danielk1977 | fd9a0a4 | 2005-05-24 12:01:00 +0000 | [diff] [blame] | 1 | /* |
| 2 | ** 2005 May 23 |
| 3 | ** |
| 4 | ** The author disclaims copyright to this source code. In place of |
| 5 | ** a legal notice, here is a blessing: |
| 6 | ** |
| 7 | ** May you do good and not evil. |
| 8 | ** May you find forgiveness for yourself and forgive others. |
| 9 | ** May you share freely, never taking more than you give. |
| 10 | ** |
| 11 | ************************************************************************* |
| 12 | ** |
| 13 | ** This file contains functions used to access the internal hash tables |
| 14 | ** of user defined functions and collation sequences. |
| 15 | ** |
drh | 55ef4d9 | 2005-08-14 01:20:37 +0000 | [diff] [blame^] | 16 | ** $Id: callback.c,v 1.3 2005/08/14 01:20:38 drh Exp $ |
danielk1977 | fd9a0a4 | 2005-05-24 12:01:00 +0000 | [diff] [blame] | 17 | */ |
| 18 | |
| 19 | #include "sqliteInt.h" |
| 20 | |
| 21 | /* |
danielk1977 | 4dade03 | 2005-05-25 10:45:10 +0000 | [diff] [blame] | 22 | ** Invoke the 'collation needed' callback to request a collation sequence |
| 23 | ** in the database text encoding of name zName, length nName. |
| 24 | ** If the collation sequence |
| 25 | */ |
| 26 | static void callCollNeeded(sqlite3 *db, const char *zName, int nName){ |
| 27 | assert( !db->xCollNeeded || !db->xCollNeeded16 ); |
| 28 | if( nName<0 ) nName = strlen(zName); |
| 29 | if( db->xCollNeeded ){ |
| 30 | char *zExternal = sqliteStrNDup(zName, nName); |
| 31 | if( !zExternal ) return; |
| 32 | db->xCollNeeded(db->pCollNeededArg, db, (int)db->enc, zExternal); |
| 33 | sqliteFree(zExternal); |
| 34 | } |
| 35 | #ifndef SQLITE_OMIT_UTF16 |
| 36 | if( db->xCollNeeded16 ){ |
| 37 | char const *zExternal; |
| 38 | sqlite3_value *pTmp = sqlite3GetTransientValue(db); |
| 39 | sqlite3ValueSetStr(pTmp, -1, zName, SQLITE_UTF8, SQLITE_STATIC); |
| 40 | zExternal = sqlite3ValueText(pTmp, SQLITE_UTF16NATIVE); |
| 41 | if( !zExternal ) return; |
| 42 | db->xCollNeeded16(db->pCollNeededArg, db, (int)db->enc, zExternal); |
| 43 | } |
| 44 | #endif |
| 45 | } |
| 46 | |
| 47 | /* |
| 48 | ** This routine is called if the collation factory fails to deliver a |
| 49 | ** collation function in the best encoding but there may be other versions |
| 50 | ** of this collation function (for other text encodings) available. Use one |
| 51 | ** of these instead if they exist. Avoid a UTF-8 <-> UTF-16 conversion if |
| 52 | ** possible. |
| 53 | */ |
| 54 | static int synthCollSeq(sqlite3 *db, CollSeq *pColl){ |
| 55 | CollSeq *pColl2; |
| 56 | char *z = pColl->zName; |
| 57 | int n = strlen(z); |
| 58 | int i; |
| 59 | static const u8 aEnc[] = { SQLITE_UTF16BE, SQLITE_UTF16LE, SQLITE_UTF8 }; |
| 60 | for(i=0; i<3; i++){ |
| 61 | pColl2 = sqlite3FindCollSeq(db, aEnc[i], z, n, 0); |
| 62 | if( pColl2->xCmp!=0 ){ |
| 63 | memcpy(pColl, pColl2, sizeof(CollSeq)); |
| 64 | return SQLITE_OK; |
| 65 | } |
| 66 | } |
| 67 | return SQLITE_ERROR; |
| 68 | } |
| 69 | |
| 70 | /* |
| 71 | ** This function is responsible for invoking the collation factory callback |
| 72 | ** or substituting a collation sequence of a different encoding when the |
| 73 | ** requested collation sequence is not available in the database native |
| 74 | ** encoding. |
| 75 | ** |
| 76 | ** If it is not NULL, then pColl must point to the database native encoding |
| 77 | ** collation sequence with name zName, length nName. |
| 78 | ** |
| 79 | ** The return value is either the collation sequence to be used in database |
| 80 | ** db for collation type name zName, length nName, or NULL, if no collation |
| 81 | ** sequence can be found. |
| 82 | */ |
| 83 | CollSeq *sqlite3GetCollSeq( |
| 84 | sqlite3* db, |
| 85 | CollSeq *pColl, |
| 86 | const char *zName, |
| 87 | int nName |
| 88 | ){ |
| 89 | CollSeq *p; |
| 90 | |
| 91 | p = pColl; |
| 92 | if( !p ){ |
| 93 | p = sqlite3FindCollSeq(db, db->enc, zName, nName, 0); |
| 94 | } |
| 95 | if( !p || !p->xCmp ){ |
| 96 | /* No collation sequence of this type for this encoding is registered. |
| 97 | ** Call the collation factory to see if it can supply us with one. |
| 98 | */ |
| 99 | callCollNeeded(db, zName, nName); |
| 100 | p = sqlite3FindCollSeq(db, db->enc, zName, nName, 0); |
| 101 | } |
| 102 | if( p && !p->xCmp && synthCollSeq(db, p) ){ |
| 103 | p = 0; |
| 104 | } |
| 105 | assert( !p || p->xCmp ); |
| 106 | return p; |
| 107 | } |
| 108 | |
| 109 | /* |
| 110 | ** This routine is called on a collation sequence before it is used to |
| 111 | ** check that it is defined. An undefined collation sequence exists when |
| 112 | ** a database is loaded that contains references to collation sequences |
| 113 | ** that have not been defined by sqlite3_create_collation() etc. |
| 114 | ** |
| 115 | ** If required, this routine calls the 'collation needed' callback to |
| 116 | ** request a definition of the collating sequence. If this doesn't work, |
| 117 | ** an equivalent collating sequence that uses a text encoding different |
| 118 | ** from the main database is substituted, if one is available. |
| 119 | */ |
| 120 | int sqlite3CheckCollSeq(Parse *pParse, CollSeq *pColl){ |
| 121 | if( pColl ){ |
| 122 | const char *zName = pColl->zName; |
| 123 | CollSeq *p = sqlite3GetCollSeq(pParse->db, pColl, zName, -1); |
| 124 | if( !p ){ |
| 125 | if( pParse->nErr==0 ){ |
| 126 | sqlite3ErrorMsg(pParse, "no such collation sequence: %s", zName); |
| 127 | } |
| 128 | pParse->nErr++; |
| 129 | return SQLITE_ERROR; |
| 130 | } |
| 131 | } |
| 132 | return SQLITE_OK; |
| 133 | } |
| 134 | |
| 135 | |
| 136 | |
| 137 | /* |
danielk1977 | fd9a0a4 | 2005-05-24 12:01:00 +0000 | [diff] [blame] | 138 | ** Locate and return an entry from the db.aCollSeq hash table. If the entry |
| 139 | ** specified by zName and nName is not found and parameter 'create' is |
| 140 | ** true, then create a new entry. Otherwise return NULL. |
| 141 | ** |
| 142 | ** Each pointer stored in the sqlite3.aCollSeq hash table contains an |
| 143 | ** array of three CollSeq structures. The first is the collation sequence |
| 144 | ** prefferred for UTF-8, the second UTF-16le, and the third UTF-16be. |
| 145 | ** |
| 146 | ** Stored immediately after the three collation sequences is a copy of |
| 147 | ** the collation sequence name. A pointer to this string is stored in |
| 148 | ** each collation sequence structure. |
| 149 | */ |
drh | 55ef4d9 | 2005-08-14 01:20:37 +0000 | [diff] [blame^] | 150 | static CollSeq *findCollSeqEntry( |
danielk1977 | fd9a0a4 | 2005-05-24 12:01:00 +0000 | [diff] [blame] | 151 | sqlite3 *db, |
| 152 | const char *zName, |
| 153 | int nName, |
| 154 | int create |
| 155 | ){ |
| 156 | CollSeq *pColl; |
| 157 | if( nName<0 ) nName = strlen(zName); |
| 158 | pColl = sqlite3HashFind(&db->aCollSeq, zName, nName); |
| 159 | |
| 160 | if( 0==pColl && create ){ |
| 161 | pColl = sqliteMalloc( 3*sizeof(*pColl) + nName + 1 ); |
| 162 | if( pColl ){ |
| 163 | CollSeq *pDel = 0; |
| 164 | pColl[0].zName = (char*)&pColl[3]; |
| 165 | pColl[0].enc = SQLITE_UTF8; |
| 166 | pColl[1].zName = (char*)&pColl[3]; |
| 167 | pColl[1].enc = SQLITE_UTF16LE; |
| 168 | pColl[2].zName = (char*)&pColl[3]; |
| 169 | pColl[2].enc = SQLITE_UTF16BE; |
| 170 | memcpy(pColl[0].zName, zName, nName); |
| 171 | pColl[0].zName[nName] = 0; |
| 172 | pDel = sqlite3HashInsert(&db->aCollSeq, pColl[0].zName, nName, pColl); |
| 173 | |
| 174 | /* If a malloc() failure occured in sqlite3HashInsert(), it will |
| 175 | ** return the pColl pointer to be deleted (because it wasn't added |
| 176 | ** to the hash table). |
| 177 | */ |
| 178 | assert( !pDel || (sqlite3_malloc_failed && pDel==pColl) ); |
| 179 | sqliteFree(pDel); |
| 180 | } |
| 181 | } |
| 182 | return pColl; |
| 183 | } |
| 184 | |
| 185 | /* |
| 186 | ** Parameter zName points to a UTF-8 encoded string nName bytes long. |
| 187 | ** Return the CollSeq* pointer for the collation sequence named zName |
| 188 | ** for the encoding 'enc' from the database 'db'. |
| 189 | ** |
| 190 | ** If the entry specified is not found and 'create' is true, then create a |
| 191 | ** new entry. Otherwise return NULL. |
| 192 | */ |
| 193 | CollSeq *sqlite3FindCollSeq( |
| 194 | sqlite3 *db, |
| 195 | u8 enc, |
| 196 | const char *zName, |
| 197 | int nName, |
| 198 | int create |
| 199 | ){ |
| 200 | CollSeq *pColl = findCollSeqEntry(db, zName, nName, create); |
| 201 | assert( SQLITE_UTF8==1 && SQLITE_UTF16LE==2 && SQLITE_UTF16BE==3 ); |
| 202 | assert( enc>=SQLITE_UTF8 && enc<=SQLITE_UTF16BE ); |
| 203 | if( pColl ) pColl += enc-1; |
| 204 | return pColl; |
| 205 | } |
| 206 | |
| 207 | /* |
| 208 | ** Locate a user function given a name, a number of arguments and a flag |
| 209 | ** indicating whether the function prefers UTF-16 over UTF-8. Return a |
| 210 | ** pointer to the FuncDef structure that defines that function, or return |
| 211 | ** NULL if the function does not exist. |
| 212 | ** |
| 213 | ** If the createFlag argument is true, then a new (blank) FuncDef |
| 214 | ** structure is created and liked into the "db" structure if a |
| 215 | ** no matching function previously existed. When createFlag is true |
| 216 | ** and the nArg parameter is -1, then only a function that accepts |
| 217 | ** any number of arguments will be returned. |
| 218 | ** |
| 219 | ** If createFlag is false and nArg is -1, then the first valid |
| 220 | ** function found is returned. A function is valid if either xFunc |
| 221 | ** or xStep is non-zero. |
| 222 | ** |
| 223 | ** If createFlag is false, then a function with the required name and |
| 224 | ** number of arguments may be returned even if the eTextRep flag does not |
| 225 | ** match that requested. |
| 226 | */ |
| 227 | FuncDef *sqlite3FindFunction( |
| 228 | sqlite3 *db, /* An open database */ |
| 229 | const char *zName, /* Name of the function. Not null-terminated */ |
| 230 | int nName, /* Number of characters in the name */ |
| 231 | int nArg, /* Number of arguments. -1 means any number */ |
| 232 | u8 enc, /* Preferred text encoding */ |
| 233 | int createFlag /* Create new entry if true and does not otherwise exist */ |
| 234 | ){ |
| 235 | FuncDef *p; /* Iterator variable */ |
| 236 | FuncDef *pFirst; /* First function with this name */ |
| 237 | FuncDef *pBest = 0; /* Best match found so far */ |
| 238 | int bestmatch = 0; |
| 239 | |
| 240 | |
| 241 | assert( enc==SQLITE_UTF8 || enc==SQLITE_UTF16LE || enc==SQLITE_UTF16BE ); |
| 242 | if( nArg<-1 ) nArg = -1; |
| 243 | |
| 244 | pFirst = (FuncDef*)sqlite3HashFind(&db->aFunc, zName, nName); |
| 245 | for(p=pFirst; p; p=p->pNext){ |
| 246 | /* During the search for the best function definition, bestmatch is set |
| 247 | ** as follows to indicate the quality of the match with the definition |
| 248 | ** pointed to by pBest: |
| 249 | ** |
| 250 | ** 0: pBest is NULL. No match has been found. |
| 251 | ** 1: A variable arguments function that prefers UTF-8 when a UTF-16 |
| 252 | ** encoding is requested, or vice versa. |
| 253 | ** 2: A variable arguments function that uses UTF-16BE when UTF-16LE is |
| 254 | ** requested, or vice versa. |
| 255 | ** 3: A variable arguments function using the same text encoding. |
| 256 | ** 4: A function with the exact number of arguments requested that |
| 257 | ** prefers UTF-8 when a UTF-16 encoding is requested, or vice versa. |
| 258 | ** 5: A function with the exact number of arguments requested that |
| 259 | ** prefers UTF-16LE when UTF-16BE is requested, or vice versa. |
| 260 | ** 6: An exact match. |
| 261 | ** |
| 262 | ** A larger value of 'matchqual' indicates a more desirable match. |
| 263 | */ |
| 264 | if( p->nArg==-1 || p->nArg==nArg || nArg==-1 ){ |
| 265 | int match = 1; /* Quality of this match */ |
| 266 | if( p->nArg==nArg || nArg==-1 ){ |
| 267 | match = 4; |
| 268 | } |
| 269 | if( enc==p->iPrefEnc ){ |
| 270 | match += 2; |
| 271 | } |
| 272 | else if( (enc==SQLITE_UTF16LE && p->iPrefEnc==SQLITE_UTF16BE) || |
| 273 | (enc==SQLITE_UTF16BE && p->iPrefEnc==SQLITE_UTF16LE) ){ |
| 274 | match += 1; |
| 275 | } |
| 276 | |
| 277 | if( match>bestmatch ){ |
| 278 | pBest = p; |
| 279 | bestmatch = match; |
| 280 | } |
| 281 | } |
| 282 | } |
| 283 | |
| 284 | /* If the createFlag parameter is true, and the seach did not reveal an |
| 285 | ** exact match for the name, number of arguments and encoding, then add a |
| 286 | ** new entry to the hash table and return it. |
| 287 | */ |
| 288 | if( createFlag && bestmatch<6 && |
drh | 55ef4d9 | 2005-08-14 01:20:37 +0000 | [diff] [blame^] | 289 | (pBest = sqliteMalloc(sizeof(*pBest)+nName)) ){ |
danielk1977 | fd9a0a4 | 2005-05-24 12:01:00 +0000 | [diff] [blame] | 290 | pBest->nArg = nArg; |
| 291 | pBest->pNext = pFirst; |
danielk1977 | fd9a0a4 | 2005-05-24 12:01:00 +0000 | [diff] [blame] | 292 | pBest->iPrefEnc = enc; |
| 293 | memcpy(pBest->zName, zName, nName); |
| 294 | pBest->zName[nName] = 0; |
| 295 | if( pBest==sqlite3HashInsert(&db->aFunc,pBest->zName,nName,(void*)pBest) ){ |
| 296 | sqliteFree(pBest); |
| 297 | return 0; |
| 298 | } |
| 299 | } |
| 300 | |
| 301 | if( pBest && (pBest->xStep || pBest->xFunc || createFlag) ){ |
| 302 | return pBest; |
| 303 | } |
| 304 | return 0; |
| 305 | } |