000001  /*
000002  ** 2005 May 23 
000003  **
000004  ** The author disclaims copyright to this source code.  In place of
000005  ** a legal notice, here is a blessing:
000006  **
000007  **    May you do good and not evil.
000008  **    May you find forgiveness for yourself and forgive others.
000009  **    May you share freely, never taking more than you give.
000010  **
000011  *************************************************************************
000012  **
000013  ** This file contains functions used to access the internal hash tables
000014  ** of user defined functions and collation sequences.
000015  */
000016  
000017  #include "sqliteInt.h"
000018  
000019  /*
000020  ** Invoke the 'collation needed' callback to request a collation sequence
000021  ** in the encoding enc of name zName, length nName.
000022  */
000023  static void callCollNeeded(sqlite3 *db, int enc, const char *zName){
000024    assert( !db->xCollNeeded || !db->xCollNeeded16 );
000025    if( db->xCollNeeded ){
000026      char *zExternal = sqlite3DbStrDup(db, zName);
000027      if( !zExternal ) return;
000028      db->xCollNeeded(db->pCollNeededArg, db, enc, zExternal);
000029      sqlite3DbFree(db, zExternal);
000030    }
000031  #ifndef SQLITE_OMIT_UTF16
000032    if( db->xCollNeeded16 ){
000033      char const *zExternal;
000034      sqlite3_value *pTmp = sqlite3ValueNew(db);
000035      sqlite3ValueSetStr(pTmp, -1, zName, SQLITE_UTF8, SQLITE_STATIC);
000036      zExternal = sqlite3ValueText(pTmp, SQLITE_UTF16NATIVE);
000037      if( zExternal ){
000038        db->xCollNeeded16(db->pCollNeededArg, db, (int)ENC(db), zExternal);
000039      }
000040      sqlite3ValueFree(pTmp);
000041    }
000042  #endif
000043  }
000044  
000045  /*
000046  ** This routine is called if the collation factory fails to deliver a
000047  ** collation function in the best encoding but there may be other versions
000048  ** of this collation function (for other text encodings) available. Use one
000049  ** of these instead if they exist. Avoid a UTF-8 <-> UTF-16 conversion if
000050  ** possible.
000051  */
000052  static int synthCollSeq(sqlite3 *db, CollSeq *pColl){
000053    CollSeq *pColl2;
000054    char *z = pColl->zName;
000055    int i;
000056    static const u8 aEnc[] = { SQLITE_UTF16BE, SQLITE_UTF16LE, SQLITE_UTF8 };
000057    for(i=0; i<3; i++){
000058      pColl2 = sqlite3FindCollSeq(db, aEnc[i], z, 0);
000059      if( pColl2->xCmp!=0 ){
000060        memcpy(pColl, pColl2, sizeof(CollSeq));
000061        pColl->xDel = 0;         /* Do not copy the destructor */
000062        return SQLITE_OK;
000063      }
000064    }
000065    return SQLITE_ERROR;
000066  }
000067  
000068  /*
000069  ** This routine is called on a collation sequence before it is used to
000070  ** check that it is defined. An undefined collation sequence exists when
000071  ** a database is loaded that contains references to collation sequences
000072  ** that have not been defined by sqlite3_create_collation() etc.
000073  **
000074  ** If required, this routine calls the 'collation needed' callback to
000075  ** request a definition of the collating sequence. If this doesn't work, 
000076  ** an equivalent collating sequence that uses a text encoding different
000077  ** from the main database is substituted, if one is available.
000078  */
000079  int sqlite3CheckCollSeq(Parse *pParse, CollSeq *pColl){
000080    if( pColl && pColl->xCmp==0 ){
000081      const char *zName = pColl->zName;
000082      sqlite3 *db = pParse->db;
000083      CollSeq *p = sqlite3GetCollSeq(pParse, ENC(db), pColl, zName);
000084      if( !p ){
000085        return SQLITE_ERROR;
000086      }
000087      assert( p==pColl );
000088    }
000089    return SQLITE_OK;
000090  }
000091  
000092  
000093  
000094  /*
000095  ** Locate and return an entry from the db.aCollSeq hash table. If the entry
000096  ** specified by zName and nName is not found and parameter 'create' is
000097  ** true, then create a new entry. Otherwise return NULL.
000098  **
000099  ** Each pointer stored in the sqlite3.aCollSeq hash table contains an
000100  ** array of three CollSeq structures. The first is the collation sequence
000101  ** preferred for UTF-8, the second UTF-16le, and the third UTF-16be.
000102  **
000103  ** Stored immediately after the three collation sequences is a copy of
000104  ** the collation sequence name. A pointer to this string is stored in
000105  ** each collation sequence structure.
000106  */
000107  static CollSeq *findCollSeqEntry(
000108    sqlite3 *db,          /* Database connection */
000109    const char *zName,    /* Name of the collating sequence */
000110    int create            /* Create a new entry if true */
000111  ){
000112    CollSeq *pColl;
000113    pColl = sqlite3HashFind(&db->aCollSeq, zName);
000114  
000115    if( 0==pColl && create ){
000116      int nName = sqlite3Strlen30(zName) + 1;
000117      pColl = sqlite3DbMallocZero(db, 3*sizeof(*pColl) + nName);
000118      if( pColl ){
000119        CollSeq *pDel = 0;
000120        pColl[0].zName = (char*)&pColl[3];
000121        pColl[0].enc = SQLITE_UTF8;
000122        pColl[1].zName = (char*)&pColl[3];
000123        pColl[1].enc = SQLITE_UTF16LE;
000124        pColl[2].zName = (char*)&pColl[3];
000125        pColl[2].enc = SQLITE_UTF16BE;
000126        memcpy(pColl[0].zName, zName, nName);
000127        pDel = sqlite3HashInsert(&db->aCollSeq, pColl[0].zName, pColl);
000128  
000129        /* If a malloc() failure occurred in sqlite3HashInsert(), it will 
000130        ** return the pColl pointer to be deleted (because it wasn't added
000131        ** to the hash table).
000132        */
000133        assert( pDel==0 || pDel==pColl );
000134        if( pDel!=0 ){
000135          sqlite3OomFault(db);
000136          sqlite3DbFree(db, pDel);
000137          pColl = 0;
000138        }
000139      }
000140    }
000141    return pColl;
000142  }
000143  
000144  /*
000145  ** Parameter zName points to a UTF-8 encoded string nName bytes long.
000146  ** Return the CollSeq* pointer for the collation sequence named zName
000147  ** for the encoding 'enc' from the database 'db'.
000148  **
000149  ** If the entry specified is not found and 'create' is true, then create a
000150  ** new entry.  Otherwise return NULL.
000151  **
000152  ** A separate function sqlite3LocateCollSeq() is a wrapper around
000153  ** this routine.  sqlite3LocateCollSeq() invokes the collation factory
000154  ** if necessary and generates an error message if the collating sequence
000155  ** cannot be found.
000156  **
000157  ** See also: sqlite3LocateCollSeq(), sqlite3GetCollSeq()
000158  */
000159  CollSeq *sqlite3FindCollSeq(
000160    sqlite3 *db,          /* Database connection to search */
000161    u8 enc,               /* Desired text encoding */
000162    const char *zName,    /* Name of the collating sequence.  Might be NULL */
000163    int create            /* True to create CollSeq if doesn't already exist */
000164  ){
000165    CollSeq *pColl;
000166    assert( SQLITE_UTF8==1 && SQLITE_UTF16LE==2 && SQLITE_UTF16BE==3 );
000167    assert( enc>=SQLITE_UTF8 && enc<=SQLITE_UTF16BE );
000168    if( zName ){
000169      pColl = findCollSeqEntry(db, zName, create);
000170      if( pColl ) pColl += enc-1;
000171    }else{
000172      pColl = db->pDfltColl;
000173    }
000174    return pColl;
000175  }
000176  
000177  /*
000178  ** Change the text encoding for a database connection. This means that
000179  ** the pDfltColl must change as well.
000180  */
000181  void sqlite3SetTextEncoding(sqlite3 *db, u8 enc){
000182    assert( enc==SQLITE_UTF8 || enc==SQLITE_UTF16LE || enc==SQLITE_UTF16BE );
000183    db->enc = enc;
000184    /* EVIDENCE-OF: R-08308-17224 The default collating function for all
000185    ** strings is BINARY. 
000186    */
000187    db->pDfltColl = sqlite3FindCollSeq(db, enc, sqlite3StrBINARY, 0);
000188    sqlite3ExpirePreparedStatements(db, 1);
000189  }
000190  
000191  /*
000192  ** This function is responsible for invoking the collation factory callback
000193  ** or substituting a collation sequence of a different encoding when the
000194  ** requested collation sequence is not available in the desired encoding.
000195  ** 
000196  ** If it is not NULL, then pColl must point to the database native encoding 
000197  ** collation sequence with name zName, length nName.
000198  **
000199  ** The return value is either the collation sequence to be used in database
000200  ** db for collation type name zName, length nName, or NULL, if no collation
000201  ** sequence can be found.  If no collation is found, leave an error message.
000202  **
000203  ** See also: sqlite3LocateCollSeq(), sqlite3FindCollSeq()
000204  */
000205  CollSeq *sqlite3GetCollSeq(
000206    Parse *pParse,        /* Parsing context */
000207    u8 enc,               /* The desired encoding for the collating sequence */
000208    CollSeq *pColl,       /* Collating sequence with native encoding, or NULL */
000209    const char *zName     /* Collating sequence name */
000210  ){
000211    CollSeq *p;
000212    sqlite3 *db = pParse->db;
000213  
000214    p = pColl;
000215    if( !p ){
000216      p = sqlite3FindCollSeq(db, enc, zName, 0);
000217    }
000218    if( !p || !p->xCmp ){
000219      /* No collation sequence of this type for this encoding is registered.
000220      ** Call the collation factory to see if it can supply us with one.
000221      */
000222      callCollNeeded(db, enc, zName);
000223      p = sqlite3FindCollSeq(db, enc, zName, 0);
000224    }
000225    if( p && !p->xCmp && synthCollSeq(db, p) ){
000226      p = 0;
000227    }
000228    assert( !p || p->xCmp );
000229    if( p==0 ){
000230      sqlite3ErrorMsg(pParse, "no such collation sequence: %s", zName);
000231      pParse->rc = SQLITE_ERROR_MISSING_COLLSEQ;
000232    }
000233    return p;
000234  }
000235  
000236  /*
000237  ** This function returns the collation sequence for database native text
000238  ** encoding identified by the string zName.
000239  **
000240  ** If the requested collation sequence is not available, or not available
000241  ** in the database native encoding, the collation factory is invoked to
000242  ** request it. If the collation factory does not supply such a sequence,
000243  ** and the sequence is available in another text encoding, then that is
000244  ** returned instead.
000245  **
000246  ** If no versions of the requested collations sequence are available, or
000247  ** another error occurs, NULL is returned and an error message written into
000248  ** pParse.
000249  **
000250  ** This routine is a wrapper around sqlite3FindCollSeq().  This routine
000251  ** invokes the collation factory if the named collation cannot be found
000252  ** and generates an error message.
000253  **
000254  ** See also: sqlite3FindCollSeq(), sqlite3GetCollSeq()
000255  */
000256  CollSeq *sqlite3LocateCollSeq(Parse *pParse, const char *zName){
000257    sqlite3 *db = pParse->db;
000258    u8 enc = ENC(db);
000259    u8 initbusy = db->init.busy;
000260    CollSeq *pColl;
000261  
000262    pColl = sqlite3FindCollSeq(db, enc, zName, initbusy);
000263    if( !initbusy && (!pColl || !pColl->xCmp) ){
000264      pColl = sqlite3GetCollSeq(pParse, enc, pColl, zName);
000265    }
000266  
000267    return pColl;
000268  }
000269  
000270  /* During the search for the best function definition, this procedure
000271  ** is called to test how well the function passed as the first argument
000272  ** matches the request for a function with nArg arguments in a system
000273  ** that uses encoding enc. The value returned indicates how well the
000274  ** request is matched. A higher value indicates a better match.
000275  **
000276  ** If nArg is -1 that means to only return a match (non-zero) if p->nArg
000277  ** is also -1.  In other words, we are searching for a function that
000278  ** takes a variable number of arguments.
000279  **
000280  ** If nArg is -2 that means that we are searching for any function 
000281  ** regardless of the number of arguments it uses, so return a positive
000282  ** match score for any
000283  **
000284  ** The returned value is always between 0 and 6, as follows:
000285  **
000286  ** 0: Not a match.
000287  ** 1: UTF8/16 conversion required and function takes any number of arguments.
000288  ** 2: UTF16 byte order change required and function takes any number of args.
000289  ** 3: encoding matches and function takes any number of arguments
000290  ** 4: UTF8/16 conversion required - argument count matches exactly
000291  ** 5: UTF16 byte order conversion required - argument count matches exactly
000292  ** 6: Perfect match:  encoding and argument count match exactly.
000293  **
000294  ** If nArg==(-2) then any function with a non-null xSFunc is
000295  ** a perfect match and any function with xSFunc NULL is
000296  ** a non-match.
000297  */
000298  #define FUNC_PERFECT_MATCH 6  /* The score for a perfect match */
000299  static int matchQuality(
000300    FuncDef *p,     /* The function we are evaluating for match quality */
000301    int nArg,       /* Desired number of arguments.  (-1)==any */
000302    u8 enc          /* Desired text encoding */
000303  ){
000304    int match;
000305    assert( p->nArg>=-1 );
000306  
000307    /* Wrong number of arguments means "no match" */
000308    if( p->nArg!=nArg ){
000309      if( nArg==(-2) ) return (p->xSFunc==0) ? 0 : FUNC_PERFECT_MATCH;
000310      if( p->nArg>=0 ) return 0;
000311    }
000312  
000313    /* Give a better score to a function with a specific number of arguments
000314    ** than to function that accepts any number of arguments. */
000315    if( p->nArg==nArg ){
000316      match = 4;
000317    }else{
000318      match = 1;
000319    }
000320  
000321    /* Bonus points if the text encoding matches */
000322    if( enc==(p->funcFlags & SQLITE_FUNC_ENCMASK) ){
000323      match += 2;  /* Exact encoding match */
000324    }else if( (enc & p->funcFlags & 2)!=0 ){
000325      match += 1;  /* Both are UTF16, but with different byte orders */
000326    }
000327  
000328    return match;
000329  }
000330  
000331  /*
000332  ** Search a FuncDefHash for a function with the given name.  Return
000333  ** a pointer to the matching FuncDef if found, or 0 if there is no match.
000334  */
000335  FuncDef *sqlite3FunctionSearch(
000336    int h,               /* Hash of the name */
000337    const char *zFunc    /* Name of function */
000338  ){
000339    FuncDef *p;
000340    for(p=sqlite3BuiltinFunctions.a[h]; p; p=p->u.pHash){
000341      assert( p->funcFlags & SQLITE_FUNC_BUILTIN );
000342      if( sqlite3StrICmp(p->zName, zFunc)==0 ){
000343        return p;
000344      }
000345    }
000346    return 0;
000347  }
000348  
000349  /*
000350  ** Insert a new FuncDef into a FuncDefHash hash table.
000351  */
000352  void sqlite3InsertBuiltinFuncs(
000353    FuncDef *aDef,      /* List of global functions to be inserted */
000354    int nDef            /* Length of the apDef[] list */
000355  ){
000356    int i;
000357    for(i=0; i<nDef; i++){
000358      FuncDef *pOther;
000359      const char *zName = aDef[i].zName;
000360      int nName = sqlite3Strlen30(zName);
000361      int h = SQLITE_FUNC_HASH(zName[0], nName);
000362      assert( aDef[i].funcFlags & SQLITE_FUNC_BUILTIN );
000363      pOther = sqlite3FunctionSearch(h, zName);
000364      if( pOther ){
000365        assert( pOther!=&aDef[i] && pOther->pNext!=&aDef[i] );
000366        aDef[i].pNext = pOther->pNext;
000367        pOther->pNext = &aDef[i];
000368      }else{
000369        aDef[i].pNext = 0;
000370        aDef[i].u.pHash = sqlite3BuiltinFunctions.a[h];
000371        sqlite3BuiltinFunctions.a[h] = &aDef[i];
000372      }
000373    }
000374  }
000375    
000376    
000377  
000378  /*
000379  ** Locate a user function given a name, a number of arguments and a flag
000380  ** indicating whether the function prefers UTF-16 over UTF-8.  Return a
000381  ** pointer to the FuncDef structure that defines that function, or return
000382  ** NULL if the function does not exist.
000383  **
000384  ** If the createFlag argument is true, then a new (blank) FuncDef
000385  ** structure is created and liked into the "db" structure if a
000386  ** no matching function previously existed.
000387  **
000388  ** If nArg is -2, then the first valid function found is returned.  A
000389  ** function is valid if xSFunc is non-zero.  The nArg==(-2)
000390  ** case is used to see if zName is a valid function name for some number
000391  ** of arguments.  If nArg is -2, then createFlag must be 0.
000392  **
000393  ** If createFlag is false, then a function with the required name and
000394  ** number of arguments may be returned even if the eTextRep flag does not
000395  ** match that requested.
000396  */
000397  FuncDef *sqlite3FindFunction(
000398    sqlite3 *db,       /* An open database */
000399    const char *zName, /* Name of the function.  zero-terminated */
000400    int nArg,          /* Number of arguments.  -1 means any number */
000401    u8 enc,            /* Preferred text encoding */
000402    u8 createFlag      /* Create new entry if true and does not otherwise exist */
000403  ){
000404    FuncDef *p;         /* Iterator variable */
000405    FuncDef *pBest = 0; /* Best match found so far */
000406    int bestScore = 0;  /* Score of best match */
000407    int h;              /* Hash value */
000408    int nName;          /* Length of the name */
000409  
000410    assert( nArg>=(-2) );
000411    assert( nArg>=(-1) || createFlag==0 );
000412    nName = sqlite3Strlen30(zName);
000413  
000414    /* First search for a match amongst the application-defined functions.
000415    */
000416    p = (FuncDef*)sqlite3HashFind(&db->aFunc, zName);
000417    while( p ){
000418      int score = matchQuality(p, nArg, enc);
000419      if( score>bestScore ){
000420        pBest = p;
000421        bestScore = score;
000422      }
000423      p = p->pNext;
000424    }
000425  
000426    /* If no match is found, search the built-in functions.
000427    **
000428    ** If the DBFLAG_PreferBuiltin flag is set, then search the built-in
000429    ** functions even if a prior app-defined function was found.  And give
000430    ** priority to built-in functions.
000431    **
000432    ** Except, if createFlag is true, that means that we are trying to
000433    ** install a new function.  Whatever FuncDef structure is returned it will
000434    ** have fields overwritten with new information appropriate for the
000435    ** new function.  But the FuncDefs for built-in functions are read-only.
000436    ** So we must not search for built-ins when creating a new function.
000437    */ 
000438    if( !createFlag && (pBest==0 || (db->mDbFlags & DBFLAG_PreferBuiltin)!=0) ){
000439      bestScore = 0;
000440      h = SQLITE_FUNC_HASH(sqlite3UpperToLower[(u8)zName[0]], nName);
000441      p = sqlite3FunctionSearch(h, zName);
000442      while( p ){
000443        int score = matchQuality(p, nArg, enc);
000444        if( score>bestScore ){
000445          pBest = p;
000446          bestScore = score;
000447        }
000448        p = p->pNext;
000449      }
000450    }
000451  
000452    /* If the createFlag parameter is true and the search did not reveal an
000453    ** exact match for the name, number of arguments and encoding, then add a
000454    ** new entry to the hash table and return it.
000455    */
000456    if( createFlag && bestScore<FUNC_PERFECT_MATCH && 
000457        (pBest = sqlite3DbMallocZero(db, sizeof(*pBest)+nName+1))!=0 ){
000458      FuncDef *pOther;
000459      u8 *z;
000460      pBest->zName = (const char*)&pBest[1];
000461      pBest->nArg = (u16)nArg;
000462      pBest->funcFlags = enc;
000463      memcpy((char*)&pBest[1], zName, nName+1);
000464      for(z=(u8*)pBest->zName; *z; z++) *z = sqlite3UpperToLower[*z];
000465      pOther = (FuncDef*)sqlite3HashInsert(&db->aFunc, pBest->zName, pBest);
000466      if( pOther==pBest ){
000467        sqlite3DbFree(db, pBest);
000468        sqlite3OomFault(db);
000469        return 0;
000470      }else{
000471        pBest->pNext = pOther;
000472      }
000473    }
000474  
000475    if( pBest && (pBest->xSFunc || createFlag) ){
000476      return pBest;
000477    }
000478    return 0;
000479  }
000480  
000481  /*
000482  ** Free all resources held by the schema structure. The void* argument points
000483  ** at a Schema struct. This function does not call sqlite3DbFree(db, ) on the 
000484  ** pointer itself, it just cleans up subsidiary resources (i.e. the contents
000485  ** of the schema hash tables).
000486  **
000487  ** The Schema.cache_size variable is not cleared.
000488  */
000489  void sqlite3SchemaClear(void *p){
000490    Hash temp1;
000491    Hash temp2;
000492    HashElem *pElem;
000493    Schema *pSchema = (Schema *)p;
000494    sqlite3 xdb;
000495  
000496    memset(&xdb, 0, sizeof(xdb));
000497    temp1 = pSchema->tblHash;
000498    temp2 = pSchema->trigHash;
000499    sqlite3HashInit(&pSchema->trigHash);
000500    sqlite3HashClear(&pSchema->idxHash);
000501    for(pElem=sqliteHashFirst(&temp2); pElem; pElem=sqliteHashNext(pElem)){
000502      sqlite3DeleteTrigger(&xdb, (Trigger*)sqliteHashData(pElem));
000503    }
000504    sqlite3HashClear(&temp2);
000505    sqlite3HashInit(&pSchema->tblHash);
000506    for(pElem=sqliteHashFirst(&temp1); pElem; pElem=sqliteHashNext(pElem)){
000507      Table *pTab = sqliteHashData(pElem);
000508      sqlite3DeleteTable(&xdb, pTab);
000509    }
000510    sqlite3HashClear(&temp1);
000511    sqlite3HashClear(&pSchema->fkeyHash);
000512    pSchema->pSeqTab = 0;
000513    if( pSchema->schemaFlags & DB_SchemaLoaded ){
000514      pSchema->iGeneration++;
000515    }
000516    pSchema->schemaFlags &= ~(DB_SchemaLoaded|DB_ResetWanted);
000517  }
000518  
000519  /*
000520  ** Find and return the schema associated with a BTree.  Create
000521  ** a new one if necessary.
000522  */
000523  Schema *sqlite3SchemaGet(sqlite3 *db, Btree *pBt){
000524    Schema * p;
000525    if( pBt ){
000526      p = (Schema *)sqlite3BtreeSchema(pBt, sizeof(Schema), sqlite3SchemaClear);
000527    }else{
000528      p = (Schema *)sqlite3DbMallocZero(0, sizeof(Schema));
000529    }
000530    if( !p ){
000531      sqlite3OomFault(db);
000532    }else if ( 0==p->file_format ){
000533      sqlite3HashInit(&p->tblHash);
000534      sqlite3HashInit(&p->idxHash);
000535      sqlite3HashInit(&p->trigHash);
000536      sqlite3HashInit(&p->fkeyHash);
000537      p->enc = SQLITE_UTF8;
000538    }
000539    return p;
000540  }