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 /* declared structs and functions */
  45
  46 int icu_check_status (UErrorCode status);
  47
  48 struct icu_buf_utf16
  49 {
  50   UChar * utf16;
  51   int32_t utf16_len;
  52   int32_t utf16_cap;
  53 };
  54
  55 struct icu_buf_utf16 * icu_buf_utf16_create(size_t capacity);
  56
  57 struct icu_buf_utf16 * icu_buf_utf16_clear(struct icu_buf_utf16 * buf16);
  58
  59 struct icu_buf_utf16 * icu_buf_utf16_resize(struct icu_buf_utf16 * buf16,
  60                                             size_t capacity);
  61
  62 struct icu_buf_utf16 * icu_buf_utf16_copy(struct icu_buf_utf16 * dest16,
  63                                           struct icu_buf_utf16 * src16);
  64
  65 void icu_buf_utf16_destroy(struct icu_buf_utf16 * buf16);
  66
  67
  68
  69 struct icu_buf_utf8
  70 {
  71     uint8_t * utf8;
  72     int32_t utf8_len;
  73     int32_t utf8_cap;
  74 };
  75
  76 struct icu_buf_utf8 * icu_buf_utf8_create(size_t capacity);
  77
  78 struct icu_buf_utf8 * icu_buf_utf8_clear(struct icu_buf_utf8 * buf8);
  79
  80 struct icu_buf_utf8 * icu_buf_utf8_resize(struct icu_buf_utf8 * buf8,
  81                                           size_t capacity);
  82
  83 void icu_buf_utf8_destroy(struct icu_buf_utf8 * buf8);
  84
  85
  86 UErrorCode icu_utf16_from_utf8(struct icu_buf_utf16 * dest16,
  87                                struct icu_buf_utf8 * src8,
  88                                UErrorCode * status);
  89
  90 UErrorCode icu_utf16_from_utf8_cstr(struct icu_buf_utf16 * dest16,
  91                                     const char * src8cstr,
  92                                     UErrorCode * status);
  93
  94
  95 UErrorCode icu_utf16_to_utf8(struct icu_buf_utf8 * dest8,
  96                              struct icu_buf_utf16 * src16,
  97                              UErrorCode * status);
  98
  99 struct icu_casemap
 100 {
 101     char locale[16];
 102     char action;
 103 };
 104
 105 struct icu_casemap * icu_casemap_create(const char *locale, char action,
 106                                             UErrorCode *status);
 107
 108 void icu_casemap_destroy(struct icu_casemap * casemap);
 109
 110 int icu_casemap_casemap(struct icu_casemap * casemap,
 111                         struct icu_buf_utf16 * dest16,
 112                         struct icu_buf_utf16 * src16,
 113                         UErrorCode *status);
 114
 115 int icu_utf16_casemap(struct icu_buf_utf16 * dest16,
 116                       struct icu_buf_utf16 * src16,
 117                       const char *locale, char action,
 118                       UErrorCode *status);
 119
 120 UErrorCode icu_sortkey8_from_utf16(UCollator *coll,
 121                                    struct icu_buf_utf8 * dest8,
 122                                    struct icu_buf_utf16 * src16,
 123                                    UErrorCode * status);
 124
 125 struct icu_tokenizer
 126 {
 127     char locale[16];
 128     char action;
 129     UBreakIterator* bi;
 130     struct icu_buf_utf16 * buf16;
 131     int32_t token_count;
 132     int32_t token_id;
 133     int32_t token_start;
 134     int32_t token_end;
 135 /*
 136   keep always invariant
 137   0 <= token_start
 138   <= token_end
 139   <= buf16->utf16_len
 140   and invariant
 141   0 <= token_id <= token_count
 142 */
 143 };
 144
 145 struct icu_tokenizer * icu_tokenizer_create(const char *locale, char action,
 146                                             UErrorCode *status);
 147
 148 void icu_tokenizer_destroy(struct icu_tokenizer * tokenizer);
 149
 150 int icu_tokenizer_attach(struct icu_tokenizer * tokenizer,
 151                          struct icu_buf_utf16 * src16, UErrorCode *status);
 152
 153 int32_t icu_tokenizer_next_token(struct icu_tokenizer * tokenizer,
 154                                  struct icu_buf_utf16 * tkn16,
 155                                  UErrorCode *status);
 156
 157 int32_t icu_tokenizer_token_id(struct icu_tokenizer * tokenizer);
 158 int32_t icu_tokenizer_token_start(struct icu_tokenizer * tokenizer);
 159 int32_t icu_tokenizer_token_end(struct icu_tokenizer * tokenizer);
 160 int32_t icu_tokenizer_token_length(struct icu_tokenizer * tokenizer);
 161 int32_t icu_tokenizer_token_count(struct icu_tokenizer * tokenizer);
 162
 163
 164
 165 struct icu_normalizer
 166 {
 167     char action;
 168     struct icu_buf_utf16 * rules16;
 169     UParseError parse_error[256];
 170     UTransliterator * trans;
 171 };
 172
 173 struct icu_normalizer * icu_normalizer_create(const char *rules, char action,
 174                                               UErrorCode *status);
 175
 176
 177 void icu_normalizer_destroy(struct icu_normalizer * normalizer);
 178
 179 int icu_normalizer_normalize(struct icu_normalizer * normalizer,
 180                              struct icu_buf_utf16 * dest16,
 181                              struct icu_buf_utf16 * src16,
 182                              UErrorCode *status);
 183
 184 enum icu_chain_step_type {
 185     ICU_chain_step_type_none,
 186     ICU_chain_step_type_display,   /* convert to utf8 display format */
 187     ICU_chain_step_type_casemap,   /* apply utf16 charmap */
 188     ICU_chain_step_type_normalize, /* apply utf16 normalization */
 189     ICU_chain_step_type_tokenize   /* apply utf16 tokenization */
 190 };
 191
 192
 193
 194 struct icu_chain_step
 195 {
 196     /* type and action object */
 197     enum icu_chain_step_type type;
 198     union {
 199         struct icu_casemap * casemap;
 200         struct icu_normalizer * normalizer;
 201         struct icu_tokenizer * tokenizer;
 202     } u;
 203     /* temprary post-action utf16 buffer */
 204     struct icu_buf_utf16 * buf16;
 205     struct icu_chain_step * previous;
 206     int more_tokens;
 207     int need_new_token;
 208 };
 209
 210
 211 struct icu_chain;
 212
 213 struct icu_chain_step * icu_chain_step_create(struct icu_chain * chain,
 214                                               enum icu_chain_step_type type,
 215                                               const uint8_t * rule,
 216                                               struct icu_buf_utf16 * buf16,
 217                                               UErrorCode *status);
 218
 219
 220 void icu_chain_step_destroy(struct icu_chain_step * step);
 221
 222
 223 struct icu_chain
 224 {
 225     uint8_t locale[16];
 226     int sort;
 227
 228     UCollator * coll;
 229
 230     /* number of tokens returned so far */
 231     int32_t token_count;
 232
 233     /* utf8 output buffers */
 234     struct icu_buf_utf8 * display8;
 235     struct icu_buf_utf8 * norm8;
 236     struct icu_buf_utf8 * sort8;
 237
 238     /* utf16 source buffer */
 239     struct icu_buf_utf16 * src16;
 240
 241     /* linked list of chain steps */
 242     struct icu_chain_step * steps;
 243 };
 244
 245 struct icu_chain * icu_chain_create(const uint8_t * locale,
 246                                     int sort,
 247                                     UErrorCode * status);
 248
 249 void icu_chain_destroy(struct icu_chain * chain);
 250
 251 struct icu_chain * icu_chain_xml_config(xmlNode *xml_node,
 252                                         const uint8_t * locale,
 253                                         int sort,
 254                                         UErrorCode * status);
 255
 256 struct icu_chain_step * icu_chain_insert_step(struct icu_chain * chain,
 257                                               enum icu_chain_step_type type,
 258                                               const uint8_t * rule,
 259                                               UErrorCode *status);
 260
 261 int icu_chain_step_next_token(struct icu_chain * chain,
 262                               struct icu_chain_step * step,
 263                               UErrorCode *status);
 264
 265 int icu_chain_assign_cstr(struct icu_chain * chain,
 266                           const char * src8cstr,
 267                           UErrorCode *status);
 268
 269 int icu_chain_next_token(struct icu_chain * chain,
 270                          UErrorCode *status);
 271
 272 int icu_chain_get_token_count(struct icu_chain * chain);
 273
 274 const char * icu_chain_get_display(struct icu_chain * chain);
 275
 276 const char * icu_chain_get_norm(struct icu_chain * chain);
 277
 278 const char * icu_chain_get_sort(struct icu_chain * chain);
 279
 280 const UCollator * icu_chain_get_coll(struct icu_chain * chain);
 281
 282 #endif /* ICU_I18NL_H */
 283
 284 /*
 285  * Local variables:
 286  * c-basic-offset: 4
 287  * indent-tabs-mode: nil
 288  * End:
 289  * vim: shiftwidth=4 tabstop=8 expandtab
 290  */