i18npool/source/breakiterator/xdictionary.cxx

   1 /* -*- Mode: C++; tab-width: 4; indent-tabs-mode: nil; c-basic-offset: 4 -*- */
   2 /*
   3  * This file is part of the LibreOffice project.
   4  *
   5  * This Source Code Form is subject to the terms of the Mozilla Public
   6  * License, v. 2.0. If a copy of the MPL was not distributed with this
   7  * file, You can obtain one at http://mozilla.org/MPL/2.0/.
   8  *
   9  * This file incorporates work covered by the following license notice:
  10  *
  11  *   Licensed to the Apache Software Foundation (ASF) under one or more
  12  *   contributor license agreements. See the NOTICE file distributed
  13  *   with this work for additional information regarding copyright
  14  *   ownership. The ASF licenses this file to you under the Apache
  15  *   License, Version 2.0 (the "License"); you may not use this file
  16  *   except in compliance with the License. You may obtain a copy of
  17  *   the License at http://www.apache.org/licenses/LICENSE-2.0 .
  18  */
  19
  20 #include <config_folders.h>
  21
  22 #include <osl/file.h>
  23 #include <osl/mutex.hxx>
  24 #include <rtl/ustrbuf.hxx>
  25 #include <rtl/bootstrap.hxx>
  26 #include <com/sun/star/i18n/WordType.hpp>
  27 #include <xdictionary.hxx>
  28 #include <unicode/uchar.h>
  29 #include <string.h>
  30 #include <breakiteratorImpl.hxx>
  31
  32 using namespace com::sun::star::i18n;
  33
  34 namespace i18npool {
  35
  36 #ifdef DICT_JA_ZH_IN_DATAFILE
  37
  38 #elif !defined DISABLE_DYNLOADING
  39
  40 extern "C" { static void SAL_CALL thisModule() {} }
  41
  42 #else
  43
  44 extern "C" {
  45
  46 sal_uInt8* getExistMark_ja();
  47 sal_Int16* getIndex1_ja();
  48 sal_Int32* getIndex2_ja();
  49 sal_Int32* getLenArray_ja();
  50 sal_Unicode* getDataArea_ja();
  51
  52 sal_uInt8* getExistMark_zh();
  53 sal_Int16* getIndex1_zh();
  54 sal_Int32* getIndex2_zh();
  55 sal_Int32* getLenArray_zh();
  56 sal_Unicode* getDataArea_zh();
  57
  58 }
  59
  60 #endif
  61
  62 xdictionary::xdictionary(const sal_Char *lang) :
  63     boundary(),
  64     japaneseWordBreak( false )
  65 {
  66
  67 #ifdef DICT_JA_ZH_IN_DATAFILE
  68
  69     if( strcmp( lang, "ja" ) == 0 || strcmp( lang, "zh" ) == 0 )
  70     {
  71         OUString sUrl( "$BRAND_BASE_DIR/" LIBO_SHARE_FOLDER "/dict_" );
  72         rtl::Bootstrap::expandMacros(sUrl);
  73
  74         if( strcmp( lang, "ja" ) == 0 )
  75             sUrl += "ja.data";
  76         else if( strcmp( lang, "zh" ) == 0 )
  77             sUrl += "zh.data";
  78
  79         oslFileHandle aFileHandle;
  80         sal_uInt64 nFileSize;
  81         char *pMapping;
  82         if( osl_openFile( sUrl.pData, &aFileHandle, osl_File_OpenFlag_Read ) == osl_File_E_None &&
  83             osl_getFileSize( aFileHandle, &nFileSize) == osl_File_E_None &&
  84             osl_mapFile( aFileHandle, (void **) &pMapping, nFileSize, 0, osl_File_MapFlag_RandomAccess ) == osl_File_E_None )
  85         {
  86             // We have the offsets to the parts of the file at its end, see gendict.cxx
  87             sal_Int64 *pEOF = (sal_Int64*)(pMapping + nFileSize);
  88
  89             data.existMark = (sal_uInt8*) (pMapping + pEOF[-1]);
  90             data.index2 = (sal_Int32*) (pMapping + pEOF[-2]);
  91             data.index1 = (sal_Int16*) (pMapping + pEOF[-3]);
  92             data.lenArray = (sal_Int32*) (pMapping + pEOF[-4]);
  93             data.dataArea = (sal_Unicode*) (pMapping + pEOF[-5]);
  94         }
  95     }
  96
  97 #elif !defined DISABLE_DYNLOADING
  98
  99     initDictionaryData( lang );
 100
 101 #else
 102
 103     if( strcmp( lang, "ja" ) == 0 ) {
 104         data.existMark = getExistMark_ja();
 105         data.index1 = getIndex1_ja();
 106         data.index2 = getIndex2_ja();
 107         data.lenArray = getLenArray_ja();
 108         data.dataArea = getDataArea_ja();
 109     }
 110     else if( strcmp( lang, "zh" ) == 0 ) {
 111         data.existMark = getExistMark_zh();
 112         data.index1 = getIndex1_zh();
 113         data.index2 = getIndex2_zh();
 114         data.lenArray = getLenArray_zh();
 115         data.dataArea = getDataArea_zh();
 116     }
 117
 118 #endif
 119
 120     for (WordBreakCache & i : cache)
 121         i.size = 0;
 122
 123     japaneseWordBreak = false;
 124 }
 125
 126 xdictionary::~xdictionary()
 127 {
 128     for (WordBreakCache & i : cache) {
 129         if (i.size > 0) {
 130             delete [] i.contents;
 131             delete [] i.wordboundary;
 132         }
 133     }
 134 }
 135
 136 namespace {
 137     struct datacache {
 138         oslModule       mhModule;
 139         OString         maLang;
 140         xdictionarydata maData;
 141     };
 142 }
 143
 144 #if !defined(DICT_JA_ZH_IN_DATAFILE) && !defined(DISABLE_DYNLOADING)
 145
 146 void xdictionary::initDictionaryData(const sal_Char *pLang)
 147 {
 148     // Global cache, never released for performance
 149     static std::vector< datacache > aLoadedCache;
 150
 151     osl::MutexGuard aGuard( osl::Mutex::getGlobalMutex() );
 152     for(datacache & i : aLoadedCache)
 153     {
 154         if( i.maLang != pLang )
 155         {
 156             data = i.maData;
 157             return;
 158         }
 159     }
 160
 161     // otherwise add to the cache, positive or negative.
 162     datacache aEntry;
 163     aEntry.maLang = OString( pLang, strlen( pLang ) );
 164
 165 #ifdef SAL_DLLPREFIX
 166     OUStringBuffer aBuf( sal::static_int_cast<int>(strlen(pLang) + 7 + 6) );    // mostly "lib*.so" (with * == dict_zh)
 167     aBuf.append( SAL_DLLPREFIX );
 168 #else
 169     OUStringBuffer aBuf( sal::static_int_cast<int>(strlen(pLang) + 7 + 4) );    // mostly "*.dll" (with * == dict_zh)
 170 #endif
 171     aBuf.append( "dict_" ).appendAscii( pLang ).append( SAL_DLLEXTENSION );
 172     aEntry.mhModule = osl_loadModuleRelative( &thisModule, aBuf.makeStringAndClear().pData, SAL_LOADMODULE_DEFAULT );
 173     if( aEntry.mhModule ) {
 174         oslGenericFunction func;
 175         func = osl_getAsciiFunctionSymbol( aEntry.mhModule, "getExistMark" );
 176         aEntry.maData.existMark = reinterpret_cast<sal_uInt8 const * (*)()>(func)();
 177         func = osl_getAsciiFunctionSymbol( aEntry.mhModule, "getIndex1" );
 178         aEntry.maData.index1 = reinterpret_cast<sal_Int16 const * (*)()>(func)();
 179         func = osl_getAsciiFunctionSymbol( aEntry.mhModule, "getIndex2" );
 180         aEntry.maData.index2 = reinterpret_cast<sal_Int32 const * (*)()>(func)();
 181         func = osl_getAsciiFunctionSymbol( aEntry.mhModule, "getLenArray" );
 182         aEntry.maData.lenArray = reinterpret_cast<sal_Int32 const * (*)()>(func)();
 183         func = osl_getAsciiFunctionSymbol( aEntry.mhModule, "getDataArea" );
 184         aEntry.maData.dataArea = reinterpret_cast<sal_Unicode const * (*)()>(func)();
 185     }
 186
 187     data = aEntry.maData;
 188     aLoadedCache.push_back( aEntry );
 189 }
 190
 191 #endif
 192
 193 void xdictionary::setJapaneseWordBreak()
 194 {
 195     japaneseWordBreak = true;
 196 }
 197
 198 bool xdictionary::exists(const sal_uInt32 c)
 199 {
 200     // 0x1FFF is the hardcoded limit in gendict for data.existMarks
 201     bool exist = data.existMark && (c>>3) < 0x1FFF && (data.existMark[c>>3] & (1<<(c&0x07))) != 0;
 202     if (!exist && japaneseWordBreak)
 203         return BreakIteratorImpl::getScriptClass(c) == css::i18n::ScriptType::ASIAN;
 204     else
 205         return exist;
 206 }
 207
 208 sal_Int32 xdictionary::getLongestMatch(const sal_Unicode* str, sal_Int32 sLen)
 209 {
 210     if ( !data.index1 ) return 0;
 211
 212     sal_Int16 idx = data.index1[str[0] >> 8];
 213
 214     if (idx == 0xFF) return 0;
 215
 216     idx = (idx<<8) | (str[0]&0xff);
 217
 218     sal_uInt32 begin = data.index2[idx], end = data.index2[idx+1];
 219
 220     if (begin == 0) return 0;
 221
 222     str++; sLen--; // first character is not stored in the dictionary
 223     for (sal_uInt32 i = end; i > begin; i--) {
 224         sal_Int32 len = data.lenArray[i] - data.lenArray[i - 1];
 225         if (sLen >= len) {
 226             const sal_Unicode *dstr = data.dataArea + data.lenArray[i-1];
 227             sal_Int32 pos = 0;
 228
 229             while (pos < len && dstr[pos] == str[pos]) { pos++; }
 230
 231             if (pos == len)
 232                 return len + 1;
 233         }
 234     }
 235     return 0;
 236 }
 237
 238
 239 /*
 240  * c-tor
 241  */
 242
 243 WordBreakCache::WordBreakCache() :
 244     length( 0 ),
 245     contents( nullptr ),
 246     wordboundary( nullptr ),
 247     size( 0 )
 248 {
 249 }
 250
 251 /*
 252  * Compare two unicode string,
 253  */
 254
 255 bool WordBreakCache::equals(const sal_Unicode* str, Boundary const & boundary)
 256 {
 257     // Different length, different string.
 258     if (length != boundary.endPos - boundary.startPos) return false;
 259
 260     for (sal_Int32 i = 0; i < length; i++)
 261         if (contents[i] != str[i + boundary.startPos]) return false;
 262
 263     return true;
 264 }
 265
 266
 267 /*
 268  * Retrieve the segment containing the character at pos.
 269  * @param pos : Position of the given character.
 270  * @return true if CJK.
 271  */
 272 bool xdictionary::seekSegment(const OUString &rText, sal_Int32 pos,
 273     Boundary& segBoundary)
 274 {
 275     sal_Int32 indexUtf16;
 276
 277     if (segmentCachedString.pData != rText.pData) {
 278         // Cache the passed text so we can avoid regenerating the segment if it's the same
 279         // (pData is refcounted and assigning the OUString references it, which ensures that
 280         // the object is the same if we get the same pointer back later)
 281         segmentCachedString = rText;
 282     } else {
 283         // If pos is within the cached boundary, use that boundary
 284         if (pos >= segmentCachedBoundary.startPos && pos <= segmentCachedBoundary.endPos) {
 285             segBoundary.startPos = segmentCachedBoundary.startPos;
 286             segBoundary.endPos = segmentCachedBoundary.endPos;
 287             indexUtf16 = segmentCachedBoundary.startPos;
 288             rText.iterateCodePoints(&indexUtf16);
 289             return segmentCachedBoundary.endPos > indexUtf16;
 290         }
 291     }
 292
 293     segBoundary.endPos = segBoundary.startPos = pos;
 294
 295     indexUtf16 = pos;
 296     while (indexUtf16 > 0)
 297     {
 298         sal_uInt32 ch = rText.iterateCodePoints(&indexUtf16, -1);
 299         if (u_isWhitespace(ch) || exists(ch))
 300             segBoundary.startPos = indexUtf16;
 301         else
 302             break;
 303     }
 304
 305     indexUtf16 = pos;
 306     while (indexUtf16 < rText.getLength())
 307     {
 308         sal_uInt32 ch = rText.iterateCodePoints(&indexUtf16);
 309         if (u_isWhitespace(ch) || exists(ch))
 310             segBoundary.endPos = indexUtf16;
 311         else
 312             break;
 313     }
 314
 315     // Cache the calculated boundary
 316     segmentCachedBoundary.startPos = segBoundary.startPos;
 317     segmentCachedBoundary.endPos = segBoundary.endPos;
 318
 319     indexUtf16 = segBoundary.startPos;
 320     rText.iterateCodePoints(&indexUtf16);
 321     return segBoundary.endPos > indexUtf16;
 322 }
 323
 324 #define KANJA       1
 325 #define KATAKANA    2
 326 #define HIRAKANA    3
 327
 328 static sal_Int16 JapaneseCharType(sal_Unicode c)
 329 {
 330     if (0x3041 <= c && c <= 0x309e)
 331         return HIRAKANA;
 332     if ((0x30a1 <= c && c <= 0x30fe) || (0xff65 <= c && c <= 0xff9f))
 333         return KATAKANA;
 334     return KANJA;
 335 }
 336
 337 WordBreakCache& xdictionary::getCache(const sal_Unicode *text, Boundary const & wordBoundary)
 338 {
 339     WordBreakCache& rCache = cache[text[0] & 0x1f];
 340
 341     if (rCache.size != 0 && rCache.equals(text, wordBoundary))
 342         return rCache;
 343
 344     sal_Int32 len = wordBoundary.endPos - wordBoundary.startPos;
 345
 346     if (rCache.size == 0 || len > rCache.size) {
 347         if (rCache.size != 0) {
 348             delete [] rCache.contents;
 349             delete [] rCache.wordboundary;
 350             rCache.size = len;
 351         }
 352         else
 353             rCache.size = len > DEFAULT_SIZE ? len : DEFAULT_SIZE;
 354         rCache.contents = new sal_Unicode[rCache.size + 1];
 355         rCache.wordboundary = new sal_Int32[rCache.size + 2];
 356     }
 357     rCache.length  = len;
 358     memcpy(rCache.contents, text + wordBoundary.startPos, len * sizeof(sal_Unicode));
 359     *(rCache.contents + len) = 0x0000;
 360     // reset the wordboundary in cache
 361     memset(rCache.wordboundary, '\0', sizeof(sal_Int32)*(len + 2));
 362
 363     sal_Int32 i = 0;        // loop variable
 364     while (rCache.wordboundary[i] < rCache.length) {
 365         len = 0;
 366         // look the continuous white space as one word and cache it
 367         while (u_isWhitespace((sal_uInt32)text[wordBoundary.startPos + rCache.wordboundary[i] + len]))
 368             len ++;
 369
 370         if (len == 0) {
 371             const sal_Unicode *str = text + wordBoundary.startPos + rCache.wordboundary[i];
 372             sal_Int32 slen = rCache.length - rCache.wordboundary[i];
 373             sal_Int16 type = 0, count = 0;
 374             for (;len == 0 && slen > 0; str++, slen--) {
 375                 len = getLongestMatch(str, slen);
 376                 if (len == 0) {
 377                     if (!japaneseWordBreak) {
 378                         len = 1;
 379                     } else {
 380                         if (count == 0)
 381                             type = JapaneseCharType(*str);
 382                         else if (type != JapaneseCharType(*str))
 383                             break;
 384                         count++;
 385                     }
 386                 }
 387             }
 388             if (count)
 389             {
 390                 rCache.wordboundary[i+1] = rCache.wordboundary[i] + count;
 391                 i++;
 392             }
 393         }
 394
 395         if (len) {
 396             rCache.wordboundary[i+1] = rCache.wordboundary[i] + len;
 397             i++;
 398         }
 399     }
 400     rCache.wordboundary[i + 1] = rCache.length + 1;
 401
 402     return rCache;
 403 }
 404
 405 Boundary xdictionary::previousWord(const OUString& rText, sal_Int32 anyPos, sal_Int16 wordType)
 406 {
 407         // looking for the first non-whitespace character from anyPos
 408         sal_uInt32 ch = 0;
 409         if (anyPos > 0)
 410             rText.iterateCodePoints(&anyPos, -1);
 411
 412         while (anyPos > 0 && u_isWhitespace(ch)) ch = rText.iterateCodePoints(&anyPos, -1);
 413
 414         return getWordBoundary(rText, anyPos, wordType, true);
 415 }
 416
 417 Boundary xdictionary::nextWord(const OUString& rText, sal_Int32 anyPos, sal_Int16 wordType)
 418 {
 419         boundary = getWordBoundary(rText, anyPos, wordType, true);
 420         anyPos = boundary.endPos;
 421         const sal_Int32 nLen = rText.getLength();
 422         if (anyPos < nLen) {
 423             // looking for the first non-whitespace character from anyPos
 424             sal_uInt32 ch = rText.iterateCodePoints(&anyPos);
 425             while (u_isWhitespace(ch) && (anyPos < nLen)) ch=rText.iterateCodePoints(&anyPos);
 426             if (anyPos > 0)
 427                 rText.iterateCodePoints(&anyPos, -1);
 428         }
 429
 430         return getWordBoundary(rText, anyPos, wordType, true);
 431 }
 432
 433 Boundary const & xdictionary::getWordBoundary(const OUString& rText, sal_Int32 anyPos, sal_Int16 wordType, bool bDirection)
 434 {
 435         const sal_Unicode *text=rText.getStr();
 436         sal_Int32 len=rText.getLength();
 437         if (anyPos >= len || anyPos < 0) {
 438             boundary.startPos = boundary.endPos = anyPos < 0 ? 0 : len;
 439         } else if (seekSegment(rText, anyPos, boundary)) {          // character in dict
 440             WordBreakCache& aCache = getCache(text, boundary);
 441             sal_Int32 i = 0;
 442
 443             while (aCache.wordboundary[i] <= anyPos - boundary.startPos) i++;
 444
 445             sal_Int32 startPos = aCache.wordboundary[i - 1];
 446             // if bDirection is false
 447             if (!bDirection && startPos > 0 && startPos == (anyPos - boundary.startPos))
 448             {
 449                 sal_Int32 indexUtf16 = anyPos-1;
 450                 sal_uInt32 ch = rText.iterateCodePoints(&indexUtf16);
 451                 if (u_isWhitespace(ch))
 452                     i--;
 453             }
 454
 455             boundary.endPos = boundary.startPos;
 456             boundary.endPos += aCache.wordboundary[i];
 457             boundary.startPos += aCache.wordboundary[i-1];
 458
 459         } else {
 460             boundary.startPos = anyPos;
 461             if (anyPos < len) rText.iterateCodePoints(&anyPos);
 462             boundary.endPos = anyPos < len ? anyPos : len;
 463         }
 464         if (wordType == WordType::WORD_COUNT) {
 465             // skip punctuation for word count.
 466             while (boundary.endPos < len)
 467             {
 468                 sal_Int32 indexUtf16 = boundary.endPos;
 469                 if (u_ispunct(rText.iterateCodePoints(&indexUtf16)))
 470                     boundary.endPos = indexUtf16;
 471                 else
 472                     break;
 473             }
 474         }
 475
 476         return boundary;
 477 }
 478
 479 }
 480
 481 /* vim:set shiftwidth=4 softtabstop=4 expandtab: */