include/yaz/icu_I18N.h

   1 /* This file is part of the YAZ toolkit.
   2  * Copyright (C) 1995-2009 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 /** \file
  29     \brief Internal header for ICU utilities
  30 */
  31
  32 #ifndef ICU_I18NL_H
  33 #define ICU_I18NL_H
  34
  35 #include <yaz/yconfig.h>
  36
  37 #include <unicode/utypes.h>   /* Basic ICU data types */
  38 #include <unicode/uchar.h>    /* char names           */
  39
  40 #include <unicode/ucol.h>
  41 #include <unicode/ubrk.h>
  42 #include <unicode/utrans.h>
  43
  44 #include <yaz/icu.h>
  45
  46 /* declared structs and functions */
  47
  48 int icu_check_status (UErrorCode status);
  49
  50 struct icu_buf_utf16
  51 {
  52     UChar * utf16;
  53     int32_t utf16_len;
  54     int32_t utf16_cap;
  55 };
  56
  57 struct icu_buf_utf16 * icu_buf_utf16_create(size_t capacity);
  58
  59 struct icu_buf_utf16 * icu_buf_utf16_clear(struct icu_buf_utf16 * buf16);
  60
  61 struct icu_buf_utf16 * icu_buf_utf16_resize(struct icu_buf_utf16 * buf16,
  62                                             size_t capacity);
  63
  64 struct icu_buf_utf16 * icu_buf_utf16_copy(struct icu_buf_utf16 * dest16,
  65                                           struct icu_buf_utf16 * src16);
  66
  67 void icu_buf_utf16_destroy(struct icu_buf_utf16 * buf16);
  68
  69
  70
  71 struct icu_buf_utf8
  72 {
  73     uint8_t * utf8;
  74     int32_t utf8_len;
  75     int32_t utf8_cap;
  76 };
  77
  78 struct icu_buf_utf8 * icu_buf_utf8_create(size_t capacity);
  79
  80 struct icu_buf_utf8 * icu_buf_utf8_clear(struct icu_buf_utf8 * buf8);
  81
  82 struct icu_buf_utf8 * icu_buf_utf8_resize(struct icu_buf_utf8 * buf8,
  83                                           size_t capacity);
  84
  85 void icu_buf_utf8_destroy(struct icu_buf_utf8 * buf8);
  86
  87
  88 UErrorCode icu_utf16_from_utf8(struct icu_buf_utf16 * dest16,
  89                                struct icu_buf_utf8 * src8,
  90                                UErrorCode * status);
  91
  92 UErrorCode icu_utf16_from_utf8_cstr(struct icu_buf_utf16 * dest16,
  93                                     const char * src8cstr,
  94                                     UErrorCode * status);
  95
  96
  97 UErrorCode icu_utf16_to_utf8(struct icu_buf_utf8 * dest8,
  98                              struct icu_buf_utf16 * src16,
  99                              UErrorCode * status);
 100
 101 struct icu_casemap
 102 {
 103     char action;
 104 };
 105
 106 struct icu_casemap * icu_casemap_create(char action, 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                         const char *locale);
 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 void 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 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_transform
 166 {
 167     char action;
 168     UParseError parse_error;
 169     UTransliterator * trans;
 170 };
 171
 172 struct icu_transform * icu_transform_create(const char *id, char action,
 173                                             const char *rules,
 174                                             UErrorCode *status);
 175
 176 void icu_transform_destroy(struct icu_transform * transform);
 177
 178 int icu_transform_trans(struct icu_transform * transform,
 179                         struct icu_buf_utf16 * dest16,
 180                         struct icu_buf_utf16 * src16,
 181                         UErrorCode *status);
 182
 183 enum icu_chain_step_type {
 184     ICU_chain_step_type_none,
 185     ICU_chain_step_type_display,   /* convert to utf8 display format */
 186     ICU_chain_step_type_casemap,   /* apply utf16 charmap */
 187     ICU_chain_step_type_transform, /* apply utf16 transform */
 188     ICU_chain_step_type_tokenize,  /* apply utf16 tokenization */
 189     ICU_chain_step_type_transliterate  /* 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_transform * transform;
 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(yaz_icu_chain_t 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     char *locale;
 226     int sort;
 227
 228     const char * src8cstr;
 229
 230     UCollator * coll;
 231
 232     /* number of tokens returned so far */
 233     int32_t token_count;
 234
 235     /* utf8 output buffers */
 236     struct icu_buf_utf8 * display8;
 237     struct icu_buf_utf8 * norm8;
 238     struct icu_buf_utf8 * sort8;
 239
 240     /* utf16 source buffer */
 241     struct icu_buf_utf16 * src16;
 242
 243     /* linked list of chain steps */
 244     struct icu_chain_step * steps;
 245 };
 246
 247 struct icu_chain_step * icu_chain_insert_step(yaz_icu_chain_t chain,
 248                                               enum icu_chain_step_type type,
 249                                               const uint8_t * rule,
 250                                               UErrorCode *status);
 251
 252 int icu_chain_step_next_token(yaz_icu_chain_t chain,
 253                               struct icu_chain_step * step,
 254                               UErrorCode *status);
 255
 256 int icu_chain_token_number(yaz_icu_chain_t chain);
 257
 258 const UCollator * icu_chain_get_coll(yaz_icu_chain_t chain);
 259
 260 #endif /* ICU_I18NL_H */
 261
 262 /*
 263  * Local variables:
 264  * c-basic-offset: 4
 265  * c-file-style: "Stroustrup"
 266  * indent-tabs-mode: nil
 267  * End:
 268  * vim: shiftwidth=4 tabstop=8 expandtab
 269  */
 270