include/yaz/icu_I18N.h

   1 /*
   2  * Copyright (c) 1995-2007, Index Data
   3  * All rights reserved.
   4  * Redistribution and use in source and binary forms, with or without
   5  * modification, are permitted provided that the following conditions are met:
   6  *
   7  *     * Redistributions of source code must retain the above copyright
   8  *       notice, this list of conditions and the following disclaimer.
   9  *     * Redistributions in binary form must reproduce the above copyright
  10  *       notice, this list of conditions and the following disclaimer in the
  11  *       documentation and/or other materials provided with the distribution.
  12  *     * Neither the name of Index Data nor the names of its contributors
  13  *       may be used to endorse or promote products derived from this
  14  *       software without specific prior written permission.
  15  *
  16  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND ANY
  17  * EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED
  18  * WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE
  19  * DISCLAIMED. IN NO EVENT SHALL THE REGENTS AND CONTRIBUTORS BE LIABLE FOR ANY
  20  * DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES
  21  * (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES;
  22  * LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND
  23  * ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
  24  * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF
  25  * THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
  26  */
  27
  28 #ifndef ICU_I18NL_H
  29 #define ICU_I18NL_H
  30
  31 #include <yaz/nmem.h>
  32
  33 #include <libxml/parser.h>
  34 #include <libxml/tree.h>
  35
  36 #include <unicode/utypes.h>   /* Basic ICU data types */
  37 #include <unicode/uchar.h>    /* char names           */
  38
  39 #include <unicode/ucol.h>
  40 #include <unicode/ubrk.h>
  41 #include <unicode/utrans.h>
  42
  43
  44 // #define ICU_CHAIN_SORTKEY
  45 #undef ICU_CHAIN_SORTKEY
  46
  47 /* declared structs and functions */
  48
  49 int icu_check_status (UErrorCode status);
  50
  51 struct icu_buf_utf16
  52 {
  53   UChar * utf16;
  54   int32_t utf16_len;
  55   int32_t utf16_cap;
  56 };
  57
  58 struct icu_buf_utf16 * icu_buf_utf16_create(size_t capacity);
  59
  60 struct icu_buf_utf16 * icu_buf_utf16_clear(struct icu_buf_utf16 * buf16);
  61
  62 struct icu_buf_utf16 * icu_buf_utf16_resize(struct icu_buf_utf16 * buf16,
  63                                             size_t capacity);
  64
  65 struct icu_buf_utf16 * icu_buf_utf16_copy(struct icu_buf_utf16 * dest16,
  66                                           struct icu_buf_utf16 * src16);
  67
  68 void icu_buf_utf16_destroy(struct icu_buf_utf16 * buf16);
  69
  70
  71
  72 struct icu_buf_utf8
  73 {
  74     uint8_t * utf8;
  75     int32_t utf8_len;
  76     int32_t utf8_cap;
  77 };
  78
  79 struct icu_buf_utf8 * icu_buf_utf8_create(size_t capacity);
  80
  81 struct icu_buf_utf8 * icu_buf_utf8_clear(struct icu_buf_utf8 * buf8);
  82
  83 struct icu_buf_utf8 * icu_buf_utf8_resize(struct icu_buf_utf8 * buf8,
  84                                           size_t capacity);
  85
  86 void icu_buf_utf8_destroy(struct icu_buf_utf8 * buf8);
  87
  88
  89 UErrorCode icu_utf16_from_utf8(struct icu_buf_utf16 * dest16,
  90                                struct icu_buf_utf8 * src8,
  91                                UErrorCode * status);
  92
  93 UErrorCode icu_utf16_from_utf8_cstr(struct icu_buf_utf16 * dest16,
  94                                     const char * src8cstr,
  95                                     UErrorCode * status);
  96
  97
  98 UErrorCode icu_utf16_to_utf8(struct icu_buf_utf8 * dest8,
  99                              struct icu_buf_utf16 * src16,
 100                              UErrorCode * status);
 101
 102 struct icu_casemap
 103 {
 104     char locale[16];
 105     char action;
 106 };
 107
 108 struct icu_casemap * icu_casemap_create(const char *locale, char action,
 109                                             UErrorCode *status);
 110
 111 void icu_casemap_destroy(struct icu_casemap * casemap);
 112
 113 int icu_casemap_casemap(struct icu_casemap * casemap,
 114                         struct icu_buf_utf16 * dest16,
 115                         struct icu_buf_utf16 * src16,
 116                         UErrorCode *status);
 117
 118 int icu_utf16_casemap(struct icu_buf_utf16 * dest16,
 119                       struct icu_buf_utf16 * src16,
 120                       const char *locale, char action,
 121                       UErrorCode *status);
 122
 123 UErrorCode icu_sortkey8_from_utf16(UCollator *coll,
 124                                    struct icu_buf_utf8 * dest8,
 125                                    struct icu_buf_utf16 * src16,
 126                                    UErrorCode * status);
 127
 128 struct icu_tokenizer
 129 {
 130     char locale[16];
 131     char action;
 132     UBreakIterator* bi;
 133     struct icu_buf_utf16 * buf16;
 134     int32_t token_count;
 135     int32_t token_id;
 136     int32_t token_start;
 137     int32_t token_end;
 138 /*
 139   keep always invariant
 140   0 <= token_start
 141   <= token_end
 142   <= buf16->utf16_len
 143   and invariant
 144   0 <= token_id <= token_count
 145 */
 146 };
 147
 148 struct icu_tokenizer * icu_tokenizer_create(const char *locale, char action,
 149                                             UErrorCode *status);
 150
 151 void icu_tokenizer_destroy(struct icu_tokenizer * tokenizer);
 152
 153 int icu_tokenizer_attach(struct icu_tokenizer * tokenizer,
 154                          struct icu_buf_utf16 * src16, UErrorCode *status);
 155
 156 int32_t icu_tokenizer_next_token(struct icu_tokenizer * tokenizer,
 157                                  struct icu_buf_utf16 * tkn16,
 158                                  UErrorCode *status);
 159
 160 int32_t icu_tokenizer_token_id(struct icu_tokenizer * tokenizer);
 161 int32_t icu_tokenizer_token_start(struct icu_tokenizer * tokenizer);
 162 int32_t icu_tokenizer_token_end(struct icu_tokenizer * tokenizer);
 163 int32_t icu_tokenizer_token_length(struct icu_tokenizer * tokenizer);
 164 int32_t icu_tokenizer_token_count(struct icu_tokenizer * tokenizer);
 165
 166
 167
 168 struct icu_normalizer
 169 {
 170     char action;
 171     struct icu_buf_utf16 * rules16;
 172     UParseError parse_error[256];
 173     UTransliterator * trans;
 174 };
 175
 176 struct icu_normalizer * icu_normalizer_create(const char *rules, char action,
 177                                               UErrorCode *status);
 178
 179
 180 void icu_normalizer_destroy(struct icu_normalizer * normalizer);
 181
 182 int icu_normalizer_normalize(struct icu_normalizer * normalizer,
 183                              struct icu_buf_utf16 * dest16,
 184                              struct icu_buf_utf16 * src16,
 185                              UErrorCode *status);
 186
 187 enum icu_chain_step_type {
 188     ICU_chain_step_type_none,
 189     ICU_chain_step_type_display,   /* convert to utf8 display format */
 190 #ifdef ICU_CHAIN_SORTKEY
 191     ICU_chain_step_type_index,     /* convert to utf8 index format  */
 192     ICU_chain_step_type_sortkey,   /* convert to utf8 sortkey format */
 193 #endif
 194     ICU_chain_step_type_casemap,   /* apply utf16 charmap */
 195     ICU_chain_step_type_normalize, /* apply utf16 normalization */
 196     ICU_chain_step_type_tokenize   /* apply utf16 tokenization */
 197 };
 198
 199
 200
 201 struct icu_chain_step
 202 {
 203     /* type and action object */
 204     enum icu_chain_step_type type;
 205     union {
 206         struct icu_casemap * casemap;
 207         struct icu_normalizer * normalizer;
 208         struct icu_tokenizer * tokenizer;
 209     } u;
 210     /* temprary post-action utf16 buffer */
 211     struct icu_buf_utf16 * buf16;
 212     struct icu_chain_step * previous;
 213     int more_tokens;
 214     int need_new_token;
 215 };
 216
 217
 218 struct icu_chain;
 219
 220 struct icu_chain_step * icu_chain_step_create(struct icu_chain * chain,
 221                                               enum icu_chain_step_type type,
 222                                               const uint8_t * rule,
 223                                               struct icu_buf_utf16 * buf16,
 224                                               UErrorCode *status);
 225
 226
 227 void icu_chain_step_destroy(struct icu_chain_step * step);
 228
 229
 230 struct icu_chain
 231 {
 232     uint8_t locale[16];
 233     int sort;
 234
 235     UCollator * coll;
 236
 237     /* number of tokens returned so far */
 238     int32_t token_count;
 239
 240     /* utf8 output buffers */
 241     struct icu_buf_utf8 * display8;
 242     struct icu_buf_utf8 * norm8;
 243     struct icu_buf_utf8 * sort8;
 244
 245     /* utf16 source buffer */
 246     struct icu_buf_utf16 * src16;
 247
 248     /* linked list of chain steps */
 249     struct icu_chain_step * steps;
 250 };
 251
 252 struct icu_chain * icu_chain_create(const uint8_t * locale,
 253                                     int sort,
 254                                     UErrorCode * status);
 255
 256 void icu_chain_destroy(struct icu_chain * chain);
 257
 258 struct icu_chain * icu_chain_xml_config(xmlNode *xml_node,
 259                                         const uint8_t * locale,
 260                                         int sort,
 261                                         UErrorCode * status);
 262
 263 struct icu_chain_step * icu_chain_insert_step(struct icu_chain * chain,
 264                                               enum icu_chain_step_type type,
 265                                               const uint8_t * rule,
 266                                               UErrorCode *status);
 267
 268 int icu_chain_step_next_token(struct icu_chain * chain,
 269                               struct icu_chain_step * step,
 270                               UErrorCode *status);
 271
 272 int icu_chain_assign_cstr(struct icu_chain * chain,
 273                           const char * src8cstr,
 274                           UErrorCode *status);
 275
 276 int icu_chain_next_token(struct icu_chain * chain,
 277                          UErrorCode *status);
 278
 279 int icu_chain_get_token_count(struct icu_chain * chain);
 280
 281 const char * icu_chain_get_display(struct icu_chain * chain);
 282
 283 const char * icu_chain_get_norm(struct icu_chain * chain);
 284
 285 const char * icu_chain_get_sort(struct icu_chain * chain);
 286
 287 const UCollator * icu_chain_get_coll(struct icu_chain * chain);
 288
 289 #endif /* ICU_I18NL_H */
 290
 291 /*
 292  * Local variables:
 293  * c-basic-offset: 4
 294  * indent-tabs-mode: nil
 295  * End:
 296  * vim: shiftwidth=4 tabstop=8 expandtab
 297  */