icuSources/tools/genbrk/genbrk.cpp

   1 /*
   2 **********************************************************************
   3 *   Copyright (C) 2002-2006, International Business Machines
   4 *   Corporation and others.  All Rights Reserved.
   5 **********************************************************************
   6 *
   7 * File genbrk.c
   8 */
   9
  10 //--------------------------------------------------------------------
  11 //
  12 //   Tool for generating RuleBasedBreakIterator data files (.brk files).
  13 //   .brk files contain the precompiled rules for standard types
  14 //   of iterators - word, line, sentence, etc.
  15 //
  16 //   Usage:  genbrk [options] -r rule-file.txt  -o output-file.brk
  17 //
  18 //       options:   -v         verbose
  19 //                  -? or -h   help
  20 //
  21 //   The input rule file is a plain text file containing break rules
  22 //    in the input format accepted by RuleBasedBreakIterators.  The
  23 //    file can be encoded as utf-8, or utf-16 (either endian), or
  24 //    in the default code page (platform dependent.).  utf encoded
  25 //    files must include a BOM.
  26 //
  27 //--------------------------------------------------------------------
  28
  29 #include "unicode/utypes.h"
  30 #include "unicode/ucnv.h"
  31 #include "unicode/unistr.h"
  32 #include "unicode/rbbi.h"
  33 #include "unicode/uclean.h"
  34 #include "unicode/udata.h"
  35 #include "unicode/putil.h"
  36
  37 #include "uoptions.h"
  38 #include "unewdata.h"
  39 #include "ucmndata.h"
  40 #include "rbbidata.h"
  41 #include "cmemory.h"
  42
  43 #include <stdio.h>
  44 #include <stdlib.h>
  45 #include <string.h>
  46
  47 U_NAMESPACE_USE
  48
  49 static char *progName;
  50 static UOption options[]={
  51     UOPTION_HELP_H,             /* 0 */
  52     UOPTION_HELP_QUESTION_MARK, /* 1 */
  53     UOPTION_VERBOSE,            /* 2 */
  54     { "rules", NULL, NULL, NULL, 'r', UOPT_REQUIRES_ARG, 0 },   /* 3 */
  55     { "out",   NULL, NULL, NULL, 'o', UOPT_REQUIRES_ARG, 0 },   /* 4 */
  56     UOPTION_ICUDATADIR,         /* 5 */
  57     UOPTION_DESTDIR,            /* 6 */
  58     UOPTION_COPYRIGHT,          /* 7 */
  59 };
  60
  61 void usageAndDie(int retCode) {
  62         printf("Usage: %s [-v] [-options] -r rule-file -o output-file\n", progName);
  63         printf("\tRead in break iteration rules text and write out the binary data\n"
  64             "options:\n"
  65             "\t-h or -? or --help  this usage text\n"
  66             "\t-V or --version     show a version message\n"
  67             "\t-c or --copyright   include a copyright notice\n"
  68             "\t-v or --verbose     turn on verbose output\n"
  69             "\t-i or --icudatadir  directory for locating any needed intermediate data files,\n"
  70             "\t                    followed by path, defaults to %s\n"
  71             "\t-d or --destdir     destination directory, followed by the path\n",
  72             u_getDataDirectory());
  73         exit (retCode);
  74 }
  75
  76
  77 #if UCONFIG_NO_BREAK_ITERATION
  78
  79 /* dummy UDataInfo cf. udata.h */
  80 static UDataInfo dummyDataInfo = {
  81     sizeof(UDataInfo),
  82     0,
  83
  84     U_IS_BIG_ENDIAN,
  85     U_CHARSET_FAMILY,
  86     U_SIZEOF_UCHAR,
  87     0,
  88
  89     { 0, 0, 0, 0 },                 /* dummy dataFormat */
  90     { 0, 0, 0, 0 },                 /* dummy formatVersion */
  91     { 0, 0, 0, 0 }                  /* dummy dataVersion */
  92 };
  93
  94 #else
  95
  96 //
  97 //  Set up the ICU data header, defined in ucmndata.h
  98 //
  99 DataHeader dh ={
 100     {sizeof(DataHeader),           // Struct MappedData
 101         0xda,
 102         0x27},
 103
 104     {                               // struct UDataInfo
 105         sizeof(UDataInfo),          //     size
 106         0,                          //     reserved
 107         U_IS_BIG_ENDIAN,
 108         U_CHARSET_FAMILY,
 109         U_SIZEOF_UCHAR,
 110         0,                          //     reserved
 111
 112     { 0x42, 0x72, 0x6b, 0x20 },     //     dataFormat="Brk "
 113     { 0xff, 0, 0, 0 },              //     formatVersion.  Filled in later with values
 114                                     //      from the RBBI rule builder.  The  values declared
 115                                     //      here should never appear in any real RBBI data.
 116         { 4, 1, 0, 0 }              //   dataVersion (Unicode version)
 117     }};
 118
 119 #endif
 120
 121 //----------------------------------------------------------------------------
 122 //
 123 //  main      for genbrk
 124 //
 125 //----------------------------------------------------------------------------
 126 int  main(int argc, char **argv) {
 127     UErrorCode  status = U_ZERO_ERROR;
 128     const char *ruleFileName;
 129     const char *outFileName;
 130     const char *outDir = NULL;
 131     const char *copyright = NULL;
 132
 133     //
 134     // Pick up and check the command line arguments,
 135     //    using the standard ICU tool utils option handling.
 136     //
 137     U_MAIN_INIT_ARGS(argc, argv);
 138     progName = argv[0];
 139     argc=u_parseArgs(argc, argv, sizeof(options)/sizeof(options[0]), options);
 140     if(argc<0) {
 141         // Unrecognized option
 142         fprintf(stderr, "error in command line argument \"%s\"\n", argv[-argc]);
 143         usageAndDie(U_ILLEGAL_ARGUMENT_ERROR);
 144     }
 145
 146     if(options[0].doesOccur || options[1].doesOccur) {
 147         //  -? or -h for help.
 148         usageAndDie(0);
 149     }
 150
 151     if (!(options[3].doesOccur && options[4].doesOccur)) {
 152         fprintf(stderr, "rule file and output file must both be specified.\n");
 153         usageAndDie(U_ILLEGAL_ARGUMENT_ERROR);
 154     }
 155     ruleFileName = options[3].value;
 156     outFileName  = options[4].value;
 157
 158     if (options[5].doesOccur) {
 159         u_setDataDirectory(options[5].value);
 160     }
 161
 162     /* Initialize ICU */
 163     u_init(&status);
 164     if (U_FAILURE(status)) {
 165         fprintf(stderr, "%s: can not initialize ICU.  status = %s\n",
 166             argv[0], u_errorName(status));
 167         exit(1);
 168     }
 169     status = U_ZERO_ERROR;
 170
 171     /* Combine the directory with the file name */
 172     if(options[6].doesOccur) {
 173         outDir = options[6].value;
 174     }
 175     if (options[7].doesOccur) {
 176         copyright = U_COPYRIGHT_STRING;
 177     }
 178
 179 #if UCONFIG_NO_BREAK_ITERATION
 180
 181     UNewDataMemory *pData;
 182     char msg[1024];
 183
 184     /* write message with just the name */
 185     sprintf(msg, "genbrk writes dummy %s because of UCONFIG_NO_BREAK_ITERATION, see uconfig.h", outFileName);
 186     fprintf(stderr, "%s\n", msg);
 187
 188     /* write the dummy data file */
 189     pData = udata_create(outDir, NULL, outFileName, &dummyDataInfo, NULL, &status);
 190     udata_writeBlock(pData, msg, strlen(msg));
 191     udata_finish(pData, &status);
 192     return (int)status;
 193
 194 #else
 195
 196     //
 197     //  Read in the rule source file
 198     //
 199     long        result;
 200     long        ruleFileSize;
 201     FILE        *file;
 202     char        *ruleBufferC;
 203
 204     file = fopen(ruleFileName, "rb");
 205     if( file == 0 ) {
 206         fprintf(stderr, "Could not open file \"%s\"\n", ruleFileName);
 207         exit(-1);
 208     }
 209     fseek(file, 0, SEEK_END);
 210     ruleFileSize = ftell(file);
 211     fseek(file, 0, SEEK_SET);
 212     ruleBufferC = new char[ruleFileSize+10];
 213
 214     result = (long)fread(ruleBufferC, 1, ruleFileSize, file);
 215     if (result != ruleFileSize)  {
 216         fprintf(stderr, "Error reading file \"%s\"\n", ruleFileName);
 217         exit (-1);
 218     }
 219     ruleBufferC[ruleFileSize]=0;
 220     fclose(file);
 221
 222     //
 223     // Look for a Unicode Signature (BOM) on the rule file
 224     //
 225     int32_t        signatureLength;
 226     const char *   ruleSourceC = ruleBufferC;
 227     const char*    encoding = ucnv_detectUnicodeSignature(
 228                            ruleSourceC, ruleFileSize, &signatureLength, &status);
 229     if (U_FAILURE(status)) {
 230         exit(status);
 231     }
 232     if(encoding!=NULL ){
 233         ruleSourceC  += signatureLength;
 234         ruleFileSize -= signatureLength;
 235     }
 236
 237     //
 238     // Open a converter to take the rule file to UTF-16
 239     //
 240     UConverter* conv;
 241     conv = ucnv_open(encoding, &status);
 242     if (U_FAILURE(status)) {
 243         fprintf(stderr, "ucnv_open: ICU Error \"%s\"\n", u_errorName(status));
 244         exit(status);
 245     }
 246
 247     //
 248     // Convert the rules to UChar.
 249     //  Preflight first to determine required buffer size.
 250     //
 251     uint32_t destCap = ucnv_toUChars(conv,
 252                        NULL,           //  dest,
 253                        0,              //  destCapacity,
 254                        ruleSourceC,
 255                        ruleFileSize,
 256                        &status);
 257     if (status != U_BUFFER_OVERFLOW_ERROR) {
 258         fprintf(stderr, "ucnv_toUChars: ICU Error \"%s\"\n", u_errorName(status));
 259         exit(status);
 260     };
 261
 262     status = U_ZERO_ERROR;
 263     UChar *ruleSourceU = new UChar[destCap+1];
 264     ucnv_toUChars(conv,
 265                   ruleSourceU,     //  dest,
 266                   destCap+1,
 267                   ruleSourceC,
 268                   ruleFileSize,
 269                   &status);
 270     if (U_FAILURE(status)) {
 271         fprintf(stderr, "ucnv_toUChars: ICU Error \"%s\"\n", u_errorName(status));
 272         exit(status);
 273     };
 274     ucnv_close(conv);
 275
 276
 277     //
 278     //  Put the source rules into a UnicodeString
 279     //
 280     UnicodeString ruleSourceS(FALSE, ruleSourceU, destCap);
 281
 282     //
 283     //  Create the break iterator from the rules
 284     //     This will compile the rules.
 285     //
 286     UParseError parseError;
 287     parseError.line = 0;
 288     parseError.offset = 0;
 289     RuleBasedBreakIterator *bi = new RuleBasedBreakIterator(ruleSourceS, parseError, status);
 290     if (U_FAILURE(status)) {
 291         fprintf(stderr, "createRuleBasedBreakIterator: ICU Error \"%s\"  at line %d, column %d\n",
 292                 u_errorName(status), (int)parseError.line, (int)parseError.offset);
 293         exit(status);
 294     };
 295
 296
 297     //
 298     //  Get the compiled rule data from the break iterator.
 299     //
 300     uint32_t        outDataSize;
 301     const uint8_t  *outData;
 302     outData = bi->getBinaryRules(outDataSize);
 303
 304     // Copy the data format version numbers from the RBBI data header into the UDataMemory header.
 305     uprv_memcpy(dh.info.formatVersion, ((RBBIDataHeader *)outData)->fFormatVersion, sizeof(dh.info.formatVersion));
 306
 307     //
 308     //  Create the output file
 309     //
 310     size_t bytesWritten;
 311     UNewDataMemory *pData;
 312     pData = udata_create(outDir, NULL, outFileName, &(dh.info), copyright, &status);
 313     if(U_FAILURE(status)) {
 314         fprintf(stderr, "genbrk: Could not open output file \"%s\", \"%s\"\n",
 315                          outFileName, u_errorName(status));
 316         exit(status);
 317     }
 318
 319
 320     //  Write the data itself.
 321     udata_writeBlock(pData, outData, outDataSize);
 322     // finish up
 323     bytesWritten = udata_finish(pData, &status);
 324     if(U_FAILURE(status)) {
 325         fprintf(stderr, "genbrk: error %d writing the output file\n", status);
 326         exit(status);
 327     }
 328
 329     if (bytesWritten != outDataSize) {
 330         fprintf(stderr, "Error writing to output file \"%s\"\n", outFileName);
 331         exit(-1);
 332     }
 333
 334     delete bi;
 335     delete[] ruleSourceU;
 336     delete[] ruleBufferC;
 337     u_cleanup();
 338
 339
 340     printf("genbrk: tool completed successfully.\n");
 341     return 0;
 342
 343 #endif /* #if !UCONFIG_NO_BREAK_ITERATION */
 344 }
 345