Commit | Line | Data |
---|---|---|
b7b65b15 | 1 | /* Low-level bidirectional buffer-scanning functions for GNU Emacs. |
b118e65f EZ |
2 | Copyright (C) 2000, 2001, 2004, 2005, 2009, 2010 |
3 | Free Software Foundation, Inc. | |
b7b65b15 EZ |
4 | |
5 | This file is part of GNU Emacs. | |
6 | ||
a8d11bd3 | 7 | GNU Emacs is free software: you can redistribute it and/or modify |
b7b65b15 | 8 | it under the terms of the GNU General Public License as published by |
a8d11bd3 EZ |
9 | the Free Software Foundation, either version 3 of the License, or |
10 | (at your option) any later version. | |
b7b65b15 EZ |
11 | |
12 | GNU Emacs 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. | |
16 | ||
b7b65b15 | 17 | You should have received a copy of the GNU General Public License |
a8d11bd3 | 18 | along with GNU Emacs. If not, see <http://www.gnu.org/licenses/>. */ |
b7b65b15 | 19 | |
2d6e4628 EZ |
20 | /* Written by Eli Zaretskii <eliz@gnu.org>. |
21 | ||
22 | A sequential implementation of the Unicode Bidirectional algorithm, | |
b7b65b15 EZ |
23 | as per UAX#9, a part of the Unicode Standard. |
24 | ||
25 | Unlike the reference and most other implementations, this one is | |
940afb59 EZ |
26 | designed to be called once for every character in the buffer or |
27 | string. | |
b7b65b15 | 28 | |
4b292a22 | 29 | The main entry point is bidi_move_to_visually_next. Each time it |
b7b65b15 EZ |
30 | is called, it finds the next character in the visual order, and |
31 | returns its information in a special structure. The caller is then | |
32 | expected to process this character for display or any other | |
4b292a22 EZ |
33 | purposes, and call bidi_move_to_visually_next for the next |
34 | character. See the comments in bidi_move_to_visually_next for more | |
35 | details about its algorithm that finds the next visual-order | |
b7b65b15 EZ |
36 | character by resolving their levels on the fly. |
37 | ||
940afb59 EZ |
38 | The two other entry points are bidi_paragraph_init and |
39 | bidi_mirror_char. The first determines the base direction of a | |
40 | paragraph, while the second returns the mirrored version of its | |
41 | argument character. | |
42 | ||
89d3374a EZ |
43 | If you want to understand the code, you will have to read it |
44 | together with the relevant portions of UAX#9. The comments include | |
45 | references to UAX#9 rules, for that very reason. | |
46 | ||
b7b65b15 EZ |
47 | A note about references to UAX#9 rules: if the reference says |
48 | something like "X9/Retaining", it means that you need to refer to | |
49 | rule X9 and to its modifications decribed in the "Implementation | |
50 | Notes" section of UAX#9, under "Retaining Format Codes". */ | |
51 | ||
b7b65b15 | 52 | #include <config.h> |
b7b65b15 | 53 | #include <stdio.h> |
b7b65b15 | 54 | #include <string.h> |
29e3d8d1 EZ |
55 | #include <setjmp.h> |
56 | ||
b7b65b15 EZ |
57 | #include "lisp.h" |
58 | #include "buffer.h" | |
59 | #include "character.h" | |
60 | #include "dispextern.h" | |
61 | ||
62 | static int bidi_initialized = 0; | |
63 | ||
cbc4fd20 | 64 | static Lisp_Object bidi_type_table, bidi_mirror_table; |
b7b65b15 | 65 | |
e7402cb2 | 66 | /* FIXME: Remove these when bidi_explicit_dir_char uses a lookup table. */ |
b7b65b15 EZ |
67 | #define LRM_CHAR 0x200E |
68 | #define RLM_CHAR 0x200F | |
69 | #define LRE_CHAR 0x202A | |
70 | #define RLE_CHAR 0x202B | |
71 | #define PDF_CHAR 0x202C | |
72 | #define LRO_CHAR 0x202D | |
73 | #define RLO_CHAR 0x202E | |
74 | ||
b7b65b15 | 75 | #define BIDI_EOB -1 |
e7402cb2 | 76 | #define BIDI_BOB -2 /* FIXME: Is this needed? */ |
b7b65b15 | 77 | |
b7b65b15 EZ |
78 | /* Local data structures. (Look in dispextern.h for the rest.) */ |
79 | ||
80 | /* What we need to know about the current paragraph. */ | |
81 | struct bidi_paragraph_info { | |
82 | int start_bytepos; /* byte position where it begins */ | |
83 | int end_bytepos; /* byte position where it ends */ | |
84 | int embedding_level; /* its basic embedding level */ | |
85 | bidi_dir_t base_dir; /* its base direction */ | |
86 | }; | |
87 | ||
88 | /* Data type for describing the bidirectional character categories. */ | |
89 | typedef enum { | |
90 | UNKNOWN_BC, | |
91 | NEUTRAL, | |
92 | WEAK, | |
93 | STRONG | |
94 | } bidi_category_t; | |
95 | ||
96 | int bidi_ignore_explicit_marks_for_paragraph_level = 1; | |
97 | ||
372b7a95 | 98 | static Lisp_Object paragraph_start_re, paragraph_separate_re; |
6bff6497 | 99 | static Lisp_Object Qparagraph_start, Qparagraph_separate; |
b7b65b15 | 100 | |
b7b65b15 | 101 | static void |
971de7fb | 102 | bidi_initialize (void) |
b7b65b15 | 103 | { |
317fbf33 EZ |
104 | |
105 | #include "biditype.h" | |
cbc4fd20 | 106 | #include "bidimirror.h" |
317fbf33 | 107 | |
b7b65b15 EZ |
108 | int i; |
109 | ||
110 | bidi_type_table = Fmake_char_table (Qnil, make_number (STRONG_L)); | |
2d6e4628 | 111 | staticpro (&bidi_type_table); |
b7b65b15 EZ |
112 | |
113 | for (i = 0; i < sizeof bidi_type / sizeof bidi_type[0]; i++) | |
317fbf33 | 114 | char_table_set_range (bidi_type_table, bidi_type[i].from, bidi_type[i].to, |
b7b65b15 | 115 | make_number (bidi_type[i].type)); |
6bff6497 | 116 | |
cbc4fd20 EZ |
117 | bidi_mirror_table = Fmake_char_table (Qnil, Qnil); |
118 | staticpro (&bidi_mirror_table); | |
119 | ||
120 | for (i = 0; i < sizeof bidi_mirror / sizeof bidi_mirror[0]; i++) | |
121 | char_table_set (bidi_mirror_table, bidi_mirror[i].from, | |
122 | make_number (bidi_mirror[i].to)); | |
123 | ||
b4bf28b7 SM |
124 | Qparagraph_start = intern ("paragraph-start"); |
125 | staticpro (&Qparagraph_start); | |
372b7a95 EZ |
126 | paragraph_start_re = Fsymbol_value (Qparagraph_start); |
127 | if (!STRINGP (paragraph_start_re)) | |
128 | paragraph_start_re = build_string ("\f\\|[ \t]*$"); | |
129 | staticpro (¶graph_start_re); | |
b4bf28b7 SM |
130 | Qparagraph_separate = intern ("paragraph-separate"); |
131 | staticpro (&Qparagraph_separate); | |
372b7a95 EZ |
132 | paragraph_separate_re = Fsymbol_value (Qparagraph_separate); |
133 | if (!STRINGP (paragraph_separate_re)) | |
134 | paragraph_separate_re = build_string ("[ \t\f]*$"); | |
135 | staticpro (¶graph_separate_re); | |
b7b65b15 EZ |
136 | bidi_initialized = 1; |
137 | } | |
138 | ||
6bff6497 EZ |
139 | /* Return the bidi type of a character CH, subject to the current |
140 | directional OVERRIDE. */ | |
fd3998ff | 141 | static INLINE bidi_type_t |
6bff6497 | 142 | bidi_get_type (int ch, bidi_dir_t override) |
b7b65b15 | 143 | { |
6bff6497 EZ |
144 | bidi_type_t default_type; |
145 | ||
e7402cb2 EZ |
146 | if (ch == BIDI_EOB) |
147 | return NEUTRAL_B; | |
e342a24d EZ |
148 | if (ch < 0 || ch > MAX_CHAR) |
149 | abort (); | |
6bff6497 EZ |
150 | |
151 | default_type = (bidi_type_t) XINT (CHAR_TABLE_REF (bidi_type_table, ch)); | |
152 | ||
153 | if (override == NEUTRAL_DIR) | |
154 | return default_type; | |
155 | ||
156 | switch (default_type) | |
157 | { | |
158 | /* Although UAX#9 does not tell, it doesn't make sense to | |
159 | override NEUTRAL_B and LRM/RLM characters. */ | |
160 | case NEUTRAL_B: | |
161 | case LRE: | |
162 | case LRO: | |
163 | case RLE: | |
164 | case RLO: | |
165 | case PDF: | |
166 | return default_type; | |
167 | default: | |
168 | switch (ch) | |
169 | { | |
170 | case LRM_CHAR: | |
171 | case RLM_CHAR: | |
172 | return default_type; | |
173 | default: | |
174 | if (override == L2R) /* X6 */ | |
175 | return STRONG_L; | |
176 | else if (override == R2L) | |
177 | return STRONG_R; | |
178 | else | |
179 | abort (); /* can't happen: handled above */ | |
180 | } | |
181 | } | |
b7b65b15 EZ |
182 | } |
183 | ||
2d6e4628 EZ |
184 | void |
185 | bidi_check_type (bidi_type_t type) | |
186 | { | |
187 | if (type < UNKNOWN_BT || type > NEUTRAL_ON) | |
188 | abort (); | |
189 | } | |
190 | ||
b7b65b15 | 191 | /* Given a bidi TYPE of a character, return its category. */ |
fd3998ff | 192 | static INLINE bidi_category_t |
b7b65b15 EZ |
193 | bidi_get_category (bidi_type_t type) |
194 | { | |
195 | switch (type) | |
196 | { | |
197 | case UNKNOWN_BT: | |
198 | return UNKNOWN_BC; | |
199 | case STRONG_L: | |
200 | case STRONG_R: | |
201 | case STRONG_AL: | |
202 | case LRE: | |
203 | case LRO: | |
204 | case RLE: | |
205 | case RLO: | |
206 | return STRONG; | |
207 | case PDF: /* ??? really?? */ | |
208 | case WEAK_EN: | |
209 | case WEAK_ES: | |
210 | case WEAK_ET: | |
211 | case WEAK_AN: | |
212 | case WEAK_CS: | |
213 | case WEAK_NSM: | |
214 | case WEAK_BN: | |
215 | return WEAK; | |
216 | case NEUTRAL_B: | |
217 | case NEUTRAL_S: | |
218 | case NEUTRAL_WS: | |
219 | case NEUTRAL_ON: | |
220 | return NEUTRAL; | |
221 | default: | |
222 | abort (); | |
223 | } | |
224 | } | |
225 | ||
a6e8d97c EZ |
226 | /* Return the mirrored character of C, if it has one. If C has no |
227 | mirrored counterpart, return C. | |
228 | Note: The conditions in UAX#9 clause L4 regarding the surrounding | |
229 | context must be tested by the caller. */ | |
b7b65b15 EZ |
230 | int |
231 | bidi_mirror_char (int c) | |
232 | { | |
cbc4fd20 EZ |
233 | Lisp_Object val; |
234 | ||
235 | if (c == BIDI_EOB) | |
236 | return c; | |
237 | if (c < 0 || c > MAX_CHAR) | |
238 | abort (); | |
b7b65b15 | 239 | |
cbc4fd20 EZ |
240 | val = CHAR_TABLE_REF (bidi_mirror_table, c); |
241 | if (INTEGERP (val)) | |
b7b65b15 | 242 | { |
cbc4fd20 | 243 | int v = XINT (val); |
b7b65b15 | 244 | |
cbc4fd20 EZ |
245 | if (v < 0 || v > MAX_CHAR) |
246 | abort (); | |
247 | ||
248 | return v; | |
b7b65b15 | 249 | } |
cbc4fd20 | 250 | |
b7b65b15 EZ |
251 | return c; |
252 | } | |
253 | ||
254 | /* Copy the bidi iterator from FROM to TO. To save cycles, this only | |
255 | copies the part of the level stack that is actually in use. */ | |
fd3998ff | 256 | static INLINE void |
b7b65b15 EZ |
257 | bidi_copy_it (struct bidi_it *to, struct bidi_it *from) |
258 | { | |
259 | int i; | |
260 | ||
e69a9370 EZ |
261 | /* Copy everything except the level stack and beyond. */ |
262 | memcpy (to, from, ((size_t)&((struct bidi_it *)0)->level_stack[0])); | |
b7b65b15 EZ |
263 | |
264 | /* Copy the active part of the level stack. */ | |
265 | to->level_stack[0] = from->level_stack[0]; /* level zero is always in use */ | |
266 | for (i = 1; i <= from->stack_idx; i++) | |
267 | to->level_stack[i] = from->level_stack[i]; | |
268 | } | |
269 | ||
270 | /* Caching the bidi iterator states. */ | |
271 | ||
2fe72643 EZ |
272 | #define BIDI_CACHE_CHUNK 200 |
273 | static struct bidi_it *bidi_cache; | |
274 | static size_t bidi_cache_size = 0; | |
0416466c | 275 | static size_t elsz = sizeof (struct bidi_it); |
2fe72643 EZ |
276 | static int bidi_cache_idx; /* next unused cache slot */ |
277 | static int bidi_cache_last_idx; /* slot of last cache hit */ | |
b7b65b15 | 278 | |
fd3998ff | 279 | static INLINE void |
b7b65b15 EZ |
280 | bidi_cache_reset (void) |
281 | { | |
282 | bidi_cache_idx = 0; | |
283 | bidi_cache_last_idx = -1; | |
284 | } | |
285 | ||
2fe72643 EZ |
286 | static INLINE void |
287 | bidi_cache_shrink (void) | |
288 | { | |
289 | if (bidi_cache_size > BIDI_CACHE_CHUNK) | |
290 | { | |
0416466c EZ |
291 | bidi_cache_size = BIDI_CACHE_CHUNK; |
292 | bidi_cache = | |
293 | (struct bidi_it *) xrealloc (bidi_cache, bidi_cache_size * elsz); | |
2fe72643 EZ |
294 | } |
295 | bidi_cache_reset (); | |
296 | } | |
297 | ||
fd3998ff | 298 | static INLINE void |
b7b65b15 EZ |
299 | bidi_cache_fetch_state (int idx, struct bidi_it *bidi_it) |
300 | { | |
301 | int current_scan_dir = bidi_it->scan_dir; | |
302 | ||
303 | if (idx < 0 || idx >= bidi_cache_idx) | |
304 | abort (); | |
305 | ||
306 | bidi_copy_it (bidi_it, &bidi_cache[idx]); | |
307 | bidi_it->scan_dir = current_scan_dir; | |
308 | bidi_cache_last_idx = idx; | |
309 | } | |
310 | ||
311 | /* Find a cached state with a given CHARPOS and resolved embedding | |
312 | level less or equal to LEVEL. if LEVEL is -1, disregard the | |
313 | resolved levels in cached states. DIR, if non-zero, means search | |
314 | in that direction from the last cache hit. */ | |
fd3998ff | 315 | static INLINE int |
b7b65b15 EZ |
316 | bidi_cache_search (int charpos, int level, int dir) |
317 | { | |
318 | int i, i_start; | |
319 | ||
320 | if (bidi_cache_idx) | |
321 | { | |
322 | if (charpos < bidi_cache[bidi_cache_last_idx].charpos) | |
323 | dir = -1; | |
324 | else if (charpos > bidi_cache[bidi_cache_last_idx].charpos) | |
325 | dir = 1; | |
326 | if (dir) | |
327 | i_start = bidi_cache_last_idx; | |
328 | else | |
329 | { | |
330 | dir = -1; | |
331 | i_start = bidi_cache_idx - 1; | |
332 | } | |
333 | ||
334 | if (dir < 0) | |
335 | { | |
336 | /* Linear search for now; FIXME! */ | |
337 | for (i = i_start; i >= 0; i--) | |
338 | if (bidi_cache[i].charpos == charpos | |
339 | && (level == -1 || bidi_cache[i].resolved_level <= level)) | |
340 | return i; | |
341 | } | |
342 | else | |
343 | { | |
344 | for (i = i_start; i < bidi_cache_idx; i++) | |
345 | if (bidi_cache[i].charpos == charpos | |
346 | && (level == -1 || bidi_cache[i].resolved_level <= level)) | |
347 | return i; | |
348 | } | |
349 | } | |
350 | ||
351 | return -1; | |
352 | } | |
353 | ||
354 | /* Find a cached state where the resolved level changes to a value | |
355 | that is lower than LEVEL, and return its cache slot index. DIR is | |
356 | the direction to search, starting with the last used cache slot. | |
357 | BEFORE, if non-zero, means return the index of the slot that is | |
358 | ``before'' the level change in the search direction. That is, | |
359 | given the cached levels like this: | |
360 | ||
361 | 1122333442211 | |
362 | AB C | |
363 | ||
364 | and assuming we are at the position cached at the slot marked with | |
365 | C, searching backwards (DIR = -1) for LEVEL = 2 will return the | |
366 | index of slot B or A, depending whether BEFORE is, respectively, | |
367 | non-zero or zero. */ | |
368 | static int | |
369 | bidi_cache_find_level_change (int level, int dir, int before) | |
370 | { | |
371 | if (bidi_cache_idx) | |
372 | { | |
373 | int i = dir ? bidi_cache_last_idx : bidi_cache_idx - 1; | |
374 | int incr = before ? 1 : 0; | |
375 | ||
376 | if (!dir) | |
377 | dir = -1; | |
378 | else if (!incr) | |
379 | i += dir; | |
380 | ||
381 | if (dir < 0) | |
382 | { | |
383 | while (i >= incr) | |
384 | { | |
385 | if (bidi_cache[i - incr].resolved_level >= 0 | |
386 | && bidi_cache[i - incr].resolved_level < level) | |
387 | return i; | |
388 | i--; | |
389 | } | |
390 | } | |
391 | else | |
392 | { | |
393 | while (i < bidi_cache_idx - incr) | |
394 | { | |
395 | if (bidi_cache[i + incr].resolved_level >= 0 | |
396 | && bidi_cache[i + incr].resolved_level < level) | |
397 | return i; | |
398 | i++; | |
399 | } | |
400 | } | |
401 | } | |
402 | ||
403 | return -1; | |
404 | } | |
405 | ||
fd3998ff | 406 | static INLINE void |
b7b65b15 EZ |
407 | bidi_cache_iterator_state (struct bidi_it *bidi_it, int resolved) |
408 | { | |
409 | int idx; | |
410 | ||
411 | /* We should never cache on backward scans. */ | |
412 | if (bidi_it->scan_dir == -1) | |
413 | abort (); | |
414 | idx = bidi_cache_search (bidi_it->charpos, -1, 1); | |
415 | ||
416 | if (idx < 0) | |
417 | { | |
418 | idx = bidi_cache_idx; | |
2fe72643 EZ |
419 | /* Enlarge the cache as needed. */ |
420 | if (idx >= bidi_cache_size) | |
421 | { | |
0416466c | 422 | bidi_cache_size += BIDI_CACHE_CHUNK; |
2fe72643 | 423 | bidi_cache = |
0416466c | 424 | (struct bidi_it *) xrealloc (bidi_cache, bidi_cache_size * elsz); |
2fe72643 | 425 | } |
bd924a5d EZ |
426 | /* Character positions should correspond to cache positions 1:1. |
427 | If we are outside the range of cached positions, the cache is | |
428 | useless and must be reset. */ | |
429 | if (idx > 0 && | |
430 | (bidi_it->charpos > bidi_cache[idx - 1].charpos + 1 | |
431 | || bidi_it->charpos < bidi_cache[0].charpos)) | |
432 | { | |
433 | bidi_cache_reset (); | |
434 | idx = 0; | |
435 | } | |
b7b65b15 EZ |
436 | bidi_copy_it (&bidi_cache[idx], bidi_it); |
437 | if (!resolved) | |
438 | bidi_cache[idx].resolved_level = -1; | |
439 | } | |
440 | else | |
441 | { | |
442 | /* Copy only the members which could have changed, to avoid | |
443 | costly copying of the entire struct. */ | |
444 | bidi_cache[idx].type = bidi_it->type; | |
2d6e4628 | 445 | bidi_check_type (bidi_it->type); |
89d3374a EZ |
446 | bidi_cache[idx].type_after_w1 = bidi_it->type_after_w1; |
447 | bidi_check_type (bidi_it->type_after_w1); | |
b7b65b15 EZ |
448 | if (resolved) |
449 | bidi_cache[idx].resolved_level = bidi_it->resolved_level; | |
450 | else | |
451 | bidi_cache[idx].resolved_level = -1; | |
452 | bidi_cache[idx].invalid_levels = bidi_it->invalid_levels; | |
453 | bidi_cache[idx].invalid_rl_levels = bidi_it->invalid_rl_levels; | |
454 | bidi_cache[idx].next_for_neutral = bidi_it->next_for_neutral; | |
455 | bidi_cache[idx].next_for_ws = bidi_it->next_for_ws; | |
456 | bidi_cache[idx].ignore_bn_limit = bidi_it->ignore_bn_limit; | |
457 | } | |
458 | ||
459 | bidi_cache_last_idx = idx; | |
460 | if (idx >= bidi_cache_idx) | |
461 | bidi_cache_idx = idx + 1; | |
462 | } | |
463 | ||
fd3998ff | 464 | static INLINE bidi_type_t |
b7b65b15 EZ |
465 | bidi_cache_find (int charpos, int level, struct bidi_it *bidi_it) |
466 | { | |
467 | int i = bidi_cache_search (charpos, level, bidi_it->scan_dir); | |
468 | ||
469 | if (i >= 0) | |
470 | { | |
471 | bidi_dir_t current_scan_dir = bidi_it->scan_dir; | |
472 | ||
5009f85e | 473 | bidi_copy_it (bidi_it, &bidi_cache[i]); |
b7b65b15 EZ |
474 | bidi_cache_last_idx = i; |
475 | /* Don't let scan direction from from the cached state override | |
476 | the current scan direction. */ | |
477 | bidi_it->scan_dir = current_scan_dir; | |
478 | return bidi_it->type; | |
479 | } | |
480 | ||
481 | return UNKNOWN_BT; | |
482 | } | |
483 | ||
fd3998ff | 484 | static INLINE int |
b7b65b15 EZ |
485 | bidi_peek_at_next_level (struct bidi_it *bidi_it) |
486 | { | |
487 | if (bidi_cache_idx == 0 || bidi_cache_last_idx == -1) | |
488 | abort (); | |
489 | return bidi_cache[bidi_cache_last_idx + bidi_it->scan_dir].resolved_level; | |
490 | } | |
491 | ||
be39f003 EZ |
492 | /* Check if buffer position CHARPOS/BYTEPOS is the end of a paragraph. |
493 | Value is the non-negative length of the paragraph separator | |
494 | following the buffer position, -1 if position is at the beginning | |
495 | of a new paragraph, or -2 if position is neither at beginning nor | |
496 | at end of a paragraph. */ | |
fd3998ff | 497 | static EMACS_INT |
6bff6497 | 498 | bidi_at_paragraph_end (EMACS_INT charpos, EMACS_INT bytepos) |
b7b65b15 | 499 | { |
b4bf28b7 | 500 | /* FIXME: Why Fbuffer_local_value rather than just Fsymbol_value? */ |
372b7a95 EZ |
501 | Lisp_Object sep_re; |
502 | Lisp_Object start_re; | |
be39f003 EZ |
503 | EMACS_INT val; |
504 | ||
372b7a95 EZ |
505 | sep_re = paragraph_separate_re; |
506 | start_re = paragraph_start_re; | |
be39f003 EZ |
507 | |
508 | val = fast_looking_at (sep_re, charpos, bytepos, ZV, ZV_BYTE, Qnil); | |
509 | if (val < 0) | |
510 | { | |
511 | if (fast_looking_at (start_re, charpos, bytepos, ZV, ZV_BYTE, Qnil) >= 0) | |
512 | val = -1; | |
513 | else | |
514 | val = -2; | |
515 | } | |
b7b65b15 | 516 | |
be39f003 | 517 | return val; |
b7b65b15 EZ |
518 | } |
519 | ||
520 | /* Determine the start-of-run (sor) directional type given the two | |
521 | embedding levels on either side of the run boundary. Also, update | |
522 | the saved info about previously seen characters, since that info is | |
523 | generally valid for a single level run. */ | |
fd3998ff | 524 | static INLINE void |
b7b65b15 EZ |
525 | bidi_set_sor_type (struct bidi_it *bidi_it, int level_before, int level_after) |
526 | { | |
527 | int higher_level = level_before > level_after ? level_before : level_after; | |
528 | ||
529 | /* The prev_was_pdf gork is required for when we have several PDFs | |
530 | in a row. In that case, we want to compute the sor type for the | |
531 | next level run only once: when we see the first PDF. That's | |
532 | because the sor type depends only on the higher of the two levels | |
533 | that we find on the two sides of the level boundary (see UAX#9, | |
534 | clause X10), and so we don't need to know the final embedding | |
535 | level to which we descend after processing all the PDFs. */ | |
e342a24d | 536 | if (!bidi_it->prev_was_pdf || level_before < level_after) |
b7b65b15 EZ |
537 | /* FIXME: should the default sor direction be user selectable? */ |
538 | bidi_it->sor = (higher_level & 1) != 0 ? R2L : L2R; | |
539 | if (level_before > level_after) | |
540 | bidi_it->prev_was_pdf = 1; | |
541 | ||
542 | bidi_it->prev.type = UNKNOWN_BT; | |
89d3374a EZ |
543 | bidi_it->last_strong.type = bidi_it->last_strong.type_after_w1 = |
544 | bidi_it->last_strong.orig_type = UNKNOWN_BT; | |
b7b65b15 EZ |
545 | bidi_it->prev_for_neutral.type = bidi_it->sor == R2L ? STRONG_R : STRONG_L; |
546 | bidi_it->prev_for_neutral.charpos = bidi_it->charpos; | |
547 | bidi_it->prev_for_neutral.bytepos = bidi_it->bytepos; | |
89d3374a EZ |
548 | bidi_it->next_for_neutral.type = bidi_it->next_for_neutral.type_after_w1 = |
549 | bidi_it->next_for_neutral.orig_type = UNKNOWN_BT; | |
b7b65b15 EZ |
550 | bidi_it->ignore_bn_limit = 0; /* meaning it's unknown */ |
551 | } | |
552 | ||
6bff6497 | 553 | static void |
be39f003 EZ |
554 | bidi_line_init (struct bidi_it *bidi_it) |
555 | { | |
556 | bidi_it->scan_dir = 1; /* FIXME: do we need to have control on this? */ | |
557 | bidi_it->resolved_level = bidi_it->level_stack[0].level; | |
558 | bidi_it->level_stack[0].override = NEUTRAL_DIR; /* X1 */ | |
559 | bidi_it->invalid_levels = 0; | |
560 | bidi_it->invalid_rl_levels = -1; | |
561 | bidi_it->next_en_pos = -1; | |
562 | bidi_it->next_for_ws.type = UNKNOWN_BT; | |
b44d9321 EZ |
563 | bidi_set_sor_type (bidi_it, |
564 | bidi_it->paragraph_dir == R2L ? 1 : 0, | |
be39f003 EZ |
565 | bidi_it->level_stack[0].level); /* X10 */ |
566 | ||
567 | bidi_cache_reset (); | |
568 | } | |
569 | ||
570 | /* Find the beginning of this paragraph by looking back in the buffer. | |
571 | Value is the byte position of the paragraph's beginning. */ | |
572 | static EMACS_INT | |
b44d9321 | 573 | bidi_find_paragraph_start (EMACS_INT pos, EMACS_INT pos_byte) |
6bff6497 | 574 | { |
372b7a95 | 575 | Lisp_Object re = paragraph_start_re; |
6bff6497 EZ |
576 | EMACS_INT limit = ZV, limit_byte = ZV_BYTE; |
577 | ||
6bff6497 EZ |
578 | while (pos_byte > BEGV_BYTE |
579 | && fast_looking_at (re, pos, pos_byte, limit, limit_byte, Qnil) < 0) | |
580 | { | |
be39f003 EZ |
581 | pos = find_next_newline_no_quit (pos - 1, -1); |
582 | pos_byte = CHAR_TO_BYTE (pos); | |
6bff6497 | 583 | } |
be39f003 | 584 | return pos_byte; |
6bff6497 EZ |
585 | } |
586 | ||
be39f003 | 587 | /* Determine the direction, a.k.a. base embedding level, of the |
b44d9321 EZ |
588 | paragraph we are about to iterate through. If DIR is either L2R or |
589 | R2L, just use that. Otherwise, determine the paragraph direction | |
590 | from the first strong character of the paragraph. | |
591 | ||
592 | Note that this gives the paragraph separator the same direction as | |
593 | the preceding paragraph, even though Emacs generally views the | |
594 | separartor as not belonging to any paragraph. */ | |
b7b65b15 EZ |
595 | void |
596 | bidi_paragraph_init (bidi_dir_t dir, struct bidi_it *bidi_it) | |
597 | { | |
6bff6497 | 598 | EMACS_INT bytepos = bidi_it->bytepos; |
e7402cb2 | 599 | |
5e65aec0 EZ |
600 | /* Special case for an empty buffer. */ |
601 | if (bytepos == BEGV_BYTE && bytepos == ZV_BYTE) | |
602 | dir = L2R; | |
9c82e145 | 603 | /* We should never be called at EOB or before BEGV. */ |
5e65aec0 | 604 | else if (bytepos >= ZV_BYTE || bytepos < BEGV_BYTE) |
9c82e145 EZ |
605 | abort (); |
606 | ||
be39f003 EZ |
607 | if (dir == L2R) |
608 | { | |
609 | bidi_it->paragraph_dir = L2R; | |
610 | bidi_it->new_paragraph = 0; | |
611 | } | |
612 | else if (dir == R2L) | |
613 | { | |
614 | bidi_it->paragraph_dir = R2L; | |
615 | bidi_it->new_paragraph = 0; | |
616 | } | |
b7b65b15 EZ |
617 | else if (dir == NEUTRAL_DIR) /* P2 */ |
618 | { | |
6bff6497 EZ |
619 | int ch, ch_len; |
620 | EMACS_INT pos; | |
621 | bidi_type_t type; | |
be39f003 | 622 | |
d20e1419 EZ |
623 | if (!bidi_initialized) |
624 | bidi_initialize (); | |
625 | ||
be39f003 EZ |
626 | /* If we are inside a paragraph separator, we are just waiting |
627 | for the separator to be exhausted; use the previous paragraph | |
e5a2fec7 EZ |
628 | direction. But don't do that if we have been just reseated, |
629 | because we need to reinitialize below in that case. */ | |
630 | if (!bidi_it->first_elt | |
631 | && bidi_it->charpos < bidi_it->separator_limit) | |
be39f003 EZ |
632 | return; |
633 | ||
b44d9321 | 634 | /* If we are on a newline, get past it to where the next |
5e65aec0 EZ |
635 | paragraph might start. But don't do that at BEGV since then |
636 | we are potentially in a new paragraph that doesn't yet | |
637 | exist. */ | |
c143c213 | 638 | pos = bidi_it->charpos; |
5e65aec0 | 639 | if (bytepos > BEGV_BYTE && FETCH_CHAR (bytepos) == '\n') |
be39f003 | 640 | { |
b44d9321 | 641 | bytepos++; |
c143c213 | 642 | pos++; |
be39f003 | 643 | } |
b44d9321 EZ |
644 | |
645 | /* We are either at the beginning of a paragraph or in the | |
646 | middle of it. Find where this paragraph starts. */ | |
647 | bytepos = bidi_find_paragraph_start (pos, bytepos); | |
6bff6497 | 648 | |
be39f003 EZ |
649 | bidi_it->separator_limit = -1; |
650 | bidi_it->new_paragraph = 0; | |
6bff6497 EZ |
651 | ch = FETCH_CHAR (bytepos); |
652 | ch_len = CHAR_BYTES (ch); | |
be39f003 | 653 | pos = BYTE_TO_CHAR (bytepos); |
6bff6497 | 654 | type = bidi_get_type (ch, NEUTRAL_DIR); |
b7b65b15 | 655 | |
e342a24d | 656 | for (pos++, bytepos += ch_len; |
b7b65b15 EZ |
657 | /* NOTE: UAX#9 says to search only for L, AL, or R types of |
658 | characters, and ignore RLE, RLO, LRE, and LRO. However, | |
659 | I'm not sure it makes sense to omit those 4; should try | |
660 | with and without that to see the effect. */ | |
661 | (bidi_get_category (type) != STRONG) | |
662 | || (bidi_ignore_explicit_marks_for_paragraph_level | |
663 | && (type == RLE || type == RLO | |
664 | || type == LRE || type == LRO)); | |
6bff6497 | 665 | type = bidi_get_type (ch, NEUTRAL_DIR)) |
b7b65b15 | 666 | { |
be39f003 | 667 | if (type == NEUTRAL_B && bidi_at_paragraph_end (pos, bytepos) >= -1) |
b7b65b15 | 668 | break; |
21fce5ab EZ |
669 | if (bytepos >= ZV_BYTE) |
670 | { | |
671 | /* Pretend there's a paragraph separator at end of buffer. */ | |
672 | type = NEUTRAL_B; | |
673 | break; | |
674 | } | |
b7b65b15 EZ |
675 | FETCH_CHAR_ADVANCE (ch, pos, bytepos); |
676 | } | |
677 | if (type == STRONG_R || type == STRONG_AL) /* P3 */ | |
be39f003 EZ |
678 | bidi_it->paragraph_dir = R2L; |
679 | else if (type == STRONG_L) | |
680 | bidi_it->paragraph_dir = L2R; | |
b7b65b15 | 681 | } |
be39f003 EZ |
682 | else |
683 | abort (); | |
b7b65b15 | 684 | |
b44d9321 EZ |
685 | /* Contrary to UAX#9 clause P3, we only default the paragraph |
686 | direction to L2R if we have no previous usable paragraph | |
687 | direction. */ | |
d20e1419 | 688 | if (bidi_it->paragraph_dir != L2R && bidi_it->paragraph_dir != R2L) |
b44d9321 | 689 | bidi_it->paragraph_dir = L2R; /* P3 and ``higher protocols'' */ |
be39f003 | 690 | if (bidi_it->paragraph_dir == R2L) |
b44d9321 | 691 | bidi_it->level_stack[0].level = 1; |
be39f003 | 692 | else |
b44d9321 | 693 | bidi_it->level_stack[0].level = 0; |
be39f003 EZ |
694 | |
695 | bidi_line_init (bidi_it); | |
b7b65b15 EZ |
696 | } |
697 | ||
6bff6497 EZ |
698 | /* Do whatever UAX#9 clause X8 says should be done at paragraph's |
699 | end. */ | |
fd3998ff | 700 | static INLINE void |
b7b65b15 EZ |
701 | bidi_set_paragraph_end (struct bidi_it *bidi_it) |
702 | { | |
703 | bidi_it->invalid_levels = 0; | |
704 | bidi_it->invalid_rl_levels = -1; | |
705 | bidi_it->stack_idx = 0; | |
706 | bidi_it->resolved_level = bidi_it->level_stack[0].level; | |
b7b65b15 EZ |
707 | } |
708 | ||
89d3374a | 709 | /* Initialize the bidi iterator from buffer position CHARPOS. */ |
b7b65b15 | 710 | void |
6bff6497 | 711 | bidi_init_it (EMACS_INT charpos, EMACS_INT bytepos, struct bidi_it *bidi_it) |
b7b65b15 EZ |
712 | { |
713 | if (! bidi_initialized) | |
714 | bidi_initialize (); | |
89d3374a EZ |
715 | bidi_it->charpos = charpos; |
716 | bidi_it->bytepos = bytepos; | |
6bff6497 EZ |
717 | bidi_it->first_elt = 1; |
718 | bidi_set_paragraph_end (bidi_it); | |
719 | bidi_it->new_paragraph = 1; | |
be39f003 | 720 | bidi_it->separator_limit = -1; |
b7b65b15 | 721 | bidi_it->type = NEUTRAL_B; |
ebb5722e EZ |
722 | bidi_it->type_after_w1 = NEUTRAL_B; |
723 | bidi_it->orig_type = NEUTRAL_B; | |
b7b65b15 | 724 | bidi_it->prev_was_pdf = 0; |
ebb5722e EZ |
725 | bidi_it->prev.type = bidi_it->prev.type_after_w1 = |
726 | bidi_it->prev.orig_type = UNKNOWN_BT; | |
89d3374a EZ |
727 | bidi_it->last_strong.type = bidi_it->last_strong.type_after_w1 = |
728 | bidi_it->last_strong.orig_type = UNKNOWN_BT; | |
b7b65b15 EZ |
729 | bidi_it->next_for_neutral.charpos = -1; |
730 | bidi_it->next_for_neutral.type = | |
89d3374a EZ |
731 | bidi_it->next_for_neutral.type_after_w1 = |
732 | bidi_it->next_for_neutral.orig_type = UNKNOWN_BT; | |
b7b65b15 EZ |
733 | bidi_it->prev_for_neutral.charpos = -1; |
734 | bidi_it->prev_for_neutral.type = | |
89d3374a EZ |
735 | bidi_it->prev_for_neutral.type_after_w1 = |
736 | bidi_it->prev_for_neutral.orig_type = UNKNOWN_BT; | |
b7b65b15 | 737 | bidi_it->sor = L2R; /* FIXME: should it be user-selectable? */ |
2fe72643 | 738 | bidi_cache_shrink (); |
b7b65b15 EZ |
739 | } |
740 | ||
741 | /* Push the current embedding level and override status; reset the | |
742 | current level to LEVEL and the current override status to OVERRIDE. */ | |
fd3998ff | 743 | static INLINE void |
b7b65b15 EZ |
744 | bidi_push_embedding_level (struct bidi_it *bidi_it, |
745 | int level, bidi_dir_t override) | |
746 | { | |
747 | bidi_it->stack_idx++; | |
748 | if (bidi_it->stack_idx >= BIDI_MAXLEVEL) | |
749 | abort (); | |
750 | bidi_it->level_stack[bidi_it->stack_idx].level = level; | |
751 | bidi_it->level_stack[bidi_it->stack_idx].override = override; | |
752 | } | |
753 | ||
754 | /* Pop the embedding level and directional override status from the | |
755 | stack, and return the new level. */ | |
fd3998ff | 756 | static INLINE int |
b7b65b15 EZ |
757 | bidi_pop_embedding_level (struct bidi_it *bidi_it) |
758 | { | |
759 | /* UAX#9 says to ignore invalid PDFs. */ | |
760 | if (bidi_it->stack_idx > 0) | |
761 | bidi_it->stack_idx--; | |
762 | return bidi_it->level_stack[bidi_it->stack_idx].level; | |
763 | } | |
764 | ||
765 | /* Record in SAVED_INFO the information about the current character. */ | |
fd3998ff | 766 | static INLINE void |
b7b65b15 EZ |
767 | bidi_remember_char (struct bidi_saved_info *saved_info, |
768 | struct bidi_it *bidi_it) | |
769 | { | |
770 | saved_info->charpos = bidi_it->charpos; | |
771 | saved_info->bytepos = bidi_it->bytepos; | |
772 | saved_info->type = bidi_it->type; | |
2d6e4628 | 773 | bidi_check_type (bidi_it->type); |
89d3374a EZ |
774 | saved_info->type_after_w1 = bidi_it->type_after_w1; |
775 | bidi_check_type (bidi_it->type_after_w1); | |
b7b65b15 | 776 | saved_info->orig_type = bidi_it->orig_type; |
2d6e4628 | 777 | bidi_check_type (bidi_it->orig_type); |
b7b65b15 EZ |
778 | } |
779 | ||
780 | /* Resolve the type of a neutral character according to the type of | |
781 | surrounding strong text and the current embedding level. */ | |
fd3998ff | 782 | static INLINE bidi_type_t |
b7b65b15 EZ |
783 | bidi_resolve_neutral_1 (bidi_type_t prev_type, bidi_type_t next_type, int lev) |
784 | { | |
785 | /* N1: European and Arabic numbers are treated as though they were R. */ | |
786 | if (next_type == WEAK_EN || next_type == WEAK_AN) | |
787 | next_type = STRONG_R; | |
788 | if (prev_type == WEAK_EN || prev_type == WEAK_AN) | |
789 | prev_type = STRONG_R; | |
790 | ||
791 | if (next_type == prev_type) /* N1 */ | |
792 | return next_type; | |
793 | else if ((lev & 1) == 0) /* N2 */ | |
794 | return STRONG_L; | |
795 | else | |
796 | return STRONG_R; | |
797 | } | |
798 | ||
fd3998ff | 799 | static INLINE int |
b7b65b15 EZ |
800 | bidi_explicit_dir_char (int c) |
801 | { | |
802 | /* FIXME: this should be replaced with a lookup table with suitable | |
803 | bits set, like standard C ctype macros do. */ | |
804 | return (c == LRE_CHAR || c == LRO_CHAR | |
805 | || c == RLE_CHAR || c == RLO_CHAR || c == PDF_CHAR); | |
806 | } | |
807 | ||
808 | /* A helper function for bidi_resolve_explicit. It advances to the | |
809 | next character in logical order and determines the new embedding | |
810 | level and directional override, but does not take into account | |
811 | empty embeddings. */ | |
812 | static int | |
813 | bidi_resolve_explicit_1 (struct bidi_it *bidi_it) | |
814 | { | |
815 | int curchar; | |
816 | bidi_type_t type; | |
817 | int current_level; | |
818 | int new_level; | |
819 | bidi_dir_t override; | |
820 | ||
9c82e145 EZ |
821 | if (bidi_it->bytepos < BEGV_BYTE /* after reseat to BEGV? */ |
822 | || bidi_it->first_elt) | |
e7402cb2 | 823 | { |
9c82e145 EZ |
824 | bidi_it->first_elt = 0; |
825 | if (bidi_it->charpos < BEGV) | |
826 | bidi_it->charpos = BEGV; | |
827 | bidi_it->bytepos = CHAR_TO_BYTE (bidi_it->charpos); | |
e7402cb2 | 828 | } |
9c82e145 | 829 | else if (bidi_it->bytepos < ZV_BYTE) /* don't move at ZV */ |
b7b65b15 EZ |
830 | { |
831 | bidi_it->charpos++; | |
e7402cb2 EZ |
832 | if (bidi_it->ch_len == 0) |
833 | abort (); | |
b7b65b15 EZ |
834 | bidi_it->bytepos += bidi_it->ch_len; |
835 | } | |
836 | ||
837 | current_level = bidi_it->level_stack[bidi_it->stack_idx].level; /* X1 */ | |
838 | override = bidi_it->level_stack[bidi_it->stack_idx].override; | |
839 | new_level = current_level; | |
840 | ||
841 | /* in case it is a unibyte character (not yet implemented) */ | |
842 | /* _fetch_multibyte_char_len = 1; */ | |
e7402cb2 EZ |
843 | if (bidi_it->bytepos >= ZV_BYTE) |
844 | { | |
845 | curchar = BIDI_EOB; | |
846 | bidi_it->ch_len = 1; | |
847 | } | |
848 | else | |
849 | { | |
850 | curchar = FETCH_CHAR (bidi_it->bytepos); | |
851 | bidi_it->ch_len = CHAR_BYTES (curchar); | |
852 | } | |
b7b65b15 | 853 | bidi_it->ch = curchar; |
b7b65b15 | 854 | |
6bff6497 EZ |
855 | /* Don't apply directional override here, as all the types we handle |
856 | below will not be affected by the override anyway, and we need | |
857 | the original type unaltered. The override will be applied in | |
858 | bidi_resolve_weak. */ | |
859 | type = bidi_get_type (curchar, NEUTRAL_DIR); | |
89d3374a EZ |
860 | bidi_it->orig_type = type; |
861 | bidi_check_type (bidi_it->orig_type); | |
b7b65b15 EZ |
862 | |
863 | if (type != PDF) | |
864 | bidi_it->prev_was_pdf = 0; | |
865 | ||
89d3374a | 866 | bidi_it->type_after_w1 = UNKNOWN_BT; |
b7b65b15 EZ |
867 | |
868 | switch (type) | |
869 | { | |
870 | case RLE: /* X2 */ | |
871 | case RLO: /* X4 */ | |
89d3374a EZ |
872 | bidi_it->type_after_w1 = type; |
873 | bidi_check_type (bidi_it->type_after_w1); | |
b7b65b15 EZ |
874 | type = WEAK_BN; /* X9/Retaining */ |
875 | if (bidi_it->ignore_bn_limit <= 0) | |
876 | { | |
877 | if (current_level <= BIDI_MAXLEVEL - 4) | |
878 | { | |
879 | /* Compute the least odd embedding level greater than | |
880 | the current level. */ | |
881 | new_level = ((current_level + 1) & ~1) + 1; | |
89d3374a | 882 | if (bidi_it->type_after_w1 == RLE) |
b7b65b15 EZ |
883 | override = NEUTRAL_DIR; |
884 | else | |
885 | override = R2L; | |
886 | if (current_level == BIDI_MAXLEVEL - 4) | |
887 | bidi_it->invalid_rl_levels = 0; | |
888 | bidi_push_embedding_level (bidi_it, new_level, override); | |
889 | } | |
890 | else | |
891 | { | |
892 | bidi_it->invalid_levels++; | |
893 | /* See the commentary about invalid_rl_levels below. */ | |
894 | if (bidi_it->invalid_rl_levels < 0) | |
895 | bidi_it->invalid_rl_levels = 0; | |
896 | bidi_it->invalid_rl_levels++; | |
897 | } | |
898 | } | |
89d3374a | 899 | else if (bidi_it->prev.type_after_w1 == WEAK_EN /* W5/Retaining */ |
b7b65b15 EZ |
900 | || bidi_it->next_en_pos > bidi_it->charpos) |
901 | type = WEAK_EN; | |
902 | break; | |
903 | case LRE: /* X3 */ | |
904 | case LRO: /* X5 */ | |
89d3374a EZ |
905 | bidi_it->type_after_w1 = type; |
906 | bidi_check_type (bidi_it->type_after_w1); | |
b7b65b15 EZ |
907 | type = WEAK_BN; /* X9/Retaining */ |
908 | if (bidi_it->ignore_bn_limit <= 0) | |
909 | { | |
910 | if (current_level <= BIDI_MAXLEVEL - 5) | |
911 | { | |
912 | /* Compute the least even embedding level greater than | |
913 | the current level. */ | |
914 | new_level = ((current_level + 2) & ~1); | |
89d3374a | 915 | if (bidi_it->type_after_w1 == LRE) |
b7b65b15 EZ |
916 | override = NEUTRAL_DIR; |
917 | else | |
918 | override = L2R; | |
919 | bidi_push_embedding_level (bidi_it, new_level, override); | |
920 | } | |
921 | else | |
922 | { | |
923 | bidi_it->invalid_levels++; | |
924 | /* invalid_rl_levels counts invalid levels encountered | |
925 | while the embedding level was already too high for | |
926 | LRE/LRO, but not for RLE/RLO. That is because | |
927 | there may be exactly one PDF which we should not | |
928 | ignore even though invalid_levels is non-zero. | |
929 | invalid_rl_levels helps to know what PDF is | |
930 | that. */ | |
931 | if (bidi_it->invalid_rl_levels >= 0) | |
932 | bidi_it->invalid_rl_levels++; | |
933 | } | |
934 | } | |
89d3374a | 935 | else if (bidi_it->prev.type_after_w1 == WEAK_EN /* W5/Retaining */ |
b7b65b15 EZ |
936 | || bidi_it->next_en_pos > bidi_it->charpos) |
937 | type = WEAK_EN; | |
938 | break; | |
939 | case PDF: /* X7 */ | |
89d3374a EZ |
940 | bidi_it->type_after_w1 = type; |
941 | bidi_check_type (bidi_it->type_after_w1); | |
b7b65b15 EZ |
942 | type = WEAK_BN; /* X9/Retaining */ |
943 | if (bidi_it->ignore_bn_limit <= 0) | |
944 | { | |
945 | if (!bidi_it->invalid_rl_levels) | |
946 | { | |
947 | new_level = bidi_pop_embedding_level (bidi_it); | |
948 | bidi_it->invalid_rl_levels = -1; | |
949 | if (bidi_it->invalid_levels) | |
950 | bidi_it->invalid_levels--; | |
951 | /* else nothing: UAX#9 says to ignore invalid PDFs */ | |
952 | } | |
953 | if (!bidi_it->invalid_levels) | |
954 | new_level = bidi_pop_embedding_level (bidi_it); | |
955 | else | |
956 | { | |
957 | bidi_it->invalid_levels--; | |
958 | bidi_it->invalid_rl_levels--; | |
959 | } | |
960 | } | |
89d3374a | 961 | else if (bidi_it->prev.type_after_w1 == WEAK_EN /* W5/Retaining */ |
b7b65b15 EZ |
962 | || bidi_it->next_en_pos > bidi_it->charpos) |
963 | type = WEAK_EN; | |
964 | break; | |
965 | default: | |
966 | /* Nothing. */ | |
967 | break; | |
968 | } | |
969 | ||
970 | bidi_it->type = type; | |
2d6e4628 | 971 | bidi_check_type (bidi_it->type); |
b7b65b15 EZ |
972 | |
973 | return new_level; | |
974 | } | |
975 | ||
976 | /* Given an iterator state in BIDI_IT, advance one character position | |
977 | in the buffer to the next character (in the logical order), resolve | |
978 | any explicit embeddings and directional overrides, and return the | |
979 | embedding level of the character after resolving explicit | |
980 | directives and ignoring empty embeddings. */ | |
981 | static int | |
982 | bidi_resolve_explicit (struct bidi_it *bidi_it) | |
983 | { | |
984 | int prev_level = bidi_it->level_stack[bidi_it->stack_idx].level; | |
985 | int new_level = bidi_resolve_explicit_1 (bidi_it); | |
986 | ||
987 | if (prev_level < new_level | |
988 | && bidi_it->type == WEAK_BN | |
989 | && bidi_it->ignore_bn_limit == 0 /* only if not already known */ | |
1502b819 | 990 | && bidi_it->bytepos < ZV_BYTE /* not already at EOB */ |
b7b65b15 EZ |
991 | && bidi_explicit_dir_char (FETCH_CHAR (bidi_it->bytepos |
992 | + bidi_it->ch_len))) | |
993 | { | |
994 | /* Avoid pushing and popping embedding levels if the level run | |
995 | is empty, as this breaks level runs where it shouldn't. | |
996 | UAX#9 removes all the explicit embedding and override codes, | |
997 | so empty embeddings disappear without a trace. We need to | |
998 | behave as if we did the same. */ | |
999 | struct bidi_it saved_it; | |
1000 | int level = prev_level; | |
1001 | ||
1002 | bidi_copy_it (&saved_it, bidi_it); | |
1003 | ||
1004 | while (bidi_explicit_dir_char (FETCH_CHAR (bidi_it->bytepos | |
1005 | + bidi_it->ch_len))) | |
1006 | { | |
1007 | level = bidi_resolve_explicit_1 (bidi_it); | |
1008 | } | |
1009 | ||
1010 | if (level == prev_level) /* empty embedding */ | |
1011 | saved_it.ignore_bn_limit = bidi_it->charpos + 1; | |
1012 | else /* this embedding is non-empty */ | |
1013 | saved_it.ignore_bn_limit = -1; | |
1014 | ||
1015 | bidi_copy_it (bidi_it, &saved_it); | |
1016 | if (bidi_it->ignore_bn_limit > 0) | |
1017 | { | |
1018 | /* We pushed a level, but we shouldn't have. Undo that. */ | |
1019 | if (!bidi_it->invalid_rl_levels) | |
1020 | { | |
1021 | new_level = bidi_pop_embedding_level (bidi_it); | |
1022 | bidi_it->invalid_rl_levels = -1; | |
1023 | if (bidi_it->invalid_levels) | |
1024 | bidi_it->invalid_levels--; | |
1025 | } | |
1026 | if (!bidi_it->invalid_levels) | |
1027 | new_level = bidi_pop_embedding_level (bidi_it); | |
1028 | else | |
1029 | { | |
1030 | bidi_it->invalid_levels--; | |
1031 | bidi_it->invalid_rl_levels--; | |
1032 | } | |
1033 | } | |
1034 | } | |
1035 | ||
b7b65b15 EZ |
1036 | if (bidi_it->type == NEUTRAL_B) /* X8 */ |
1037 | { | |
21fce5ab | 1038 | bidi_set_paragraph_end (bidi_it); |
6bff6497 EZ |
1039 | /* This is needed by bidi_resolve_weak below, and in L1. */ |
1040 | bidi_it->type_after_w1 = bidi_it->type; | |
89d3374a | 1041 | bidi_check_type (bidi_it->type_after_w1); |
b7b65b15 EZ |
1042 | } |
1043 | ||
1044 | return new_level; | |
1045 | } | |
1046 | ||
1047 | /* Advance in the buffer, resolve weak types and return the type of | |
1048 | the next character after weak type resolution. */ | |
fd3998ff | 1049 | static bidi_type_t |
b7b65b15 EZ |
1050 | bidi_resolve_weak (struct bidi_it *bidi_it) |
1051 | { | |
1052 | bidi_type_t type; | |
1053 | bidi_dir_t override; | |
1054 | int prev_level = bidi_it->level_stack[bidi_it->stack_idx].level; | |
1055 | int new_level = bidi_resolve_explicit (bidi_it); | |
1056 | int next_char; | |
1057 | bidi_type_t type_of_next; | |
1058 | struct bidi_it saved_it; | |
1059 | ||
1060 | type = bidi_it->type; | |
1061 | override = bidi_it->level_stack[bidi_it->stack_idx].override; | |
1062 | ||
1063 | if (type == UNKNOWN_BT | |
1064 | || type == LRE | |
1065 | || type == LRO | |
1066 | || type == RLE | |
1067 | || type == RLO | |
1068 | || type == PDF) | |
1069 | abort (); | |
1070 | ||
1071 | if (new_level != prev_level | |
1072 | || bidi_it->type == NEUTRAL_B) | |
1073 | { | |
1074 | /* We've got a new embedding level run, compute the directional | |
1075 | type of sor and initialize per-run variables (UAX#9, clause | |
1076 | X10). */ | |
1077 | bidi_set_sor_type (bidi_it, prev_level, new_level); | |
1078 | } | |
1079 | else if (type == NEUTRAL_S || type == NEUTRAL_WS | |
1080 | || type == WEAK_BN || type == STRONG_AL) | |
89d3374a EZ |
1081 | bidi_it->type_after_w1 = type; /* needed in L1 */ |
1082 | bidi_check_type (bidi_it->type_after_w1); | |
b7b65b15 EZ |
1083 | |
1084 | /* Level and directional override status are already recorded in | |
1085 | bidi_it, and do not need any change; see X6. */ | |
1086 | if (override == R2L) /* X6 */ | |
1087 | type = STRONG_R; | |
1088 | else if (override == L2R) | |
1089 | type = STRONG_L; | |
bc5a45f3 | 1090 | else |
b7b65b15 | 1091 | { |
bc5a45f3 | 1092 | if (type == WEAK_NSM) /* W1 */ |
b7b65b15 | 1093 | { |
bc5a45f3 | 1094 | /* Note that we don't need to consider the case where the |
5930fe97 EZ |
1095 | prev character has its type overridden by an RLO or LRO, |
1096 | because then either the type of this NSM would have been | |
1097 | also overridden, or the previous character is outside the | |
1098 | current level run, and thus not relevant to this NSM. | |
1099 | This is why NSM gets the type_after_w1 of the previous | |
1100 | character. */ | |
ebb5722e EZ |
1101 | if (bidi_it->prev.type_after_w1 != UNKNOWN_BT |
1102 | /* if type_after_w1 is NEUTRAL_B, this NSM is at sor */ | |
1103 | && bidi_it->prev.type_after_w1 != NEUTRAL_B) | |
5930fe97 | 1104 | type = bidi_it->prev.type_after_w1; |
bc5a45f3 EZ |
1105 | else if (bidi_it->sor == R2L) |
1106 | type = STRONG_R; | |
1107 | else if (bidi_it->sor == L2R) | |
1108 | type = STRONG_L; | |
1109 | else /* shouldn't happen! */ | |
1110 | abort (); | |
b7b65b15 | 1111 | } |
bc5a45f3 EZ |
1112 | if (type == WEAK_EN /* W2 */ |
1113 | && bidi_it->last_strong.type_after_w1 == STRONG_AL) | |
1114 | type = WEAK_AN; | |
1115 | else if (type == STRONG_AL) /* W3 */ | |
1116 | type = STRONG_R; | |
1117 | else if ((type == WEAK_ES /* W4 */ | |
1118 | && bidi_it->prev.type_after_w1 == WEAK_EN | |
1119 | && bidi_it->prev.orig_type == WEAK_EN) | |
1120 | || (type == WEAK_CS | |
1121 | && ((bidi_it->prev.type_after_w1 == WEAK_EN | |
1122 | && bidi_it->prev.orig_type == WEAK_EN) | |
1123 | || bidi_it->prev.type_after_w1 == WEAK_AN))) | |
b7b65b15 | 1124 | { |
e7402cb2 EZ |
1125 | next_char = |
1126 | bidi_it->bytepos + bidi_it->ch_len >= ZV_BYTE | |
1127 | ? BIDI_EOB : FETCH_CHAR (bidi_it->bytepos + bidi_it->ch_len); | |
6bff6497 | 1128 | type_of_next = bidi_get_type (next_char, override); |
b7b65b15 | 1129 | |
bc5a45f3 | 1130 | if (type_of_next == WEAK_BN |
b7b65b15 EZ |
1131 | || bidi_explicit_dir_char (next_char)) |
1132 | { | |
1133 | bidi_copy_it (&saved_it, bidi_it); | |
1134 | while (bidi_resolve_explicit (bidi_it) == new_level | |
bc5a45f3 | 1135 | && bidi_it->type == WEAK_BN) |
b7b65b15 EZ |
1136 | ; |
1137 | type_of_next = bidi_it->type; | |
b7b65b15 EZ |
1138 | bidi_copy_it (bidi_it, &saved_it); |
1139 | } | |
bc5a45f3 EZ |
1140 | |
1141 | /* If the next character is EN, but the last strong-type | |
1142 | character is AL, that next EN will be changed to AN when | |
1143 | we process it in W2 above. So in that case, this ES | |
1144 | should not be changed into EN. */ | |
1145 | if (type == WEAK_ES | |
1146 | && type_of_next == WEAK_EN | |
1147 | && bidi_it->last_strong.type_after_w1 != STRONG_AL) | |
1148 | type = WEAK_EN; | |
1149 | else if (type == WEAK_CS) | |
b7b65b15 | 1150 | { |
bc5a45f3 EZ |
1151 | if (bidi_it->prev.type_after_w1 == WEAK_AN |
1152 | && (type_of_next == WEAK_AN | |
1153 | /* If the next character is EN, but the last | |
1154 | strong-type character is AL, EN will be later | |
1155 | changed to AN when we process it in W2 above. | |
1156 | So in that case, this ES should not be | |
1157 | changed into EN. */ | |
1158 | || (type_of_next == WEAK_EN | |
1159 | && bidi_it->last_strong.type_after_w1 == STRONG_AL))) | |
1160 | type = WEAK_AN; | |
1161 | else if (bidi_it->prev.type_after_w1 == WEAK_EN | |
1162 | && type_of_next == WEAK_EN | |
1163 | && bidi_it->last_strong.type_after_w1 != STRONG_AL) | |
1164 | type = WEAK_EN; | |
1165 | } | |
1166 | } | |
1167 | else if (type == WEAK_ET /* W5: ET with EN before or after it */ | |
1168 | || type == WEAK_BN) /* W5/Retaining */ | |
1169 | { | |
1170 | if (bidi_it->prev.type_after_w1 == WEAK_EN /* ET/BN w/EN before it */ | |
1171 | || bidi_it->next_en_pos > bidi_it->charpos) | |
1172 | type = WEAK_EN; | |
1173 | else /* W5: ET/BN with EN after it. */ | |
1174 | { | |
1175 | EMACS_INT en_pos = bidi_it->charpos + 1; | |
1176 | ||
1177 | next_char = | |
1178 | bidi_it->bytepos + bidi_it->ch_len >= ZV_BYTE | |
1179 | ? BIDI_EOB : FETCH_CHAR (bidi_it->bytepos + bidi_it->ch_len); | |
1180 | type_of_next = bidi_get_type (next_char, override); | |
1181 | ||
1182 | if (type_of_next == WEAK_ET | |
1183 | || type_of_next == WEAK_BN | |
1184 | || bidi_explicit_dir_char (next_char)) | |
1185 | { | |
1186 | bidi_copy_it (&saved_it, bidi_it); | |
1187 | while (bidi_resolve_explicit (bidi_it) == new_level | |
1188 | && (bidi_it->type == WEAK_BN | |
1189 | || bidi_it->type == WEAK_ET)) | |
1190 | ; | |
1191 | type_of_next = bidi_it->type; | |
1192 | en_pos = bidi_it->charpos; | |
1193 | bidi_copy_it (bidi_it, &saved_it); | |
1194 | } | |
1195 | if (type_of_next == WEAK_EN) | |
b7b65b15 | 1196 | { |
bc5a45f3 EZ |
1197 | /* If the last strong character is AL, the EN we've |
1198 | found will become AN when we get to it (W2). */ | |
1199 | if (bidi_it->last_strong.type_after_w1 != STRONG_AL) | |
1200 | { | |
1201 | type = WEAK_EN; | |
1202 | /* Remember this EN position, to speed up processing | |
1203 | of the next ETs. */ | |
1204 | bidi_it->next_en_pos = en_pos; | |
1205 | } | |
1206 | else if (type == WEAK_BN) | |
1207 | type = NEUTRAL_ON; /* W6/Retaining */ | |
b7b65b15 | 1208 | } |
b7b65b15 EZ |
1209 | } |
1210 | } | |
1211 | } | |
1212 | ||
1213 | if (type == WEAK_ES || type == WEAK_ET || type == WEAK_CS /* W6 */ | |
89d3374a EZ |
1214 | || (type == WEAK_BN |
1215 | && (bidi_it->prev.type_after_w1 == WEAK_CS /* W6/Retaining */ | |
1216 | || bidi_it->prev.type_after_w1 == WEAK_ES | |
1217 | || bidi_it->prev.type_after_w1 == WEAK_ET))) | |
b7b65b15 EZ |
1218 | type = NEUTRAL_ON; |
1219 | ||
1220 | /* Store the type we've got so far, before we clobber it with strong | |
1221 | types in W7 and while resolving neutral types. But leave alone | |
1222 | the original types that were recorded above, because we will need | |
1223 | them for the L1 clause. */ | |
89d3374a EZ |
1224 | if (bidi_it->type_after_w1 == UNKNOWN_BT) |
1225 | bidi_it->type_after_w1 = type; | |
1226 | bidi_check_type (bidi_it->type_after_w1); | |
b7b65b15 EZ |
1227 | |
1228 | if (type == WEAK_EN) /* W7 */ | |
1229 | { | |
89d3374a | 1230 | if ((bidi_it->last_strong.type_after_w1 == STRONG_L) |
b7b65b15 EZ |
1231 | || (bidi_it->last_strong.type == UNKNOWN_BT && bidi_it->sor == L2R)) |
1232 | type = STRONG_L; | |
1233 | } | |
1234 | ||
1235 | bidi_it->type = type; | |
2d6e4628 | 1236 | bidi_check_type (bidi_it->type); |
b7b65b15 EZ |
1237 | return type; |
1238 | } | |
1239 | ||
fd3998ff | 1240 | static bidi_type_t |
b7b65b15 EZ |
1241 | bidi_resolve_neutral (struct bidi_it *bidi_it) |
1242 | { | |
1243 | int prev_level = bidi_it->level_stack[bidi_it->stack_idx].level; | |
1244 | bidi_type_t type = bidi_resolve_weak (bidi_it); | |
1245 | int current_level = bidi_it->level_stack[bidi_it->stack_idx].level; | |
1246 | ||
1247 | if (!(type == STRONG_R | |
1248 | || type == STRONG_L | |
1249 | || type == WEAK_BN | |
1250 | || type == WEAK_EN | |
1251 | || type == WEAK_AN | |
1252 | || type == NEUTRAL_B | |
1253 | || type == NEUTRAL_S | |
1254 | || type == NEUTRAL_WS | |
1255 | || type == NEUTRAL_ON)) | |
1256 | abort (); | |
1257 | ||
1258 | if (bidi_get_category (type) == NEUTRAL | |
1259 | || (type == WEAK_BN && prev_level == current_level)) | |
1260 | { | |
1261 | if (bidi_it->next_for_neutral.type != UNKNOWN_BT) | |
1262 | type = bidi_resolve_neutral_1 (bidi_it->prev_for_neutral.type, | |
1263 | bidi_it->next_for_neutral.type, | |
1264 | current_level); | |
1265 | else | |
1266 | { | |
1267 | /* Arrrgh!! The UAX#9 algorithm is too deeply entrenched in | |
1268 | the assumption of batch-style processing; see clauses W4, | |
1269 | W5, and especially N1, which require to look far forward | |
1270 | (as well as back) in the buffer. May the fleas of a | |
1271 | thousand camels infest the armpits of those who design | |
1272 | supposedly general-purpose algorithms by looking at their | |
1273 | own implementations, and fail to consider other possible | |
1274 | implementations! */ | |
1275 | struct bidi_it saved_it; | |
1276 | bidi_type_t next_type; | |
1277 | ||
1278 | if (bidi_it->scan_dir == -1) | |
1279 | abort (); | |
1280 | ||
1281 | bidi_copy_it (&saved_it, bidi_it); | |
1282 | /* Scan the text forward until we find the first non-neutral | |
1283 | character, and then use that to resolve the neutral we | |
1284 | are dealing with now. We also cache the scanned iterator | |
1285 | states, to salvage some of the effort later. */ | |
1286 | bidi_cache_iterator_state (bidi_it, 0); | |
1287 | do { | |
1288 | /* Record the info about the previous character, so that | |
1289 | it will be cached below with this state. */ | |
89d3374a | 1290 | if (bidi_it->type_after_w1 != WEAK_BN /* W1/Retaining */ |
b7b65b15 EZ |
1291 | && bidi_it->type != WEAK_BN) |
1292 | bidi_remember_char (&bidi_it->prev, bidi_it); | |
1293 | type = bidi_resolve_weak (bidi_it); | |
1294 | /* Paragraph separators have their levels fully resolved | |
1295 | at this point, so cache them as resolved. */ | |
1296 | bidi_cache_iterator_state (bidi_it, type == NEUTRAL_B); | |
1297 | /* FIXME: implement L1 here, by testing for a newline and | |
1298 | resetting the level for any sequence of whitespace | |
1299 | characters adjacent to it. */ | |
1300 | } while (!(type == NEUTRAL_B | |
1301 | || (type != WEAK_BN | |
1302 | && bidi_get_category (type) != NEUTRAL) | |
1303 | /* This is all per level run, so stop when we | |
1304 | reach the end of this level run. */ | |
1305 | || bidi_it->level_stack[bidi_it->stack_idx].level != | |
1306 | current_level)); | |
1307 | ||
1308 | bidi_remember_char (&saved_it.next_for_neutral, bidi_it); | |
1309 | ||
1310 | switch (type) | |
1311 | { | |
1312 | case STRONG_L: | |
1313 | case STRONG_R: | |
1314 | case STRONG_AL: | |
1315 | next_type = type; | |
1316 | break; | |
1317 | case WEAK_EN: | |
1318 | case WEAK_AN: | |
1319 | /* N1: ``European and Arabic numbers are treated as | |
1320 | though they were R.'' */ | |
1321 | next_type = STRONG_R; | |
1322 | saved_it.next_for_neutral.type = STRONG_R; | |
1323 | break; | |
1324 | case WEAK_BN: | |
1325 | if (!bidi_explicit_dir_char (bidi_it->ch)) | |
1326 | abort (); /* can't happen: BNs are skipped */ | |
1327 | /* FALLTHROUGH */ | |
1328 | case NEUTRAL_B: | |
1329 | /* Marched all the way to the end of this level run. | |
1330 | We need to use the eor type, whose information is | |
1331 | stored by bidi_set_sor_type in the prev_for_neutral | |
1332 | member. */ | |
1333 | if (saved_it.type != WEAK_BN | |
89d3374a | 1334 | || bidi_get_category (bidi_it->prev.type_after_w1) == NEUTRAL) |
b7b65b15 EZ |
1335 | { |
1336 | next_type = bidi_it->prev_for_neutral.type; | |
1337 | saved_it.next_for_neutral.type = next_type; | |
2d6e4628 | 1338 | bidi_check_type (next_type); |
b7b65b15 EZ |
1339 | } |
1340 | else | |
1341 | { | |
1342 | /* This is a BN which does not adjoin neutrals. | |
1343 | Leave its type alone. */ | |
1344 | bidi_copy_it (bidi_it, &saved_it); | |
1345 | return bidi_it->type; | |
1346 | } | |
1347 | break; | |
1348 | default: | |
1349 | abort (); | |
1350 | } | |
1351 | type = bidi_resolve_neutral_1 (saved_it.prev_for_neutral.type, | |
1352 | next_type, current_level); | |
1353 | saved_it.type = type; | |
2d6e4628 | 1354 | bidi_check_type (type); |
b7b65b15 EZ |
1355 | bidi_copy_it (bidi_it, &saved_it); |
1356 | } | |
1357 | } | |
1358 | return type; | |
1359 | } | |
1360 | ||
1361 | /* Given an iterator state in BIDI_IT, advance one character position | |
1362 | in the buffer to the next character (in the logical order), resolve | |
1363 | the bidi type of that next character, and return that type. */ | |
fd3998ff | 1364 | static bidi_type_t |
b7b65b15 EZ |
1365 | bidi_type_of_next_char (struct bidi_it *bidi_it) |
1366 | { | |
1367 | bidi_type_t type; | |
1368 | ||
1369 | /* This should always be called during a forward scan. */ | |
1370 | if (bidi_it->scan_dir != 1) | |
1371 | abort (); | |
1372 | ||
1373 | /* Reset the limit until which to ignore BNs if we step out of the | |
1374 | area where we found only empty levels. */ | |
1375 | if ((bidi_it->ignore_bn_limit > 0 | |
1376 | && bidi_it->ignore_bn_limit <= bidi_it->charpos) | |
1377 | || (bidi_it->ignore_bn_limit == -1 | |
1378 | && !bidi_explicit_dir_char (bidi_it->ch))) | |
1379 | bidi_it->ignore_bn_limit = 0; | |
1380 | ||
1381 | type = bidi_resolve_neutral (bidi_it); | |
1382 | ||
1383 | return type; | |
1384 | } | |
1385 | ||
1386 | /* Given an iterator state BIDI_IT, advance one character position in | |
1387 | the buffer to the next character (in the logical order), resolve | |
1388 | the embedding and implicit levels of that next character, and | |
1389 | return the resulting level. */ | |
fd3998ff | 1390 | static int |
b7b65b15 EZ |
1391 | bidi_level_of_next_char (struct bidi_it *bidi_it) |
1392 | { | |
1393 | bidi_type_t type; | |
1394 | int level, prev_level = -1; | |
1395 | struct bidi_saved_info next_for_neutral; | |
1396 | ||
1397 | if (bidi_it->scan_dir == 1) | |
1398 | { | |
1399 | /* There's no sense in trying to advance if we hit end of text. */ | |
1502b819 | 1400 | if (bidi_it->bytepos >= ZV_BYTE) |
b7b65b15 EZ |
1401 | return bidi_it->resolved_level; |
1402 | ||
1403 | /* Record the info about the previous character. */ | |
89d3374a | 1404 | if (bidi_it->type_after_w1 != WEAK_BN /* W1/Retaining */ |
b7b65b15 EZ |
1405 | && bidi_it->type != WEAK_BN) |
1406 | bidi_remember_char (&bidi_it->prev, bidi_it); | |
89d3374a EZ |
1407 | if (bidi_it->type_after_w1 == STRONG_R |
1408 | || bidi_it->type_after_w1 == STRONG_L | |
1409 | || bidi_it->type_after_w1 == STRONG_AL) | |
b7b65b15 EZ |
1410 | bidi_remember_char (&bidi_it->last_strong, bidi_it); |
1411 | /* FIXME: it sounds like we don't need both prev and | |
1412 | prev_for_neutral members, but I'm leaving them both for now. */ | |
1413 | if (bidi_it->type == STRONG_R || bidi_it->type == STRONG_L | |
1414 | || bidi_it->type == WEAK_EN || bidi_it->type == WEAK_AN) | |
1415 | bidi_remember_char (&bidi_it->prev_for_neutral, bidi_it); | |
1416 | ||
1417 | /* If we overstepped the characters used for resolving neutrals | |
1418 | and whitespace, invalidate their info in the iterator. */ | |
1419 | if (bidi_it->charpos >= bidi_it->next_for_neutral.charpos) | |
1420 | bidi_it->next_for_neutral.type = UNKNOWN_BT; | |
1421 | if (bidi_it->next_en_pos >= 0 | |
1422 | && bidi_it->charpos >= bidi_it->next_en_pos) | |
1423 | bidi_it->next_en_pos = -1; | |
1424 | if (bidi_it->next_for_ws.type != UNKNOWN_BT | |
1425 | && bidi_it->charpos >= bidi_it->next_for_ws.charpos) | |
1426 | bidi_it->next_for_ws.type = UNKNOWN_BT; | |
1427 | ||
1428 | /* This must be taken before we fill the iterator with the info | |
1429 | about the next char. If we scan backwards, the iterator | |
1430 | state must be already cached, so there's no need to know the | |
1431 | embedding level of the previous character, since we will be | |
1432 | returning to our caller shortly. */ | |
1433 | prev_level = bidi_it->level_stack[bidi_it->stack_idx].level; | |
1434 | } | |
1435 | next_for_neutral = bidi_it->next_for_neutral; | |
1436 | ||
1437 | /* Perhaps it is already cached. */ | |
1438 | type = bidi_cache_find (bidi_it->charpos + bidi_it->scan_dir, -1, bidi_it); | |
1439 | if (type != UNKNOWN_BT) | |
1440 | { | |
1441 | /* Don't lose the information for resolving neutrals! The | |
1442 | cached states could have been cached before their | |
1443 | next_for_neutral member was computed. If we are on our way | |
1444 | forward, we can simply take the info from the previous | |
1445 | state. */ | |
1446 | if (bidi_it->scan_dir == 1 | |
1447 | && bidi_it->next_for_neutral.type == UNKNOWN_BT) | |
1448 | bidi_it->next_for_neutral = next_for_neutral; | |
1449 | ||
1450 | /* If resolved_level is -1, it means this state was cached | |
1451 | before it was completely resolved, so we cannot return | |
1452 | it. */ | |
1453 | if (bidi_it->resolved_level != -1) | |
1454 | return bidi_it->resolved_level; | |
1455 | } | |
1456 | if (bidi_it->scan_dir == -1) | |
1457 | /* If we are going backwards, the iterator state is already cached | |
1458 | from previous scans, and should be fully resolved. */ | |
1459 | abort (); | |
1460 | ||
1461 | if (type == UNKNOWN_BT) | |
1462 | type = bidi_type_of_next_char (bidi_it); | |
1463 | ||
1464 | if (type == NEUTRAL_B) | |
1465 | return bidi_it->resolved_level; | |
1466 | ||
1467 | level = bidi_it->level_stack[bidi_it->stack_idx].level; | |
1468 | if ((bidi_get_category (type) == NEUTRAL /* && type != NEUTRAL_B */) | |
1469 | || (type == WEAK_BN && prev_level == level)) | |
1470 | { | |
1471 | if (bidi_it->next_for_neutral.type == UNKNOWN_BT) | |
1472 | abort (); | |
1473 | ||
1474 | /* If the cached state shows a neutral character, it was not | |
1475 | resolved by bidi_resolve_neutral, so do it now. */ | |
1476 | type = bidi_resolve_neutral_1 (bidi_it->prev_for_neutral.type, | |
1477 | bidi_it->next_for_neutral.type, | |
1478 | level); | |
1479 | } | |
1480 | ||
1481 | if (!(type == STRONG_R | |
1482 | || type == STRONG_L | |
1483 | || type == WEAK_BN | |
1484 | || type == WEAK_EN | |
1485 | || type == WEAK_AN)) | |
1486 | abort (); | |
1487 | bidi_it->type = type; | |
2d6e4628 | 1488 | bidi_check_type (bidi_it->type); |
b7b65b15 EZ |
1489 | |
1490 | /* For L1 below, we need to know, for each WS character, whether | |
1491 | it belongs to a sequence of WS characters preceeding a newline | |
1492 | or a TAB or a paragraph separator. */ | |
89d3374a | 1493 | if (bidi_it->orig_type == NEUTRAL_WS |
b7b65b15 EZ |
1494 | && bidi_it->next_for_ws.type == UNKNOWN_BT) |
1495 | { | |
1496 | int ch; | |
1497 | int clen = bidi_it->ch_len; | |
6bff6497 EZ |
1498 | EMACS_INT bpos = bidi_it->bytepos; |
1499 | EMACS_INT cpos = bidi_it->charpos; | |
b7b65b15 EZ |
1500 | bidi_type_t chtype; |
1501 | ||
1502 | do { | |
1503 | /*_fetch_multibyte_char_len = 1;*/ | |
e7402cb2 | 1504 | ch = bpos + clen >= ZV_BYTE ? BIDI_EOB : FETCH_CHAR (bpos + clen); |
b7b65b15 EZ |
1505 | bpos += clen; |
1506 | cpos++; | |
e7402cb2 EZ |
1507 | clen = (ch == BIDI_EOB ? 1 : CHAR_BYTES (ch)); |
1508 | if (ch == '\n' || ch == BIDI_EOB /* || ch == LINESEP_CHAR */) | |
b7b65b15 EZ |
1509 | chtype = NEUTRAL_B; |
1510 | else | |
6bff6497 | 1511 | chtype = bidi_get_type (ch, NEUTRAL_DIR); |
b7b65b15 EZ |
1512 | } while (chtype == NEUTRAL_WS || chtype == WEAK_BN |
1513 | || bidi_explicit_dir_char (ch)); /* L1/Retaining */ | |
1514 | bidi_it->next_for_ws.type = chtype; | |
2d6e4628 | 1515 | bidi_check_type (bidi_it->next_for_ws.type); |
b7b65b15 EZ |
1516 | bidi_it->next_for_ws.charpos = cpos; |
1517 | bidi_it->next_for_ws.bytepos = bpos; | |
1518 | } | |
1519 | ||
1520 | /* Resolve implicit levels, with a twist: PDFs get the embedding | |
1521 | level of the enbedding they terminate. See below for the | |
1522 | reason. */ | |
89d3374a | 1523 | if (bidi_it->orig_type == PDF |
b7b65b15 EZ |
1524 | /* Don't do this if this formatting code didn't change the |
1525 | embedding level due to invalid or empty embeddings. */ | |
1526 | && prev_level != level) | |
1527 | { | |
1528 | /* Don't look in UAX#9 for the reason for this: it's our own | |
1529 | private quirk. The reason is that we want the formatting | |
1530 | codes to be delivered so that they bracket the text of their | |
1531 | embedding. For example, given the text | |
1532 | ||
1533 | {RLO}teST{PDF} | |
1534 | ||
1535 | we want it to be displayed as | |
1536 | ||
1537 | {RLO}STet{PDF} | |
1538 | ||
1539 | not as | |
1540 | ||
1541 | STet{RLO}{PDF} | |
1542 | ||
1543 | which will result because we bump up the embedding level as | |
1544 | soon as we see the RLO and pop it as soon as we see the PDF, | |
1545 | so RLO itself has the same embedding level as "teST", and | |
1546 | thus would be normally delivered last, just before the PDF. | |
1547 | The switch below fiddles with the level of PDF so that this | |
1548 | ugly side effect does not happen. | |
1549 | ||
1550 | (This is, of course, only important if the formatting codes | |
e7402cb2 EZ |
1551 | are actually displayed, but Emacs does need to display them |
1552 | if the user wants to.) */ | |
b7b65b15 EZ |
1553 | level = prev_level; |
1554 | } | |
89d3374a EZ |
1555 | else if (bidi_it->orig_type == NEUTRAL_B /* L1 */ |
1556 | || bidi_it->orig_type == NEUTRAL_S | |
e7402cb2 EZ |
1557 | || bidi_it->ch == '\n' || bidi_it->ch == BIDI_EOB |
1558 | /* || bidi_it->ch == LINESEP_CHAR */ | |
89d3374a | 1559 | || (bidi_it->orig_type == NEUTRAL_WS |
b7b65b15 EZ |
1560 | && (bidi_it->next_for_ws.type == NEUTRAL_B |
1561 | || bidi_it->next_for_ws.type == NEUTRAL_S))) | |
1562 | level = bidi_it->level_stack[0].level; | |
1563 | else if ((level & 1) == 0) /* I1 */ | |
1564 | { | |
1565 | if (type == STRONG_R) | |
1566 | level++; | |
1567 | else if (type == WEAK_EN || type == WEAK_AN) | |
1568 | level += 2; | |
1569 | } | |
1570 | else /* I2 */ | |
1571 | { | |
1572 | if (type == STRONG_L || type == WEAK_EN || type == WEAK_AN) | |
1573 | level++; | |
1574 | } | |
1575 | ||
1576 | bidi_it->resolved_level = level; | |
1577 | return level; | |
1578 | } | |
1579 | ||
1580 | /* Move to the other edge of a level given by LEVEL. If END_FLAG is | |
1581 | non-zero, we are at the end of a level, and we need to prepare to | |
1582 | resume the scan of the lower level. | |
1583 | ||
1584 | If this level's other edge is cached, we simply jump to it, filling | |
1585 | the iterator structure with the iterator state on the other edge. | |
1586 | Otherwise, we walk the buffer until we come back to the same level | |
1587 | as LEVEL. | |
1588 | ||
1589 | Note: we are not talking here about a ``level run'' in the UAX#9 | |
1590 | sense of the term, but rather about a ``level'' which includes | |
1591 | all the levels higher than it. In other words, given the levels | |
1592 | like this: | |
1593 | ||
1594 | 11111112222222333333334443343222222111111112223322111 | |
1595 | A B C | |
1596 | ||
1597 | and assuming we are at point A scanning left to right, this | |
1598 | function moves to point C, whereas the UAX#9 ``level 2 run'' ends | |
1599 | at point B. */ | |
1600 | static void | |
1601 | bidi_find_other_level_edge (struct bidi_it *bidi_it, int level, int end_flag) | |
1602 | { | |
1603 | int dir = end_flag ? -bidi_it->scan_dir : bidi_it->scan_dir; | |
1604 | int idx; | |
1605 | ||
1606 | /* Try the cache first. */ | |
1607 | if ((idx = bidi_cache_find_level_change (level, dir, end_flag)) >= 0) | |
1608 | bidi_cache_fetch_state (idx, bidi_it); | |
1609 | else | |
1610 | { | |
1611 | int new_level; | |
1612 | ||
1613 | if (end_flag) | |
1614 | abort (); /* if we are at end of level, its edges must be cached */ | |
1615 | ||
1616 | bidi_cache_iterator_state (bidi_it, 1); | |
1617 | do { | |
1618 | new_level = bidi_level_of_next_char (bidi_it); | |
1619 | bidi_cache_iterator_state (bidi_it, 1); | |
1620 | } while (new_level >= level); | |
1621 | } | |
1622 | } | |
1623 | ||
1624 | void | |
4b292a22 | 1625 | bidi_move_to_visually_next (struct bidi_it *bidi_it) |
b7b65b15 EZ |
1626 | { |
1627 | int old_level, new_level, next_level; | |
9c82e145 | 1628 | struct bidi_it sentinel; |
b7b65b15 EZ |
1629 | |
1630 | if (bidi_it->scan_dir == 0) | |
1631 | { | |
1632 | bidi_it->scan_dir = 1; /* default to logical order */ | |
1633 | } | |
1634 | ||
be39f003 EZ |
1635 | /* If we just passed a newline, initialize for the next line. */ |
1636 | if (!bidi_it->first_elt && bidi_it->orig_type == NEUTRAL_B) | |
1637 | bidi_line_init (bidi_it); | |
1638 | ||
6dcfd253 EZ |
1639 | /* Prepare the sentinel iterator state, and cache it. When we bump |
1640 | into it, scanning backwards, we'll know that the last non-base | |
1641 | level is exhausted. */ | |
b7b65b15 | 1642 | if (bidi_cache_idx == 0) |
9c82e145 EZ |
1643 | { |
1644 | bidi_copy_it (&sentinel, bidi_it); | |
1645 | if (bidi_it->first_elt) | |
1646 | { | |
1647 | sentinel.charpos--; /* cached charpos needs to be monotonic */ | |
1648 | sentinel.bytepos--; | |
1649 | sentinel.ch = '\n'; /* doesn't matter, but why not? */ | |
1650 | sentinel.ch_len = 1; | |
1651 | } | |
6dcfd253 | 1652 | bidi_cache_iterator_state (&sentinel, 1); |
9c82e145 | 1653 | } |
b7b65b15 EZ |
1654 | |
1655 | old_level = bidi_it->resolved_level; | |
1656 | new_level = bidi_level_of_next_char (bidi_it); | |
b7b65b15 EZ |
1657 | |
1658 | /* Reordering of resolved levels (clause L2) is implemented by | |
1659 | jumping to the other edge of the level and flipping direction of | |
c0546589 | 1660 | scanning the text whenever we find a level change. */ |
b7b65b15 EZ |
1661 | if (new_level != old_level) |
1662 | { | |
1663 | int ascending = new_level > old_level; | |
1664 | int level_to_search = ascending ? old_level + 1 : old_level; | |
1665 | int incr = ascending ? 1 : -1; | |
1666 | int expected_next_level = old_level + incr; | |
1667 | ||
b7b65b15 EZ |
1668 | /* Jump (or walk) to the other edge of this level. */ |
1669 | bidi_find_other_level_edge (bidi_it, level_to_search, !ascending); | |
1670 | /* Switch scan direction and peek at the next character in the | |
1671 | new direction. */ | |
1672 | bidi_it->scan_dir = -bidi_it->scan_dir; | |
1673 | ||
1674 | /* The following loop handles the case where the resolved level | |
1675 | jumps by more than one. This is typical for numbers inside a | |
1676 | run of text with left-to-right embedding direction, but can | |
1677 | also happen in other situations. In those cases the decision | |
1678 | where to continue after a level change, and in what direction, | |
1679 | is tricky. For example, given a text like below: | |
1680 | ||
1681 | abcdefgh | |
1682 | 11336622 | |
1683 | ||
1684 | (where the numbers below the text show the resolved levels), | |
1685 | the result of reordering according to UAX#9 should be this: | |
1686 | ||
1687 | efdcghba | |
1688 | ||
1689 | This is implemented by the loop below which flips direction | |
1690 | and jumps to the other edge of the level each time it finds | |
1691 | the new level not to be the expected one. The expected level | |
1692 | is always one more or one less than the previous one. */ | |
1693 | next_level = bidi_peek_at_next_level (bidi_it); | |
1694 | while (next_level != expected_next_level) | |
1695 | { | |
1696 | expected_next_level += incr; | |
1697 | level_to_search += incr; | |
1698 | bidi_find_other_level_edge (bidi_it, level_to_search, !ascending); | |
1699 | bidi_it->scan_dir = -bidi_it->scan_dir; | |
1700 | next_level = bidi_peek_at_next_level (bidi_it); | |
1701 | } | |
1702 | ||
1703 | /* Finally, deliver the next character in the new direction. */ | |
1704 | next_level = bidi_level_of_next_char (bidi_it); | |
1705 | } | |
1706 | ||
b44d9321 EZ |
1707 | /* Take note when we have just processed the newline that precedes |
1708 | the end of the paragraph. The next time we are about to be | |
1709 | called, set_iterator_to_next will automatically reinit the | |
1710 | paragraph direction, if needed. We do this at the newline before | |
1711 | the paragraph separator, because the next character might not be | |
1712 | the first character of the next paragraph, due to the bidi | |
c0546589 EZ |
1713 | reordering, whereas we _must_ know the paragraph base direction |
1714 | _before_ we process the paragraph's text, since the base | |
1715 | direction affects the reordering. */ | |
6bff6497 | 1716 | if (bidi_it->scan_dir == 1 |
be39f003 EZ |
1717 | && bidi_it->orig_type == NEUTRAL_B |
1718 | && bidi_it->bytepos < ZV_BYTE) | |
1719 | { | |
1720 | EMACS_INT sep_len = | |
1721 | bidi_at_paragraph_end (bidi_it->charpos + 1, | |
1722 | bidi_it->bytepos + bidi_it->ch_len); | |
1723 | if (sep_len >= 0) | |
1724 | { | |
1725 | bidi_it->new_paragraph = 1; | |
b44d9321 EZ |
1726 | /* Record the buffer position of the last character of the |
1727 | paragraph separator. */ | |
1728 | bidi_it->separator_limit = bidi_it->charpos + 1 + sep_len; | |
be39f003 EZ |
1729 | } |
1730 | } | |
6bff6497 | 1731 | |
b7b65b15 EZ |
1732 | if (bidi_it->scan_dir == 1 && bidi_cache_idx) |
1733 | { | |
1734 | /* If we are at paragraph's base embedding level and beyond the | |
1735 | last cached position, the cache's job is done and we can | |
1736 | discard it. */ | |
1737 | if (bidi_it->resolved_level == bidi_it->level_stack[0].level | |
1738 | && bidi_it->charpos > bidi_cache[bidi_cache_idx - 1].charpos) | |
1739 | bidi_cache_reset (); | |
1740 | /* But as long as we are caching during forward scan, we must | |
1741 | cache each state, or else the cache integrity will be | |
1742 | compromised: it assumes cached states correspond to buffer | |
1743 | positions 1:1. */ | |
1744 | else | |
1745 | bidi_cache_iterator_state (bidi_it, 1); | |
1746 | } | |
1747 | } | |
1748 | ||
1749 | /* This is meant to be called from within the debugger, whenever you | |
1750 | wish to examine the cache contents. */ | |
1751 | void | |
1752 | bidi_dump_cached_states (void) | |
1753 | { | |
1754 | int i; | |
1755 | int ndigits = 1; | |
1756 | ||
1757 | if (bidi_cache_idx == 0) | |
1758 | { | |
1759 | fprintf (stderr, "The cache is empty.\n"); | |
1760 | return; | |
1761 | } | |
1762 | fprintf (stderr, "Total of %d state%s in cache:\n", | |
1763 | bidi_cache_idx, bidi_cache_idx == 1 ? "" : "s"); | |
1764 | ||
1765 | for (i = bidi_cache[bidi_cache_idx - 1].charpos; i > 0; i /= 10) | |
1766 | ndigits++; | |
1767 | fputs ("ch ", stderr); | |
1768 | for (i = 0; i < bidi_cache_idx; i++) | |
1769 | fprintf (stderr, "%*c", ndigits, bidi_cache[i].ch); | |
1770 | fputs ("\n", stderr); | |
1771 | fputs ("lvl ", stderr); | |
1772 | for (i = 0; i < bidi_cache_idx; i++) | |
1773 | fprintf (stderr, "%*d", ndigits, bidi_cache[i].resolved_level); | |
1774 | fputs ("\n", stderr); | |
1775 | fputs ("pos ", stderr); | |
1776 | for (i = 0; i < bidi_cache_idx; i++) | |
3d608a86 | 1777 | fprintf (stderr, "%*ld", ndigits, (long)bidi_cache[i].charpos); |
b7b65b15 EZ |
1778 | fputs ("\n", stderr); |
1779 | } |