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