icuSources/tools/gendict/gendict.cpp

   1 /*
   2 **********************************************************************
   3 *   Copyright (C) 2002-2014, 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 };
  59
  60 enum arguments {
  61     ARG_HELP = 0,
  62     ARG_QMARK,
  63     ARG_VERBOSE,
  64     ARG_ICUDATADIR,
  65     ARG_COPYRIGHT,
  66     ARG_UCHARS,
  67     ARG_BYTES,
  68     ARG_TRANSFORM
  69 };
  70
  71 // prints out the standard usage method describing command line arguments,
  72 // then bails out with the desired exit code
  73 static void usageAndDie(UErrorCode retCode) {
  74     fprintf((U_SUCCESS(retCode) ? stdout : stderr), "Usage: %s -trietype [-options] input-dictionary-file output-file\n", progName);
  75     fprintf((U_SUCCESS(retCode) ? stdout : stderr),
  76            "\tRead in a word list and write out a string trie dictionary\n"
  77            "options:\n"
  78            "\t-h or -? or --help  this usage text\n"
  79            "\t-V or --version     show a version message\n"
  80            "\t-c or --copyright   include a copyright notice\n"
  81            "\t-v or --verbose     turn on verbose output\n"
  82            "\t-i or --icudatadir  directory for locating any needed intermediate data files,\n" // TODO: figure out if we need this option
  83            "\t                    followed by path, defaults to %s\n"
  84            "\t--uchars            output a UCharsTrie (mutually exclusive with -b!)\n"
  85            "\t--bytes             output a BytesTrie (mutually exclusive with -u!)\n"
  86            "\t--transform         the kind of transform to use (eg --transform offset-40A3,\n"
  87            "\t                    which specifies an offset transform with constant 0x40A3)\n",
  88             u_getDataDirectory());
  89     exit(retCode);
  90 }
  91
  92
  93 /* UDataInfo cf. udata.h */
  94 static UDataInfo dataInfo = {
  95     sizeof(UDataInfo),
  96     0,
  97
  98     U_IS_BIG_ENDIAN,
  99     U_CHARSET_FAMILY,
 100     U_SIZEOF_UCHAR,
 101     0,
 102
 103     { 0x44, 0x69, 0x63, 0x74 },     /* "Dict" */
 104     { 1, 0, 0, 0 },                 /* format version */
 105     { 0, 0, 0, 0 }                  /* data version */
 106 };
 107
 108 #if !UCONFIG_NO_BREAK_ITERATION
 109
 110 // A wrapper for both BytesTrieBuilder and UCharsTrieBuilder.
 111 // may want to put this somewhere in ICU, as it could be useful outside
 112 // of this tool?
 113 class DataDict {
 114 private:
 115     BytesTrieBuilder *bt;
 116     UCharsTrieBuilder *ut;
 117     UChar32 transformConstant;
 118     int32_t transformType;
 119 public:
 120     // constructs a new data dictionary. if there is an error,
 121     // it will be returned in status
 122     // isBytesTrie != 0 will produce a BytesTrieBuilder,
 123     // isBytesTrie == 0 will produce a UCharsTrieBuilder
 124     DataDict(UBool isBytesTrie, UErrorCode &status) : bt(NULL), ut(NULL),
 125         transformConstant(0), transformType(DictionaryData::TRANSFORM_NONE) {
 126         if (isBytesTrie) {
 127             bt = new BytesTrieBuilder(status);
 128         } else {
 129             ut = new UCharsTrieBuilder(status);
 130         }
 131     }
 132
 133     ~DataDict() {
 134         delete bt;
 135         delete ut;
 136     }
 137
 138 private:
 139     char transform(UChar32 c, UErrorCode &status) {
 140         if (transformType == DictionaryData::TRANSFORM_TYPE_OFFSET) {
 141             if (c == 0x200D) { return (char)0xFF; }
 142             else if (c == 0x200C) { return (char)0xFE; }
 143             int32_t delta = c - transformConstant;
 144             if (delta < 0 || 0xFD < delta) {
 145                 fprintf(stderr, "Codepoint U+%04lx out of range for --transform offset-%04lx!\n",
 146                         (long)c, (long)transformConstant);
 147                 exit(U_ILLEGAL_ARGUMENT_ERROR); // TODO: should return and print the line number
 148             }
 149             return (char)delta;
 150         } else { // no such transform type
 151             status = U_INTERNAL_PROGRAM_ERROR;
 152             return (char)c; // it should be noted this transform type will not generally work
 153         }
 154     }
 155
 156     void transform(const UnicodeString &word, CharString &buf, UErrorCode &errorCode) {
 157         UChar32 c = 0;
 158         int32_t len = word.length();
 159         for (int32_t i = 0; i < len; i += U16_LENGTH(c)) {
 160             c = word.char32At(i);
 161             buf.append(transform(c, errorCode), errorCode);
 162         }
 163     }
 164
 165 public:
 166     // sets the desired transformation data.
 167     // should be populated from a command line argument
 168     // so far the only acceptable format is offset-<hex constant>
 169     // eventually others (mask-<hex constant>?) may be enabled
 170     // more complex functions may be more difficult
 171     void setTransform(const char *t) {
 172         if (strncmp(t, "offset-", 7) == 0) {
 173             char *end;
 174             unsigned long base = uprv_strtoul(t + 7, &end, 16);
 175             if (end == (t + 7) || *end != 0 || base > 0x10FF80) {
 176                 fprintf(stderr, "Syntax for offset value in --transform offset-%s invalid!\n", t + 7);
 177                 usageAndDie(U_ILLEGAL_ARGUMENT_ERROR);
 178             }
 179             transformType = DictionaryData::TRANSFORM_TYPE_OFFSET;
 180             transformConstant = (UChar32)base;
 181         }
 182         else {
 183             fprintf(stderr, "Invalid transform specified: %s\n", t);
 184             usageAndDie(U_ILLEGAL_ARGUMENT_ERROR);
 185         }
 186     }
 187
 188     // add a word to the trie
 189     void addWord(const UnicodeString &word, int32_t value, UErrorCode &status) {
 190         if (bt) {
 191             CharString buf;
 192             transform(word, buf, status);
 193             bt->add(buf.toStringPiece(), value, status);
 194         }
 195         if (ut) { ut->add(word, value, status); }
 196     }
 197
 198     // if we are a bytestrie, give back the StringPiece representing the serialized version of us
 199     StringPiece serializeBytes(UErrorCode &status) {
 200         return bt->buildStringPiece(USTRINGTRIE_BUILD_SMALL, status);
 201     }
 202
 203     // if we are a ucharstrie, produce the UnicodeString representing the serialized version of us
 204     void serializeUChars(UnicodeString &s, UErrorCode &status) {
 205         ut->buildUnicodeString(USTRINGTRIE_BUILD_SMALL, s, status);
 206     }
 207
 208     int32_t getTransform() {
 209         return (int32_t)(transformType | transformConstant);
 210     }
 211 };
 212 #endif
 213
 214 static const UChar LINEFEED_CHARACTER = 0x000A;
 215 static const UChar CARRIAGE_RETURN_CHARACTER = 0x000D;
 216
 217 static UBool readLine(UCHARBUF *f, UnicodeString &fileLine, IcuToolErrorCode &errorCode) {
 218     int32_t lineLength;
 219     const UChar *line = ucbuf_readline(f, &lineLength, errorCode);
 220     if(line == NULL || errorCode.isFailure()) { return FALSE; }
 221     // Strip trailing CR/LF, comments, and spaces.
 222     const UChar *comment = u_memchr(line, 0x23, lineLength);  // '#'
 223     if(comment != NULL) {
 224         lineLength = (int32_t)(comment - line);
 225     } else {
 226         while(lineLength > 0 && (line[lineLength - 1] == CARRIAGE_RETURN_CHARACTER || line[lineLength - 1] == LINEFEED_CHARACTER)) { --lineLength; }
 227     }
 228     while(lineLength > 0 && u_isspace(line[lineLength - 1])) { --lineLength; }
 229     fileLine.setTo(FALSE, line, lineLength);
 230     return TRUE;
 231 }
 232
 233 //----------------------------------------------------------------------------
 234 //
 235 //  main      for gendict
 236 //
 237 //----------------------------------------------------------------------------
 238 int  main(int argc, char **argv) {
 239     //
 240     // Pick up and check the command line arguments,
 241     //    using the standard ICU tool utils option handling.
 242     //
 243     U_MAIN_INIT_ARGS(argc, argv);
 244     progName = argv[0];
 245     argc=u_parseArgs(argc, argv, sizeof(options)/sizeof(options[0]), options);
 246     if(argc<0) {
 247         // Unrecognized option
 248         fprintf(stderr, "error in command line argument \"%s\"\n", argv[-argc]);
 249         usageAndDie(U_ILLEGAL_ARGUMENT_ERROR);
 250     }
 251
 252     if(options[ARG_HELP].doesOccur || options[ARG_QMARK].doesOccur) {
 253         //  -? or -h for help.
 254         usageAndDie(U_ZERO_ERROR);
 255     }
 256
 257     UBool verbose = options[ARG_VERBOSE].doesOccur;
 258
 259     if (argc < 3) {
 260         fprintf(stderr, "input and output file must both be specified.\n");
 261         usageAndDie(U_ILLEGAL_ARGUMENT_ERROR);
 262     }
 263     const char *outFileName  = argv[2];
 264     const char *wordFileName = argv[1];
 265
 266     startTime = uprv_getRawUTCtime(); // initialize start timer
 267
 268         if (options[ARG_ICUDATADIR].doesOccur) {
 269         u_setDataDirectory(options[ARG_ICUDATADIR].value);
 270     }
 271
 272     const char *copyright = NULL;
 273     if (options[ARG_COPYRIGHT].doesOccur) {
 274         copyright = U_COPYRIGHT_STRING;
 275     }
 276
 277     if (options[ARG_UCHARS].doesOccur == options[ARG_BYTES].doesOccur) {
 278         fprintf(stderr, "you must specify exactly one type of trie to output!\n");
 279         usageAndDie(U_ILLEGAL_ARGUMENT_ERROR);
 280     }
 281     UBool isBytesTrie = options[ARG_BYTES].doesOccur;
 282     if (isBytesTrie != options[ARG_TRANSFORM].doesOccur) {
 283         fprintf(stderr, "you must provide a transformation for a bytes trie, and must not provide one for a uchars trie!\n");
 284         usageAndDie(U_ILLEGAL_ARGUMENT_ERROR);
 285     }
 286
 287     IcuToolErrorCode status("gendict/main()");
 288
 289 #if UCONFIG_NO_BREAK_ITERATION || UCONFIG_NO_FILE_IO
 290     const char* outDir=NULL;
 291
 292     UNewDataMemory *pData;
 293     char msg[1024];
 294     UErrorCode tempstatus = U_ZERO_ERROR;
 295
 296     /* write message with just the name */ // potential for a buffer overflow here...
 297     sprintf(msg, "gendict writes dummy %s because of UCONFIG_NO_BREAK_ITERATION and/or UCONFIG_NO_FILE_IO, see uconfig.h", outFileName);
 298     fprintf(stderr, "%s\n", msg);
 299
 300     /* write the dummy data file */
 301     pData = udata_create(outDir, NULL, outFileName, &dataInfo, NULL, &tempstatus);
 302     udata_writeBlock(pData, msg, strlen(msg));
 303     udata_finish(pData, &tempstatus);
 304     return (int)tempstatus;
 305
 306 #else
 307     //  Read in the dictionary source file
 308     if (verbose) { printf("Opening file %s...\n", wordFileName); }
 309     const char *codepage = "UTF-8";
 310     UCHARBUF *f = ucbuf_open(wordFileName, &codepage, TRUE, FALSE, status);
 311     if (status.isFailure()) {
 312         fprintf(stderr, "error opening input file: ICU Error \"%s\"\n", status.errorName());
 313         exit(status.reset());
 314     }
 315     if (verbose) { printf("Initializing dictionary builder of type %s...\n", (isBytesTrie ? "BytesTrie" : "UCharsTrie")); }
 316     DataDict dict(isBytesTrie, status);
 317     if (status.isFailure()) {
 318         fprintf(stderr, "new DataDict: ICU Error \"%s\"\n", status.errorName());
 319         exit(status.reset());
 320     }
 321     if (options[ARG_TRANSFORM].doesOccur) {
 322         dict.setTransform(options[ARG_TRANSFORM].value);
 323     }
 324
 325     UnicodeString fileLine;
 326     if (verbose) { puts("Adding words to dictionary..."); }
 327     UBool hasValues = FALSE;
 328     UBool hasValuelessContents = FALSE;
 329     int lineCount = 0;
 330     int wordCount = 0;
 331     int minlen = 255;
 332     int maxlen = 0;
 333     UBool isOk = TRUE;
 334     while (readLine(f, fileLine, status)) {
 335         lineCount++;
 336         if (fileLine.isEmpty()) continue;
 337
 338         // Parse word [spaces value].
 339         int32_t keyLen;
 340         for (keyLen = 0; keyLen < fileLine.length() && !u_isspace(fileLine[keyLen]); ++keyLen) {}
 341         if (keyLen == 0) {
 342             fprintf(stderr, "Error: no word on line %i!\n", lineCount);
 343             isOk = FALSE;
 344             continue;
 345         }
 346         int32_t valueStart;
 347         for (valueStart = keyLen;
 348             valueStart < fileLine.length() && u_isspace(fileLine[valueStart]);
 349             ++valueStart) {}
 350
 351         if (keyLen < valueStart) {
 352             int32_t valueLength = fileLine.length() - valueStart;
 353             if (valueLength > 15) {
 354                 fprintf(stderr, "Error: value too long on line %i!\n", lineCount);
 355                 isOk = FALSE;
 356                 continue;
 357             }
 358             char s[16];
 359             fileLine.extract(valueStart, valueLength, s, 16, US_INV);
 360             char *end;
 361             unsigned long value = uprv_strtoul(s, &end, 0);
 362             if (end == s || *end != 0 || (int32_t)uprv_strlen(s) != valueLength || value > 0xffffffff) {
 363                 fprintf(stderr, "Error: value syntax error or value too large on line %i!\n", lineCount);
 364                 isOk = FALSE;
 365                 continue;
 366             }
 367             dict.addWord(fileLine.tempSubString(0, keyLen), (int32_t)value, status);
 368             hasValues = TRUE;
 369             wordCount++;
 370             if (keyLen < minlen) minlen = keyLen;
 371             if (keyLen > maxlen) maxlen = keyLen;
 372         } else {
 373             dict.addWord(fileLine.tempSubString(0, keyLen), 0, status);
 374             hasValuelessContents = TRUE;
 375             wordCount++;
 376             if (keyLen < minlen) minlen = keyLen;
 377             if (keyLen > maxlen) maxlen = keyLen;
 378         }
 379
 380         if (status.isFailure()) {
 381             fprintf(stderr, "ICU Error \"%s\": Failed to add word to trie at input line %d in input file\n",
 382                 status.errorName(), lineCount);
 383             exit(status.reset());
 384         }
 385     }
 386     if (verbose) { printf("Processed %d lines, added %d words, minlen %d, maxlen %d\n", lineCount, wordCount, minlen, maxlen); }
 387
 388     if (!isOk && status.isSuccess()) {
 389         status.set(U_ILLEGAL_ARGUMENT_ERROR);
 390     }
 391     if (hasValues && hasValuelessContents) {
 392         fprintf(stderr, "warning: file contained both valued and unvalued strings!\n");
 393     }
 394
 395     if (verbose) { printf("Serializing data...isBytesTrie? %d\n", isBytesTrie); }
 396     int32_t outDataSize;
 397     const void *outData;
 398     UnicodeString usp;
 399     if (isBytesTrie) {
 400         StringPiece sp = dict.serializeBytes(status);
 401         outDataSize = sp.size();
 402         outData = sp.data();
 403     } else {
 404         dict.serializeUChars(usp, status);
 405         outDataSize = usp.length() * U_SIZEOF_UCHAR;
 406         outData = usp.getBuffer();
 407     }
 408     if (status.isFailure()) {
 409         fprintf(stderr, "gendict: got failure of type %s while serializing, if U_ILLEGAL_ARGUMENT_ERROR possibly due to duplicate dictionary entries\n", status.errorName());
 410         exit(status.reset());
 411     }
 412     if (verbose) { puts("Opening output file..."); }
 413     UNewDataMemory *pData = udata_create(NULL, NULL, outFileName, &dataInfo, copyright, status);
 414     if (status.isFailure()) {
 415         fprintf(stderr, "gendict: could not open output file \"%s\", \"%s\"\n", outFileName, status.errorName());
 416         exit(status.reset());
 417     }
 418
 419     if (verbose) { puts("Writing to output file..."); }
 420     int32_t indexes[DictionaryData::IX_COUNT] = {
 421         DictionaryData::IX_COUNT * sizeof(int32_t), 0, 0, 0, 0, 0, 0, 0
 422     };
 423     int32_t size = outDataSize + indexes[DictionaryData::IX_STRING_TRIE_OFFSET];
 424     indexes[DictionaryData::IX_RESERVED1_OFFSET] = size;
 425     indexes[DictionaryData::IX_RESERVED2_OFFSET] = size;
 426     indexes[DictionaryData::IX_TOTAL_SIZE] = size;
 427
 428     indexes[DictionaryData::IX_TRIE_TYPE] = isBytesTrie ? DictionaryData::TRIE_TYPE_BYTES : DictionaryData::TRIE_TYPE_UCHARS;
 429     if (hasValues) {
 430         indexes[DictionaryData::IX_TRIE_TYPE] |= DictionaryData::TRIE_HAS_VALUES;
 431     }
 432
 433     indexes[DictionaryData::IX_TRANSFORM] = dict.getTransform();
 434     udata_writeBlock(pData, indexes, sizeof(indexes));
 435     udata_writeBlock(pData, outData, outDataSize);
 436     size_t bytesWritten = udata_finish(pData, status);
 437     if (status.isFailure()) {
 438         fprintf(stderr, "gendict: error \"%s\" writing the output file\n", status.errorName());
 439         exit(status.reset());
 440     }
 441
 442     if (bytesWritten != (size_t)size) {
 443         fprintf(stderr, "Error writing to output file \"%s\"\n", outFileName);
 444         exit(U_INTERNAL_PROGRAM_ERROR);
 445     }
 446
 447     printf("%s: done writing\t%s (%ds).\n", progName, outFileName, elapsedTime());
 448
 449 #ifdef TEST_GENDICT
 450     if (isBytesTrie) {
 451         BytesTrie::Iterator it(outData, outDataSize, status);
 452         while (it.hasNext()) {
 453             it.next(status);
 454             const StringPiece s = it.getString();
 455             int32_t val = it.getValue();
 456             printf("%s -> %i\n", s.data(), val);
 457         }
 458     } else {
 459         UCharsTrie::Iterator it((const UChar *)outData, outDataSize, status);
 460         while (it.hasNext()) {
 461             it.next(status);
 462             const UnicodeString s = it.getString();
 463             int32_t val = it.getValue();
 464             char tmp[1024];
 465             s.extract(0, s.length(), tmp, 1024);
 466             printf("%s -> %i\n", tmp, val);
 467         }
 468     }
 469 #endif
 470
 471     return 0;
 472 #endif /* #if !UCONFIG_NO_BREAK_ITERATION */
 473 }