icuSources/tools/gendict/gendict.cpp

   1 // © 2016 and later: Unicode, Inc. and others.
   2 // License & terms of use: http://www.unicode.org/copyright.html
   3 /*
   4 **********************************************************************
   5 *   Copyright (C) 2002-2016, International Business Machines
   6 *   Corporation and others.  All Rights Reserved.
   7 **********************************************************************
   8 *
   9 * File gendict.cpp
  10 */
  11
  12 #include "unicode/utypes.h"
  13 #include "unicode/uchar.h"
  14 #include "unicode/ucnv.h"
  15 #include "unicode/uniset.h"
  16 #include "unicode/unistr.h"
  17 #include "unicode/uclean.h"
  18 #include "unicode/udata.h"
  19 #include "unicode/putil.h"
  20 #include "unicode/ucharstriebuilder.h"
  21 #include "unicode/bytestriebuilder.h"
  22 #include "unicode/ucharstrie.h"
  23 #include "unicode/bytestrie.h"
  24 #include "unicode/ucnv.h"
  25 #include "unicode/ustring.h"
  26 #include "unicode/utf16.h"
  27
  28 #include "charstr.h"
  29 #include "dictionarydata.h"
  30 #include "uoptions.h"
  31 #include "unewdata.h"
  32 #include "cmemory.h"
  33 #include "uassert.h"
  34 #include "ucbuf.h"
  35 #include "toolutil.h"
  36 #include "cstring.h"
  37
  38 #include <stdio.h>
  39 #include <stdlib.h>
  40 #include <string.h>
  41
  42 #include "putilimp.h"
  43 UDate startTime;
  44
  45 static int elapsedTime() {
  46   return (int)uprv_floor((uprv_getRawUTCtime()-startTime)/1000.0);
  47 }
  48
  49 U_NAMESPACE_USE
  50
  51 static char *progName;
  52 static UOption options[]={
  53     UOPTION_HELP_H,             /* 0 */
  54     UOPTION_HELP_QUESTION_MARK, /* 1 */
  55     UOPTION_VERBOSE,            /* 2 */
  56     UOPTION_ICUDATADIR,         /* 4 */
  57     UOPTION_COPYRIGHT,          /* 5 */
  58     { "uchars", NULL, NULL, NULL, '\1', UOPT_NO_ARG, 0}, /* 6 */
  59     { "bytes", NULL, NULL, NULL, '\1', UOPT_NO_ARG, 0}, /* 7 */
  60     { "transform", NULL, NULL, NULL, '\1', UOPT_REQUIRES_ARG, 0}, /* 8 */
  61     UOPTION_QUIET,              /* 9 */
  62 };
  63
  64 enum arguments {
  65     ARG_HELP = 0,
  66     ARG_QMARK,
  67     ARG_VERBOSE,
  68     ARG_ICUDATADIR,
  69     ARG_COPYRIGHT,
  70     ARG_UCHARS,
  71     ARG_BYTES,
  72     ARG_TRANSFORM,
  73     ARG_QUIET
  74 };
  75
  76 // prints out the standard usage method describing command line arguments,
  77 // then bails out with the desired exit code
  78 static void usageAndDie(UErrorCode retCode) {
  79     fprintf((U_SUCCESS(retCode) ? stdout : stderr), "Usage: %s -trietype [-options] input-dictionary-file output-file\n", progName);
  80     fprintf((U_SUCCESS(retCode) ? stdout : stderr),
  81            "\tRead in a word list and write out a string trie dictionary\n"
  82            "options:\n"
  83            "\t-h or -? or --help  this usage text\n"
  84            "\t-V or --version     show a version message\n"
  85            "\t-c or --copyright   include a copyright notice\n"
  86            "\t-v or --verbose     turn on verbose output\n"
  87            "\t-q or --quiet       do not display warnings and progress\n"
  88            "\t-i or --icudatadir  directory for locating any needed intermediate data files,\n" // TODO: figure out if we need this option
  89            "\t                    followed by path, defaults to %s\n"
  90            "\t--uchars            output a UCharsTrie (mutually exclusive with -b!)\n"
  91            "\t--bytes             output a BytesTrie (mutually exclusive with -u!)\n"
  92            "\t--transform         the kind of transform to use (eg --transform offset-40A3,\n"
  93            "\t                    which specifies an offset transform with constant 0x40A3)\n",
  94             u_getDataDirectory());
  95     exit(retCode);
  96 }
  97
  98
  99 /* UDataInfo cf. udata.h */
 100 static UDataInfo dataInfo = {
 101     sizeof(UDataInfo),
 102     0,
 103
 104     U_IS_BIG_ENDIAN,
 105     U_CHARSET_FAMILY,
 106     U_SIZEOF_UCHAR,
 107     0,
 108
 109     { 0x44, 0x69, 0x63, 0x74 },     /* "Dict" */
 110     { 1, 0, 0, 0 },                 /* format version */
 111     { 0, 0, 0, 0 }                  /* data version */
 112 };
 113
 114 #if !UCONFIG_NO_BREAK_ITERATION
 115
 116 // A wrapper for both BytesTrieBuilder and UCharsTrieBuilder.
 117 // may want to put this somewhere in ICU, as it could be useful outside
 118 // of this tool?
 119 class DataDict {
 120 private:
 121     BytesTrieBuilder *bt;
 122     UCharsTrieBuilder *ut;
 123     UChar32 transformConstant;
 124     int32_t transformType;
 125 public:
 126     // constructs a new data dictionary. if there is an error,
 127     // it will be returned in status
 128     // isBytesTrie != 0 will produce a BytesTrieBuilder,
 129     // isBytesTrie == 0 will produce a UCharsTrieBuilder
 130     DataDict(UBool isBytesTrie, UErrorCode &status) : bt(NULL), ut(NULL),
 131         transformConstant(0), transformType(DictionaryData::TRANSFORM_NONE) {
 132         if (isBytesTrie) {
 133             bt = new BytesTrieBuilder(status);
 134         } else {
 135             ut = new UCharsTrieBuilder(status);
 136         }
 137     }
 138
 139     ~DataDict() {
 140         delete bt;
 141         delete ut;
 142     }
 143
 144 private:
 145     char transform(UChar32 c, UErrorCode &status) {
 146         if (transformType == DictionaryData::TRANSFORM_TYPE_OFFSET) {
 147             if (c == 0x200D) { return (char)0xFF; }
 148             else if (c == 0x200C) { return (char)0xFE; }
 149             int32_t delta = c - transformConstant;
 150             if (delta < 0 || 0xFD < delta) {
 151                 fprintf(stderr, "Codepoint U+%04lx out of range for --transform offset-%04lx!\n",
 152                         (long)c, (long)transformConstant);
 153                 exit(U_ILLEGAL_ARGUMENT_ERROR); // TODO: should return and print the line number
 154             }
 155             return (char)delta;
 156         } else { // no such transform type
 157             status = U_INTERNAL_PROGRAM_ERROR;
 158             return (char)c; // it should be noted this transform type will not generally work
 159         }
 160     }
 161
 162     void transform(const UnicodeString &word, CharString &buf, UErrorCode &errorCode) {
 163         UChar32 c = 0;
 164         int32_t len = word.length();
 165         for (int32_t i = 0; i < len; i += U16_LENGTH(c)) {
 166             c = word.char32At(i);
 167             buf.append(transform(c, errorCode), errorCode);
 168         }
 169     }
 170
 171 public:
 172     // sets the desired transformation data.
 173     // should be populated from a command line argument
 174     // so far the only acceptable format is offset-<hex constant>
 175     // eventually others (mask-<hex constant>?) may be enabled
 176     // more complex functions may be more difficult
 177     void setTransform(const char *t) {
 178         if (strncmp(t, "offset-", 7) == 0) {
 179             char *end;
 180             unsigned long base = uprv_strtoul(t + 7, &end, 16);
 181             if (end == (t + 7) || *end != 0 || base > 0x10FF80) {
 182                 fprintf(stderr, "Syntax for offset value in --transform offset-%s invalid!\n", t + 7);
 183                 usageAndDie(U_ILLEGAL_ARGUMENT_ERROR);
 184             }
 185             transformType = DictionaryData::TRANSFORM_TYPE_OFFSET;
 186             transformConstant = (UChar32)base;
 187         }
 188         else {
 189             fprintf(stderr, "Invalid transform specified: %s\n", t);
 190             usageAndDie(U_ILLEGAL_ARGUMENT_ERROR);
 191         }
 192     }
 193
 194     // add a word to the trie
 195     void addWord(const UnicodeString &word, int32_t value, UErrorCode &status) {
 196         if (bt) {
 197             CharString buf;
 198             transform(word, buf, status);
 199             bt->add(buf.toStringPiece(), value, status);
 200         }
 201         if (ut) { ut->add(word, value, status); }
 202     }
 203
 204     // if we are a bytestrie, give back the StringPiece representing the serialized version of us
 205     StringPiece serializeBytes(UErrorCode &status) {
 206         return bt->buildStringPiece(USTRINGTRIE_BUILD_SMALL, status);
 207     }
 208
 209     // if we are a ucharstrie, produce the UnicodeString representing the serialized version of us
 210     void serializeUChars(UnicodeString &s, UErrorCode &status) {
 211         ut->buildUnicodeString(USTRINGTRIE_BUILD_SMALL, s, status);
 212     }
 213
 214     int32_t getTransform() {
 215         return (int32_t)(transformType | transformConstant);
 216     }
 217 };
 218 #endif
 219
 220 static const UChar LINEFEED_CHARACTER = 0x000A;
 221 static const UChar CARRIAGE_RETURN_CHARACTER = 0x000D;
 222
 223 static UBool readLine(UCHARBUF *f, UnicodeString &fileLine, IcuToolErrorCode &errorCode) {
 224     int32_t lineLength;
 225     const UChar *line = ucbuf_readline(f, &lineLength, errorCode);
 226     if(line == NULL || errorCode.isFailure()) { return FALSE; }
 227     // Strip trailing CR/LF, comments, and spaces.
 228     const UChar *comment = u_memchr(line, 0x23, lineLength);  // '#'
 229     if(comment != NULL) {
 230         lineLength = (int32_t)(comment - line);
 231     } else {
 232         while(lineLength > 0 && (line[lineLength - 1] == CARRIAGE_RETURN_CHARACTER || line[lineLength - 1] == LINEFEED_CHARACTER)) { --lineLength; }
 233     }
 234     while(lineLength > 0 && u_isspace(line[lineLength - 1])) { --lineLength; }
 235     fileLine.setTo(FALSE, line, lineLength);
 236     return TRUE;
 237 }
 238
 239 //----------------------------------------------------------------------------
 240 //
 241 //  main      for gendict
 242 //
 243 //----------------------------------------------------------------------------
 244 int  main(int argc, char **argv) {
 245     //
 246     // Pick up and check the command line arguments,
 247     //    using the standard ICU tool utils option handling.
 248     //
 249     U_MAIN_INIT_ARGS(argc, argv);
 250     progName = argv[0];
 251     argc=u_parseArgs(argc, argv, UPRV_LENGTHOF(options), options);
 252     if(argc<0) {
 253         // Unrecognized option
 254         fprintf(stderr, "error in command line argument \"%s\"\n", argv[-argc]);
 255         usageAndDie(U_ILLEGAL_ARGUMENT_ERROR);
 256     }
 257
 258     if(options[ARG_HELP].doesOccur || options[ARG_QMARK].doesOccur) {
 259         //  -? or -h for help.
 260         usageAndDie(U_ZERO_ERROR);
 261     }
 262
 263     UBool verbose = options[ARG_VERBOSE].doesOccur;
 264     UBool quiet = options[ARG_QUIET].doesOccur;
 265
 266     if (argc < 3) {
 267         fprintf(stderr, "input and output file must both be specified.\n");
 268         usageAndDie(U_ILLEGAL_ARGUMENT_ERROR);
 269     }
 270     const char *outFileName  = argv[2];
 271     const char *wordFileName = argv[1];
 272
 273     startTime = uprv_getRawUTCtime(); // initialize start timer
 274
 275         if (options[ARG_ICUDATADIR].doesOccur) {
 276         u_setDataDirectory(options[ARG_ICUDATADIR].value);
 277     }
 278
 279     const char *copyright = NULL;
 280     if (options[ARG_COPYRIGHT].doesOccur) {
 281         copyright = U_COPYRIGHT_STRING;
 282     }
 283
 284     if (options[ARG_UCHARS].doesOccur == options[ARG_BYTES].doesOccur) {
 285         fprintf(stderr, "you must specify exactly one type of trie to output!\n");
 286         usageAndDie(U_ILLEGAL_ARGUMENT_ERROR);
 287     }
 288     UBool isBytesTrie = options[ARG_BYTES].doesOccur;
 289     if (isBytesTrie != options[ARG_TRANSFORM].doesOccur) {
 290         fprintf(stderr, "you must provide a transformation for a bytes trie, and must not provide one for a uchars trie!\n");
 291         usageAndDie(U_ILLEGAL_ARGUMENT_ERROR);
 292     }
 293
 294     IcuToolErrorCode status("gendict/main()");
 295
 296 #if UCONFIG_NO_BREAK_ITERATION || UCONFIG_NO_FILE_IO
 297     const char* outDir=NULL;
 298
 299     UNewDataMemory *pData;
 300     char msg[1024];
 301     UErrorCode tempstatus = U_ZERO_ERROR;
 302
 303     /* write message with just the name */ // potential for a buffer overflow here...
 304     sprintf(msg, "gendict writes dummy %s because of UCONFIG_NO_BREAK_ITERATION and/or UCONFIG_NO_FILE_IO, see uconfig.h", outFileName);
 305     fprintf(stderr, "%s\n", msg);
 306
 307     /* write the dummy data file */
 308     pData = udata_create(outDir, NULL, outFileName, &dataInfo, NULL, &tempstatus);
 309     udata_writeBlock(pData, msg, strlen(msg));
 310     udata_finish(pData, &tempstatus);
 311     return (int)tempstatus;
 312
 313 #else
 314     //  Read in the dictionary source file
 315     if (verbose) { printf("Opening file %s...\n", wordFileName); }
 316     const char *codepage = "UTF-8";
 317     UCHARBUF *f = ucbuf_open(wordFileName, &codepage, TRUE, FALSE, status);
 318     if (status.isFailure()) {
 319         fprintf(stderr, "error opening input file: ICU Error \"%s\"\n", status.errorName());
 320         exit(status.reset());
 321     }
 322     if (verbose) { printf("Initializing dictionary builder of type %s...\n", (isBytesTrie ? "BytesTrie" : "UCharsTrie")); }
 323     DataDict dict(isBytesTrie, status);
 324     if (status.isFailure()) {
 325         fprintf(stderr, "new DataDict: ICU Error \"%s\"\n", status.errorName());
 326         exit(status.reset());
 327     }
 328     if (options[ARG_TRANSFORM].doesOccur) {
 329         dict.setTransform(options[ARG_TRANSFORM].value);
 330     }
 331
 332     UnicodeString fileLine;
 333     if (verbose) { puts("Adding words to dictionary..."); }
 334     UBool hasValues = FALSE;
 335     UBool hasValuelessContents = FALSE;
 336     int lineCount = 0;
 337     int wordCount = 0;
 338     int minlen = 255;
 339     int maxlen = 0;
 340     UBool isOk = TRUE;
 341     while (readLine(f, fileLine, status)) {
 342         lineCount++;
 343         if (fileLine.isEmpty()) continue;
 344
 345         // Parse word [spaces value].
 346         int32_t keyLen;
 347         for (keyLen = 0; keyLen < fileLine.length() && !u_isspace(fileLine[keyLen]); ++keyLen) {}
 348         if (keyLen == 0) {
 349             fprintf(stderr, "Error: no word on line %i!\n", lineCount);
 350             isOk = FALSE;
 351             continue;
 352         }
 353         int32_t valueStart;
 354         for (valueStart = keyLen;
 355             valueStart < fileLine.length() && u_isspace(fileLine[valueStart]);
 356             ++valueStart) {}
 357
 358         if (keyLen < valueStart) {
 359             int32_t valueLength = fileLine.length() - valueStart;
 360             if (valueLength > 15) {
 361                 fprintf(stderr, "Error: value too long on line %i!\n", lineCount);
 362                 isOk = FALSE;
 363                 continue;
 364             }
 365             char s[16];
 366             fileLine.extract(valueStart, valueLength, s, 16, US_INV);
 367             char *end;
 368             unsigned long value = uprv_strtoul(s, &end, 0);
 369             if (end == s || *end != 0 || (int32_t)uprv_strlen(s) != valueLength || value > 0xffffffff) {
 370                 fprintf(stderr, "Error: value syntax error or value too large on line %i!\n", lineCount);
 371                 isOk = FALSE;
 372                 continue;
 373             }
 374             dict.addWord(fileLine.tempSubString(0, keyLen), (int32_t)value, status);
 375             hasValues = TRUE;
 376             wordCount++;
 377             if (keyLen < minlen) minlen = keyLen;
 378             if (keyLen > maxlen) maxlen = keyLen;
 379         } else {
 380             dict.addWord(fileLine.tempSubString(0, keyLen), 0, status);
 381             hasValuelessContents = TRUE;
 382             wordCount++;
 383             if (keyLen < minlen) minlen = keyLen;
 384             if (keyLen > maxlen) maxlen = keyLen;
 385         }
 386
 387         if (status.isFailure()) {
 388             fprintf(stderr, "ICU Error \"%s\": Failed to add word to trie at input line %d in input file\n",
 389                 status.errorName(), lineCount);
 390             exit(status.reset());
 391         }
 392     }
 393     if (verbose) { printf("Processed %d lines, added %d words, minlen %d, maxlen %d\n", lineCount, wordCount, minlen, maxlen); }
 394
 395     if (!isOk && status.isSuccess()) {
 396         status.set(U_ILLEGAL_ARGUMENT_ERROR);
 397     }
 398     if (hasValues && hasValuelessContents) {
 399         fprintf(stderr, "warning: file contained both valued and unvalued strings!\n");
 400     }
 401
 402     if (verbose) { printf("Serializing data...isBytesTrie? %d\n", isBytesTrie); }
 403     int32_t outDataSize;
 404     const void *outData;
 405     UnicodeString usp;
 406     if (isBytesTrie) {
 407         StringPiece sp = dict.serializeBytes(status);
 408         outDataSize = sp.size();
 409         outData = sp.data();
 410     } else {
 411         dict.serializeUChars(usp, status);
 412         outDataSize = usp.length() * U_SIZEOF_UCHAR;
 413         outData = usp.getBuffer();
 414     }
 415     if (status.isFailure()) {
 416         fprintf(stderr, "gendict: got failure of type %s while serializing, if U_ILLEGAL_ARGUMENT_ERROR possibly due to duplicate dictionary entries\n", status.errorName());
 417         exit(status.reset());
 418     }
 419     if (verbose) { puts("Opening output file..."); }
 420     UNewDataMemory *pData = udata_create(NULL, NULL, outFileName, &dataInfo, copyright, status);
 421     if (status.isFailure()) {
 422         fprintf(stderr, "gendict: could not open output file \"%s\", \"%s\"\n", outFileName, status.errorName());
 423         exit(status.reset());
 424     }
 425
 426     if (verbose) { puts("Writing to output file..."); }
 427     int32_t indexes[DictionaryData::IX_COUNT] = {
 428         DictionaryData::IX_COUNT * sizeof(int32_t), 0, 0, 0, 0, 0, 0, 0
 429     };
 430     int32_t size = outDataSize + indexes[DictionaryData::IX_STRING_TRIE_OFFSET];
 431     indexes[DictionaryData::IX_RESERVED1_OFFSET] = size;
 432     indexes[DictionaryData::IX_RESERVED2_OFFSET] = size;
 433     indexes[DictionaryData::IX_TOTAL_SIZE] = size;
 434
 435     indexes[DictionaryData::IX_TRIE_TYPE] = isBytesTrie ? DictionaryData::TRIE_TYPE_BYTES : DictionaryData::TRIE_TYPE_UCHARS;
 436     if (hasValues) {
 437         indexes[DictionaryData::IX_TRIE_TYPE] |= DictionaryData::TRIE_HAS_VALUES;
 438     }
 439
 440     indexes[DictionaryData::IX_TRANSFORM] = dict.getTransform();
 441     udata_writeBlock(pData, indexes, sizeof(indexes));
 442     udata_writeBlock(pData, outData, outDataSize);
 443     size_t bytesWritten = udata_finish(pData, status);
 444     if (status.isFailure()) {
 445         fprintf(stderr, "gendict: error \"%s\" writing the output file\n", status.errorName());
 446         exit(status.reset());
 447     }
 448
 449     if (bytesWritten != (size_t)size) {
 450         fprintf(stderr, "Error writing to output file \"%s\"\n", outFileName);
 451         exit(U_INTERNAL_PROGRAM_ERROR);
 452     }
 453
 454     if (!quiet) { printf("%s: done writing\t%s (%ds).\n", progName, outFileName, elapsedTime()); }
 455
 456 #ifdef TEST_GENDICT
 457     if (isBytesTrie) {
 458         BytesTrie::Iterator it(outData, outDataSize, status);
 459         while (it.hasNext()) {
 460             it.next(status);
 461             const StringPiece s = it.getString();
 462             int32_t val = it.getValue();
 463             printf("%s -> %i\n", s.data(), val);
 464         }
 465     } else {
 466         UCharsTrie::Iterator it((const UChar *)outData, outDataSize, status);
 467         while (it.hasNext()) {
 468             it.next(status);
 469             const UnicodeString s = it.getString();
 470             int32_t val = it.getValue();
 471             char tmp[1024];
 472             s.extract(0, s.length(), tmp, 1024);
 473             printf("%s -> %i\n", tmp, val);
 474         }
 475     }
 476 #endif
 477
 478     return 0;
 479 #endif /* #if !UCONFIG_NO_BREAK_ITERATION */
 480 }