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