icuSources/tools/gendict/gendict.cpp

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