util/yaz-icu.c

   1 /* This file is part of the YAZ toolkit.
   2  * Copyright (C) 1995-2013 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
  30 /* commando line and config parameters */
  31 struct config_t {
  32     char conffile[1024];
  33     char print[1024];
  34     int xmloutput;
  35     int sortoutput;
  36     yaz_icu_chain_t chain;
  37     FILE * infile;
  38     FILE * outfile;
  39 };
  40
  41 void print_option_error(const struct config_t *p_config)
  42 {
  43     fprintf(stderr, "yaz-icu [options] [infile]\n"
  44             "Options:\n"
  45             "   -c file         XML configuration\n"
  46             "   -p a|c|l|t      Print ICU info \n"
  47             "   -s              Show sort normalization key\n"
  48             "   -x              XML output instread of text\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 = 0;
  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         case 0:
 100             if (p_config->infile)
 101             {
 102                 fprintf(stderr, "yaz-icu: only one input file may be given\n");
 103                 print_option_error(p_config);
 104             }
 105             p_config->infile = fopen(arg, "r");
 106             if (!p_config->infile)
 107             {
 108                 fprintf(stderr, "yaz-icu: cannot open %s : %s\n",
 109                         arg, strerror(errno));
 110                 exit(1);
 111             }
 112             break;
 113         default:
 114             fprintf(stderr, "yaz_icu: invalid option: %s\n", arg);
 115             print_option_error(p_config);
 116         }
 117     }
 118
 119     if (p_config->infile == 0)
 120         p_config->infile = stdin;
 121
 122     if (!strlen(p_config->conffile) && !strlen(p_config->print))
 123         print_option_error(p_config);
 124 }
 125
 126 static void print_icu_converters(const struct config_t *p_config)
 127 {
 128     int32_t count;
 129     int32_t i;
 130
 131     count = ucnv_countAvailable();
 132     if (p_config->xmloutput)
 133         fprintf(p_config->outfile, "<converters count=\"%d\" default=\"%s\">\n",
 134                 count, ucnv_getDefaultName());
 135     else
 136     {
 137         fprintf(p_config->outfile, "Available ICU converters: %d\n", count);
 138         fprintf(p_config->outfile, "Default ICU Converter is: '%s'\n",
 139                 ucnv_getDefaultName());
 140     }
 141
 142     for (i = 0; i < count; i++)
 143     {
 144         if (p_config->xmloutput)
 145             fprintf(p_config->outfile, "<converter id=\"%s\"/>\n",
 146                     ucnv_getAvailableName(i));
 147         else
 148             fprintf(p_config->outfile, "%s\n", ucnv_getAvailableName(i));
 149     }
 150
 151     if (p_config->xmloutput)
 152         fprintf(p_config->outfile, "</converters>\n");
 153     else
 154         fprintf(p_config->outfile, "\n");
 155 }
 156
 157 static void print_icu_transliterators(const struct config_t *p_config)
 158 {
 159     UErrorCode status;
 160     UEnumeration *en = utrans_openIDs(&status);
 161     int32_t count = uenum_count(en, &status);
 162     const char *name;
 163     int32_t length;
 164
 165     if (p_config->xmloutput)
 166         fprintf(p_config->outfile, "<transliterators count=\"%d\">\n",  count);
 167     else
 168         fprintf(p_config->outfile, "Available ICU transliterators: %d\n", count);
 169
 170     while ((name = uenum_next(en, &length, &status)))
 171     {
 172         if (p_config->xmloutput)
 173             fprintf(p_config->outfile, "<transliterator id=\"%s\"/>\n", name);
 174         else
 175             fprintf(p_config->outfile, "%s\n", name);
 176     }
 177     uenum_close(en);
 178     if (p_config->xmloutput)
 179         fprintf(p_config->outfile, "</transliterators>\n");
 180     else
 181     {
 182         fprintf(p_config->outfile, "\n\nUnicode Set Patterns:\n"
 183                 "   Pattern         Description\n"
 184                 "   Ranges          [a-z]       The lower case letters a through z\n"
 185                 "   Named Chars     [abc123] The six characters a,b,c,1,2 and 3\n"
 186                 "   String          [abc{def}] chars a, b and c, and string 'def'\n"
 187                 "   Categories      [\\p{Letter}] Perl General Category 'Letter'.\n"
 188                 "   Categories      [:Letter:] Posix General Category 'Letter'.\n"
 189                 "\n"
 190                 "   Combination     Example\n"
 191                 "   Union           [[:Greek:] [:letter:]]\n"
 192                 "   Intersection    [[:Greek:] & [:letter:]]\n"
 193                 "   Set Complement  [[:Greek:] - [:letter:]]\n"
 194                 "   Complement      [^[:Greek:] [:letter:]]\n"
 195                 "\n"
 196              "see: http://icu.sourceforge.net/userguide/unicodeSet.html\n"
 197                 "\n"
 198                 "Examples:\n"
 199                 "   [:Punctuation:] Any-Remove\n"
 200                 "   [:Cased-Letter:] Any-Upper\n"
 201                 "   [:Control:] Any-Remove\n"
 202                 "   [:Decimal_Number:] Any-Remove\n"
 203                 "   [:Final_Punctuation:] Any-Remove\n"
 204                 "   [:Georgian:] Any-Upper\n"
 205                 "   [:Katakana:] Any-Remove\n"
 206                 "   [:Arabic:] Any-Remove\n"
 207                 "   [:Punctuation:] Remove\n"
 208                 "   [[:Punctuation:]-[.,]] Remove\n"
 209                 "   [:Line_Separator:] Any-Remove\n"
 210                 "   [:Math_Symbol:] Any-Remove\n"
 211                 "   Lower; [:^Letter:] Remove (word tokenization)\n"
 212                 "   [:^Number:] Remove (numeric tokenization)\n"
 213                 "   [:^Katagana:] Remove (remove everything except Katagana)\n"
 214                 "   Lower;[[:WhiteSpace:][:Punctuation:]] Remove (word tokenization)\n"
 215                 "   NFD; [:Nonspacing Mark:] Remove; NFC   (removes accents from characters)\n"
 216                 "   [A-Za-z]; Lower(); Latin-Katakana; Katakana-Hiragana (transforms latin and katagana to hiragana)\n"
 217                 "   [[:separator:][:start punctuation:][:initial punctuation:]] Remove \n"
 218                 "\n"
 219                 "see http://userguide.icu-project.org/transforms/general\n"
 220                 "    http://www.unicode.org/reports/tr44/\n"
 221             );
 222
 223
 224         fprintf(p_config->outfile, "\n\n");
 225
 226     }
 227 }
 228
 229 static void print_icu_xml_locales(const struct config_t *p_config)
 230 {
 231     int32_t count;
 232     int32_t i;
 233     UErrorCode status = U_ZERO_ERROR;
 234
 235     UChar keyword[64];
 236     int32_t keyword_len = 0;
 237     char keyword_str[128];
 238     int32_t keyword_str_len = 0;
 239
 240     UChar language[64];
 241     int32_t language_len = 0;
 242     char lang_str[128];
 243     int32_t lang_str_len = 0;
 244
 245     UChar script[64];
 246     int32_t script_len = 0;
 247     char script_str[128];
 248     int32_t script_str_len = 0;
 249
 250     UChar location[64];
 251     int32_t location_len = 0;
 252     char location_str[128];
 253     int32_t location_str_len = 0;
 254
 255     UChar variant[64];
 256     int32_t variant_len = 0;
 257     char variant_str[128];
 258     int32_t variant_str_len = 0;
 259
 260     UChar name[64];
 261     int32_t name_len = 0;
 262     char name_str[128];
 263     int32_t name_str_len = 0;
 264
 265     UChar localname[64];
 266     int32_t localname_len = 0;
 267     char localname_str[128];
 268     int32_t localname_str_len = 0;
 269
 270     count = uloc_countAvailable() ;
 271
 272     if (p_config->xmloutput)
 273     {
 274         fprintf(p_config->outfile, "<locales count=\"%d\" default=\"%s\" collations=\"%d\">\n",
 275                 count, uloc_getDefault(), ucol_countAvailable());
 276     }
 277     else
 278     {
 279         fprintf(p_config->outfile, "Available ICU locales: %d\n", count);
 280         fprintf(p_config->outfile, "Default locale is: %s\n",  uloc_getDefault());
 281     }
 282
 283     for (i = 0; i < count; i++)
 284     {
 285
 286         keyword_len
 287             = uloc_getDisplayKeyword(uloc_getAvailable(i), "en",
 288                                      keyword, 64,
 289                                      &status);
 290
 291         u_strToUTF8(keyword_str, 128, &keyword_str_len,
 292                     keyword, keyword_len,
 293                     &status);
 294
 295
 296         language_len
 297             = uloc_getDisplayLanguage(uloc_getAvailable(i), "en",
 298                                       language, 64,
 299                                       &status);
 300
 301         u_strToUTF8(lang_str, 128, &lang_str_len,
 302                     language, language_len,
 303                     &status);
 304
 305
 306         script_len
 307             = uloc_getDisplayScript(uloc_getAvailable(i), "en",
 308                                     script, 64,
 309                                     &status);
 310
 311         u_strToUTF8(script_str, 128, &script_str_len,
 312                     script, script_len,
 313                     &status);
 314
 315         location_len
 316             = uloc_getDisplayCountry(uloc_getAvailable(i), "en",
 317                                      location, 64,
 318                                      &status);
 319
 320         u_strToUTF8(location_str, 128, &location_str_len,
 321                     location, location_len,
 322                     &status);
 323
 324         variant_len
 325             = uloc_getDisplayVariant(uloc_getAvailable(i), "en",
 326                                      variant, 64,
 327                                      &status);
 328
 329         u_strToUTF8(variant_str, 128, &variant_str_len,
 330                     variant, variant_len,
 331                     &status);
 332
 333         name_len
 334             = uloc_getDisplayName(uloc_getAvailable(i), "en",
 335                                   name, 64,
 336                                   &status);
 337
 338         u_strToUTF8(name_str, 128, &name_str_len,
 339                     name, name_len,
 340                     &status);
 341
 342         localname_len
 343             = uloc_getDisplayName(uloc_getAvailable(i), uloc_getAvailable(i),
 344                                   localname, 64,
 345                                   &status);
 346
 347         u_strToUTF8(localname_str, 128, &localname_str_len,
 348                     localname, localname_len,
 349                     &status);
 350
 351
 352         if (p_config->xmloutput)
 353         {
 354             fprintf(p_config->outfile, "<locale id=\"%s\"", uloc_getAvailable(i));
 355             if (strlen(lang_str))
 356                 fprintf(p_config->outfile, " language=\"%s\"", lang_str);
 357             if (strlen(script_str))
 358                 fprintf(p_config->outfile, " script=\"%s\"", script_str);
 359             if (strlen(location_str))
 360                 fprintf(p_config->outfile, " location=\"%s\"", location_str);
 361             if (strlen(variant_str))
 362                 fprintf(p_config->outfile, " variant=\"%s\"", variant_str);
 363             if (strlen(name_str))
 364                 fprintf(p_config->outfile, " name=\"%s\"", name_str);
 365             if (strlen(localname_str))
 366                 fprintf(p_config->outfile, " localname=\"%s\"", localname_str);
 367             fprintf(p_config->outfile, ">");
 368             if (strlen(localname_str))
 369                 fprintf(p_config->outfile, "%s", localname_str);
 370             fprintf(p_config->outfile, "</locale>\n");
 371         }
 372         else if (1 == p_config->xmloutput)
 373         {
 374             fprintf(p_config->outfile, "%s", uloc_getAvailable(i));
 375             fprintf(p_config->outfile, " | ");
 376             if (strlen(name_str))
 377                 fprintf(p_config->outfile, "%s", name_str);
 378             fprintf(p_config->outfile, " | ");
 379             if (strlen(localname_str))
 380                 fprintf(p_config->outfile, "%s", localname_str);
 381             fprintf(p_config->outfile, "\n");
 382         }
 383         else
 384             fprintf(p_config->outfile, "%s\n", uloc_getAvailable(i));
 385     }
 386     if (p_config->xmloutput)
 387         fprintf(p_config->outfile, "</locales>\n");
 388     else
 389         fprintf(p_config->outfile, "\n");
 390
 391     if (U_FAILURE(status))
 392     {
 393         fprintf(stderr, "ICU Error: %d %s\n", status, u_errorName(status));
 394         exit(2);
 395     }
 396 }
 397
 398
 399 static void print_info(const struct config_t *p_config)
 400 {
 401     if (p_config->xmloutput)
 402         fprintf(p_config->outfile, "<?xml version=\"1.0\" encoding=\"UTF-8\"?>\n"
 403                 "<icu>\n");
 404
 405     if ('c' == p_config->print[0])
 406         print_icu_converters(p_config);
 407     else if ('l' == p_config->print[0])
 408         print_icu_xml_locales(p_config);
 409     else if ('t' == p_config->print[0])
 410         print_icu_transliterators(p_config);
 411     else {
 412         print_icu_converters(p_config);
 413         print_icu_xml_locales(p_config);
 414         print_icu_transliterators(p_config);
 415     }
 416
 417     if (p_config->xmloutput)
 418         fprintf(p_config->outfile, "</icu>\n");
 419
 420     exit(0);
 421 }
 422
 423
 424
 425 static void process_text_file(struct config_t *p_config)
 426 {
 427     char *line = 0;
 428     char linebuf[1024];
 429
 430     xmlDoc *doc = xmlParseFile(p_config->conffile);
 431     xmlNode *xml_node = xmlDocGetRootElement(doc);
 432
 433     long unsigned int token_count = 0;
 434     long unsigned int line_count = 0;
 435
 436     UErrorCode status = U_ZERO_ERROR;
 437
 438     if (!xml_node)
 439     {
 440         printf("Could not parse XML config file '%s' \n",
 441                 p_config->conffile);
 442         exit(1);
 443     }
 444
 445     p_config->chain = icu_chain_xml_config(xml_node, 1, &status);
 446
 447     if (!p_config->chain || !U_SUCCESS(status))
 448     {
 449         printf("Could not set up ICU chain from config file '%s' \n",
 450                 p_config->conffile);
 451         if (!U_SUCCESS(status))
 452             printf("ICU Error: %d %s\n", status, u_errorName(status));
 453         exit(1);
 454     }
 455
 456     if (p_config->xmloutput)
 457         fprintf(p_config->outfile,
 458                 "<?xml version=\"1.0\" encoding=\"UTF-8\"?>\n"
 459                 "<icu>\n"
 460                 "<tokens>\n");
 461
 462     /* read input lines for processing */
 463     while ((line=fgets(linebuf, sizeof(linebuf)-1, p_config->infile)))
 464     {
 465         WRBUF sw = wrbuf_alloc();
 466         WRBUF cdata = wrbuf_alloc();
 467         int success = icu_chain_assign_cstr(p_config->chain, line, &status);
 468         line_count++;
 469
 470         while (success && icu_chain_next_token(p_config->chain, &status))
 471         {
 472             if (U_FAILURE(status))
 473                 success = 0;
 474             else
 475             {
 476                 const char *sortkey = icu_chain_token_sortkey(p_config->chain);
 477                 wrbuf_rewind(sw);
 478                 wrbuf_puts_escaped(sw, sortkey);
 479                 token_count++;
 480                 if (p_config->xmloutput)
 481                 {
 482                     fprintf(p_config->outfile,
 483                             "<token id=\"%lu\" line=\"%lu\"",
 484                             token_count, line_count);
 485
 486                     wrbuf_rewind(cdata);
 487                     wrbuf_xmlputs(cdata, icu_chain_token_norm(p_config->chain));
 488                     fprintf(p_config->outfile, " norm=\"%s\"",
 489                             wrbuf_cstr(cdata));
 490
 491                     wrbuf_rewind(cdata);
 492                     wrbuf_xmlputs(cdata, icu_chain_token_display(p_config->chain));
 493                     fprintf(p_config->outfile, " display=\"%s\"",
 494                             wrbuf_cstr(cdata));
 495
 496                     if (p_config->sortoutput)
 497                     {
 498                         wrbuf_rewind(cdata);
 499                         wrbuf_xmlputs(cdata, wrbuf_cstr(sw));
 500                         fprintf(p_config->outfile, " sortkey=\"%s\"",
 501                                 wrbuf_cstr(cdata));
 502                     }
 503                     fprintf(p_config->outfile, "/>\n");
 504                 }
 505                 else
 506                 {
 507                     fprintf(p_config->outfile, "%lu %lu '%s' '%s'",
 508                             token_count,
 509                             line_count,
 510                             icu_chain_token_norm(p_config->chain),
 511                             icu_chain_token_display(p_config->chain));
 512                     if (p_config->sortoutput)
 513                     {
 514                         fprintf(p_config->outfile, " '%s'", wrbuf_cstr(sw));
 515                     }
 516                     fprintf(p_config->outfile, "\n");
 517                 }
 518             }
 519         }
 520         wrbuf_destroy(sw);
 521         wrbuf_destroy(cdata);
 522     }
 523
 524     if (p_config->xmloutput)
 525         fprintf(p_config->outfile,
 526                 "</tokens>\n"
 527                 "</icu>\n");
 528
 529     icu_chain_destroy(p_config->chain);
 530     xmlFreeDoc(doc);
 531     if (line)
 532         free(line);
 533 }
 534
 535 #endif /* YAZ_HAVE_ICU */
 536
 537
 538 int main(int argc, char **argv)
 539 {
 540 #if YAZ_HAVE_ICU
 541     struct config_t config;
 542
 543     read_params(argc, argv, &config);
 544
 545     if (config.conffile && strlen(config.conffile))
 546         process_text_file(&config);
 547
 548     if (config.print && strlen(config.print))
 549         print_info(&config);
 550
 551     u_cleanup();
 552 #else /* YAZ_HAVE_ICU */
 553
 554     printf("ICU not available on your system.\n"
 555            "Please install libicu-dev and icu-doc or similar, "
 556            "re-configure and re-compile\n");
 557
 558
 559     exit(3);
 560 #endif /* YAZ_HAVE_ICU */
 561
 562     return 0;
 563 }
 564
 565
 566 /*
 567  * Local variables:
 568  * c-basic-offset: 4
 569  * c-file-style: "Stroustrup"
 570  * indent-tabs-mode: nil
 571  * End:
 572  * vim: shiftwidth=4 tabstop=8 expandtab
 573  */
 574