util/yaz-icu.c

   1 /* This file is part of the YAZ toolkit.
   2  * Copyright (C) 1995-2011 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     {
 143         fprintf(config.outfile, "Available ICU converters: %d\n", count);
 144         fprintf(config.outfile, "Default ICU Converter is: '%s'\n",
 145                 ucnv_getDefaultName());
 146     }
 147
 148     for (i = 0; i < count; i++)
 149     {
 150         if (p_config->xmloutput)
 151             fprintf(config.outfile, "<converter id=\"%s\"/>\n",
 152                     ucnv_getAvailableName(i));
 153         else
 154             fprintf(config.outfile, "%s\n", ucnv_getAvailableName(i));
 155     }
 156
 157     if (p_config->xmloutput)
 158         fprintf(config.outfile, "</converters>\n");
 159     else
 160         fprintf(config.outfile, "\n");
 161 }
 162
 163 static void print_icu_transliterators(const struct config_t *p_config)
 164 {
 165     UErrorCode status;
 166     UEnumeration *en = utrans_openIDs(&status);
 167     int32_t count = uenum_count(en, &status);
 168     const char *name;
 169     int32_t length;
 170
 171     if (p_config->xmloutput)
 172         fprintf(config.outfile, "<transliterators count=\"%d\">\n",  count);
 173     else
 174         fprintf(config.outfile, "Available ICU transliterators: %d\n", count);
 175
 176     while ((name = uenum_next(en, &length, &status)))
 177     {
 178         if (p_config->xmloutput)
 179             fprintf(config.outfile, "<transliterator id=\"%s\"/>\n", name);
 180         else
 181             fprintf(config.outfile, "%s\n", name);
 182     }
 183     uenum_close(en);
 184     if (p_config->xmloutput)
 185         fprintf(config.outfile, "</transliterators>\n");
 186     else
 187     {
 188         fprintf(config.outfile, "\n\nUnicode Set Patterns:\n"
 189                 "   Pattern         Description\n"
 190                 "   Ranges          [a-z]       The lower case letters a through z\n"
 191                 "   Named Chars     [abc123] The six characters a,b,c,1,2 and 3\n"
 192                 "   String          [abc{def}] chars a, b and c, and string 'def'\n"
 193                 "   Categories      [\\p{Letter}] Perl General Category 'Letter'.\n"
 194                 "   Categories      [:Letter:] Posix General Category 'Letter'.\n"
 195                 "\n"
 196                 "   Combination     Example\n"
 197                 "   Union           [[:Greek:] [:letter:]]\n"
 198                 "   Intersection    [[:Greek:] & [:letter:]]\n"
 199                 "   Set Complement  [[:Greek:] - [:letter:]]\n"
 200                 "   Complement      [^[:Greek:] [:letter:]]\n"
 201                 "\n"
 202              "see: http://icu.sourceforge.net/userguide/unicodeSet.html\n"
 203                 "\n"
 204                 "Examples:\n"
 205                 "   [:Punctuation:] Any-Remove\n"
 206                 "   [:Cased-Letter:] Any-Upper\n"
 207                 "   [:Control:] Any-Remove\n"
 208                 "   [:Decimal_Number:] Any-Remove\n"
 209                 "   [:Final_Punctuation:] Any-Remove\n"
 210                 "   [:Georgian:] Any-Upper\n"
 211                 "   [:Katakana:] Any-Remove\n"
 212                 "   [:Arabic:] Any-Remove\n"
 213                 "   [:Punctuation:] Remove\n"
 214                 "   [[:Punctuation:]-[.,]] Remove\n"
 215                 "   [:Line_Separator:] Any-Remove\n"
 216                 "   [:Math_Symbol:] Any-Remove\n"
 217                 "   Lower; [:^Letter:] Remove (word tokenization)\n"
 218                 "   [:^Number:] Remove (numeric tokenization)\n"
 219                 "   [:^Katagana:] Remove (remove everything except Katagana)\n"
 220                 "   Lower;[[:WhiteSpace:][:Punctuation:]] Remove (word tokenization)\n"
 221                 "   NFD; [:Nonspacing Mark:] Remove; NFC   (removes accents from characters)\n"
 222                 "   [A-Za-z]; Lower(); Latin-Katakana; Katakana-Hiragana (transforms latin and katagana to hiragana)\n"
 223                 "   [[:separator:][:start punctuation:][:initial punctuation:]] Remove \n"
 224                 "\n"
 225                 "see http://userguide.icu-project.org/transforms/general\n"
 226                 "    http://www.unicode.org/reports/tr44/\n"
 227             );
 228
 229
 230         fprintf(config.outfile, "\n\n");
 231
 232     }
 233 }
 234
 235 static void print_icu_xml_locales(const struct config_t *p_config)
 236 {
 237     int32_t count;
 238     int32_t i;
 239     UErrorCode status = U_ZERO_ERROR;
 240
 241     UChar keyword[64];
 242     int32_t keyword_len = 0;
 243     char keyword_str[128];
 244     int32_t keyword_str_len = 0;
 245
 246     UChar language[64];
 247     int32_t language_len = 0;
 248     char lang_str[128];
 249     int32_t lang_str_len = 0;
 250
 251     UChar script[64];
 252     int32_t script_len = 0;
 253     char script_str[128];
 254     int32_t script_str_len = 0;
 255
 256     UChar location[64];
 257     int32_t location_len = 0;
 258     char location_str[128];
 259     int32_t location_str_len = 0;
 260
 261     UChar variant[64];
 262     int32_t variant_len = 0;
 263     char variant_str[128];
 264     int32_t variant_str_len = 0;
 265
 266     UChar name[64];
 267     int32_t name_len = 0;
 268     char name_str[128];
 269     int32_t name_str_len = 0;
 270
 271     UChar localname[64];
 272     int32_t localname_len = 0;
 273     char localname_str[128];
 274     int32_t localname_str_len = 0;
 275
 276     count = uloc_countAvailable() ;
 277
 278     if (p_config->xmloutput)
 279     {
 280         fprintf(config.outfile, "<locales count=\"%d\" default=\"%s\" collations=\"%d\">\n",
 281                 count, uloc_getDefault(), ucol_countAvailable());
 282     }
 283     else
 284     {
 285         fprintf(config.outfile, "Available ICU locales: %d\n", count);
 286         fprintf(config.outfile, "Default locale is: %s\n",  uloc_getDefault());
 287     }
 288
 289     for (i = 0; i < count; i++)
 290     {
 291
 292         keyword_len
 293             = uloc_getDisplayKeyword(uloc_getAvailable(i), "en",
 294                                      keyword, 64,
 295                                      &status);
 296
 297         u_strToUTF8(keyword_str, 128, &keyword_str_len,
 298                     keyword, keyword_len,
 299                     &status);
 300
 301
 302         language_len
 303             = uloc_getDisplayLanguage(uloc_getAvailable(i), "en",
 304                                       language, 64,
 305                                       &status);
 306
 307         u_strToUTF8(lang_str, 128, &lang_str_len,
 308                     language, language_len,
 309                     &status);
 310
 311
 312         script_len
 313             = uloc_getDisplayScript(uloc_getAvailable(i), "en",
 314                                     script, 64,
 315                                     &status);
 316
 317         u_strToUTF8(script_str, 128, &script_str_len,
 318                     script, script_len,
 319                     &status);
 320
 321         location_len
 322             = uloc_getDisplayCountry(uloc_getAvailable(i), "en",
 323                                      location, 64,
 324                                      &status);
 325
 326         u_strToUTF8(location_str, 128, &location_str_len,
 327                     location, location_len,
 328                     &status);
 329
 330         variant_len
 331             = uloc_getDisplayVariant(uloc_getAvailable(i), "en",
 332                                      variant, 64,
 333                                      &status);
 334
 335         u_strToUTF8(variant_str, 128, &variant_str_len,
 336                     variant, variant_len,
 337                     &status);
 338
 339         name_len
 340             = uloc_getDisplayName(uloc_getAvailable(i), "en",
 341                                   name, 64,
 342                                   &status);
 343
 344         u_strToUTF8(name_str, 128, &name_str_len,
 345                     name, name_len,
 346                     &status);
 347
 348         localname_len
 349             = uloc_getDisplayName(uloc_getAvailable(i), uloc_getAvailable(i),
 350                                   localname, 64,
 351                                   &status);
 352
 353         u_strToUTF8(localname_str, 128, &localname_str_len,
 354                     localname, localname_len,
 355                     &status);
 356
 357
 358         if (p_config->xmloutput)
 359         {
 360             fprintf(config.outfile, "<locale id=\"%s\"", uloc_getAvailable(i));
 361             /* fprintf(config.outfile, " locale=\"%s\"", uloc_getAvailable(i)); */
 362             /* if (strlen(keyword_str)) */
 363             /*   fprintf(config.outfile, " keyword=\"%s\"", keyword_str); */
 364             /* if (ucol_getAvailable(i)) */
 365             /*   fprintf(config.outfile, " collation=\"1\""); */
 366             if (strlen(lang_str))
 367                 fprintf(config.outfile, " language=\"%s\"", lang_str);
 368             if (strlen(script_str))
 369                 fprintf(config.outfile, " script=\"%s\"", script_str);
 370             if (strlen(location_str))
 371                 fprintf(config.outfile, " location=\"%s\"", location_str);
 372             if (strlen(variant_str))
 373                 fprintf(config.outfile, " variant=\"%s\"", variant_str);
 374             if (strlen(name_str))
 375                 fprintf(config.outfile, " name=\"%s\"", name_str);
 376             if (strlen(localname_str))
 377                 fprintf(config.outfile, " localname=\"%s\"", localname_str);
 378             fprintf(config.outfile, ">");
 379             if (strlen(localname_str))
 380                 fprintf(config.outfile, "%s", localname_str);
 381             fprintf(config.outfile, "</locale>\n");
 382         }
 383         else if (1 == p_config->xmloutput)
 384         {
 385             fprintf(config.outfile, "%s", uloc_getAvailable(i));
 386             fprintf(config.outfile, " | ");
 387             if (strlen(name_str))
 388                 fprintf(config.outfile, "%s", name_str);
 389             fprintf(config.outfile, " | ");
 390             if (strlen(localname_str))
 391                 fprintf(config.outfile, "%s", localname_str);
 392             fprintf(config.outfile, "\n");
 393         }
 394         else
 395             fprintf(config.outfile, "%s\n", uloc_getAvailable(i));
 396     }
 397     if (p_config->xmloutput)
 398         fprintf(config.outfile, "</locales>\n");
 399     else
 400         fprintf(config.outfile, "\n");
 401
 402     if (U_FAILURE(status))
 403     {
 404         fprintf(stderr, "ICU Error: %d %s\n", status, u_errorName(status));
 405         exit(2);
 406     }
 407 }
 408
 409
 410 static void print_info(const struct config_t *p_config)
 411 {
 412     if (p_config->xmloutput)
 413         fprintf(config.outfile, "<?xml version=\"1.0\" encoding=\"UTF-8\"?>\n"
 414                 "<icu>\n");
 415
 416     if ('c' == config.print[0])
 417         print_icu_converters(&config);
 418     else if ('l' == config.print[0])
 419         print_icu_xml_locales(&config);
 420     else if ('t' == config.print[0])
 421         print_icu_transliterators(&config);
 422     else {
 423         print_icu_converters(&config);
 424         print_icu_xml_locales(&config);
 425         print_icu_transliterators(&config);
 426     }
 427
 428     if (p_config->xmloutput)
 429         fprintf(config.outfile, "</icu>\n");
 430
 431     exit(0);
 432 }
 433
 434
 435
 436 static void process_text_file(const struct config_t *p_config)
 437 {
 438     char *line = 0;
 439     char linebuf[1024];
 440
 441     xmlDoc *doc = xmlParseFile(config.conffile);
 442     xmlNode *xml_node = xmlDocGetRootElement(doc);
 443
 444     long unsigned int token_count = 0;
 445     long unsigned int line_count = 0;
 446
 447     UErrorCode status = U_ZERO_ERROR;
 448
 449     if (!xml_node)
 450     {
 451         printf("Could not parse XML config file '%s' \n",
 452                 config.conffile);
 453         exit(1);
 454     }
 455
 456     config.chain = icu_chain_xml_config(xml_node, 1, &status);
 457
 458     if (!config.chain || !U_SUCCESS(status))
 459     {
 460         printf("Could not set up ICU chain from config file '%s' \n",
 461                 config.conffile);
 462         if (!U_SUCCESS(status))
 463             printf("ICU Error: %d %s\n", status, u_errorName(status));
 464         exit(1);
 465     }
 466
 467     if (p_config->xmloutput)
 468         fprintf(config.outfile,
 469                 "<?xml version=\"1.0\" encoding=\"UTF-8\"?>\n"
 470                 "<icu>\n"
 471                 "<tokens>\n");
 472
 473     /* read input lines for processing */
 474     while ((line=fgets(linebuf, sizeof(linebuf)-1, config.infile)))
 475     {
 476         WRBUF sw = wrbuf_alloc();
 477         WRBUF cdata = wrbuf_alloc();
 478         int success = icu_chain_assign_cstr(config.chain, line, &status);
 479         line_count++;
 480
 481         while (success && icu_chain_next_token(config.chain, &status))
 482         {
 483             if (U_FAILURE(status))
 484                 success = 0;
 485             else
 486             {
 487                 const char *sortkey = icu_chain_token_sortkey(config.chain);
 488                 wrbuf_rewind(sw);
 489                 wrbuf_puts_escaped(sw, sortkey);
 490                 token_count++;
 491                 if (p_config->xmloutput)
 492                 {
 493                     fprintf(config.outfile,
 494                             "<token id=\"%lu\" line=\"%lu\"",
 495                             token_count, line_count);
 496
 497                     wrbuf_rewind(cdata);
 498                     wrbuf_xmlputs(cdata, icu_chain_token_norm(config.chain));
 499                     fprintf(config.outfile, " norm=\"%s\"",
 500                             wrbuf_cstr(cdata));
 501
 502                     wrbuf_rewind(cdata);
 503                     wrbuf_xmlputs(cdata, icu_chain_token_display(config.chain));
 504                     fprintf(config.outfile, " display=\"%s\"",
 505                             wrbuf_cstr(cdata));
 506
 507                     if (p_config->sortoutput)
 508                     {
 509                         wrbuf_rewind(cdata);
 510                         wrbuf_xmlputs(cdata, wrbuf_cstr(sw));
 511                         fprintf(config.outfile, " sortkey=\"%s\"",
 512                                 wrbuf_cstr(cdata));
 513                     }
 514                     fprintf(config.outfile, "/>\n");
 515                 }
 516                 else
 517                 {
 518                     fprintf(config.outfile, "%lu %lu '%s' '%s'",
 519                             token_count,
 520                             line_count,
 521                             icu_chain_token_norm(config.chain),
 522                             icu_chain_token_display(config.chain));
 523                     if (p_config->sortoutput)
 524                     {
 525                         fprintf(config.outfile, " '%s'", wrbuf_cstr(sw));
 526                     }
 527                     fprintf(config.outfile, "\n");
 528                 }
 529             }
 530         }
 531         wrbuf_destroy(sw);
 532         wrbuf_destroy(cdata);
 533     }
 534
 535     if (p_config->xmloutput)
 536         fprintf(config.outfile,
 537                 "</tokens>\n"
 538                 "</icu>\n");
 539
 540     icu_chain_destroy(config.chain);
 541     xmlFreeDoc(doc);
 542     if (line)
 543         free(line);
 544 }
 545
 546 #endif /* YAZ_HAVE_ICU */
 547
 548
 549 int main(int argc, char **argv)
 550 {
 551
 552 #if YAZ_HAVE_ICU
 553
 554     read_params(argc, argv, &config);
 555
 556     if (config.conffile && strlen(config.conffile))
 557         process_text_file(&config);
 558
 559     if (config.print && strlen(config.print))
 560         print_info(&config);
 561
 562 #else /* YAZ_HAVE_ICU */
 563
 564     printf("ICU not available on your system.\n"
 565            "Please install libicu-dev and icu-doc or similar, "
 566            "re-configure and re-compile\n");
 567
 568
 569     exit(3);
 570 #endif /* YAZ_HAVE_ICU */
 571
 572     return 0;
 573 }
 574
 575
 576 /*
 577  * Local variables:
 578  * c-basic-offset: 4
 579  * c-file-style: "Stroustrup"
 580  * indent-tabs-mode: nil
 581  * End:
 582  * vim: shiftwidth=4 tabstop=8 expandtab
 583  */
 584