /* pMatch is owned by the cache */ - return pMatch; - } - - /* The input JSON was not found anywhere in the cache. We will need - ** to parse it ourselves and generate a new JsonParse object. - */ - bJsonRCStr = sqlite3ValueIsOfClass(pJson,sqlite3RCStrUnref); - p = sqlite3_malloc64( sizeof(*p) + (bJsonRCStr ? 0 : nJson+1) ); - if( p==0 ){ - sqlite3_result_error_nomem(pCtx); +static u32 jsonbPayloadSize(const JsonParse *pParse, u32 i, u32 *pSz){ + u8 x; + u32 sz; + u32 n; + if( NEVER(i>pParse->nBlob) ){ + *pSz = 0; return 0; } - memset(p, 0, sizeof(*p)); - if( bJsonRCStr ){ - p->zJson = sqlite3RCStrRef(zJson); - p->bJsonIsRCStr = 1; - }else{ - p->zJson = (char*)&p[1]; - memcpy(p->zJson, zJson, nJson+1); - } - p->nJPRef = 1; - if( jsonParse(p, pErrCtx) ){ - if( pErrCtx==0 ){ - p->nErr = 1; - assert( p->nJPRef==1 ); /* Caller will own the new JsonParse object p */ - return p; - } - jsonParseFree(p); - return 0; - } - p->nJson = nJson; - p->iHold = iMaxHold+1; - /* Transfer ownership of the new JsonParse to the cache */ - sqlite3_set_auxdata(pCtx, JSON_CACHE_ID+iMinKey, p, - (void(*)(void*))jsonParseFree); - return (JsonParse*)sqlite3_get_auxdata(pCtx, JSON_CACHE_ID+iMinKey); -} - -/* -** Compare the OBJECT label at pNode against zKey,nKey. Return true on -** a match. -*/ -static int jsonLabelCompare(const JsonNode *pNode, const char *zKey, u32 nKey){ - assert( pNode->eU==1 ); - if( pNode->jnFlags & JNODE_RAW ){ - if( pNode->n!=nKey ) return 0; - return strncmp(pNode->u.zJContent, zKey, nKey)==0; - }else{ - if( pNode->n!=nKey+2 ) return 0; - return strncmp(pNode->u.zJContent+1, zKey, nKey)==0; - } -} -static int jsonSameLabel(const JsonNode *p1, const JsonNode *p2){ - if( p1->jnFlags & JNODE_RAW ){ - return jsonLabelCompare(p2, p1->u.zJContent, p1->n); - }else if( p2->jnFlags & JNODE_RAW ){ - return jsonLabelCompare(p1, p2->u.zJContent, p2->n); - }else{ - return p1->n==p2->n && strncmp(p1->u.zJContent,p2->u.zJContent,p1->n)==0; - } -} - -/* forward declaration */ -static JsonNode *jsonLookupAppend(JsonParse*,const char*,int*,const char**); - -/* -** Search along zPath to find the node specified. Return a pointer -** to that node, or NULL if zPath is malformed or if there is no such -** node. -** -** If pApnd!=0, then try to append new nodes to complete zPath if it is -** possible to do so and if no existing node corresponds to zPath. If -** new nodes are appended *pApnd is set to 1. -*/ -static JsonNode *jsonLookupStep( - JsonParse *pParse, /* The JSON to search */ - u32 iRoot, /* Begin the search at this node */ - const char *zPath, /* The path to search */ - int *pApnd, /* Append nodes to complete path if not NULL */ - const char **pzErr /* Make *pzErr point to any syntax error in zPath */ -){ - u32 i, j, nKey; - const char *zKey; - JsonNode *pRoot; - if( pParse->oom ) return 0; - pRoot = &pParse->aNode[iRoot]; - if( pRoot->jnFlags & (JNODE_REPLACE|JNODE_REMOVE) && pParse->useMod ){ - while( (pRoot->jnFlags & JNODE_REPLACE)!=0 ){ - u32 idx = (u32)(pRoot - pParse->aNode); - i = pParse->iSubst; - while( 1 /*exit-by-break*/ ){ - assert( inNode ); - assert( pParse->aNode[i].eType==JSON_SUBST ); - assert( pParse->aNode[i].eU==4 ); - assert( pParse->aNode[i].u.iPrevaNode[i].n==idx ){ - pRoot = &pParse->aNode[i+1]; - iRoot = i+1; - break; - } - i = pParse->aNode[i].u.iPrev; - } - } - if( pRoot->jnFlags & JNODE_REMOVE ){ + x = pParse->aBlob[i]>>4; + if( x<=11 ){ + sz = x; + n = 1; + }else if( x==12 ){ + if( i+1>=pParse->nBlob ){ + *pSz = 0; return 0; } + sz = pParse->aBlob[i+1]; + n = 2; + }else if( x==13 ){ + if( i+2>=pParse->nBlob ){ + *pSz = 0; + return 0; + } + sz = (pParse->aBlob[i+1]<<8) + pParse->aBlob[i+2]; + n = 3; + }else if( x==14 ){ + if( i+4>=pParse->nBlob ){ + *pSz = 0; + return 0; + } + sz = (pParse->aBlob[i+1]<<24) + (pParse->aBlob[i+2]<<16) + + (pParse->aBlob[i+3]<<8) + pParse->aBlob[i+4]; + n = 5; + }else{ + if( i+8>=pParse->nBlob + || pParse->aBlob[i+1]!=0 + || pParse->aBlob[i+2]!=0 + || pParse->aBlob[i+3]!=0 + || pParse->aBlob[i+4]!=0 + ){ + *pSz = 0; + return 0; + } + sz = (pParse->aBlob[i+5]<<24) + (pParse->aBlob[i+6]<<16) + + (pParse->aBlob[i+7]<<8) + pParse->aBlob[i+8]; + n = 9; + } + if( i+sz+n > pParse->nBlob + && i+sz+n > pParse->nBlob-pParse->delta + ){ + sz = 0; + n = 0; + } + *pSz = sz; + return n; +} + + +/* +** Translate the binary JSONB representation of JSON beginning at +** pParse->aBlob[i] into a JSON text string. Append the JSON +** text onto the end of pOut. Return the index in pParse->aBlob[] +** of the first byte past the end of the element that is translated. +** +** If an error is detected in the BLOB input, the pOut->eErr flag +** might get set to JSTRING_MALFORMED. But not all BLOB input errors +** are detected. So a malformed JSONB input might either result +** in an error, or in incorrect JSON. +** +** The pOut->eErr JSTRING_OOM flag is set on a OOM. +*/ +static u32 jsonXlateBlobToText( + const JsonParse *pParse, /* the complete parse of the JSON */ + u32 i, /* Start rendering at this index */ + JsonString *pOut /* Write JSON here */ +){ + u32 sz, n, j, iEnd; + + n = jsonbPayloadSize(pParse, i, &sz); + if( n==0 ){ + pOut->eErr |= JSTRING_MALFORMED; + return pParse->nBlob+1; + } + switch( pParse->aBlob[i] & 0x0f ){ + case JSONB_NULL: { + jsonAppendRawNZ(pOut, "null", 4); + return i+1; + } + case JSONB_TRUE: { + jsonAppendRawNZ(pOut, "true", 4); + return i+1; + } + case JSONB_FALSE: { + jsonAppendRawNZ(pOut, "false", 5); + return i+1; + } + case JSONB_INT: + case JSONB_FLOAT: { + jsonAppendRaw(pOut, (const char*)&pParse->aBlob[i+n], sz); + break; + } + case JSONB_INT5: { /* Integer literal in hexadecimal notation */ + u32 k = 2; + sqlite3_uint64 u = 0; + const char *zIn = (const char*)&pParse->aBlob[i+n]; + int bOverflow = 0; + if( zIn[0]=='-' ){ + jsonAppendChar(pOut, '-'); + k++; + }else if( zIn[0]=='+' ){ + k++; + } + for(; keErr |= JSTRING_MALFORMED; + break; + }else if( (u>>60)!=0 ){ + bOverflow = 1; + }else{ + u = u*16 + sqlite3HexToInt(zIn[k]); + } + } + jsonPrintf(100,pOut,bOverflow?"9.0e999":"%llu", u); + break; + } + case JSONB_FLOAT5: { /* Float literal missing digits beside "." */ + u32 k = 0; + const char *zIn = (const char*)&pParse->aBlob[i+n]; + if( zIn[0]=='-' ){ + jsonAppendChar(pOut, '-'); + k++; + } + if( zIn[k]=='.' ){ + jsonAppendChar(pOut, '0'); + } + for(; kaBlob[i+n], sz); + jsonAppendChar(pOut, '"'); + break; + } + case JSONB_TEXT: + case JSONB_TEXT5: { + const char *zIn; + u32 k; + u32 sz2 = sz; + zIn = (const char*)&pParse->aBlob[i+n]; + jsonAppendChar(pOut, '"'); + while( sz2>0 ){ + for(k=0; k0 ){ + jsonAppendRawNZ(pOut, zIn, k); + if( k>=sz2 ){ + break; + } + zIn += k; + sz2 -= k; + } + if( zIn[0]=='"' ){ + jsonAppendRawNZ(pOut, "\\\"", 2); + zIn++; + sz2--; + continue; + } + assert( zIn[0]=='\\' ); + assert( sz2>=1 ); + if( sz2<2 ){ + pOut->eErr |= JSTRING_MALFORMED; + break; + } + switch( (u8)zIn[1] ){ + case '\'': + jsonAppendChar(pOut, '\''); + break; + case 'v': + jsonAppendRawNZ(pOut, "\\u0009", 6); + break; + case 'x': + if( sz2<4 ){ + pOut->eErr |= JSTRING_MALFORMED; + sz2 = 2; + break; + } + jsonAppendRawNZ(pOut, "\\u00", 4); + jsonAppendRawNZ(pOut, &zIn[2], 2); + zIn += 2; + sz2 -= 2; + break; + case '0': + jsonAppendRawNZ(pOut, "\\u0000", 6); + break; + case '\r': + if( sz2>2 && zIn[2]=='\n' ){ + zIn++; + sz2--; + } + break; + case '\n': + break; + case 0xe2: + /* '\' followed by either U+2028 or U+2029 is ignored as + ** whitespace. Not that in UTF8, U+2028 is 0xe2 0x80 0x29. + ** U+2029 is the same except for the last byte */ + if( sz2<4 + || 0x80!=(u8)zIn[2] + || (0xa8!=(u8)zIn[3] && 0xa9!=(u8)zIn[3]) + ){ + pOut->eErr |= JSTRING_MALFORMED; + sz2 = 2; + break; + } + zIn += 2; + sz2 -= 2; + break; + default: + jsonAppendRawNZ(pOut, zIn, 2); + break; + } + assert( sz2>=2 ); + zIn += 2; + sz2 -= 2; + } + jsonAppendChar(pOut, '"'); + break; + } + case JSONB_TEXTRAW: { + jsonAppendString(pOut, (const char*)&pParse->aBlob[i+n], sz); + break; + } + case JSONB_ARRAY: { + jsonAppendChar(pOut, '['); + j = i+n; + iEnd = j+sz; + while( j0 ) pOut->nUsed--; + jsonAppendChar(pOut, ']'); + break; + } + case JSONB_OBJECT: { + int x = 0; + jsonAppendChar(pOut, '{'); + j = i+n; + iEnd = j+sz; + while( jeErr |= JSTRING_MALFORMED; + if( sz>0 ) pOut->nUsed--; + jsonAppendChar(pOut, '}'); + break; + } + + default: { + pOut->eErr |= JSTRING_MALFORMED; + break; + } + } + return i+n+sz; +} + +/* Return true if the input pJson +** +** For performance reasons, this routine does not do a detailed check of the +** input BLOB to ensure that it is well-formed. Hence, false positives are +** possible. False negatives should never occur, however. +*/ +static int jsonFuncArgMightBeBinary(sqlite3_value *pJson){ + u32 sz, n; + const u8 *aBlob; + int nBlob; + JsonParse s; + if( sqlite3_value_type(pJson)!=SQLITE_BLOB ) return 0; + aBlob = sqlite3_value_blob(pJson); + nBlob = sqlite3_value_bytes(pJson); + if( nBlob<1 ) return 0; + if( NEVER(aBlob==0) || (aBlob[0] & 0x0f)>JSONB_OBJECT ) return 0; + memset(&s, 0, sizeof(s)); + s.aBlob = (u8*)aBlob; + s.nBlob = nBlob; + n = jsonbPayloadSize(&s, 0, &sz); + if( n==0 ) return 0; + if( sz+n!=(u32)nBlob ) return 0; + if( (aBlob[0] & 0x0f)<=JSONB_FALSE && sz>0 ) return 0; + return sz+n==(u32)nBlob; +} + +/* +** Given that a JSONB_ARRAY object starts at offset i, return +** the number of entries in that array. +*/ +static u32 jsonbArrayCount(JsonParse *pParse, u32 iRoot){ + u32 n, sz, i, iEnd; + u32 k = 0; + n = jsonbPayloadSize(pParse, iRoot, &sz); + iEnd = iRoot+n+sz; + for(i=iRoot+n; n>0 && idelta. +*/ +static void jsonAfterEditSizeAdjust(JsonParse *pParse, u32 iRoot){ + u32 sz = 0; + u32 nBlob; + assert( pParse->delta!=0 ); + assert( pParse->nBlobAlloc >= pParse->nBlob ); + nBlob = pParse->nBlob; + pParse->nBlob = pParse->nBlobAlloc; + (void)jsonbPayloadSize(pParse, iRoot, &sz); + pParse->nBlob = nBlob; + sz += pParse->delta; + pParse->delta += jsonBlobChangePayloadSize(pParse, iRoot, sz); +} + +/* +** Modify the JSONB blob at pParse->aBlob by removing nDel bytes of +** content beginning at iDel, and replacing them with nIns bytes of +** content given by aIns. +** +** nDel may be zero, in which case no bytes are removed. But iDel is +** still important as new bytes will be insert beginning at iDel. +** +** aIns may be zero, in which case space is created to hold nIns bytes +** beginning at iDel, but that space is uninitialized. +** +** Set pParse->oom if an OOM occurs. +*/ +static void jsonBlobEdit( + JsonParse *pParse, /* The JSONB to be modified is in pParse->aBlob */ + u32 iDel, /* First byte to be removed */ + u32 nDel, /* Number of bytes to remove */ + const u8 *aIns, /* Content to insert */ + u32 nIns /* Bytes of content to insert */ +){ + i64 d = (i64)nIns - (i64)nDel; + if( d!=0 ){ + if( pParse->nBlob + d > pParse->nBlobAlloc ){ + jsonBlobExpand(pParse, pParse->nBlob+d); + if( pParse->oom ) return; + } + memmove(&pParse->aBlob[iDel+nIns], + &pParse->aBlob[iDel+nDel], + pParse->nBlob - (iDel+nDel)); + pParse->nBlob += d; + pParse->delta += d; + } + if( nIns && aIns ) memcpy(&pParse->aBlob[iDel], aIns, nIns); +} + +/* +** Return the number of escaped newlines to be ignored. +** An escaped newline is a one of the following byte sequences: +** +** 0x5c 0x0a +** 0x5c 0x0d +** 0x5c 0x0d 0x0a +** 0x5c 0xe2 0x80 0xa8 +** 0x5c 0xe2 0x80 0xa9 +*/ +static u32 jsonBytesToBypass(const char *z, u32 n){ + u32 i = 0; + while( i+10 ); + assert( z[0]=='\\' ); + if( n<2 ){ + *piOut = 0xFFFD; + return n; + } + switch( (u8)z[1] ){ + case 'u': { + u32 v, vlo; + if( n<6 ){ + *piOut = 0xFFFD; + return n; + } + v = jsonHexToInt4(&z[2]); + if( (v & 0xfc00)==0xd800 + && n>=12 + && z[6]=='\\' + && z[7]=='u' + && ((vlo = jsonHexToInt4(&z[8]))&0xfc00)==0xdc00 + ){ + *piOut = ((v&0x3ff)<<10) + (vlo&0x3ff) + 0x10000; + return 12; + }else{ + *piOut = v; + return 6; + } + } + case 'b': { *piOut = '\b'; return 2; } + case 'f': { *piOut = '\f'; return 2; } + case 'n': { *piOut = '\n'; return 2; } + case 'r': { *piOut = '\r'; return 2; } + case 't': { *piOut = '\t'; return 2; } + case 'v': { *piOut = '\v'; return 2; } + case '0': { *piOut = 0; return 2; } + case '\'': + case '"': + case '/': + case '\\':{ *piOut = z[1]; return 2; } + case 'x': { + if( n<4 ){ + *piOut = 0xFFFD; + return n; + } + *piOut = (jsonHexToInt(z[2])<<4) | jsonHexToInt(z[3]); + return 4; + } + case 0xe2: + case '\r': + case '\n': { + u32 nSkip = jsonBytesToBypass(z, n); + if( nSkip==0 ){ + *piOut = 0xFFFD; + return n; + }else if( nSkip==n ){ + *piOut = 0; + return n; + }else if( z[nSkip]=='\\' ){ + return nSkip + jsonUnescapeOneChar(&z[nSkip], n-nSkip, piOut); + }else{ + *piOut = z[nSkip]; + return nSkip+1; + } + } + default: { + *piOut = 0xFFFD; + return 2; + } + } +} + + +/* +** Compare two object labels. Return 1 if they are equal and +** 0 if they differ. +** +** In this version, we know that one or the other or both of the +** two comparands contains an escape sequence. +*/ +static SQLITE_NOINLINE int jsonLabelCompareEscaped( + const char *zLeft, /* The left label */ + u32 nLeft, /* Size of the left label in bytes */ + int rawLeft, /* True if zLeft contains no escapes */ + const char *zRight, /* The right label */ + u32 nRight, /* Size of the right label in bytes */ + int rawRight /* True if zRight is escape-free */ +){ + u32 cLeft, cRight; + assert( rawLeft==0 || rawRight==0 ); + while( nLeft>0 && nRight>0 ){ + if( rawLeft || zLeft[0]!='\\' ){ + cLeft = ((u8*)zLeft)[0]; + zLeft++; + nLeft--; + }else{ + u32 n = jsonUnescapeOneChar(zLeft, nLeft, &cLeft); + zLeft += n; + assert( n<=nLeft ); + nLeft -= n; + } + if( rawRight || zRight[0]!='\\' ){ + cRight = ((u8*)zRight)[0]; + zRight++; + nRight--; + }else{ + u32 n = jsonUnescapeOneChar(zRight, nRight, &cRight); + zRight += n; + assert( n<=nRight ); + nRight -= n; + } + if( cLeft!=cRight ) return 0; + } + return nLeft==0 && nRight==0; +} + +/* +** Compare two object labels. Return 1 if they are equal and +** 0 if they differ. Return -1 if an OOM occurs. +*/ +static int jsonLabelCompare( + const char *zLeft, /* The left label */ + u32 nLeft, /* Size of the left label in bytes */ + int rawLeft, /* True if zLeft contains no escapes */ + const char *zRight, /* The right label */ + u32 nRight, /* Size of the right label in bytes */ + int rawRight /* True if zRight is escape-free */ +){ + if( rawLeft && rawRight ){ + /* Simpliest case: Neither label contains escapes. A simple + ** memcmp() is sufficient. */ + if( nLeft!=nRight ) return 0; + return memcmp(zLeft, zRight, nLeft)==0; + }else{ + return jsonLabelCompareEscaped(zLeft, nLeft, rawLeft, + zRight, nRight, rawRight); + } +} + +/* +** Error returns from jsonLookupStep() +*/ +#define JSON_LOOKUP_ERROR 0xffffffff +#define JSON_LOOKUP_NOTFOUND 0xfffffffe +#define JSON_LOOKUP_PATHERROR 0xfffffffd +#define JSON_LOOKUP_ISERROR(x) ((x)>=JSON_LOOKUP_PATHERROR) + +/* Forward declaration */ +static u32 jsonLookupStep(JsonParse*,u32,const char*,u32); + + +/* This helper routine for jsonLookupStep() populates pIns with +** binary data that is to be inserted into pParse. +** +** In the common case, pIns just points to pParse->aIns and pParse->nIns. +** But if the zPath of the original edit operation includes path elements +** that go deeper, additional substructure must be created. +** +** For example: +** +** json_insert('{}', '$.a.b.c', 123); +** +** The search stops at '$.a' But additional substructure must be +** created for the ".b.c" part of the patch so that the final result +** is: {"a":{"b":{"c"::123}}}. This routine populates pIns with +** the binary equivalent of {"b":{"c":123}} so that it can be inserted. +** +** The caller is responsible for resetting pIns when it has finished +** using the substructure. +*/ +static u32 jsonCreateEditSubstructure( + JsonParse *pParse, /* The original JSONB that is being edited */ + JsonParse *pIns, /* Populate this with the blob data to insert */ + const char *zTail /* Tail of the path that determins substructure */ +){ + static const u8 emptyObject[] = { JSONB_ARRAY, JSONB_OBJECT }; + int rc; + memset(pIns, 0, sizeof(*pIns)); + if( zTail[0]==0 ){ + /* No substructure. Just insert what is given in pParse. */ + pIns->aBlob = pParse->aIns; + pIns->nBlob = pParse->nIns; + rc = 0; + }else{ + /* Construct the binary substructure */ + pIns->nBlob = 1; + pIns->aBlob = (u8*)&emptyObject[zTail[0]=='.']; + pIns->eEdit = pParse->eEdit; + pIns->nIns = pParse->nIns; + pIns->aIns = pParse->aIns; + rc = jsonLookupStep(pIns, 0, zTail, 0); + pParse->oom |= pIns->oom; + } + return rc; /* Error code only */ +} + +/* +** Search along zPath to find the Json element specified. Return an +** index into pParse->aBlob[] for the start of that element's value. +** +** If the value found by this routine is the value half of label/value pair +** within an object, then set pPath->iLabel to the start of the corresponding +** label, before returning. +** +** Return one of the JSON_LOOKUP error codes if problems are seen. +** +** This routine will also modify the blob. If pParse->eEdit is one of +** JEDIT_DEL, JEDIT_REPL, JEDIT_INS, or JEDIT_SET, then changes might be +** made to the selected value. If an edit is performed, then the return +** value does not necessarily point to the select element. If an edit +** is performed, the return value is only useful for detecting error +** conditions. +*/ +static u32 jsonLookupStep( + JsonParse *pParse, /* The JSON to search */ + u32 iRoot, /* Begin the search at this element of aBlob[] */ + const char *zPath, /* The path to search */ + u32 iLabel /* Label if iRoot is a value of in an object */ +){ + u32 i, j, k, nKey, sz, n, iEnd, rc; + const char *zKey; + u8 x; + + if( zPath[0]==0 ){ + if( pParse->eEdit && jsonBlobMakeEditable(pParse, pParse->nIns) ){ + n = jsonbPayloadSize(pParse, iRoot, &sz); + sz += n; + if( pParse->eEdit==JEDIT_DEL ){ + if( iLabel>0 ){ + sz += iRoot - iLabel; + iRoot = iLabel; + } + jsonBlobEdit(pParse, iRoot, sz, 0, 0); + }else if( pParse->eEdit==JEDIT_INS ){ + /* Already exists, so json_insert() is a no-op */ + }else{ + /* json_set() or json_replace() */ + jsonBlobEdit(pParse, iRoot, sz, pParse->aIns, pParse->nIns); + } + } + pParse->iLabel = iLabel; + return iRoot; } - if( zPath[0]==0 ) return pRoot; if( zPath[0]=='.' ){ - if( pRoot->eType!=JSON_OBJECT ) return 0; + int rawKey = 1; + x = pParse->aBlob[iRoot]; zPath++; if( zPath[0]=='"' ){ zKey = zPath + 1; @@ -2062,251 +2429,635 @@ static JsonNode *jsonLookupStep( if( zPath[i] ){ i++; }else{ - *pzErr = zPath; - return 0; + return JSON_LOOKUP_PATHERROR; } testcase( nKey==0 ); + rawKey = memchr(zKey, '\\', nKey)==0; }else{ zKey = zPath; for(i=0; zPath[i] && zPath[i]!='.' && zPath[i]!='['; i++){} nKey = i; if( nKey==0 ){ - *pzErr = zPath; - return 0; + return JSON_LOOKUP_PATHERROR; } } - j = 1; - for(;;){ - while( j<=pRoot->n ){ - if( jsonLabelCompare(pRoot+j, zKey, nKey) ){ - return jsonLookupStep(pParse, iRoot+j+1, &zPath[i], pApnd, pzErr); + if( (x & 0x0f)!=JSONB_OBJECT ) return JSON_LOOKUP_NOTFOUND; + n = jsonbPayloadSize(pParse, iRoot, &sz); + j = iRoot + n; /* j is the index of a label */ + iEnd = j+sz; + while( jaBlob[j] & 0x0f; + if( xJSONB_TEXTRAW ) return JSON_LOOKUP_ERROR; + n = jsonbPayloadSize(pParse, j, &sz); + if( n==0 ) return JSON_LOOKUP_ERROR; + k = j+n; /* k is the index of the label text */ + if( k+sz>=iEnd ) return JSON_LOOKUP_ERROR; + zLabel = (const char*)&pParse->aBlob[k]; + rawLabel = x==JSONB_TEXT || x==JSONB_TEXTRAW; + if( jsonLabelCompare(zKey, nKey, rawKey, zLabel, sz, rawLabel) ){ + u32 v = k+sz; /* v is the index of the value */ + if( ((pParse->aBlob[v])&0x0f)>JSONB_OBJECT ) return JSON_LOOKUP_ERROR; + n = jsonbPayloadSize(pParse, v, &sz); + if( n==0 || v+n+sz>iEnd ) return JSON_LOOKUP_ERROR; + assert( j>0 ); + rc = jsonLookupStep(pParse, v, &zPath[i], j); + if( pParse->delta ) jsonAfterEditSizeAdjust(pParse, iRoot); + return rc; + } + j = k+sz; + if( ((pParse->aBlob[j])&0x0f)>JSONB_OBJECT ) return JSON_LOOKUP_ERROR; + n = jsonbPayloadSize(pParse, j, &sz); + if( n==0 ) return JSON_LOOKUP_ERROR; + j += n+sz; + } + if( j>iEnd ) return JSON_LOOKUP_ERROR; + if( pParse->eEdit>=JEDIT_INS ){ + u32 nIns; /* Total bytes to insert (label+value) */ + JsonParse v; /* BLOB encoding of the value to be inserted */ + JsonParse ix; /* Header of the label to be inserted */ + testcase( pParse->eEdit==JEDIT_INS ); + testcase( pParse->eEdit==JEDIT_SET ); + memset(&ix, 0, sizeof(ix)); + jsonBlobAppendNode(&ix, rawKey?JSONB_TEXTRAW:JSONB_TEXT5, nKey, 0); + pParse->oom |= ix.oom; + rc = jsonCreateEditSubstructure(pParse, &v, &zPath[i]); + if( !JSON_LOOKUP_ISERROR(rc) + && jsonBlobMakeEditable(pParse, ix.nBlob+nKey+v.nBlob) + ){ + assert( !pParse->oom ); + nIns = ix.nBlob + nKey + v.nBlob; + jsonBlobEdit(pParse, j, 0, 0, nIns); + if( !pParse->oom ){ + assert( pParse->aBlob!=0 ); /* Because pParse->oom!=0 */ + assert( ix.aBlob!=0 ); /* Because pPasre->oom!=0 */ + memcpy(&pParse->aBlob[j], ix.aBlob, ix.nBlob); + k = j + ix.nBlob; + memcpy(&pParse->aBlob[k], zKey, nKey); + k += nKey; + memcpy(&pParse->aBlob[k], v.aBlob, v.nBlob); + if( ALWAYS(pParse->delta) ) jsonAfterEditSizeAdjust(pParse, iRoot); } - j++; - j += jsonNodeSize(&pRoot[j]); } - if( (pRoot->jnFlags & JNODE_APPEND)==0 ) break; - if( pParse->useMod==0 ) break; - assert( pRoot->eU==2 ); - iRoot = pRoot->u.iAppend; - pRoot = &pParse->aNode[iRoot]; - j = 1; - } - if( pApnd ){ - u32 iStart, iLabel; - JsonNode *pNode; - assert( pParse->useMod ); - iStart = jsonParseAddNode(pParse, JSON_OBJECT, 2, 0); - iLabel = jsonParseAddNode(pParse, JSON_STRING, nKey, zKey); - zPath += i; - pNode = jsonLookupAppend(pParse, zPath, pApnd, pzErr); - if( pParse->oom ) return 0; - if( pNode ){ - pRoot = &pParse->aNode[iRoot]; - assert( pRoot->eU==0 ); - pRoot->u.iAppend = iStart; - pRoot->jnFlags |= JNODE_APPEND; - VVA( pRoot->eU = 2 ); - pParse->aNode[iLabel].jnFlags |= JNODE_RAW; - } - return pNode; + jsonParseReset(&v); + jsonParseReset(&ix); + return rc; } }else if( zPath[0]=='[' ){ - i = 0; - j = 1; - while( sqlite3Isdigit(zPath[j]) ){ - i = i*10 + zPath[j] - '0'; - j++; + x = pParse->aBlob[iRoot] & 0x0f; + if( x!=JSONB_ARRAY ) return JSON_LOOKUP_NOTFOUND; + n = jsonbPayloadSize(pParse, iRoot, &sz); + k = 0; + i = 1; + while( sqlite3Isdigit(zPath[i]) ){ + k = k*10 + zPath[i] - '0'; + i++; } - if( j<2 || zPath[j]!=']' ){ + if( i<2 || zPath[i]!=']' ){ if( zPath[1]=='#' ){ - JsonNode *pBase = pRoot; - int iBase = iRoot; - if( pRoot->eType!=JSON_ARRAY ) return 0; - for(;;){ - while( j<=pBase->n ){ - if( (pBase[j].jnFlags & JNODE_REMOVE)==0 || pParse->useMod==0 ) i++; - j += jsonNodeSize(&pBase[j]); - } - if( (pBase->jnFlags & JNODE_APPEND)==0 ) break; - if( pParse->useMod==0 ) break; - assert( pBase->eU==2 ); - iBase = pBase->u.iAppend; - pBase = &pParse->aNode[iBase]; - j = 1; - } - j = 2; + k = jsonbArrayCount(pParse, iRoot); + i = 2; if( zPath[2]=='-' && sqlite3Isdigit(zPath[3]) ){ - unsigned int x = 0; - j = 3; + unsigned int nn = 0; + i = 3; do{ - x = x*10 + zPath[j] - '0'; - j++; - }while( sqlite3Isdigit(zPath[j]) ); - if( x>i ) return 0; - i -= x; + nn = nn*10 + zPath[i] - '0'; + i++; + }while( sqlite3Isdigit(zPath[i]) ); + if( nn>k ) return JSON_LOOKUP_NOTFOUND; + k -= nn; } - if( zPath[j]!=']' ){ - *pzErr = zPath; - return 0; + if( zPath[i]!=']' ){ + return JSON_LOOKUP_PATHERROR; } }else{ - *pzErr = zPath; - return 0; + return JSON_LOOKUP_PATHERROR; } } - if( pRoot->eType!=JSON_ARRAY ) return 0; - zPath += j + 1; - j = 1; - for(;;){ - while( j<=pRoot->n - && (i>0 || ((pRoot[j].jnFlags & JNODE_REMOVE)!=0 && pParse->useMod)) + j = iRoot+n; + iEnd = j+sz; + while( jdelta ) jsonAfterEditSizeAdjust(pParse, iRoot); + return rc; + } + k--; + n = jsonbPayloadSize(pParse, j, &sz); + if( n==0 ) return JSON_LOOKUP_ERROR; + j += n+sz; + } + if( j>iEnd ) return JSON_LOOKUP_ERROR; + if( k>0 ) return JSON_LOOKUP_NOTFOUND; + if( pParse->eEdit>=JEDIT_INS ){ + JsonParse v; + testcase( pParse->eEdit==JEDIT_INS ); + testcase( pParse->eEdit==JEDIT_SET ); + rc = jsonCreateEditSubstructure(pParse, &v, &zPath[i+1]); + if( !JSON_LOOKUP_ISERROR(rc) + && jsonBlobMakeEditable(pParse, v.nBlob) ){ - if( (pRoot[j].jnFlags & JNODE_REMOVE)==0 || pParse->useMod==0 ) i--; - j += jsonNodeSize(&pRoot[j]); + assert( !pParse->oom ); + jsonBlobEdit(pParse, j, 0, v.aBlob, v.nBlob); } - if( i==0 && j<=pRoot->n ) break; - if( (pRoot->jnFlags & JNODE_APPEND)==0 ) break; - if( pParse->useMod==0 ) break; - assert( pRoot->eU==2 ); - iRoot = pRoot->u.iAppend; - pRoot = &pParse->aNode[iRoot]; - j = 1; - } - if( j<=pRoot->n ){ - return jsonLookupStep(pParse, iRoot+j, zPath, pApnd, pzErr); - } - if( i==0 && pApnd ){ - u32 iStart; - JsonNode *pNode; - assert( pParse->useMod ); - iStart = jsonParseAddNode(pParse, JSON_ARRAY, 1, 0); - pNode = jsonLookupAppend(pParse, zPath, pApnd, pzErr); - if( pParse->oom ) return 0; - if( pNode ){ - pRoot = &pParse->aNode[iRoot]; - assert( pRoot->eU==0 ); - pRoot->u.iAppend = iStart; - pRoot->jnFlags |= JNODE_APPEND; - VVA( pRoot->eU = 2 ); - } - return pNode; + jsonParseReset(&v); + if( pParse->delta ) jsonAfterEditSizeAdjust(pParse, iRoot); + return rc; } }else{ - *pzErr = zPath; + return JSON_LOOKUP_PATHERROR; } - return 0; + return JSON_LOOKUP_NOTFOUND; } /* -** Append content to pParse that will complete zPath. Return a pointer -** to the inserted node, or return NULL if the append fails. +** Convert a JSON BLOB into text and make that text the return value +** of an SQL function. */ -static JsonNode *jsonLookupAppend( - JsonParse *pParse, /* Append content to the JSON parse */ - const char *zPath, /* Description of content to append */ - int *pApnd, /* Set this flag to 1 */ - const char **pzErr /* Make this point to any syntax error */ +static void jsonReturnTextJsonFromBlob( + sqlite3_context *ctx, + const u8 *aBlob, + u32 nBlob ){ - *pApnd = 1; - if( zPath[0]==0 ){ - jsonParseAddNode(pParse, JSON_NULL, 0, 0); - return pParse->oom ? 0 : &pParse->aNode[pParse->nNode-1]; + JsonParse x; + JsonString s; + + if( NEVER(aBlob==0) ) return; + memset(&x, 0, sizeof(x)); + x.aBlob = (u8*)aBlob; + x.nBlob = nBlob; + jsonStringInit(&s, ctx); + jsonXlateBlobToText(&x, 0, &s); + jsonReturnString(&s, 0, 0); +} + + +/* +** Return the value of the BLOB node at index i. +** +** If the value is a primitive, return it as an SQL value. +** If the value is an array or object, return it as either +** JSON text or the BLOB encoding, depending on the JSON_B flag +** on the userdata. +*/ +static void jsonReturnFromBlob( + JsonParse *pParse, /* Complete JSON parse tree */ + u32 i, /* Index of the node */ + sqlite3_context *pCtx, /* Return value for this function */ + int textOnly /* return text JSON. Disregard user-data */ +){ + u32 n, sz; + int rc; + sqlite3 *db = sqlite3_context_db_handle(pCtx); + + n = jsonbPayloadSize(pParse, i, &sz); + if( n==0 ){ + sqlite3_result_error(pCtx, "malformed JSON", -1); + return; } - if( zPath[0]=='.' ){ - jsonParseAddNode(pParse, JSON_OBJECT, 0, 0); - }else if( strncmp(zPath,"[0]",3)==0 ){ - jsonParseAddNode(pParse, JSON_ARRAY, 0, 0); + switch( pParse->aBlob[i] & 0x0f ){ + case JSONB_NULL: { + sqlite3_result_null(pCtx); + break; + } + case JSONB_TRUE: { + sqlite3_result_int(pCtx, 1); + break; + } + case JSONB_FALSE: { + sqlite3_result_int(pCtx, 0); + break; + } + case JSONB_INT5: + case JSONB_INT: { + sqlite3_int64 iRes = 0; + char *z; + int bNeg = 0; + char x = (char)pParse->aBlob[i+n]; + if( x=='-' && ALWAYS(sz>0) ){ n++; sz--; bNeg = 1; } + z = sqlite3DbStrNDup(db, (const char*)&pParse->aBlob[i+n], (int)sz); + if( z==0 ) return; + rc = sqlite3DecOrHexToI64(z, &iRes); + sqlite3DbFree(db, z); + if( rc<=1 ){ + sqlite3_result_int64(pCtx, bNeg ? -iRes : iRes); + }else if( rc==3 && bNeg ){ + sqlite3_result_int64(pCtx, SMALLEST_INT64); + }else{ + if( bNeg ){ n--; sz++; } + goto to_double; + } + break; + } + case JSONB_FLOAT5: + case JSONB_FLOAT: { + double r; + char *z; + to_double: + z = sqlite3DbStrNDup(db, (const char*)&pParse->aBlob[i+n], (int)sz); + if( z==0 ) return; + sqlite3AtoF(z, &r, sqlite3Strlen30(z), SQLITE_UTF8); + sqlite3DbFree(db, z); + sqlite3_result_double(pCtx, r); + break; + } + case JSONB_TEXTRAW: + case JSONB_TEXT: { + sqlite3_result_text(pCtx, (char*)&pParse->aBlob[i+n], sz, + SQLITE_TRANSIENT); + break; + } + case JSONB_TEXT5: + case JSONB_TEXTJ: { + /* Translate JSON formatted string into raw text */ + u32 iIn, iOut; + const char *z; + char *zOut; + u32 nOut = sz; + z = (const char*)&pParse->aBlob[i+n]; + zOut = sqlite3_malloc( nOut+1 ); + if( zOut==0 ){ + sqlite3_result_error_nomem(pCtx); + break; + } + for(iIn=iOut=0; iIn>6)); + zOut[iOut++] = 0x80 | (v&0x3f); + }else if( v<0x10000 ){ + zOut[iOut++] = 0xe0 | (v>>12); + zOut[iOut++] = 0x80 | ((v>>6)&0x3f); + zOut[iOut++] = 0x80 | (v&0x3f); + }else{ + zOut[iOut++] = 0xf0 | (v>>18); + zOut[iOut++] = 0x80 | ((v>>12)&0x3f); + zOut[iOut++] = 0x80 | ((v>>6)&0x3f); + zOut[iOut++] = 0x80 | (v&0x3f); + } + iIn += szEscape - 1; + }else{ + zOut[iOut++] = c; + } + } /* end for() */ + zOut[iOut] = 0; + sqlite3_result_text(pCtx, zOut, iOut, sqlite3_free); + break; + } + case JSONB_ARRAY: + case JSONB_OBJECT: { + int flags = textOnly ? 0 : SQLITE_PTR_TO_INT(sqlite3_user_data(pCtx)); + if( flags & JSON_BLOB ){ + sqlite3_result_blob(pCtx, &pParse->aBlob[i], sz+n, SQLITE_TRANSIENT); + }else{ + jsonReturnTextJsonFromBlob(pCtx, &pParse->aBlob[i], sz+n); + } + break; + } + default: { + sqlite3_result_error(pCtx, "malformed JSON", -1); + break; + } + } +} + +/* +** pArg is a function argument that might be an SQL value or a JSON +** value. Figure out what it is and encode it as a JSONB blob. +** Return the results in pParse. +** +** pParse is uninitialized upon entry. This routine will handle the +** initialization of pParse. The result will be contained in +** pParse->aBlob and pParse->nBlob. pParse->aBlob might be dynamically +** allocated (if pParse->nBlobAlloc is greater than zero) in which case +** the caller is responsible for freeing the space allocated to pParse->aBlob +** when it has finished with it. Or pParse->aBlob might be a static string +** or a value obtained from sqlite3_value_blob(pArg). +** +** If the argument is a BLOB that is clearly not a JSONB, then this +** function might set an error message in ctx and return non-zero. +** It might also set an error message and return non-zero on an OOM error. +*/ +static int jsonFunctionArgToBlob( + sqlite3_context *ctx, + sqlite3_value *pArg, + JsonParse *pParse +){ + int eType = sqlite3_value_type(pArg); + static u8 aNull[] = { 0x00 }; + memset(pParse, 0, sizeof(pParse[0])); + switch( eType ){ + default: { + pParse->aBlob = aNull; + pParse->nBlob = 1; + return 0; + } + case SQLITE_BLOB: { + if( jsonFuncArgMightBeBinary(pArg) ){ + pParse->aBlob = (u8*)sqlite3_value_blob(pArg); + pParse->nBlob = sqlite3_value_bytes(pArg); + }else{ + sqlite3_result_error(ctx, "JSON cannot hold BLOB values", -1); + return 1; + } + break; + } + case SQLITE_TEXT: { + const char *zJson = (const char*)sqlite3_value_text(pArg); + int nJson = sqlite3_value_bytes(pArg); + if( zJson==0 ) return 1; + if( sqlite3_value_subtype(pArg)==JSON_SUBTYPE ){ + pParse->zJson = (char*)zJson; + pParse->nJson = nJson; + if( jsonConvertTextToBlob(pParse, ctx) ){ + sqlite3_result_error(ctx, "malformed JSON", -1); + sqlite3_free(pParse->aBlob); + memset(pParse, 0, sizeof(pParse[0])); + return 1; + } + }else{ + jsonBlobAppendNode(pParse, JSONB_TEXTRAW, nJson, zJson); + } + break; + } + case SQLITE_FLOAT: + case SQLITE_INTEGER: { + int n = sqlite3_value_bytes(pArg); + const char *z = (const char*)sqlite3_value_text(pArg); + int e = eType==SQLITE_INTEGER ? JSONB_INT : JSONB_FLOAT; + if( z==0 ) return 1; + jsonBlobAppendNode(pParse, e, n, z); + break; + } + } + if( pParse->oom ){ + sqlite3_result_error_nomem(ctx); + return 1; }else{ return 0; } - if( pParse->oom ) return 0; - return jsonLookupStep(pParse, pParse->nNode-1, zPath, pApnd, pzErr); } /* -** Return the text of a syntax error message on a JSON path. Space is -** obtained from sqlite3_malloc(). -*/ -static char *jsonPathSyntaxError(const char *zErr){ - return sqlite3_mprintf("JSON path error near '%q'", zErr); -} - -/* -** Do a node lookup using zPath. Return a pointer to the node on success. -** Return NULL if not found or if there is an error. +** Generate a bad path error. ** -** On an error, write an error message into pCtx and increment the -** pParse->nErr counter. -** -** If pApnd!=NULL then try to append missing nodes and set *pApnd = 1 if -** nodes are appended. +** If ctx is not NULL then push the error message into ctx and return NULL. +** If ctx is NULL, then return the text of the error message. */ -static JsonNode *jsonLookup( - JsonParse *pParse, /* The JSON to search */ - const char *zPath, /* The path to search */ - int *pApnd, /* Append nodes to complete path if not NULL */ - sqlite3_context *pCtx /* Report errors here, if not NULL */ +static char *jsonBadPathError( + sqlite3_context *ctx, /* The function call containing the error */ + const char *zPath /* The path with the problem */ ){ - const char *zErr = 0; - JsonNode *pNode = 0; - char *zMsg; - - if( zPath==0 ) return 0; - if( zPath[0]!='$' ){ - zErr = zPath; - goto lookup_err; - } - zPath++; - pNode = jsonLookupStep(pParse, 0, zPath, pApnd, &zErr); - if( zErr==0 ) return pNode; - -lookup_err: - pParse->nErr++; - assert( zErr!=0 && pCtx!=0 ); - zMsg = jsonPathSyntaxError(zErr); + char *zMsg = sqlite3_mprintf("bad JSON path: %Q", zPath); + if( ctx==0 ) return zMsg; if( zMsg ){ - sqlite3_result_error(pCtx, zMsg, -1); + sqlite3_result_error(ctx, zMsg, -1); sqlite3_free(zMsg); }else{ - sqlite3_result_error_nomem(pCtx); + sqlite3_result_error_nomem(ctx); } return 0; } - -/* -** Report the wrong number of arguments for json_insert(), json_replace() -** or json_set(). +/* argv[0] is a BLOB that seems likely to be a JSONB. Subsequent +** arguments come in parse where each pair contains a JSON path and +** content to insert or set at that patch. Do the updates +** and return the result. +** +** The specific operation is determined by eEdit, which can be one +** of JEDIT_INS, JEDIT_REPL, or JEDIT_SET. */ -static void jsonWrongNumArgs( - sqlite3_context *pCtx, - const char *zFuncName +static void jsonInsertIntoBlob( + sqlite3_context *ctx, + int argc, + sqlite3_value **argv, + int eEdit /* JEDIT_INS, JEDIT_REPL, or JEDIT_SET */ ){ - char *zMsg = sqlite3_mprintf("json_%s() needs an odd number of arguments", - zFuncName); - sqlite3_result_error(pCtx, zMsg, -1); - sqlite3_free(zMsg); + int i; + u32 rc = 0; + const char *zPath = 0; + int flgs; + JsonParse *p; + JsonParse ax; + + assert( (argc&1)==1 ); + flgs = argc==1 ? 0 : JSON_EDITABLE; + p = jsonParseFuncArg(ctx, argv[0], flgs); + if( p==0 ) return; + for(i=1; inBlob, ax.aBlob, ax.nBlob); + } + rc = 0; + }else{ + p->eEdit = eEdit; + p->nIns = ax.nBlob; + p->aIns = ax.aBlob; + p->delta = 0; + rc = jsonLookupStep(p, 0, zPath+1, 0); + } + jsonParseReset(&ax); + if( rc==JSON_LOOKUP_NOTFOUND ) continue; + if( JSON_LOOKUP_ISERROR(rc) ) goto jsonInsertIntoBlob_patherror; + } + jsonReturnParse(ctx, p); + jsonParseFree(p); + return; + +jsonInsertIntoBlob_patherror: + jsonParseFree(p); + if( rc==JSON_LOOKUP_ERROR ){ + sqlite3_result_error(ctx, "malformed JSON", -1); + }else{ + jsonBadPathError(ctx, zPath); + } + return; } /* -** Mark all NULL entries in the Object passed in as JNODE_REMOVE. +** Make a copy of a JsonParse object. The copy will be editable. */ -static void jsonRemoveAllNulls(JsonNode *pNode){ - int i, n; - assert( pNode->eType==JSON_OBJECT ); - n = pNode->n; - for(i=2; i<=n; i += jsonNodeSize(&pNode[i])+1){ - switch( pNode[i].eType ){ - case JSON_NULL: - pNode[i].jnFlags |= JNODE_REMOVE; - break; - case JSON_OBJECT: - jsonRemoveAllNulls(&pNode[i]); - break; + + +/* +** Generate a JsonParse object, containing valid JSONB in aBlob and nBlob, +** from the SQL function argument pArg. Return a pointer to the new +** JsonParse object. +** +** Ownership of the new JsonParse object is passed to the caller. The +** caller should invoke jsonParseFree() on the return value when it +** has finished using it. +** +** If any errors are detected, an appropriate error messages is set +** using sqlite3_result_error() or the equivalent and this routine +** returns NULL. This routine also returns NULL if the pArg argument +** is an SQL NULL value, but no error message is set in that case. This +** is so that SQL functions that are given NULL arguments will return +** a NULL value. +*/ +static JsonParse *jsonParseFuncArg( + sqlite3_context *ctx, + sqlite3_value *pArg, + u32 flgs +){ + int eType; /* Datatype of pArg */ + JsonParse *p = 0; /* Value to be returned */ + JsonParse *pFromCache = 0; /* Value taken from cache */ + + assert( ctx!=0 ); + eType = sqlite3_value_type(pArg); + if( eType==SQLITE_NULL ){ + return 0; + } + pFromCache = jsonCacheSearch(ctx, pArg); + if( pFromCache ){ + pFromCache->nJPRef++; + if( (flgs & JSON_EDITABLE)==0 ){ + return pFromCache; } } +rebuild_from_cache: + p = sqlite3_malloc64( sizeof(*p) ); + if( p==0 ) goto json_pfa_oom; + memset(p, 0, sizeof(*p)); + p->nJPRef = 1; + if( pFromCache!=0 ){ + u32 nBlob = pFromCache->nBlob; + p->aBlob = sqlite3_malloc64( nBlob ); + if( p->aBlob==0 ) goto json_pfa_oom; + memcpy(p->aBlob, pFromCache->aBlob, nBlob); + p->nBlobAlloc = p->nBlob = nBlob; + p->hasNonstd = pFromCache->hasNonstd; + jsonParseFree(pFromCache); + return p; + } + if( eType==SQLITE_BLOB ){ + u32 n, sz = 0; + p->aBlob = (u8*)sqlite3_value_blob(pArg); + p->nBlob = (u32)sqlite3_value_bytes(pArg); + if( p->nBlob==0 ){ + goto json_pfa_malformed; + } + if( NEVER(p->aBlob==0) ){ + goto json_pfa_oom; + } + if( (p->aBlob[0] & 0x0f)>JSONB_OBJECT ){ + goto json_pfa_malformed; + } + n = jsonbPayloadSize(p, 0, &sz); + if( n==0 + || sz+n!=p->nBlob + || ((p->aBlob[0] & 0x0f)<=JSONB_FALSE && sz>0) + ){ + goto json_pfa_malformed; + } + if( (flgs & JSON_EDITABLE)!=0 && jsonBlobMakeEditable(p, 0)==0 ){ + goto json_pfa_oom; + } + return p; + } + p->zJson = (char*)sqlite3_value_text(pArg); + p->nJson = sqlite3_value_bytes(pArg); + if( p->nJson==0 ) goto json_pfa_malformed; + if( NEVER(p->zJson==0) ) goto json_pfa_oom; + if( jsonConvertTextToBlob(p, (flgs & JSON_KEEPERROR) ? 0 : ctx) ){ + if( flgs & JSON_KEEPERROR ){ + p->nErr = 1; + return p; + }else{ + jsonParseFree(p); + return 0; + } + }else{ + int isRCStr = sqlite3ValueIsOfClass(pArg, sqlite3RCStrUnref); + int rc; + if( !isRCStr ){ + char *zNew = sqlite3RCStrNew( p->nJson ); + if( zNew==0 ) goto json_pfa_oom; + memcpy(zNew, p->zJson, p->nJson); + p->zJson = zNew; + p->zJson[p->nJson] = 0; + }else{ + sqlite3RCStrRef(p->zJson); + } + p->bJsonIsRCStr = 1; + rc = jsonCacheInsert(ctx, p); + if( rc==SQLITE_NOMEM ) goto json_pfa_oom; + if( flgs & JSON_EDITABLE ){ + pFromCache = p; + p = 0; + goto rebuild_from_cache; + } + } + return p; + +json_pfa_malformed: + if( flgs & JSON_KEEPERROR ){ + p->nErr = 1; + return p; + }else{ + jsonParseFree(p); + sqlite3_result_error(ctx, "malformed JSON", -1); + return 0; + } + +json_pfa_oom: + jsonParseFree(pFromCache); + jsonParseFree(p); + sqlite3_result_error_nomem(ctx); + return 0; } +/* +** Make the return value of a JSON function either the raw JSONB blob +** or make it JSON text, depending on whether the JSON_BLOB flag is +** set on the function. +*/ +static void jsonReturnParse( + sqlite3_context *ctx, + JsonParse *p +){ + int flgs; + if( p->oom ){ + sqlite3_result_error_nomem(ctx); + return; + } + flgs = SQLITE_PTR_TO_INT(sqlite3_user_data(ctx)); + if( flgs & JSON_BLOB ){ + if( p->nBlobAlloc>0 && !p->bReadOnly ){ + sqlite3_result_blob(ctx, p->aBlob, p->nBlob, SQLITE_DYNAMIC); + p->nBlobAlloc = 0; + }else{ + sqlite3_result_blob(ctx, p->aBlob, p->nBlob, SQLITE_TRANSIENT); + } + }else{ + JsonString s; + jsonStringInit(&s, ctx); + jsonXlateBlobToText(p, 0, &s); + jsonReturnString(&s, p, ctx); + sqlite3_result_subtype(ctx, JSON_SUBTYPE); + } +} /**************************************************************************** ** SQL functions used for testing and debugging @@ -2314,63 +3065,116 @@ static void jsonRemoveAllNulls(JsonNode *pNode){ #if SQLITE_DEBUG /* -** Print N node entries. +** Decode JSONB bytes in aBlob[] starting at iStart through but not +** including iEnd. Indent the +** content by nIndent spaces. */ -static void jsonDebugPrintNodeEntries( - JsonNode *aNode, /* First node entry to print */ - int N /* Number of node entries to print */ +static void jsonDebugPrintBlob( + JsonParse *pParse, /* JSON content */ + u32 iStart, /* Start rendering here */ + u32 iEnd, /* Do not render this byte or any byte after this one */ + int nIndent /* Indent by this many spaces */ ){ - int i; - for(i=0; iaBlob[iStart] & 0x0f; + u32 savedNBlob = pParse->nBlob; + printf("%5d:%*s", iStart, nIndent, ""); + if( pParse->nBlobAlloc>pParse->nBlob ){ + pParse->nBlob = pParse->nBlobAlloc; } - printf("node %4u: %-7s n=%-5d", i, zType, aNode[i].n); - if( (aNode[i].jnFlags & ~JNODE_LABEL)!=0 ){ - u8 f = aNode[i].jnFlags; - if( f & JNODE_RAW ) printf(" RAW"); - if( f & JNODE_ESCAPE ) printf(" ESCAPE"); - if( f & JNODE_REMOVE ) printf(" REMOVE"); - if( f & JNODE_REPLACE ) printf(" REPLACE"); - if( f & JNODE_APPEND ) printf(" APPEND"); - if( f & JNODE_JSON5 ) printf(" JSON5"); + nn = n = jsonbPayloadSize(pParse, iStart, &sz); + if( nn==0 ) nn = 1; + if( sz>0 && xaBlob[iStart+i]); + if( n==0 ){ + printf(" ERROR invalid node size\n"); + iStart = n==0 ? iStart+1 : iEnd; + continue; } + pParse->nBlob = savedNBlob; + if( iStart+n+sz>iEnd ){ + iEnd = iStart+n+sz; + if( iEnd>pParse->nBlob ){ + if( pParse->nBlobAlloc>0 && iEnd>pParse->nBlobAlloc ){ + iEnd = pParse->nBlobAlloc; + }else{ + iEnd = pParse->nBlob; + } + } + } + printf(" <-- "); + switch( x ){ + case JSONB_NULL: printf("null"); break; + case JSONB_TRUE: printf("true"); break; + case JSONB_FALSE: printf("false"); break; + case JSONB_INT: printf("int"); break; + case JSONB_INT5: printf("int5"); break; + case JSONB_FLOAT: printf("float"); break; + case JSONB_FLOAT5: printf("float5"); break; + case JSONB_TEXT: printf("text"); break; + case JSONB_TEXTJ: printf("textj"); break; + case JSONB_TEXT5: printf("text5"); break; + case JSONB_TEXTRAW: printf("textraw"); break; + case JSONB_ARRAY: { + printf("array, %u bytes\n", sz); + jsonDebugPrintBlob(pParse, iStart+n, iStart+n+sz, nIndent+2); + showContent = 0; + break; + } + case JSONB_OBJECT: { + printf("object, %u bytes\n", sz); + jsonDebugPrintBlob(pParse, iStart+n, iStart+n+sz, nIndent+2); + showContent = 0; + break; + } + default: { + printf("ERROR: unknown node type\n"); + showContent = 0; + break; + } + } + if( showContent ){ + if( sz==0 && x<=JSONB_FALSE ){ + printf("\n"); + }else{ + u32 i; + printf(": \""); + for(i=iStart+n; iaBlob[i]; + if( c<0x20 || c>=0x7f ) c = '.'; + putchar(c); + } + printf("\"\n"); + } + } + iStart += n + sz; } } +static void jsonShowParse(JsonParse *pParse){ + if( pParse==0 ){ + printf("NULL pointer\n"); + return; + }else{ + printf("nBlobAlloc = %u\n", pParse->nBlobAlloc); + printf("nBlob = %u\n", pParse->nBlob); + printf("delta = %d\n", pParse->delta); + if( pParse->nBlob==0 ) return; + printf("content (bytes 0..%u):\n", pParse->nBlob-1); + } + jsonDebugPrintBlob(pParse, 0, pParse->nBlob, 0); +} #endif /* SQLITE_DEBUG */ - -#if 0 /* 1 for debugging. 0 normally. Requires -DSQLITE_DEBUG too */ -static void jsonDebugPrintParse(JsonParse *p){ - jsonDebugPrintNodeEntries(p->aNode, p->nNode); -} -static void jsonDebugPrintNode(JsonNode *pNode){ - jsonDebugPrintNodeEntries(pNode, jsonNodeSize(pNode)); -} -#else - /* The usual case */ -# define jsonDebugPrintNode(X) -# define jsonDebugPrintParse(X) -#endif - #ifdef SQLITE_DEBUG /* ** SQL function: json_parse(JSON) ** -** Parse JSON using jsonParseCached(). Then print a dump of that -** parse on standard output. Return the mimified JSON result, just -** like the json() function. +** Parse JSON using jsonParseFuncArg(). Then print a dump of that +** parse on standard output. */ static void jsonParseFunc( sqlite3_context *ctx, @@ -2380,36 +3184,9 @@ static void jsonParseFunc( JsonParse *p; /* The parse */ assert( argc==1 ); - p = jsonParseCached(ctx, argv[0], ctx, 0); - if( p==0 ) return; - printf("nNode = %u\n", p->nNode); - printf("nAlloc = %u\n", p->nAlloc); - printf("nJson = %d\n", p->nJson); - printf("nAlt = %d\n", p->nAlt); - printf("nErr = %u\n", p->nErr); - printf("oom = %u\n", p->oom); - printf("hasNonstd = %u\n", p->hasNonstd); - printf("useMod = %u\n", p->useMod); - printf("hasMod = %u\n", p->hasMod); - printf("nJPRef = %u\n", p->nJPRef); - printf("iSubst = %u\n", p->iSubst); - printf("iHold = %u\n", p->iHold); - jsonDebugPrintNodeEntries(p->aNode, p->nNode); - jsonReturnJson(p, p->aNode, ctx, 1, 0); -} - -/* -** The json_test1(JSON) function return true (1) if the input is JSON -** text generated by another json function. It returns (0) if the input -** is not known to be JSON. -*/ -static void jsonTest1Func( - sqlite3_context *ctx, - int argc, - sqlite3_value **argv -){ - UNUSED_PARAMETER(argc); - sqlite3_result_int(ctx, sqlite3_value_subtype(argv[0])==JSON_SUBTYPE); + p = jsonParseFuncArg(ctx, argv[0], 0); + jsonShowParse(p); + jsonParseFree(p); } #endif /* SQLITE_DEBUG */ @@ -2418,7 +3195,7 @@ static void jsonTest1Func( ****************************************************************************/ /* -** Implementation of the json_QUOTE(VALUE) function. Return a JSON value +** Implementation of the json_quote(VALUE) function. Return a JSON value ** corresponding to the SQL value input. Mostly this means putting ** double-quotes around strings and returning the unquoted string "null" ** when given a NULL input. @@ -2431,9 +3208,9 @@ static void jsonQuoteFunc( JsonString jx; UNUSED_PARAMETER(argc); - jsonInit(&jx, ctx); - jsonAppendValue(&jx, argv[0]); - jsonResult(&jx); + jsonStringInit(&jx, ctx); + jsonAppendSqlValue(&jx, argv[0]); + jsonReturnString(&jx, 0, 0); sqlite3_result_subtype(ctx, JSON_SUBTYPE); } @@ -2450,18 +3227,17 @@ static void jsonArrayFunc( int i; JsonString jx; - jsonInit(&jx, ctx); + jsonStringInit(&jx, ctx); jsonAppendChar(&jx, '['); for(i=0; inNode ); if( argc==2 ){ const char *zPath = (const char*)sqlite3_value_text(argv[1]); - pNode = jsonLookup(p, zPath, 0, ctx); - }else{ - pNode = p->aNode; - } - if( pNode==0 ){ - return; - } - if( pNode->eType==JSON_ARRAY ){ - while( 1 /*exit-by-break*/ ){ - i = 1; - while( i<=pNode->n ){ - if( (pNode[i].jnFlags & JNODE_REMOVE)==0 ) n++; - i += jsonNodeSize(&pNode[i]); - } - if( (pNode->jnFlags & JNODE_APPEND)==0 ) break; - if( p->useMod==0 ) break; - assert( pNode->eU==2 ); - pNode = &p->aNode[pNode->u.iAppend]; + if( zPath==0 ){ + jsonParseFree(p); + return; } + i = jsonLookupStep(p, 0, zPath[0]=='$' ? zPath+1 : "@", 0); + if( JSON_LOOKUP_ISERROR(i) ){ + if( i==JSON_LOOKUP_NOTFOUND ){ + /* no-op */ + }else if( i==JSON_LOOKUP_PATHERROR ){ + jsonBadPathError(ctx, zPath); + }else{ + sqlite3_result_error(ctx, "malformed JSON", -1); + } + eErr = 1; + i = 0; + } + }else{ + i = 0; } - sqlite3_result_int64(ctx, n); + if( (p->aBlob[i] & 0x0f)==JSONB_ARRAY ){ + cnt = jsonbArrayCount(p, i); + } + if( !eErr ) sqlite3_result_int64(ctx, cnt); + jsonParseFree(p); } -/* -** Bit values for the flags passed into jsonExtractFunc() or -** jsonSetFunc() via the user-data value. -*/ -#define JSON_JSON 0x01 /* Result is always JSON */ -#define JSON_SQL 0x02 /* Result is always SQL */ -#define JSON_ABPATH 0x03 /* Allow abbreviated JSON path specs */ -#define JSON_ISSET 0x04 /* json_set(), not json_insert() */ +/* True if the string is all digits */ +static int jsonAllDigits(const char *z, int n){ + int i; + for(i=0; i and ->> operators accept abbreviated PATH arguments. This - ** is mostly for compatibility with PostgreSQL, but also for - ** convenience. - ** - ** NUMBER ==> $[NUMBER] // PG compatible - ** LABEL ==> $.LABEL // PG compatible - ** [NUMBER] ==> $[NUMBER] // Not PG. Purely for convenience - */ - jsonInit(&jx, ctx); - if( sqlite3Isdigit(zPath[0]) ){ - jsonAppendRawNZ(&jx, "$[", 2); - jsonAppendRaw(&jx, zPath, (int)strlen(zPath)); - jsonAppendRawNZ(&jx, "]", 2); - }else{ - jsonAppendRawNZ(&jx, "$.", 1 + (zPath[0]!='[')); - jsonAppendRaw(&jx, zPath, (int)strlen(zPath)); - jsonAppendChar(&jx, 0); - } - pNode = jx.bErr ? 0 : jsonLookup(p, jx.zBuf, 0, ctx); - jsonReset(&jx); - }else{ - pNode = jsonLookup(p, zPath, 0, ctx); - } - if( pNode ){ - if( flags & JSON_JSON ){ - jsonReturnJson(p, pNode, ctx, 0, 0); - }else{ - jsonReturn(p, pNode, ctx, 1); - } - } - }else{ - pNode = jsonLookup(p, zPath, 0, ctx); - if( p->nErr==0 && pNode ) jsonReturn(p, pNode, ctx, 0); - } - }else{ - /* Two or more PATH arguments results in a JSON array with each - ** element of the array being the value selected by one of the PATHs */ - int i; - jsonInit(&jx, ctx); + flags = SQLITE_PTR_TO_INT(sqlite3_user_data(ctx)); + jsonStringInit(&jx, ctx); + if( argc>2 ){ jsonAppendChar(&jx, '['); - for(i=1; inErr ) break; - jsonAppendSeparator(&jx); - if( pNode ){ - jsonRenderNode(p, pNode, &jx); + } + for(i=1; i and ->> operators accept abbreviated PATH arguments. This + ** is mostly for compatibility with PostgreSQL, but also for + ** convenience. + ** + ** NUMBER ==> $[NUMBER] // PG compatible + ** LABEL ==> $.LABEL // PG compatible + ** [NUMBER] ==> $[NUMBER] // Not PG. Purely for convenience + */ + jsonStringInit(&jx, ctx); + if( jsonAllDigits(zPath, nPath) ){ + jsonAppendRawNZ(&jx, "[", 1); + jsonAppendRaw(&jx, zPath, nPath); + jsonAppendRawNZ(&jx, "]", 2); + }else if( jsonAllAlphanum(zPath, nPath) ){ + jsonAppendRawNZ(&jx, ".", 1); + jsonAppendRaw(&jx, zPath, nPath); + }else if( zPath[0]=='[' && nPath>=3 && zPath[nPath-1]==']' ){ + jsonAppendRaw(&jx, zPath, nPath); }else{ + jsonAppendRawNZ(&jx, ".\"", 2); + jsonAppendRaw(&jx, zPath, nPath); + jsonAppendRawNZ(&jx, "\"", 1); + } + jsonStringTerminate(&jx); + j = jsonLookupStep(p, 0, jx.zBuf, 0); + jsonStringReset(&jx); + }else{ + jsonBadPathError(ctx, zPath); + goto json_extract_error; + } + if( jnBlob ){ + if( argc==2 ){ + if( flags & JSON_JSON ){ + jsonStringInit(&jx, ctx); + jsonXlateBlobToText(p, j, &jx); + jsonReturnString(&jx, 0, 0); + jsonStringReset(&jx); + assert( (flags & JSON_BLOB)==0 ); + sqlite3_result_subtype(ctx, JSON_SUBTYPE); + }else{ + jsonReturnFromBlob(p, j, ctx, 0); + if( (flags & (JSON_SQL|JSON_BLOB))==0 + && (p->aBlob[j]&0x0f)>=JSONB_ARRAY + ){ + sqlite3_result_subtype(ctx, JSON_SUBTYPE); + } + } + }else{ + jsonAppendSeparator(&jx); + jsonXlateBlobToText(p, j, &jx); + } + }else if( j==JSON_LOOKUP_NOTFOUND ){ + if( argc==2 ){ + goto json_extract_error; /* Return NULL if not found */ + }else{ + jsonAppendSeparator(&jx); jsonAppendRawNZ(&jx, "null", 4); } + }else if( j==JSON_LOOKUP_ERROR ){ + sqlite3_result_error(ctx, "malformed JSON", -1); + goto json_extract_error; + }else{ + jsonBadPathError(ctx, zPath); + goto json_extract_error; } - if( i==argc ){ - jsonAppendChar(&jx, ']'); - jsonResult(&jx); + } + if( argc>2 ){ + jsonAppendChar(&jx, ']'); + jsonReturnString(&jx, 0, 0); + if( (flags & JSON_BLOB)==0 ){ sqlite3_result_subtype(ctx, JSON_SUBTYPE); } - jsonReset(&jx); } +json_extract_error: + jsonStringReset(&jx); + jsonParseFree(p); + return; } -/* This is the RFC 7396 MergePatch algorithm. +/* +** Return codes for jsonMergePatch() */ -static JsonNode *jsonMergePatch( - JsonParse *pParse, /* The JSON parser that contains the TARGET */ - u32 iTarget, /* Node of the TARGET in pParse */ - JsonNode *pPatch /* The PATCH */ +#define JSON_MERGE_OK 0 /* Success */ +#define JSON_MERGE_BADTARGET 1 /* Malformed TARGET blob */ +#define JSON_MERGE_BADPATCH 2 /* Malformed PATCH blob */ +#define JSON_MERGE_OOM 3 /* Out-of-memory condition */ + +/* +** RFC-7396 MergePatch for two JSONB blobs. +** +** pTarget is the target. pPatch is the patch. The target is updated +** in place. The patch is read-only. +** +** The original RFC-7396 algorithm is this: +** +** define MergePatch(Target, Patch): +** if Patch is an Object: +** if Target is not an Object: +** Target = {} # Ignore the contents and set it to an empty Object +** for each Name/Value pair in Patch: +** if Value is null: +** if Name exists in Target: +** remove the Name/Value pair from Target +** else: +** Target[Name] = MergePatch(Target[Name], Value) +** return Target +** else: +** return Patch +** +** Here is an equivalent algorithm restructured to show the actual +** implementation: +** +** 01 define MergePatch(Target, Patch): +** 02 if Patch is not an Object: +** 03 return Patch +** 04 else: // if Patch is an Object +** 05 if Target is not an Object: +** 06 Target = {} +** 07 for each Name/Value pair in Patch: +** 08 if Name exists in Target: +** 09 if Value is null: +** 10 remove the Name/Value pair from Target +** 11 else +** 12 Target[name] = MergePatch(Target[Name], Value) +** 13 else if Value is not NULL: +** 14 if Value is not an Object: +** 15 Target[name] = Value +** 16 else: +** 17 Target[name] = MergePatch('{}',value) +** 18 return Target +** | +** ^---- Line numbers referenced in comments in the implementation +*/ +static int jsonMergePatch( + JsonParse *pTarget, /* The JSON parser that contains the TARGET */ + u32 iTarget, /* Index of TARGET in pTarget->aBlob[] */ + const JsonParse *pPatch, /* The PATCH */ + u32 iPatch /* Index of PATCH in pPatch->aBlob[] */ ){ - u32 i, j; - u32 iRoot; - JsonNode *pTarget; - if( pPatch->eType!=JSON_OBJECT ){ - return pPatch; + u8 x; /* Type of a single node */ + u32 n, sz=0; /* Return values from jsonbPayloadSize() */ + u32 iTCursor; /* Cursor position while scanning the target object */ + u32 iTStart; /* First label in the target object */ + u32 iTEndBE; /* Original first byte past end of target, before edit */ + u32 iTEnd; /* Current first byte past end of target */ + u8 eTLabel; /* Node type of the target label */ + u32 iTLabel = 0; /* Index of the label */ + u32 nTLabel = 0; /* Header size in bytes for the target label */ + u32 szTLabel = 0; /* Size of the target label payload */ + u32 iTValue = 0; /* Index of the target value */ + u32 nTValue = 0; /* Header size of the target value */ + u32 szTValue = 0; /* Payload size for the target value */ + + u32 iPCursor; /* Cursor position while scanning the patch */ + u32 iPEnd; /* First byte past the end of the patch */ + u8 ePLabel; /* Node type of the patch label */ + u32 iPLabel; /* Start of patch label */ + u32 nPLabel; /* Size of header on the patch label */ + u32 szPLabel; /* Payload size of the patch label */ + u32 iPValue; /* Start of patch value */ + u32 nPValue; /* Header size for the patch value */ + u32 szPValue; /* Payload size of the patch value */ + + assert( iTarget>=0 && iTargetnBlob ); + assert( iPatch>=0 && iPatchnBlob ); + x = pPatch->aBlob[iPatch] & 0x0f; + if( x!=JSONB_OBJECT ){ /* Algorithm line 02 */ + u32 szPatch; /* Total size of the patch, header+payload */ + u32 szTarget; /* Total size of the target, header+payload */ + n = jsonbPayloadSize(pPatch, iPatch, &sz); + szPatch = n+sz; + sz = 0; + n = jsonbPayloadSize(pTarget, iTarget, &sz); + szTarget = n+sz; + jsonBlobEdit(pTarget, iTarget, szTarget, pPatch->aBlob+iPatch, szPatch); + return pTarget->oom ? JSON_MERGE_OOM : JSON_MERGE_OK; /* Line 03 */ } - assert( iTargetnNode ); - pTarget = &pParse->aNode[iTarget]; - assert( (pPatch->jnFlags & JNODE_APPEND)==0 ); - if( pTarget->eType!=JSON_OBJECT ){ - jsonRemoveAllNulls(pPatch); - return pPatch; + x = pTarget->aBlob[iTarget] & 0x0f; + if( x!=JSONB_OBJECT ){ /* Algorithm line 05 */ + n = jsonbPayloadSize(pTarget, iTarget, &sz); + jsonBlobEdit(pTarget, iTarget+n, sz, 0, 0); + x = pTarget->aBlob[iTarget]; + pTarget->aBlob[iTarget] = (x & 0xf0) | JSONB_OBJECT; } - iRoot = iTarget; - for(i=1; in; i += jsonNodeSize(&pPatch[i+1])+1){ - u32 nKey; - const char *zKey; - assert( pPatch[i].eType==JSON_STRING ); - assert( pPatch[i].jnFlags & JNODE_LABEL ); - assert( pPatch[i].eU==1 ); - nKey = pPatch[i].n; - zKey = pPatch[i].u.zJContent; - for(j=1; jn; j += jsonNodeSize(&pTarget[j+1])+1 ){ - assert( pTarget[j].eType==JSON_STRING ); - assert( pTarget[j].jnFlags & JNODE_LABEL ); - if( jsonSameLabel(&pPatch[i], &pTarget[j]) ){ - if( pTarget[j+1].jnFlags & (JNODE_REMOVE|JNODE_REPLACE) ) break; - if( pPatch[i+1].eType==JSON_NULL ){ - pTarget[j+1].jnFlags |= JNODE_REMOVE; - }else{ - JsonNode *pNew = jsonMergePatch(pParse, iTarget+j+1, &pPatch[i+1]); - if( pNew==0 ) return 0; - if( pNew!=&pParse->aNode[iTarget+j+1] ){ - jsonParseAddSubstNode(pParse, iTarget+j+1); - jsonParseAddNodeArray(pParse, pNew, jsonNodeSize(pNew)); - } - pTarget = &pParse->aNode[iTarget]; - } - break; + n = jsonbPayloadSize(pPatch, iPatch, &sz); + if( NEVER(n==0) ) return JSON_MERGE_BADPATCH; + iPCursor = iPatch+n; + iPEnd = iPCursor+sz; + n = jsonbPayloadSize(pTarget, iTarget, &sz); + if( NEVER(n==0) ) return JSON_MERGE_BADTARGET; + iTStart = iTarget+n; + iTEndBE = iTStart+sz; + + while( iPCursoraBlob[iPCursor] & 0x0f; + if( ePLabelJSONB_TEXTRAW ){ + return JSON_MERGE_BADPATCH; + } + nPLabel = jsonbPayloadSize(pPatch, iPCursor, &szPLabel); + if( nPLabel==0 ) return JSON_MERGE_BADPATCH; + iPValue = iPCursor + nPLabel + szPLabel; + if( iPValue>=iPEnd ) return JSON_MERGE_BADPATCH; + nPValue = jsonbPayloadSize(pPatch, iPValue, &szPValue); + if( nPValue==0 ) return JSON_MERGE_BADPATCH; + iPCursor = iPValue + nPValue + szPValue; + if( iPCursor>iPEnd ) return JSON_MERGE_BADPATCH; + + iTCursor = iTStart; + iTEnd = iTEndBE + pTarget->delta; + while( iTCursoraBlob[iTCursor] & 0x0f; + if( eTLabelJSONB_TEXTRAW ){ + return JSON_MERGE_BADTARGET; + } + nTLabel = jsonbPayloadSize(pTarget, iTCursor, &szTLabel); + if( nTLabel==0 ) return JSON_MERGE_BADTARGET; + iTValue = iTLabel + nTLabel + szTLabel; + if( iTValue>=iTEnd ) return JSON_MERGE_BADTARGET; + nTValue = jsonbPayloadSize(pTarget, iTValue, &szTValue); + if( nTValue==0 ) return JSON_MERGE_BADTARGET; + if( iTValue + nTValue + szTValue > iTEnd ) return JSON_MERGE_BADTARGET; + isEqual = jsonLabelCompare( + (const char*)&pPatch->aBlob[iPLabel+nPLabel], + szPLabel, + (ePLabel==JSONB_TEXT || ePLabel==JSONB_TEXTRAW), + (const char*)&pTarget->aBlob[iTLabel+nTLabel], + szTLabel, + (eTLabel==JSONB_TEXT || eTLabel==JSONB_TEXTRAW)); + if( isEqual ) break; + iTCursor = iTValue + nTValue + szTValue; + } + x = pPatch->aBlob[iPValue] & 0x0f; + if( iTCursoroom) ) return JSON_MERGE_OOM; + }else{ + /* Algorithm line 12 */ + int rc, savedDelta = pTarget->delta; + pTarget->delta = 0; + rc = jsonMergePatch(pTarget, iTValue, pPatch, iPValue); + if( rc ) return rc; + pTarget->delta += savedDelta; + } + }else if( x>0 ){ /* Algorithm line 13 */ + /* No match and patch value is not NULL */ + u32 szNew = szPLabel+nPLabel; + if( (pPatch->aBlob[iPValue] & 0x0f)!=JSONB_OBJECT ){ /* Line 14 */ + jsonBlobEdit(pTarget, iTEnd, 0, 0, szPValue+nPValue+szNew); + if( pTarget->oom ) return JSON_MERGE_OOM; + memcpy(&pTarget->aBlob[iTEnd], &pPatch->aBlob[iPLabel], szNew); + memcpy(&pTarget->aBlob[iTEnd+szNew], + &pPatch->aBlob[iPValue], szPValue+nPValue); + }else{ + int rc, savedDelta; + jsonBlobEdit(pTarget, iTEnd, 0, 0, szNew+1); + if( pTarget->oom ) return JSON_MERGE_OOM; + memcpy(&pTarget->aBlob[iTEnd], &pPatch->aBlob[iPLabel], szNew); + pTarget->aBlob[iTEnd+szNew] = 0x00; + savedDelta = pTarget->delta; + pTarget->delta = 0; + rc = jsonMergePatch(pTarget, iTEnd+szNew,pPatch,iPValue); + if( rc ) return rc; + pTarget->delta += savedDelta; } } - if( j>=pTarget->n && pPatch[i+1].eType!=JSON_NULL ){ - int iStart; - JsonNode *pApnd; - u32 nApnd; - iStart = jsonParseAddNode(pParse, JSON_OBJECT, 0, 0); - jsonParseAddNode(pParse, JSON_STRING, nKey, zKey); - pApnd = &pPatch[i+1]; - if( pApnd->eType==JSON_OBJECT ) jsonRemoveAllNulls(pApnd); - nApnd = jsonNodeSize(pApnd); - jsonParseAddNodeArray(pParse, pApnd, jsonNodeSize(pApnd)); - if( pParse->oom ) return 0; - pParse->aNode[iStart].n = 1+nApnd; - pParse->aNode[iRoot].jnFlags |= JNODE_APPEND; - pParse->aNode[iRoot].u.iAppend = iStart; - VVA( pParse->aNode[iRoot].eU = 2 ); - iRoot = iStart; - pTarget = &pParse->aNode[iTarget]; - } } - return pTarget; + if( pTarget->delta ) jsonAfterEditSizeAdjust(pTarget, iTarget); + return pTarget->oom ? JSON_MERGE_OOM : JSON_MERGE_OK; } + /* ** Implementation of the json_mergepatch(JSON1,JSON2) function. Return a JSON ** object that is the result of running the RFC 7396 MergePatch() algorithm @@ -2696,28 +3635,27 @@ static void jsonPatchFunc( int argc, sqlite3_value **argv ){ - JsonParse *pX; /* The JSON that is being patched */ - JsonParse *pY; /* The patch */ - JsonNode *pResult; /* The result of the merge */ + JsonParse *pTarget; /* The TARGET */ + JsonParse *pPatch; /* The PATCH */ + int rc; /* Result code */ UNUSED_PARAMETER(argc); - pX = jsonParseCached(ctx, argv[0], ctx, 1); - if( pX==0 ) return; - assert( pX->hasMod==0 ); - pX->hasMod = 1; - pY = jsonParseCached(ctx, argv[1], ctx, 1); - if( pY==0 ) return; - pX->useMod = 1; - pY->useMod = 1; - pResult = jsonMergePatch(pX, 0, pY->aNode); - assert( pResult!=0 || pX->oom ); - if( pResult && pX->oom==0 ){ - jsonDebugPrintParse(pX); - jsonDebugPrintNode(pResult); - jsonReturnJson(pX, pResult, ctx, 0, 0); - }else{ - sqlite3_result_error_nomem(ctx); + assert( argc==2 ); + pTarget = jsonParseFuncArg(ctx, argv[0], JSON_EDITABLE); + if( pTarget==0 ) return; + pPatch = jsonParseFuncArg(ctx, argv[1], 0); + if( pPatch ){ + rc = jsonMergePatch(pTarget, 0, pPatch, 0); + if( rc==JSON_MERGE_OK ){ + jsonReturnParse(ctx, pTarget); + }else if( rc==JSON_MERGE_OOM ){ + sqlite3_result_error_nomem(ctx); + }else{ + sqlite3_result_error(ctx, "malformed JSON", -1); + } + jsonParseFree(pPatch); } + jsonParseFree(pTarget); } @@ -2741,23 +3679,23 @@ static void jsonObjectFunc( "of arguments", -1); return; } - jsonInit(&jx, ctx); + jsonStringInit(&jx, ctx); jsonAppendChar(&jx, '{'); for(i=0; i1); - if( pParse==0 ) return; - for(i=1; i<(u32)argc; i++){ + p = jsonParseFuncArg(ctx, argv[0], argc>1 ? JSON_EDITABLE : 0); + if( p==0 ) return; + for(i=1; inErr ) goto remove_done; - if( pNode ){ - pNode->jnFlags |= JNODE_REMOVE; - pParse->hasMod = 1; - pParse->useMod = 1; + if( zPath==0 ){ + goto json_remove_done; } - } - if( (pParse->aNode[0].jnFlags & JNODE_REMOVE)==0 ){ - jsonReturnJson(pParse, pParse->aNode, ctx, 1, 0); - } -remove_done: - jsonDebugPrintParse(p); -} - -/* -** Substitute the value at iNode with the pValue parameter. -*/ -static void jsonReplaceNode( - sqlite3_context *pCtx, - JsonParse *p, - int iNode, - sqlite3_value *pValue -){ - int idx = jsonParseAddSubstNode(p, iNode); - if( idx<=0 ){ - assert( p->oom ); - return; - } - switch( sqlite3_value_type(pValue) ){ - case SQLITE_NULL: { - jsonParseAddNode(p, JSON_NULL, 0, 0); - break; + if( zPath[0]!='$' ){ + goto json_remove_patherror; } - case SQLITE_FLOAT: { - char *z = sqlite3_mprintf("%!0.15g", sqlite3_value_double(pValue)); - int n; - if( z==0 ){ - p->oom = 1; - break; - } - n = sqlite3Strlen30(z); - jsonParseAddNode(p, JSON_REAL, n, z); - jsonParseAddCleanup(p, sqlite3_free, z); - break; + if( zPath[1]==0 ){ + /* json_remove(j,'$') returns NULL */ + goto json_remove_done; } - case SQLITE_INTEGER: { - char *z = sqlite3_mprintf("%lld", sqlite3_value_int64(pValue)); - int n; - if( z==0 ){ - p->oom = 1; - break; - } - n = sqlite3Strlen30(z); - jsonParseAddNode(p, JSON_INT, n, z); - jsonParseAddCleanup(p, sqlite3_free, z); - - break; - } - case SQLITE_TEXT: { - const char *z = (const char*)sqlite3_value_text(pValue); - u32 n = (u32)sqlite3_value_bytes(pValue); - if( z==0 ){ - p->oom = 1; - break; - } - if( sqlite3_value_subtype(pValue)!=JSON_SUBTYPE ){ - char *zCopy = sqlite3_malloc64( n+1 ); - int k; - if( zCopy ){ - memcpy(zCopy, z, n); - zCopy[n] = 0; - jsonParseAddCleanup(p, sqlite3_free, zCopy); - }else{ - p->oom = 1; - sqlite3_result_error_nomem(pCtx); - } - k = jsonParseAddNode(p, JSON_STRING, n, zCopy); - assert( k>0 || p->oom ); - if( p->oom==0 ) p->aNode[k].jnFlags |= JNODE_RAW; + p->eEdit = JEDIT_DEL; + p->delta = 0; + rc = jsonLookupStep(p, 0, zPath+1, 0); + if( JSON_LOOKUP_ISERROR(rc) ){ + if( rc==JSON_LOOKUP_NOTFOUND ){ + continue; /* No-op */ + }else if( rc==JSON_LOOKUP_PATHERROR ){ + jsonBadPathError(ctx, zPath); }else{ - JsonParse *pPatch = jsonParseCached(pCtx, pValue, pCtx, 1); - if( pPatch==0 ){ - p->oom = 1; - break; - } - jsonParseAddNodeArray(p, pPatch->aNode, pPatch->nNode); - /* The nodes copied out of pPatch and into p likely contain - ** u.zJContent pointers into pPatch->zJson. So preserve the - ** content of pPatch until p is destroyed. */ - assert( pPatch->nJPRef>=1 ); - pPatch->nJPRef++; - jsonParseAddCleanup(p, (void(*)(void*))jsonParseFree, pPatch); + sqlite3_result_error(ctx, "malformed JSON", -1); } - break; - } - default: { - jsonParseAddNode(p, JSON_NULL, 0, 0); - sqlite3_result_error(pCtx, "JSON cannot hold BLOB values", -1); - p->nErr++; - break; + goto json_remove_done; } } + jsonReturnParse(ctx, p); + jsonParseFree(p); + return; + +json_remove_patherror: + jsonBadPathError(ctx, zPath); + +json_remove_done: + jsonParseFree(p); + return; } /* @@ -2900,32 +3768,12 @@ static void jsonReplaceFunc( int argc, sqlite3_value **argv ){ - JsonParse *pParse; /* The parse */ - JsonNode *pNode; - const char *zPath; - u32 i; - if( argc<1 ) return; if( (argc&1)==0 ) { jsonWrongNumArgs(ctx, "replace"); return; } - pParse = jsonParseCached(ctx, argv[0], ctx, argc>1); - if( pParse==0 ) return; - pParse->nJPRef++; - for(i=1; i<(u32)argc; i+=2){ - zPath = (const char*)sqlite3_value_text(argv[i]); - pParse->useMod = 1; - pNode = jsonLookup(pParse, zPath, 0, ctx); - if( pParse->nErr ) goto replace_err; - if( pNode ){ - jsonReplaceNode(ctx, pParse, (u32)(pNode - pParse->aNode), argv[i+1]); - } - } - jsonReturnJson(pParse, pParse->aNode, ctx, 1, 0); -replace_err: - jsonDebugPrintParse(pParse); - jsonParseFree(pParse); + jsonInsertIntoBlob(ctx, argc, argv, JEDIT_REPL); } @@ -2946,39 +3794,16 @@ static void jsonSetFunc( int argc, sqlite3_value **argv ){ - JsonParse *pParse; /* The parse */ - JsonNode *pNode; - const char *zPath; - u32 i; - int bApnd; - int bIsSet = sqlite3_user_data(ctx)!=0; + + int flags = SQLITE_PTR_TO_INT(sqlite3_user_data(ctx)); + int bIsSet = (flags&JSON_ISSET)!=0; if( argc<1 ) return; if( (argc&1)==0 ) { jsonWrongNumArgs(ctx, bIsSet ? "set" : "insert"); return; } - pParse = jsonParseCached(ctx, argv[0], ctx, argc>1); - if( pParse==0 ) return; - pParse->nJPRef++; - for(i=1; i<(u32)argc; i+=2){ - zPath = (const char*)sqlite3_value_text(argv[i]); - bApnd = 0; - pParse->useMod = 1; - pNode = jsonLookup(pParse, zPath, &bApnd, ctx); - if( pParse->oom ){ - sqlite3_result_error_nomem(ctx); - goto jsonSetDone; - }else if( pParse->nErr ){ - goto jsonSetDone; - }else if( pNode && (bApnd || bIsSet) ){ - jsonReplaceNode(ctx, pParse, (u32)(pNode - pParse->aNode), argv[i+1]); - } - } - jsonDebugPrintParse(pParse); - jsonReturnJson(pParse, pParse->aNode, ctx, 1, 0); -jsonSetDone: - jsonParseFree(pParse); + jsonInsertIntoBlob(ctx, argc, argv, bIsSet ? JEDIT_SET : JEDIT_INS); } /* @@ -2994,27 +3819,93 @@ static void jsonTypeFunc( sqlite3_value **argv ){ JsonParse *p; /* The parse */ - const char *zPath; - JsonNode *pNode; + const char *zPath = 0; + u32 i; - p = jsonParseCached(ctx, argv[0], ctx, 0); + p = jsonParseFuncArg(ctx, argv[0], 0); if( p==0 ) return; if( argc==2 ){ zPath = (const char*)sqlite3_value_text(argv[1]); - pNode = jsonLookup(p, zPath, 0, ctx); + if( zPath==0 ) goto json_type_done; + if( zPath[0]!='$' ){ + jsonBadPathError(ctx, zPath); + goto json_type_done; + } + i = jsonLookupStep(p, 0, zPath+1, 0); + if( JSON_LOOKUP_ISERROR(i) ){ + if( i==JSON_LOOKUP_NOTFOUND ){ + /* no-op */ + }else if( i==JSON_LOOKUP_PATHERROR ){ + jsonBadPathError(ctx, zPath); + }else{ + sqlite3_result_error(ctx, "malformed JSON", -1); + } + goto json_type_done; + } }else{ - pNode = p->aNode; - } - if( pNode ){ - sqlite3_result_text(ctx, jsonType[pNode->eType], -1, SQLITE_STATIC); + i = 0; } + sqlite3_result_text(ctx, jsonbType[p->aBlob[i]&0x0f], -1, SQLITE_STATIC); +json_type_done: + jsonParseFree(p); } /* ** json_valid(JSON) +** json_valid(JSON, FLAGS) ** -** Return 1 if JSON is a well-formed canonical JSON string according -** to RFC-7159. Return 0 otherwise. +** Check the JSON argument to see if it is well-formed. The FLAGS argument +** encodes the various constraints on what is meant by "well-formed": +** +** 0x01 Canonical RFC-8259 JSON text +** 0x02 JSON text with optional JSON-5 extensions +** 0x04 Superficially appears to be JSONB +** 0x08 Strictly well-formed JSONB +** +** If the FLAGS argument is omitted, it defaults to 1. Useful values for +** FLAGS include: +** +** 1 Strict canonical JSON text +** 2 JSON text perhaps with JSON-5 extensions +** 4 Superficially appears to be JSONB +** 5 Canonical JSON text or superficial JSONB +** 6 JSON-5 text or superficial JSONB +** 8 Strict JSONB +** 9 Canonical JSON text or strict JSONB +** 10 JSON-5 text or strict JSONB +** +** Other flag combinations are redundant. For example, every canonical +** JSON text is also well-formed JSON-5 text, so FLAG values 2 and 3 +** are the same. Similarly, any input that passes a strict JSONB validation +** will also pass the superficial validation so 12 through 15 are the same +** as 8 through 11 respectively. +** +** This routine runs in linear time to validate text and when doing strict +** JSONB validation. Superficial JSONB validation is constant time, +** assuming the BLOB is already in memory. The performance advantage +** of superficial JSONB validation is why that option is provided. +** Application developers can choose to do fast superficial validation or +** slower strict validation, according to their specific needs. +** +** Only the lower four bits of the FLAGS argument are currently used. +** Higher bits are reserved for future expansion. To facilitate +** compatibility, the current implementation raises an error if any bit +** in FLAGS is set other than the lower four bits. +** +** The original circa 2015 implementation of the JSON routines in +** SQLite only supported canonical RFC-8259 JSON text and the json_valid() +** function only accepted one argument. That is why the default value +** for the FLAGS argument is 1, since FLAGS=1 causes this routine to only +** recognize canonical RFC-8259 JSON text as valid. The extra FLAGS +** argument was added when the JSON routines were extended to support +** JSON5-like extensions and binary JSONB stored in BLOBs. +** +** Return Values: +** +** * Raise an error if FLAGS is outside the range of 1 to 15. +** * Return NULL if the input is NULL +** * Return 1 if the input is well-formed. +** * Return 0 if the input is not well-formed. */ static void jsonValidFunc( sqlite3_context *ctx, @@ -3022,79 +3913,148 @@ static void jsonValidFunc( sqlite3_value **argv ){ JsonParse *p; /* The parse */ - UNUSED_PARAMETER(argc); - if( sqlite3_value_type(argv[0])==SQLITE_NULL ){ + u8 flags = 1; + u8 res = 0; + if( argc==2 ){ + i64 f = sqlite3_value_int64(argv[1]); + if( f<1 || f>15 ){ + sqlite3_result_error(ctx, "FLAGS parameter to json_valid() must be" + " between 1 and 15", -1); + return; + } + flags = f & 0x0f; + } + switch( sqlite3_value_type(argv[0]) ){ + case SQLITE_NULL: { #ifdef SQLITE_LEGACY_JSON_VALID - /* Incorrect legacy behavior was to return FALSE for a NULL input */ - sqlite3_result_int(ctx, 0); + /* Incorrect legacy behavior was to return FALSE for a NULL input */ + sqlite3_result_int(ctx, 0); #endif - return; - } - p = jsonParseCached(ctx, argv[0], 0, 0); - if( p==0 || p->oom ){ - sqlite3_result_error_nomem(ctx); - sqlite3_free(p); - }else{ - sqlite3_result_int(ctx, p->nErr==0 && (p->hasNonstd==0 || p->useMod)); - if( p->nErr ) jsonParseFree(p); + return; + } + case SQLITE_BLOB: { + if( (flags & 0x0c)!=0 && jsonFuncArgMightBeBinary(argv[0]) ){ + if( flags & 0x04 ){ + /* Superficial checking only - accomplisehd by the + ** jsonFuncArgMightBeBinary() call above. */ + res = 1; + }else{ + /* Strict checking. Check by translating BLOB->TEXT->BLOB. If + ** no errors occur, call that a "strict check". */ + JsonParse px; + JsonString sx; + u8 oom = 0; + memset(&px, 0, sizeof(px)); + px.aBlob = (u8*)sqlite3_value_blob(argv[0]); + px.nBlob = sqlite3_value_bytes(argv[0]); + jsonStringInit(&sx, 0); + jsonXlateBlobToText(&px, 0, &sx); + jsonParseReset(&px); + if( sx.eErr & JSTRING_OOM ) oom = 1; + if( sx.eErr==0 ){ + memset(&px, 0, sizeof(px)); + px.zJson = sx.zBuf; + px.nJson = sx.nUsed; + if( jsonXlateTextToBlob(&px, 0)==px.nJson ){ + res = 1; + } + oom |= px.oom; + jsonParseReset(&px); + } + jsonStringReset(&sx); + if( oom ){ + sqlite3_result_error_nomem(ctx); + return; + } + } + } + break; + } + default: { + JsonParse px; + if( (flags & 0x3)==0 ) break; + memset(&px, 0, sizeof(px)); + + p = jsonParseFuncArg(ctx, argv[0], JSON_KEEPERROR); + if( p ){ + if( p->oom ){ + sqlite3_result_error_nomem(ctx); + }else if( p->nErr ){ + /* no-op */ + }else if( (flags & 0x02)!=0 || p->hasNonstd==0 ){ + res = 1; + } + jsonParseFree(p); + }else{ + sqlite3_result_error_nomem(ctx); + } + break; + } } + sqlite3_result_int(ctx, res); } /* ** json_error_position(JSON) ** -** If the argument is not an interpretable JSON string, then return the 1-based -** character position at which the parser first recognized that the input -** was in error. The left-most character is 1. If the string is valid -** JSON, then return 0. +** If the argument is NULL, return NULL ** -** Note that json_valid() is only true for strictly conforming canonical JSON. -** But this routine returns zero if the input contains extension. Thus: +** If the argument is BLOB, do a fast validity check and return non-zero +** if the check fails. The returned value does not indicate where in the +** BLOB the error occurs. ** -** (1) If the input X is strictly conforming canonical JSON: -** -** json_valid(X) returns true -** json_error_position(X) returns 0 -** -** (2) If the input X is JSON but it includes extension (such as JSON5) that -** are not part of RFC-8259: -** -** json_valid(X) returns false -** json_error_position(X) return 0 -** -** (3) If the input X cannot be interpreted as JSON even taking extensions -** into account: -** -** json_valid(X) return false -** json_error_position(X) returns 1 or more +** Otherwise interpret the argument is TEXT (even if it is numeric) and +** return the 1-based character position for where the parser first recognized +** that the input was not valid JSON, or return 0 if the input text looks +** ok. JSON-5 extensions are accepted. */ static void jsonErrorFunc( sqlite3_context *ctx, int argc, sqlite3_value **argv ){ - JsonParse *p; /* The parse */ + i64 iErrPos = 0; /* Error position to be returned */ + JsonParse s; + + assert( argc==1 ); UNUSED_PARAMETER(argc); - if( sqlite3_value_type(argv[0])==SQLITE_NULL ) return; - p = jsonParseCached(ctx, argv[0], 0, 0); - if( p==0 || p->oom ){ - sqlite3_result_error_nomem(ctx); - sqlite3_free(p); - }else if( p->nErr==0 ){ - sqlite3_result_int(ctx, 0); - }else{ - int n = 1; - u32 i; - const char *z = (const char*)sqlite3_value_text(argv[0]); - for(i=0; iiErr && ALWAYS(z[i]); i++){ - if( (z[i]&0xc0)!=0x80 ) n++; + memset(&s, 0, sizeof(s)); + if( jsonFuncArgMightBeBinary(argv[0]) ){ + JsonString out; + jsonStringInit(&out, 0); + s.aBlob = (u8*)sqlite3_value_blob(argv[0]); + s.nBlob = sqlite3_value_bytes(argv[0]); + jsonXlateBlobToText(&s, 0, &out); + if( out.eErr ){ + iErrPos = (out.eErr & JSTRING_MALFORMED)!=0 ? 1 : -1; } - sqlite3_result_int(ctx, n); - jsonParseFree(p); + jsonStringReset(&out); + }else{ + s.zJson = (char*)sqlite3_value_text(argv[0]); + if( s.zJson==0 ) return; /* NULL input or OOM */ + s.nJson = sqlite3_value_bytes(argv[0]); + if( jsonConvertTextToBlob(&s,0) ){ + if( s.oom ){ + iErrPos = -1; + }else{ + /* Convert byte-offset s.iErr into a character offset */ + u32 k; + assert( s.zJson!=0 ); /* Because s.oom is false */ + for(k=0; kzBuf==0 ){ - jsonInit(pStr, ctx); + jsonStringInit(pStr, ctx); jsonAppendChar(pStr, '['); }else if( pStr->nUsed>1 ){ jsonAppendChar(pStr, ','); } pStr->pCtx = ctx; - jsonAppendValue(pStr, argv[0]); + jsonAppendSqlValue(pStr, argv[0]); } } static void jsonArrayCompute(sqlite3_context *ctx, int isFinal){ JsonString *pStr; pStr = (JsonString*)sqlite3_aggregate_context(ctx, 0); if( pStr ){ + int flags; pStr->pCtx = ctx; jsonAppendChar(pStr, ']'); - if( pStr->bErr ){ - if( pStr->bErr==1 ) sqlite3_result_error_nomem(ctx); - assert( pStr->bStatic ); + flags = SQLITE_PTR_TO_INT(sqlite3_user_data(ctx)); + if( pStr->eErr ){ + jsonReturnString(pStr, 0, 0); + return; + }else if( flags & JSON_BLOB ){ + jsonReturnStringAsBlob(pStr); + if( isFinal ){ + sqlite3RCStrUnref(pStr->zBuf); + }else{ + pStr->nUsed--; + } + return; }else if( isFinal ){ sqlite3_result_text(ctx, pStr->zBuf, (int)pStr->nUsed, pStr->bStatic ? SQLITE_TRANSIENT : @@ -3219,27 +4189,38 @@ static void jsonObjectStep( pStr = (JsonString*)sqlite3_aggregate_context(ctx, sizeof(*pStr)); if( pStr ){ if( pStr->zBuf==0 ){ - jsonInit(pStr, ctx); + jsonStringInit(pStr, ctx); jsonAppendChar(pStr, '{'); }else if( pStr->nUsed>1 ){ jsonAppendChar(pStr, ','); } pStr->pCtx = ctx; z = (const char*)sqlite3_value_text(argv[0]); - n = (u32)sqlite3_value_bytes(argv[0]); + n = sqlite3Strlen30(z); jsonAppendString(pStr, z, n); jsonAppendChar(pStr, ':'); - jsonAppendValue(pStr, argv[1]); + jsonAppendSqlValue(pStr, argv[1]); } } static void jsonObjectCompute(sqlite3_context *ctx, int isFinal){ JsonString *pStr; pStr = (JsonString*)sqlite3_aggregate_context(ctx, 0); if( pStr ){ + int flags; jsonAppendChar(pStr, '}'); - if( pStr->bErr ){ - if( pStr->bErr==1 ) sqlite3_result_error_nomem(ctx); - assert( pStr->bStatic ); + pStr->pCtx = ctx; + flags = SQLITE_PTR_TO_INT(sqlite3_user_data(ctx)); + if( pStr->eErr ){ + jsonReturnString(pStr, 0, 0); + return; + }else if( flags & JSON_BLOB ){ + jsonReturnStringAsBlob(pStr); + if( isFinal ){ + sqlite3RCStrUnref(pStr->zBuf); + }else{ + pStr->nUsed--; + } + return; }else if( isFinal ){ sqlite3_result_text(ctx, pStr->zBuf, (int)pStr->nUsed, pStr->bStatic ? SQLITE_TRANSIENT : @@ -3267,19 +4248,37 @@ static void jsonObjectFinal(sqlite3_context *ctx){ /**************************************************************************** ** The json_each virtual table ****************************************************************************/ +typedef struct JsonParent JsonParent; +struct JsonParent { + u32 iHead; /* Start of object or array */ + u32 iValue; /* Start of the value */ + u32 iEnd; /* First byte past the end */ + u32 nPath; /* Length of path */ + i64 iKey; /* Key for JSONB_ARRAY */ +}; + typedef struct JsonEachCursor JsonEachCursor; struct JsonEachCursor { sqlite3_vtab_cursor base; /* Base class - must be first */ u32 iRowid; /* The rowid */ - u32 iBegin; /* The first node of the scan */ - u32 i; /* Index in sParse.aNode[] of current row */ + u32 i; /* Index in sParse.aBlob[] of current row */ u32 iEnd; /* EOF when i equals or exceeds this value */ - u8 eType; /* Type of top-level element */ + u32 nRoot; /* Size of the root path in bytes */ + u8 eType; /* Type of the container for element i */ u8 bRecursive; /* True for json_tree(). False for json_each() */ - char *zJson; /* Input JSON */ - char *zRoot; /* Path by which to filter zJson */ + u32 nParent; /* Current nesting depth */ + u32 nParentAlloc; /* Space allocated for aParent[] */ + JsonParent *aParent; /* Parent elements of i */ + sqlite3 *db; /* Database connection */ + JsonString path; /* Current path */ JsonParse sParse; /* Parse of the input JSON */ }; +typedef struct JsonEachConnection JsonEachConnection; +struct JsonEachConnection { + sqlite3_vtab base; /* Base class - must be first */ + sqlite3 *db; /* Database connection */ +}; + /* Constructor for the json_each virtual table */ static int jsonEachConnect( @@ -3289,7 +4288,7 @@ static int jsonEachConnect( sqlite3_vtab **ppVtab, char **pzErr ){ - sqlite3_vtab *pNew; + JsonEachConnection *pNew; int rc; /* Column numbers */ @@ -3315,10 +4314,11 @@ static int jsonEachConnect( "CREATE TABLE x(key,value,type,atom,id,parent,fullkey,path," "json HIDDEN,root HIDDEN)"); if( rc==SQLITE_OK ){ - pNew = *ppVtab = sqlite3_malloc( sizeof(*pNew) ); + pNew = (JsonEachConnection*)(*ppVtab = sqlite3_malloc( sizeof(*pNew) )); if( pNew==0 ) return SQLITE_NOMEM; memset(pNew, 0, sizeof(*pNew)); sqlite3_vtab_config(db, SQLITE_VTAB_INNOCUOUS); + pNew->db = db; } return rc; } @@ -3331,12 +4331,15 @@ static int jsonEachDisconnect(sqlite3_vtab *pVtab){ /* constructor for a JsonEachCursor object for json_each(). */ static int jsonEachOpenEach(sqlite3_vtab *p, sqlite3_vtab_cursor **ppCursor){ + JsonEachConnection *pVtab = (JsonEachConnection*)p; JsonEachCursor *pCur; UNUSED_PARAMETER(p); pCur = sqlite3_malloc( sizeof(*pCur) ); if( pCur==0 ) return SQLITE_NOMEM; memset(pCur, 0, sizeof(*pCur)); + pCur->db = pVtab->db; + jsonStringZero(&pCur->path); *ppCursor = &pCur->base; return SQLITE_OK; } @@ -3354,20 +4357,23 @@ static int jsonEachOpenTree(sqlite3_vtab *p, sqlite3_vtab_cursor **ppCursor){ /* Reset a JsonEachCursor back to its original state. Free any memory ** held. */ static void jsonEachCursorReset(JsonEachCursor *p){ - sqlite3_free(p->zRoot); jsonParseReset(&p->sParse); + jsonStringReset(&p->path); + sqlite3DbFree(p->db, p->aParent); p->iRowid = 0; p->i = 0; + p->aParent = 0; + p->nParent = 0; + p->nParentAlloc = 0; p->iEnd = 0; p->eType = 0; - p->zJson = 0; - p->zRoot = 0; } /* Destructor for a jsonEachCursor object */ static int jsonEachClose(sqlite3_vtab_cursor *cur){ JsonEachCursor *p = (JsonEachCursor*)cur; jsonEachCursorReset(p); + sqlite3_free(cur); return SQLITE_OK; } @@ -3379,200 +4385,233 @@ static int jsonEachEof(sqlite3_vtab_cursor *cur){ return p->i >= p->iEnd; } -/* Advance the cursor to the next element for json_tree() */ -static int jsonEachNext(sqlite3_vtab_cursor *cur){ - JsonEachCursor *p = (JsonEachCursor*)cur; - if( p->bRecursive ){ - if( p->sParse.aNode[p->i].jnFlags & JNODE_LABEL ) p->i++; - p->i++; - p->iRowid++; - if( p->iiEnd ){ - u32 iUp = p->sParse.aUp[p->i]; - JsonNode *pUp = &p->sParse.aNode[iUp]; - p->eType = pUp->eType; - if( pUp->eType==JSON_ARRAY ){ - assert( pUp->eU==0 || pUp->eU==3 ); - testcase( pUp->eU==3 ); - VVA( pUp->eU = 3 ); - if( iUp==p->i-1 ){ - pUp->u.iKey = 0; - }else{ - pUp->u.iKey++; +/* +** If the cursor is currently pointing at the label of a object entry, +** then return the index of the value. For all other cases, return the +** current pointer position, which is the value. +*/ +static int jsonSkipLabel(JsonEachCursor *p){ + if( p->eType==JSONB_OBJECT ){ + u32 sz = 0; + u32 n = jsonbPayloadSize(&p->sParse, p->i, &sz); + return p->i + n + sz; + }else{ + return p->i; + } +} + +/* +** Append the path name for the current element. +*/ +static void jsonAppendPathName(JsonEachCursor *p){ + assert( p->nParent>0 ); + assert( p->eType==JSONB_ARRAY || p->eType==JSONB_OBJECT ); + if( p->eType==JSONB_ARRAY ){ + jsonPrintf(30, &p->path, "[%lld]", p->aParent[p->nParent-1].iKey); + }else{ + u32 n, sz = 0, k, i; + const char *z; + int needQuote = 0; + n = jsonbPayloadSize(&p->sParse, p->i, &sz); + k = p->i + n; + z = (const char*)&p->sParse.aBlob[k]; + if( sz==0 || !sqlite3Isalpha(z[0]) ){ + needQuote = 1; + }else{ + for(i=0; ieType ){ - case JSON_ARRAY: { - p->i += jsonNodeSize(&p->sParse.aNode[p->i]); - p->iRowid++; - break; - } - case JSON_OBJECT: { - p->i += 1 + jsonNodeSize(&p->sParse.aNode[p->i+1]); - p->iRowid++; - break; - } - default: { - p->i = p->iEnd; - break; - } + if( needQuote ){ + jsonPrintf(sz+4,&p->path,".\"%.*s\"", sz, z); + }else{ + jsonPrintf(sz+2,&p->path,".%.*s", sz, z); } } - return SQLITE_OK; } -/* Append an object label to the JSON Path being constructed -** in pStr. -*/ -static void jsonAppendObjectPathElement( - JsonString *pStr, - JsonNode *pNode -){ - int jj, nn; - const char *z; - assert( pNode->eType==JSON_STRING ); - assert( pNode->jnFlags & JNODE_LABEL ); - assert( pNode->eU==1 ); - z = pNode->u.zJContent; - nn = pNode->n; - if( (pNode->jnFlags & JNODE_RAW)==0 ){ - assert( nn>=2 ); - assert( z[0]=='"' || z[0]=='\'' ); - assert( z[nn-1]=='"' || z[0]=='\'' ); - if( nn>2 && sqlite3Isalpha(z[1]) ){ - for(jj=2; jjbRecursive ){ + u8 x; + u8 levelChange = 0; + u32 n, sz = 0; + u32 i = jsonSkipLabel(p); + x = p->sParse.aBlob[i] & 0x0f; + n = jsonbPayloadSize(&p->sParse, i, &sz); + if( x==JSONB_OBJECT || x==JSONB_ARRAY ){ + JsonParent *pParent; + if( p->nParent>=p->nParentAlloc ){ + JsonParent *pNew; + u64 nNew; + nNew = p->nParentAlloc*2 + 3; + pNew = sqlite3DbRealloc(p->db, p->aParent, sizeof(JsonParent)*nNew); + if( pNew==0 ) return SQLITE_NOMEM; + p->nParentAlloc = (u32)nNew; + p->aParent = pNew; + } + levelChange = 1; + pParent = &p->aParent[p->nParent]; + pParent->iHead = p->i; + pParent->iValue = i; + pParent->iEnd = i + n + sz; + pParent->iKey = -1; + pParent->nPath = (u32)p->path.nUsed; + if( p->eType && p->nParent ){ + jsonAppendPathName(p); + if( p->path.eErr ) rc = SQLITE_NOMEM; + } + p->nParent++; + p->i = i + n; + }else{ + p->i = i + n + sz; + } + while( p->nParent>0 && p->i >= p->aParent[p->nParent-1].iEnd ){ + p->nParent--; + p->path.nUsed = p->aParent[p->nParent].nPath; + levelChange = 1; + } + if( levelChange ){ + if( p->nParent>0 ){ + JsonParent *pParent = &p->aParent[p->nParent-1]; + u32 iVal = pParent->iValue; + p->eType = p->sParse.aBlob[iVal] & 0x0f; + }else{ + p->eType = 0; } } + }else{ + u32 n, sz = 0; + u32 i = jsonSkipLabel(p); + n = jsonbPayloadSize(&p->sParse, i, &sz); + p->i = i + n + sz; } - jsonPrintf(nn+2, pStr, ".%.*s", nn, z); + if( p->eType==JSONB_ARRAY && p->nParent ){ + p->aParent[p->nParent-1].iKey++; + } + p->iRowid++; + return rc; } -/* Append the name of the path for element i to pStr +/* Length of the path for rowid==0 in bRecursive mode. */ -static void jsonEachComputePath( - JsonEachCursor *p, /* The cursor */ - JsonString *pStr, /* Write the path here */ - u32 i /* Path to this element */ -){ - JsonNode *pNode, *pUp; - u32 iUp; - if( i==0 ){ - jsonAppendChar(pStr, '$'); - return; - } - iUp = p->sParse.aUp[i]; - jsonEachComputePath(p, pStr, iUp); - pNode = &p->sParse.aNode[i]; - pUp = &p->sParse.aNode[iUp]; - if( pUp->eType==JSON_ARRAY ){ - assert( pUp->eU==3 || (pUp->eU==0 && pUp->u.iKey==0) ); - testcase( pUp->eU==0 ); - jsonPrintf(30, pStr, "[%d]", pUp->u.iKey); - }else{ - assert( pUp->eType==JSON_OBJECT ); - if( (pNode->jnFlags & JNODE_LABEL)==0 ) pNode--; - jsonAppendObjectPathElement(pStr, pNode); +static int jsonEachPathLength(JsonEachCursor *p){ + u32 n = p->path.nUsed; + const char *z = p->path.zBuf; + if( p->iRowid==0 && p->bRecursive && n>1 ){ + if( z[n-1]==']' ){ + do{ + assert( n>1 ); + n--; + }while( z[n]!='[' ); + }else if( z[n-1]=='"' ){ + do{ + assert( n>1 ); + n--; + }while( z[n]!='.' || z[n+1]!='"' ); + }else{ + do{ + assert( n>1 ); + n--; + }while( z[n]!='.' ); + } } + return n; } /* Return the value of a column */ static int jsonEachColumn( sqlite3_vtab_cursor *cur, /* The cursor */ sqlite3_context *ctx, /* First argument to sqlite3_result_...() */ - int i /* Which column to return */ + int iColumn /* Which column to return */ ){ JsonEachCursor *p = (JsonEachCursor*)cur; - JsonNode *pThis = &p->sParse.aNode[p->i]; - switch( i ){ + switch( iColumn ){ case JEACH_KEY: { - if( p->i==0 ) break; - if( p->eType==JSON_OBJECT ){ - jsonReturn(&p->sParse, pThis, ctx, 0); - }else if( p->eType==JSON_ARRAY ){ - u32 iKey; - if( p->bRecursive ){ - if( p->iRowid==0 ) break; - assert( p->sParse.aNode[p->sParse.aUp[p->i]].eU==3 ); - iKey = p->sParse.aNode[p->sParse.aUp[p->i]].u.iKey; + if( p->nParent==0 ){ + u32 n, j; + if( p->nRoot==1 ) break; + j = jsonEachPathLength(p); + n = p->nRoot - j; + if( n==0 ){ + break; + }else if( p->path.zBuf[j]=='[' ){ + i64 x; + sqlite3Atoi64(&p->path.zBuf[j+1], &x, n-1, SQLITE_UTF8); + sqlite3_result_int64(ctx, x); + }else if( p->path.zBuf[j+1]=='"' ){ + sqlite3_result_text(ctx, &p->path.zBuf[j+2], n-3, SQLITE_TRANSIENT); }else{ - iKey = p->iRowid; + sqlite3_result_text(ctx, &p->path.zBuf[j+1], n-1, SQLITE_TRANSIENT); } - sqlite3_result_int64(ctx, (sqlite3_int64)iKey); + break; + } + if( p->eType==JSONB_OBJECT ){ + jsonReturnFromBlob(&p->sParse, p->i, ctx, 1); + }else{ + assert( p->eType==JSONB_ARRAY ); + sqlite3_result_int64(ctx, p->aParent[p->nParent-1].iKey); } break; } case JEACH_VALUE: { - if( pThis->jnFlags & JNODE_LABEL ) pThis++; - jsonReturn(&p->sParse, pThis, ctx, 0); + u32 i = jsonSkipLabel(p); + jsonReturnFromBlob(&p->sParse, i, ctx, 1); break; } case JEACH_TYPE: { - if( pThis->jnFlags & JNODE_LABEL ) pThis++; - sqlite3_result_text(ctx, jsonType[pThis->eType], -1, SQLITE_STATIC); + u32 i = jsonSkipLabel(p); + u8 eType = p->sParse.aBlob[i] & 0x0f; + sqlite3_result_text(ctx, jsonbType[eType], -1, SQLITE_STATIC); break; } case JEACH_ATOM: { - if( pThis->jnFlags & JNODE_LABEL ) pThis++; - if( pThis->eType>=JSON_ARRAY ) break; - jsonReturn(&p->sParse, pThis, ctx, 0); + u32 i = jsonSkipLabel(p); + if( (p->sParse.aBlob[i] & 0x0f)sParse, i, ctx, 1); + } break; } case JEACH_ID: { - sqlite3_result_int64(ctx, - (sqlite3_int64)p->i + ((pThis->jnFlags & JNODE_LABEL)!=0)); + sqlite3_result_int64(ctx, (sqlite3_int64)p->i); break; } case JEACH_PARENT: { - if( p->i>p->iBegin && p->bRecursive ){ - sqlite3_result_int64(ctx, (sqlite3_int64)p->sParse.aUp[p->i]); + if( p->nParent>0 && p->bRecursive ){ + sqlite3_result_int64(ctx, p->aParent[p->nParent-1].iHead); } break; } case JEACH_FULLKEY: { - JsonString x; - jsonInit(&x, ctx); - if( p->bRecursive ){ - jsonEachComputePath(p, &x, p->i); - }else{ - if( p->zRoot ){ - jsonAppendRaw(&x, p->zRoot, (int)strlen(p->zRoot)); - }else{ - jsonAppendChar(&x, '$'); - } - if( p->eType==JSON_ARRAY ){ - jsonPrintf(30, &x, "[%d]", p->iRowid); - }else if( p->eType==JSON_OBJECT ){ - jsonAppendObjectPathElement(&x, pThis); - } - } - jsonResult(&x); + u64 nBase = p->path.nUsed; + if( p->nParent ) jsonAppendPathName(p); + sqlite3_result_text64(ctx, p->path.zBuf, p->path.nUsed, + SQLITE_TRANSIENT, SQLITE_UTF8); + p->path.nUsed = nBase; break; } case JEACH_PATH: { - if( p->bRecursive ){ - JsonString x; - jsonInit(&x, ctx); - jsonEachComputePath(p, &x, p->sParse.aUp[p->i]); - jsonResult(&x); - break; - } - /* For json_each() path and root are the same so fall through - ** into the root case */ - /* no break */ deliberate_fall_through + u32 n = jsonEachPathLength(p); + sqlite3_result_text64(ctx, p->path.zBuf, n, + SQLITE_TRANSIENT, SQLITE_UTF8); + break; } default: { - const char *zRoot = p->zRoot; - if( zRoot==0 ) zRoot = "$"; - sqlite3_result_text(ctx, zRoot, -1, SQLITE_STATIC); + sqlite3_result_text(ctx, p->path.zBuf, p->nRoot, SQLITE_STATIC); break; } case JEACH_JSON: { - assert( i==JEACH_JSON ); - sqlite3_result_text(ctx, p->sParse.zJson, -1, SQLITE_STATIC); + if( p->sParse.zJson==0 ){ + sqlite3_result_blob(ctx, p->sParse.aBlob, p->sParse.nBlob, + SQLITE_STATIC); + }else{ + sqlite3_result_text(ctx, p->sParse.zJson, -1, SQLITE_STATIC); + } break; } } @@ -3663,86 +4702,100 @@ static int jsonEachFilter( int argc, sqlite3_value **argv ){ JsonEachCursor *p = (JsonEachCursor*)cur; - const char *z; const char *zRoot = 0; - sqlite3_int64 n; + u32 i, n, sz; UNUSED_PARAMETER(idxStr); UNUSED_PARAMETER(argc); jsonEachCursorReset(p); if( idxNum==0 ) return SQLITE_OK; - z = (const char*)sqlite3_value_text(argv[0]); - if( z==0 ) return SQLITE_OK; memset(&p->sParse, 0, sizeof(p->sParse)); p->sParse.nJPRef = 1; - if( sqlite3ValueIsOfClass(argv[0], sqlite3RCStrUnref) ){ - p->sParse.zJson = sqlite3RCStrRef((char*)z); - }else{ - n = sqlite3_value_bytes(argv[0]); - p->sParse.zJson = sqlite3RCStrNew( n+1 ); - if( p->sParse.zJson==0 ) return SQLITE_NOMEM; - memcpy(p->sParse.zJson, z, (size_t)n+1); - } - p->sParse.bJsonIsRCStr = 1; - p->zJson = p->sParse.zJson; - if( jsonParse(&p->sParse, 0) ){ - int rc = SQLITE_NOMEM; - if( p->sParse.oom==0 ){ - sqlite3_free(cur->pVtab->zErrMsg); - cur->pVtab->zErrMsg = sqlite3_mprintf("malformed JSON"); - if( cur->pVtab->zErrMsg ) rc = SQLITE_ERROR; + if( sqlite3_value_type(argv[0])==SQLITE_BLOB ){ + if( jsonFuncArgMightBeBinary(argv[0]) ){ + p->sParse.nBlob = sqlite3_value_bytes(argv[0]); + p->sParse.aBlob = (u8*)sqlite3_value_blob(argv[0]); + }else{ + goto json_each_malformed_input; } - jsonEachCursorReset(p); - return rc; - }else if( p->bRecursive && jsonParseFindParents(&p->sParse) ){ - jsonEachCursorReset(p); - return SQLITE_NOMEM; }else{ - JsonNode *pNode = 0; - if( idxNum==3 ){ - const char *zErr = 0; - zRoot = (const char*)sqlite3_value_text(argv[1]); - if( zRoot==0 ) return SQLITE_OK; - n = sqlite3_value_bytes(argv[1]); - p->zRoot = sqlite3_malloc64( n+1 ); - if( p->zRoot==0 ) return SQLITE_NOMEM; - memcpy(p->zRoot, zRoot, (size_t)n+1); - if( zRoot[0]!='$' ){ - zErr = zRoot; - }else{ - pNode = jsonLookupStep(&p->sParse, 0, p->zRoot+1, 0, &zErr); + p->sParse.zJson = (char*)sqlite3_value_text(argv[0]); + p->sParse.nJson = sqlite3_value_bytes(argv[0]); + if( p->sParse.zJson==0 ){ + p->i = p->iEnd = 0; + return SQLITE_OK; + } + if( jsonConvertTextToBlob(&p->sParse, 0) ){ + if( p->sParse.oom ){ + return SQLITE_NOMEM; } - if( zErr ){ + goto json_each_malformed_input; + } + } + if( idxNum==3 ){ + zRoot = (const char*)sqlite3_value_text(argv[1]); + if( zRoot==0 ) return SQLITE_OK; + if( zRoot[0]!='$' ){ + sqlite3_free(cur->pVtab->zErrMsg); + cur->pVtab->zErrMsg = jsonBadPathError(0, zRoot); + jsonEachCursorReset(p); + return cur->pVtab->zErrMsg ? SQLITE_ERROR : SQLITE_NOMEM; + } + p->nRoot = sqlite3Strlen30(zRoot); + if( zRoot[1]==0 ){ + i = p->i = 0; + p->eType = 0; + }else{ + i = jsonLookupStep(&p->sParse, 0, zRoot+1, 0); + if( JSON_LOOKUP_ISERROR(i) ){ + if( i==JSON_LOOKUP_NOTFOUND ){ + p->i = 0; + p->eType = 0; + p->iEnd = 0; + return SQLITE_OK; + } sqlite3_free(cur->pVtab->zErrMsg); - cur->pVtab->zErrMsg = jsonPathSyntaxError(zErr); + cur->pVtab->zErrMsg = jsonBadPathError(0, zRoot); jsonEachCursorReset(p); return cur->pVtab->zErrMsg ? SQLITE_ERROR : SQLITE_NOMEM; - }else if( pNode==0 ){ - return SQLITE_OK; } - }else{ - pNode = p->sParse.aNode; - } - p->iBegin = p->i = (int)(pNode - p->sParse.aNode); - p->eType = pNode->eType; - if( p->eType>=JSON_ARRAY ){ - assert( pNode->eU==0 ); - VVA( pNode->eU = 3 ); - pNode->u.iKey = 0; - p->iEnd = p->i + pNode->n + 1; - if( p->bRecursive ){ - p->eType = p->sParse.aNode[p->sParse.aUp[p->i]].eType; - if( p->i>0 && (p->sParse.aNode[p->i-1].jnFlags & JNODE_LABEL)!=0 ){ - p->i--; - } + if( p->sParse.iLabel ){ + p->i = p->sParse.iLabel; + p->eType = JSONB_OBJECT; }else{ - p->i++; + p->i = i; + p->eType = JSONB_ARRAY; } - }else{ - p->iEnd = p->i+1; } + jsonAppendRaw(&p->path, zRoot, p->nRoot); + }else{ + i = p->i = 0; + p->eType = 0; + p->nRoot = 1; + jsonAppendRaw(&p->path, "$", 1); + } + p->nParent = 0; + n = jsonbPayloadSize(&p->sParse, i, &sz); + p->iEnd = i+n+sz; + if( (p->sParse.aBlob[i] & 0x0f)>=JSONB_ARRAY && !p->bRecursive ){ + p->i = i + n; + p->eType = p->sParse.aBlob[i] & 0x0f; + p->aParent = sqlite3DbMallocZero(p->db, sizeof(JsonParent)); + if( p->aParent==0 ) return SQLITE_NOMEM; + p->nParent = 1; + p->nParentAlloc = 1; + p->aParent[0].iKey = 0; + p->aParent[0].iEnd = p->iEnd; + p->aParent[0].iHead = p->i; + p->aParent[0].iValue = i; } return SQLITE_OK; + +json_each_malformed_input: + sqlite3_free(cur->pVtab->zErrMsg); + cur->pVtab->zErrMsg = sqlite3_mprintf("malformed JSON"); + jsonEachCursorReset(p); + return cur->pVtab->zErrMsg ? SQLITE_ERROR : SQLITE_NOMEM; } /* The methods of the json_each virtual table */ @@ -3811,40 +4864,54 @@ static sqlite3_module jsonTreeModule = { void sqlite3RegisterJsonFunctions(void){ #ifndef SQLITE_OMIT_JSON static FuncDef aJsonFunc[] = { - /* calls sqlite3_result_subtype() */ - /* | */ - /* Uses cache ______ | __ calls sqlite3_value_subtype() */ - /* | | | */ - /* Num args _________ | | | ___ Flags */ - /* | | | | | */ - /* | | | | | */ - JFUNCTION(json, 1, 1, 1, 0, 0, jsonRemoveFunc), - JFUNCTION(json_array, -1, 0, 1, 1, 0, jsonArrayFunc), - JFUNCTION(json_array_length, 1, 1, 0, 0, 0, jsonArrayLengthFunc), - JFUNCTION(json_array_length, 2, 1, 0, 0, 0, jsonArrayLengthFunc), - JFUNCTION(json_error_position,1, 1, 0, 0, 0, jsonErrorFunc), - JFUNCTION(json_extract, -1, 1, 1, 0, 0, jsonExtractFunc), - JFUNCTION(->, 2, 1, 1, 0, JSON_JSON, jsonExtractFunc), - JFUNCTION(->>, 2, 1, 0, 0, JSON_SQL, jsonExtractFunc), - JFUNCTION(json_insert, -1, 1, 1, 1, 0, jsonSetFunc), - JFUNCTION(json_object, -1, 0, 1, 1, 0, jsonObjectFunc), - JFUNCTION(json_patch, 2, 1, 1, 0, 0, jsonPatchFunc), - JFUNCTION(json_quote, 1, 0, 1, 1, 0, jsonQuoteFunc), - JFUNCTION(json_remove, -1, 1, 1, 0, 0, jsonRemoveFunc), - JFUNCTION(json_replace, -1, 1, 1, 1, 0, jsonReplaceFunc), - JFUNCTION(json_set, -1, 1, 1, 1, JSON_ISSET, jsonSetFunc), - JFUNCTION(json_type, 1, 1, 0, 0, 0, jsonTypeFunc), - JFUNCTION(json_type, 2, 1, 0, 0, 0, jsonTypeFunc), - JFUNCTION(json_valid, 1, 1, 0, 0, 0, jsonValidFunc), + /* sqlite3_result_subtype() ----, ,--- sqlite3_value_subtype() */ + /* | | */ + /* Uses cache ------, | | ,---- Returns JSONB */ + /* | | | | */ + /* Number of arguments ---, | | | | ,--- Flags */ + /* | | | | | | */ + JFUNCTION(json, 1,1,1, 0,0,0, jsonRemoveFunc), + JFUNCTION(jsonb, 1,1,0, 0,1,0, jsonRemoveFunc), + JFUNCTION(json_array, -1,0,1, 1,0,0, jsonArrayFunc), + JFUNCTION(jsonb_array, -1,0,1, 1,1,0, jsonArrayFunc), + JFUNCTION(json_array_length, 1,1,0, 0,0,0, jsonArrayLengthFunc), + JFUNCTION(json_array_length, 2,1,0, 0,0,0, jsonArrayLengthFunc), + JFUNCTION(json_error_position,1,1,0, 0,0,0, jsonErrorFunc), + JFUNCTION(json_extract, -1,1,1, 0,0,0, jsonExtractFunc), + JFUNCTION(jsonb_extract, -1,1,0, 0,1,0, jsonExtractFunc), + JFUNCTION(->, 2,1,1, 0,0,JSON_JSON, jsonExtractFunc), + JFUNCTION(->>, 2,1,0, 0,0,JSON_SQL, jsonExtractFunc), + JFUNCTION(json_insert, -1,1,1, 1,0,0, jsonSetFunc), + JFUNCTION(jsonb_insert, -1,1,0, 1,1,0, jsonSetFunc), + JFUNCTION(json_object, -1,0,1, 1,0,0, jsonObjectFunc), + JFUNCTION(jsonb_object, -1,0,1, 1,1,0, jsonObjectFunc), + JFUNCTION(json_patch, 2,1,1, 0,0,0, jsonPatchFunc), + JFUNCTION(jsonb_patch, 2,1,0, 0,1,0, jsonPatchFunc), + JFUNCTION(json_quote, 1,0,1, 1,0,0, jsonQuoteFunc), + JFUNCTION(json_remove, -1,1,1, 0,0,0, jsonRemoveFunc), + JFUNCTION(jsonb_remove, -1,1,0, 0,1,0, jsonRemoveFunc), + JFUNCTION(json_replace, -1,1,1, 1,0,0, jsonReplaceFunc), + JFUNCTION(jsonb_replace, -1,1,0, 1,1,0, jsonReplaceFunc), + JFUNCTION(json_set, -1,1,1, 1,0,JSON_ISSET, jsonSetFunc), + JFUNCTION(jsonb_set, -1,1,0, 1,1,JSON_ISSET, jsonSetFunc), + JFUNCTION(json_type, 1,1,0, 0,0,0, jsonTypeFunc), + JFUNCTION(json_type, 2,1,0, 0,0,0, jsonTypeFunc), + JFUNCTION(json_valid, 1,1,0, 0,0,0, jsonValidFunc), + JFUNCTION(json_valid, 2,1,0, 0,0,0, jsonValidFunc), #if SQLITE_DEBUG - JFUNCTION(json_parse, 1, 1, 1, 0, 0, jsonParseFunc), - JFUNCTION(json_test1, 1, 1, 0, 1, 0, jsonTest1Func), + JFUNCTION(json_parse, 1,1,0, 0,0,0, jsonParseFunc), #endif WAGGREGATE(json_group_array, 1, 0, 0, jsonArrayStep, jsonArrayFinal, jsonArrayValue, jsonGroupInverse, SQLITE_SUBTYPE|SQLITE_RESULT_SUBTYPE|SQLITE_UTF8| SQLITE_DETERMINISTIC), + WAGGREGATE(jsonb_group_array, 1, JSON_BLOB, 0, + jsonArrayStep, jsonArrayFinal, jsonArrayValue, jsonGroupInverse, + SQLITE_SUBTYPE|SQLITE_RESULT_SUBTYPE|SQLITE_UTF8|SQLITE_DETERMINISTIC), WAGGREGATE(json_group_object, 2, 0, 0, + jsonObjectStep, jsonObjectFinal, jsonObjectValue, jsonGroupInverse, + SQLITE_SUBTYPE|SQLITE_RESULT_SUBTYPE|SQLITE_UTF8|SQLITE_DETERMINISTIC), + WAGGREGATE(jsonb_group_object,2, JSON_BLOB, 0, jsonObjectStep, jsonObjectFinal, jsonObjectValue, jsonGroupInverse, SQLITE_SUBTYPE|SQLITE_RESULT_SUBTYPE|SQLITE_UTF8| SQLITE_DETERMINISTIC) diff --git a/src/pragma.c b/src/pragma.c index 38c564753a..83bc7757d8 100644 --- a/src/pragma.c +++ b/src/pragma.c @@ -1780,7 +1780,8 @@ void sqlite3Pragma( if( pVTab->pModule->iVersion<4 ) continue; if( pVTab->pModule->xIntegrity==0 ) continue; sqlite3VdbeAddOp3(v, OP_VCheck, i, 3, isQuick); - sqlite3VdbeAppendP4(v, pTab, P4_TABLE); + pTab->nTabRef++; + sqlite3VdbeAppendP4(v, pTab, P4_TABLEREF); a1 = sqlite3VdbeAddOp1(v, OP_IsNull, 3); VdbeCoverage(v); integrityCheckResultRow(v); sqlite3VdbeJumpHere(v, a1); diff --git a/src/printf.c b/src/printf.c index 3c0b182d39..c6b3803ca9 100644 --- a/src/printf.c +++ b/src/printf.c @@ -1369,7 +1369,7 @@ void sqlite3_str_appendf(StrAccum *p, const char *zFormat, ...){ /***************************************************************************** -** Reference counted string storage +** Reference counted string/blob storage *****************************************************************************/ /* diff --git a/src/select.c b/src/select.c index e5312c7ee7..f61b34f5fe 100644 --- a/src/select.c +++ b/src/select.c @@ -184,6 +184,9 @@ Select *sqlite3SelectNew( void sqlite3SelectDelete(sqlite3 *db, Select *p){ if( OK_IF_ALWAYS_TRUE(p) ) clearSelect(db, p, 1); } +void sqlite3SelectDeleteGeneric(sqlite3 *db, void *p){ + if( p ) clearSelect(db, (Select*)p, 1); +} /* ** Return a pointer to the right-most SELECT statement in a compound. @@ -3204,9 +3207,7 @@ multi_select_end: pDest->iSdst = dest.iSdst; pDest->nSdst = dest.nSdst; if( pDelete ){ - sqlite3ParserAddCleanup(pParse, - (void(*)(sqlite3*,void*))sqlite3SelectDelete, - pDelete); + sqlite3ParserAddCleanup(pParse, sqlite3SelectDeleteGeneric, pDelete); } return rc; } @@ -3757,8 +3758,7 @@ static int multiSelectOrderBy( /* Make arrangements to free the 2nd and subsequent arms of the compound ** after the parse has finished */ if( pSplit->pPrior ){ - sqlite3ParserAddCleanup(pParse, - (void(*)(sqlite3*,void*))sqlite3SelectDelete, pSplit->pPrior); + sqlite3ParserAddCleanup(pParse, sqlite3SelectDeleteGeneric, pSplit->pPrior); } pSplit->pPrior = pPrior; pPrior->pNext = pSplit; @@ -4579,9 +4579,7 @@ static int flattenSubquery( Table *pTabToDel = pSubitem->pTab; if( pTabToDel->nTabRef==1 ){ Parse *pToplevel = sqlite3ParseToplevel(pParse); - sqlite3ParserAddCleanup(pToplevel, - (void(*)(sqlite3*,void*))sqlite3DeleteTable, - pTabToDel); + sqlite3ParserAddCleanup(pToplevel, sqlite3DeleteTableGeneric, pTabToDel); testcase( pToplevel->earlyCleanup ); }else{ pTabToDel->nTabRef--; @@ -5628,8 +5626,7 @@ static struct Cte *searchWith( With *sqlite3WithPush(Parse *pParse, With *pWith, u8 bFree){ if( pWith ){ if( bFree ){ - pWith = (With*)sqlite3ParserAddCleanup(pParse, - (void(*)(sqlite3*,void*))sqlite3WithDelete, + pWith = (With*)sqlite3ParserAddCleanup(pParse, sqlite3WithDeleteGeneric, pWith); if( pWith==0 ) return 0; } @@ -7037,7 +7034,8 @@ static SrcItem *isSelfJoinView( /* ** Deallocate a single AggInfo object */ -static void agginfoFree(sqlite3 *db, AggInfo *p){ +static void agginfoFree(sqlite3 *db, void *pArg){ + AggInfo *p = (AggInfo*)pArg; sqlite3DbFree(db, p->aCol); sqlite3DbFree(db, p->aFunc); sqlite3DbFreeNN(db, p); @@ -7291,9 +7289,8 @@ int sqlite3Select( sqlite3TreeViewExprList(0, p->pOrderBy, 0, "ORDERBY"); } #endif - sqlite3ParserAddCleanup(pParse, - (void(*)(sqlite3*,void*))sqlite3ExprListDelete, - p->pOrderBy); + sqlite3ParserAddCleanup(pParse, sqlite3ExprListDeleteGeneric, + p->pOrderBy); testcase( pParse->earlyCleanup ); p->pOrderBy = 0; } @@ -7485,9 +7482,8 @@ int sqlite3Select( ){ TREETRACE(0x800,pParse,p, ("omit superfluous ORDER BY on %r FROM-clause subquery\n",i+1)); - sqlite3ParserAddCleanup(pParse, - (void(*)(sqlite3*,void*))sqlite3ExprListDelete, - pSub->pOrderBy); + sqlite3ParserAddCleanup(pParse, sqlite3ExprListDeleteGeneric, + pSub->pOrderBy); pSub->pOrderBy = 0; } @@ -8016,8 +8012,7 @@ int sqlite3Select( */ pAggInfo = sqlite3DbMallocZero(db, sizeof(*pAggInfo) ); if( pAggInfo ){ - sqlite3ParserAddCleanup(pParse, - (void(*)(sqlite3*,void*))agginfoFree, pAggInfo); + sqlite3ParserAddCleanup(pParse, agginfoFree, pAggInfo); testcase( pParse->earlyCleanup ); } if( db->mallocFailed ){ diff --git a/src/shell.c.in b/src/shell.c.in index 4a4a0e1fef..6d1312d0f8 100644 --- a/src/shell.c.in +++ b/src/shell.c.in @@ -12061,10 +12061,6 @@ int SQLITE_CDECL wmain(int argc, wchar_t **wargv){ }else if( cli_strcmp(z,"-init")==0 ){ zInitFile = cmdline_option_value(argc, argv, ++i); }else if( cli_strcmp(z,"-interactive")==0 ){ - /* Need to check for interactive override here to so that it can - ** affect console setup (for Windows only) and testing thereof. - */ - stdin_is_interactive = 1; }else if( cli_strcmp(z,"-batch")==0 ){ /* Need to check for batch mode here to so we can avoid printing ** informational messages (like from process_sqliterc) before @@ -12338,7 +12334,10 @@ int SQLITE_CDECL wmain(int argc, wchar_t **wargv){ sqlite3_libversion(), sqlite3_sourceid(), 8*(int)sizeof(char*)); return 0; }else if( cli_strcmp(z,"-interactive")==0 ){ - /* already handled */ + /* Need to check for interactive override here to so that it can + ** affect console setup (for Windows only) and testing thereof. + */ + stdin_is_interactive = 1; }else if( cli_strcmp(z,"-batch")==0 ){ /* already handled */ }else if( cli_strcmp(z,"-utf8")==0 ){ diff --git a/src/sqliteInt.h b/src/sqliteInt.h index bb61cb6916..a0d5962005 100644 --- a/src/sqliteInt.h +++ b/src/sqliteInt.h @@ -2114,11 +2114,11 @@ struct FuncDestructor { #define MFUNCTION(zName, nArg, xPtr, xFunc) \ {nArg, SQLITE_FUNC_BUILTIN|SQLITE_FUNC_CONSTANT|SQLITE_UTF8, \ xPtr, 0, xFunc, 0, 0, 0, #zName, {0} } -#define JFUNCTION(zName, nArg, bUseCache, bWS, bRS, iArg, xFunc) \ +#define JFUNCTION(zName, nArg, bUseCache, bWS, bRS, bJsonB, iArg, xFunc) \ {nArg, SQLITE_FUNC_BUILTIN|SQLITE_DETERMINISTIC|SQLITE_FUNC_CONSTANT|\ SQLITE_UTF8|((bUseCache)*SQLITE_FUNC_RUNONLY)|\ ((bRS)*SQLITE_SUBTYPE)|((bWS)*SQLITE_RESULT_SUBTYPE), \ - SQLITE_INT_TO_PTR(iArg), 0, xFunc, 0, 0, 0, #zName, {0} } + SQLITE_INT_TO_PTR(iArg|((bJsonB)*JSON_BLOB)),0,xFunc,0, 0, 0, #zName, {0} } #define INLINE_FUNC(zName, nArg, iArg, mFlags) \ {nArg, SQLITE_FUNC_BUILTIN|\ SQLITE_UTF8|SQLITE_FUNC_INLINE|SQLITE_FUNC_CONSTANT|(mFlags), \ @@ -4116,6 +4116,9 @@ struct sqlite3_str { ** ** 3. Make a (read-only) copy of a read-only RCStr string using ** sqlite3RCStrRef(). +** +** "String" is in the name, but an RCStr object can also be used to hold +** binary data. */ struct RCStr { u64 nRCRef; /* Number of references */ @@ -4800,6 +4803,7 @@ void sqlite3ExprOrderByAggregateError(Parse*,Expr*); void sqlite3ExprFunctionUsable(Parse*,const Expr*,const FuncDef*); void sqlite3ExprAssignVarNumber(Parse*, Expr*, u32); void sqlite3ExprDelete(sqlite3*, Expr*); +void sqlite3ExprDeleteGeneric(sqlite3*,void*); void sqlite3ExprDeferredDelete(Parse*, Expr*); void sqlite3ExprUnmapAndDelete(Parse*, Expr*); ExprList *sqlite3ExprListAppend(Parse*,ExprList*,Expr*); @@ -4809,6 +4813,7 @@ void sqlite3ExprListSetSortOrder(ExprList*,int,int); void sqlite3ExprListSetName(Parse*,ExprList*,const Token*,int); void sqlite3ExprListSetSpan(Parse*,ExprList*,const char*,const char*); void sqlite3ExprListDelete(sqlite3*, ExprList*); +void sqlite3ExprListDeleteGeneric(sqlite3*,void*); u32 sqlite3ExprListFlags(const ExprList*); int sqlite3IndexHasDuplicateRootPage(Index*); int sqlite3Init(sqlite3*, char**); @@ -4899,6 +4904,7 @@ void sqlite3CreateView(Parse*,Token*,Token*,Token*,ExprList*,Select*,int,int); void sqlite3DropTable(Parse*, SrcList*, int, int); void sqlite3CodeDropTable(Parse*, Table*, int, int); void sqlite3DeleteTable(sqlite3*, Table*); +void sqlite3DeleteTableGeneric(sqlite3*, void*); void sqlite3FreeIndex(sqlite3*, Index*); #ifndef SQLITE_OMIT_AUTOINCREMENT void sqlite3AutoincrementBegin(Parse *pParse); @@ -4935,6 +4941,7 @@ int sqlite3Select(Parse*, Select*, SelectDest*); Select *sqlite3SelectNew(Parse*,ExprList*,SrcList*,Expr*,ExprList*, Expr*,ExprList*,u32,Expr*); void sqlite3SelectDelete(sqlite3*, Select*); +void sqlite3SelectDeleteGeneric(sqlite3*,void*); Table *sqlite3SrcListLookup(Parse*, SrcList*); int sqlite3IsReadOnly(Parse*, Table*, Trigger*); void sqlite3OpenTable(Parse*, int iCur, int iDb, Table*, int); @@ -5507,6 +5514,7 @@ const char *sqlite3JournalModename(int); void sqlite3CteDelete(sqlite3*,Cte*); With *sqlite3WithAdd(Parse*,With*,Cte*); void sqlite3WithDelete(sqlite3*,With*); + void sqlite3WithDeleteGeneric(sqlite3*,void*); With *sqlite3WithPush(Parse*, With*, u8); #else # define sqlite3CteNew(P,T,E,S) ((void*)0) diff --git a/src/vdbe.c b/src/vdbe.c index 5bc94d5278..121e4ce663 100644 --- a/src/vdbe.c +++ b/src/vdbe.c @@ -8193,9 +8193,10 @@ case OP_VCheck: { /* out2 */ pOut = &aMem[pOp->p2]; sqlite3VdbeMemSetNull(pOut); /* Innocent until proven guilty */ - assert( pOp->p4type==P4_TABLE ); + assert( pOp->p4type==P4_TABLEREF ); pTab = pOp->p4.pTab; assert( pTab!=0 ); + assert( pTab->nTabRef>0 ); assert( IsVirtual(pTab) ); if( pTab->u.vtab.p==0 ) break; pVtab = pTab->u.vtab.p->pVtab; @@ -8204,13 +8205,11 @@ case OP_VCheck: { /* out2 */ assert( pModule!=0 ); assert( pModule->iVersion>=4 ); assert( pModule->xIntegrity!=0 ); - pTab->nTabRef++; sqlite3VtabLock(pTab->u.vtab.p); assert( pOp->p1>=0 && pOp->p1nDb ); rc = pModule->xIntegrity(pVtab, db->aDb[pOp->p1].zDbSName, pTab->zName, pOp->p3, &zErr); sqlite3VtabUnlock(pTab->u.vtab.p); - sqlite3DeleteTable(db, pTab); if( rc ){ sqlite3_free(zErr); goto abort_due_to_error; @@ -8335,6 +8334,7 @@ case OP_VColumn: { /* ncycle */ const sqlite3_module *pModule; Mem *pDest; sqlite3_context sContext; + FuncDef nullFunc; VdbeCursor *pCur = p->apCsr[pOp->p1]; assert( pCur!=0 ); @@ -8352,6 +8352,9 @@ case OP_VColumn: { /* ncycle */ memset(&sContext, 0, sizeof(sContext)); sContext.pOut = pDest; sContext.enc = encoding; + nullFunc.pUserData = 0; + nullFunc.funcFlags = SQLITE_RESULT_SUBTYPE; + sContext.pFunc = &nullFunc; assert( pOp->p5==OPFLAG_NOCHNG || pOp->p5==0 ); if( pOp->p5 & OPFLAG_NOCHNG ){ sqlite3VdbeMemSetNull(pDest); diff --git a/src/vdbe.h b/src/vdbe.h index f44f24f93e..25bda6be7a 100644 --- a/src/vdbe.h +++ b/src/vdbe.h @@ -126,6 +126,7 @@ typedef struct VdbeOpList VdbeOpList; #define P4_INT64 (-13) /* P4 is a 64-bit signed integer */ #define P4_INTARRAY (-14) /* P4 is a vector of 32-bit integers */ #define P4_FUNCCTX (-15) /* P4 is a pointer to an sqlite3_context object */ +#define P4_TABLEREF (-16) /* Like P4_TABLE, but reference counted */ /* Error message codes for OP_Halt */ #define P5_ConstraintNotNull 1 diff --git a/src/vdbeapi.c b/src/vdbeapi.c index 570cb3d8b6..14c6091e0c 100644 --- a/src/vdbeapi.c +++ b/src/vdbeapi.c @@ -950,9 +950,8 @@ int sqlite3_step(sqlite3_stmt *pStmt){ void *sqlite3_user_data(sqlite3_context *p){ #ifdef SQLITE_ENABLE_API_ARMOR if( p==0 ) return 0; -#else - assert( p && p->pFunc ); #endif + assert( p && p->pFunc ); return p->pFunc->pUserData; } diff --git a/src/vdbeaux.c b/src/vdbeaux.c index 718d30dca0..e45c2d00c4 100644 --- a/src/vdbeaux.c +++ b/src/vdbeaux.c @@ -1400,6 +1400,10 @@ static void freeP4(sqlite3 *db, int p4type, void *p4){ if( db->pnBytesFreed==0 ) sqlite3VtabUnlock((VTable *)p4); break; } + case P4_TABLEREF: { + if( db->pnBytesFreed==0 ) sqlite3DeleteTable(db, (Table*)p4); + break; + } } } diff --git a/test/json/json-q1-b.txt b/test/json/json-q1-b.txt new file mode 100644 index 0000000000..e78c63670a --- /dev/null +++ b/test/json/json-q1-b.txt @@ -0,0 +1,24 @@ +.mode qbox +.timer on +.param set $label 'q87' +SELECT rowid, x->>$label FROM data1 WHERE x->>$label IS NOT NULL; + +CREATE TEMP TABLE t2(x JSON TEXT); +WITH RECURSIVE + c(x) AS (VALUES(1) UNION ALL SELECT x+1 FROM c WHERE x<25000), + array1(y) AS ( + SELECT json_group_array( + json_object('x',x,'y',random(),'z',hex(randomblob(50))) + ) + FROM c + ), + c2(n) AS (VALUES(1) UNION ALL SELECT n+1 FROM c2 WHERE n<5) +INSERT INTO t2(x) + SELECT jsonb_object('a',n,'b',n*2,'c',y,'d',3,'e',5,'f',6) FROM array1, c2; +CREATE INDEX t2x1 ON t2(x->>'a'); +CREATE INDEX t2x2 ON t2(x->>'b'); +CREATE INDEX t2x3 ON t2(x->>'e'); +CREATE INDEX t2x4 ON t2(x->>'f'); +UPDATE t2 SET x=jsonb_replace(x,'$.f',(x->>'f')+1); +UPDATE t2 SET x=jsonb_set(x,'$.e',(x->>'f')-1); +UPDATE t2 SET x=jsonb_remove(x,'$.d'); diff --git a/test/json/json-speed-check.sh b/test/json/json-speed-check.sh index f7e7f1be58..c57bfecd7a 100755 --- a/test/json/json-speed-check.sh +++ b/test/json/json-speed-check.sh @@ -40,6 +40,7 @@ doCachegrind=1 doVdbeProfile=0 doWal=1 doDiff=1 +doJsonB=0 while test "$1" != ""; do case $1 in --nodiff) @@ -54,6 +55,9 @@ while test "$1" != ""; do --gcc7) CC=gcc-7 ;; + --jsonb) + doJsonB=1 + ;; -*) CC_OPTS="$CC_OPTS $1" ;; @@ -69,12 +73,18 @@ rm -f cachegrind.out.* jsonshell $CC -g -Os -Wall -I. $CC_OPTS ./shell.c ./sqlite3.c -o jsonshell -ldl -lpthread ls -l jsonshell | tee -a summary-$NAME.txt home=`echo $0 | sed -e 's,/[^/]*$,,'` -echo ./jsonshell json100mb.db "<$home/json-q1.txt" -valgrind --tool=cachegrind ./jsonshell json100mb.db <$home/json-q1.txt \ - 2>&1 | tee -a summary-$NAME.txt +if test $doJsonB -eq 1; then + echo ./jsonshell json100mb_b.db "<$home/json-q1-b.txt" + valgrind --tool=cachegrind ./jsonshell json100mb_b.db <$home/json-q1-b.txt \ + 2>&1 | tee -a summary-$NAME.txt +else + echo ./jsonshell json100mb.db "<$home/json-q1.txt" + valgrind --tool=cachegrind ./jsonshell json100mb.db <$home/json-q1.txt \ + 2>&1 | tee -a summary-$NAME.txt +fi cg_anno.tcl cachegrind.out.* >jout-$NAME.txt echo '*****************************************************' >>jout-$NAME.txt sed 's/^[0-9=-]\{9\}/==00000==/' summary-$NAME.txt >>jout-$NAME.txt -if test "$NAME" != "$BASELINE"; then +if test "$NAME" != "$BASELINE" -a $doDiff -ne 0; then fossil xdiff --tk -c 20 jout-$BASELINE.txt jout-$NAME.txt fi diff --git a/test/json101.test b/test/json101.test index c62991bbbf..bae68e7344 100644 --- a/test/json101.test +++ b/test/json101.test @@ -36,6 +36,9 @@ do_execsql_test json101-1.2 { do_catchsql_test json101-1.3 { SELECT json_array(1,printf('%.1000c','x'),x'abcd',3); } {1 {JSON cannot hold BLOB values}} +do_catchsql_test json101-1.3b { + SELECT jsonb_array(1,printf('%.1000c','x'),x'abcd',3); +} {1 {JSON cannot hold BLOB values}} do_execsql_test json101-1.4 { SELECT json_array(-9223372036854775808,9223372036854775807,0,1,-1, 0.0, 1.0, -1.0, -1e99, +2e100, @@ -47,36 +50,83 @@ do_execsql_test json101-1.4 { 'abcdefghijklmnopqrstuvwyxzABCDEFGHIJKLMNOPQRSTUVWXYZ', 99); } {[-9223372036854775808,9223372036854775807,0,1,-1,0.0,1.0,-1.0,-1.0e+99,2.0e+100,"one","two","three",4,5,6,7,8,9,10,11,12,13,14,15,16,17,18,19,null,21,22,23,24,25,26,27,28,29,30,31,"abcdefghijklmnopqrstuvwyxzABCDEFGHIJKLMNOPQRSTUVWXYZ","abcdefghijklmnopqrstuvwyxzABCDEFGHIJKLMNOPQRSTUVWXYZ","abcdefghijklmnopqrstuvwyxzABCDEFGHIJKLMNOPQRSTUVWXYZ",99]} +do_execsql_test json101-1.4b { + SELECT json(jsonb_array(-9223372036854775808,9223372036854775807,0,1,-1, + 0.0, 1.0, -1.0, -1e99, +2e100, + 'one','two','three', + 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, + 19, NULL, 21, 22, 23, 24, 25, 26, 27, 28, 29, 30, 31, + 'abcdefghijklmnopqrstuvwyxzABCDEFGHIJKLMNOPQRSTUVWXYZ', + 'abcdefghijklmnopqrstuvwyxzABCDEFGHIJKLMNOPQRSTUVWXYZ', + 'abcdefghijklmnopqrstuvwyxzABCDEFGHIJKLMNOPQRSTUVWXYZ', + 99)); +} {[-9223372036854775808,9223372036854775807,0,1,-1,0.0,1.0,-1.0,-1.0e+99,2.0e+100,"one","two","three",4,5,6,7,8,9,10,11,12,13,14,15,16,17,18,19,null,21,22,23,24,25,26,27,28,29,30,31,"abcdefghijklmnopqrstuvwyxzABCDEFGHIJKLMNOPQRSTUVWXYZ","abcdefghijklmnopqrstuvwyxzABCDEFGHIJKLMNOPQRSTUVWXYZ","abcdefghijklmnopqrstuvwyxzABCDEFGHIJKLMNOPQRSTUVWXYZ",99]} do_execsql_test json101-2.1 { SELECT json_object('a',1,'b',2.5,'c',null,'d','String Test'); } {{{"a":1,"b":2.5,"c":null,"d":"String Test"}}} +do_execsql_test json101-2.1b { + SELECT json(jsonb_object('a',1,'b',2.5,'c',null,'d','String Test')); +} {{{"a":1,"b":2.5,"c":null,"d":"String Test"}}} do_catchsql_test json101-2.2 { SELECT json_object('a',printf('%.1000c','x'),2,2.5); } {1 {json_object() labels must be TEXT}} +do_catchsql_test json101-2.2b { + SELECT jsonb_object('a',printf('%.1000c','x'),2,2.5); +} {1 {json_object() labels must be TEXT}} +do_execsql_test json101-2.2.2 { + SELECT json_object('a',json_array('xyx',77,4.5),'x',2.5); +} {{{"a":["xyx",77,4.5],"x":2.5}}} +do_execsql_test json101-2.2.2b { + SELECT json(jsonb_object('a',json_array('xyx',77,4.5),'x',2.5)); +} {{{"a":["xyx",77,4.5],"x":2.5}}} +do_execsql_test json101-2.2.3 { + SELECT json_object('a',jsonb_array('xyx',77,4.5),'x',2.5); +} {{{"a":["xyx",77,4.5],"x":2.5}}} +do_execsql_test json101-2.2.3b { + SELECT json(jsonb_object('a',jsonb_array('xyx',77,4.5),'x',2.5)); +} {{{"a":["xyx",77,4.5],"x":2.5}}} do_catchsql_test json101-2.3 { SELECT json_object('a',1,'b'); } {1 {json_object() requires an even number of arguments}} do_catchsql_test json101-2.4 { SELECT json_object('a',printf('%.1000c','x'),'b',x'abcd'); } {1 {JSON cannot hold BLOB values}} +do_execsql_test json101-2.5 { + SELECT json_object('a',printf('%.10c','x'),'b',jsonb_array(1,2,3)); +} {{{"a":"xxxxxxxxxx","b":[1,2,3]}}} do_execsql_test json101-3.1 { SELECT json_replace('{"a":1,"b":2}','$.a','[3,4,5]'); } {{{"a":"[3,4,5]","b":2}}} +do_execsql_test json101-3.1b { + SELECT json(jsonb_replace('{"a":1,"b":2}','$.a','[3,4,5]')); +} {{{"a":"[3,4,5]","b":2}}} do_execsql_test json101-3.2 { SELECT json_replace('{"a":1,"b":2}','$.a',json('[3,4,5]')); } {{{"a":[3,4,5],"b":2}}} +do_execsql_test json101-3.2b { + SELECT json_replace('{"a":1,"b":2}','$.a',jsonb('[3,4,5]')); +} {{{"a":[3,4,5],"b":2}}} do_execsql_test json101-3.3 { SELECT json_type(json_set('{"a":1,"b":2}','$.b','{"x":3,"y":4}'),'$.b'); } {text} +do_execsql_test json101-3.3b { + SELECT json_type(jsonb_set('{"a":1,"b":2}','$.b','{"x":3,"y":4}'),'$.b'); +} {text} do_execsql_test json101-3.4 { SELECT json_type(json_set('{"a":1,"b":2}','$.b',json('{"x":3,"y":4}')),'$.b'); } {object} +do_execsql_test json101-3.4b { + SELECT json_type(jsonb_set('{"a":1,"b":2}','$.b',jsonb('{"x":3,"y":4}')),'$.b'); +} {object} ifcapable vtab { -do_execsql_test json101-3.5 { - SELECT fullkey, atom, '|' FROM json_tree(json_set('{}','$.x',123,'$.x',456)); -} {{$} {} | {$.x} 456 |} + do_execsql_test json101-3.5 { + SELECT fullkey, atom, '|' FROM json_tree(json_set('{}','$.x',123,'$.x',456)); + } {{$} {} | {$.x} 456 |} + do_execsql_test json101-3.5b { + SELECT fullkey, atom, '|' FROM json_tree(jsonb_set('{}','$.x',123,'$.x',456)); + } {{$} {} | {$.x} 456 |} } # Per rfc7159, any JSON value is allowed at the top level, and whitespace @@ -131,6 +181,13 @@ do_execsql_test json101-4.10 { WHERE json_extract(x,'$')<>x AND json_type(x) IN ('object','array'); } {4} +do_execsql_test json101-4.10b { + CREATE TABLE j1b AS SELECT jsonb(x) AS "x" FROM j1; + SELECT count(*) FROM j1b WHERE json_type(x) IN ('object','array'); + SELECT json(x) FROM j1b + WHERE json_extract(x,'$')<>json(x) + AND json_type(x) IN ('object','array'); +} {4} do_execsql_test json101-5.1 { CREATE TABLE j2(id INTEGER PRIMARY KEY, json, src); @@ -258,11 +315,17 @@ do_execsql_test json101-5.1 { } ]','https://adobe.github.io/Spry/samples/data_region/JSONDataSetSample.html'); SELECT count(*) FROM j2; -} {3} + CREATE TABLE j2b(id INTEGER PRIMARY KEY, json, src); + INSERT INTO J2b(id,json,src) SELECT id, jsonb(json), src FROM j2; + SELECT count(*) FROM j2b; +} {3 3} do_execsql_test json101-5.2 { SELECT id, json_valid(json), json_type(json), '|' FROM j2 ORDER BY id; } {1 1 object | 2 1 object | 3 1 array |} +do_execsql_test json101-5.2b { + SELECT id, json_valid(json,5), json_type(json), '|' FROM j2b ORDER BY id; +} {1 1 object | 2 1 object | 3 1 array |} ifcapable !vtab { finish_test @@ -277,6 +340,12 @@ do_execsql_test json101-5.3 { WHERE fullkey!=(path || CASE WHEN typeof(key)=='integer' THEN '['||key||']' ELSE '.'||key END); } {} +do_execsql_test json101-5.3b { + SELECT j2b.rowid, jx.rowid, fullkey, path, key + FROM j2b, json_tree(j2b.json) AS jx + WHERE fullkey!=(path || CASE WHEN typeof(key)=='integer' THEN '['||key||']' + ELSE '.'||key END); +} {} do_execsql_test json101-5.4 { SELECT j2.rowid, jx.rowid, fullkey, path, key FROM j2, json_each(j2.json) AS jx @@ -371,6 +440,13 @@ do_execsql_test json101-8.1 { UPDATE t8 SET b=json_array(a); SELECT b FROM t8; } {{["abc\u0001\u0002\u0003\u0004\u0005\u0006\u0007\b\t\n\u000b\f\r\u000e\u000f\u0010\u0011\u0012\u0013\u0014\u0015\u0016\u0017\u0018\u0019\u001a\u001b\u001c\u001d\u001e\u001f !\"#xyz"]}} +do_execsql_test json101-8.1b { + DROP TABLE IF EXISTS t8; + CREATE TABLE t8(a,b); + INSERT INTO t8(a) VALUES('abc' || char(1,2,3,4,5,6,7,8,9,10,11,12,13,14,15,16,17,18,19,20,21,22,23,24,25,26,27,28,29,30,31,32,33,34,35) || 'xyz'); + UPDATE t8 SET b=jsonb_array(a); + SELECT json(b) FROM t8; +} {{["abc\u0001\u0002\u0003\u0004\u0005\u0006\u0007\b\t\n\u000b\f\r\u000e\u000f\u0010\u0011\u0012\u0013\u0014\u0015\u0016\u0017\u0018\u0019\u001a\u001b\u001c\u001d\u001e\u001f !\"#xyz"]}} do_execsql_test json101-8.2 { SELECT a=json_extract(b,'$[0]') FROM t8; } {1} @@ -401,7 +477,7 @@ do_execsql_test json101-9.4 { SELECT json_quote(null); } {"null"} do_catchsql_test json101-9.5 { - SELECT json_quote(x'30313233'); + SELECT json_quote(x'3031323334'); } {1 {JSON cannot hold BLOB values}} do_catchsql_test json101-9.6 { SELECT json_quote(123,456) @@ -769,14 +845,23 @@ do_execsql_test json101-12.100 { } }'); } {} + do_execsql_test json101-12.110 { SELECT json_remove(x, '$.settings.layer2."dis.legomenon".forceDisplay') FROM t12; } {{{"settings":{"layer2":{"hapax.legomenon":{"forceDisplay":true,"transliterate":true,"add.footnote":true,"summary.report":true},"dis.legomenon":{"transliterate":false,"add.footnote":false,"summary.report":true},"tris.legomenon":{"forceDisplay":true,"transliterate":false,"add.footnote":false,"summary.report":false}}}}}} +do_execsql_test json101-12.110b { + SELECT json_remove(jsonb(x), '$.settings.layer2."dis.legomenon".forceDisplay') + FROM t12; +} {{{"settings":{"layer2":{"hapax.legomenon":{"forceDisplay":true,"transliterate":true,"add.footnote":true,"summary.report":true},"dis.legomenon":{"transliterate":false,"add.footnote":false,"summary.report":true},"tris.legomenon":{"forceDisplay":true,"transliterate":false,"add.footnote":false,"summary.report":false}}}}}} do_execsql_test json101-12.120 { SELECT json_extract(x, '$.settings.layer2."tris.legomenon"."summary.report"') FROM t12; } {0} +do_execsql_test json101-12.120b { + SELECT json_extract(jsonb(x), '$.settings.layer2."tris.legomenon"."summary.report"') + FROM t12; +} {0} # 2018-01-26 # ticket https://www.sqlite.org/src/tktview/80177f0c226ff54f6ddd41 @@ -840,16 +925,16 @@ do_execsql_test json101-14.170 { # do_execsql_test json101-15.100 { SELECT * FROM JSON_EACH('{"a":1, "b":2}'); -} {a 1 integer 1 2 {} {$.a} {$} b 2 integer 2 4 {} {$.b} {$}} +} {a 1 integer 1 1 {} {$.a} {$} b 2 integer 2 5 {} {$.b} {$}} do_execsql_test json101-15.110 { SELECT xyz.* FROM JSON_EACH('{"a":1, "b":2}') AS xyz; -} {a 1 integer 1 2 {} {$.a} {$} b 2 integer 2 4 {} {$.b} {$}} +} {a 1 integer 1 1 {} {$.a} {$} b 2 integer 2 5 {} {$.b} {$}} do_execsql_test json101-15.120 { SELECT * FROM (JSON_EACH('{"a":1, "b":2}')); -} {a 1 integer 1 2 {} {$.a} {$} b 2 integer 2 4 {} {$.b} {$}} +} {a 1 integer 1 1 {} {$.a} {$} b 2 integer 2 5 {} {$.b} {$}} do_execsql_test json101-15.130 { SELECT xyz.* FROM (JSON_EACH('{"a":1, "b":2}')) AS xyz; -} {a 1 integer 1 2 {} {$.a} {$} b 2 integer 2 4 {} {$.b} {$}} +} {a 1 integer 1 1 {} {$.a} {$} b 2 integer 2 5 {} {$.b} {$}} # 2019-11-10 # Mailing list bug report on the handling of surrogate pairs @@ -889,7 +974,7 @@ do_execsql_test json101-18.4 { } {6} do_catchsql_test json101-18.5 { SELECT json_extract('{"":8}', '$.'); -} {1 {JSON path error near ''}} +} {1 {bad JSON path: '$.'}} # 2022-08-29 https://sqlite.org/forum/forumpost/9b9e4716c0d7bbd1 # This is not a problem specifically with JSON functions. It is @@ -1054,6 +1139,29 @@ do_execsql_test json101-23.2 { FROM (SELECT json_set('[]','$[#]',0,'$[#]',1) AS j); } {{[0,1]} 0 1} - +# Insert/Set/Replace where the path specifies substructure that +# does not yet exist +# +proc tx x {return [string map [list ( \173 ) \175 ' \042 < \133 > \135] $x]} +foreach {id start path ins set repl} { + 1 {{}} {$.a.b.c} ('a':('b':('c':9))) ('a':('b':('c':9))) () + 2 {{a:4}} {$.a.b.c} ('a':4) ('a':4) ('a':4) + 3 {{a:{}}} {$.a.b.c} ('a':('b':('c':9))) ('a':('b':('c':9))) ('a':()) + 4 {[0,1,2]} {$[3].a[0].b} <0,1,2,('a':<('b':9)>)> <0,1,2,('a':<('b':9)>)> <0,1,2> + 5 {[0,1,2]} {$[1].a[0].b} <0,1,2> <0,1,2> <0,1,2> + 6 {[0,{},2]} {$[1].a[0].b} <0,('a':<('b':9)>),2> <0,('a':<('b':9)>),2> <0,(),2> + 7 {[0,1,2]} {$[3][0].b} <0,1,2,<('b':9)>> <0,1,2,<('b':9)>> <0,1,2> + 8 {[0,1,2]} {$[1][0].b} <0,1,2> <0,1,2> <0,1,2> +} { + do_execsql_test json101-24.$id.insert { + SELECT json_insert($start,$path,9); + } [list [tx $ins]] + do_execsql_test json101-24.$id.set { + SELECT json_set($start,$path,9); + } [list [tx $set]] + do_execsql_test json101-24.$id.replace { + SELECT json_replace($start,$path,9); + } [list [tx $repl]] +} finish_test diff --git a/test/json102.test b/test/json102.test index ce901efeb1..15a54b47c4 100644 --- a/test/json102.test +++ b/test/json102.test @@ -21,159 +21,492 @@ source $testdir/tester.tcl do_execsql_test json102-100 { SELECT json_object('ex','[52,3.14159]'); } {{{"ex":"[52,3.14159]"}}} +do_execsql_test json102-100b { + SELECT json(jsonb_object('ex','[52,3.14159]')); +} {{{"ex":"[52,3.14159]"}}} do_execsql_test json102-110 { SELECT json_object('ex',json('[52,3.14159]')); } {{{"ex":[52,3.14159]}}} +do_execsql_test json102-110-2 { + SELECT json(jsonb_object('ex',json('[52,3.14159]'))); +} {{{"ex":[52,3.14159]}}} +do_execsql_test json102-110-3 { + SELECT json_object('ex',jsonb('[52,3.14159]')); +} {{{"ex":[52,3.14159]}}} +do_execsql_test json102-110-3 { + SELECT json(jsonb_object('ex',jsonb('[52,3.14159]'))); +} {{{"ex":[52,3.14159]}}} do_execsql_test json102-120 { SELECT json_object('ex',json_array(52,3.14159)); } {{{"ex":[52,3.14159]}}} +do_execsql_test json102-120-2 { + SELECT json(jsonb_object('ex',json_array(52,3.14159))); +} {{{"ex":[52,3.14159]}}} +do_execsql_test json102-120-3 { + SELECT json_object('ex',jsonb_array(52,3.14159)); +} {{{"ex":[52,3.14159]}}} +do_execsql_test json102-120-4 { + SELECT json(jsonb_object('ex',jsonb_array(52,3.14159))); +} {{{"ex":[52,3.14159]}}} do_execsql_test json102-130 { SELECT json(' { "this" : "is", "a": [ "test" ] } '); } {{{"this":"is","a":["test"]}}} +do_execsql_test json102-130b { + SELECT json(jsonb(' { "this" : "is", "a": [ "test" ] } ')); +} {{{"this":"is","a":["test"]}}} do_execsql_test json102-140 { SELECT json_array(1,2,'3',4); } {{[1,2,"3",4]}} +do_execsql_test json102-140b { + SELECT json(jsonb_array(1,2,'3',4)); +} {{[1,2,"3",4]}} do_execsql_test json102-150 { SELECT json_array('[1,2]'); } {{["[1,2]"]}} +do_execsql_test json102-150b { + SELECT json(jsonb_array('[1,2]')); +} {{["[1,2]"]}} do_execsql_test json102-160 { SELECT json_array(json_array(1,2)); } {{[[1,2]]}} +do_execsql_test json102-160-2 { + SELECT json_array(jsonb_array(1,2)); +} {{[[1,2]]}} +do_execsql_test json102-160-3 { + SELECT json(jsonb_array(json_array(1,2))); +} {{[[1,2]]}} +do_execsql_test json102-160-4 { + SELECT json(jsonb_array(jsonb_array(1,2))); +} {{[[1,2]]}} do_execsql_test json102-170 { SELECT json_array(1,null,'3','[4,5]','{"six":7.7}'); } {{[1,null,"3","[4,5]","{\"six\":7.7}"]}} +do_execsql_test json102-170b { + SELECT json(jsonb_array(1,null,'3','[4,5]','{"six":7.7}')); +} {{[1,null,"3","[4,5]","{\"six\":7.7}"]}} do_execsql_test json102-180 { SELECT json_array(1,null,'3',json('[4,5]'),json('{"six":7.7}')); } {{[1,null,"3",[4,5],{"six":7.7}]}} +do_execsql_test json102-180-2 { + SELECT json_array(1,null,'3',jsonb('[4,5]'),json('{"six":7.7}')); +} {{[1,null,"3",[4,5],{"six":7.7}]}} +do_execsql_test json102-180-3 { + SELECT json(jsonb_array(1,null,'3',json('[4,5]'),json('{"six":7.7}'))); +} {{[1,null,"3",[4,5],{"six":7.7}]}} +do_execsql_test json102-180-4 { + SELECT json(jsonb_array(1,null,'3',jsonb('[4,5]'),jsonb('{"six":7.7}'))); +} {{[1,null,"3",[4,5],{"six":7.7}]}} do_execsql_test json102-190 { SELECT json_array_length('[1,2,3,4]'); } {{4}} +do_execsql_test json102-190b { + SELECT json_array_length(jsonb('[1,2,3,4]')); +} {{4}} do_execsql_test json102-191 { SELECT json_array_length( json_remove('[1,2,3,4]','$[2]') ); } {{3}} +do_execsql_test json102-191b { + SELECT json_array_length( jsonb_remove('[1,2,3,4]','$[2]') ); +} {{3}} do_execsql_test json102-200 { SELECT json_array_length('[1,2,3,4]', '$'); } {{4}} +do_execsql_test json102-200b { + SELECT json_array_length(jsonb('[1,2,3,4]'), '$'); +} {{4}} do_execsql_test json102-210 { SELECT json_array_length('[1,2,3,4]', '$[2]'); } {{0}} +do_execsql_test json102-210b { + SELECT json_array_length(jsonb('[1,2,3,4]'), '$[2]'); +} {{0}} do_execsql_test json102-220 { SELECT json_array_length('{"one":[1,2,3]}'); } {{0}} -do_execsql_test json102-230 { - SELECT json_array_length('{"one":[1,2,3]}', '$.one'); +do_execsql_test json102-220 { + SELECT json_array_length('{"one":[1,2,3]}'); +} {{0}} +do_execsql_test json102-230b { + SELECT json_array_length(jsonb('{"one":[1,2,3]}'), '$.one'); } {{3}} do_execsql_test json102-240 { SELECT json_array_length('{"one":[1,2,3]}', '$.two'); } {{}} +do_execsql_test json102-240b { + SELECT json_array_length(jsonb('{"one":[1,2,3]}'), '$.two'); +} {{}} do_execsql_test json102-250 { SELECT json_extract('{"a":2,"c":[4,5,{"f":7}]}', '$'); } {{{"a":2,"c":[4,5,{"f":7}]}}} +do_execsql_test json102-250-2 { + SELECT json_extract(jsonb('{"a":2,"c":[4,5,{"f":7}]}'), '$'); +} {{{"a":2,"c":[4,5,{"f":7}]}}} +do_execsql_test json102-250-3 { + SELECT json(jsonb_extract('{"a":2,"c":[4,5,{"f":7}]}', '$')); +} {{{"a":2,"c":[4,5,{"f":7}]}}} +do_execsql_test json102-250-4 { + SELECT json(jsonb_extract(jsonb('{"a":2,"c":[4,5,{"f":7}]}'), '$')); +} {{{"a":2,"c":[4,5,{"f":7}]}}} do_execsql_test json102-260 { SELECT json_extract('{"a":2,"c":[4,5,{"f":7}]}', '$.c'); } {{[4,5,{"f":7}]}} +do_execsql_test json102-260-2 { + SELECT json_extract(jsonb('{"a":2,"c":[4,5,{"f":7}]}'), '$.c'); +} {{[4,5,{"f":7}]}} +do_execsql_test json102-260-3 { + SELECT json(jsonb_extract('{"a":2,"c":[4,5,{"f":7}]}', '$.c')); +} {{[4,5,{"f":7}]}} +do_execsql_test json102-260-4 { + SELECT json(jsonb_extract(jsonb('{"a":2,"c":[4,5,{"f":7}]}'), '$.c')); +} {{[4,5,{"f":7}]}} do_execsql_test json102-270 { SELECT json_extract('{"a":2,"c":[4,5,{"f":7}]}', '$.c[2]'); } {{{"f":7}}} +do_execsql_test json102-270-2 { + SELECT json_extract(jsonb('{"a":2,"c":[4,5,{"f":7}]}'), '$.c[2]'); +} {{{"f":7}}} +do_execsql_test json102-270-3 { + SELECT json(jsonb_extract(jsonb('{"a":2,"c":[4,5,{"f":7}]}'), '$.c[2]')); +} {{{"f":7}}} +do_execsql_test json102-270-4 { + SELECT json(jsonb_extract('{"a":2,"c":[4,5,{"f":7}]}', '$.c[2]')); +} {{{"f":7}}} do_execsql_test json102-280 { SELECT json_extract('{"a":2,"c":[4,5,{"f":7}]}', '$.c[2].f'); } {{7}} +do_execsql_test json102-280b { + SELECT jsonb_extract('{"a":2,"c":[4,5,{"f":7}]}', '$.c[2].f'); +} {{7}} do_execsql_test json102-290 { SELECT json_extract('{"a":2,"c":[4,5],"f":7}','$.c','$.a'); } {{[[4,5],2]}} +do_execsql_test json102-290-2 { + SELECT json_extract(jsonb('{"a":2,"c":[4,5],"f":7}'),'$.c','$.a'); +} {{[[4,5],2]}} +do_execsql_test json102-290-3 { + SELECT json(jsonb_extract('{"a":2,"c":[4,5],"f":7}','$.c','$.a')); +} {{[[4,5],2]}} +do_execsql_test json102-290-4 { + SELECT json(jsonb_extract(jsonb('{"a":2,"c":[4,5],"f":7}'),'$.c','$.a')); +} {{[[4,5],2]}} do_execsql_test json102-300 { SELECT json_extract('{"a":2,"c":[4,5,{"f":7}]}', '$.x'); } {{}} +do_execsql_test json102-300b { + SELECT jsonb_extract('{"a":2,"c":[4,5,{"f":7}]}', '$.x'); +} {{}} do_execsql_test json102-310 { SELECT json_extract('{"a":2,"c":[4,5,{"f":7}]}', '$.x', '$.a'); } {{[null,2]}} +do_execsql_test json102-310-2 { + SELECT json_extract(jsonb('{"a":2,"c":[4,5,{"f":7}]}'), '$.x', '$.a'); +} {{[null,2]}} +do_execsql_test json102-310-3 { + SELECT json(jsonb_extract(jsonb('{"a":2,"c":[4,5,{"f":7}]}'), '$.x', '$.a')); +} {{[null,2]}} +do_execsql_test json102-310-43 { + SELECT json(jsonb_extract('{"a":2,"c":[4,5,{"f":7}]}', '$.x', '$.a')); +} {{[null,2]}} do_execsql_test json102-320 { SELECT json_insert('{"a":2,"c":4}', '$.a', 99); } {{{"a":2,"c":4}}} +do_execsql_test json102-320-2 { + SELECT json_insert(jsonb('{"a":2,"c":4}'), '$.a', 99); +} {{{"a":2,"c":4}}} +do_execsql_test json102-320-3 { + SELECT json(jsonb_insert('{"a":2,"c":4}', '$.a', 99)); +} {{{"a":2,"c":4}}} +do_execsql_test json102-320-4 { + SELECT json(jsonb_insert(jsonb('{"a":2,"c":4}'), '$.a', 99)); +} {{{"a":2,"c":4}}} do_execsql_test json102-330 { SELECT json_insert('{"a":2,"c":4}', '$.e', 99); } {{{"a":2,"c":4,"e":99}}} +do_execsql_test json102-330-2 { + SELECT json_insert(jsonb('{"a":2,"c":4}'), '$.e', 99); +} {{{"a":2,"c":4,"e":99}}} +do_execsql_test json102-330-3 { + SELECT json(jsonb_insert('{"a":2,"c":4}', '$.e', 99)); +} {{{"a":2,"c":4,"e":99}}} +do_execsql_test json102-330-4 { + SELECT json(jsonb_insert(jsonb('{"a":2,"c":4}'), '$.e', 99)); +} {{{"a":2,"c":4,"e":99}}} do_execsql_test json102-340 { SELECT json_replace('{"a":2,"c":4}', '$.a', 99); } {{{"a":99,"c":4}}} +do_execsql_test json102-340-2 { + SELECT json_replace(jsonb('{"a":2,"c":4}'), '$.a', 99); +} {{{"a":99,"c":4}}} +do_execsql_test json102-340-3 { + SELECT json(jsonb_replace('{"a":2,"c":4}', '$.a', 99)); +} {{{"a":99,"c":4}}} +do_execsql_test json102-340-4 { + SELECT json(jsonb_replace(jsonb('{"a":2,"c":4}'), '$.a', 99)); +} {{{"a":99,"c":4}}} do_execsql_test json102-350 { SELECT json_replace('{"a":2,"c":4}', '$.e', 99); } {{{"a":2,"c":4}}} +do_execsql_test json102-350-2 { + SELECT json_replace(jsonb('{"a":2,"c":4}'), '$.e', 99); +} {{{"a":2,"c":4}}} +do_execsql_test json102-350-3 { + SELECT json(jsonb_replace('{"a":2,"c":4}', '$.e', 99)); +} {{{"a":2,"c":4}}} +do_execsql_test json102-350-4 { + SELECT json(jsonb_replace(jsonb('{"a":2,"c":4}'), '$.e', 99)); +} {{{"a":2,"c":4}}} do_execsql_test json102-360 { SELECT json_set('{"a":2,"c":4}', '$.a', 99); } {{{"a":99,"c":4}}} +do_execsql_test json102-360-2 { + SELECT json_set(jsonb('{"a":2,"c":4}'), '$.a', 99); +} {{{"a":99,"c":4}}} +do_execsql_test json102-360-3 { + SELECT json(jsonb_set('{"a":2,"c":4}', '$.a', 99)); +} {{{"a":99,"c":4}}} +do_execsql_test json102-360-4 { + SELECT json(jsonb_set(jsonb('{"a":2,"c":4}'), '$.a', 99)); +} {{{"a":99,"c":4}}} do_execsql_test json102-370 { SELECT json_set('{"a":2,"c":4}', '$.e', 99); } {{{"a":2,"c":4,"e":99}}} +do_execsql_test json102-370-2 { + SELECT json_set(jsonb('{"a":2,"c":4}'), '$.e', 99); +} {{{"a":2,"c":4,"e":99}}} +do_execsql_test json102-370-3 { + SELECT json(jsonb_set('{"a":2,"c":4}', '$.e', 99)); +} {{{"a":2,"c":4,"e":99}}} +do_execsql_test json102-370-4 { + SELECT json(jsonb_set(jsonb('{"a":2,"c":4}'), '$.e', 99)); +} {{{"a":2,"c":4,"e":99}}} do_execsql_test json102-380 { SELECT json_set('{"a":2,"c":4}', '$.c', '[97,96]'); } {{{"a":2,"c":"[97,96]"}}} +do_execsql_test json102-380-2 { + SELECT json_set(jsonb('{"a":2,"c":4}'), '$.c', '[97,96]'); +} {{{"a":2,"c":"[97,96]"}}} +do_execsql_test json102-380-3 { + SELECT json(jsonb_set('{"a":2,"c":4}', '$.c', '[97,96]')); +} {{{"a":2,"c":"[97,96]"}}} +do_execsql_test json102-380-4 { + SELECT json(jsonb_set(jsonb('{"a":2,"c":4}'), '$.c', '[97,96]')); +} {{{"a":2,"c":"[97,96]"}}} do_execsql_test json102-390 { SELECT json_set('{"a":2,"c":4}', '$.c', json('[97,96]')); } {{{"a":2,"c":[97,96]}}} +do_execsql_test json102-390-2 { + SELECT json_set(jsonb('{"a":2,"c":4}'), '$.c', json('[97,96]')); +} {{{"a":2,"c":[97,96]}}} +do_execsql_test json102-390-3 { + SELECT json(jsonb_set('{"a":2,"c":4}', '$.c', json('[97,96]'))); +} {{{"a":2,"c":[97,96]}}} +do_execsql_test json102-390-4 { + SELECT json(jsonb_set(jsonb('{"a":2,"c":4}'), '$.c', json('[97,96]'))); +} {{{"a":2,"c":[97,96]}}} +do_execsql_test json102-390-5 { + SELECT json_set('{"a":2,"c":4}', '$.c', jsonb('[97,96]')); +} {{{"a":2,"c":[97,96]}}} +do_execsql_test json102-390-6 { + SELECT json_set(jsonb('{"a":2,"c":4}'), '$.c', jsonb('[97,96]')); +} {{{"a":2,"c":[97,96]}}} +do_execsql_test json102-390-7 { + SELECT json(jsonb_set('{"a":2,"c":4}', '$.c', jsonb('[97,96]'))); +} {{{"a":2,"c":[97,96]}}} +do_execsql_test json102-390-8 { + SELECT json(jsonb_set(jsonb('{"a":2,"c":4}'), '$.c', jsonb('[97,96]'))); +} {{{"a":2,"c":[97,96]}}} do_execsql_test json102-400 { SELECT json_set('{"a":2,"c":4}', '$.c', json_array(97,96)); } {{{"a":2,"c":[97,96]}}} +do_execsql_test json102-400-2 { + SELECT json_set(jsonb('{"a":2,"c":4}'), '$.c', json_array(97,96)); +} {{{"a":2,"c":[97,96]}}} +do_execsql_test json102-400-3 { + SELECT json(jsonb_set('{"a":2,"c":4}', '$.c', json_array(97,96))); +} {{{"a":2,"c":[97,96]}}} +do_execsql_test json102-400-4 { + SELECT json(jsonb_set(jsonb('{"a":2,"c":4}'), '$.c', json_array(97,96))); +} {{{"a":2,"c":[97,96]}}} +do_execsql_test json102-400-5 { + SELECT json_set('{"a":2,"c":4}', '$.c', jsonb_array(97,96)); +} {{{"a":2,"c":[97,96]}}} +do_execsql_test json102-400-6 { + SELECT json_set(jsonb('{"a":2,"c":4}'), '$.c', jsonb_array(97,96)); +} {{{"a":2,"c":[97,96]}}} +do_execsql_test json102-400-7 { + SELECT json(jsonb_set('{"a":2,"c":4}', '$.c', jsonb_array(97,96))); +} {{{"a":2,"c":[97,96]}}} +do_execsql_test json102-400-8 { + SELECT json(jsonb_set(jsonb('{"a":2,"c":4}'), '$.c', jsonb_array(97,96))); +} {{{"a":2,"c":[97,96]}}} do_execsql_test json102-410 { SELECT json_object('a',2,'c',4); } {{{"a":2,"c":4}}} +do_execsql_test json102-410b { + SELECT json(jsonb_object('a',2,'c',4)); +} {{{"a":2,"c":4}}} do_execsql_test json102-420 { SELECT json_object('a',2,'c','{e:5}'); } {{{"a":2,"c":"{e:5}"}}} +do_execsql_test json102-420b { + SELECT json(jsonb_object('a',2,'c','{e:5}')); +} {{{"a":2,"c":"{e:5}"}}} do_execsql_test json102-430 { SELECT json_object('a',2,'c',json_object('e',5)); } {{{"a":2,"c":{"e":5}}}} +do_execsql_test json102-430-2 { + SELECT json(jsonb_object('a',2,'c',json_object('e',5))); +} {{{"a":2,"c":{"e":5}}}} +do_execsql_test json102-430-3 { + SELECT json_object('a',2,'c',jsonb_object('e',5)); +} {{{"a":2,"c":{"e":5}}}} +do_execsql_test json102-430-4 { + SELECT json(jsonb_object('a',2,'c',jsonb_object('e',5))); +} {{{"a":2,"c":{"e":5}}}} do_execsql_test json102-440 { SELECT json_remove('[0,1,2,3,4]','$[2]'); } {{[0,1,3,4]}} +do_execsql_test json102-440-2 { + SELECT json_remove(jsonb('[0,1,2,3,4]'),'$[2]'); +} {{[0,1,3,4]}} +do_execsql_test json102-440-3 { + SELECT json(jsonb_remove('[0,1,2,3,4]','$[2]')); +} {{[0,1,3,4]}} +do_execsql_test json102-440-4 { + SELECT json(jsonb_remove(jsonb('[0,1,2,3,4]'),'$[2]')); +} {{[0,1,3,4]}} do_execsql_test json102-450 { SELECT json_remove('[0,1,2,3,4]','$[2]','$[0]'); } {{[1,3,4]}} +do_execsql_test json102-450-2 { + SELECT json_remove(jsonb('[0,1,2,3,4]'),'$[2]','$[0]'); +} {{[1,3,4]}} +do_execsql_test json102-450-3 { + SELECT json(jsonb_remove('[0,1,2,3,4]','$[2]','$[0]')); +} {{[1,3,4]}} +do_execsql_test json102-450-4 { + SELECT json(jsonb_remove(jsonb('[0,1,2,3,4]'),'$[2]','$[0]')); +} {{[1,3,4]}} do_execsql_test json102-460 { SELECT json_remove('[0,1,2,3,4]','$[0]','$[2]'); } {{[1,2,4]}} +do_execsql_test json102-460-2 { + SELECT json_remove(jsonb('[0,1,2,3,4]'),'$[0]','$[2]'); +} {{[1,2,4]}} +do_execsql_test json102-460-3 { + SELECT json(jsonb_remove('[0,1,2,3,4]','$[0]','$[2]')); +} {{[1,2,4]}} +do_execsql_test json102-460-4 { + SELECT json(jsonb_remove(jsonb('[0,1,2,3,4]'),'$[0]','$[2]')); +} {{[1,2,4]}} do_execsql_test json102-470 { SELECT json_remove('{"x":25,"y":42}'); } {{{"x":25,"y":42}}} +do_execsql_test json102-470-2 { + SELECT json_remove(jsonb('{"x":25,"y":42}')); +} {{{"x":25,"y":42}}} +do_execsql_test json102-470-3 { + SELECT json(jsonb_remove('{"x":25,"y":42}')); +} {{{"x":25,"y":42}}} +do_execsql_test json102-470-4 { + SELECT json(jsonb_remove(jsonb('{"x":25,"y":42}'))); +} {{{"x":25,"y":42}}} do_execsql_test json102-480 { SELECT json_remove('{"x":25,"y":42}','$.z'); } {{{"x":25,"y":42}}} +do_execsql_test json102-480-2 { + SELECT json_remove(jsonb('{"x":25,"y":42}'),'$.z'); +} {{{"x":25,"y":42}}} +do_execsql_test json102-480-3 { + SELECT json(jsonb_remove('{"x":25,"y":42}','$.z')); +} {{{"x":25,"y":42}}} +do_execsql_test json102-480-4 { + SELECT json(jsonb_remove(jsonb('{"x":25,"y":42}'),'$.z')); +} {{{"x":25,"y":42}}} do_execsql_test json102-490 { SELECT json_remove('{"x":25,"y":42}','$.y'); } {{{"x":25}}} +do_execsql_test json102-490-2 { + SELECT json_remove(jsonb('{"x":25,"y":42}'),'$.y'); +} {{{"x":25}}} +do_execsql_test json102-490-3 { + SELECT json(jsonb_remove('{"x":25,"y":42}','$.y')); +} {{{"x":25}}} +do_execsql_test json102-490-4 { + SELECT json(jsonb_remove(jsonb('{"x":25,"y":42}'),'$.y')); +} {{{"x":25}}} do_execsql_test json102-500 { SELECT json_remove('{"x":25,"y":42}','$'); } {{}} +do_execsql_test json102-500-2 { + SELECT json_remove(jsonb('{"x":25,"y":42}'),'$'); +} {{}} +do_execsql_test json102-500-3 { + SELECT json(jsonb_remove('{"x":25,"y":42}','$')); +} {{}} +do_execsql_test json102-500-4 { + SELECT json(jsonb_remove(jsonb('{"x":25,"y":42}'),'$')); +} {{}} do_execsql_test json102-510 { SELECT json_type('{"a":[2,3.5,true,false,null,"x"]}'); } {{object}} +do_execsql_test json102-510b { + SELECT json_type(x'cc0f1761cb0b133235332e350102001778'); +} {{object}} do_execsql_test json102-520 { SELECT json_type('{"a":[2,3.5,true,false,null,"x"]}','$'); } {{object}} +do_execsql_test json102-520b { + SELECT json_type(x'cc0f1761cb0b133235332e350102001778','$'); +} {{object}} do_execsql_test json102-530 { SELECT json_type('{"a":[2,3.5,true,false,null,"x"]}','$.a'); } {{array}} +do_execsql_test json102-530b { + SELECT json_type(x'cc0f1761cb0b133235332e350102001778','$.a'); +} {{array}} do_execsql_test json102-540 { SELECT json_type('{"a":[2,3.5,true,false,null,"x"]}','$.a[0]'); } {{integer}} +do_execsql_test json102-540b { + SELECT json_type(x'cc0f1761cb0b133235332e350102001778','$.a[0]'); +} {{integer}} do_execsql_test json102-550 { SELECT json_type('{"a":[2,3.5,true,false,null,"x"]}','$.a[1]'); } {{real}} +do_execsql_test json102-550b { + SELECT json_type(x'cc0f1761cb0b133235332e350102001778','$.a[1]'); +} {{real}} do_execsql_test json102-560 { SELECT json_type('{"a":[2,3.5,true,false,null,"x"]}','$.a[2]'); } {{true}} +do_execsql_test json102-560b { + SELECT json_type(x'cc0f1761cb0b133235332e350102001778','$.a[2]'); +} {{true}} do_execsql_test json102-570 { SELECT json_type('{"a":[2,3.5,true,false,null,"x"]}','$.a[3]'); } {{false}} +do_execsql_test json102-570b { + SELECT json_type(x'cc0f1761cb0b133235332e350102001778','$.a[3]'); +} {{false}} do_execsql_test json102-580 { SELECT json_type('{"a":[2,3.5,true,false,null,"x"]}','$.a[4]'); } {{null}} +do_execsql_test json102-580b { + SELECT json_type(x'cc0f1761cb0b133235332e350102001778','$.a[4]'); +} {{null}} do_execsql_test json102-590 { SELECT json_type('{"a":[2,3.5,true,false,null,"x"]}','$.a[5]'); } {{text}} +do_execsql_test json102-590b { + SELECT json_type(x'cc0f1761cb0b133235332e350102001778','$.a[5]'); +} {{text}} do_execsql_test json102-600 { SELECT json_type('{"a":[2,3.5,true,false,null,"x"]}','$.a[6]'); } {{}} +do_execsql_test json102-600b { + SELECT json_type(x'cc0f1761cb0b133235332e350102001778','$.a[6]'); +} {{}} do_execsql_test json102-610 { SELECT json_valid(char(123)||'"x":35'||char(125)); } {{1}} @@ -183,17 +516,24 @@ do_execsql_test json102-620 { ifcapable vtab { do_execsql_test json102-1000 { - CREATE TABLE user(name,phone); + CREATE TABLE user(name,phone,phoneb); INSERT INTO user(name,phone) VALUES ('Alice','["919-555-2345","804-555-3621"]'), ('Bob','["201-555-8872"]'), ('Cindy','["704-555-9983"]'), ('Dave','["336-555-8421","704-555-4321","803-911-4421"]'); + UPDATE user SET phoneb=jsonb(phone); SELECT DISTINCT user.name FROM user, json_each(user.phone) WHERE json_each.value LIKE '704-%' ORDER BY 1; } {Cindy Dave} +do_execsql_test json102-1000b { + SELECT DISTINCT user.name + FROM user, json_each(user.phoneb) + WHERE json_each.value LIKE '704-%' + ORDER BY 1; +} {Cindy Dave} do_execsql_test json102-1010 { UPDATE user @@ -253,6 +593,12 @@ do_execsql_test json102-1110 { WHERE json_tree.type NOT IN ('object','array') ORDER BY +big.rowid, +json_tree.id } $correct_answer +do_execsql_test json102-1110b { + SELECT big.rowid, fullkey, value + FROM big, json_tree(jsonb(big.json)) + WHERE json_tree.type NOT IN ('object','array') + ORDER BY +big.rowid, +json_tree.id +} $correct_answer do_execsql_test json102-1120 { SELECT big.rowid, fullkey, atom FROM big, json_tree(big.json) diff --git a/test/json105.test b/test/json105.test index 83face188e..5592191c22 100644 --- a/test/json105.test +++ b/test/json105.test @@ -96,18 +96,18 @@ json_replace_test 80 {'$.b[#-1]','AAA','$.b[#-1]','BBB'} \ do_catchsql_test json105-6.10 { SELECT json_extract(j, '$.b[#-]') FROM t1; -} {1 {JSON path error near '[#-]'}} +} {1 {bad JSON path: '$.b[#-]'}} do_catchsql_test json105-6.20 { SELECT json_extract(j, '$.b[#9]') FROM t1; -} {1 {JSON path error near '[#9]'}} +} {1 {bad JSON path: '$.b[#9]'}} do_catchsql_test json105-6.30 { SELECT json_extract(j, '$.b[#+2]') FROM t1; -} {1 {JSON path error near '[#+2]'}} +} {1 {bad JSON path: '$.b[#+2]'}} do_catchsql_test json105-6.40 { SELECT json_extract(j, '$.b[#-1') FROM t1; -} {1 {JSON path error near '[#-1'}} +} {1 {bad JSON path: '$.b[#-1'}} do_catchsql_test json105-6.50 { SELECT json_extract(j, '$.b[#-1x]') FROM t1; -} {1 {JSON path error near '[#-1x]'}} +} {1 {bad JSON path: '$.b[#-1x]'}} finish_test diff --git a/test/json501.test b/test/json501.test index 3318eea7f2..40b3b563db 100644 --- a/test/json501.test +++ b/test/json501.test @@ -252,13 +252,13 @@ do_execsql_test 8.11 { do_execsql_test 9.1 { WITH c(x) AS (VALUES('{x: +Infinity}')) SELECT x->>'x', json(x) FROM c; -} {Inf {{"x":9.0e999}}} +} {Inf {{"x":9e999}}} do_execsql_test 9.2 { WITH c(x) AS (VALUES('{x: -Infinity}')) SELECT x->>'x', json(x) FROM c; -} {-Inf {{"x":-9.0e999}}} +} {-Inf {{"x":-9e999}}} do_execsql_test 9.3 { WITH c(x) AS (VALUES('{x: Infinity}')) SELECT x->>'x', json(x) FROM c; -} {Inf {{"x":9.0e999}}} +} {Inf {{"x":9e999}}} do_execsql_test 9.4 { WITH c(x) AS (VALUES('{x: NaN}')) SELECT x->>'x', json(x) FROM c; } {{} {{"x":null}}} diff --git a/test/json502.test b/test/json502.test index 595bf63318..48c372c4f8 100644 --- a/test/json502.test +++ b/test/json502.test @@ -36,5 +36,30 @@ do_catchsql_test 2.3 { SELECT '{a:null,{"h":[1,[1,2,3]],"j":"abc"}:true}'->'$h[#-1]'; } {1 {malformed JSON}} +# Verify that escaped label names are compared correctly. +# +do_execsql_test 3.1 { + SELECT '{"a\x62c":123}' ->> 'abc'; +} 123 +do_execsql_test 3.2 { + SELECT '{"abc":123}' ->> 'a\x62c'; +} 123 + +db null null +do_execsql_test 3.3 { + DROP TABLE t1; + CREATE TABLE t1(x); + INSERT INTO t1 VALUES(json_insert('{}','$.a\',111,'$."b\\"',222)); + INSERT INTO t1 VALUES(jsonb_insert('{}','$.a\',111,'$."b\\"',222)); + SELECT x->'$.a\', x->'$.a\\', x->'$."a\\"', x->'$."b\\"' FROM t1; +} {111 null 111 222 111 null 111 222} + +do_execsql_test 3.4 { + SELECT json_patch('{"a\x62c":123}','{"ab\x63":456}') ->> 'abc'; +} 456 + +do_execsql_test 4.1 { + SELECT * FROM json_tree('{"\u0017":1}','$."\x17"'); +} {{\x17} 1 integer 1 1 null {$."\x17"} {$}} finish_test diff --git a/test/jsonb01.test b/test/jsonb01.test new file mode 100644 index 0000000000..d1b53ae6cc --- /dev/null +++ b/test/jsonb01.test @@ -0,0 +1,49 @@ +# 2023-11-15 +# +# 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. +# +#*********************************************************************** +# Test cases for JSONB +# + +set testdir [file dirname $argv0] +source $testdir/tester.tcl + +do_execsql_test jsonb01-1.1 { + CREATE TABLE t1(x JSON BLOB); + INSERT INTO t1 VALUES(jsonb('{a:5,b:{x:10,y:11},c:[1,2,3,4]}')); +} +foreach {id path res} { + 1 {$.a} {{{"b":{"x":10,"y":11},"c":[1,2,3,4]}}} + 2 {$.b} {{{"a":5,"c":[1,2,3,4]}}} + 3 {$.c} {{{"a":5,"b":{"x":10,"y":11}}}} + 4 {$.d} {{{"a":5,"b":{"x":10,"y":11},"c":[1,2,3,4]}}} + 5 {$.b.x} {{{"a":5,"b":{"y":11},"c":[1,2,3,4]}}} + 6 {$.b.y} {{{"a":5,"b":{"x":10},"c":[1,2,3,4]}}} + 7 {$.c[0]} {{{"a":5,"b":{"x":10,"y":11},"c":[2,3,4]}}} + 8 {$.c[1]} {{{"a":5,"b":{"x":10,"y":11},"c":[1,3,4]}}} + 9 {$.c[2]} {{{"a":5,"b":{"x":10,"y":11},"c":[1,2,4]}}} + 10 {$.c[3]} {{{"a":5,"b":{"x":10,"y":11},"c":[1,2,3]}}} + 11 {$.c[4]} {{{"a":5,"b":{"x":10,"y":11},"c":[1,2,3,4]}}} + 12 {$.c[#]} {{{"a":5,"b":{"x":10,"y":11},"c":[1,2,3,4]}}} + 13 {$.c[#-1]} {{{"a":5,"b":{"x":10,"y":11},"c":[1,2,3]}}} + 14 {$.c[#-2]} {{{"a":5,"b":{"x":10,"y":11},"c":[1,2,4]}}} + 15 {$.c[#-3]} {{{"a":5,"b":{"x":10,"y":11},"c":[1,3,4]}}} + 16 {$.c[#-4]} {{{"a":5,"b":{"x":10,"y":11},"c":[2,3,4]}}} + 17 {$.c[#-5]} {{{"a":5,"b":{"x":10,"y":11},"c":[1,2,3,4]}}} + 18 {$.c[#-6]} {{{"a":5,"b":{"x":10,"y":11},"c":[1,2,3,4]}}} +} { + do_execsql_test jsonb01-1.2.$id.1 { + SELECT json(jsonb_remove(x,$path)) FROM t1; + } $res + do_execsql_test jsonb01-1.2.$id.2 { + SELECT json_remove(x,$path) FROM t1; + } $res +} + +finish_test