util/yaz-icu.c

   1 /* This file is part of the YAZ toolkit.
   2  * Copyright (C) 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 #include <errno.h>
  15
  16 #include <yaz/options.h>
  17
  18 #if YAZ_HAVE_ICU
  19
  20 #include <unicode/ucnv.h>
  21 #include <unicode/ustring.h>
  22 #include <unicode/ucol.h>
  23 #include <unicode/ubrk.h>
  24 #include <unicode/utrans.h>
  25 #include <unicode/uclean.h>
  26
  27 #include <yaz/icu.h>
  28 #include <yaz/wrbuf.h>
  29 #include <yaz/backtrace.h>
  30
  31 /* commando line and config parameters */
  32 struct config_t {
  33     char conffile[1024];
  34     char print[1024];
  35     int xmloutput;
  36     int sortoutput;
  37     int org_output;
  38     yaz_icu_chain_t chain;
  39     FILE * infile;
  40     FILE * outfile;
  41 };
  42
  43 void print_option_error(const struct config_t *p_config)
  44 {
  45     fprintf(stderr, "yaz-icu [options] [infile]\n"
  46             "Options:\n"
  47             "   -c file         XML configuration\n"
  48             "   -p a|c|l|t      Print ICU info \n"
  49             "   -s              Show sort normalization key\n"
  50             "   -o              Show org positions\n"
  51             "   -x              XML output instread of text\n"
  52             "\n"
  53             "Examples:\n"
  54             "cat hugetextfile.txt | ./yaz-icu -c config.xml \n"
  55             "./yaz-icu -p c\n"
  56             "./yaz-icu -p l -x\n"
  57             "./yaz-icu -p t -x\n"
  58             "\n"
  59             "Example ICU chain XML configuration file:\n"
  60             "<icu_chain locale=\"en\">\n"
  61             "  <transform rule=\"[:Control:] Any-Remove\"/>\n"
  62             "  <tokenize rule=\"l\"/>\n"
  63             "  <transform rule=\"[[:WhiteSpace:][:Punctuation:]] Remove\"/>\n"
  64             "  <casemap rule=\"l\"/>\n"
  65             "</icu_chain>\n"
  66           );
  67     exit(1);
  68 }
  69
  70 void read_params(int argc, char **argv, struct config_t *p_config)
  71 {
  72     char *arg;
  73     int ret;
  74
  75     /* set default parameters */
  76     p_config->conffile[0] = 0;
  77     p_config->print[0] = 0;
  78     p_config->xmloutput = 0;
  79     p_config->sortoutput = 0;
  80     p_config->chain = 0;
  81     p_config->infile = 0;
  82     p_config->outfile = stdout;
  83     p_config->org_output = 0;
  84
  85     /* set up command line parameters */
  86
  87     while ((ret = options("c:op:sx", argv, argc, &arg)) != -2)
  88     {
  89         switch (ret)
  90         {
  91         case 'c':
  92             strcpy(p_config->conffile, arg);
  93             break;
  94         case 'p':
  95             strcpy(p_config->print, arg);
  96             break;
  97         case 's':
  98             p_config->sortoutput = 1;
  99             break;
 100         case 'x':
 101             p_config->xmloutput = 1;
 102             break;
 103         case 'o':
 104             p_config->org_output = 1;
 105             break;
 106         case 0:
 107             if (p_config->infile)
 108             {
 109                 fprintf(stderr, "yaz-icu: only one input file may be given\n");
 110                 print_option_error(p_config);
 111             }
 112             p_config->infile = fopen(arg, "r");
 113             if (!p_config->infile)
 114             {
 115                 fprintf(stderr, "yaz-icu: cannot open %s : %s\n",
 116                         arg, strerror(errno));
 117                 exit(1);
 118             }
 119             break;
 120         default:
 121             fprintf(stderr, "yaz_icu: invalid option: %s\n", arg);
 122             print_option_error(p_config);
 123         }
 124     }
 125
 126     if (p_config->infile == 0)
 127         p_config->infile = stdin;
 128
 129     if (!strlen(p_config->conffile) && !strlen(p_config->print))
 130         print_option_error(p_config);
 131 }
 132
 133 static void print_icu_converters(const struct config_t *p_config)
 134 {
 135     int32_t count;
 136     int32_t i;
 137
 138     count = ucnv_countAvailable();
 139     if (p_config->xmloutput)
 140         fprintf(p_config->outfile, "<converters count=\"%d\" default=\"%s\">\n",
 141                 count, ucnv_getDefaultName());
 142     else
 143     {
 144         fprintf(p_config->outfile, "Available ICU converters: %d\n", count);
 145         fprintf(p_config->outfile, "Default ICU Converter is: '%s'\n",
 146                 ucnv_getDefaultName());
 147     }
 148
 149     for (i = 0; i < count; i++)
 150     {
 151         if (p_config->xmloutput)
 152             fprintf(p_config->outfile, "<converter id=\"%s\"/>\n",
 153                     ucnv_getAvailableName(i));
 154         else
 155             fprintf(p_config->outfile, "%s\n", ucnv_getAvailableName(i));
 156     }
 157
 158     if (p_config->xmloutput)
 159         fprintf(p_config->outfile, "</converters>\n");
 160     else
 161         fprintf(p_config->outfile, "\n");
 162 }
 163
 164 static void print_icu_transliterators(const struct config_t *p_config)
 165 {
 166     UErrorCode status;
 167     UEnumeration *en = utrans_openIDs(&status);
 168     int32_t count = uenum_count(en, &status);
 169     const char *name;
 170     int32_t length;
 171
 172     if (p_config->xmloutput)
 173         fprintf(p_config->outfile, "<transliterators count=\"%d\">\n",  count);
 174     else
 175         fprintf(p_config->outfile, "Available ICU transliterators: %d\n", count);
 176
 177     while ((name = uenum_next(en, &length, &status)))
 178     {
 179         if (p_config->xmloutput)
 180             fprintf(p_config->outfile, "<transliterator id=\"%s\"/>\n", name);
 181         else
 182             fprintf(p_config->outfile, "%s\n", name);
 183     }
 184     uenum_close(en);
 185     if (p_config->xmloutput)
 186         fprintf(p_config->outfile, "</transliterators>\n");
 187     else
 188     {
 189         fprintf(p_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://userguide.icu-project.org/transforms/general\n"
 227                 "    http://www.unicode.org/reports/tr44/\n"
 228             );
 229
 230
 231         fprintf(p_config->outfile, "\n\n");
 232
 233     }
 234 }
 235
 236 static void print_icu_xml_locales(const struct config_t *p_config)
 237 {
 238     int32_t count;
 239     int32_t i;
 240     UErrorCode status = U_ZERO_ERROR;
 241
 242     UChar keyword[64];
 243     int32_t keyword_len = 0;
 244     char keyword_str[128];
 245     int32_t keyword_str_len = 0;
 246
 247     UChar language[64];
 248     int32_t language_len = 0;
 249     char lang_str[128];
 250     int32_t lang_str_len = 0;
 251
 252     UChar script[64];
 253     int32_t script_len = 0;
 254     char script_str[128];
 255     int32_t script_str_len = 0;
 256
 257     UChar location[64];
 258     int32_t location_len = 0;
 259     char location_str[128];
 260     int32_t location_str_len = 0;
 261
 262     UChar variant[64];
 263     int32_t variant_len = 0;
 264     char variant_str[128];
 265     int32_t variant_str_len = 0;
 266
 267     UChar name[64];
 268     int32_t name_len = 0;
 269     char name_str[128];
 270     int32_t name_str_len = 0;
 271
 272     UChar localname[64];
 273     int32_t localname_len = 0;
 274     char localname_str[128];
 275     int32_t localname_str_len = 0;
 276
 277     count = uloc_countAvailable() ;
 278
 279     if (p_config->xmloutput)
 280     {
 281         fprintf(p_config->outfile, "<locales count=\"%d\" default=\"%s\" collations=\"%d\">\n",
 282                 count, uloc_getDefault(), ucol_countAvailable());
 283     }
 284     else
 285     {
 286         fprintf(p_config->outfile, "Available ICU locales: %d\n", count);
 287         fprintf(p_config->outfile, "Default locale is: %s\n",  uloc_getDefault());
 288     }
 289
 290     for (i = 0; i < count; i++)
 291     {
 292
 293         keyword_len
 294             = uloc_getDisplayKeyword(uloc_getAvailable(i), "en",
 295                                      keyword, 64,
 296                                      &status);
 297
 298         u_strToUTF8(keyword_str, 128, &keyword_str_len,
 299                     keyword, keyword_len,
 300                     &status);
 301
 302
 303         language_len
 304             = uloc_getDisplayLanguage(uloc_getAvailable(i), "en",
 305                                       language, 64,
 306                                       &status);
 307
 308         u_strToUTF8(lang_str, 128, &lang_str_len,
 309                     language, language_len,
 310                     &status);
 311
 312
 313         script_len
 314             = uloc_getDisplayScript(uloc_getAvailable(i), "en",
 315                                     script, 64,
 316                                     &status);
 317
 318         u_strToUTF8(script_str, 128, &script_str_len,
 319                     script, script_len,
 320                     &status);
 321
 322         location_len
 323             = uloc_getDisplayCountry(uloc_getAvailable(i), "en",
 324                                      location, 64,
 325                                      &status);
 326
 327         u_strToUTF8(location_str, 128, &location_str_len,
 328                     location, location_len,
 329                     &status);
 330
 331         variant_len
 332             = uloc_getDisplayVariant(uloc_getAvailable(i), "en",
 333                                      variant, 64,
 334                                      &status);
 335
 336         u_strToUTF8(variant_str, 128, &variant_str_len,
 337                     variant, variant_len,
 338                     &status);
 339
 340         name_len
 341             = uloc_getDisplayName(uloc_getAvailable(i), "en",
 342                                   name, 64,
 343                                   &status);
 344
 345         u_strToUTF8(name_str, 128, &name_str_len,
 346                     name, name_len,
 347                     &status);
 348
 349         localname_len
 350             = uloc_getDisplayName(uloc_getAvailable(i), uloc_getAvailable(i),
 351                                   localname, 64,
 352                                   &status);
 353
 354         u_strToUTF8(localname_str, 128, &localname_str_len,
 355                     localname, localname_len,
 356                     &status);
 357
 358
 359         if (p_config->xmloutput)
 360         {
 361             fprintf(p_config->outfile, "<locale id=\"%s\"", uloc_getAvailable(i));
 362             if (strlen(lang_str))
 363                 fprintf(p_config->outfile, " language=\"%s\"", lang_str);
 364             if (strlen(script_str))
 365                 fprintf(p_config->outfile, " script=\"%s\"", script_str);
 366             if (strlen(location_str))
 367                 fprintf(p_config->outfile, " location=\"%s\"", location_str);
 368             if (strlen(variant_str))
 369                 fprintf(p_config->outfile, " variant=\"%s\"", variant_str);
 370             if (strlen(name_str))
 371                 fprintf(p_config->outfile, " name=\"%s\"", name_str);
 372             if (strlen(localname_str))
 373                 fprintf(p_config->outfile, " localname=\"%s\"", localname_str);
 374             fprintf(p_config->outfile, ">");
 375             if (strlen(localname_str))
 376                 fprintf(p_config->outfile, "%s", localname_str);
 377             fprintf(p_config->outfile, "</locale>\n");
 378         }
 379         else if (1 == p_config->xmloutput)
 380         {
 381             fprintf(p_config->outfile, "%s", uloc_getAvailable(i));
 382             fprintf(p_config->outfile, " | ");
 383             if (strlen(name_str))
 384                 fprintf(p_config->outfile, "%s", name_str);
 385             fprintf(p_config->outfile, " | ");
 386             if (strlen(localname_str))
 387                 fprintf(p_config->outfile, "%s", localname_str);
 388             fprintf(p_config->outfile, "\n");
 389         }
 390         else
 391             fprintf(p_config->outfile, "%s\n", uloc_getAvailable(i));
 392     }
 393     if (p_config->xmloutput)
 394         fprintf(p_config->outfile, "</locales>\n");
 395     else
 396         fprintf(p_config->outfile, "\n");
 397
 398     if (U_FAILURE(status))
 399     {
 400         fprintf(stderr, "ICU Error: %d %s\n", status, u_errorName(status));
 401         exit(2);
 402     }
 403 }
 404
 405
 406 static void print_info(const struct config_t *p_config)
 407 {
 408     if (p_config->xmloutput)
 409         fprintf(p_config->outfile, "<?xml version=\"1.0\" encoding=\"UTF-8\"?>\n"
 410                 "<icu>\n");
 411
 412     if ('c' == p_config->print[0])
 413         print_icu_converters(p_config);
 414     else if ('l' == p_config->print[0])
 415         print_icu_xml_locales(p_config);
 416     else if ('t' == p_config->print[0])
 417         print_icu_transliterators(p_config);
 418     else {
 419         print_icu_converters(p_config);
 420         print_icu_xml_locales(p_config);
 421         print_icu_transliterators(p_config);
 422     }
 423
 424     if (p_config->xmloutput)
 425         fprintf(p_config->outfile, "</icu>\n");
 426
 427     exit(0);
 428 }
 429
 430
 431
 432 static void process_text_file(struct config_t *p_config)
 433 {
 434     char *line = 0;
 435     char linebuf[1024];
 436
 437     xmlDoc *doc = xmlParseFile(p_config->conffile);
 438     xmlNode *xml_node = xmlDocGetRootElement(doc);
 439
 440     long unsigned int token_count = 0;
 441     long unsigned int line_count = 0;
 442
 443     UErrorCode status = U_ZERO_ERROR;
 444
 445     if (!xml_node)
 446     {
 447         printf("Could not parse XML config file '%s' \n",
 448                 p_config->conffile);
 449         exit(1);
 450     }
 451
 452     p_config->chain = icu_chain_xml_config(xml_node, 1, &status);
 453
 454     if (!p_config->chain || !U_SUCCESS(status))
 455     {
 456         printf("Could not set up ICU chain from config file '%s' \n",
 457                 p_config->conffile);
 458         exit(1);
 459     }
 460
 461     if (p_config->xmloutput)
 462         fprintf(p_config->outfile,
 463                 "<?xml version=\"1.0\" encoding=\"UTF-8\"?>\n"
 464                 "<icu>\n"
 465                 "<tokens>\n");
 466
 467     /* read input lines for processing */
 468     while ((line=fgets(linebuf, sizeof(linebuf)-1, p_config->infile)))
 469     {
 470         WRBUF sw = wrbuf_alloc();
 471         WRBUF cdata = wrbuf_alloc();
 472         int success = icu_chain_assign_cstr(p_config->chain, line, &status);
 473         line_count++;
 474
 475         while (success && icu_chain_next_token(p_config->chain, &status))
 476         {
 477             if (U_FAILURE(status))
 478                 success = 0;
 479             else
 480             {
 481                 size_t start, len;
 482                 const char *org_string = 0;
 483                 const char *sortkey = icu_chain_token_sortkey(p_config->chain);
 484
 485                 icu_chain_get_org_info2(p_config->chain, &start, &len,
 486                                         &org_string);
 487                 wrbuf_rewind(sw);
 488                 wrbuf_puts_escaped(sw, sortkey);
 489                 token_count++;
 490                 if (p_config->xmloutput)
 491                 {
 492                     fprintf(p_config->outfile,
 493                             "<token id=\"%lu\" line=\"%lu\"",
 494                             token_count, line_count);
 495
 496                     wrbuf_rewind(cdata);
 497                     wrbuf_xmlputs(cdata, icu_chain_token_norm(p_config->chain));
 498                     fprintf(p_config->outfile, " norm=\"%s\"",
 499                             wrbuf_cstr(cdata));
 500
 501                     wrbuf_rewind(cdata);
 502                     wrbuf_xmlputs(cdata, icu_chain_token_display(p_config->chain));
 503                     fprintf(p_config->outfile, " display=\"%s\"",
 504                             wrbuf_cstr(cdata));
 505
 506                     if (p_config->sortoutput)
 507                     {
 508                         wrbuf_rewind(cdata);
 509                         wrbuf_xmlputs(cdata, wrbuf_cstr(sw));
 510                         fprintf(p_config->outfile, " sortkey=\"%s\"",
 511                                 wrbuf_cstr(cdata));
 512                     }
 513                     fprintf(p_config->outfile, "/>\n");
 514                 }
 515                 else
 516                 {
 517                     fprintf(p_config->outfile, "%lu %lu '%s' '%s'",
 518                             token_count,
 519                             line_count,
 520                             icu_chain_token_norm(p_config->chain),
 521                             icu_chain_token_display(p_config->chain));
 522                     if (p_config->sortoutput)
 523                     {
 524                         fprintf(p_config->outfile, " '%s'", wrbuf_cstr(sw));
 525                     }
 526                     if (p_config->org_output)
 527                     {
 528                         fprintf(p_config->outfile, " %ld+%ld",
 529                                 (long) start, (long) len);
 530                         fputc(' ', p_config->outfile);
 531                         fwrite(org_string, 1, start, p_config->outfile);
 532                         fputc('*', p_config->outfile);
 533                         fwrite(org_string + start, 1, len, p_config->outfile);
 534                         fputc('*', p_config->outfile);
 535                         fputs(org_string + start + len, p_config->outfile);
 536                     }
 537                     fprintf(p_config->outfile, "\n");
 538                 }
 539             }
 540         }
 541         wrbuf_destroy(sw);
 542         wrbuf_destroy(cdata);
 543     }
 544
 545     if (p_config->xmloutput)
 546         fprintf(p_config->outfile,
 547                 "</tokens>\n"
 548                 "</icu>\n");
 549
 550     icu_chain_destroy(p_config->chain);
 551     xmlFreeDoc(doc);
 552     if (line)
 553         free(line);
 554 }
 555
 556 #endif /* YAZ_HAVE_ICU */
 557
 558
 559 int main(int argc, char **argv)
 560 {
 561 #if YAZ_HAVE_ICU
 562     struct config_t config;
 563
 564     yaz_enable_panic_backtrace(*argv);
 565     read_params(argc, argv, &config);
 566
 567     if (config.conffile && strlen(config.conffile))
 568         process_text_file(&config);
 569
 570     if (config.print && strlen(config.print))
 571         print_info(&config);
 572
 573     u_cleanup();
 574 #else /* YAZ_HAVE_ICU */
 575
 576     printf("ICU not available on your system.\n"
 577            "Please install libicu-dev and icu-doc or similar, "
 578            "re-configure and re-compile\n");
 579
 580
 581     exit(3);
 582 #endif /* YAZ_HAVE_ICU */
 583
 584     return 0;
 585 }
 586
 587
 588 /*
 589  * Local variables:
 590  * c-basic-offset: 4
 591  * c-file-style: "Stroustrup"
 592  * indent-tabs-mode: nil
 593  * End:
 594  * vim: shiftwidth=4 tabstop=8 expandtab
 595  */
 596