sl@0: /*
sl@0: ** 2007 May 6
sl@0: **
sl@0: ** The author disclaims copyright to this source code.  In place of
sl@0: ** a legal notice, here is a blessing:
sl@0: **
sl@0: **    May you do good and not evil.
sl@0: **    May you find forgiveness for yourself and forgive others.
sl@0: **    May you share freely, never taking more than you give.
sl@0: **
sl@0: *************************************************************************
sl@0: ** $Id: icu.c,v 1.7 2007/12/13 21:54:11 drh Exp $
sl@0: **
sl@0: ** This file implements an integration between the ICU library 
sl@0: ** ("International Components for Unicode", an open-source library 
sl@0: ** for handling unicode data) and SQLite. The integration uses 
sl@0: ** ICU to provide the following to SQLite:
sl@0: **
sl@0: **   * An implementation of the SQL regexp() function (and hence REGEXP
sl@0: **     operator) using the ICU uregex_XX() APIs.
sl@0: **
sl@0: **   * Implementations of the SQL scalar upper() and lower() functions
sl@0: **     for case mapping.
sl@0: **
sl@0: **   * Integration of ICU and SQLite collation seqences.
sl@0: **
sl@0: **   * An implementation of the LIKE operator that uses ICU to 
sl@0: **     provide case-independent matching.
sl@0: */
sl@0: 
sl@0: #if !defined(SQLITE_CORE) || defined(SQLITE_ENABLE_ICU)
sl@0: 
sl@0: /* Include ICU headers */
sl@0: #include <unicode/utypes.h>
sl@0: #include <unicode/uregex.h>
sl@0: #include <unicode/ustring.h>
sl@0: #include <unicode/ucol.h>
sl@0: 
sl@0: #include <assert.h>
sl@0: 
sl@0: #ifndef SQLITE_CORE
sl@0:   #include "sqlite3ext.h"
sl@0:   SQLITE_EXTENSION_INIT1
sl@0: #else
sl@0:   #include "sqlite3.h"
sl@0: #endif
sl@0: 
sl@0: /*
sl@0: ** Maximum length (in bytes) of the pattern in a LIKE or GLOB
sl@0: ** operator.
sl@0: */
sl@0: #ifndef SQLITE_MAX_LIKE_PATTERN_LENGTH
sl@0: # define SQLITE_MAX_LIKE_PATTERN_LENGTH 50000
sl@0: #endif
sl@0: 
sl@0: /*
sl@0: ** Version of sqlite3_free() that is always a function, never a macro.
sl@0: */
sl@0: static void xFree(void *p){
sl@0:   sqlite3_free(p);
sl@0: }
sl@0: 
sl@0: /*
sl@0: ** Compare two UTF-8 strings for equality where the first string is
sl@0: ** a "LIKE" expression. Return true (1) if they are the same and 
sl@0: ** false (0) if they are different.
sl@0: */
sl@0: static int icuLikeCompare(
sl@0:   const uint8_t *zPattern,   /* LIKE pattern */
sl@0:   const uint8_t *zString,    /* The UTF-8 string to compare against */
sl@0:   const UChar32 uEsc         /* The escape character */
sl@0: ){
sl@0:   static const int MATCH_ONE = (UChar32)'_';
sl@0:   static const int MATCH_ALL = (UChar32)'%';
sl@0: 
sl@0:   int iPattern = 0;       /* Current byte index in zPattern */
sl@0:   int iString = 0;        /* Current byte index in zString */
sl@0: 
sl@0:   int prevEscape = 0;     /* True if the previous character was uEsc */
sl@0: 
sl@0:   while( zPattern[iPattern]!=0 ){
sl@0: 
sl@0:     /* Read (and consume) the next character from the input pattern. */
sl@0:     UChar32 uPattern;
sl@0:     U8_NEXT_UNSAFE(zPattern, iPattern, uPattern);
sl@0:     assert(uPattern!=0);
sl@0: 
sl@0:     /* There are now 4 possibilities:
sl@0:     **
sl@0:     **     1. uPattern is an unescaped match-all character "%",
sl@0:     **     2. uPattern is an unescaped match-one character "_",
sl@0:     **     3. uPattern is an unescaped escape character, or
sl@0:     **     4. uPattern is to be handled as an ordinary character
sl@0:     */
sl@0:     if( !prevEscape && uPattern==MATCH_ALL ){
sl@0:       /* Case 1. */
sl@0:       uint8_t c;
sl@0: 
sl@0:       /* Skip any MATCH_ALL or MATCH_ONE characters that follow a
sl@0:       ** MATCH_ALL. For each MATCH_ONE, skip one character in the 
sl@0:       ** test string.
sl@0:       */
sl@0:       while( (c=zPattern[iPattern]) == MATCH_ALL || c == MATCH_ONE ){
sl@0:         if( c==MATCH_ONE ){
sl@0:           if( zString[iString]==0 ) return 0;
sl@0:           U8_FWD_1_UNSAFE(zString, iString);
sl@0:         }
sl@0:         iPattern++;
sl@0:       }
sl@0: 
sl@0:       if( zPattern[iPattern]==0 ) return 1;
sl@0: 
sl@0:       while( zString[iString] ){
sl@0:         if( icuLikeCompare(&zPattern[iPattern], &zString[iString], uEsc) ){
sl@0:           return 1;
sl@0:         }
sl@0:         U8_FWD_1_UNSAFE(zString, iString);
sl@0:       }
sl@0:       return 0;
sl@0: 
sl@0:     }else if( !prevEscape && uPattern==MATCH_ONE ){
sl@0:       /* Case 2. */
sl@0:       if( zString[iString]==0 ) return 0;
sl@0:       U8_FWD_1_UNSAFE(zString, iString);
sl@0: 
sl@0:     }else if( !prevEscape && uPattern==uEsc){
sl@0:       /* Case 3. */
sl@0:       prevEscape = 1;
sl@0: 
sl@0:     }else{
sl@0:       /* Case 4. */
sl@0:       UChar32 uString;
sl@0:       U8_NEXT_UNSAFE(zString, iString, uString);
sl@0:       uString = u_foldCase(uString, U_FOLD_CASE_DEFAULT);
sl@0:       uPattern = u_foldCase(uPattern, U_FOLD_CASE_DEFAULT);
sl@0:       if( uString!=uPattern ){
sl@0:         return 0;
sl@0:       }
sl@0:       prevEscape = 0;
sl@0:     }
sl@0:   }
sl@0: 
sl@0:   return zString[iString]==0;
sl@0: }
sl@0: 
sl@0: /*
sl@0: ** Implementation of the like() SQL function.  This function implements
sl@0: ** the build-in LIKE operator.  The first argument to the function is the
sl@0: ** pattern and the second argument is the string.  So, the SQL statements:
sl@0: **
sl@0: **       A LIKE B
sl@0: **
sl@0: ** is implemented as like(B, A). If there is an escape character E, 
sl@0: **
sl@0: **       A LIKE B ESCAPE E
sl@0: **
sl@0: ** is mapped to like(B, A, E).
sl@0: */
sl@0: static void icuLikeFunc(
sl@0:   sqlite3_context *context, 
sl@0:   int argc, 
sl@0:   sqlite3_value **argv
sl@0: ){
sl@0:   const unsigned char *zA = sqlite3_value_text(argv[0]);
sl@0:   const unsigned char *zB = sqlite3_value_text(argv[1]);
sl@0:   UChar32 uEsc = 0;
sl@0: 
sl@0:   /* Limit the length of the LIKE or GLOB pattern to avoid problems
sl@0:   ** of deep recursion and N*N behavior in patternCompare().
sl@0:   */
sl@0:   if( sqlite3_value_bytes(argv[0])>SQLITE_MAX_LIKE_PATTERN_LENGTH ){
sl@0:     sqlite3_result_error(context, "LIKE or GLOB pattern too complex", -1);
sl@0:     return;
sl@0:   }
sl@0: 
sl@0: 
sl@0:   if( argc==3 ){
sl@0:     /* The escape character string must consist of a single UTF-8 character.
sl@0:     ** Otherwise, return an error.
sl@0:     */
sl@0:     int nE= sqlite3_value_bytes(argv[2]);
sl@0:     const unsigned char *zE = sqlite3_value_text(argv[2]);
sl@0:     int i = 0;
sl@0:     if( zE==0 ) return;
sl@0:     U8_NEXT(zE, i, nE, uEsc);
sl@0:     if( i!=nE){
sl@0:       sqlite3_result_error(context, 
sl@0:           "ESCAPE expression must be a single character", -1);
sl@0:       return;
sl@0:     }
sl@0:   }
sl@0: 
sl@0:   if( zA && zB ){
sl@0:     sqlite3_result_int(context, icuLikeCompare(zA, zB, uEsc));
sl@0:   }
sl@0: }
sl@0: 
sl@0: /*
sl@0: ** This function is called when an ICU function called from within
sl@0: ** the implementation of an SQL scalar function returns an error.
sl@0: **
sl@0: ** The scalar function context passed as the first argument is 
sl@0: ** loaded with an error message based on the following two args.
sl@0: */
sl@0: static void icuFunctionError(
sl@0:   sqlite3_context *pCtx,       /* SQLite scalar function context */
sl@0:   const char *zName,           /* Name of ICU function that failed */
sl@0:   UErrorCode e                 /* Error code returned by ICU function */
sl@0: ){
sl@0:   char zBuf[128];
sl@0:   sqlite3_snprintf(128, zBuf, "ICU error: %s(): %s", zName, u_errorName(e));
sl@0:   zBuf[127] = '\0';
sl@0:   sqlite3_result_error(pCtx, zBuf, -1);
sl@0: }
sl@0: 
sl@0: /*
sl@0: ** Function to delete compiled regexp objects. Registered as
sl@0: ** a destructor function with sqlite3_set_auxdata().
sl@0: */
sl@0: static void icuRegexpDelete(void *p){
sl@0:   URegularExpression *pExpr = (URegularExpression *)p;
sl@0:   uregex_close(pExpr);
sl@0: }
sl@0: 
sl@0: /*
sl@0: ** Implementation of SQLite REGEXP operator. This scalar function takes
sl@0: ** two arguments. The first is a regular expression pattern to compile
sl@0: ** the second is a string to match against that pattern. If either 
sl@0: ** argument is an SQL NULL, then NULL Is returned. Otherwise, the result
sl@0: ** is 1 if the string matches the pattern, or 0 otherwise.
sl@0: **
sl@0: ** SQLite maps the regexp() function to the regexp() operator such
sl@0: ** that the following two are equivalent:
sl@0: **
sl@0: **     zString REGEXP zPattern
sl@0: **     regexp(zPattern, zString)
sl@0: **
sl@0: ** Uses the following ICU regexp APIs:
sl@0: **
sl@0: **     uregex_open()
sl@0: **     uregex_matches()
sl@0: **     uregex_close()
sl@0: */
sl@0: static void icuRegexpFunc(sqlite3_context *p, int nArg, sqlite3_value **apArg){
sl@0:   UErrorCode status = U_ZERO_ERROR;
sl@0:   URegularExpression *pExpr;
sl@0:   UBool res;
sl@0:   const UChar *zString = sqlite3_value_text16(apArg[1]);
sl@0: 
sl@0:   /* If the left hand side of the regexp operator is NULL, 
sl@0:   ** then the result is also NULL. 
sl@0:   */
sl@0:   if( !zString ){
sl@0:     return;
sl@0:   }
sl@0: 
sl@0:   pExpr = sqlite3_get_auxdata(p, 0);
sl@0:   if( !pExpr ){
sl@0:     const UChar *zPattern = sqlite3_value_text16(apArg[0]);
sl@0:     if( !zPattern ){
sl@0:       return;
sl@0:     }
sl@0:     pExpr = uregex_open(zPattern, -1, 0, 0, &status);
sl@0: 
sl@0:     if( U_SUCCESS(status) ){
sl@0:       sqlite3_set_auxdata(p, 0, pExpr, icuRegexpDelete);
sl@0:     }else{
sl@0:       assert(!pExpr);
sl@0:       icuFunctionError(p, "uregex_open", status);
sl@0:       return;
sl@0:     }
sl@0:   }
sl@0: 
sl@0:   /* Configure the text that the regular expression operates on. */
sl@0:   uregex_setText(pExpr, zString, -1, &status);
sl@0:   if( !U_SUCCESS(status) ){
sl@0:     icuFunctionError(p, "uregex_setText", status);
sl@0:     return;
sl@0:   }
sl@0: 
sl@0:   /* Attempt the match */
sl@0:   res = uregex_matches(pExpr, 0, &status);
sl@0:   if( !U_SUCCESS(status) ){
sl@0:     icuFunctionError(p, "uregex_matches", status);
sl@0:     return;
sl@0:   }
sl@0: 
sl@0:   /* Set the text that the regular expression operates on to a NULL
sl@0:   ** pointer. This is not really necessary, but it is tidier than 
sl@0:   ** leaving the regular expression object configured with an invalid
sl@0:   ** pointer after this function returns.
sl@0:   */
sl@0:   uregex_setText(pExpr, 0, 0, &status);
sl@0: 
sl@0:   /* Return 1 or 0. */
sl@0:   sqlite3_result_int(p, res ? 1 : 0);
sl@0: }
sl@0: 
sl@0: /*
sl@0: ** Implementations of scalar functions for case mapping - upper() and 
sl@0: ** lower(). Function upper() converts its input to upper-case (ABC).
sl@0: ** Function lower() converts to lower-case (abc).
sl@0: **
sl@0: ** ICU provides two types of case mapping, "general" case mapping and
sl@0: ** "language specific". Refer to ICU documentation for the differences
sl@0: ** between the two.
sl@0: **
sl@0: ** To utilise "general" case mapping, the upper() or lower() scalar 
sl@0: ** functions are invoked with one argument:
sl@0: **
sl@0: **     upper('ABC') -> 'abc'
sl@0: **     lower('abc') -> 'ABC'
sl@0: **
sl@0: ** To access ICU "language specific" case mapping, upper() or lower()
sl@0: ** should be invoked with two arguments. The second argument is the name
sl@0: ** of the locale to use. Passing an empty string ("") or SQL NULL value
sl@0: ** as the second argument is the same as invoking the 1 argument version
sl@0: ** of upper() or lower().
sl@0: **
sl@0: **     lower('I', 'en_us') -> 'i'
sl@0: **     lower('I', 'tr_tr') -> 'ı' (small dotless i)
sl@0: **
sl@0: ** http://www.icu-project.org/userguide/posix.html#case_mappings
sl@0: */
sl@0: static void icuCaseFunc16(sqlite3_context *p, int nArg, sqlite3_value **apArg){
sl@0:   const UChar *zInput;
sl@0:   UChar *zOutput;
sl@0:   int nInput;
sl@0:   int nOutput;
sl@0: 
sl@0:   UErrorCode status = U_ZERO_ERROR;
sl@0:   const char *zLocale = 0;
sl@0: 
sl@0:   assert(nArg==1 || nArg==2);
sl@0:   if( nArg==2 ){
sl@0:     zLocale = (const char *)sqlite3_value_text(apArg[1]);
sl@0:   }
sl@0: 
sl@0:   zInput = sqlite3_value_text16(apArg[0]);
sl@0:   if( !zInput ){
sl@0:     return;
sl@0:   }
sl@0:   nInput = sqlite3_value_bytes16(apArg[0]);
sl@0: 
sl@0:   nOutput = nInput * 2 + 2;
sl@0:   zOutput = sqlite3_malloc(nOutput);
sl@0:   if( !zOutput ){
sl@0:     return;
sl@0:   }
sl@0: 
sl@0:   if( sqlite3_user_data(p) ){
sl@0:     u_strToUpper(zOutput, nOutput/2, zInput, nInput/2, zLocale, &status);
sl@0:   }else{
sl@0:     u_strToLower(zOutput, nOutput/2, zInput, nInput/2, zLocale, &status);
sl@0:   }
sl@0: 
sl@0:   if( !U_SUCCESS(status) ){
sl@0:     icuFunctionError(p, "u_strToLower()/u_strToUpper", status);
sl@0:     return;
sl@0:   }
sl@0: 
sl@0:   sqlite3_result_text16(p, zOutput, -1, xFree);
sl@0: }
sl@0: 
sl@0: /*
sl@0: ** Collation sequence destructor function. The pCtx argument points to
sl@0: ** a UCollator structure previously allocated using ucol_open().
sl@0: */
sl@0: static void icuCollationDel(void *pCtx){
sl@0:   UCollator *p = (UCollator *)pCtx;
sl@0:   ucol_close(p);
sl@0: }
sl@0: 
sl@0: /*
sl@0: ** Collation sequence comparison function. The pCtx argument points to
sl@0: ** a UCollator structure previously allocated using ucol_open().
sl@0: */
sl@0: static int icuCollationColl(
sl@0:   void *pCtx,
sl@0:   int nLeft,
sl@0:   const void *zLeft,
sl@0:   int nRight,
sl@0:   const void *zRight
sl@0: ){
sl@0:   UCollationResult res;
sl@0:   UCollator *p = (UCollator *)pCtx;
sl@0:   res = ucol_strcoll(p, (UChar *)zLeft, nLeft/2, (UChar *)zRight, nRight/2);
sl@0:   switch( res ){
sl@0:     case UCOL_LESS:    return -1;
sl@0:     case UCOL_GREATER: return +1;
sl@0:     case UCOL_EQUAL:   return 0;
sl@0:   }
sl@0:   assert(!"Unexpected return value from ucol_strcoll()");
sl@0:   return 0;
sl@0: }
sl@0: 
sl@0: /*
sl@0: ** Implementation of the scalar function icu_load_collation().
sl@0: **
sl@0: ** This scalar function is used to add ICU collation based collation 
sl@0: ** types to an SQLite database connection. It is intended to be called
sl@0: ** as follows:
sl@0: **
sl@0: **     SELECT icu_load_collation(<locale>, <collation-name>);
sl@0: **
sl@0: ** Where <locale> is a string containing an ICU locale identifier (i.e.
sl@0: ** "en_AU", "tr_TR" etc.) and <collation-name> is the name of the
sl@0: ** collation sequence to create.
sl@0: */
sl@0: static void icuLoadCollation(
sl@0:   sqlite3_context *p, 
sl@0:   int nArg, 
sl@0:   sqlite3_value **apArg
sl@0: ){
sl@0:   sqlite3 *db = (sqlite3 *)sqlite3_user_data(p);
sl@0:   UErrorCode status = U_ZERO_ERROR;
sl@0:   const char *zLocale;      /* Locale identifier - (eg. "jp_JP") */
sl@0:   const char *zName;        /* SQL Collation sequence name (eg. "japanese") */
sl@0:   UCollator *pUCollator;    /* ICU library collation object */
sl@0:   int rc;                   /* Return code from sqlite3_create_collation_x() */
sl@0: 
sl@0:   assert(nArg==2);
sl@0:   zLocale = (const char *)sqlite3_value_text(apArg[0]);
sl@0:   zName = (const char *)sqlite3_value_text(apArg[1]);
sl@0: 
sl@0:   if( !zLocale || !zName ){
sl@0:     return;
sl@0:   }
sl@0: 
sl@0:   pUCollator = ucol_open(zLocale, &status);
sl@0:   if( !U_SUCCESS(status) ){
sl@0:     icuFunctionError(p, "ucol_open", status);
sl@0:     return;
sl@0:   }
sl@0:   assert(p);
sl@0: 
sl@0:   rc = sqlite3_create_collation_v2(db, zName, SQLITE_UTF16, (void *)pUCollator, 
sl@0:       icuCollationColl, icuCollationDel
sl@0:   );
sl@0:   if( rc!=SQLITE_OK ){
sl@0:     ucol_close(pUCollator);
sl@0:     sqlite3_result_error(p, "Error registering collation function", -1);
sl@0:   }
sl@0: }
sl@0: 
sl@0: /*
sl@0: ** Register the ICU extension functions with database db.
sl@0: */
sl@0: int sqlite3IcuInit(sqlite3 *db){
sl@0:   struct IcuScalar {
sl@0:     const char *zName;                        /* Function name */
sl@0:     int nArg;                                 /* Number of arguments */
sl@0:     int enc;                                  /* Optimal text encoding */
sl@0:     void *pContext;                           /* sqlite3_user_data() context */
sl@0:     void (*xFunc)(sqlite3_context*,int,sqlite3_value**);
sl@0:   } scalars[] = {
sl@0:     {"regexp",-1, SQLITE_ANY,          0, icuRegexpFunc},
sl@0: 
sl@0:     {"lower",  1, SQLITE_UTF16,        0, icuCaseFunc16},
sl@0:     {"lower",  2, SQLITE_UTF16,        0, icuCaseFunc16},
sl@0:     {"upper",  1, SQLITE_UTF16, (void*)1, icuCaseFunc16},
sl@0:     {"upper",  2, SQLITE_UTF16, (void*)1, icuCaseFunc16},
sl@0: 
sl@0:     {"lower",  1, SQLITE_UTF8,         0, icuCaseFunc16},
sl@0:     {"lower",  2, SQLITE_UTF8,         0, icuCaseFunc16},
sl@0:     {"upper",  1, SQLITE_UTF8,  (void*)1, icuCaseFunc16},
sl@0:     {"upper",  2, SQLITE_UTF8,  (void*)1, icuCaseFunc16},
sl@0: 
sl@0:     {"like",   2, SQLITE_UTF8,         0, icuLikeFunc},
sl@0:     {"like",   3, SQLITE_UTF8,         0, icuLikeFunc},
sl@0: 
sl@0:     {"icu_load_collation",  2, SQLITE_UTF8, (void*)db, icuLoadCollation},
sl@0:   };
sl@0: 
sl@0:   int rc = SQLITE_OK;
sl@0:   int i;
sl@0: 
sl@0:   for(i=0; rc==SQLITE_OK && i<(sizeof(scalars)/sizeof(struct IcuScalar)); i++){
sl@0:     struct IcuScalar *p = &scalars[i];
sl@0:     rc = sqlite3_create_function(
sl@0:         db, p->zName, p->nArg, p->enc, p->pContext, p->xFunc, 0, 0
sl@0:     );
sl@0:   }
sl@0: 
sl@0:   return rc;
sl@0: }
sl@0: 
sl@0: #if !SQLITE_CORE
sl@0: int sqlite3_extension_init(
sl@0:   sqlite3 *db, 
sl@0:   char **pzErrMsg,
sl@0:   const sqlite3_api_routines *pApi
sl@0: ){
sl@0:   SQLITE_EXTENSION_INIT2(pApi)
sl@0:   return sqlite3IcuInit(db);
sl@0: }
sl@0: #endif
sl@0: 
sl@0: #endif