1 /* 2 ** 2007 May 6 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 ** $Id: icu.c,v 1.7 2007/12/13 21:54:11 drh Exp $ 13 ** 14 ** This file implements an integration between the ICU library 15 ** ("International Components for Unicode", an open-source library 16 ** for handling unicode data) and SQLite. The integration uses 17 ** ICU to provide the following to SQLite: 18 ** 19 ** * An implementation of the SQL regexp() function (and hence REGEXP 20 ** operator) using the ICU uregex_XX() APIs. 21 ** 22 ** * Implementations of the SQL scalar upper() and lower() functions 23 ** for case mapping. 24 ** 25 ** * Integration of ICU and SQLite collation seqences. 26 ** 27 ** * An implementation of the LIKE operator that uses ICU to 28 ** provide case-independent matching. 29 */ 30 31 #if !defined(SQLITE_CORE) || defined(SQLITE_ENABLE_ICU) 32 33 /* Include ICU headers */ 34 #include <unicode/utypes.h> 35 #include <unicode/uregex.h> 36 #include <unicode/ustring.h> 37 #include <unicode/ucol.h> 38 39 #include <assert.h> 40 41 #ifndef SQLITE_CORE 42 #include "sqlite3ext.h" 43 SQLITE_EXTENSION_INIT1 44 #else 45 #include "sqlite3.h" 46 #endif 47 48 /* 49 ** Maximum length (in bytes) of the pattern in a LIKE or GLOB 50 ** operator. 51 */ 52 #ifndef SQLITE_MAX_LIKE_PATTERN_LENGTH 53 # define SQLITE_MAX_LIKE_PATTERN_LENGTH 50000 54 #endif 55 56 /* 57 ** Version of sqlite3_free() that is always a function, never a macro. 58 */ 59 static void xFree(void *p){ 60 sqlite3_free(p); 61 } 62 63 /* 64 ** Compare two UTF-8 strings for equality where the first string is 65 ** a "LIKE" expression. Return true (1) if they are the same and 66 ** false (0) if they are different. 67 */ 68 static int icuLikeCompare( 69 const uint8_t *zPattern, /* LIKE pattern */ 70 const uint8_t *zString, /* The UTF-8 string to compare against */ 71 const UChar32 uEsc /* The escape character */ 72 ){ 73 static const int MATCH_ONE = (UChar32)'_'; 74 static const int MATCH_ALL = (UChar32)'%'; 75 76 int iPattern = 0; /* Current byte index in zPattern */ 77 int iString = 0; /* Current byte index in zString */ 78 79 int prevEscape = 0; /* True if the previous character was uEsc */ 80 81 while( zPattern[iPattern]!=0 ){ 82 83 /* Read (and consume) the next character from the input pattern. */ 84 UChar32 uPattern; 85 U8_NEXT_UNSAFE(zPattern, iPattern, uPattern); 86 assert(uPattern!=0); 87 88 /* There are now 4 possibilities: 89 ** 90 ** 1. uPattern is an unescaped match-all character "%", 91 ** 2. uPattern is an unescaped match-one character "_", 92 ** 3. uPattern is an unescaped escape character, or 93 ** 4. uPattern is to be handled as an ordinary character 94 */ 95 if( !prevEscape && uPattern==MATCH_ALL ){ 96 /* Case 1. */ 97 uint8_t c; 98 99 /* Skip any MATCH_ALL or MATCH_ONE characters that follow a 100 ** MATCH_ALL. For each MATCH_ONE, skip one character in the 101 ** test string. 102 */ 103 while( (c=zPattern[iPattern]) == MATCH_ALL || c == MATCH_ONE ){ 104 if( c==MATCH_ONE ){ 105 if( zString[iString]==0 ) return 0; 106 U8_FWD_1_UNSAFE(zString, iString); 107 } 108 iPattern++; 109 } 110 111 if( zPattern[iPattern]==0 ) return 1; 112 113 while( zString[iString] ){ 114 if( icuLikeCompare(&zPattern[iPattern], &zString[iString], uEsc) ){ 115 return 1; 116 } 117 U8_FWD_1_UNSAFE(zString, iString); 118 } 119 return 0; 120 121 }else if( !prevEscape && uPattern==MATCH_ONE ){ 122 /* Case 2. */ 123 if( zString[iString]==0 ) return 0; 124 U8_FWD_1_UNSAFE(zString, iString); 125 126 }else if( !prevEscape && uPattern==uEsc){ 127 /* Case 3. */ 128 prevEscape = 1; 129 130 }else{ 131 /* Case 4. */ 132 UChar32 uString; 133 U8_NEXT_UNSAFE(zString, iString, uString); 134 uString = u_foldCase(uString, U_FOLD_CASE_DEFAULT); 135 uPattern = u_foldCase(uPattern, U_FOLD_CASE_DEFAULT); 136 if( uString!=uPattern ){ 137 return 0; 138 } 139 prevEscape = 0; 140 } 141 } 142 143 return zString[iString]==0; 144 } 145 146 /* 147 ** Implementation of the like() SQL function. This function implements 148 ** the build-in LIKE operator. The first argument to the function is the 149 ** pattern and the second argument is the string. So, the SQL statements: 150 ** 151 ** A LIKE B 152 ** 153 ** is implemented as like(B, A). If there is an escape character E, 154 ** 155 ** A LIKE B ESCAPE E 156 ** 157 ** is mapped to like(B, A, E). 158 */ 159 static void icuLikeFunc( 160 sqlite3_context *context, 161 int argc, 162 sqlite3_value **argv 163 ){ 164 const unsigned char *zA = sqlite3_value_text(argv[0]); 165 const unsigned char *zB = sqlite3_value_text(argv[1]); 166 UChar32 uEsc = 0; 167 168 /* Limit the length of the LIKE or GLOB pattern to avoid problems 169 ** of deep recursion and N*N behavior in patternCompare(). 170 */ 171 if( sqlite3_value_bytes(argv[0])>SQLITE_MAX_LIKE_PATTERN_LENGTH ){ 172 sqlite3_result_error(context, "LIKE or GLOB pattern too complex", -1); 173 return; 174 } 175 176 177 if( argc==3 ){ 178 /* The escape character string must consist of a single UTF-8 character. 179 ** Otherwise, return an error. 180 */ 181 int nE= sqlite3_value_bytes(argv[2]); 182 const unsigned char *zE = sqlite3_value_text(argv[2]); 183 int i = 0; 184 if( zE==0 ) return; 185 U8_NEXT(zE, i, nE, uEsc); 186 if( i!=nE){ 187 sqlite3_result_error(context, 188 "ESCAPE expression must be a single character", -1); 189 return; 190 } 191 } 192 193 if( zA && zB ){ 194 sqlite3_result_int(context, icuLikeCompare(zA, zB, uEsc)); 195 } 196 } 197 198 /* 199 ** This function is called when an ICU function called from within 200 ** the implementation of an SQL scalar function returns an error. 201 ** 202 ** The scalar function context passed as the first argument is 203 ** loaded with an error message based on the following two args. 204 */ 205 static void icuFunctionError( 206 sqlite3_context *pCtx, /* SQLite scalar function context */ 207 const char *zName, /* Name of ICU function that failed */ 208 UErrorCode e /* Error code returned by ICU function */ 209 ){ 210 char zBuf[128]; 211 sqlite3_snprintf(128, zBuf, "ICU error: %s(): %s", zName, u_errorName(e)); 212 zBuf[127] = '\0'; 213 sqlite3_result_error(pCtx, zBuf, -1); 214 } 215 216 /* 217 ** Function to delete compiled regexp objects. Registered as 218 ** a destructor function with sqlite3_set_auxdata(). 219 */ 220 static void icuRegexpDelete(void *p){ 221 URegularExpression *pExpr = (URegularExpression *)p; 222 uregex_close(pExpr); 223 } 224 225 /* 226 ** Implementation of SQLite REGEXP operator. This scalar function takes 227 ** two arguments. The first is a regular expression pattern to compile 228 ** the second is a string to match against that pattern. If either 229 ** argument is an SQL NULL, then NULL Is returned. Otherwise, the result 230 ** is 1 if the string matches the pattern, or 0 otherwise. 231 ** 232 ** SQLite maps the regexp() function to the regexp() operator such 233 ** that the following two are equivalent: 234 ** 235 ** zString REGEXP zPattern 236 ** regexp(zPattern, zString) 237 ** 238 ** Uses the following ICU regexp APIs: 239 ** 240 ** uregex_open() 241 ** uregex_matches() 242 ** uregex_close() 243 */ 244 static void icuRegexpFunc(sqlite3_context *p, int nArg, sqlite3_value **apArg){ 245 UErrorCode status = U_ZERO_ERROR; 246 URegularExpression *pExpr; 247 UBool res; 248 const UChar *zString = sqlite3_value_text16(apArg[1]); 249 250 (void)nArg; /* Unused parameter */ 251 252 /* If the left hand side of the regexp operator is NULL, 253 ** then the result is also NULL. 254 */ 255 if( !zString ){ 256 return; 257 } 258 259 pExpr = sqlite3_get_auxdata(p, 0); 260 if( !pExpr ){ 261 const UChar *zPattern = sqlite3_value_text16(apArg[0]); 262 if( !zPattern ){ 263 return; 264 } 265 pExpr = uregex_open(zPattern, -1, 0, 0, &status); 266 267 if( U_SUCCESS(status) ){ 268 sqlite3_set_auxdata(p, 0, pExpr, icuRegexpDelete); 269 }else{ 270 assert(!pExpr); 271 icuFunctionError(p, "uregex_open", status); 272 return; 273 } 274 } 275 276 /* Configure the text that the regular expression operates on. */ 277 uregex_setText(pExpr, zString, -1, &status); 278 if( !U_SUCCESS(status) ){ 279 icuFunctionError(p, "uregex_setText", status); 280 return; 281 } 282 283 /* Attempt the match */ 284 res = uregex_matches(pExpr, 0, &status); 285 if( !U_SUCCESS(status) ){ 286 icuFunctionError(p, "uregex_matches", status); 287 return; 288 } 289 290 /* Set the text that the regular expression operates on to a NULL 291 ** pointer. This is not really necessary, but it is tidier than 292 ** leaving the regular expression object configured with an invalid 293 ** pointer after this function returns. 294 */ 295 uregex_setText(pExpr, 0, 0, &status); 296 297 /* Return 1 or 0. */ 298 sqlite3_result_int(p, res ? 1 : 0); 299 } 300 301 /* 302 ** Implementations of scalar functions for case mapping - upper() and 303 ** lower(). Function upper() converts its input to upper-case (ABC). 304 ** Function lower() converts to lower-case (abc). 305 ** 306 ** ICU provides two types of case mapping, "general" case mapping and 307 ** "language specific". Refer to ICU documentation for the differences 308 ** between the two. 309 ** 310 ** To utilise "general" case mapping, the upper() or lower() scalar 311 ** functions are invoked with one argument: 312 ** 313 ** upper('ABC') -> 'abc' 314 ** lower('abc') -> 'ABC' 315 ** 316 ** To access ICU "language specific" case mapping, upper() or lower() 317 ** should be invoked with two arguments. The second argument is the name 318 ** of the locale to use. Passing an empty string ("") or SQL NULL value 319 ** as the second argument is the same as invoking the 1 argument version 320 ** of upper() or lower(). 321 ** 322 ** lower('I', 'en_us') -> 'i' 323 ** lower('I', 'tr_tr') -> '' (small dotless i) 324 ** 325 ** http://www.icu-project.org/userguide/posix.html#case_mappings 326 */ 327 static void icuCaseFunc16(sqlite3_context *p, int nArg, sqlite3_value **apArg){ 328 const UChar *zInput; 329 UChar *zOutput; 330 int nInput; 331 int nOutput; 332 333 UErrorCode status = U_ZERO_ERROR; 334 const char *zLocale = 0; 335 336 assert(nArg==1 || nArg==2); 337 if( nArg==2 ){ 338 zLocale = (const char *)sqlite3_value_text(apArg[1]); 339 } 340 341 zInput = sqlite3_value_text16(apArg[0]); 342 if( !zInput ){ 343 return; 344 } 345 nInput = sqlite3_value_bytes16(apArg[0]); 346 347 nOutput = nInput * 2 + 2; 348 zOutput = sqlite3_malloc(nOutput); 349 if( !zOutput ){ 350 return; 351 } 352 353 if( sqlite3_user_data(p) ){ 354 u_strToUpper(zOutput, nOutput/2, zInput, nInput/2, zLocale, &status); 355 }else{ 356 u_strToLower(zOutput, nOutput/2, zInput, nInput/2, zLocale, &status); 357 } 358 359 if( !U_SUCCESS(status) ){ 360 icuFunctionError(p, "u_strToLower()/u_strToUpper", status); 361 return; 362 } 363 364 sqlite3_result_text16(p, zOutput, -1, xFree); 365 } 366 367 /* 368 ** Collation sequence destructor function. The pCtx argument points to 369 ** a UCollator structure previously allocated using ucol_open(). 370 */ 371 static void icuCollationDel(void *pCtx){ 372 UCollator *p = (UCollator *)pCtx; 373 ucol_close(p); 374 } 375 376 /* 377 ** Collation sequence comparison function. The pCtx argument points to 378 ** a UCollator structure previously allocated using ucol_open(). 379 */ 380 static int icuCollationColl( 381 void *pCtx, 382 int nLeft, 383 const void *zLeft, 384 int nRight, 385 const void *zRight 386 ){ 387 UCollationResult res; 388 UCollator *p = (UCollator *)pCtx; 389 res = ucol_strcoll(p, (UChar *)zLeft, nLeft/2, (UChar *)zRight, nRight/2); 390 switch( res ){ 391 case UCOL_LESS: return -1; 392 case UCOL_GREATER: return +1; 393 case UCOL_EQUAL: return 0; 394 } 395 assert(!"Unexpected return value from ucol_strcoll()"); 396 return 0; 397 } 398 399 /* 400 ** Implementation of the scalar function icu_load_collation(). 401 ** 402 ** This scalar function is used to add ICU collation based collation 403 ** types to an SQLite database connection. It is intended to be called 404 ** as follows: 405 ** 406 ** SELECT icu_load_collation(<locale>, <collation-name>); 407 ** 408 ** Where <locale> is a string containing an ICU locale identifier (i.e. 409 ** "en_AU", "tr_TR" etc.) and <collation-name> is the name of the 410 ** collation sequence to create. 411 */ 412 static void icuLoadCollation( 413 sqlite3_context *p, 414 int nArg, 415 sqlite3_value **apArg 416 ){ 417 sqlite3 *db = (sqlite3 *)sqlite3_user_data(p); 418 UErrorCode status = U_ZERO_ERROR; 419 const char *zLocale; /* Locale identifier - (eg. "jp_JP") */ 420 const char *zName; /* SQL Collation sequence name (eg. "japanese") */ 421 UCollator *pUCollator; /* ICU library collation object */ 422 int rc; /* Return code from sqlite3_create_collation_x() */ 423 424 assert(nArg==2); 425 zLocale = (const char *)sqlite3_value_text(apArg[0]); 426 zName = (const char *)sqlite3_value_text(apArg[1]); 427 428 if( !zLocale || !zName ){ 429 return; 430 } 431 432 pUCollator = ucol_open(zLocale, &status); 433 if( !U_SUCCESS(status) ){ 434 icuFunctionError(p, "ucol_open", status); 435 return; 436 } 437 assert(p); 438 439 rc = sqlite3_create_collation_v2(db, zName, SQLITE_UTF16, (void *)pUCollator, 440 icuCollationColl, icuCollationDel 441 ); 442 if( rc!=SQLITE_OK ){ 443 ucol_close(pUCollator); 444 sqlite3_result_error(p, "Error registering collation function", -1); 445 } 446 } 447 448 /* 449 ** Register the ICU extension functions with database db. 450 */ 451 int sqlite3IcuInit(sqlite3 *db){ 452 struct IcuScalar { 453 const char *zName; /* Function name */ 454 int nArg; /* Number of arguments */ 455 int enc; /* Optimal text encoding */ 456 void *pContext; /* sqlite3_user_data() context */ 457 void (*xFunc)(sqlite3_context*,int,sqlite3_value**); 458 } scalars[] = { 459 {"regexp", 2, SQLITE_ANY, 0, icuRegexpFunc}, 460 461 {"lower", 1, SQLITE_UTF16, 0, icuCaseFunc16}, 462 {"lower", 2, SQLITE_UTF16, 0, icuCaseFunc16}, 463 {"upper", 1, SQLITE_UTF16, (void*)1, icuCaseFunc16}, 464 {"upper", 2, SQLITE_UTF16, (void*)1, icuCaseFunc16}, 465 466 {"lower", 1, SQLITE_UTF8, 0, icuCaseFunc16}, 467 {"lower", 2, SQLITE_UTF8, 0, icuCaseFunc16}, 468 {"upper", 1, SQLITE_UTF8, (void*)1, icuCaseFunc16}, 469 {"upper", 2, SQLITE_UTF8, (void*)1, icuCaseFunc16}, 470 471 {"like", 2, SQLITE_UTF8, 0, icuLikeFunc}, 472 {"like", 3, SQLITE_UTF8, 0, icuLikeFunc}, 473 474 {"icu_load_collation", 2, SQLITE_UTF8, (void*)db, icuLoadCollation}, 475 }; 476 477 int rc = SQLITE_OK; 478 int i; 479 480 for(i=0; rc==SQLITE_OK && i<(int)(sizeof(scalars)/sizeof(scalars[0])); i++){ 481 struct IcuScalar *p = &scalars[i]; 482 rc = sqlite3_create_function( 483 db, p->zName, p->nArg, p->enc, p->pContext, p->xFunc, 0, 0 484 ); 485 } 486 487 return rc; 488 } 489 490 #if !SQLITE_CORE 491 int sqlite3_extension_init( 492 sqlite3 *db, 493 char **pzErrMsg, 494 const sqlite3_api_routines *pApi 495 ){ 496 SQLITE_EXTENSION_INIT2(pApi) 497 return sqlite3IcuInit(db); 498 } 499 #endif 500 501 #endif 502