util/yaz-icu.c

   1 /* This file is part of the YAZ toolkit.
   2  * Copyright (C) 1995-2009 Index Data
   3  * See the file LICENSE for details.
   4  */
   5
   6 #if HAVE_CONFIG_H
   7 #include "config.h"
   8 #endif
   9
  10 #include <string.h>
  11
  12 #include <stdio.h>
  13 #include <stdlib.h>
  14
  15 #include <yaz/options.h>
  16
  17 #if YAZ_HAVE_ICU
  18
  19 #include <unicode/ucnv.h>
  20 #include <unicode/ustring.h>
  21 #include <unicode/ucol.h>
  22 #include <unicode/ubrk.h>
  23 #include <unicode/utrans.h>
  24
  25 #include <yaz/icu.h>
  26 #include <yaz/wrbuf.h>
  27
  28 /* commando line and config parameters */
  29 static struct config_t {
  30     char conffile[1024];
  31     char print[1024];
  32     int xmloutput;
  33     int sortoutput;
  34     yaz_icu_chain_t chain;
  35     FILE * infile;
  36     FILE * outfile;
  37 } config;
  38
  39
  40
  41 void print_option_error(const struct config_t *p_config)
  42 {
  43     fprintf(stderr, "Calling error, valid options are :\n");
  44     fprintf(stderr, "yaz-icu\n"
  45             "   [-c (path/to/config/file.xml)]\n"
  46             "   [-p (a|c|l|t)] print ICU info \n"
  47             "   [-s] Show sort normalization key\n"
  48             "   [-x] XML output\n"
  49             "\n"
  50             "Examples:\n"
  51             "cat hugetextfile.txt | ./yaz-icu -c config.xml \n"
  52             "./yaz-icu -p c\n"
  53             "./yaz-icu -p l -x\n"
  54             "./yaz-icu -p t -x\n"
  55             "\n"
  56             "Example ICU chain XML configuration file:\n"
  57             "<icu_chain locale=\"en\">\n"
  58             "  <transform rule=\"[:Control:] Any-Remove\"/>\n"
  59             "  <tokenize rule=\"l\"/>\n"
  60             "  <transform rule=\"[[:WhiteSpace:][:Punctuation:]] Remove\"/>\n"
  61             "  <casemap rule=\"l\"/>\n"
  62             "</icu_chain>\n"
  63           );
  64     exit(1);
  65 }
  66
  67 void read_params(int argc, char **argv, struct config_t *p_config)
  68 {
  69     char *arg;
  70     int ret;
  71
  72     /* set default parameters */
  73     p_config->conffile[0] = 0;
  74     p_config->print[0] = 0;
  75     p_config->xmloutput = 0;
  76     p_config->sortoutput = 0;
  77     p_config->chain = 0;
  78     p_config->infile = stdin;
  79     p_config->outfile = stdout;
  80
  81     /* set up command line parameters */
  82
  83     while ((ret = options("c:p:xs", argv, argc, &arg)) != -2)
  84     {
  85         switch (ret)
  86         {
  87         case 'c':
  88             strcpy(p_config->conffile, arg);
  89             break;
  90         case 'p':
  91             strcpy(p_config->print, arg);
  92             break;
  93         case 's':
  94             p_config->sortoutput = 1;
  95             break;
  96         case 'x':
  97             p_config->xmloutput = 1;
  98             break;
  99         default:
 100             printf("Got %d\n", ret);
 101             print_option_error(p_config);
 102         }
 103     }
 104
 105     if ((!strlen(p_config->conffile)
 106          && !strlen(p_config->print))
 107         || !config.infile
 108         || !config.outfile)
 109
 110         print_option_error(p_config);
 111 }
 112
 113
 114 /*     UConverter *conv; */
 115 /*     conv = ucnv_open("utf-8", &status); */
 116 /*     assert(U_SUCCESS(status)); */
 117
 118 /*     *ustr16_len  */
 119 /*       = ucnv_toUChars(conv, ustr16, 1024,  */
 120 /*                       (const char *) *xstr8, strlen((const char *) *xstr8), */
 121 /*                       &status); */
 122
 123
 124
 125 /*      ucnv_fromUChars(conv, */
 126 /*                      (char *) *xstr8, strlen((const char *) *xstr8), */
 127 /*                      ustr16, *ustr16_len, */
 128 /*                      &status); */
 129 /*      ucnv_close(conv); */
 130
 131
 132 static void print_icu_converters(const struct config_t *p_config)
 133 {
 134     int32_t count;
 135     int32_t i;
 136
 137     count = ucnv_countAvailable();
 138     if (p_config->xmloutput)
 139         fprintf(config.outfile, "<converters count=\"%d\" default=\"%s\">\n",
 140                 count, ucnv_getDefaultName());
 141     else {
 142         fprintf(config.outfile, "Available ICU converters: %d\n", count);
 143         fprintf(config.outfile, "Default ICU Converter is: '%s'\n",
 144                 ucnv_getDefaultName());
 145     }
 146
 147     for(i=0;i<count;i++)
 148     {
 149         if (p_config->xmloutput)
 150             fprintf(config.outfile, "<converter id=\"%s\"/>\n",
 151                     ucnv_getAvailableName(i));
 152         else
 153             fprintf(config.outfile, "%s ", ucnv_getAvailableName(i));
 154     }
 155
 156     if (p_config->xmloutput)
 157         fprintf(config.outfile, "</converters>\n");
 158     else
 159         fprintf(config.outfile, "\n");
 160 }
 161
 162 static void print_icu_transliterators(const struct config_t *p_config)
 163 {
 164     int32_t buf_cap = 128;
 165     char buf[128];
 166     int32_t i;
 167     int32_t count = utrans_countAvailableIDs();
 168
 169     if (p_config->xmloutput)
 170         fprintf(config.outfile, "<transliterators count=\"%d\">\n",  count);
 171     else
 172         fprintf(config.outfile, "Available ICU transliterators: %d\n", count);
 173
 174     for(i = 0; i <count; i++)
 175     {
 176         utrans_getAvailableID(i, buf, buf_cap);
 177         if (p_config->xmloutput)
 178             fprintf(config.outfile, "<transliterator id=\"%s\"/>\n", buf);
 179         else
 180             fprintf(config.outfile, " %s", buf);
 181     }
 182
 183     if (p_config->xmloutput)
 184     {
 185         fprintf(config.outfile, "</transliterators>\n");
 186     }
 187     else
 188     {
 189         fprintf(config.outfile, "\n\nUnicode Set Patterns:\n"
 190                 "   Pattern         Description\n"
 191                 "   Ranges          [a-z]       The lower case letters a through z\n"
 192                 "   Named Chars     [abc123] The six characters a,b,c,1,2 and 3\n"
 193                 "   String          [abc{def}] chars a, b and c, and string 'def'\n"
 194                 "   Categories      [\\p{Letter}] Perl General Category 'Letter'.\n"
 195                 "   Categories      [:Letter:] Posix General Category 'Letter'.\n"
 196                 "\n"
 197                 "   Combination     Example\n"
 198                 "   Union           [[:Greek:] [:letter:]]\n"
 199                 "   Intersection    [[:Greek:] & [:letter:]]\n"
 200                 "   Set Complement  [[:Greek:] - [:letter:]]\n"
 201                 "   Complement      [^[:Greek:] [:letter:]]\n"
 202                 "\n"
 203              "see: http://icu.sourceforge.net/userguide/unicodeSet.html\n"
 204                 "\n"
 205                 "Examples:\n"
 206                 "   [:Punctuation:] Any-Remove\n"
 207                 "   [:Cased-Letter:] Any-Upper\n"
 208                 "   [:Control:] Any-Remove\n"
 209                 "   [:Decimal_Number:] Any-Remove\n"
 210                 "   [:Final_Punctuation:] Any-Remove\n"
 211                 "   [:Georgian:] Any-Upper\n"
 212                 "   [:Katakana:] Any-Remove\n"
 213                 "   [:Arabic:] Any-Remove\n"
 214                 "   [:Punctuation:] Remove\n"
 215                 "   [[:Punctuation:]-[.,]] Remove\n"
 216                 "   [:Line_Separator:] Any-Remove\n"
 217                 "   [:Math_Symbol:] Any-Remove\n"
 218                 "   Lower; [:^Letter:] Remove (word tokenization)\n"
 219                 "   [:^Number:] Remove (numeric tokenization)\n"
 220                 "   [:^Katagana:] Remove (remove everything except Katagana)\n"
 221                 "   Lower;[[:WhiteSpace:][:Punctuation:]] Remove (word tokenization)\n"
 222                 "   NFD; [:Nonspacing Mark:] Remove; NFC   (removes accents from characters)\n"
 223                 "   [A-Za-z]; Lower(); Latin-Katakana; Katakana-Hiragana (transforms latin and katagana to hiragana)\n"
 224                 "   [[:separator:][:start punctuation:][:initial punctuation:]] Remove \n"
 225                 "\n"
 226                 "see http://icu.sourceforge.net/userguide/Transform.html\n"
 227                 "    http://www.unicode.org/Public/UNIDATA/UCD.html\n"
 228                 "    http://icu.sourceforge.net/userguide/Transform.html\n"
 229                 "    http://icu.sourceforge.net/userguide/TransformRule.html\n"
 230             );
 231
 232
 233         fprintf(config.outfile, "\n\n");
 234
 235     }
 236 }
 237
 238 static void print_icu_xml_locales(const struct config_t *p_config)
 239 {
 240     int32_t count;
 241     int32_t i;
 242     UErrorCode status = U_ZERO_ERROR;
 243
 244     UChar keyword[64];
 245     int32_t keyword_len = 0;
 246     char keyword_str[128];
 247     int32_t keyword_str_len = 0;
 248
 249     UChar language[64];
 250     int32_t language_len = 0;
 251     char lang_str[128];
 252     int32_t lang_str_len = 0;
 253
 254     UChar script[64];
 255     int32_t script_len = 0;
 256     char script_str[128];
 257     int32_t script_str_len = 0;
 258
 259     UChar location[64];
 260     int32_t location_len = 0;
 261     char location_str[128];
 262     int32_t location_str_len = 0;
 263
 264     UChar variant[64];
 265     int32_t variant_len = 0;
 266     char variant_str[128];
 267     int32_t variant_str_len = 0;
 268
 269     UChar name[64];
 270     int32_t name_len = 0;
 271     char name_str[128];
 272     int32_t name_str_len = 0;
 273
 274     UChar localname[64];
 275     int32_t localname_len = 0;
 276     char localname_str[128];
 277     int32_t localname_str_len = 0;
 278
 279     count = uloc_countAvailable() ;
 280
 281     if (p_config->xmloutput)
 282     {
 283         fprintf(config.outfile, "<locales count=\"%d\" default=\"%s\" collations=\"%d\">\n",
 284                 count, uloc_getDefault(), ucol_countAvailable());
 285     }
 286
 287     for(i=0;i<count;i++)
 288     {
 289
 290         keyword_len
 291             = uloc_getDisplayKeyword(uloc_getAvailable(i), "en",
 292                                      keyword, 64,
 293                                      &status);
 294
 295         u_strToUTF8(keyword_str, 128, &keyword_str_len,
 296                     keyword, keyword_len,
 297                     &status);
 298
 299
 300         language_len
 301             = uloc_getDisplayLanguage(uloc_getAvailable(i), "en",
 302                                       language, 64,
 303                                       &status);
 304
 305         u_strToUTF8(lang_str, 128, &lang_str_len,
 306                     language, language_len,
 307                     &status);
 308
 309
 310         script_len
 311             = uloc_getDisplayScript(uloc_getAvailable(i), "en",
 312                                     script, 64,
 313                                     &status);
 314
 315         u_strToUTF8(script_str, 128, &script_str_len,
 316                     script, script_len,
 317                     &status);
 318
 319         location_len
 320             = uloc_getDisplayCountry(uloc_getAvailable(i), "en",
 321                                      location, 64,
 322                                      &status);
 323
 324         u_strToUTF8(location_str, 128, &location_str_len,
 325                     location, location_len,
 326                     &status);
 327
 328         variant_len
 329             = uloc_getDisplayVariant(uloc_getAvailable(i), "en",
 330                                      variant, 64,
 331                                      &status);
 332
 333         u_strToUTF8(variant_str, 128, &variant_str_len,
 334                     variant, variant_len,
 335                     &status);
 336
 337         name_len
 338             = uloc_getDisplayName(uloc_getAvailable(i), "en",
 339                                   name, 64,
 340                                   &status);
 341
 342         u_strToUTF8(name_str, 128, &name_str_len,
 343                     name, name_len,
 344                     &status);
 345
 346         localname_len
 347             = uloc_getDisplayName(uloc_getAvailable(i), uloc_getAvailable(i),
 348                                   localname, 64,
 349                                   &status);
 350
 351         u_strToUTF8(localname_str, 128, &localname_str_len,
 352                     localname, localname_len,
 353                     &status);
 354
 355
 356         if (p_config->xmloutput)
 357         {
 358             fprintf(config.outfile, "<locale id=\"%s\"", uloc_getAvailable(i));
 359             /* fprintf(config.outfile, " locale=\"%s\"", uloc_getAvailable(i)); */
 360             /* if (strlen(keyword_str)) */
 361             /*   fprintf(config.outfile, " keyword=\"%s\"", keyword_str); */
 362             /* if (ucol_getAvailable(i)) */
 363             /*   fprintf(config.outfile, " collation=\"1\""); */
 364             if (strlen(lang_str))
 365                 fprintf(config.outfile, " language=\"%s\"", lang_str);
 366             if (strlen(script_str))
 367                 fprintf(config.outfile, " script=\"%s\"", script_str);
 368             if (strlen(location_str))
 369                 fprintf(config.outfile, " location=\"%s\"", location_str);
 370             if (strlen(variant_str))
 371                 fprintf(config.outfile, " variant=\"%s\"", variant_str);
 372             if (strlen(name_str))
 373                 fprintf(config.outfile, " name=\"%s\"", name_str);
 374             if (strlen(localname_str))
 375                 fprintf(config.outfile, " localname=\"%s\"", localname_str);
 376             fprintf(config.outfile, ">");
 377             if (strlen(localname_str))
 378                 fprintf(config.outfile, "%s", localname_str);
 379             fprintf(config.outfile, "</locale>\n");
 380         }
 381         else if (1 == p_config->xmloutput)
 382         {
 383             fprintf(config.outfile, "%s", uloc_getAvailable(i));
 384             fprintf(config.outfile, " | ");
 385             if (strlen(name_str))
 386                 fprintf(config.outfile, "%s", name_str);
 387             fprintf(config.outfile, " | ");
 388             if (strlen(localname_str))
 389                 fprintf(config.outfile, "%s", localname_str);
 390             fprintf(config.outfile, "\n");
 391         }
 392         else
 393             fprintf(config.outfile, "%s ", uloc_getAvailable(i));
 394     }
 395     if (p_config->xmloutput)
 396         fprintf(config.outfile, "</locales>\n");
 397     else
 398         fprintf(config.outfile, "\n");
 399
 400     if(U_FAILURE(status))
 401     {
 402         fprintf(stderr, "ICU Error: %d %s\n", status, u_errorName(status));
 403         exit(2);
 404     }
 405 }
 406
 407
 408 static void print_info(const struct config_t *p_config)
 409 {
 410     if (p_config->xmloutput)
 411         fprintf(config.outfile, "<?xml version=\"1.0\" encoding=\"UTF-8\"?>\n"
 412                 "<icu>\n");
 413
 414     if ('c' == config.print[0])
 415         print_icu_converters(&config);
 416     else if ('l' == config.print[0])
 417         print_icu_xml_locales(&config);
 418     else if ('t' == config.print[0])
 419         print_icu_transliterators(&config);
 420     else {
 421         print_icu_converters(&config);
 422         print_icu_xml_locales(&config);
 423         print_icu_transliterators(&config);
 424     }
 425
 426     if (p_config->xmloutput)
 427         fprintf(config.outfile, "</icu>\n");
 428
 429     exit(0);
 430 }
 431
 432
 433
 434 static void process_text_file(const struct config_t *p_config)
 435 {
 436     char *line = 0;
 437     char linebuf[1024];
 438
 439     xmlDoc *doc = xmlParseFile(config.conffile);
 440     xmlNode *xml_node = xmlDocGetRootElement(doc);
 441
 442     long unsigned int token_count = 0;
 443     long unsigned int line_count = 0;
 444
 445     UErrorCode status = U_ZERO_ERROR;
 446     int success = 0;
 447
 448     if (!xml_node)
 449     {
 450         printf("Could not parse XML config file '%s' \n",
 451                 config.conffile);
 452         exit(1);
 453     }
 454
 455     config.chain = icu_chain_xml_config(xml_node, 1, &status);
 456
 457     if (config.chain && U_SUCCESS(status))
 458         success = 1;
 459     else {
 460         printf("Could not set up ICU chain from config file '%s' \n",
 461                 config.conffile);
 462         exit(1);
 463     }
 464
 465     if (p_config->xmloutput)
 466         fprintf(config.outfile,
 467                 "<?xml version=\"1.0\" encoding=\"UTF-8\"?>\n"
 468                 "<icu>\n"
 469                 "<tokens>\n");
 470
 471     /* read input lines for processing */
 472     while ((line=fgets(linebuf, sizeof(linebuf)-1, config.infile)))
 473     {
 474         success = icu_chain_assign_cstr(config.chain, line, &status);
 475         line_count++;
 476
 477         while (success && icu_chain_next_token(config.chain, &status))
 478         {
 479             WRBUF sw = wrbuf_alloc();
 480             if (U_FAILURE(status))
 481                 success = 0;
 482             else {
 483                 const char *sortkey = icu_chain_token_sortkey(config.chain);
 484                 wrbuf_rewind(sw);
 485                 wrbuf_puts_escaped(sw, sortkey);
 486                 token_count++;
 487                 if (p_config->xmloutput)
 488                 {
 489                     /* should XML encode this. Bug #1902 */
 490                     fprintf(config.outfile,
 491                             "<token id=\"%lu\" line=\"%lu\""
 492                             " norm=\"%s\" display=\"%s\"",
 493                             token_count,
 494                             line_count,
 495                             icu_chain_token_norm(config.chain),
 496                             icu_chain_token_display(config.chain));
 497                     if (p_config->sortoutput)
 498                     {
 499                         fprintf(config.outfile, " sortkey=\"%s\"",
 500                                 wrbuf_cstr(sw));
 501                     }
 502                     fprintf(config.outfile, "/>\n");
 503                 }
 504                 else
 505                 {
 506                     fprintf(config.outfile, "%lu %lu '%s' '%s'",
 507                             token_count,
 508                             line_count,
 509                             icu_chain_token_norm(config.chain),
 510                             icu_chain_token_display(config.chain));
 511                     if (p_config->sortoutput)
 512                     {
 513                         fprintf(config.outfile, " '%s'", wrbuf_cstr(sw));
 514                     }
 515                     fprintf(config.outfile, "\n");
 516                 }
 517             }
 518             wrbuf_destroy(sw);
 519         }
 520
 521     }
 522
 523     if (p_config->xmloutput)
 524         fprintf(config.outfile,
 525                 "</tokens>\n"
 526                 "</icu>\n");
 527
 528     icu_chain_destroy(config.chain);
 529     xmlFreeDoc(doc);
 530     if (line)
 531         free(line);
 532 }
 533
 534 #endif /* YAZ_HAVE_ICU */
 535
 536
 537 int main(int argc, char **argv)
 538 {
 539
 540 #if YAZ_HAVE_ICU
 541
 542     read_params(argc, argv, &config);
 543
 544     if (config.conffile && strlen(config.conffile))
 545         process_text_file(&config);
 546
 547     if (config.print && strlen(config.print))
 548         print_info(&config);
 549
 550 #else /* YAZ_HAVE_ICU */
 551
 552     printf("ICU not available on your system.\n"
 553            "Please install libicu36-dev and icu-doc or similar, "
 554            "re-configure and re-compile\n");
 555
 556
 557     exit(3);
 558 #endif /* YAZ_HAVE_ICU */
 559
 560     return 0;
 561 }
 562
 563
 564 /*
 565  * Local variables:
 566  * c-basic-offset: 4
 567  * c-file-style: "Stroustrup"
 568  * indent-tabs-mode: nil
 569  * End:
 570  * vim: shiftwidth=4 tabstop=8 expandtab
 571  */
 572