include/yaz/icu_I18N.h

   1 /* $Id: icu_I18N.h,v 1.1 2007-10-22 12:21:39 adam Exp $
   2    Copyright (c) 2006-2007, Index Data.
   3
   4    This file is part of Pazpar2.
   5
   6    Pazpar2 is free software; you can redistribute it and/or modify it under
   7    the terms of the GNU General Public License as published by the Free
   8    Software Foundation; either version 2, or (at your option) any later
   9    version.
  10
  11    Pazpar2 is distributed in the hope that it will be useful, but WITHOUT ANY
  12    WARRANTY; without even the implied warranty of MERCHANTABILITY or
  13    FITNESS FOR A PARTICULAR PURPOSE.  See the GNU General Public License
  14    for more details.
  15
  16    You should have received a copy of the GNU General Public License
  17    along with Pazpar2; see the file LICENSE.  If not, write to the
  18    Free Software Foundation, 59 Temple Place - Suite 330, Boston, MA
  19    02111-1307, USA.
  20 */
  21
  22 #ifndef ICU_I18NL_H
  23 #define ICU_I18NL_H
  24
  25 #include <yaz/nmem.h>
  26
  27 #include <libxml/parser.h>
  28 #include <libxml/tree.h>
  29
  30 #include <unicode/utypes.h>   /* Basic ICU data types */
  31 #include <unicode/uchar.h>    /* char names           */
  32
  33 //#include <unicode/ustdio.h>
  34 #include <unicode/ucol.h>
  35 //#include <unicode/ucnv.h>     /* C   Converter API    */
  36 //#include <unicode/ustring.h>  /* some more string fcns*/
  37 //#include <unicode/uloc.h>
  38 #include <unicode/ubrk.h>
  39 //#include <unicode/unistr.h>
  40 #include <unicode/utrans.h>
  41
  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 struct icu_buf_utf16 * icu_buf_utf16_resize(struct icu_buf_utf16 * buf16,
  57                                             size_t capacity);
  58 struct icu_buf_utf16 * icu_buf_utf16_copy(struct icu_buf_utf16 * dest16,
  59                                           struct icu_buf_utf16 * src16);
  60 void icu_buf_utf16_destroy(struct icu_buf_utf16 * buf16);
  61
  62
  63
  64 struct icu_buf_utf8
  65 {
  66   uint8_t * utf8;
  67   int32_t utf8_len;
  68   int32_t utf8_cap;
  69 };
  70
  71 struct icu_buf_utf8 * icu_buf_utf8_create(size_t capacity);
  72 struct icu_buf_utf8 * icu_buf_utf8_resize(struct icu_buf_utf8 * buf8,
  73                                           size_t capacity);
  74 void icu_buf_utf8_destroy(struct icu_buf_utf8 * buf8);
  75
  76
  77 UErrorCode icu_utf16_from_utf8(struct icu_buf_utf16 * dest16,
  78                                struct icu_buf_utf8 * src8,
  79                                UErrorCode * status);
  80
  81 UErrorCode icu_utf16_from_utf8_cstr(struct icu_buf_utf16 * dest16,
  82                                     const char * src8cstr,
  83                                     UErrorCode * status);
  84
  85
  86 UErrorCode icu_utf16_to_utf8(struct icu_buf_utf8 * dest8,
  87                              struct icu_buf_utf16 * src16,
  88                              UErrorCode * status);
  89
  90 struct icu_casemap
  91 {
  92   char locale[16];
  93   char action;
  94 };
  95
  96 struct icu_casemap * icu_casemap_create(const char *locale, char action,
  97                                             UErrorCode *status);
  98
  99 void icu_casemap_destroy(struct icu_casemap * casemap);
 100
 101 int icu_casemap_casemap(struct icu_casemap * casemap,
 102                         struct icu_buf_utf16 * dest16,
 103                         struct icu_buf_utf16 * src16,
 104                         UErrorCode *status);
 105
 106 int icu_utf16_casemap(struct icu_buf_utf16 * dest16,
 107                       struct icu_buf_utf16 * src16,
 108                       const char *locale, char action,
 109                       UErrorCode *status);
 110
 111 UErrorCode icu_sortkey8_from_utf16(UCollator *coll,
 112                                    struct icu_buf_utf8 * dest8,
 113                                    struct icu_buf_utf16 * src16,
 114                                    UErrorCode * status);
 115
 116 struct icu_tokenizer
 117 {
 118   char locale[16];
 119   char action;
 120   UBreakIterator* bi;
 121   struct icu_buf_utf16 * buf16;
 122   int32_t token_count;
 123   int32_t token_id;
 124   int32_t token_start;
 125   int32_t token_end;
 126   // keep always invariant
 127   // 0 <= token_start
 128   //   <= token_end
 129   //   <= buf16->utf16_len
 130   // and invariant
 131   // 0 <= token_id <= token_count
 132 };
 133
 134 struct icu_tokenizer * icu_tokenizer_create(const char *locale, char action,
 135                                             UErrorCode *status);
 136
 137 void icu_tokenizer_destroy(struct icu_tokenizer * tokenizer);
 138
 139 int icu_tokenizer_attach(struct icu_tokenizer * tokenizer,
 140                          struct icu_buf_utf16 * src16, UErrorCode *status);
 141
 142 int32_t icu_tokenizer_next_token(struct icu_tokenizer * tokenizer,
 143                                  struct icu_buf_utf16 * tkn16,
 144                                  UErrorCode *status);
 145
 146 int32_t icu_tokenizer_token_id(struct icu_tokenizer * tokenizer);
 147 int32_t icu_tokenizer_token_start(struct icu_tokenizer * tokenizer);
 148 int32_t icu_tokenizer_token_end(struct icu_tokenizer * tokenizer);
 149 int32_t icu_tokenizer_token_length(struct icu_tokenizer * tokenizer);
 150 int32_t icu_tokenizer_token_count(struct icu_tokenizer * tokenizer);
 151
 152
 153
 154 struct icu_normalizer
 155 {
 156   char action;
 157   struct icu_buf_utf16 * rules16;
 158   UParseError parse_error[256];
 159   UTransliterator * trans;
 160 };
 161
 162 struct icu_normalizer * icu_normalizer_create(const char *rules, char action,
 163                                               UErrorCode *status);
 164
 165
 166 void icu_normalizer_destroy(struct icu_normalizer * normalizer);
 167
 168 int icu_normalizer_normalize(struct icu_normalizer * normalizer,
 169                              struct icu_buf_utf16 * dest16,
 170                              struct icu_buf_utf16 * src16,
 171                              UErrorCode *status);
 172
 173
 174 #if 0
 175 struct icu_token
 176 {
 177   int32_t token_id;
 178   uint8_t * display8;
 179   uint8_t * norm8;
 180   uint8_t * sort8;
 181 }
 182 #endif
 183
 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                                         UErrorCode * status);
 253
 254
 255 struct icu_chain_step * icu_chain_insert_step(struct icu_chain * chain,
 256                                               enum icu_chain_step_type type,
 257                                               const uint8_t * rule,
 258                                               UErrorCode *status);
 259
 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
 281
 282
 283
 284 #endif // ICU_I18NL_H