3 * Library to deal with pinyin.
5 * Copyright (C) 2011 Peng Wu <alexepico@gmail.com>
7 * This program is free software; you can redistribute it and/or modify
8 * it under the terms of the GNU General Public License as published by
9 * the Free Software Foundation; either version 2 of the License, or
10 * (at your option) any later version.
12 * This program is distributed in the hope that it will be useful,
13 * but WITHOUT ANY WARRANTY; without even the implied warranty of
14 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
15 * GNU General Public License for more details.
17 * You should have received a copy of the GNU General Public License
18 * along with this program; if not, write to the Free Software
19 * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301, USA.
23 #include "pinyin_parser2.h"
29 #include "pinyin_phrase2.h"
30 #include "pinyin_custom2.h"
31 #include "chewing_key.h"
32 #include "pinyin_parser_table.h"
33 #include "double_pinyin_table.h"
34 #include "chewing_table.h"
37 using namespace pinyin;
39 static bool check_pinyin_options(pinyin_option_t options, const pinyin_index_item_t * item) {
40 guint32 flags = item->m_flags;
41 assert (flags & IS_PINYIN);
43 /* handle incomplete pinyin. */
44 if (flags & PINYIN_INCOMPLETE) {
45 if (!(options & PINYIN_INCOMPLETE))
49 /* handle correct pinyin, currently only one flag per item. */
50 flags &= PINYIN_CORRECT_ALL;
51 options &= PINYIN_CORRECT_ALL;
54 if ((flags & options) != flags)
61 static bool check_chewing_options(pinyin_option_t options, const chewing_index_item_t * item) {
62 guint32 flags = item->m_flags;
63 assert (flags & IS_CHEWING);
65 /* handle incomplete chewing. */
66 if (flags & CHEWING_INCOMPLETE) {
67 if (!(options & CHEWING_INCOMPLETE))
75 gint ChewingKey::get_table_index() {
76 assert(m_initial < CHEWING_NUMBER_OF_INITIALS);
77 assert(m_middle < CHEWING_NUMBER_OF_MIDDLES);
78 assert(m_final < CHEWING_NUMBER_OF_FINALS);
80 gint index = chewing_key_table[(m_initial * CHEWING_NUMBER_OF_MIDDLES + m_middle) * CHEWING_NUMBER_OF_FINALS + m_final];
81 return index == -1 ? 0 : index;
84 gchar * ChewingKey::get_pinyin_string() {
85 assert(m_tone < CHEWING_NUMBER_OF_TONES);
86 gint index = get_table_index();
87 assert(index < G_N_ELEMENTS(content_table));
88 const content_table_item_t & item = content_table[index];
90 if (CHEWING_ZERO_TONE == m_tone) {
91 return g_strdup(item.m_pinyin_str);
93 return g_strdup_printf("%s%d", item.m_pinyin_str, m_tone);
97 gchar * ChewingKey::get_chewing_string() {
98 assert(m_tone < CHEWING_NUMBER_OF_TONES);
99 gint index = get_table_index();
100 assert(index < G_N_ELEMENTS(content_table));
101 const content_table_item_t & item = content_table[index];
103 if (CHEWING_ZERO_TONE == m_tone) {
104 return g_strdup(item.m_chewing_str);
106 return g_strdup_printf("%s%s", item.m_chewing_str,
107 chewing_tone_table[m_tone]);
114 /* internal information for pinyin parsers. */
115 struct parse_value_t{
117 ChewingKeyRest m_key_rest;
131 const guint16 max_full_pinyin_length = 7; /* include tone. */
133 const guint16 max_double_pinyin_length = 3; /* include tone. */
135 const guint16 max_chewing_length = 4; /* include tone. */
137 static bool compare_pinyin_less_than(const pinyin_index_item_t & lhs,
138 const pinyin_index_item_t & rhs){
139 return 0 > strcmp(lhs.m_pinyin_input, rhs.m_pinyin_input);
142 static inline bool search_pinyin_index(pinyin_option_t options,
145 pinyin_index_item_t item;
146 memset(&item, 0, sizeof(item));
147 item.m_pinyin_input = pinyin;
149 std_lite::pair<const pinyin_index_item_t *,
150 const pinyin_index_item_t *> range;
151 range = std_lite::equal_range
152 (pinyin_index, pinyin_index + G_N_ELEMENTS(pinyin_index),
153 item, compare_pinyin_less_than);
155 guint16 range_len = range.second - range.first;
156 assert(range_len <= 1);
157 if (range_len == 1) {
158 const pinyin_index_item_t * index = range.first;
160 if (!check_pinyin_options(options, index))
163 key = content_table[index->m_table_index].m_chewing_key;
164 assert(key.get_table_index() == index->m_table_index);
171 static bool compare_chewing_less_than(const chewing_index_item_t & lhs,
172 const chewing_index_item_t & rhs){
173 return 0 > strcmp(lhs.m_chewing_input, rhs.m_chewing_input);
176 static inline bool search_chewing_index(pinyin_option_t options,
177 const char * chewing,
179 chewing_index_item_t item;
180 memset(&item, 0, sizeof(item));
181 item.m_chewing_input = chewing;
183 std_lite::pair<const chewing_index_item_t *,
184 const chewing_index_item_t *> range;
185 range = std_lite::equal_range
186 (chewing_index, chewing_index + G_N_ELEMENTS(chewing_index),
187 item, compare_chewing_less_than);
189 guint16 range_len = range.second - range.first;
190 assert (range_len <= 1);
192 if (range_len == 1) {
193 const chewing_index_item_t * index = range.first;
195 if (!check_chewing_options(options, index))
198 key = content_table[index->m_table_index].m_chewing_key;
199 assert(key.get_table_index() == index->m_table_index);
206 /* Full Pinyin Parser */
207 FullPinyinParser2::FullPinyinParser2 (){
208 m_parse_steps = g_array_new(TRUE, FALSE, sizeof(parse_value_t));
212 bool FullPinyinParser2::parse_one_key (pinyin_option_t options,
214 const char * pinyin, int len) const {
215 /* "'" are not accepted in parse_one_key. */
216 gchar * input = g_strndup(pinyin, len);
217 assert(NULL == strchr(input, '\''));
219 guint16 tone = CHEWING_ZERO_TONE; guint16 tone_pos = 0;
220 guint16 parsed_len = len;
223 if (options & USE_TONE) {
224 /* find the tone in the last character. */
225 char chr = input[parsed_len - 1];
226 if ( '0' < chr && chr <= '5' ) {
229 tone_pos = parsed_len;
233 /* parse pinyin core staff here. */
235 /* Note: optimize here? */
236 input[parsed_len] = '\0';
237 if (!search_pinyin_index(options, input, key)) {
242 if (options & USE_TONE) {
243 /* post processing tone. */
244 if ( parsed_len == tone_pos ) {
245 if (tone != CHEWING_ZERO_TONE) {
253 return parsed_len == len;
257 int FullPinyinParser2::parse (pinyin_option_t options, ChewingKeyVector & keys,
258 ChewingKeyRestVector & key_rests,
259 const char *str, int len) const {
262 g_array_set_size(keys, 0);
263 g_array_set_size(key_rests, 0);
265 /* init m_parse_steps, and prepare dynamic programming. */
266 int step_len = len + 1;
267 g_array_set_size(m_parse_steps, 0);
269 for (i = 0; i < step_len; ++i) {
270 g_array_append_val(m_parse_steps, value);
274 gchar * input = g_strndup(str, len);
275 parse_value_t * curstep = NULL, * nextstep = NULL;
277 for (i = 0; i < len; ++i) {
278 if (input[i] == '\'') {
279 curstep = &g_array_index(m_parse_steps, parse_value_t, i);
280 nextstep = &g_array_index(m_parse_steps, parse_value_t, i + 1);
282 /* propagate current step into next step. */
283 nextstep->m_key = ChewingKey();
284 nextstep->m_key_rest = ChewingKeyRest();
285 nextstep->m_num_keys = curstep->m_num_keys;
286 nextstep->m_parsed_len = curstep->m_parsed_len + 1;
287 nextstep->m_last_step = i;
292 /* forward to next "'" */
293 if ( 0 == next_sep ) {
295 for (k = i; k < len; ++k) {
296 if (input[k] == '\'')
304 * do maximum forward match first. */
305 for (size_t pos = i; pos < next_sep; ++pos) {
306 curstep = &g_array_index(m_parse_steps, parse_value_t, pos);
307 size_t try_len = std_lite::min
308 (pos + max_full_pinyin_length, next_sep);
309 for (size_t n = try_len; n > pos; --n) {
310 nextstep = &g_array_index(m_parse_steps, parse_value_t, n);
313 const char * onepinyin = input + pos;
314 gint16 onepinyinlen = n - pos;
315 value = parse_value_t();
317 ChewingKey key; ChewingKeyRest rest;
318 bool parsed = parse_one_key
319 (options, key, onepinyin, onepinyinlen);
320 rest.m_raw_begin = pos; rest.m_raw_end = n;
325 //printf("onepinyin:%s len:%d\n", onepinyin, onepinyinlen);
326 value.m_key = key; value.m_key_rest = rest;
327 value.m_num_keys = curstep->m_num_keys + 1;
328 value.m_parsed_len = curstep->m_parsed_len + onepinyinlen;
329 value.m_last_step = pos;
332 if (-1 == nextstep->m_last_step)
334 if (value.m_parsed_len > nextstep->m_parsed_len)
336 if (value.m_parsed_len == nextstep->m_parsed_len &&
337 value.m_num_keys < nextstep->m_num_keys)
340 /* maximum forward, set pos to n in next iteration. */
347 /* dynamic programming here. */
348 /* for (size_t m = i; m < next_sep; ++m) */
351 curstep = &g_array_index(m_parse_steps, parse_value_t, m);
352 size_t try_len = std_lite::min
353 (m + max_full_pinyin_length, next_sep);
354 for (size_t n = m + 1; n < try_len + 1; ++n) {
355 nextstep = &g_array_index(m_parse_steps, parse_value_t, n);
358 const char * onepinyin = input + m;
359 gint16 onepinyinlen = n - m;
360 value = parse_value_t();
362 ChewingKey key; ChewingKeyRest rest;
363 bool parsed = parse_one_key
364 (options, key, onepinyin, onepinyinlen);
365 rest.m_raw_begin = m; rest.m_raw_end = n;
369 //printf("onepinyin:%s len:%d\n", onepinyin, onepinyinlen);
371 value.m_key = key; value.m_key_rest = rest;
372 value.m_num_keys = curstep->m_num_keys + 1;
373 value.m_parsed_len = curstep->m_parsed_len + onepinyinlen;
374 value.m_last_step = m;
377 /* no previous result */
378 if (-1 == nextstep->m_last_step)
380 /* prefer the longest pinyin */
381 if (value.m_parsed_len > nextstep->m_parsed_len)
383 /* prefer the shortest keys with the same pinyin length */
384 if (value.m_parsed_len == nextstep->m_parsed_len &&
385 value.m_num_keys < nextstep->m_num_keys)
388 /* handle with the same pinyin length and the number of keys */
389 if (value.m_parsed_len == nextstep->m_parsed_len &&
390 value.m_num_keys == nextstep->m_num_keys) {
393 /* prefer the complete pinyin with shengmu
394 * over without shengmu,
395 * ex: "kaneiji" -> "ka'nei'ji".
397 if ((value.m_key.m_initial != CHEWING_ZERO_INITIAL &&
398 !(value.m_key.m_middle == CHEWING_ZERO_MIDDLE &&
399 value.m_key.m_final == CHEWING_ZERO_FINAL)) &&
400 nextstep->m_key.m_initial == CHEWING_ZERO_INITIAL)
403 /* prefer the complete pinyin 'er'
404 * over the in-complete pinyin 'r',
405 * ex: "xierqi" -> "xi'er'qi."
407 if ((value.m_key.m_initial == CHEWING_ZERO_INITIAL &&
408 value.m_key.m_middle == CHEWING_ZERO_MIDDLE &&
409 value.m_key.m_final == CHEWING_ER) &&
410 (nextstep->m_key.m_initial == CHEWING_R &&
411 nextstep->m_key.m_middle == CHEWING_ZERO_MIDDLE &&
412 nextstep->m_key.m_final == CHEWING_ZERO_FINAL))
416 /* prefer the 'a' at the end of clause,
417 * ex: "zheyanga$" -> "zhe'yang'a$".
419 if (value.m_parsed_len == len &&
420 (nextstep->m_key.m_initial != CHEWING_ZERO_INITIAL &&
421 nextstep->m_key.m_final == CHEWING_A) &&
422 (value.m_key.m_initial == CHEWING_ZERO_INITIAL &&
423 value.m_key.m_middle == CHEWING_ZERO_MIDDLE &&
424 value.m_key.m_final == CHEWING_A))
431 /* final step for back tracing. */
432 gint16 parsed_len = final_step(step_len, keys, key_rests);
434 /* post processing for re-split table. */
435 if (options & USE_RESPLIT_TABLE) {
436 post_process(options, keys, key_rests);
443 int FullPinyinParser2::final_step(size_t step_len, ChewingKeyVector & keys,
444 ChewingKeyRestVector & key_rests) const{
446 gint16 parsed_len = 0;
447 parse_value_t * curstep = NULL;
449 /* find longest match, which starts from the beginning of input. */
450 for (i = step_len - 1; i >= 0; --i) {
451 curstep = &g_array_index(m_parse_steps, parse_value_t, i);
452 if (i == curstep->m_parsed_len)
455 /* prepare saving. */
456 parsed_len = curstep->m_parsed_len;
457 gint16 num_keys = curstep->m_num_keys;
458 g_array_set_size(keys, num_keys);
459 g_array_set_size(key_rests, num_keys);
461 /* save the match. */
462 while (curstep->m_last_step != -1) {
463 gint16 pos = curstep->m_num_keys - 1;
466 if (0 != curstep->m_key.get_table_index()) {
467 ChewingKey * key = &g_array_index(keys, ChewingKey, pos);
468 ChewingKeyRest * rest = &g_array_index
469 (key_rests, ChewingKeyRest, pos);
470 *key = curstep->m_key; *rest = curstep->m_key_rest;
474 curstep = &g_array_index(m_parse_steps, parse_value_t,
475 curstep->m_last_step);
481 bool FullPinyinParser2::post_process(pinyin_option_t options,
482 ChewingKeyVector & keys,
483 ChewingKeyRestVector & key_rests) const {
485 assert(keys->len == key_rests->len);
486 gint16 num_keys = keys->len;
488 ChewingKey * cur_key = NULL, * next_key = NULL;
489 ChewingKeyRest * cur_rest = NULL, * next_rest = NULL;
490 guint16 next_tone = CHEWING_ZERO_TONE;
492 for (i = 0; i < num_keys - 1; ++i) {
493 cur_rest = &g_array_index(key_rests, ChewingKeyRest, i);
494 next_rest = &g_array_index(key_rests, ChewingKeyRest, i + 1);
497 if (cur_rest->m_raw_end != next_rest->m_raw_begin)
500 cur_key = &g_array_index(keys, ChewingKey, i);
501 next_key = &g_array_index(keys, ChewingKey, i + 1);
504 if (CHEWING_ZERO_TONE != cur_key->m_tone)
507 if (options & USE_TONE) {
508 next_tone = next_key->m_tone;
509 next_key->m_tone = CHEWING_ZERO_TONE;
512 /* lookup re-split table */
514 const resplit_table_item_t * item = NULL;
515 for (k = 0; k < G_N_ELEMENTS(resplit_table); ++k) {
516 item = resplit_table + k;
518 if (item->m_orig_freq >= item->m_new_freq)
521 /* use pinyin_exact_compare2 here. */
522 if (0 == pinyin_exact_compare2(item->m_orig_keys,
529 if (k < G_N_ELEMENTS(resplit_table)) {
531 item = resplit_table + k;
532 *cur_key = item->m_new_keys[0];
533 *next_key = item->m_new_keys[1];
534 /* assumes only moved one char in gen_all_resplit script. */
535 cur_rest->m_raw_end ++;
536 next_rest->m_raw_begin ++;
539 /* save back tones */
540 if (options & USE_TONE) {
541 next_key->m_tone = next_tone;
548 #define IS_KEY(x) (('a' <= x && x <= 'z') || x == ';')
550 bool DoublePinyinParser2::parse_one_key(pinyin_option_t options,
552 const char *str, int len) const {
553 options &= ~(PINYIN_CORRECT_ALL|PINYIN_AMB_ALL);
556 if (!(options & PINYIN_INCOMPLETE))
563 int charid = ch == ';' ? 26 : ch - 'a';
564 const char * sheng = m_shengmu_table[charid].m_shengmu;
565 if (NULL == sheng || strcmp(sheng, "'") == 0)
568 if (search_pinyin_index(options, sheng, key)) {
575 ChewingTone tone = CHEWING_ZERO_TONE;
576 options &= ~(PINYIN_INCOMPLETE|CHEWING_INCOMPLETE);
580 if (!(options & USE_TONE))
583 if (!('0' < ch && ch <= '5'))
585 tone = (ChewingTone) (ch - '0');
588 if (2 == len || 3 == len) {
589 /* parse shengmu here. */
594 int charid = ch == ';' ? 26 : ch - 'a';
595 const char * sheng = m_shengmu_table[charid].m_shengmu;
598 if (strcmp(sheng, "'") == 0)
601 /* parse yunmu here. */
606 charid = ch == ';' ? 26 : ch - 'a';
608 const char * yun = m_yunmu_table[charid].m_yunmus[0];
612 gchar * pinyin = g_strdup_printf("%s%s", sheng, yun);
613 if (search_pinyin_index(options, pinyin, key)) {
621 yun = m_yunmu_table[charid].m_yunmus[1];
625 pinyin = g_strdup_printf("%s%s", sheng, yun);
626 if (search_pinyin_index(options, pinyin, key)) {
639 /* only 'a'-'z' and ';' are accepted here. */
640 int DoublePinyinParser2::parse(pinyin_option_t options, ChewingKeyVector & keys,
641 ChewingKeyRestVector & key_rests,
642 const char *str, int len) const {
643 g_array_set_size(keys, 0);
644 g_array_set_size(key_rests, 0);
646 int maximum_len = 0; int i;
647 /* probe the longest possible double pinyin string. */
648 for (i = 0; i < len; ++i) {
649 const char ch = str[i];
650 if (!(IS_KEY(ch) || ('0' < ch && ch <= '5')))
655 /* maximum forward match for double pinyin. */
657 while (parsed_len < maximum_len) {
658 const char * cur_str = str + parsed_len;
659 i = std_lite::min(maximum_len - parsed_len,
660 (int)max_double_pinyin_length);
662 ChewingKey key; ChewingKeyRest key_rest;
664 bool success = parse_one_key(options, key, cur_str, i);
669 if (0 == i) /* no more possible double pinyins. */
672 key_rest.m_raw_begin = parsed_len; key_rest.m_raw_end = parsed_len + i;
675 /* save the pinyin */
676 g_array_append_val(keys, key);
677 g_array_append_val(key_rests, key_rest);
685 bool DoublePinyinParser2::set_scheme(DoublePinyinScheme scheme) {
688 case DOUBLE_PINYIN_ZRM:
689 m_shengmu_table = double_pinyin_zrm_sheng;
690 m_yunmu_table = double_pinyin_zrm_yun;
692 case DOUBLE_PINYIN_MS:
693 m_shengmu_table = double_pinyin_mspy_sheng;
694 m_yunmu_table = double_pinyin_mspy_yun;
696 case DOUBLE_PINYIN_ZIGUANG:
697 m_shengmu_table = double_pinyin_zgpy_sheng;
698 m_yunmu_table = double_pinyin_zgpy_yun;
700 case DOUBLE_PINYIN_ABC:
701 m_shengmu_table = double_pinyin_abc_sheng;
702 m_yunmu_table = double_pinyin_abc_yun;
704 case DOUBLE_PINYIN_PYJJ:
705 m_shengmu_table = double_pinyin_pyjj_sheng;
706 m_yunmu_table = double_pinyin_pyjj_yun;
708 case DOUBLE_PINYIN_XHE:
709 m_shengmu_table = double_pinyin_xhe_sheng;
710 m_yunmu_table = double_pinyin_xhe_yun;
712 case DOUBLE_PINYIN_CUSTOMIZED:
716 return false; /* no such scheme. */
719 /* the chewing string must be freed with g_free. */
720 static bool search_chewing_symbols(const chewing_symbol_item_t * symbol_table,
721 const char key, const char ** chewing) {
723 /* just iterate the table, as we only have < 50 items. */
724 while (symbol_table->m_input != '\0') {
725 if (symbol_table->m_input == key) {
726 *chewing = symbol_table->m_chewing;
734 static bool search_chewing_tones(const chewing_tone_item_t * tone_table,
735 const char key, char * tone) {
736 *tone = CHEWING_ZERO_TONE;
737 /* just iterate the table, as we only have < 10 items. */
738 while (tone_table->m_input != '\0') {
739 if (tone_table->m_input == key) {
740 *tone = tone_table->m_tone;
749 bool ChewingParser2::parse_one_key(pinyin_option_t options,
751 const char *str, int len) const {
752 options &= ~(PINYIN_CORRECT_ALL|PINYIN_AMB_ALL);
753 char tone = CHEWING_ZERO_TONE;
755 int symbols_len = len;
756 /* probe whether the last key is tone key in str. */
757 if (options & USE_TONE) {
758 char ch = str[len - 1];
759 /* remove tone from input */
760 if (search_chewing_tones(m_tone_table, ch, &tone))
765 gchar * chewing = NULL; const char * onechar = NULL;
767 /* probe the possible chewing map in the rest of str. */
768 for (i = 0; i < symbols_len; ++i) {
769 if (!search_chewing_symbols(m_symbol_table, str[i], &onechar)) {
775 chewing = g_strdup(onechar);
777 gchar * tmp = chewing;
778 chewing = g_strconcat(chewing, onechar, NULL);
783 /* search the chewing in the chewing index table. */
784 if (chewing && search_chewing_index(options, chewing, key)) {
785 /* save back tone if available. */
796 /* only characters in chewing keyboard scheme are accepted here. */
797 int ChewingParser2::parse(pinyin_option_t options, ChewingKeyVector & keys,
798 ChewingKeyRestVector & key_rests,
799 const char *str, int len) const {
800 g_array_set_size(keys, 0);
801 g_array_set_size(key_rests, 0);
803 int maximum_len = 0; int i;
804 /* probe the longest possible chewing string. */
805 for (i = 0; i < len; ++i) {
806 if (!in_chewing_scheme(options, str[i], NULL))
811 /* maximum forward match for chewing. */
813 while (parsed_len < maximum_len) {
814 const char * cur_str = str + parsed_len;
815 i = std_lite::min(maximum_len - parsed_len,
816 (int)max_chewing_length);
818 ChewingKey key; ChewingKeyRest key_rest;
820 bool success = parse_one_key(options, key, cur_str, i);
825 if (0 == i) /* no more possible chewings. */
828 key_rest.m_raw_begin = parsed_len; key_rest.m_raw_end = parsed_len + i;
831 /* save the pinyin. */
832 g_array_append_val(keys, key);
833 g_array_append_val(key_rests, key_rest);
840 bool ChewingParser2::set_scheme(ChewingScheme scheme) {
842 case CHEWING_STANDARD:
843 m_symbol_table = chewing_standard_symbols;
844 m_tone_table = chewing_standard_tones;
847 m_symbol_table = chewing_ibm_symbols;
848 m_tone_table = chewing_ibm_tones;
850 case CHEWING_GINYIEH:
851 m_symbol_table = chewing_ginyieh_symbols;
852 m_tone_table = chewing_ginyieh_tones;
855 m_symbol_table = chewing_eten_symbols;
856 m_tone_table = chewing_eten_tones;
864 bool ChewingParser2::in_chewing_scheme(pinyin_option_t options,
865 const char key, const char ** symbol)
867 const gchar * chewing = NULL;
868 char tone = CHEWING_ZERO_TONE;
870 if (search_chewing_symbols(m_symbol_table, key, &chewing)) {
876 if (!(options & USE_TONE))
879 if (search_chewing_tones(m_tone_table, key, &tone)) {
881 *symbol = chewing_tone_table[tone];