| OLD | NEW |
| (Empty) |
| 1 Index: src/hunspell/affixmgr.cxx | |
| 2 =================================================================== | |
| 3 RCS file: /cvsroot/hunspell/hunspell/src/hunspell/affixmgr.cxx,v | |
| 4 retrieving revision 1.41 | |
| 5 diff -u -r1.41 affixmgr.cxx | |
| 6 --- src/hunspell/affixmgr.cxx 16 Dec 2011 09:15:34 -0000 1.41 | |
| 7 +++ src/hunspell/affixmgr.cxx 29 May 2014 01:05:07 -0000 | |
| 8 @@ -14,8 +14,14 @@ | |
| 9 | |
| 10 #include "csutil.hxx" | |
| 11 | |
| 12 +#ifdef HUNSPELL_CHROME_CLIENT | |
| 13 +AffixMgr::AffixMgr(hunspell::BDictReader* reader, HashMgr** ptr, int * md) | |
| 14 +{ | |
| 15 + bdict_reader = reader; | |
| 16 +#else | |
| 17 AffixMgr::AffixMgr(const char * affpath, HashMgr** ptr, int * md, const char *
key) | |
| 18 { | |
| 19 +#endif | |
| 20 // register hash manager and load affix data from aff file | |
| 21 pHMgr = ptr[0]; | |
| 22 alldic = ptr; | |
| 23 @@ -107,9 +113,17 @@ | |
| 24 sFlag[i] = NULL; | |
| 25 } | |
| 26 | |
| 27 +#ifdef HUNSPELL_CHROME_CLIENT | |
| 28 + // Define dummy parameters for parse_file() to avoid changing the parameters | |
| 29 + // of parse_file(). This may make it easier to merge the changes of the | |
| 30 + // original hunspell. | |
| 31 + const char* affpath = NULL; | |
| 32 + const char* key = NULL; | |
| 33 +#else | |
| 34 for (int j=0; j < CONTSIZE; j++) { | |
| 35 contclasses[j] = 0; | |
| 36 } | |
| 37 +#endif | |
| 38 | |
| 39 if (parse_file(affpath, key)) { | |
| 40 HUNSPELL_WARNING(stderr, "Failure loading aff file %s\n",affpath); | |
| 41 @@ -269,6 +283,43 @@ | |
| 42 char * line; // io buffers | |
| 43 char ft; // affix type | |
| 44 | |
| 45 +#ifdef HUNSPELL_CHROME_CLIENT | |
| 46 + // open the affix file | |
| 47 + // We're always UTF-8 | |
| 48 + utf8 = 1; | |
| 49 + | |
| 50 + // A BDICT file stores PFX and SFX lines in a special section and it provides | |
| 51 + // a special line iterator for reading PFX and SFX lines. | |
| 52 + // We create a FileMgr object from this iterator and parse PFX and SFX lines | |
| 53 + // before parsing other lines. | |
| 54 + hunspell::LineIterator affix_iterator = bdict_reader->GetAffixLineIterator(); | |
| 55 + FileMgr* iterator = new FileMgr(&affix_iterator); | |
| 56 + if (!iterator) { | |
| 57 + HUNSPELL_WARNING(stderr, | |
| 58 + "error: could not create a FileMgr from an affix line iterator.\n"); | |
| 59 + return 1; | |
| 60 + } | |
| 61 + | |
| 62 + while ((line = iterator->getline())) { | |
| 63 + ft = ' '; | |
| 64 + if (strncmp(line,"PFX",3) == 0) ft = complexprefixes ? 'S' : 'P'; | |
| 65 + if (strncmp(line,"SFX",3) == 0) ft = complexprefixes ? 'P' : 'S'; | |
| 66 + if (ft != ' ') | |
| 67 + parse_affix(line, ft, iterator, NULL); | |
| 68 + } | |
| 69 + delete iterator; | |
| 70 + | |
| 71 + // Create a FileMgr object for reading lines except PFX and SFX lines. | |
| 72 + // We don't need to change the loop below since our FileMgr emulates the | |
| 73 + // original one. | |
| 74 + hunspell::LineIterator other_iterator = bdict_reader->GetOtherLineIterator(); | |
| 75 + FileMgr * afflst = new FileMgr(&other_iterator); | |
| 76 + if (!afflst) { | |
| 77 + HUNSPELL_WARNING(stderr, | |
| 78 + "error: could not create a FileMgr from an other line iterator.\n"); | |
| 79 + return 1; | |
| 80 + } | |
| 81 +#else | |
| 82 // checking flag duplication | |
| 83 char dupflags[CONTSIZE]; | |
| 84 char dupflags_ini = 1; | |
| 85 @@ -282,6 +333,7 @@ | |
| 86 HUNSPELL_WARNING(stderr, "error: could not open affix description file %s\n
",affpath); | |
| 87 return 1; | |
| 88 } | |
| 89 +#endif | |
| 90 | |
| 91 // step one is to parse the affix file building up the internal | |
| 92 // affix data structures | |
| 93 @@ -291,6 +343,7 @@ | |
| 94 while ((line = afflst->getline()) != NULL) { | |
| 95 mychomp(line); | |
| 96 | |
| 97 +#ifndef HUNSPELL_CHROME_CLIENT | |
| 98 /* remove byte order mark */ | |
| 99 if (firstline) { | |
| 100 firstline = 0; | |
| 101 @@ -299,6 +352,7 @@ | |
| 102 memmove(line, line+3, strlen(line+3)+1); | |
| 103 } | |
| 104 } | |
| 105 +#endif | |
| 106 | |
| 107 /* parse in the keyboard string */ | |
| 108 if (strncmp(line,"KEY",3) == 0) { | |
| 109 @@ -545,6 +599,7 @@ | |
| 110 } | |
| 111 } | |
| 112 | |
| 113 +#ifndef HUNSPELL_CHROME_CLIENT | |
| 114 /* parse in the typical fault correcting table */ | |
| 115 if (strncmp(line,"REP",3) == 0) { | |
| 116 if (parse_reptable(line, afflst)) { | |
| 117 @@ -552,6 +607,7 @@ | |
| 118 return 1; | |
| 119 } | |
| 120 } | |
| 121 +#endif | |
| 122 | |
| 123 /* parse in the input conversion table */ | |
| 124 if (strncmp(line,"ICONV",5) == 0) { | |
| 125 @@ -699,6 +755,7 @@ | |
| 126 checksharps=1; | |
| 127 } | |
| 128 | |
| 129 +#ifndef HUNSPELL_CHROME_CLIENT | |
| 130 /* parse this affix: P - prefix, S - suffix */ | |
| 131 ft = ' '; | |
| 132 if (strncmp(line,"PFX",3) == 0) ft = complexprefixes ? 'S' : 'P'; | |
| 133 @@ -713,6 +770,7 @@ | |
| 134 return 1; | |
| 135 } | |
| 136 } | |
| 137 +#endif | |
| 138 } | |
| 139 | |
| 140 finishFileMgr(afflst); | |
| 141 @@ -1307,6 +1365,26 @@ | |
| 142 const char * r; | |
| 143 int lenr, lenp; | |
| 144 | |
| 145 +#ifdef HUNSPELL_CHROME_CLIENT | |
| 146 + const char *pattern, *pattern2; | |
| 147 + hunspell::ReplacementIterator iterator = bdict_reader->GetReplacementIterator
(); | |
| 148 + while (iterator.GetNext(&pattern, &pattern2)) { | |
| 149 + r = word; | |
| 150 + lenr = strlen(pattern2); | |
| 151 + lenp = strlen(pattern); | |
| 152 + | |
| 153 + // search every occurence of the pattern in the word | |
| 154 + while ((r=strstr(r, pattern)) != NULL) { | |
| 155 + strcpy(candidate, word); | |
| 156 + if (r-word + lenr + strlen(r+lenp) >= MAXLNLEN) break; | |
| 157 + strcpy(candidate+(r-word), pattern2); | |
| 158 + strcpy(candidate+(r-word)+lenr, r+lenp); | |
| 159 + if (candidate_check(candidate,strlen(candidate))) return 1; | |
| 160 + r++; // search for the next letter | |
| 161 + } | |
| 162 + } | |
| 163 + | |
| 164 +#else | |
| 165 if ((wl < 2) || !numrep) return 0; | |
| 166 | |
| 167 for (int i=0; i < numrep; i++ ) { | |
| 168 @@ -1323,6 +1401,7 @@ | |
| 169 r++; // search for the next letter | |
| 170 } | |
| 171 } | |
| 172 +#endif | |
| 173 return 0; | |
| 174 } | |
| 175 | |
| 176 @@ -4219,6 +4298,7 @@ | |
| 177 case 1: { | |
| 178 np++; | |
| 179 aflag = pHMgr->decode_flag(piece); | |
| 180 +#ifndef HUNSPELL_CHROME_CLIENT // We don't check for duplicates. | |
| 181 if (((at == 'S') && (dupflags[aflag] & dupSFX)) || | |
| 182 ((at == 'P') && (dupflags[aflag] & dupPFX))) { | |
| 183 HUNSPELL_WARNING(stderr, "error: line %d: multiple defi
nitions of an affix flag\n", | |
| 184 @@ -4226,6 +4306,7 @@ | |
| 185 // return 1; XXX permissive mode for bad dictionaries | |
| 186 } | |
| 187 dupflags[aflag] += (char) ((at == 'S') ? dupSFX : dupPFX); | |
| 188 +#endif | |
| 189 break; | |
| 190 } | |
| 191 // piece 3 - is cross product indicator | |
| 192 Index: src/hunspell/affixmgr.hxx | |
| 193 =================================================================== | |
| 194 RCS file: /cvsroot/hunspell/hunspell/src/hunspell/affixmgr.hxx,v | |
| 195 retrieving revision 1.15 | |
| 196 diff -u -r1.15 affixmgr.hxx | |
| 197 --- src/hunspell/affixmgr.hxx 13 Oct 2011 13:41:54 -0000 1.15 | |
| 198 +++ src/hunspell/affixmgr.hxx 29 May 2014 01:05:07 -0000 | |
| 199 @@ -18,6 +18,40 @@ | |
| 200 class PfxEntry; | |
| 201 class SfxEntry; | |
| 202 | |
| 203 +#ifdef HUNSPELL_CHROME_CLIENT | |
| 204 + | |
| 205 +#include <vector> | |
| 206 + | |
| 207 +// This class provides an implementation of the contclasses array in AffixMgr | |
| 208 +// that is normally a large static array. We should almost never need more than | |
| 209 +// 256 elements, so this class only allocates that much to start off with. If | |
| 210 +// elements higher than that are actually used, we'll automatically expand. | |
| 211 +class ContClasses { | |
| 212 + public: | |
| 213 + ContClasses() { | |
| 214 + // Pre-allocate a buffer so that typically, we'll never have to resize. | |
| 215 + EnsureSizeIs(256); | |
| 216 + } | |
| 217 + | |
| 218 + char& operator[](size_t index) { | |
| 219 + EnsureSizeIs(index + 1); | |
| 220 + return data[index]; | |
| 221 + } | |
| 222 + | |
| 223 + void EnsureSizeIs(size_t new_size) { | |
| 224 + if (data.size() >= new_size) | |
| 225 + return; // Nothing to do. | |
| 226 + | |
| 227 + size_t old_size = data.size(); | |
| 228 + data.resize(new_size); | |
| 229 + memset(&data[old_size], 0, new_size - old_size); | |
| 230 + } | |
| 231 + | |
| 232 + std::vector<char> data; | |
| 233 +}; | |
| 234 + | |
| 235 +#endif // HUNSPELL_CHROME_CLIENT | |
| 236 + | |
| 237 class LIBHUNSPELL_DLL_EXPORTED AffixMgr | |
| 238 { | |
| 239 | |
| 240 @@ -106,12 +140,20 @@ | |
| 241 int fullstrip; | |
| 242 | |
| 243 int havecontclass; // boolean variable | |
| 244 +#ifdef HUNSPELL_CHROME_CLIENT | |
| 245 + ContClasses contclasses; | |
| 246 +#else | |
| 247 char contclasses[CONTSIZE]; // flags of possible continuing cl
asses (twofold affix) | |
| 248 +#endif | |
| 249 | |
| 250 public: | |
| 251 | |
| 252 +#ifdef HUNSPELL_CHROME_CLIENT | |
| 253 + AffixMgr(hunspell::BDictReader* reader, HashMgr** ptr, int * md); | |
| 254 +#else | |
| 255 AffixMgr(const char * affpath, HashMgr** ptr, int * md, | |
| 256 const char * key = NULL); | |
| 257 +#endif | |
| 258 ~AffixMgr(); | |
| 259 struct hentry * affix_check(const char * word, int len, | |
| 260 const unsigned short needflag = (unsigned short) 0, | |
| 261 @@ -218,6 +260,10 @@ | |
| 262 int get_fullstrip() const; | |
| 263 | |
| 264 private: | |
| 265 +#ifdef HUNSPELL_CHROME_CLIENT | |
| 266 + // Not owned by us, owned by the Hunspell object. | |
| 267 + hunspell::BDictReader* bdict_reader; | |
| 268 +#endif | |
| 269 int parse_file(const char * affpath, const char * key); | |
| 270 int parse_flag(char * line, unsigned short * out, FileMgr * af); | |
| 271 int parse_num(char * line, int * out, FileMgr * af); | |
| 272 @@ -249,4 +295,3 @@ | |
| 273 }; | |
| 274 | |
| 275 #endif | |
| 276 - | |
| 277 Index: src/hunspell/filemgr.cxx | |
| 278 =================================================================== | |
| 279 RCS file: /cvsroot/hunspell/hunspell/src/hunspell/filemgr.cxx,v | |
| 280 retrieving revision 1.5 | |
| 281 diff -u -r1.5 filemgr.cxx | |
| 282 --- src/hunspell/filemgr.cxx 23 Jun 2011 09:21:50 -0000 1.5 | |
| 283 +++ src/hunspell/filemgr.cxx 29 May 2014 01:05:07 -0000 | |
| 284 @@ -7,6 +7,32 @@ | |
| 285 | |
| 286 #include "filemgr.hxx" | |
| 287 | |
| 288 +#ifdef HUNSPELL_CHROME_CLIENT | |
| 289 +#include "third_party/hunspell/google/bdict_reader.h" | |
| 290 + | |
| 291 +FileMgr::FileMgr(hunspell::LineIterator* iterator) : iterator_(iterator) { | |
| 292 +} | |
| 293 + | |
| 294 +FileMgr::~FileMgr() { | |
| 295 +} | |
| 296 + | |
| 297 +char * FileMgr::getline() { | |
| 298 + // Read one line from a BDICT file and store the line to our line buffer. | |
| 299 + // To emulate the original FileMgr::getline(), this function returns | |
| 300 + // the pointer to our line buffer if we can read a line without errors. | |
| 301 + // Otherwise, this function returns NULL. | |
| 302 + bool result = iterator_->AdvanceAndCopy(line_, BUFSIZE - 1); | |
| 303 + return result ? line_ : NULL; | |
| 304 +} | |
| 305 + | |
| 306 +int FileMgr::getlinenum() { | |
| 307 + // This function is used only for displaying a line number that causes a | |
| 308 + // parser error. For a BDICT file, providing a line number doesn't help | |
| 309 + // identifying the place where causes a parser error so much since it is a | |
| 310 + // binary file. So, we just return 0. | |
| 311 + return 0; | |
| 312 +} | |
| 313 +#else | |
| 314 int FileMgr::fail(const char * err, const char * par) { | |
| 315 fprintf(stderr, err, par); | |
| 316 return -1; | |
| 317 @@ -47,3 +73,4 @@ | |
| 318 int FileMgr::getlinenum() { | |
| 319 return linenum; | |
| 320 } | |
| 321 +#endif | |
| 322 Index: src/hunspell/filemgr.hxx | |
| 323 =================================================================== | |
| 324 RCS file: /cvsroot/hunspell/hunspell/src/hunspell/filemgr.hxx,v | |
| 325 retrieving revision 1.3 | |
| 326 diff -u -r1.3 filemgr.hxx | |
| 327 --- src/hunspell/filemgr.hxx 15 Apr 2010 11:22:08 -0000 1.3 | |
| 328 +++ src/hunspell/filemgr.hxx 29 May 2014 01:05:07 -0000 | |
| 329 @@ -7,6 +7,30 @@ | |
| 330 #include "hunzip.hxx" | |
| 331 #include <stdio.h> | |
| 332 | |
| 333 +#ifdef HUNSPELL_CHROME_CLIENT | |
| 334 +namespace hunspell { | |
| 335 +class LineIterator; | |
| 336 +} // namespace hunspell | |
| 337 + | |
| 338 +// A class which encapsulates operations of reading a BDICT file. | |
| 339 +// Chrome uses a BDICT file to compress hunspell dictionaries. A BDICT file is | |
| 340 +// a binary file converted from a DIC file and an AFF file. (See | |
| 341 +// "bdict_reader.h" for its format.) | |
| 342 +// This class encapsulates the operations of reading a BDICT file and emulates | |
| 343 +// the original FileMgr operations for AffixMgr so that it can read a BDICT | |
| 344 +// file without so many changes. | |
| 345 +class FileMgr { | |
| 346 + public: | |
| 347 + FileMgr(hunspell::LineIterator* iterator); | |
| 348 + ~FileMgr(); | |
| 349 + char* getline(); | |
| 350 + int getlinenum(); | |
| 351 + | |
| 352 + protected: | |
| 353 + hunspell::LineIterator* iterator_; | |
| 354 + char line_[BUFSIZE + 50]; // input buffer | |
| 355 +}; | |
| 356 +#else | |
| 357 class LIBHUNSPELL_DLL_EXPORTED FileMgr | |
| 358 { | |
| 359 protected: | |
| 360 @@ -23,3 +47,4 @@ | |
| 361 int getlinenum(); | |
| 362 }; | |
| 363 #endif | |
| 364 +#endif | |
| 365 Index: src/hunspell/hashmgr.cxx | |
| 366 =================================================================== | |
| 367 RCS file: /cvsroot/hunspell/hunspell/src/hunspell/hashmgr.cxx,v | |
| 368 retrieving revision 1.12 | |
| 369 diff -u -r1.12 hashmgr.cxx | |
| 370 --- src/hunspell/hashmgr.cxx 23 Jun 2011 09:21:50 -0000 1.12 | |
| 371 +++ src/hunspell/hashmgr.cxx 29 May 2014 01:05:07 -0000 | |
| 372 @@ -12,8 +12,14 @@ | |
| 373 | |
| 374 // build a hash table from a munched word list | |
| 375 | |
| 376 +#ifdef HUNSPELL_CHROME_CLIENT | |
| 377 +HashMgr::HashMgr(hunspell::BDictReader* reader) | |
| 378 +{ | |
| 379 + bdict_reader = reader; | |
| 380 +#else | |
| 381 HashMgr::HashMgr(const char * tpath, const char * apath, const char * key) | |
| 382 { | |
| 383 +#endif | |
| 384 tablesize = 0; | |
| 385 tableptr = NULL; | |
| 386 flag_mode = FLAG_CHAR; | |
| 387 @@ -31,8 +37,14 @@ | |
| 388 numaliasm = 0; | |
| 389 aliasm = NULL; | |
| 390 forbiddenword = FORBIDDENWORD; // forbidden word signing flag | |
| 391 +#ifdef HUNSPELL_CHROME_CLIENT | |
| 392 + // No tables to load, just the AF lines. | |
| 393 + load_config(NULL, NULL); | |
| 394 + int ec = LoadAFLines(); | |
| 395 +#else | |
| 396 load_config(apath, key); | |
| 397 int ec = load_tables(tpath, key); | |
| 398 +#endif | |
| 399 if (ec) { | |
| 400 /* error condition - what should we do here */ | |
| 401 HUNSPELL_WARNING(stderr, "Hash Manager Error : %d\n",ec); | |
| 402 @@ -91,15 +103,58 @@ | |
| 403 if (ignorechars) free(ignorechars); | |
| 404 if (ignorechars_utf16) free(ignorechars_utf16); | |
| 405 | |
| 406 +#ifdef HUNSPELL_CHROME_CLIENT | |
| 407 + EmptyHentryCache(); | |
| 408 + for (std::vector<std::string*>::iterator it = pointer_to_strings_.begin(); | |
| 409 + it != pointer_to_strings_.end(); ++it) { | |
| 410 + delete *it; | |
| 411 + } | |
| 412 +#endif | |
| 413 #ifdef MOZILLA_CLIENT | |
| 414 delete [] csconv; | |
| 415 #endif | |
| 416 } | |
| 417 | |
| 418 +#ifdef HUNSPELL_CHROME_CLIENT | |
| 419 +void HashMgr::EmptyHentryCache() { | |
| 420 + // We need to delete each cache entry, and each additional one in the linked | |
| 421 + // list of homonyms. | |
| 422 + for (HEntryCache::iterator i = hentry_cache.begin(); | |
| 423 + i != hentry_cache.end(); ++i) { | |
| 424 + hentry* cur = i->second; | |
| 425 + while (cur) { | |
| 426 + hentry* next = cur->next_homonym; | |
| 427 + DeleteHashEntry(cur); | |
| 428 + cur = next; | |
| 429 + } | |
| 430 + } | |
| 431 + hentry_cache.clear(); | |
| 432 +} | |
| 433 +#endif | |
| 434 + | |
| 435 // lookup a root word in the hashtable | |
| 436 | |
| 437 struct hentry * HashMgr::lookup(const char *word) const | |
| 438 { | |
| 439 +#ifdef HUNSPELL_CHROME_CLIENT | |
| 440 + int affix_ids[hunspell::BDict::MAX_AFFIXES_PER_WORD]; | |
| 441 + int affix_count = bdict_reader->FindWord(word, affix_ids); | |
| 442 + if (affix_count == 0) { // look for custom added word | |
| 443 + std::map<base::StringPiece, int>::const_iterator iter = | |
| 444 + custom_word_to_affix_id_map_.find(word); | |
| 445 + if (iter != custom_word_to_affix_id_map_.end()) { | |
| 446 + affix_count = 1; | |
| 447 + affix_ids[0] = iter->second; | |
| 448 + } | |
| 449 + } | |
| 450 + | |
| 451 + static const int kMaxWordLen = 128; | |
| 452 + static char word_buf[kMaxWordLen]; | |
| 453 + // To take account of null-termination, we use upto 127. | |
| 454 + strncpy(word_buf, word, kMaxWordLen - 1); | |
| 455 + | |
| 456 + return AffixIDsToHentry(word_buf, affix_ids, affix_count); | |
| 457 +#else | |
| 458 struct hentry * dp; | |
| 459 if (tableptr) { | |
| 460 dp = tableptr[hash(word)]; | |
| 461 @@ -109,12 +164,14 @@ | |
| 462 } | |
| 463 } | |
| 464 return NULL; | |
| 465 +#endif | |
| 466 } | |
| 467 | |
| 468 // add a word to the hash table (private) | |
| 469 int HashMgr::add_word(const char * word, int wbl, int wcl, unsigned short * aff
, | |
| 470 int al, const char * desc, bool onlyupcase) | |
| 471 { | |
| 472 +#ifndef HUNSPELL_CHROME_CLIENT | |
| 473 bool upcasehomonym = false; | |
| 474 int descl = desc ? (aliasm ? sizeof(short) : strlen(desc) + 1) : 0; | |
| 475 // variable-length hash record with word and optional fields | |
| 476 @@ -206,6 +263,17 @@ | |
| 477 if (hp->astr) free(hp->astr); | |
| 478 free(hp); | |
| 479 } | |
| 480 +#else | |
| 481 + std::map<base::StringPiece, int>::iterator iter = | |
| 482 + custom_word_to_affix_id_map_.find(word); | |
| 483 + if(iter == custom_word_to_affix_id_map_.end()) { // word needs to be added | |
| 484 + std::string* new_string_word = new std::string(word); | |
| 485 + pointer_to_strings_.push_back(new_string_word); | |
| 486 + base::StringPiece sp(*(new_string_word)); | |
| 487 + custom_word_to_affix_id_map_[sp] = 0; // no affixes for custom words | |
| 488 + return 1; | |
| 489 + } | |
| 490 +#endif | |
| 491 return 0; | |
| 492 } | |
| 493 | |
| 494 @@ -256,6 +324,12 @@ | |
| 495 // remove word (personal dictionary function for standalone applications) | |
| 496 int HashMgr::remove(const char * word) | |
| 497 { | |
| 498 +#ifdef HUNSPELL_CHROME_CLIENT | |
| 499 + std::map<base::StringPiece, int>::iterator iter = | |
| 500 + custom_word_to_affix_id_map_.find(word); | |
| 501 + if (iter != custom_word_to_affix_id_map_.end()) | |
| 502 + custom_word_to_affix_id_map_.erase(iter); | |
| 503 +#else | |
| 504 struct hentry * dp = lookup(word); | |
| 505 while (dp) { | |
| 506 if (dp->alen == 0 || !TESTAFF(dp->astr, forbiddenword, dp->alen)) { | |
| 507 @@ -270,6 +344,7 @@ | |
| 508 } | |
| 509 dp = dp->next_homonym; | |
| 510 } | |
| 511 +#endif | |
| 512 return 0; | |
| 513 } | |
| 514 | |
| 515 @@ -339,6 +414,44 @@ | |
| 516 // initialize: col=-1; hp = NULL; hp = walk_hashtable(&col, hp); | |
| 517 struct hentry * HashMgr::walk_hashtable(int &col, struct hentry * hp) const | |
| 518 { | |
| 519 +#ifdef HUNSPELL_CHROME_CLIENT | |
| 520 + // Return NULL if dictionary is not valid. | |
| 521 + if (!bdict_reader->IsValid()) | |
| 522 + return NULL; | |
| 523 + | |
| 524 + // This function is only ever called by one place and not nested. We can | |
| 525 + // therefore keep static state between calls and use |col| as a "reset" flag | |
| 526 + // to avoid changing the API. It is set to -1 for the first call. | |
| 527 + // Allocate the iterator on the heap to prevent an exit time destructor. | |
| 528 + static hunspell::WordIterator& word_iterator = | |
| 529 + *new hunspell::WordIterator(bdict_reader->GetAllWordIterator()); | |
| 530 + if (col < 0) { | |
| 531 + col = 1; | |
| 532 + word_iterator = bdict_reader->GetAllWordIterator(); | |
| 533 + } | |
| 534 + | |
| 535 + int affix_ids[hunspell::BDict::MAX_AFFIXES_PER_WORD]; | |
| 536 + static const int kMaxWordLen = 128; | |
| 537 + static char word[kMaxWordLen]; | |
| 538 + int affix_count = word_iterator.Advance(word, kMaxWordLen, affix_ids); | |
| 539 + if (affix_count == 0) | |
| 540 + return NULL; | |
| 541 + short word_len = static_cast<short>(strlen(word)); | |
| 542 + | |
| 543 + // Since hunspell 1.2.8, an hentry struct becomes a variable-length struct, | |
| 544 + // i.e. a struct which uses its array 'word[1]' as a variable-length array. | |
| 545 + // As noted above, this function is not nested. So, we just use a static | |
| 546 + // struct which consists of an hentry and a char[kMaxWordLen], and initialize | |
| 547 + // the static struct and return it for now. | |
| 548 + // No need to create linked lists for the extra affixes. | |
| 549 + static struct { | |
| 550 + hentry entry; | |
| 551 + char word[kMaxWordLen]; | |
| 552 + } hash_entry; | |
| 553 + | |
| 554 + return InitHashEntry(&hash_entry.entry, sizeof(hash_entry), | |
| 555 + &word[0], word_len, affix_ids[0]); | |
| 556 +#else | |
| 557 if (hp && hp->next != NULL) return hp->next; | |
| 558 for (col++; col < tablesize; col++) { | |
| 559 if (tableptr[col]) return tableptr[col]; | |
| 560 @@ -346,11 +459,13 @@ | |
| 561 // null at end and reset to start | |
| 562 col = -1; | |
| 563 return NULL; | |
| 564 +#endif | |
| 565 } | |
| 566 | |
| 567 // load a munched word list and build a hash table on the fly | |
| 568 int HashMgr::load_tables(const char * tpath, const char * key) | |
| 569 { | |
| 570 +#ifndef HUNSPELL_CHROME_CLIENT | |
| 571 int al; | |
| 572 char * ap; | |
| 573 char * dp; | |
| 574 @@ -471,6 +586,7 @@ | |
| 575 } | |
| 576 | |
| 577 delete dict; | |
| 578 +#endif | |
| 579 return 0; | |
| 580 } | |
| 581 | |
| 582 @@ -479,6 +595,9 @@ | |
| 583 | |
| 584 int HashMgr::hash(const char * word) const | |
| 585 { | |
| 586 +#ifdef HUNSPELL_CHROME_CLIENT | |
| 587 + return 0; | |
| 588 +#else | |
| 589 long hv = 0; | |
| 590 for (int i=0; i < 4 && *word != 0; i++) | |
| 591 hv = (hv << 8) | (*word++); | |
| 592 @@ -487,6 +606,7 @@ | |
| 593 hv ^= (*word++); | |
| 594 } | |
| 595 return (unsigned long) hv % tablesize; | |
| 596 +#endif | |
| 597 } | |
| 598 | |
| 599 int HashMgr::decode_flags(unsigned short ** result, char * flags, FileMgr * af)
{ | |
| 600 @@ -607,7 +727,12 @@ | |
| 601 int firstline = 1; | |
| 602 | |
| 603 // open the affix file | |
| 604 +#ifdef HUNSPELL_CHROME_CLIENT | |
| 605 + hunspell::LineIterator iterator = bdict_reader->GetOtherLineIterator(); | |
| 606 + FileMgr * afflst = new FileMgr(&iterator); | |
| 607 +#else | |
| 608 FileMgr * afflst = new FileMgr(affpath, key); | |
| 609 +#endif | |
| 610 if (!afflst) { | |
| 611 HUNSPELL_WARNING(stderr, "Error - could not open affix description file %s\
n",affpath); | |
| 612 return 1; | |
| 613 @@ -802,6 +927,121 @@ | |
| 614 return 0; | |
| 615 } | |
| 616 | |
| 617 +#ifdef HUNSPELL_CHROME_CLIENT | |
| 618 +int HashMgr::LoadAFLines() | |
| 619 +{ | |
| 620 + utf8 = 1; // We always use UTF-8. | |
| 621 + | |
| 622 + // Read in all the AF lines which tell us the rules for each affix group ID. | |
| 623 + hunspell::LineIterator iterator = bdict_reader->GetAfLineIterator(); | |
| 624 + FileMgr afflst(&iterator); | |
| 625 + while (char* line = afflst.getline()) { | |
| 626 + int rv = parse_aliasf(line, &afflst); | |
| 627 + if (rv) | |
| 628 + return rv; | |
| 629 + } | |
| 630 + | |
| 631 + return 0; | |
| 632 +} | |
| 633 + | |
| 634 +hentry* HashMgr::InitHashEntry(hentry* entry, | |
| 635 + size_t item_size, | |
| 636 + const char* word, | |
| 637 + int word_length, | |
| 638 + int affix_index) const { | |
| 639 + // Return if the given buffer doesn't have enough space for a hentry struct | |
| 640 + // or the given word is too long. | |
| 641 + // Our BDICT cannot handle words longer than (128 - 1) bytes. So, it is | |
| 642 + // better to return an error if the given word is too long and prevent | |
| 643 + // an unexpected result caused by a long word. | |
| 644 + const int kMaxWordLen = 128; | |
| 645 + if (item_size < sizeof(hentry) + word_length + 1 || | |
| 646 + word_length >= kMaxWordLen) | |
| 647 + return NULL; | |
| 648 + | |
| 649 + // Initialize a hentry struct with the given parameters, and | |
| 650 + // append the given string at the end of this hentry struct. | |
| 651 + memset(entry, 0, item_size); | |
| 652 + FileMgr af(NULL); | |
| 653 + entry->alen = static_cast<short>( | |
| 654 + const_cast<HashMgr*>(this)->get_aliasf(affix_index, &entry->astr, &af)); | |
| 655 + entry->blen = static_cast<unsigned char>(word_length); | |
| 656 + memcpy(&entry->word, word, word_length); | |
| 657 + | |
| 658 + return entry; | |
| 659 +} | |
| 660 + | |
| 661 +hentry* HashMgr::CreateHashEntry(const char* word, | |
| 662 + int word_length, | |
| 663 + int affix_index) const { | |
| 664 + // Return if the given word is too long. | |
| 665 + // (See the comment in HashMgr::InitHashEntry().) | |
| 666 + const int kMaxWordLen = 128; | |
| 667 + if (word_length >= kMaxWordLen) | |
| 668 + return NULL; | |
| 669 + | |
| 670 + const size_t kEntrySize = sizeof(hentry) + word_length + 1; | |
| 671 + struct hentry* entry = reinterpret_cast<hentry*>(malloc(kEntrySize)); | |
| 672 + if (entry) | |
| 673 + InitHashEntry(entry, kEntrySize, word, word_length, affix_index); | |
| 674 + | |
| 675 + return entry; | |
| 676 +} | |
| 677 + | |
| 678 +void HashMgr::DeleteHashEntry(hentry* entry) const { | |
| 679 + free(entry); | |
| 680 +} | |
| 681 + | |
| 682 +hentry* HashMgr::AffixIDsToHentry(char* word, | |
| 683 + int* affix_ids, | |
| 684 + int affix_count) const | |
| 685 +{ | |
| 686 + if (affix_count == 0) | |
| 687 + return NULL; | |
| 688 + | |
| 689 + HEntryCache& cache = const_cast<HashMgr*>(this)->hentry_cache; | |
| 690 + std::string std_word(word); | |
| 691 + HEntryCache::iterator found = cache.find(std_word); | |
| 692 + if (found != cache.end()) { | |
| 693 + // We must return an existing hentry for the same word if we've previously | |
| 694 + // handed one out. Hunspell will compare pointers in some cases to see if | |
| 695 + // two words it has found are the same. | |
| 696 + return found->second; | |
| 697 + } | |
| 698 + | |
| 699 + short word_len = static_cast<short>(strlen(word)); | |
| 700 + | |
| 701 + // We can get a number of prefixes per word. There will normally be only one, | |
| 702 + // but if not, there will be a linked list of "hentry"s for the "homonym"s | |
| 703 + // for the word. | |
| 704 + struct hentry* first_he = NULL; | |
| 705 + struct hentry* prev_he = NULL; // For making linked list. | |
| 706 + for (int i = 0; i < affix_count; i++) { | |
| 707 + struct hentry* he = CreateHashEntry(word, word_len, affix_ids[i]); | |
| 708 + if (!he) | |
| 709 + break; | |
| 710 + if (i == 0) | |
| 711 + first_he = he; | |
| 712 + if (prev_he) | |
| 713 + prev_he->next_homonym = he; | |
| 714 + prev_he = he; | |
| 715 + } | |
| 716 + | |
| 717 + cache[std_word] = first_he; // Save this word in the cache for later. | |
| 718 + return first_he; | |
| 719 +} | |
| 720 + | |
| 721 +hentry* HashMgr::GetHentryFromHEntryCache(char* word) { | |
| 722 + HEntryCache& cache = const_cast<HashMgr*>(this)->hentry_cache; | |
| 723 + std::string std_word(word); | |
| 724 + HEntryCache::iterator found = cache.find(std_word); | |
| 725 + if (found != cache.end()) | |
| 726 + return found->second; | |
| 727 + else | |
| 728 + return NULL; | |
| 729 +} | |
| 730 +#endif | |
| 731 + | |
| 732 int HashMgr::is_aliasf() { | |
| 733 return (aliasf != NULL); | |
| 734 } | |
| 735 Index: src/hunspell/hashmgr.hxx | |
| 736 =================================================================== | |
| 737 RCS file: /cvsroot/hunspell/hunspell/src/hunspell/hashmgr.hxx,v | |
| 738 retrieving revision 1.3 | |
| 739 diff -u -r1.3 hashmgr.hxx | |
| 740 --- src/hunspell/hashmgr.hxx 15 Apr 2010 11:22:08 -0000 1.3 | |
| 741 +++ src/hunspell/hashmgr.hxx 29 May 2014 01:05:07 -0000 | |
| 742 @@ -8,10 +8,25 @@ | |
| 743 #include "htypes.hxx" | |
| 744 #include "filemgr.hxx" | |
| 745 | |
| 746 +#ifdef HUNSPELL_CHROME_CLIENT | |
| 747 +#include <string> | |
| 748 +#include <map> | |
| 749 + | |
| 750 +#include "base/stl_util.h" | |
| 751 +#include "base/strings/string_piece.h" | |
| 752 +#include "third_party/hunspell/google/bdict_reader.h" | |
| 753 +#endif | |
| 754 + | |
| 755 enum flag { FLAG_CHAR, FLAG_LONG, FLAG_NUM, FLAG_UNI }; | |
| 756 | |
| 757 class LIBHUNSPELL_DLL_EXPORTED HashMgr | |
| 758 { | |
| 759 +#ifdef HUNSPELL_CHROME_CLIENT | |
| 760 + // Not owned by this class, owned by the Hunspell object. | |
| 761 + hunspell::BDictReader* bdict_reader; | |
| 762 + std::map<base::StringPiece, int> custom_word_to_affix_id_map_; | |
| 763 + std::vector<std::string*> pointer_to_strings_; | |
| 764 +#endif | |
| 765 int tablesize; | |
| 766 struct hentry ** tableptr; | |
| 767 int userword; | |
| 768 @@ -34,7 +49,23 @@ | |
| 769 | |
| 770 | |
| 771 public: | |
| 772 +#ifdef HUNSPELL_CHROME_CLIENT | |
| 773 + HashMgr(hunspell::BDictReader* reader); | |
| 774 + | |
| 775 + // Return the hentry corresponding to the given word. Returns NULL if the | |
| 776 + // word is not there in the cache. | |
| 777 + hentry* GetHentryFromHEntryCache(char* word); | |
| 778 + | |
| 779 + // Called before we do a new operation. This will empty the cache of pointers | |
| 780 + // to hentries that we have cached. In Chrome, we make these on-demand, but | |
| 781 + // they must live as long as the single spellcheck operation that they're par
t | |
| 782 + // of since Hunspell will save pointers to various ones as it works. | |
| 783 + // | |
| 784 + // This function allows that cache to be emptied and not grow infinitely. | |
| 785 + void EmptyHentryCache(); | |
| 786 +#else | |
| 787 HashMgr(const char * tpath, const char * apath, const char * key = NULL); | |
| 788 +#endif | |
| 789 ~HashMgr(); | |
| 790 | |
| 791 struct hentry * lookup(const char *) const; | |
| 792 @@ -59,6 +90,40 @@ | |
| 793 int al, const char * desc, bool onlyupcase); | |
| 794 int load_config(const char * affpath, const char * key); | |
| 795 int parse_aliasf(char * line, FileMgr * af); | |
| 796 + | |
| 797 +#ifdef HUNSPELL_CHROME_CLIENT | |
| 798 + // Loads the AF lines from a BDICT. | |
| 799 + // A BDICT file compresses its AF lines to save memory. | |
| 800 + // This function decompresses each AF line and call parse_aliasf(). | |
| 801 + int LoadAFLines(); | |
| 802 + | |
| 803 + // Helper functions that create a new hentry struct, initialize it, and | |
| 804 + // delete it. | |
| 805 + // These functions encapsulate non-trivial operations in creating and | |
| 806 + // initializing a hentry struct from BDICT data to avoid changing code so muc
h | |
| 807 + // even when a hentry struct is changed. | |
| 808 + hentry* InitHashEntry(hentry* entry, | |
| 809 + size_t item_size, | |
| 810 + const char* word, | |
| 811 + int word_length, | |
| 812 + int affix_index) const; | |
| 813 + hentry* CreateHashEntry(const char* word, | |
| 814 + int word_length, | |
| 815 + int affix_index) const; | |
| 816 + void DeleteHashEntry(hentry* entry) const; | |
| 817 + | |
| 818 + // Converts the list of affix IDs to a linked list of hentry structures. The | |
| 819 + // hentry structures will point to the given word. The returned pointer will | |
| 820 + // be a statically allocated variable that will change for the next call. The | |
| 821 + // |word| buffer must be the same. | |
| 822 + hentry* AffixIDsToHentry(char* word, int* affix_ids, int affix_count) const; | |
| 823 + | |
| 824 + // See EmptyHentryCache above. Note that each one is actually a linked list | |
| 825 + // followed by the homonym pointer. | |
| 826 + typedef std::map<std::string, hentry*> HEntryCache; | |
| 827 + HEntryCache hentry_cache; | |
| 828 +#endif | |
| 829 + | |
| 830 int add_hidden_capitalized_word(char * word, int wbl, int wcl, | |
| 831 unsigned short * flags, int al, char * dp, int captype); | |
| 832 int parse_aliasm(char * line, FileMgr * af); | |
| 833 Index: src/hunspell/htypes.hxx | |
| 834 =================================================================== | |
| 835 RCS file: /cvsroot/hunspell/hunspell/src/hunspell/htypes.hxx,v | |
| 836 retrieving revision 1.3 | |
| 837 diff -u -r1.3 htypes.hxx | |
| 838 --- src/hunspell/htypes.hxx 6 Sep 2010 07:58:53 -0000 1.3 | |
| 839 +++ src/hunspell/htypes.hxx 29 May 2014 01:05:07 -0000 | |
| 840 @@ -1,6 +1,16 @@ | |
| 841 #ifndef _HTYPES_HXX_ | |
| 842 #define _HTYPES_HXX_ | |
| 843 | |
| 844 +#ifdef HUNSPELL_CHROME_CLIENT | |
| 845 +// This is a workaround for preventing errors in parsing Turkish BDICs, which | |
| 846 +// contain very long AF lines (~ 12,000 chars). | |
| 847 +// TODO(hbono) change the HashMgr::parse_aliasf() function to be able to parse | |
| 848 +// longer lines than MAXDELEN. | |
| 849 +#define MAXDELEN (8192 * 2) | |
| 850 +#else | |
| 851 +#define MAXDELEN 8192 | |
| 852 +#endif // HUNSPELL_CHROME_CLIENT | |
| 853 + | |
| 854 #define ROTATE_LEN 5 | |
| 855 | |
| 856 #define ROTATE(v,q) \ | |
| 857 Index: src/hunspell/hunspell.cxx | |
| 858 =================================================================== | |
| 859 RCS file: /cvsroot/hunspell/hunspell/src/hunspell/hunspell.cxx,v | |
| 860 retrieving revision 1.29 | |
| 861 diff -u -r1.29 hunspell.cxx | |
| 862 --- src/hunspell/hunspell.cxx 23 Jun 2011 09:21:50 -0000 1.29 | |
| 863 +++ src/hunspell/hunspell.cxx 29 May 2014 01:05:07 -0000 | |
| 864 @@ -7,20 +7,37 @@ | |
| 865 | |
| 866 #include "hunspell.hxx" | |
| 867 #include "hunspell.h" | |
| 868 +#ifndef HUNSPELL_CHROME_CLIENT | |
| 869 #ifndef MOZILLA_CLIENT | |
| 870 # include "config.h" | |
| 871 #endif | |
| 872 +#endif | |
| 873 #include "csutil.hxx" | |
| 874 | |
| 875 +#ifdef HUNSPELL_CHROME_CLIENT | |
| 876 +Hunspell::Hunspell(const unsigned char* bdict_data, size_t bdict_length) | |
| 877 +#else | |
| 878 Hunspell::Hunspell(const char * affpath, const char * dpath, const char * key) | |
| 879 +#endif | |
| 880 { | |
| 881 encoding = NULL; | |
| 882 csconv = NULL; | |
| 883 utf8 = 0; | |
| 884 complexprefixes = 0; | |
| 885 +#ifndef HUNSPELL_CHROME_CLIENT | |
| 886 affixpath = mystrdup(affpath); | |
| 887 +#endif | |
| 888 maxdic = 0; | |
| 889 | |
| 890 +#ifdef HUNSPELL_CHROME_CLIENT | |
| 891 + bdict_reader = new hunspell::BDictReader; | |
| 892 + bdict_reader->Init(bdict_data, bdict_length); | |
| 893 + | |
| 894 + pHMgr[0] = new HashMgr(bdict_reader); | |
| 895 + if (pHMgr[0]) maxdic = 1; | |
| 896 + | |
| 897 + pAMgr = new AffixMgr(bdict_reader, pHMgr, &maxdic); | |
| 898 +#else | |
| 899 /* first set up the hash manager */ | |
| 900 pHMgr[0] = new HashMgr(dpath, affpath, key); | |
| 901 if (pHMgr[0]) maxdic = 1; | |
| 902 @@ -28,6 +45,7 @@ | |
| 903 /* next set up the affix manager */ | |
| 904 /* it needs access to the hash manager lookup methods */ | |
| 905 pAMgr = new AffixMgr(affpath, pHMgr, &maxdic, key); | |
| 906 +#endif | |
| 907 | |
| 908 /* get the preferred try string and the dictionary */ | |
| 909 /* encoding from the Affix Manager for that dictionary */ | |
| 910 @@ -41,7 +59,11 @@ | |
| 911 wordbreak = pAMgr->get_breaktable(); | |
| 912 | |
| 913 /* and finally set up the suggestion manager */ | |
| 914 +#ifdef HUNSPELL_CHROME_CLIENT | |
| 915 + pSMgr = new SuggestMgr(bdict_reader, try_string, MAXSUGGESTION, pAMgr); | |
| 916 +#else | |
| 917 pSMgr = new SuggestMgr(try_string, MAXSUGGESTION, pAMgr); | |
| 918 +#endif | |
| 919 if (try_string) free(try_string); | |
| 920 } | |
| 921 | |
| 922 @@ -59,10 +81,16 @@ | |
| 923 csconv= NULL; | |
| 924 if (encoding) free(encoding); | |
| 925 encoding = NULL; | |
| 926 +#ifdef HUNSPELL_CHROME_CLIENT | |
| 927 + if (bdict_reader) delete bdict_reader; | |
| 928 + bdict_reader = NULL; | |
| 929 +#else | |
| 930 if (affixpath) free(affixpath); | |
| 931 affixpath = NULL; | |
| 932 +#endif | |
| 933 } | |
| 934 | |
| 935 +#ifndef HUNSPELL_CHROME_CLIENT | |
| 936 // load extra dictionaries | |
| 937 int Hunspell::add_dic(const char * dpath, const char * key) { | |
| 938 if (maxdic == MAXDIC || !affixpath) return 1; | |
| 939 @@ -70,6 +98,7 @@ | |
| 940 if (pHMgr[maxdic]) maxdic++; else return 1; | |
| 941 return 0; | |
| 942 } | |
| 943 +#endif | |
| 944 | |
| 945 // make a copy of src at destination while removing all leading | |
| 946 // blanks and removing any trailing periods after recording | |
| 947 @@ -322,6 +351,9 @@ | |
| 948 | |
| 949 int Hunspell::spell(const char * word, int * info, char ** root) | |
| 950 { | |
| 951 +#ifdef HUNSPELL_CHROME_CLIENT | |
| 952 + if (pHMgr[0]) pHMgr[0]->EmptyHentryCache(); | |
| 953 +#endif | |
| 954 struct hentry * rv=NULL; | |
| 955 // need larger vector. For example, Turkish capital letter I converted a | |
| 956 // 2-byte UTF-8 character (dotless i) by mkallsmall. | |
| 957 @@ -586,6 +618,13 @@ | |
| 958 if (!len) | |
| 959 return NULL; | |
| 960 | |
| 961 +#ifdef HUNSPELL_CHROME_CLIENT | |
| 962 + // We need to check if the word length is valid to make coverity (Event | |
| 963 + // fixed_size_dest: Possible overrun of N byte fixed size buffer) happy. | |
| 964 + if ((utf8 && strlen(word) >= MAXWORDUTF8LEN) || (!utf8 && strlen(word) >= MAX
WORDLEN)) | |
| 965 + return NULL; | |
| 966 +#endif | |
| 967 + | |
| 968 // word reversing wrapper for complex prefixes | |
| 969 if (complexprefixes) { | |
| 970 if (word != w2) { | |
| 971 @@ -675,6 +714,9 @@ | |
| 972 | |
| 973 int Hunspell::suggest(char*** slst, const char * word) | |
| 974 { | |
| 975 +#ifdef HUNSPELL_CHROME_CLIENT | |
| 976 + if (pHMgr[0]) pHMgr[0]->EmptyHentryCache(); | |
| 977 +#endif | |
| 978 int onlycmpdsug = 0; | |
| 979 char cw[MAXWORDUTF8LEN]; | |
| 980 char wspace[MAXWORDUTF8LEN]; | |
| 981 @@ -1921,13 +1963,21 @@ | |
| 982 | |
| 983 Hunhandle *Hunspell_create(const char * affpath, const char * dpath) | |
| 984 { | |
| 985 +#ifdef HUNSPELL_CHROME_CLIENT | |
| 986 + return NULL; | |
| 987 +#else | |
| 988 return (Hunhandle*)(new Hunspell(affpath, dpath)); | |
| 989 +#endif | |
| 990 } | |
| 991 | |
| 992 Hunhandle *Hunspell_create_key(const char * affpath, const char * dpath, | |
| 993 const char * key) | |
| 994 { | |
| 995 +#ifdef HUNSPELL_CHROME_CLIENT | |
| 996 + return NULL; | |
| 997 +#else | |
| 998 return (Hunhandle*)(new Hunspell(affpath, dpath, key)); | |
| 999 +#endif | |
| 1000 } | |
| 1001 | |
| 1002 void Hunspell_destroy(Hunhandle *pHunspell) | |
| 1003 Index: src/hunspell/hunspell.hxx | |
| 1004 =================================================================== | |
| 1005 RCS file: /cvsroot/hunspell/hunspell/src/hunspell/hunspell.hxx,v | |
| 1006 retrieving revision 1.6 | |
| 1007 diff -u -r1.6 hunspell.hxx | |
| 1008 --- src/hunspell/hunspell.hxx 21 Jan 2011 17:30:41 -0000 1.6 | |
| 1009 +++ src/hunspell/hunspell.hxx 29 May 2014 01:05:07 -0000 | |
| 1010 @@ -5,6 +5,10 @@ | |
| 1011 #include "suggestmgr.hxx" | |
| 1012 #include "langnum.hxx" | |
| 1013 | |
| 1014 +#ifdef HUNSPELL_CHROME_CLIENT | |
| 1015 +#include "third_party/hunspell/google/bdict_reader.h" | |
| 1016 +#endif | |
| 1017 + | |
| 1018 #define SPELL_XML "<?xml?>" | |
| 1019 | |
| 1020 #define MAXDIC 20 | |
| 1021 @@ -23,7 +27,9 @@ | |
| 1022 HashMgr* pHMgr[MAXDIC]; | |
| 1023 int maxdic; | |
| 1024 SuggestMgr* pSMgr; | |
| 1025 +#ifndef HUNSPELL_CHROME_CLIENT // We are using BDict instead. | |
| 1026 char * affixpath; | |
| 1027 +#endif | |
| 1028 char * encoding; | |
| 1029 struct cs_info * csconv; | |
| 1030 int langnum; | |
| 1031 @@ -31,17 +37,28 @@ | |
| 1032 int complexprefixes; | |
| 1033 char** wordbreak; | |
| 1034 | |
| 1035 +#ifdef HUNSPELL_CHROME_CLIENT | |
| 1036 + // Not owned by us, owned by the Hunspell object. | |
| 1037 + hunspell::BDictReader* bdict_reader; | |
| 1038 +#endif | |
| 1039 + | |
| 1040 public: | |
| 1041 | |
| 1042 /* Hunspell(aff, dic) - constructor of Hunspell class | |
| 1043 * input: path of affix file and dictionary file | |
| 1044 */ | |
| 1045 | |
| 1046 +#ifdef HUNSPELL_CHROME_CLIENT | |
| 1047 + Hunspell(const unsigned char* bdict_data, size_t bdict_length); | |
| 1048 +#else | |
| 1049 Hunspell(const char * affpath, const char * dpath, const char * key = NULL); | |
| 1050 +#endif | |
| 1051 ~Hunspell(); | |
| 1052 | |
| 1053 +#ifndef HUNSPELL_CHROME_CLIENT | |
| 1054 /* load extra dictionaries (only dic files) */ | |
| 1055 int add_dic(const char * dpath, const char * key = NULL); | |
| 1056 +#endif | |
| 1057 | |
| 1058 /* spell(word) - spellcheck word | |
| 1059 * output: 0 = bad word, not 0 = good word | |
| 1060 Index: src/hunspell/replist.hxx | |
| 1061 =================================================================== | |
| 1062 RCS file: /cvsroot/hunspell/hunspell/src/hunspell/replist.hxx,v | |
| 1063 retrieving revision 1.2 | |
| 1064 diff -u -r1.2 replist.hxx | |
| 1065 --- src/hunspell/replist.hxx 15 Apr 2010 11:22:09 -0000 1.2 | |
| 1066 +++ src/hunspell/replist.hxx 29 May 2014 01:05:07 -0000 | |
| 1067 @@ -2,6 +2,12 @@ | |
| 1068 #ifndef _REPLIST_HXX_ | |
| 1069 #define _REPLIST_HXX_ | |
| 1070 | |
| 1071 +#ifdef HUNSPELL_CHROME_CLIENT | |
| 1072 +// Compilation issues in spellchecker.cc think near is a macro, therefore | |
| 1073 +// removing it here solves that problem. | |
| 1074 +#undef near | |
| 1075 +#endif | |
| 1076 + | |
| 1077 #include "hunvisapi.h" | |
| 1078 | |
| 1079 #include "w_char.hxx" | |
| 1080 Index: src/hunspell/suggestmgr.cxx | |
| 1081 =================================================================== | |
| 1082 RCS file: /cvsroot/hunspell/hunspell/src/hunspell/suggestmgr.cxx,v | |
| 1083 retrieving revision 1.24 | |
| 1084 diff -u -r1.24 suggestmgr.cxx | |
| 1085 --- src/hunspell/suggestmgr.cxx 14 Feb 2011 21:47:24 -0000 1.24 | |
| 1086 +++ src/hunspell/suggestmgr.cxx 29 May 2014 01:05:07 -0000 | |
| 1087 @@ -12,9 +12,114 @@ | |
| 1088 | |
| 1089 const w_char W_VLINE = { '\0', '|' }; | |
| 1090 | |
| 1091 +#ifdef HUNSPELL_CHROME_CLIENT | |
| 1092 +namespace { | |
| 1093 +// A simple class which creates temporary hentry objects which are available | |
| 1094 +// only in a scope. To conceal memory operations from SuggestMgr functions, | |
| 1095 +// this object automatically deletes all hentry objects created through | |
| 1096 +// CreateScopedHashEntry() calls in its destructor. So, the following snippet | |
| 1097 +// raises a memory error. | |
| 1098 +// | |
| 1099 +// hentry* bad_copy = NULL; | |
| 1100 +// { | |
| 1101 +// ScopedHashEntryFactory factory; | |
| 1102 +// hentry* scoped_copy = factory.CreateScopedHashEntry(0, source); | |
| 1103 +// ... | |
| 1104 +// bad_copy = scoped_copy; | |
| 1105 +// } | |
| 1106 +// if (bad_copy->word[0]) // memory for scoped_copy has been deleted! | |
| 1107 +// | |
| 1108 +// As listed in the above snippet, it is simple to use this class. | |
| 1109 +// 1. Declare an instance of this ScopedHashEntryFactory, and; | |
| 1110 +// 2. Call its CreateHashEntry() member instead of using 'new hentry' or | |
| 1111 +// 'operator='. | |
| 1112 +// | |
| 1113 +class ScopedHashEntryFactory { | |
| 1114 + public: | |
| 1115 + ScopedHashEntryFactory(); | |
| 1116 + ~ScopedHashEntryFactory(); | |
| 1117 + | |
| 1118 + // Creates a temporary copy of the given hentry struct. | |
| 1119 + // The returned copy is available only while this object is available. | |
| 1120 + // NOTE: this function just calls memcpy() in creating a copy of the given | |
| 1121 + // hentry struct, i.e. it does NOT copy objects referred by pointers of the | |
| 1122 + // given hentry struct. | |
| 1123 + hentry* CreateScopedHashEntry(int index, const hentry* source); | |
| 1124 + | |
| 1125 + private: | |
| 1126 + // A struct which encapsulates the new hentry struct introduced in hunspell | |
| 1127 + // 1.2.8. For a pointer to an hentry struct 'h', hunspell 1.2.8 stores a word | |
| 1128 + // (including a NUL character) into 'h->word[0]',...,'h->word[h->blen]' even | |
| 1129 + // though arraysize(h->word[]) is 1. Also, it changed 'astr' to a pointer so | |
| 1130 + // it can store affix flags into 'h->astr[0]',...,'h->astr[alen-1]'. To handl
e | |
| 1131 + // this new hentry struct, we define a struct which combines three values: an | |
| 1132 + // hentry struct 'hentry'; a char array 'word[kMaxWordLen]', and; an unsigned | |
| 1133 + // short array 'astr' so a hentry struct 'h' returned from | |
| 1134 + // CreateScopedHashEntry() satisfies the following equations: | |
| 1135 + // hentry* h = factory.CreateScopedHashEntry(0, source); | |
| 1136 + // h->word[0] == ((HashEntryItem*)h)->entry.word[0]. | |
| 1137 + // h->word[1] == ((HashEntryItem*)h)->word[0]. | |
| 1138 + // ... | |
| 1139 + // h->word[h->blen] == ((HashEntryItem*)h)->word[h->blen-1]. | |
| 1140 + // h->astr[0] == ((HashEntryItem*)h)->astr[0]. | |
| 1141 + // h->astr[1] == ((HashEntryItem*)h)->astr[1]. | |
| 1142 + // ... | |
| 1143 + // h->astr[h->alen-1] == ((HashEntryItem*)h)->astr[h->alen-1]. | |
| 1144 + enum { | |
| 1145 + kMaxWordLen = 128, | |
| 1146 + kMaxAffixLen = 8, | |
| 1147 + }; | |
| 1148 + struct HashEntryItem { | |
| 1149 + hentry entry; | |
| 1150 + char word[kMaxWordLen]; | |
| 1151 + unsigned short astr[kMaxAffixLen]; | |
| 1152 + }; | |
| 1153 + | |
| 1154 + HashEntryItem hash_items_[MAX_ROOTS]; | |
| 1155 +}; | |
| 1156 + | |
| 1157 +ScopedHashEntryFactory::ScopedHashEntryFactory() { | |
| 1158 + memset(&hash_items_[0], 0, sizeof(hash_items_)); | |
| 1159 +} | |
| 1160 + | |
| 1161 +ScopedHashEntryFactory::~ScopedHashEntryFactory() { | |
| 1162 +} | |
| 1163 + | |
| 1164 +hentry* ScopedHashEntryFactory::CreateScopedHashEntry(int index, | |
| 1165 + const hentry* source) { | |
| 1166 + if (index >= MAX_ROOTS || source->blen >= kMaxWordLen) | |
| 1167 + return NULL; | |
| 1168 + | |
| 1169 + // Retrieve a HashEntryItem struct from our spool, initialize it, and | |
| 1170 + // returns the address of its 'hentry' member. | |
| 1171 + size_t source_size = sizeof(hentry) + source->blen + 1; | |
| 1172 + HashEntryItem* hash_item = &hash_items_[index]; | |
| 1173 + memcpy(&hash_item->entry, source, source_size); | |
| 1174 + if (source->astr) { | |
| 1175 + hash_item->entry.alen = source->alen; | |
| 1176 + if (hash_item->entry.alen > kMaxAffixLen) | |
| 1177 + hash_item->entry.alen = kMaxAffixLen; | |
| 1178 + memcpy(hash_item->astr, source->astr, hash_item->entry.alen * sizeof(hash_i
tem->astr[0])); | |
| 1179 + hash_item->entry.astr = &hash_item->astr[0]; | |
| 1180 + } | |
| 1181 + return &hash_item->entry; | |
| 1182 +} | |
| 1183 + | |
| 1184 +} // namespace | |
| 1185 +#endif | |
| 1186 + | |
| 1187 + | |
| 1188 +#ifdef HUNSPELL_CHROME_CLIENT | |
| 1189 +SuggestMgr::SuggestMgr(hunspell::BDictReader* reader, | |
| 1190 + const char * tryme, int maxn, | |
| 1191 + AffixMgr * aptr) | |
| 1192 +{ | |
| 1193 + bdict_reader = reader; | |
| 1194 +#else | |
| 1195 SuggestMgr::SuggestMgr(const char * tryme, int maxn, | |
| 1196 AffixMgr * aptr) | |
| 1197 { | |
| 1198 +#endif | |
| 1199 | |
| 1200 // register affix manager and check in string of chars to | |
| 1201 // try when building candidate suggestions | |
| 1202 @@ -407,6 +512,49 @@ | |
| 1203 int lenr, lenp; | |
| 1204 int wl = strlen(word); | |
| 1205 if (wl < 2 || ! pAMgr) return ns; | |
| 1206 + | |
| 1207 +#ifdef HUNSPELL_CHROME_CLIENT | |
| 1208 + const char *pattern, *pattern2; | |
| 1209 + hunspell::ReplacementIterator iterator = bdict_reader->GetReplacementIterator
(); | |
| 1210 + while (iterator.GetNext(&pattern, &pattern2)) { | |
| 1211 + r = word; | |
| 1212 + lenr = strlen(pattern2); | |
| 1213 + lenp = strlen(pattern); | |
| 1214 + | |
| 1215 + // search every occurence of the pattern in the word | |
| 1216 + while ((r=strstr(r, pattern)) != NULL) { | |
| 1217 + strcpy(candidate, word); | |
| 1218 + if (r-word + lenr + strlen(r+lenp) >= MAXLNLEN) break; | |
| 1219 + strcpy(candidate+(r-word), pattern2); | |
| 1220 + strcpy(candidate+(r-word)+lenr, r+lenp); | |
| 1221 + ns = testsug(wlst, candidate, wl-lenp+lenr, ns, cpdsuggest, NULL, NUL
L); | |
| 1222 + if (ns == -1) return -1; | |
| 1223 + // check REP suggestions with space | |
| 1224 + char * sp = strchr(candidate, ' '); | |
| 1225 + if (sp) { | |
| 1226 + char * prev = candidate; | |
| 1227 + while (sp) { | |
| 1228 + *sp = '\0'; | |
| 1229 + if (checkword(prev, strlen(prev), 0, NULL, NULL)) { | |
| 1230 + int oldns = ns; | |
| 1231 + *sp = ' '; | |
| 1232 + ns = testsug(wlst, sp + 1, strlen(sp + 1), ns, cpdsuggest, NULL
, NULL); | |
| 1233 + if (ns == -1) return -1; | |
| 1234 + if (oldns < ns) { | |
| 1235 + free(wlst[ns - 1]); | |
| 1236 + wlst[ns - 1] = mystrdup(candidate); | |
| 1237 + if (!wlst[ns - 1]) return -1; | |
| 1238 + } | |
| 1239 + } | |
| 1240 + *sp = ' '; | |
| 1241 + prev = sp + 1; | |
| 1242 + sp = strchr(prev, ' '); | |
| 1243 + } | |
| 1244 + } | |
| 1245 + r++; // search for the next letter | |
| 1246 + } | |
| 1247 + } | |
| 1248 +#else | |
| 1249 int numrep = pAMgr->get_numrep(); | |
| 1250 struct replentry* reptable = pAMgr->get_reptable(); | |
| 1251 if (reptable==NULL) return ns; | |
| 1252 @@ -448,6 +596,7 @@ | |
| 1253 r++; // search for the next letter | |
| 1254 } | |
| 1255 } | |
| 1256 +#endif | |
| 1257 return ns; | |
| 1258 } | |
| 1259 | |
| 1260 @@ -678,7 +827,9 @@ | |
| 1261 // error is missing a letter it needs | |
| 1262 int SuggestMgr::forgotchar(char ** wlst, const char * word, int ns, int cpdsugg
est) | |
| 1263 { | |
| 1264 - char candidate[MAXSWUTF8L]; | |
| 1265 + // TODO(rouslan): Remove the interim change below when this patch lands: | |
| 1266 + // http://sf.net/tracker/?func=detail&aid=3595024&group_id=143754&atid=75639
5 | |
| 1267 + char candidate[MAXSWUTF8L + 4]; | |
| 1268 char * p; | |
| 1269 clock_t timelimit = clock(); | |
| 1270 int timer = MINTIMER; | |
| 1271 @@ -700,8 +851,10 @@ | |
| 1272 // error is missing a letter it needs | |
| 1273 int SuggestMgr::forgotchar_utf(char ** wlst, const w_char * word, int wl, int n
s, int cpdsuggest) | |
| 1274 { | |
| 1275 - w_char candidate_utf[MAXSWL]; | |
| 1276 - char candidate[MAXSWUTF8L]; | |
| 1277 + // TODO(rouslan): Remove the interim change below when this patch lands: | |
| 1278 + // http://sf.net/tracker/?func=detail&aid=3595024&group_id=143754&atid=75639
5 | |
| 1279 + w_char candidate_utf[MAXSWL + 1]; | |
| 1280 + char candidate[MAXSWUTF8L + 4]; | |
| 1281 w_char * p; | |
| 1282 clock_t timelimit = clock(); | |
| 1283 int timer = MINTIMER; | |
| 1284 @@ -1057,6 +1210,9 @@ | |
| 1285 | |
| 1286 struct hentry* hp = NULL; | |
| 1287 int col = -1; | |
| 1288 +#ifdef HUNSPELL_CHROME_CLIENT | |
| 1289 + ScopedHashEntryFactory hash_entry_factory; | |
| 1290 +#endif | |
| 1291 phonetable * ph = (pAMgr) ? pAMgr->get_phonetable() : NULL; | |
| 1292 char target[MAXSWUTF8L]; | |
| 1293 char candidate[MAXSWUTF8L]; | |
| 1294 @@ -1115,7 +1271,11 @@ | |
| 1295 | |
| 1296 if (sc > scores[lp]) { | |
| 1297 scores[lp] = sc; | |
| 1298 +#ifdef HUNSPELL_CHROME_CLIENT | |
| 1299 + roots[lp] = hash_entry_factory.CreateScopedHashEntry(lp, hp); | |
| 1300 +#else | |
| 1301 roots[lp] = hp; | |
| 1302 +#endif | |
| 1303 lval = sc; | |
| 1304 for (j=0; j < MAX_ROOTS; j++) | |
| 1305 if (scores[j] < lval) { | |
| 1306 @@ -1948,16 +2108,14 @@ | |
| 1307 m = strlen(s); | |
| 1308 n = strlen(s2); | |
| 1309 } | |
| 1310 - c = (char *) malloc((m + 1) * (n + 1)); | |
| 1311 - b = (char *) malloc((m + 1) * (n + 1)); | |
| 1312 + c = (char *) calloc(m + 1, n + 1); | |
| 1313 + b = (char *) calloc(m + 1, n + 1); | |
| 1314 if (!c || !b) { | |
| 1315 if (c) free(c); | |
| 1316 if (b) free(b); | |
| 1317 *result = NULL; | |
| 1318 return; | |
| 1319 } | |
| 1320 - for (i = 1; i <= m; i++) c[i*(n+1)] = 0; | |
| 1321 - for (j = 0; j <= n; j++) c[j] = 0; | |
| 1322 for (i = 1; i <= m; i++) { | |
| 1323 for (j = 1; j <= n; j++) { | |
| 1324 if ( ((utf8) && (*((short *) su+i-1) == *((short *)su2+j-1))) | |
| 1325 Index: src/hunspell/suggestmgr.hxx | |
| 1326 =================================================================== | |
| 1327 RCS file: /cvsroot/hunspell/hunspell/src/hunspell/suggestmgr.hxx,v | |
| 1328 retrieving revision 1.5 | |
| 1329 diff -u -r1.5 suggestmgr.hxx | |
| 1330 --- src/hunspell/suggestmgr.hxx 21 Jan 2011 22:10:24 -0000 1.5 | |
| 1331 +++ src/hunspell/suggestmgr.hxx 29 May 2014 01:05:07 -0000 | |
| 1332 @@ -52,7 +52,11 @@ | |
| 1333 | |
| 1334 | |
| 1335 public: | |
| 1336 +#ifdef HUNSPELL_CHROME_CLIENT | |
| 1337 + SuggestMgr(hunspell::BDictReader* reader, const char * tryme, int maxn, Affix
Mgr *aptr); | |
| 1338 +#else | |
| 1339 SuggestMgr(const char * tryme, int maxn, AffixMgr *aptr); | |
| 1340 +#endif | |
| 1341 ~SuggestMgr(); | |
| 1342 | |
| 1343 int suggest(char*** slst, const char * word, int nsug, int * onlycmpdsug); | |
| 1344 @@ -66,6 +70,10 @@ | |
| 1345 char * suggest_morph_for_spelling_error(const char * word); | |
| 1346 | |
| 1347 private: | |
| 1348 +#ifdef HUNSPELL_CHROME_CLIENT | |
| 1349 + // Not owned by us, owned by the Hunspell object. | |
| 1350 + hunspell::BDictReader* bdict_reader; | |
| 1351 +#endif | |
| 1352 int testsug(char** wlst, const char * candidate, int wl, int ns, int cpdsugg
est, | |
| 1353 int * timer, clock_t * timelimit); | |
| 1354 int checkword(const char *, int, int, int *, clock_t *); | |
| OLD | NEW |