Commit | Line | Data |
---|---|---|
cbb09f04 | 1 | /* Low-level bidirectional buffer/string-scanning functions for GNU Emacs. |
73b0cd50 | 2 | Copyright (C) 2000-2001, 2004-2005, 2009-2011 |
b118e65f | 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, | |
cbb09f04 | 23 | (UBA) as per UAX#9, a part of the Unicode Standard. |
b7b65b15 EZ |
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 | ||
95e1afc6 | 38 | Two other entry points are bidi_paragraph_init and |
940afb59 EZ |
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 | ||
95e1afc6 EZ |
43 | A few auxiliary entry points are used to initialize the bidi |
44 | iterator for iterating an object (buffer or string), push and pop | |
45 | the bidi iterator state, and save and restore the state of the bidi | |
46 | cache. | |
47 | ||
89d3374a EZ |
48 | If you want to understand the code, you will have to read it |
49 | together with the relevant portions of UAX#9. The comments include | |
50 | references to UAX#9 rules, for that very reason. | |
51 | ||
b7b65b15 EZ |
52 | A note about references to UAX#9 rules: if the reference says |
53 | something like "X9/Retaining", it means that you need to refer to | |
54 | rule X9 and to its modifications decribed in the "Implementation | |
55 | Notes" section of UAX#9, under "Retaining Format Codes". */ | |
56 | ||
b7b65b15 | 57 | #include <config.h> |
b7b65b15 | 58 | #include <stdio.h> |
29e3d8d1 EZ |
59 | #include <setjmp.h> |
60 | ||
b7b65b15 EZ |
61 | #include "lisp.h" |
62 | #include "buffer.h" | |
63 | #include "character.h" | |
64 | #include "dispextern.h" | |
65 | ||
66 | static int bidi_initialized = 0; | |
67 | ||
cbc4fd20 | 68 | static Lisp_Object bidi_type_table, bidi_mirror_table; |
b7b65b15 EZ |
69 | |
70 | #define LRM_CHAR 0x200E | |
71 | #define RLM_CHAR 0x200F | |
b7b65b15 | 72 | #define BIDI_EOB -1 |
b7b65b15 | 73 | |
b7b65b15 EZ |
74 | /* Data type for describing the bidirectional character categories. */ |
75 | typedef enum { | |
76 | UNKNOWN_BC, | |
77 | NEUTRAL, | |
78 | WEAK, | |
79 | STRONG | |
80 | } bidi_category_t; | |
81 | ||
e78aecca | 82 | extern int bidi_ignore_explicit_marks_for_paragraph_level EXTERNALLY_VISIBLE; |
b7b65b15 EZ |
83 | int bidi_ignore_explicit_marks_for_paragraph_level = 1; |
84 | ||
372b7a95 | 85 | static Lisp_Object paragraph_start_re, paragraph_separate_re; |
6bff6497 | 86 | static Lisp_Object Qparagraph_start, Qparagraph_separate; |
b7b65b15 | 87 | |
cbb09f04 EZ |
88 | \f |
89 | /*********************************************************************** | |
90 | Utilities | |
91 | ***********************************************************************/ | |
b7b65b15 | 92 | |
6bff6497 EZ |
93 | /* Return the bidi type of a character CH, subject to the current |
94 | directional OVERRIDE. */ | |
55d4c1b2 | 95 | static inline bidi_type_t |
6bff6497 | 96 | bidi_get_type (int ch, bidi_dir_t override) |
b7b65b15 | 97 | { |
6bff6497 EZ |
98 | bidi_type_t default_type; |
99 | ||
e7402cb2 EZ |
100 | if (ch == BIDI_EOB) |
101 | return NEUTRAL_B; | |
e342a24d EZ |
102 | if (ch < 0 || ch > MAX_CHAR) |
103 | abort (); | |
6bff6497 EZ |
104 | |
105 | default_type = (bidi_type_t) XINT (CHAR_TABLE_REF (bidi_type_table, ch)); | |
106 | ||
107 | if (override == NEUTRAL_DIR) | |
108 | return default_type; | |
109 | ||
110 | switch (default_type) | |
111 | { | |
112 | /* Although UAX#9 does not tell, it doesn't make sense to | |
113 | override NEUTRAL_B and LRM/RLM characters. */ | |
114 | case NEUTRAL_B: | |
115 | case LRE: | |
116 | case LRO: | |
117 | case RLE: | |
118 | case RLO: | |
119 | case PDF: | |
120 | return default_type; | |
121 | default: | |
122 | switch (ch) | |
123 | { | |
124 | case LRM_CHAR: | |
125 | case RLM_CHAR: | |
126 | return default_type; | |
127 | default: | |
128 | if (override == L2R) /* X6 */ | |
129 | return STRONG_L; | |
130 | else if (override == R2L) | |
131 | return STRONG_R; | |
132 | else | |
133 | abort (); /* can't happen: handled above */ | |
134 | } | |
135 | } | |
b7b65b15 EZ |
136 | } |
137 | ||
7d3b3862 | 138 | static void |
2d6e4628 EZ |
139 | bidi_check_type (bidi_type_t type) |
140 | { | |
141 | if (type < UNKNOWN_BT || type > NEUTRAL_ON) | |
142 | abort (); | |
143 | } | |
144 | ||
b7b65b15 | 145 | /* Given a bidi TYPE of a character, return its category. */ |
55d4c1b2 | 146 | static inline bidi_category_t |
b7b65b15 EZ |
147 | bidi_get_category (bidi_type_t type) |
148 | { | |
149 | switch (type) | |
150 | { | |
151 | case UNKNOWN_BT: | |
152 | return UNKNOWN_BC; | |
153 | case STRONG_L: | |
154 | case STRONG_R: | |
155 | case STRONG_AL: | |
156 | case LRE: | |
157 | case LRO: | |
158 | case RLE: | |
159 | case RLO: | |
160 | return STRONG; | |
161 | case PDF: /* ??? really?? */ | |
162 | case WEAK_EN: | |
163 | case WEAK_ES: | |
164 | case WEAK_ET: | |
165 | case WEAK_AN: | |
166 | case WEAK_CS: | |
167 | case WEAK_NSM: | |
168 | case WEAK_BN: | |
169 | return WEAK; | |
170 | case NEUTRAL_B: | |
171 | case NEUTRAL_S: | |
172 | case NEUTRAL_WS: | |
173 | case NEUTRAL_ON: | |
174 | return NEUTRAL; | |
175 | default: | |
176 | abort (); | |
177 | } | |
178 | } | |
179 | ||
a6e8d97c EZ |
180 | /* Return the mirrored character of C, if it has one. If C has no |
181 | mirrored counterpart, return C. | |
182 | Note: The conditions in UAX#9 clause L4 regarding the surrounding | |
183 | context must be tested by the caller. */ | |
b7b65b15 EZ |
184 | int |
185 | bidi_mirror_char (int c) | |
186 | { | |
cbc4fd20 EZ |
187 | Lisp_Object val; |
188 | ||
189 | if (c == BIDI_EOB) | |
190 | return c; | |
191 | if (c < 0 || c > MAX_CHAR) | |
192 | abort (); | |
b7b65b15 | 193 | |
cbc4fd20 EZ |
194 | val = CHAR_TABLE_REF (bidi_mirror_table, c); |
195 | if (INTEGERP (val)) | |
b7b65b15 | 196 | { |
cbc4fd20 | 197 | int v = XINT (val); |
b7b65b15 | 198 | |
cbc4fd20 EZ |
199 | if (v < 0 || v > MAX_CHAR) |
200 | abort (); | |
201 | ||
202 | return v; | |
b7b65b15 | 203 | } |
cbc4fd20 | 204 | |
b7b65b15 EZ |
205 | return c; |
206 | } | |
207 | ||
cbb09f04 EZ |
208 | /* Determine the start-of-run (sor) directional type given the two |
209 | embedding levels on either side of the run boundary. Also, update | |
210 | the saved info about previously seen characters, since that info is | |
211 | generally valid for a single level run. */ | |
58b9f433 | 212 | static inline void |
cbb09f04 EZ |
213 | bidi_set_sor_type (struct bidi_it *bidi_it, int level_before, int level_after) |
214 | { | |
215 | int higher_level = level_before > level_after ? level_before : level_after; | |
216 | ||
217 | /* The prev_was_pdf gork is required for when we have several PDFs | |
218 | in a row. In that case, we want to compute the sor type for the | |
219 | next level run only once: when we see the first PDF. That's | |
220 | because the sor type depends only on the higher of the two levels | |
221 | that we find on the two sides of the level boundary (see UAX#9, | |
222 | clause X10), and so we don't need to know the final embedding | |
223 | level to which we descend after processing all the PDFs. */ | |
224 | if (!bidi_it->prev_was_pdf || level_before < level_after) | |
225 | /* FIXME: should the default sor direction be user selectable? */ | |
226 | bidi_it->sor = (higher_level & 1) != 0 ? R2L : L2R; | |
227 | if (level_before > level_after) | |
228 | bidi_it->prev_was_pdf = 1; | |
229 | ||
230 | bidi_it->prev.type = UNKNOWN_BT; | |
231 | bidi_it->last_strong.type = bidi_it->last_strong.type_after_w1 = | |
232 | bidi_it->last_strong.orig_type = UNKNOWN_BT; | |
233 | bidi_it->prev_for_neutral.type = bidi_it->sor == R2L ? STRONG_R : STRONG_L; | |
234 | bidi_it->prev_for_neutral.charpos = bidi_it->charpos; | |
235 | bidi_it->prev_for_neutral.bytepos = bidi_it->bytepos; | |
236 | bidi_it->next_for_neutral.type = bidi_it->next_for_neutral.type_after_w1 = | |
237 | bidi_it->next_for_neutral.orig_type = UNKNOWN_BT; | |
238 | bidi_it->ignore_bn_limit = -1; /* meaning it's unknown */ | |
239 | } | |
240 | ||
241 | /* Push the current embedding level and override status; reset the | |
242 | current level to LEVEL and the current override status to OVERRIDE. */ | |
58b9f433 | 243 | static inline void |
cbb09f04 EZ |
244 | bidi_push_embedding_level (struct bidi_it *bidi_it, |
245 | int level, bidi_dir_t override) | |
246 | { | |
247 | bidi_it->stack_idx++; | |
7e2ad32c | 248 | xassert (bidi_it->stack_idx < BIDI_MAXLEVEL); |
cbb09f04 EZ |
249 | bidi_it->level_stack[bidi_it->stack_idx].level = level; |
250 | bidi_it->level_stack[bidi_it->stack_idx].override = override; | |
251 | } | |
252 | ||
253 | /* Pop the embedding level and directional override status from the | |
254 | stack, and return the new level. */ | |
58b9f433 | 255 | static inline int |
cbb09f04 EZ |
256 | bidi_pop_embedding_level (struct bidi_it *bidi_it) |
257 | { | |
258 | /* UAX#9 says to ignore invalid PDFs. */ | |
259 | if (bidi_it->stack_idx > 0) | |
260 | bidi_it->stack_idx--; | |
261 | return bidi_it->level_stack[bidi_it->stack_idx].level; | |
262 | } | |
263 | ||
264 | /* Record in SAVED_INFO the information about the current character. */ | |
58b9f433 | 265 | static inline void |
cbb09f04 EZ |
266 | bidi_remember_char (struct bidi_saved_info *saved_info, |
267 | struct bidi_it *bidi_it) | |
268 | { | |
269 | saved_info->charpos = bidi_it->charpos; | |
270 | saved_info->bytepos = bidi_it->bytepos; | |
271 | saved_info->type = bidi_it->type; | |
272 | bidi_check_type (bidi_it->type); | |
273 | saved_info->type_after_w1 = bidi_it->type_after_w1; | |
274 | bidi_check_type (bidi_it->type_after_w1); | |
275 | saved_info->orig_type = bidi_it->orig_type; | |
276 | bidi_check_type (bidi_it->orig_type); | |
277 | } | |
278 | ||
b7b65b15 EZ |
279 | /* Copy the bidi iterator from FROM to TO. To save cycles, this only |
280 | copies the part of the level stack that is actually in use. */ | |
55d4c1b2 | 281 | static inline void |
b7b65b15 EZ |
282 | bidi_copy_it (struct bidi_it *to, struct bidi_it *from) |
283 | { | |
284 | int i; | |
285 | ||
e69a9370 | 286 | /* Copy everything except the level stack and beyond. */ |
182ce2d2 | 287 | memcpy (to, from, offsetof (struct bidi_it, level_stack[0])); |
b7b65b15 EZ |
288 | |
289 | /* Copy the active part of the level stack. */ | |
290 | to->level_stack[0] = from->level_stack[0]; /* level zero is always in use */ | |
291 | for (i = 1; i <= from->stack_idx; i++) | |
292 | to->level_stack[i] = from->level_stack[i]; | |
293 | } | |
294 | ||
cbb09f04 EZ |
295 | \f |
296 | /*********************************************************************** | |
297 | Caching the bidi iterator states | |
298 | ***********************************************************************/ | |
b7b65b15 | 299 | |
2fe72643 EZ |
300 | #define BIDI_CACHE_CHUNK 200 |
301 | static struct bidi_it *bidi_cache; | |
39e378da | 302 | static ptrdiff_t bidi_cache_size = 0; |
ad6042bb | 303 | enum { elsz = sizeof (struct bidi_it) }; |
39e378da PE |
304 | static ptrdiff_t bidi_cache_idx; /* next unused cache slot */ |
305 | static ptrdiff_t bidi_cache_last_idx; /* slot of last cache hit */ | |
306 | static ptrdiff_t bidi_cache_start = 0; /* start of cache for this | |
87e67904 EZ |
307 | "stack" level */ |
308 | ||
309 | /* Reset the cache state to the empty state. We only reset the part | |
310 | of the cache relevant to iteration of the current object. Previous | |
311 | objects, which are pushed on the display iterator's stack, are left | |
312 | intact. This is called when the cached information is no more | |
313 | useful for the current iteration, e.g. when we were reseated to a | |
314 | new position on the same object. */ | |
55d4c1b2 | 315 | static inline void |
b7b65b15 EZ |
316 | bidi_cache_reset (void) |
317 | { | |
87e67904 | 318 | bidi_cache_idx = bidi_cache_start; |
b7b65b15 EZ |
319 | bidi_cache_last_idx = -1; |
320 | } | |
321 | ||
87e67904 EZ |
322 | /* Shrink the cache to its minimal size. Called when we init the bidi |
323 | iterator for reordering a buffer or a string that does not come | |
324 | from display properties, because that means all the previously | |
325 | cached info is of no further use. */ | |
55d4c1b2 | 326 | static inline void |
2fe72643 EZ |
327 | bidi_cache_shrink (void) |
328 | { | |
329 | if (bidi_cache_size > BIDI_CACHE_CHUNK) | |
330 | { | |
0416466c EZ |
331 | bidi_cache_size = BIDI_CACHE_CHUNK; |
332 | bidi_cache = | |
333 | (struct bidi_it *) xrealloc (bidi_cache, bidi_cache_size * elsz); | |
2fe72643 EZ |
334 | } |
335 | bidi_cache_reset (); | |
336 | } | |
337 | ||
55d4c1b2 | 338 | static inline void |
39e378da | 339 | bidi_cache_fetch_state (ptrdiff_t idx, struct bidi_it *bidi_it) |
b7b65b15 EZ |
340 | { |
341 | int current_scan_dir = bidi_it->scan_dir; | |
342 | ||
87e67904 | 343 | if (idx < bidi_cache_start || idx >= bidi_cache_idx) |
b7b65b15 EZ |
344 | abort (); |
345 | ||
346 | bidi_copy_it (bidi_it, &bidi_cache[idx]); | |
347 | bidi_it->scan_dir = current_scan_dir; | |
348 | bidi_cache_last_idx = idx; | |
349 | } | |
350 | ||
351 | /* Find a cached state with a given CHARPOS and resolved embedding | |
352 | level less or equal to LEVEL. if LEVEL is -1, disregard the | |
353 | resolved levels in cached states. DIR, if non-zero, means search | |
354 | in that direction from the last cache hit. */ | |
39e378da | 355 | static inline ptrdiff_t |
61bfec98 | 356 | bidi_cache_search (EMACS_INT charpos, int level, int dir) |
b7b65b15 | 357 | { |
39e378da | 358 | ptrdiff_t i, i_start; |
b7b65b15 | 359 | |
6eec7596 | 360 | if (bidi_cache_idx > bidi_cache_start) |
b7b65b15 | 361 | { |
6eec7596 EZ |
362 | if (bidi_cache_last_idx == -1) |
363 | bidi_cache_last_idx = bidi_cache_idx - 1; | |
b7b65b15 | 364 | if (charpos < bidi_cache[bidi_cache_last_idx].charpos) |
182ce2d2 EZ |
365 | { |
366 | dir = -1; | |
367 | i_start = bidi_cache_last_idx - 1; | |
368 | } | |
369 | else if (charpos > (bidi_cache[bidi_cache_last_idx].charpos | |
370 | + bidi_cache[bidi_cache_last_idx].nchars - 1)) | |
371 | { | |
372 | dir = 1; | |
373 | i_start = bidi_cache_last_idx + 1; | |
374 | } | |
375 | else if (dir) | |
b7b65b15 EZ |
376 | i_start = bidi_cache_last_idx; |
377 | else | |
378 | { | |
379 | dir = -1; | |
380 | i_start = bidi_cache_idx - 1; | |
381 | } | |
382 | ||
383 | if (dir < 0) | |
384 | { | |
385 | /* Linear search for now; FIXME! */ | |
87e67904 | 386 | for (i = i_start; i >= bidi_cache_start; i--) |
182ce2d2 EZ |
387 | if (bidi_cache[i].charpos <= charpos |
388 | && charpos < bidi_cache[i].charpos + bidi_cache[i].nchars | |
b7b65b15 EZ |
389 | && (level == -1 || bidi_cache[i].resolved_level <= level)) |
390 | return i; | |
391 | } | |
392 | else | |
393 | { | |
394 | for (i = i_start; i < bidi_cache_idx; i++) | |
182ce2d2 EZ |
395 | if (bidi_cache[i].charpos <= charpos |
396 | && charpos < bidi_cache[i].charpos + bidi_cache[i].nchars | |
b7b65b15 EZ |
397 | && (level == -1 || bidi_cache[i].resolved_level <= level)) |
398 | return i; | |
399 | } | |
400 | } | |
401 | ||
402 | return -1; | |
403 | } | |
404 | ||
405 | /* Find a cached state where the resolved level changes to a value | |
406 | that is lower than LEVEL, and return its cache slot index. DIR is | |
407 | the direction to search, starting with the last used cache slot. | |
87e67904 EZ |
408 | If DIR is zero, we search backwards from the last occupied cache |
409 | slot. BEFORE, if non-zero, means return the index of the slot that | |
410 | is ``before'' the level change in the search direction. That is, | |
b7b65b15 EZ |
411 | given the cached levels like this: |
412 | ||
413 | 1122333442211 | |
414 | AB C | |
415 | ||
416 | and assuming we are at the position cached at the slot marked with | |
417 | C, searching backwards (DIR = -1) for LEVEL = 2 will return the | |
418 | index of slot B or A, depending whether BEFORE is, respectively, | |
419 | non-zero or zero. */ | |
39e378da | 420 | static ptrdiff_t |
b7b65b15 EZ |
421 | bidi_cache_find_level_change (int level, int dir, int before) |
422 | { | |
423 | if (bidi_cache_idx) | |
424 | { | |
39e378da | 425 | ptrdiff_t i = dir ? bidi_cache_last_idx : bidi_cache_idx - 1; |
b7b65b15 EZ |
426 | int incr = before ? 1 : 0; |
427 | ||
6eec7596 EZ |
428 | xassert (!dir || bidi_cache_last_idx >= 0); |
429 | ||
b7b65b15 EZ |
430 | if (!dir) |
431 | dir = -1; | |
432 | else if (!incr) | |
433 | i += dir; | |
434 | ||
435 | if (dir < 0) | |
436 | { | |
87e67904 | 437 | while (i >= bidi_cache_start + incr) |
b7b65b15 EZ |
438 | { |
439 | if (bidi_cache[i - incr].resolved_level >= 0 | |
440 | && bidi_cache[i - incr].resolved_level < level) | |
441 | return i; | |
442 | i--; | |
443 | } | |
444 | } | |
445 | else | |
446 | { | |
447 | while (i < bidi_cache_idx - incr) | |
448 | { | |
449 | if (bidi_cache[i + incr].resolved_level >= 0 | |
450 | && bidi_cache[i + incr].resolved_level < level) | |
451 | return i; | |
452 | i++; | |
453 | } | |
454 | } | |
455 | } | |
456 | ||
457 | return -1; | |
458 | } | |
459 | ||
58b9f433 | 460 | static inline void |
39e378da | 461 | bidi_cache_ensure_space (ptrdiff_t idx) |
58b9f433 EZ |
462 | { |
463 | /* Enlarge the cache as needed. */ | |
464 | if (idx >= bidi_cache_size) | |
465 | { | |
39e378da | 466 | ptrdiff_t new_size; |
f0eb61e9 PE |
467 | |
468 | /* The bidi cache cannot be larger than the largest Lisp string | |
469 | or buffer. */ | |
470 | ptrdiff_t string_or_buffer_bound = | |
471 | max (BUF_BYTES_MAX, STRING_BYTES_BOUND); | |
472 | ||
473 | /* Also, it cannot be larger than what C can represent. */ | |
474 | ptrdiff_t c_bound = min (PTRDIFF_MAX, SIZE_MAX) / elsz; | |
475 | ||
476 | if (min (string_or_buffer_bound, c_bound) <= idx) | |
39e378da PE |
477 | memory_full (SIZE_MAX); |
478 | new_size = idx - idx % BIDI_CACHE_CHUNK + BIDI_CACHE_CHUNK; | |
479 | bidi_cache = (struct bidi_it *) xrealloc (bidi_cache, new_size * elsz); | |
480 | bidi_cache_size = new_size; | |
58b9f433 EZ |
481 | } |
482 | } | |
483 | ||
55d4c1b2 | 484 | static inline void |
b7b65b15 EZ |
485 | bidi_cache_iterator_state (struct bidi_it *bidi_it, int resolved) |
486 | { | |
39e378da | 487 | ptrdiff_t idx; |
b7b65b15 EZ |
488 | |
489 | /* We should never cache on backward scans. */ | |
490 | if (bidi_it->scan_dir == -1) | |
491 | abort (); | |
492 | idx = bidi_cache_search (bidi_it->charpos, -1, 1); | |
493 | ||
494 | if (idx < 0) | |
495 | { | |
496 | idx = bidi_cache_idx; | |
58b9f433 | 497 | bidi_cache_ensure_space (idx); |
bd924a5d EZ |
498 | /* Character positions should correspond to cache positions 1:1. |
499 | If we are outside the range of cached positions, the cache is | |
500 | useless and must be reset. */ | |
87e67904 | 501 | if (idx > bidi_cache_start && |
182ce2d2 EZ |
502 | (bidi_it->charpos > (bidi_cache[idx - 1].charpos |
503 | + bidi_cache[idx - 1].nchars) | |
a1344e7d | 504 | || bidi_it->charpos < bidi_cache[bidi_cache_start].charpos)) |
bd924a5d EZ |
505 | { |
506 | bidi_cache_reset (); | |
87e67904 | 507 | idx = bidi_cache_start; |
bd924a5d | 508 | } |
102ebb00 EZ |
509 | if (bidi_it->nchars <= 0) |
510 | abort (); | |
b7b65b15 EZ |
511 | bidi_copy_it (&bidi_cache[idx], bidi_it); |
512 | if (!resolved) | |
513 | bidi_cache[idx].resolved_level = -1; | |
514 | } | |
515 | else | |
516 | { | |
517 | /* Copy only the members which could have changed, to avoid | |
518 | costly copying of the entire struct. */ | |
519 | bidi_cache[idx].type = bidi_it->type; | |
2d6e4628 | 520 | bidi_check_type (bidi_it->type); |
89d3374a EZ |
521 | bidi_cache[idx].type_after_w1 = bidi_it->type_after_w1; |
522 | bidi_check_type (bidi_it->type_after_w1); | |
b7b65b15 EZ |
523 | if (resolved) |
524 | bidi_cache[idx].resolved_level = bidi_it->resolved_level; | |
525 | else | |
526 | bidi_cache[idx].resolved_level = -1; | |
527 | bidi_cache[idx].invalid_levels = bidi_it->invalid_levels; | |
528 | bidi_cache[idx].invalid_rl_levels = bidi_it->invalid_rl_levels; | |
529 | bidi_cache[idx].next_for_neutral = bidi_it->next_for_neutral; | |
530 | bidi_cache[idx].next_for_ws = bidi_it->next_for_ws; | |
531 | bidi_cache[idx].ignore_bn_limit = bidi_it->ignore_bn_limit; | |
532 | } | |
533 | ||
534 | bidi_cache_last_idx = idx; | |
535 | if (idx >= bidi_cache_idx) | |
536 | bidi_cache_idx = idx + 1; | |
537 | } | |
538 | ||
55d4c1b2 | 539 | static inline bidi_type_t |
61bfec98 | 540 | bidi_cache_find (EMACS_INT charpos, int level, struct bidi_it *bidi_it) |
b7b65b15 | 541 | { |
39e378da | 542 | ptrdiff_t i = bidi_cache_search (charpos, level, bidi_it->scan_dir); |
b7b65b15 | 543 | |
87e67904 | 544 | if (i >= bidi_cache_start) |
b7b65b15 EZ |
545 | { |
546 | bidi_dir_t current_scan_dir = bidi_it->scan_dir; | |
547 | ||
5009f85e | 548 | bidi_copy_it (bidi_it, &bidi_cache[i]); |
b7b65b15 EZ |
549 | bidi_cache_last_idx = i; |
550 | /* Don't let scan direction from from the cached state override | |
551 | the current scan direction. */ | |
552 | bidi_it->scan_dir = current_scan_dir; | |
553 | return bidi_it->type; | |
554 | } | |
555 | ||
556 | return UNKNOWN_BT; | |
557 | } | |
558 | ||
55d4c1b2 | 559 | static inline int |
b7b65b15 EZ |
560 | bidi_peek_at_next_level (struct bidi_it *bidi_it) |
561 | { | |
87e67904 | 562 | if (bidi_cache_idx == bidi_cache_start || bidi_cache_last_idx == -1) |
b7b65b15 EZ |
563 | abort (); |
564 | return bidi_cache[bidi_cache_last_idx + bidi_it->scan_dir].resolved_level; | |
565 | } | |
566 | ||
cbb09f04 | 567 | \f |
58b9f433 EZ |
568 | /*********************************************************************** |
569 | Pushing and popping the bidi iterator state | |
570 | ***********************************************************************/ | |
0c22566f EZ |
571 | /* 5-slot stack for saving the start of the previous level of the |
572 | cache. xdisp.c maintains a 5-slot stack for its iterator state, | |
573 | and we need the same size of our stack. */ | |
39e378da | 574 | static ptrdiff_t bidi_cache_start_stack[IT_STACK_SIZE]; |
58b9f433 EZ |
575 | static int bidi_cache_sp; |
576 | ||
577 | /* Push the bidi iterator state in preparation for reordering a | |
578 | different object, e.g. display string found at certain buffer | |
7e2ad32c EZ |
579 | position. Pushing the bidi iterator boils down to saving its |
580 | entire state on the cache and starting a new cache "stacked" on top | |
581 | of the current cache. */ | |
58b9f433 EZ |
582 | void |
583 | bidi_push_it (struct bidi_it *bidi_it) | |
b7b65b15 | 584 | { |
58b9f433 EZ |
585 | /* Save the current iterator state in its entirety after the last |
586 | used cache slot. */ | |
587 | bidi_cache_ensure_space (bidi_cache_idx); | |
588 | memcpy (&bidi_cache[bidi_cache_idx++], bidi_it, sizeof (struct bidi_it)); | |
be39f003 | 589 | |
58b9f433 | 590 | /* Push the current cache start onto the stack. */ |
7e2ad32c | 591 | xassert (bidi_cache_sp < IT_STACK_SIZE); |
58b9f433 | 592 | bidi_cache_start_stack[bidi_cache_sp++] = bidi_cache_start; |
be39f003 | 593 | |
58b9f433 EZ |
594 | /* Start a new level of cache, and make it empty. */ |
595 | bidi_cache_start = bidi_cache_idx; | |
596 | bidi_cache_last_idx = -1; | |
597 | } | |
598 | ||
599 | /* Restore the iterator state saved by bidi_push_it and return the | |
600 | cache to the corresponding state. */ | |
601 | void | |
602 | bidi_pop_it (struct bidi_it *bidi_it) | |
603 | { | |
604 | if (bidi_cache_start <= 0) | |
605 | abort (); | |
606 | ||
607 | /* Reset the next free cache slot index to what it was before the | |
608 | call to bidi_push_it. */ | |
609 | bidi_cache_idx = bidi_cache_start - 1; | |
610 | ||
611 | /* Restore the bidi iterator state saved in the cache. */ | |
612 | memcpy (bidi_it, &bidi_cache[bidi_cache_idx], sizeof (struct bidi_it)); | |
613 | ||
614 | /* Pop the previous cache start from the stack. */ | |
615 | if (bidi_cache_sp <= 0) | |
616 | abort (); | |
617 | bidi_cache_start = bidi_cache_start_stack[--bidi_cache_sp]; | |
618 | ||
619 | /* Invalidate the last-used cache slot data. */ | |
620 | bidi_cache_last_idx = -1; | |
621 | } | |
622 | ||
ed94e6d7 EZ |
623 | /* Stash away a copy of the cache and its control variables. */ |
624 | void * | |
625 | bidi_shelve_cache (void) | |
626 | { | |
627 | unsigned char *databuf; | |
628 | ||
629 | if (bidi_cache_idx == 0) | |
630 | return NULL; | |
631 | ||
632 | databuf = xmalloc (sizeof (bidi_cache_idx) | |
633 | + bidi_cache_idx * sizeof (struct bidi_it) | |
634 | + sizeof (bidi_cache_start_stack) | |
635 | + sizeof (bidi_cache_sp) + sizeof (bidi_cache_start) | |
636 | + sizeof (bidi_cache_last_idx)); | |
637 | memcpy (databuf, &bidi_cache_idx, sizeof (bidi_cache_idx)); | |
638 | memcpy (databuf + sizeof (bidi_cache_idx), | |
639 | bidi_cache, bidi_cache_idx * sizeof (struct bidi_it)); | |
640 | memcpy (databuf + sizeof (bidi_cache_idx) | |
641 | + bidi_cache_idx * sizeof (struct bidi_it), | |
642 | bidi_cache_start_stack, sizeof (bidi_cache_start_stack)); | |
643 | memcpy (databuf + sizeof (bidi_cache_idx) | |
644 | + bidi_cache_idx * sizeof (struct bidi_it) | |
645 | + sizeof (bidi_cache_start_stack), | |
646 | &bidi_cache_sp, sizeof (bidi_cache_sp)); | |
647 | memcpy (databuf + sizeof (bidi_cache_idx) | |
648 | + bidi_cache_idx * sizeof (struct bidi_it) | |
649 | + sizeof (bidi_cache_start_stack) + sizeof (bidi_cache_sp), | |
650 | &bidi_cache_start, sizeof (bidi_cache_start)); | |
651 | memcpy (databuf + sizeof (bidi_cache_idx) | |
652 | + bidi_cache_idx * sizeof (struct bidi_it) | |
653 | + sizeof (bidi_cache_start_stack) + sizeof (bidi_cache_sp) | |
654 | + sizeof (bidi_cache_start), | |
655 | &bidi_cache_last_idx, sizeof (bidi_cache_last_idx)); | |
656 | ||
657 | return databuf; | |
658 | } | |
659 | ||
660 | /* Restore the cache state from a copy stashed away by bidi_shelve_cache. */ | |
661 | void | |
662 | bidi_unshelve_cache (void *databuf) | |
663 | { | |
664 | unsigned char *p = databuf; | |
665 | ||
666 | if (!p) | |
be39f003 | 667 | { |
ed94e6d7 EZ |
668 | /* A NULL pointer means an empty cache. */ |
669 | bidi_cache_start = 0; | |
670 | bidi_cache_sp = 0; | |
671 | bidi_cache_reset (); | |
be39f003 | 672 | } |
ed94e6d7 EZ |
673 | else |
674 | { | |
675 | memcpy (&bidi_cache_idx, p, sizeof (bidi_cache_idx)); | |
57b3e30b | 676 | bidi_cache_ensure_space (bidi_cache_idx); |
ed94e6d7 EZ |
677 | memcpy (bidi_cache, p + sizeof (bidi_cache_idx), |
678 | bidi_cache_idx * sizeof (struct bidi_it)); | |
679 | memcpy (bidi_cache_start_stack, | |
680 | p + sizeof (bidi_cache_idx) | |
681 | + bidi_cache_idx * sizeof (struct bidi_it), | |
682 | sizeof (bidi_cache_start_stack)); | |
683 | memcpy (&bidi_cache_sp, | |
684 | p + sizeof (bidi_cache_idx) | |
685 | + bidi_cache_idx * sizeof (struct bidi_it) | |
686 | + sizeof (bidi_cache_start_stack), | |
687 | sizeof (bidi_cache_sp)); | |
688 | memcpy (&bidi_cache_start, | |
689 | p + sizeof (bidi_cache_idx) | |
690 | + bidi_cache_idx * sizeof (struct bidi_it) | |
691 | + sizeof (bidi_cache_start_stack) + sizeof (bidi_cache_sp), | |
692 | sizeof (bidi_cache_start)); | |
693 | memcpy (&bidi_cache_last_idx, | |
694 | p + sizeof (bidi_cache_idx) | |
695 | + bidi_cache_idx * sizeof (struct bidi_it) | |
696 | + sizeof (bidi_cache_start_stack) + sizeof (bidi_cache_sp) | |
697 | + sizeof (bidi_cache_start), | |
698 | sizeof (bidi_cache_last_idx)); | |
699 | ||
700 | xfree (p); | |
701 | } | |
702 | } | |
b7b65b15 | 703 | |
58b9f433 | 704 | \f |
cbb09f04 EZ |
705 | /*********************************************************************** |
706 | Initialization | |
707 | ***********************************************************************/ | |
708 | static void | |
709 | bidi_initialize (void) | |
b7b65b15 | 710 | { |
be39f003 | 711 | |
cbb09f04 EZ |
712 | #include "biditype.h" |
713 | #include "bidimirror.h" | |
be39f003 | 714 | |
cbb09f04 | 715 | int i; |
b7b65b15 | 716 | |
cbb09f04 EZ |
717 | bidi_type_table = Fmake_char_table (Qnil, make_number (STRONG_L)); |
718 | staticpro (&bidi_type_table); | |
719 | ||
720 | for (i = 0; i < sizeof bidi_type / sizeof bidi_type[0]; i++) | |
721 | char_table_set_range (bidi_type_table, bidi_type[i].from, bidi_type[i].to, | |
722 | make_number (bidi_type[i].type)); | |
723 | ||
724 | bidi_mirror_table = Fmake_char_table (Qnil, Qnil); | |
725 | staticpro (&bidi_mirror_table); | |
726 | ||
727 | for (i = 0; i < sizeof bidi_mirror / sizeof bidi_mirror[0]; i++) | |
728 | char_table_set (bidi_mirror_table, bidi_mirror[i].from, | |
729 | make_number (bidi_mirror[i].to)); | |
730 | ||
731 | Qparagraph_start = intern ("paragraph-start"); | |
732 | staticpro (&Qparagraph_start); | |
733 | paragraph_start_re = Fsymbol_value (Qparagraph_start); | |
734 | if (!STRINGP (paragraph_start_re)) | |
735 | paragraph_start_re = build_string ("\f\\|[ \t]*$"); | |
736 | staticpro (¶graph_start_re); | |
737 | Qparagraph_separate = intern ("paragraph-separate"); | |
738 | staticpro (&Qparagraph_separate); | |
739 | paragraph_separate_re = Fsymbol_value (Qparagraph_separate); | |
740 | if (!STRINGP (paragraph_separate_re)) | |
741 | paragraph_separate_re = build_string ("[ \t\f]*$"); | |
742 | staticpro (¶graph_separate_re); | |
58b9f433 EZ |
743 | |
744 | bidi_cache_sp = 0; | |
745 | ||
cbb09f04 | 746 | bidi_initialized = 1; |
b7b65b15 EZ |
747 | } |
748 | ||
cbb09f04 EZ |
749 | /* Do whatever UAX#9 clause X8 says should be done at paragraph's |
750 | end. */ | |
55d4c1b2 | 751 | static inline void |
cbb09f04 | 752 | bidi_set_paragraph_end (struct bidi_it *bidi_it) |
b7b65b15 | 753 | { |
cbb09f04 EZ |
754 | bidi_it->invalid_levels = 0; |
755 | bidi_it->invalid_rl_levels = -1; | |
756 | bidi_it->stack_idx = 0; | |
757 | bidi_it->resolved_level = bidi_it->level_stack[0].level; | |
758 | } | |
b7b65b15 | 759 | |
cbb09f04 EZ |
760 | /* Initialize the bidi iterator from buffer/string position CHARPOS. */ |
761 | void | |
762 | bidi_init_it (EMACS_INT charpos, EMACS_INT bytepos, int frame_window_p, | |
763 | struct bidi_it *bidi_it) | |
764 | { | |
765 | if (! bidi_initialized) | |
766 | bidi_initialize (); | |
767 | if (charpos >= 0) | |
768 | bidi_it->charpos = charpos; | |
769 | if (bytepos >= 0) | |
770 | bidi_it->bytepos = bytepos; | |
771 | bidi_it->frame_window_p = frame_window_p; | |
772 | bidi_it->nchars = -1; /* to be computed in bidi_resolve_explicit_1 */ | |
773 | bidi_it->first_elt = 1; | |
774 | bidi_set_paragraph_end (bidi_it); | |
775 | bidi_it->new_paragraph = 1; | |
776 | bidi_it->separator_limit = -1; | |
777 | bidi_it->type = NEUTRAL_B; | |
778 | bidi_it->type_after_w1 = NEUTRAL_B; | |
779 | bidi_it->orig_type = NEUTRAL_B; | |
780 | bidi_it->prev_was_pdf = 0; | |
781 | bidi_it->prev.type = bidi_it->prev.type_after_w1 = | |
782 | bidi_it->prev.orig_type = UNKNOWN_BT; | |
89d3374a EZ |
783 | bidi_it->last_strong.type = bidi_it->last_strong.type_after_w1 = |
784 | bidi_it->last_strong.orig_type = UNKNOWN_BT; | |
cbb09f04 EZ |
785 | bidi_it->next_for_neutral.charpos = -1; |
786 | bidi_it->next_for_neutral.type = | |
787 | bidi_it->next_for_neutral.type_after_w1 = | |
89d3374a | 788 | bidi_it->next_for_neutral.orig_type = UNKNOWN_BT; |
cbb09f04 EZ |
789 | bidi_it->prev_for_neutral.charpos = -1; |
790 | bidi_it->prev_for_neutral.type = | |
791 | bidi_it->prev_for_neutral.type_after_w1 = | |
792 | bidi_it->prev_for_neutral.orig_type = UNKNOWN_BT; | |
793 | bidi_it->sor = L2R; /* FIXME: should it be user-selectable? */ | |
794 | bidi_it->disp_pos = -1; /* invalid/unknown */ | |
795 | /* We can only shrink the cache if we are at the bottom level of its | |
796 | "stack". */ | |
797 | if (bidi_cache_start == 0) | |
798 | bidi_cache_shrink (); | |
58b9f433 EZ |
799 | else |
800 | bidi_cache_reset (); | |
b7b65b15 EZ |
801 | } |
802 | ||
182ce2d2 | 803 | /* Perform initializations for reordering a new line of bidi text. */ |
6bff6497 | 804 | static void |
be39f003 EZ |
805 | bidi_line_init (struct bidi_it *bidi_it) |
806 | { | |
807 | bidi_it->scan_dir = 1; /* FIXME: do we need to have control on this? */ | |
808 | bidi_it->resolved_level = bidi_it->level_stack[0].level; | |
809 | bidi_it->level_stack[0].override = NEUTRAL_DIR; /* X1 */ | |
810 | bidi_it->invalid_levels = 0; | |
811 | bidi_it->invalid_rl_levels = -1; | |
812 | bidi_it->next_en_pos = -1; | |
813 | bidi_it->next_for_ws.type = UNKNOWN_BT; | |
b44d9321 EZ |
814 | bidi_set_sor_type (bidi_it, |
815 | bidi_it->paragraph_dir == R2L ? 1 : 0, | |
be39f003 EZ |
816 | bidi_it->level_stack[0].level); /* X10 */ |
817 | ||
818 | bidi_cache_reset (); | |
819 | } | |
820 | ||
cbb09f04 EZ |
821 | \f |
822 | /*********************************************************************** | |
823 | Fetching characters | |
824 | ***********************************************************************/ | |
825 | ||
f3014ef5 EZ |
826 | /* Count bytes in string S between BEG/BEGBYTE and END. BEG and END |
827 | are zero-based character positions in S, BEGBYTE is byte position | |
828 | corresponding to BEG. UNIBYTE, if non-zero, means S is a unibyte | |
829 | string. */ | |
87e67904 EZ |
830 | static inline EMACS_INT |
831 | bidi_count_bytes (const unsigned char *s, const EMACS_INT beg, | |
f3014ef5 | 832 | const EMACS_INT begbyte, const EMACS_INT end, int unibyte) |
87e67904 EZ |
833 | { |
834 | EMACS_INT pos = beg; | |
835 | const unsigned char *p = s + begbyte, *start = p; | |
836 | ||
f3014ef5 EZ |
837 | if (unibyte) |
838 | p = s + end; | |
839 | else | |
87e67904 | 840 | { |
f3014ef5 EZ |
841 | if (!CHAR_HEAD_P (*p)) |
842 | abort (); | |
843 | ||
844 | while (pos < end) | |
845 | { | |
846 | p += BYTES_BY_CHAR_HEAD (*p); | |
847 | pos++; | |
848 | } | |
87e67904 EZ |
849 | } |
850 | ||
851 | return p - start; | |
852 | } | |
853 | ||
854 | /* Fetch and returns the character at byte position BYTEPOS. If S is | |
855 | non-NULL, fetch the character from string S; otherwise fetch the | |
f3014ef5 EZ |
856 | character from the current buffer. UNIBYTE non-zero means S is a |
857 | unibyte string. */ | |
87e67904 | 858 | static inline int |
f3014ef5 | 859 | bidi_char_at_pos (EMACS_INT bytepos, const unsigned char *s, int unibyte) |
87e67904 EZ |
860 | { |
861 | if (s) | |
f3014ef5 EZ |
862 | { |
863 | if (unibyte) | |
864 | return s[bytepos]; | |
865 | else | |
866 | return STRING_CHAR (s + bytepos); | |
867 | } | |
87e67904 EZ |
868 | else |
869 | return FETCH_MULTIBYTE_CHAR (bytepos); | |
870 | } | |
871 | ||
102ebb00 EZ |
872 | /* Fetch and return the character at BYTEPOS/CHARPOS. If that |
873 | character is covered by a display string, treat the entire run of | |
874 | covered characters as a single character u+FFFC, and return their | |
875 | combined length in CH_LEN and NCHARS. DISP_POS specifies the | |
876 | character position of the next display string, or -1 if not yet | |
877 | computed. When the next character is at or beyond that position, | |
878 | the function updates DISP_POS with the position of the next display | |
9f257352 EZ |
879 | string. STRING->s is the C string to iterate, or NULL if iterating |
880 | over a buffer or a Lisp string; in the latter case, STRING->lstring | |
881 | is the Lisp string. */ | |
fec2107c | 882 | static inline int |
102ebb00 | 883 | bidi_fetch_char (EMACS_INT bytepos, EMACS_INT charpos, EMACS_INT *disp_pos, |
87e67904 | 884 | struct bidi_string_data *string, |
fc6f18ce | 885 | int frame_window_p, EMACS_INT *ch_len, EMACS_INT *nchars) |
182ce2d2 EZ |
886 | { |
887 | int ch; | |
9f257352 EZ |
888 | EMACS_INT endpos = |
889 | (string->s || STRINGP (string->lstring)) ? string->schars : ZV; | |
6db161be | 890 | struct text_pos pos; |
182ce2d2 | 891 | |
182ce2d2 | 892 | /* If we got past the last known position of display string, compute |
87e67904 EZ |
893 | the position of the next one. That position could be at CHARPOS. */ |
894 | if (charpos < endpos && charpos > *disp_pos) | |
6db161be EZ |
895 | { |
896 | SET_TEXT_POS (pos, charpos, bytepos); | |
897 | *disp_pos = compute_display_string_pos (&pos, string, frame_window_p); | |
898 | } | |
102ebb00 EZ |
899 | |
900 | /* Fetch the character at BYTEPOS. */ | |
87e67904 | 901 | if (charpos >= endpos) |
182ce2d2 EZ |
902 | { |
903 | ch = BIDI_EOB; | |
904 | *ch_len = 1; | |
905 | *nchars = 1; | |
87e67904 | 906 | *disp_pos = endpos; |
182ce2d2 | 907 | } |
102ebb00 | 908 | else if (charpos >= *disp_pos) |
182ce2d2 | 909 | { |
7b600102 EZ |
910 | EMACS_INT disp_end_pos; |
911 | ||
912 | /* We don't expect to find ourselves in the middle of a display | |
913 | property. Hopefully, it will never be needed. */ | |
914 | if (charpos > *disp_pos) | |
915 | abort (); | |
916 | /* Return the Unicode Object Replacement Character to represent | |
87e67904 | 917 | the entire run of characters covered by the display string. */ |
7b600102 | 918 | ch = 0xFFFC; |
87e67904 | 919 | disp_end_pos = compute_display_string_end (*disp_pos, string); |
7b600102 | 920 | *nchars = disp_end_pos - *disp_pos; |
f3014ef5 EZ |
921 | if (*nchars <= 0) |
922 | abort (); | |
87e67904 EZ |
923 | if (string->s) |
924 | *ch_len = bidi_count_bytes (string->s, *disp_pos, bytepos, | |
f3014ef5 | 925 | disp_end_pos, string->unibyte); |
9f257352 EZ |
926 | else if (STRINGP (string->lstring)) |
927 | *ch_len = bidi_count_bytes (SDATA (string->lstring), *disp_pos, | |
f3014ef5 | 928 | bytepos, disp_end_pos, string->unibyte); |
87e67904 EZ |
929 | else |
930 | *ch_len = CHAR_TO_BYTE (disp_end_pos) - bytepos; | |
182ce2d2 | 931 | } |
182ce2d2 EZ |
932 | else |
933 | { | |
87e67904 EZ |
934 | if (string->s) |
935 | { | |
66ff7ddf | 936 | int len; |
87e67904 | 937 | |
f3014ef5 EZ |
938 | if (!string->unibyte) |
939 | { | |
940 | ch = STRING_CHAR_AND_LENGTH (string->s + bytepos, len); | |
941 | *ch_len = len; | |
942 | } | |
943 | else | |
944 | { | |
945 | ch = UNIBYTE_TO_CHAR (string->s[bytepos]); | |
946 | *ch_len = 1; | |
947 | } | |
87e67904 | 948 | } |
9f257352 EZ |
949 | else if (STRINGP (string->lstring)) |
950 | { | |
66ff7ddf | 951 | int len; |
9f257352 | 952 | |
f3014ef5 EZ |
953 | if (!string->unibyte) |
954 | { | |
955 | ch = STRING_CHAR_AND_LENGTH (SDATA (string->lstring) + bytepos, | |
956 | len); | |
957 | *ch_len = len; | |
958 | } | |
959 | else | |
960 | { | |
961 | ch = UNIBYTE_TO_CHAR (SREF (string->lstring, bytepos)); | |
962 | *ch_len = 1; | |
963 | } | |
9f257352 | 964 | } |
87e67904 EZ |
965 | else |
966 | { | |
967 | ch = FETCH_MULTIBYTE_CHAR (bytepos); | |
968 | *ch_len = CHAR_BYTES (ch); | |
969 | } | |
182ce2d2 EZ |
970 | *nchars = 1; |
971 | } | |
972 | ||
973 | /* If we just entered a run of characters covered by a display | |
974 | string, compute the position of the next display string. */ | |
87e67904 | 975 | if (charpos + *nchars <= endpos && charpos + *nchars > *disp_pos) |
6db161be EZ |
976 | { |
977 | SET_TEXT_POS (pos, charpos + *nchars, bytepos + *ch_len); | |
978 | *disp_pos = compute_display_string_pos (&pos, string, frame_window_p); | |
979 | } | |
182ce2d2 EZ |
980 | |
981 | return ch; | |
982 | } | |
983 | ||
cbb09f04 EZ |
984 | \f |
985 | /*********************************************************************** | |
986 | Determining paragraph direction | |
987 | ***********************************************************************/ | |
988 | ||
989 | /* Check if buffer position CHARPOS/BYTEPOS is the end of a paragraph. | |
990 | Value is the non-negative length of the paragraph separator | |
991 | following the buffer position, -1 if position is at the beginning | |
992 | of a new paragraph, or -2 if position is neither at beginning nor | |
993 | at end of a paragraph. */ | |
994 | static EMACS_INT | |
995 | bidi_at_paragraph_end (EMACS_INT charpos, EMACS_INT bytepos) | |
996 | { | |
997 | Lisp_Object sep_re; | |
998 | Lisp_Object start_re; | |
999 | EMACS_INT val; | |
1000 | ||
1001 | sep_re = paragraph_separate_re; | |
1002 | start_re = paragraph_start_re; | |
1003 | ||
1004 | val = fast_looking_at (sep_re, charpos, bytepos, ZV, ZV_BYTE, Qnil); | |
1005 | if (val < 0) | |
1006 | { | |
1007 | if (fast_looking_at (start_re, charpos, bytepos, ZV, ZV_BYTE, Qnil) >= 0) | |
1008 | val = -1; | |
1009 | else | |
1010 | val = -2; | |
1011 | } | |
1012 | ||
1013 | return val; | |
1014 | } | |
1015 | ||
be39f003 EZ |
1016 | /* Find the beginning of this paragraph by looking back in the buffer. |
1017 | Value is the byte position of the paragraph's beginning. */ | |
1018 | static EMACS_INT | |
b44d9321 | 1019 | bidi_find_paragraph_start (EMACS_INT pos, EMACS_INT pos_byte) |
6bff6497 | 1020 | { |
372b7a95 | 1021 | Lisp_Object re = paragraph_start_re; |
6bff6497 EZ |
1022 | EMACS_INT limit = ZV, limit_byte = ZV_BYTE; |
1023 | ||
6bff6497 EZ |
1024 | while (pos_byte > BEGV_BYTE |
1025 | && fast_looking_at (re, pos, pos_byte, limit, limit_byte, Qnil) < 0) | |
1026 | { | |
182ce2d2 EZ |
1027 | /* FIXME: What if the paragraph beginning is covered by a |
1028 | display string? And what if a display string covering some | |
1029 | of the text over which we scan back includes | |
1030 | paragraph_start_re? */ | |
be39f003 EZ |
1031 | pos = find_next_newline_no_quit (pos - 1, -1); |
1032 | pos_byte = CHAR_TO_BYTE (pos); | |
6bff6497 | 1033 | } |
be39f003 | 1034 | return pos_byte; |
6bff6497 EZ |
1035 | } |
1036 | ||
bea4f10c | 1037 | /* Determine the base direction, a.k.a. base embedding level, of the |
b44d9321 EZ |
1038 | paragraph we are about to iterate through. If DIR is either L2R or |
1039 | R2L, just use that. Otherwise, determine the paragraph direction | |
bea4f10c EZ |
1040 | from the first strong directional character of the paragraph. |
1041 | ||
182ce2d2 | 1042 | NO_DEFAULT_P non-zero means don't default to L2R if the paragraph |
bea4f10c EZ |
1043 | has no strong directional characters and both DIR and |
1044 | bidi_it->paragraph_dir are NEUTRAL_DIR. In that case, search back | |
1045 | in the buffer until a paragraph is found with a strong character, | |
1046 | or until hitting BEGV. In the latter case, fall back to L2R. This | |
1047 | flag is used in current-bidi-paragraph-direction. | |
1048 | ||
1049 | Note that this function gives the paragraph separator the same | |
1050 | direction as the preceding paragraph, even though Emacs generally | |
1051 | views the separartor as not belonging to any paragraph. */ | |
b7b65b15 | 1052 | void |
bea4f10c | 1053 | bidi_paragraph_init (bidi_dir_t dir, struct bidi_it *bidi_it, int no_default_p) |
b7b65b15 | 1054 | { |
6bff6497 | 1055 | EMACS_INT bytepos = bidi_it->bytepos; |
9f257352 | 1056 | int string_p = bidi_it->string.s != NULL || STRINGP (bidi_it->string.lstring); |
bea4f10c | 1057 | EMACS_INT pstartbyte; |
87e67904 EZ |
1058 | /* Note that begbyte is a byte position, while end is a character |
1059 | position. Yes, this is ugly, but we are trying to avoid costly | |
1060 | calls to BYTE_TO_CHAR and its ilk. */ | |
1061 | EMACS_INT begbyte = string_p ? 0 : BEGV_BYTE; | |
1062 | EMACS_INT end = string_p ? bidi_it->string.schars : ZV; | |
e7402cb2 | 1063 | |
5e65aec0 | 1064 | /* Special case for an empty buffer. */ |
87e67904 | 1065 | if (bytepos == begbyte && bidi_it->charpos == end) |
5e65aec0 | 1066 | dir = L2R; |
9c82e145 | 1067 | /* We should never be called at EOB or before BEGV. */ |
87e67904 | 1068 | else if (bidi_it->charpos >= end || bytepos < begbyte) |
9c82e145 EZ |
1069 | abort (); |
1070 | ||
be39f003 EZ |
1071 | if (dir == L2R) |
1072 | { | |
1073 | bidi_it->paragraph_dir = L2R; | |
1074 | bidi_it->new_paragraph = 0; | |
1075 | } | |
1076 | else if (dir == R2L) | |
1077 | { | |
1078 | bidi_it->paragraph_dir = R2L; | |
1079 | bidi_it->new_paragraph = 0; | |
1080 | } | |
b7b65b15 EZ |
1081 | else if (dir == NEUTRAL_DIR) /* P2 */ |
1082 | { | |
182ce2d2 EZ |
1083 | int ch; |
1084 | EMACS_INT ch_len, nchars; | |
1085 | EMACS_INT pos, disp_pos = -1; | |
6bff6497 | 1086 | bidi_type_t type; |
9f257352 | 1087 | const unsigned char *s; |
be39f003 | 1088 | |
d20e1419 EZ |
1089 | if (!bidi_initialized) |
1090 | bidi_initialize (); | |
1091 | ||
be39f003 EZ |
1092 | /* If we are inside a paragraph separator, we are just waiting |
1093 | for the separator to be exhausted; use the previous paragraph | |
e5a2fec7 EZ |
1094 | direction. But don't do that if we have been just reseated, |
1095 | because we need to reinitialize below in that case. */ | |
1096 | if (!bidi_it->first_elt | |
1097 | && bidi_it->charpos < bidi_it->separator_limit) | |
be39f003 EZ |
1098 | return; |
1099 | ||
b44d9321 | 1100 | /* If we are on a newline, get past it to where the next |
5e65aec0 EZ |
1101 | paragraph might start. But don't do that at BEGV since then |
1102 | we are potentially in a new paragraph that doesn't yet | |
1103 | exist. */ | |
c143c213 | 1104 | pos = bidi_it->charpos; |
9f257352 EZ |
1105 | s = STRINGP (bidi_it->string.lstring) ? |
1106 | SDATA (bidi_it->string.lstring) : bidi_it->string.s; | |
f3014ef5 EZ |
1107 | if (bytepos > begbyte |
1108 | && bidi_char_at_pos (bytepos, s, bidi_it->string.unibyte) == '\n') | |
be39f003 | 1109 | { |
b44d9321 | 1110 | bytepos++; |
c143c213 | 1111 | pos++; |
be39f003 | 1112 | } |
b44d9321 EZ |
1113 | |
1114 | /* We are either at the beginning of a paragraph or in the | |
1115 | middle of it. Find where this paragraph starts. */ | |
87e67904 EZ |
1116 | if (string_p) |
1117 | { | |
1118 | /* We don't support changes of paragraph direction inside a | |
1119 | string. It is treated as a single paragraph. */ | |
1120 | pstartbyte = 0; | |
1121 | } | |
1122 | else | |
1123 | pstartbyte = bidi_find_paragraph_start (pos, bytepos); | |
be39f003 EZ |
1124 | bidi_it->separator_limit = -1; |
1125 | bidi_it->new_paragraph = 0; | |
bea4f10c EZ |
1126 | |
1127 | /* The following loop is run more than once only if NO_DEFAULT_P | |
87e67904 | 1128 | is non-zero, and only if we are iterating on a buffer. */ |
bea4f10c EZ |
1129 | do { |
1130 | bytepos = pstartbyte; | |
87e67904 EZ |
1131 | if (!string_p) |
1132 | pos = BYTE_TO_CHAR (bytepos); | |
1133 | ch = bidi_fetch_char (bytepos, pos, &disp_pos, &bidi_it->string, | |
1134 | bidi_it->frame_window_p, &ch_len, &nchars); | |
bea4f10c EZ |
1135 | type = bidi_get_type (ch, NEUTRAL_DIR); |
1136 | ||
182ce2d2 | 1137 | for (pos += nchars, bytepos += ch_len; |
bea4f10c EZ |
1138 | /* NOTE: UAX#9 says to search only for L, AL, or R types |
1139 | of characters, and ignore RLE, RLO, LRE, and LRO. | |
1140 | However, I'm not sure it makes sense to omit those 4; | |
1141 | should try with and without that to see the effect. */ | |
1142 | (bidi_get_category (type) != STRONG) | |
1143 | || (bidi_ignore_explicit_marks_for_paragraph_level | |
1144 | && (type == RLE || type == RLO | |
1145 | || type == LRE || type == LRO)); | |
1146 | type = bidi_get_type (ch, NEUTRAL_DIR)) | |
1147 | { | |
87e67904 | 1148 | if (pos >= end) |
bea4f10c EZ |
1149 | { |
1150 | /* Pretend there's a paragraph separator at end of | |
87e67904 | 1151 | buffer/string. */ |
bea4f10c EZ |
1152 | type = NEUTRAL_B; |
1153 | break; | |
1154 | } | |
0bb23927 EZ |
1155 | if (!string_p |
1156 | && type == NEUTRAL_B | |
1157 | && bidi_at_paragraph_end (pos, bytepos) >= -1) | |
cf99dcf8 | 1158 | break; |
102ebb00 | 1159 | /* Fetch next character and advance to get past it. */ |
87e67904 | 1160 | ch = bidi_fetch_char (bytepos, pos, &disp_pos, &bidi_it->string, |
fc6f18ce | 1161 | bidi_it->frame_window_p, &ch_len, &nchars); |
182ce2d2 EZ |
1162 | pos += nchars; |
1163 | bytepos += ch_len; | |
bea4f10c EZ |
1164 | } |
1165 | if (type == STRONG_R || type == STRONG_AL) /* P3 */ | |
1166 | bidi_it->paragraph_dir = R2L; | |
1167 | else if (type == STRONG_L) | |
1168 | bidi_it->paragraph_dir = L2R; | |
87e67904 EZ |
1169 | if (!string_p |
1170 | && no_default_p && bidi_it->paragraph_dir == NEUTRAL_DIR) | |
bea4f10c EZ |
1171 | { |
1172 | /* If this paragraph is at BEGV, default to L2R. */ | |
1173 | if (pstartbyte == BEGV_BYTE) | |
1174 | bidi_it->paragraph_dir = L2R; /* P3 and HL1 */ | |
1175 | else | |
1176 | { | |
1177 | EMACS_INT prevpbyte = pstartbyte; | |
1178 | EMACS_INT p = BYTE_TO_CHAR (pstartbyte), pbyte = pstartbyte; | |
1179 | ||
1180 | /* Find the beginning of the previous paragraph, if any. */ | |
1181 | while (pbyte > BEGV_BYTE && prevpbyte >= pstartbyte) | |
1182 | { | |
182ce2d2 EZ |
1183 | /* FXIME: What if p is covered by a display |
1184 | string? See also a FIXME inside | |
1185 | bidi_find_paragraph_start. */ | |
bea4f10c EZ |
1186 | p--; |
1187 | pbyte = CHAR_TO_BYTE (p); | |
1188 | prevpbyte = bidi_find_paragraph_start (p, pbyte); | |
1189 | } | |
1190 | pstartbyte = prevpbyte; | |
1191 | } | |
1192 | } | |
87e67904 EZ |
1193 | } while (!string_p |
1194 | && no_default_p && bidi_it->paragraph_dir == NEUTRAL_DIR); | |
b7b65b15 | 1195 | } |
be39f003 EZ |
1196 | else |
1197 | abort (); | |
b7b65b15 | 1198 | |
b44d9321 EZ |
1199 | /* Contrary to UAX#9 clause P3, we only default the paragraph |
1200 | direction to L2R if we have no previous usable paragraph | |
bea4f10c | 1201 | direction. This is allowed by the HL1 clause. */ |
d20e1419 | 1202 | if (bidi_it->paragraph_dir != L2R && bidi_it->paragraph_dir != R2L) |
bea4f10c | 1203 | bidi_it->paragraph_dir = L2R; /* P3 and HL1 ``higher-level protocols'' */ |
be39f003 | 1204 | if (bidi_it->paragraph_dir == R2L) |
b44d9321 | 1205 | bidi_it->level_stack[0].level = 1; |
be39f003 | 1206 | else |
b44d9321 | 1207 | bidi_it->level_stack[0].level = 0; |
be39f003 EZ |
1208 | |
1209 | bidi_line_init (bidi_it); | |
b7b65b15 EZ |
1210 | } |
1211 | ||
cbb09f04 EZ |
1212 | \f |
1213 | /*********************************************************************** | |
1214 | Resolving explicit and implicit levels. | |
58b9f433 | 1215 | The rest of this file constitutes the core of the UBA implementation. |
cbb09f04 | 1216 | ***********************************************************************/ |
b7b65b15 | 1217 | |
55d4c1b2 | 1218 | static inline int |
182ce2d2 | 1219 | bidi_explicit_dir_char (int ch) |
b7b65b15 | 1220 | { |
182ce2d2 EZ |
1221 | bidi_type_t ch_type; |
1222 | ||
1223 | if (!bidi_initialized) | |
1224 | abort (); | |
1225 | ch_type = (bidi_type_t) XINT (CHAR_TABLE_REF (bidi_type_table, ch)); | |
1226 | return (ch_type == LRE || ch_type == LRO | |
1227 | || ch_type == RLE || ch_type == RLO | |
1228 | || ch_type == PDF); | |
b7b65b15 EZ |
1229 | } |
1230 | ||
1231 | /* A helper function for bidi_resolve_explicit. It advances to the | |
1232 | next character in logical order and determines the new embedding | |
1233 | level and directional override, but does not take into account | |
1234 | empty embeddings. */ | |
1235 | static int | |
1236 | bidi_resolve_explicit_1 (struct bidi_it *bidi_it) | |
1237 | { | |
1238 | int curchar; | |
1239 | bidi_type_t type; | |
1240 | int current_level; | |
1241 | int new_level; | |
1242 | bidi_dir_t override; | |
9f257352 | 1243 | int string_p = bidi_it->string.s != NULL || STRINGP (bidi_it->string.lstring); |
b7b65b15 | 1244 | |
182ce2d2 EZ |
1245 | /* If reseat()'ed, don't advance, so as to start iteration from the |
1246 | position where we were reseated. bidi_it->bytepos can be less | |
1247 | than BEGV_BYTE after reseat to BEGV. */ | |
87e67904 | 1248 | if (bidi_it->bytepos < (string_p ? 0 : BEGV_BYTE) |
9c82e145 | 1249 | || bidi_it->first_elt) |
e7402cb2 | 1250 | { |
9c82e145 | 1251 | bidi_it->first_elt = 0; |
87e67904 EZ |
1252 | if (string_p) |
1253 | { | |
9f257352 EZ |
1254 | const unsigned char *p = |
1255 | STRINGP (bidi_it->string.lstring) | |
1256 | ? SDATA (bidi_it->string.lstring) : bidi_it->string.s; | |
1257 | ||
87e67904 EZ |
1258 | if (bidi_it->charpos < 0) |
1259 | bidi_it->charpos = 0; | |
f3014ef5 EZ |
1260 | bidi_it->bytepos = bidi_count_bytes (p, 0, 0, bidi_it->charpos, |
1261 | bidi_it->string.unibyte); | |
87e67904 EZ |
1262 | } |
1263 | else | |
1264 | { | |
1265 | if (bidi_it->charpos < BEGV) | |
1266 | bidi_it->charpos = BEGV; | |
1267 | bidi_it->bytepos = CHAR_TO_BYTE (bidi_it->charpos); | |
1268 | } | |
e7402cb2 | 1269 | } |
87e67904 EZ |
1270 | /* Don't move at end of buffer/string. */ |
1271 | else if (bidi_it->charpos < (string_p ? bidi_it->string.schars : ZV)) | |
b7b65b15 | 1272 | { |
182ce2d2 EZ |
1273 | /* Advance to the next character, skipping characters covered by |
1274 | display strings (nchars > 1). */ | |
102ebb00 EZ |
1275 | if (bidi_it->nchars <= 0) |
1276 | abort (); | |
182ce2d2 | 1277 | bidi_it->charpos += bidi_it->nchars; |
e7402cb2 EZ |
1278 | if (bidi_it->ch_len == 0) |
1279 | abort (); | |
b7b65b15 EZ |
1280 | bidi_it->bytepos += bidi_it->ch_len; |
1281 | } | |
1282 | ||
1283 | current_level = bidi_it->level_stack[bidi_it->stack_idx].level; /* X1 */ | |
1284 | override = bidi_it->level_stack[bidi_it->stack_idx].override; | |
1285 | new_level = current_level; | |
1286 | ||
87e67904 | 1287 | if (bidi_it->charpos >= (string_p ? bidi_it->string.schars : ZV)) |
e7402cb2 EZ |
1288 | { |
1289 | curchar = BIDI_EOB; | |
1290 | bidi_it->ch_len = 1; | |
182ce2d2 | 1291 | bidi_it->nchars = 1; |
87e67904 | 1292 | bidi_it->disp_pos = (string_p ? bidi_it->string.schars : ZV); |
e7402cb2 EZ |
1293 | } |
1294 | else | |
1295 | { | |
182ce2d2 EZ |
1296 | /* Fetch the character at BYTEPOS. If it is covered by a |
1297 | display string, treat the entire run of covered characters as | |
1298 | a single character u+FFFC. */ | |
102ebb00 | 1299 | curchar = bidi_fetch_char (bidi_it->bytepos, bidi_it->charpos, |
87e67904 EZ |
1300 | &bidi_it->disp_pos, &bidi_it->string, |
1301 | bidi_it->frame_window_p, | |
102ebb00 | 1302 | &bidi_it->ch_len, &bidi_it->nchars); |
e7402cb2 | 1303 | } |
b7b65b15 | 1304 | bidi_it->ch = curchar; |
b7b65b15 | 1305 | |
6bff6497 EZ |
1306 | /* Don't apply directional override here, as all the types we handle |
1307 | below will not be affected by the override anyway, and we need | |
1308 | the original type unaltered. The override will be applied in | |
1309 | bidi_resolve_weak. */ | |
1310 | type = bidi_get_type (curchar, NEUTRAL_DIR); | |
89d3374a EZ |
1311 | bidi_it->orig_type = type; |
1312 | bidi_check_type (bidi_it->orig_type); | |
b7b65b15 EZ |
1313 | |
1314 | if (type != PDF) | |
1315 | bidi_it->prev_was_pdf = 0; | |
1316 | ||
89d3374a | 1317 | bidi_it->type_after_w1 = UNKNOWN_BT; |
b7b65b15 EZ |
1318 | |
1319 | switch (type) | |
1320 | { | |
1321 | case RLE: /* X2 */ | |
1322 | case RLO: /* X4 */ | |
89d3374a EZ |
1323 | bidi_it->type_after_w1 = type; |
1324 | bidi_check_type (bidi_it->type_after_w1); | |
b7b65b15 | 1325 | type = WEAK_BN; /* X9/Retaining */ |
87e67904 | 1326 | if (bidi_it->ignore_bn_limit <= -1) |
b7b65b15 EZ |
1327 | { |
1328 | if (current_level <= BIDI_MAXLEVEL - 4) | |
1329 | { | |
1330 | /* Compute the least odd embedding level greater than | |
1331 | the current level. */ | |
1332 | new_level = ((current_level + 1) & ~1) + 1; | |
89d3374a | 1333 | if (bidi_it->type_after_w1 == RLE) |
b7b65b15 EZ |
1334 | override = NEUTRAL_DIR; |
1335 | else | |
1336 | override = R2L; | |
1337 | if (current_level == BIDI_MAXLEVEL - 4) | |
1338 | bidi_it->invalid_rl_levels = 0; | |
1339 | bidi_push_embedding_level (bidi_it, new_level, override); | |
1340 | } | |
1341 | else | |
1342 | { | |
1343 | bidi_it->invalid_levels++; | |
1344 | /* See the commentary about invalid_rl_levels below. */ | |
1345 | if (bidi_it->invalid_rl_levels < 0) | |
1346 | bidi_it->invalid_rl_levels = 0; | |
1347 | bidi_it->invalid_rl_levels++; | |
1348 | } | |
1349 | } | |
89d3374a | 1350 | else if (bidi_it->prev.type_after_w1 == WEAK_EN /* W5/Retaining */ |
b7b65b15 EZ |
1351 | || bidi_it->next_en_pos > bidi_it->charpos) |
1352 | type = WEAK_EN; | |
1353 | break; | |
1354 | case LRE: /* X3 */ | |
1355 | case LRO: /* X5 */ | |
89d3374a EZ |
1356 | bidi_it->type_after_w1 = type; |
1357 | bidi_check_type (bidi_it->type_after_w1); | |
b7b65b15 | 1358 | type = WEAK_BN; /* X9/Retaining */ |
87e67904 | 1359 | if (bidi_it->ignore_bn_limit <= -1) |
b7b65b15 EZ |
1360 | { |
1361 | if (current_level <= BIDI_MAXLEVEL - 5) | |
1362 | { | |
1363 | /* Compute the least even embedding level greater than | |
1364 | the current level. */ | |
1365 | new_level = ((current_level + 2) & ~1); | |
89d3374a | 1366 | if (bidi_it->type_after_w1 == LRE) |
b7b65b15 EZ |
1367 | override = NEUTRAL_DIR; |
1368 | else | |
1369 | override = L2R; | |
1370 | bidi_push_embedding_level (bidi_it, new_level, override); | |
1371 | } | |
1372 | else | |
1373 | { | |
1374 | bidi_it->invalid_levels++; | |
1375 | /* invalid_rl_levels counts invalid levels encountered | |
1376 | while the embedding level was already too high for | |
1377 | LRE/LRO, but not for RLE/RLO. That is because | |
1378 | there may be exactly one PDF which we should not | |
1379 | ignore even though invalid_levels is non-zero. | |
1380 | invalid_rl_levels helps to know what PDF is | |
1381 | that. */ | |
1382 | if (bidi_it->invalid_rl_levels >= 0) | |
1383 | bidi_it->invalid_rl_levels++; | |
1384 | } | |
1385 | } | |
89d3374a | 1386 | else if (bidi_it->prev.type_after_w1 == WEAK_EN /* W5/Retaining */ |
b7b65b15 EZ |
1387 | || bidi_it->next_en_pos > bidi_it->charpos) |
1388 | type = WEAK_EN; | |
1389 | break; | |
1390 | case PDF: /* X7 */ | |
89d3374a EZ |
1391 | bidi_it->type_after_w1 = type; |
1392 | bidi_check_type (bidi_it->type_after_w1); | |
b7b65b15 | 1393 | type = WEAK_BN; /* X9/Retaining */ |
87e67904 | 1394 | if (bidi_it->ignore_bn_limit <= -1) |
b7b65b15 EZ |
1395 | { |
1396 | if (!bidi_it->invalid_rl_levels) | |
1397 | { | |
1398 | new_level = bidi_pop_embedding_level (bidi_it); | |
1399 | bidi_it->invalid_rl_levels = -1; | |
1400 | if (bidi_it->invalid_levels) | |
1401 | bidi_it->invalid_levels--; | |
1402 | /* else nothing: UAX#9 says to ignore invalid PDFs */ | |
1403 | } | |
1404 | if (!bidi_it->invalid_levels) | |
1405 | new_level = bidi_pop_embedding_level (bidi_it); | |
1406 | else | |
1407 | { | |
1408 | bidi_it->invalid_levels--; | |
1409 | bidi_it->invalid_rl_levels--; | |
1410 | } | |
1411 | } | |
89d3374a | 1412 | else if (bidi_it->prev.type_after_w1 == WEAK_EN /* W5/Retaining */ |
b7b65b15 EZ |
1413 | || bidi_it->next_en_pos > bidi_it->charpos) |
1414 | type = WEAK_EN; | |
1415 | break; | |
1416 | default: | |
1417 | /* Nothing. */ | |
1418 | break; | |
1419 | } | |
1420 | ||
1421 | bidi_it->type = type; | |
2d6e4628 | 1422 | bidi_check_type (bidi_it->type); |
b7b65b15 EZ |
1423 | |
1424 | return new_level; | |
1425 | } | |
1426 | ||
1427 | /* Given an iterator state in BIDI_IT, advance one character position | |
182ce2d2 EZ |
1428 | in the buffer/string to the next character (in the logical order), |
1429 | resolve any explicit embeddings and directional overrides, and | |
1430 | return the embedding level of the character after resolving | |
1431 | explicit directives and ignoring empty embeddings. */ | |
b7b65b15 EZ |
1432 | static int |
1433 | bidi_resolve_explicit (struct bidi_it *bidi_it) | |
1434 | { | |
1435 | int prev_level = bidi_it->level_stack[bidi_it->stack_idx].level; | |
1436 | int new_level = bidi_resolve_explicit_1 (bidi_it); | |
87e67904 | 1437 | EMACS_INT eob = bidi_it->string.s ? bidi_it->string.schars : ZV; |
9f257352 EZ |
1438 | const unsigned char *s = STRINGP (bidi_it->string.lstring) |
1439 | ? SDATA (bidi_it->string.lstring) : bidi_it->string.s; | |
b7b65b15 EZ |
1440 | |
1441 | if (prev_level < new_level | |
1442 | && bidi_it->type == WEAK_BN | |
87e67904 EZ |
1443 | && bidi_it->ignore_bn_limit == -1 /* only if not already known */ |
1444 | && bidi_it->charpos < eob /* not already at EOB */ | |
1445 | && bidi_explicit_dir_char (bidi_char_at_pos (bidi_it->bytepos | |
f3014ef5 EZ |
1446 | + bidi_it->ch_len, s, |
1447 | bidi_it->string.unibyte))) | |
b7b65b15 EZ |
1448 | { |
1449 | /* Avoid pushing and popping embedding levels if the level run | |
1450 | is empty, as this breaks level runs where it shouldn't. | |
1451 | UAX#9 removes all the explicit embedding and override codes, | |
1452 | so empty embeddings disappear without a trace. We need to | |
1453 | behave as if we did the same. */ | |
1454 | struct bidi_it saved_it; | |
1455 | int level = prev_level; | |
1456 | ||
1457 | bidi_copy_it (&saved_it, bidi_it); | |
1458 | ||
87e67904 | 1459 | while (bidi_explicit_dir_char (bidi_char_at_pos (bidi_it->bytepos |
f3014ef5 EZ |
1460 | + bidi_it->ch_len, s, |
1461 | bidi_it->string.unibyte))) | |
b7b65b15 | 1462 | { |
182ce2d2 EZ |
1463 | /* This advances to the next character, skipping any |
1464 | characters covered by display strings. */ | |
b7b65b15 | 1465 | level = bidi_resolve_explicit_1 (bidi_it); |
9f257352 EZ |
1466 | /* If string.lstring was relocated inside bidi_resolve_explicit_1, |
1467 | a pointer to its data is no longer valid. */ | |
1468 | if (STRINGP (bidi_it->string.lstring)) | |
1469 | s = SDATA (bidi_it->string.lstring); | |
b7b65b15 EZ |
1470 | } |
1471 | ||
102ebb00 EZ |
1472 | if (bidi_it->nchars <= 0) |
1473 | abort (); | |
b7b65b15 | 1474 | if (level == prev_level) /* empty embedding */ |
182ce2d2 | 1475 | saved_it.ignore_bn_limit = bidi_it->charpos + bidi_it->nchars; |
b7b65b15 | 1476 | else /* this embedding is non-empty */ |
87e67904 | 1477 | saved_it.ignore_bn_limit = -2; |
b7b65b15 EZ |
1478 | |
1479 | bidi_copy_it (bidi_it, &saved_it); | |
87e67904 | 1480 | if (bidi_it->ignore_bn_limit > -1) |
b7b65b15 EZ |
1481 | { |
1482 | /* We pushed a level, but we shouldn't have. Undo that. */ | |
1483 | if (!bidi_it->invalid_rl_levels) | |
1484 | { | |
1485 | new_level = bidi_pop_embedding_level (bidi_it); | |
1486 | bidi_it->invalid_rl_levels = -1; | |
1487 | if (bidi_it->invalid_levels) | |
1488 | bidi_it->invalid_levels--; | |
1489 | } | |
1490 | if (!bidi_it->invalid_levels) | |
1491 | new_level = bidi_pop_embedding_level (bidi_it); | |
1492 | else | |
1493 | { | |
1494 | bidi_it->invalid_levels--; | |
1495 | bidi_it->invalid_rl_levels--; | |
1496 | } | |
1497 | } | |
1498 | } | |
1499 | ||
b7b65b15 EZ |
1500 | if (bidi_it->type == NEUTRAL_B) /* X8 */ |
1501 | { | |
21fce5ab | 1502 | bidi_set_paragraph_end (bidi_it); |
6bff6497 EZ |
1503 | /* This is needed by bidi_resolve_weak below, and in L1. */ |
1504 | bidi_it->type_after_w1 = bidi_it->type; | |
89d3374a | 1505 | bidi_check_type (bidi_it->type_after_w1); |
b7b65b15 EZ |
1506 | } |
1507 | ||
1508 | return new_level; | |
1509 | } | |
1510 | ||
182ce2d2 EZ |
1511 | /* Advance in the buffer/string, resolve weak types and return the |
1512 | type of the next character after weak type resolution. */ | |
fd3998ff | 1513 | static bidi_type_t |
b7b65b15 EZ |
1514 | bidi_resolve_weak (struct bidi_it *bidi_it) |
1515 | { | |
1516 | bidi_type_t type; | |
1517 | bidi_dir_t override; | |
1518 | int prev_level = bidi_it->level_stack[bidi_it->stack_idx].level; | |
1519 | int new_level = bidi_resolve_explicit (bidi_it); | |
1520 | int next_char; | |
1521 | bidi_type_t type_of_next; | |
1522 | struct bidi_it saved_it; | |
9f257352 EZ |
1523 | EMACS_INT eob = |
1524 | (STRINGP (bidi_it->string.lstring) || bidi_it->string.s) | |
1525 | ? bidi_it->string.schars : ZV; | |
b7b65b15 EZ |
1526 | |
1527 | type = bidi_it->type; | |
1528 | override = bidi_it->level_stack[bidi_it->stack_idx].override; | |
1529 | ||
1530 | if (type == UNKNOWN_BT | |
1531 | || type == LRE | |
1532 | || type == LRO | |
1533 | || type == RLE | |
1534 | || type == RLO | |
1535 | || type == PDF) | |
1536 | abort (); | |
1537 | ||
1538 | if (new_level != prev_level | |
1539 | || bidi_it->type == NEUTRAL_B) | |
1540 | { | |
1541 | /* We've got a new embedding level run, compute the directional | |
1542 | type of sor and initialize per-run variables (UAX#9, clause | |
1543 | X10). */ | |
1544 | bidi_set_sor_type (bidi_it, prev_level, new_level); | |
1545 | } | |
1546 | else if (type == NEUTRAL_S || type == NEUTRAL_WS | |
1547 | || type == WEAK_BN || type == STRONG_AL) | |
89d3374a EZ |
1548 | bidi_it->type_after_w1 = type; /* needed in L1 */ |
1549 | bidi_check_type (bidi_it->type_after_w1); | |
b7b65b15 EZ |
1550 | |
1551 | /* Level and directional override status are already recorded in | |
1552 | bidi_it, and do not need any change; see X6. */ | |
1553 | if (override == R2L) /* X6 */ | |
1554 | type = STRONG_R; | |
1555 | else if (override == L2R) | |
1556 | type = STRONG_L; | |
bc5a45f3 | 1557 | else |
b7b65b15 | 1558 | { |
bc5a45f3 | 1559 | if (type == WEAK_NSM) /* W1 */ |
b7b65b15 | 1560 | { |
bc5a45f3 | 1561 | /* Note that we don't need to consider the case where the |
5930fe97 EZ |
1562 | prev character has its type overridden by an RLO or LRO, |
1563 | because then either the type of this NSM would have been | |
1564 | also overridden, or the previous character is outside the | |
1565 | current level run, and thus not relevant to this NSM. | |
1566 | This is why NSM gets the type_after_w1 of the previous | |
1567 | character. */ | |
ebb5722e EZ |
1568 | if (bidi_it->prev.type_after_w1 != UNKNOWN_BT |
1569 | /* if type_after_w1 is NEUTRAL_B, this NSM is at sor */ | |
1570 | && bidi_it->prev.type_after_w1 != NEUTRAL_B) | |
5930fe97 | 1571 | type = bidi_it->prev.type_after_w1; |
bc5a45f3 EZ |
1572 | else if (bidi_it->sor == R2L) |
1573 | type = STRONG_R; | |
1574 | else if (bidi_it->sor == L2R) | |
1575 | type = STRONG_L; | |
1576 | else /* shouldn't happen! */ | |
1577 | abort (); | |
b7b65b15 | 1578 | } |
bc5a45f3 EZ |
1579 | if (type == WEAK_EN /* W2 */ |
1580 | && bidi_it->last_strong.type_after_w1 == STRONG_AL) | |
1581 | type = WEAK_AN; | |
1582 | else if (type == STRONG_AL) /* W3 */ | |
1583 | type = STRONG_R; | |
1584 | else if ((type == WEAK_ES /* W4 */ | |
1585 | && bidi_it->prev.type_after_w1 == WEAK_EN | |
1586 | && bidi_it->prev.orig_type == WEAK_EN) | |
1587 | || (type == WEAK_CS | |
1588 | && ((bidi_it->prev.type_after_w1 == WEAK_EN | |
1589 | && bidi_it->prev.orig_type == WEAK_EN) | |
1590 | || bidi_it->prev.type_after_w1 == WEAK_AN))) | |
b7b65b15 | 1591 | { |
9f257352 EZ |
1592 | const unsigned char *s = |
1593 | STRINGP (bidi_it->string.lstring) | |
1594 | ? SDATA (bidi_it->string.lstring) : bidi_it->string.s; | |
1595 | ||
e7402cb2 | 1596 | next_char = |
87e67904 EZ |
1597 | bidi_it->charpos + bidi_it->nchars >= eob |
1598 | ? BIDI_EOB | |
f3014ef5 EZ |
1599 | : bidi_char_at_pos (bidi_it->bytepos + bidi_it->ch_len, s, |
1600 | bidi_it->string.unibyte); | |
6bff6497 | 1601 | type_of_next = bidi_get_type (next_char, override); |
b7b65b15 | 1602 | |
bc5a45f3 | 1603 | if (type_of_next == WEAK_BN |
b7b65b15 EZ |
1604 | || bidi_explicit_dir_char (next_char)) |
1605 | { | |
1606 | bidi_copy_it (&saved_it, bidi_it); | |
1607 | while (bidi_resolve_explicit (bidi_it) == new_level | |
bc5a45f3 | 1608 | && bidi_it->type == WEAK_BN) |
b7b65b15 EZ |
1609 | ; |
1610 | type_of_next = bidi_it->type; | |
b7b65b15 EZ |
1611 | bidi_copy_it (bidi_it, &saved_it); |
1612 | } | |
bc5a45f3 EZ |
1613 | |
1614 | /* If the next character is EN, but the last strong-type | |
1615 | character is AL, that next EN will be changed to AN when | |
1616 | we process it in W2 above. So in that case, this ES | |
1617 | should not be changed into EN. */ | |
1618 | if (type == WEAK_ES | |
1619 | && type_of_next == WEAK_EN | |
1620 | && bidi_it->last_strong.type_after_w1 != STRONG_AL) | |
1621 | type = WEAK_EN; | |
1622 | else if (type == WEAK_CS) | |
b7b65b15 | 1623 | { |
bc5a45f3 EZ |
1624 | if (bidi_it->prev.type_after_w1 == WEAK_AN |
1625 | && (type_of_next == WEAK_AN | |
1626 | /* If the next character is EN, but the last | |
1627 | strong-type character is AL, EN will be later | |
1628 | changed to AN when we process it in W2 above. | |
1629 | So in that case, this ES should not be | |
1630 | changed into EN. */ | |
1631 | || (type_of_next == WEAK_EN | |
1632 | && bidi_it->last_strong.type_after_w1 == STRONG_AL))) | |
1633 | type = WEAK_AN; | |
1634 | else if (bidi_it->prev.type_after_w1 == WEAK_EN | |
1635 | && type_of_next == WEAK_EN | |
1636 | && bidi_it->last_strong.type_after_w1 != STRONG_AL) | |
1637 | type = WEAK_EN; | |
1638 | } | |
1639 | } | |
1640 | else if (type == WEAK_ET /* W5: ET with EN before or after it */ | |
1641 | || type == WEAK_BN) /* W5/Retaining */ | |
1642 | { | |
1643 | if (bidi_it->prev.type_after_w1 == WEAK_EN /* ET/BN w/EN before it */ | |
1644 | || bidi_it->next_en_pos > bidi_it->charpos) | |
1645 | type = WEAK_EN; | |
1646 | else /* W5: ET/BN with EN after it. */ | |
1647 | { | |
182ce2d2 | 1648 | EMACS_INT en_pos = bidi_it->charpos + bidi_it->nchars; |
9f257352 EZ |
1649 | const unsigned char *s = |
1650 | STRINGP (bidi_it->string.lstring) | |
1651 | ? SDATA (bidi_it->string.lstring) : bidi_it->string.s; | |
bc5a45f3 | 1652 | |
102ebb00 EZ |
1653 | if (bidi_it->nchars <= 0) |
1654 | abort (); | |
bc5a45f3 | 1655 | next_char = |
87e67904 EZ |
1656 | bidi_it->charpos + bidi_it->nchars >= eob |
1657 | ? BIDI_EOB | |
f3014ef5 EZ |
1658 | : bidi_char_at_pos (bidi_it->bytepos + bidi_it->ch_len, s, |
1659 | bidi_it->string.unibyte); | |
bc5a45f3 EZ |
1660 | type_of_next = bidi_get_type (next_char, override); |
1661 | ||
1662 | if (type_of_next == WEAK_ET | |
1663 | || type_of_next == WEAK_BN | |
1664 | || bidi_explicit_dir_char (next_char)) | |
1665 | { | |
1666 | bidi_copy_it (&saved_it, bidi_it); | |
1667 | while (bidi_resolve_explicit (bidi_it) == new_level | |
1668 | && (bidi_it->type == WEAK_BN | |
1669 | || bidi_it->type == WEAK_ET)) | |
1670 | ; | |
1671 | type_of_next = bidi_it->type; | |
1672 | en_pos = bidi_it->charpos; | |
1673 | bidi_copy_it (bidi_it, &saved_it); | |
1674 | } | |
1675 | if (type_of_next == WEAK_EN) | |
b7b65b15 | 1676 | { |
bc5a45f3 EZ |
1677 | /* If the last strong character is AL, the EN we've |
1678 | found will become AN when we get to it (W2). */ | |
1679 | if (bidi_it->last_strong.type_after_w1 != STRONG_AL) | |
1680 | { | |
1681 | type = WEAK_EN; | |
1682 | /* Remember this EN position, to speed up processing | |
1683 | of the next ETs. */ | |
1684 | bidi_it->next_en_pos = en_pos; | |
1685 | } | |
1686 | else if (type == WEAK_BN) | |
1687 | type = NEUTRAL_ON; /* W6/Retaining */ | |
b7b65b15 | 1688 | } |
b7b65b15 EZ |
1689 | } |
1690 | } | |
1691 | } | |
1692 | ||
1693 | if (type == WEAK_ES || type == WEAK_ET || type == WEAK_CS /* W6 */ | |
89d3374a EZ |
1694 | || (type == WEAK_BN |
1695 | && (bidi_it->prev.type_after_w1 == WEAK_CS /* W6/Retaining */ | |
1696 | || bidi_it->prev.type_after_w1 == WEAK_ES | |
1697 | || bidi_it->prev.type_after_w1 == WEAK_ET))) | |
b7b65b15 EZ |
1698 | type = NEUTRAL_ON; |
1699 | ||
1700 | /* Store the type we've got so far, before we clobber it with strong | |
1701 | types in W7 and while resolving neutral types. But leave alone | |
1702 | the original types that were recorded above, because we will need | |
1703 | them for the L1 clause. */ | |
89d3374a EZ |
1704 | if (bidi_it->type_after_w1 == UNKNOWN_BT) |
1705 | bidi_it->type_after_w1 = type; | |
1706 | bidi_check_type (bidi_it->type_after_w1); | |
b7b65b15 EZ |
1707 | |
1708 | if (type == WEAK_EN) /* W7 */ | |
1709 | { | |
89d3374a | 1710 | if ((bidi_it->last_strong.type_after_w1 == STRONG_L) |
b7b65b15 EZ |
1711 | || (bidi_it->last_strong.type == UNKNOWN_BT && bidi_it->sor == L2R)) |
1712 | type = STRONG_L; | |
1713 | } | |
1714 | ||
1715 | bidi_it->type = type; | |
2d6e4628 | 1716 | bidi_check_type (bidi_it->type); |
b7b65b15 EZ |
1717 | return type; |
1718 | } | |
1719 | ||
cbb09f04 EZ |
1720 | /* Resolve the type of a neutral character according to the type of |
1721 | surrounding strong text and the current embedding level. */ | |
58b9f433 | 1722 | static inline bidi_type_t |
cbb09f04 EZ |
1723 | bidi_resolve_neutral_1 (bidi_type_t prev_type, bidi_type_t next_type, int lev) |
1724 | { | |
1725 | /* N1: European and Arabic numbers are treated as though they were R. */ | |
1726 | if (next_type == WEAK_EN || next_type == WEAK_AN) | |
1727 | next_type = STRONG_R; | |
1728 | if (prev_type == WEAK_EN || prev_type == WEAK_AN) | |
1729 | prev_type = STRONG_R; | |
1730 | ||
1731 | if (next_type == prev_type) /* N1 */ | |
1732 | return next_type; | |
1733 | else if ((lev & 1) == 0) /* N2 */ | |
1734 | return STRONG_L; | |
1735 | else | |
1736 | return STRONG_R; | |
1737 | } | |
1738 | ||
fd3998ff | 1739 | static bidi_type_t |
b7b65b15 EZ |
1740 | bidi_resolve_neutral (struct bidi_it *bidi_it) |
1741 | { | |
1742 | int prev_level = bidi_it->level_stack[bidi_it->stack_idx].level; | |
1743 | bidi_type_t type = bidi_resolve_weak (bidi_it); | |
1744 | int current_level = bidi_it->level_stack[bidi_it->stack_idx].level; | |
1745 | ||
1746 | if (!(type == STRONG_R | |
1747 | || type == STRONG_L | |
1748 | || type == WEAK_BN | |
1749 | || type == WEAK_EN | |
1750 | || type == WEAK_AN | |
1751 | || type == NEUTRAL_B | |
1752 | || type == NEUTRAL_S | |
1753 | || type == NEUTRAL_WS | |
1754 | || type == NEUTRAL_ON)) | |
1755 | abort (); | |
1756 | ||
1757 | if (bidi_get_category (type) == NEUTRAL | |
1758 | || (type == WEAK_BN && prev_level == current_level)) | |
1759 | { | |
1760 | if (bidi_it->next_for_neutral.type != UNKNOWN_BT) | |
1761 | type = bidi_resolve_neutral_1 (bidi_it->prev_for_neutral.type, | |
1762 | bidi_it->next_for_neutral.type, | |
1763 | current_level); | |
1764 | else | |
1765 | { | |
1766 | /* Arrrgh!! The UAX#9 algorithm is too deeply entrenched in | |
1767 | the assumption of batch-style processing; see clauses W4, | |
1768 | W5, and especially N1, which require to look far forward | |
182ce2d2 EZ |
1769 | (as well as back) in the buffer/string. May the fleas of |
1770 | a thousand camels infest the armpits of those who design | |
b7b65b15 EZ |
1771 | supposedly general-purpose algorithms by looking at their |
1772 | own implementations, and fail to consider other possible | |
1773 | implementations! */ | |
1774 | struct bidi_it saved_it; | |
1775 | bidi_type_t next_type; | |
1776 | ||
1777 | if (bidi_it->scan_dir == -1) | |
1778 | abort (); | |
1779 | ||
1780 | bidi_copy_it (&saved_it, bidi_it); | |
1781 | /* Scan the text forward until we find the first non-neutral | |
1782 | character, and then use that to resolve the neutral we | |
1783 | are dealing with now. We also cache the scanned iterator | |
1784 | states, to salvage some of the effort later. */ | |
1785 | bidi_cache_iterator_state (bidi_it, 0); | |
1786 | do { | |
1787 | /* Record the info about the previous character, so that | |
1788 | it will be cached below with this state. */ | |
89d3374a | 1789 | if (bidi_it->type_after_w1 != WEAK_BN /* W1/Retaining */ |
b7b65b15 EZ |
1790 | && bidi_it->type != WEAK_BN) |
1791 | bidi_remember_char (&bidi_it->prev, bidi_it); | |
1792 | type = bidi_resolve_weak (bidi_it); | |
1793 | /* Paragraph separators have their levels fully resolved | |
1794 | at this point, so cache them as resolved. */ | |
1795 | bidi_cache_iterator_state (bidi_it, type == NEUTRAL_B); | |
1796 | /* FIXME: implement L1 here, by testing for a newline and | |
1797 | resetting the level for any sequence of whitespace | |
1798 | characters adjacent to it. */ | |
1799 | } while (!(type == NEUTRAL_B | |
1800 | || (type != WEAK_BN | |
1801 | && bidi_get_category (type) != NEUTRAL) | |
1802 | /* This is all per level run, so stop when we | |
1803 | reach the end of this level run. */ | |
1804 | || bidi_it->level_stack[bidi_it->stack_idx].level != | |
1805 | current_level)); | |
1806 | ||
1807 | bidi_remember_char (&saved_it.next_for_neutral, bidi_it); | |
1808 | ||
1809 | switch (type) | |
1810 | { | |
1811 | case STRONG_L: | |
1812 | case STRONG_R: | |
1813 | case STRONG_AL: | |
1814 | next_type = type; | |
1815 | break; | |
1816 | case WEAK_EN: | |
1817 | case WEAK_AN: | |
1818 | /* N1: ``European and Arabic numbers are treated as | |
1819 | though they were R.'' */ | |
1820 | next_type = STRONG_R; | |
1821 | saved_it.next_for_neutral.type = STRONG_R; | |
1822 | break; | |
1823 | case WEAK_BN: | |
1824 | if (!bidi_explicit_dir_char (bidi_it->ch)) | |
1825 | abort (); /* can't happen: BNs are skipped */ | |
1826 | /* FALLTHROUGH */ | |
1827 | case NEUTRAL_B: | |
1828 | /* Marched all the way to the end of this level run. | |
1829 | We need to use the eor type, whose information is | |
1830 | stored by bidi_set_sor_type in the prev_for_neutral | |
1831 | member. */ | |
1832 | if (saved_it.type != WEAK_BN | |
89d3374a | 1833 | || bidi_get_category (bidi_it->prev.type_after_w1) == NEUTRAL) |
b7b65b15 EZ |
1834 | { |
1835 | next_type = bidi_it->prev_for_neutral.type; | |
1836 | saved_it.next_for_neutral.type = next_type; | |
2d6e4628 | 1837 | bidi_check_type (next_type); |
b7b65b15 EZ |
1838 | } |
1839 | else | |
1840 | { | |
1841 | /* This is a BN which does not adjoin neutrals. | |
1842 | Leave its type alone. */ | |
1843 | bidi_copy_it (bidi_it, &saved_it); | |
1844 | return bidi_it->type; | |
1845 | } | |
1846 | break; | |
1847 | default: | |
1848 | abort (); | |
1849 | } | |
1850 | type = bidi_resolve_neutral_1 (saved_it.prev_for_neutral.type, | |
1851 | next_type, current_level); | |
1852 | saved_it.type = type; | |
2d6e4628 | 1853 | bidi_check_type (type); |
b7b65b15 EZ |
1854 | bidi_copy_it (bidi_it, &saved_it); |
1855 | } | |
1856 | } | |
1857 | return type; | |
1858 | } | |
1859 | ||
1860 | /* Given an iterator state in BIDI_IT, advance one character position | |
182ce2d2 EZ |
1861 | in the buffer/string to the next character (in the logical order), |
1862 | resolve the bidi type of that next character, and return that | |
1863 | type. */ | |
fd3998ff | 1864 | static bidi_type_t |
b7b65b15 EZ |
1865 | bidi_type_of_next_char (struct bidi_it *bidi_it) |
1866 | { | |
1867 | bidi_type_t type; | |
1868 | ||
1869 | /* This should always be called during a forward scan. */ | |
1870 | if (bidi_it->scan_dir != 1) | |
1871 | abort (); | |
1872 | ||
1873 | /* Reset the limit until which to ignore BNs if we step out of the | |
1874 | area where we found only empty levels. */ | |
87e67904 | 1875 | if ((bidi_it->ignore_bn_limit > -1 |
b7b65b15 | 1876 | && bidi_it->ignore_bn_limit <= bidi_it->charpos) |
87e67904 | 1877 | || (bidi_it->ignore_bn_limit == -2 |
b7b65b15 | 1878 | && !bidi_explicit_dir_char (bidi_it->ch))) |
87e67904 | 1879 | bidi_it->ignore_bn_limit = -1; |
b7b65b15 EZ |
1880 | |
1881 | type = bidi_resolve_neutral (bidi_it); | |
1882 | ||
1883 | return type; | |
1884 | } | |
1885 | ||
1886 | /* Given an iterator state BIDI_IT, advance one character position in | |
182ce2d2 EZ |
1887 | the buffer/string to the next character (in the current scan |
1888 | direction), resolve the embedding and implicit levels of that next | |
1889 | character, and return the resulting level. */ | |
fd3998ff | 1890 | static int |
b7b65b15 EZ |
1891 | bidi_level_of_next_char (struct bidi_it *bidi_it) |
1892 | { | |
1893 | bidi_type_t type; | |
1894 | int level, prev_level = -1; | |
1895 | struct bidi_saved_info next_for_neutral; | |
578b494e | 1896 | EMACS_INT next_char_pos = -2; |
b7b65b15 EZ |
1897 | |
1898 | if (bidi_it->scan_dir == 1) | |
1899 | { | |
9f257352 EZ |
1900 | EMACS_INT eob = |
1901 | (bidi_it->string.s || STRINGP (bidi_it->string.lstring)) | |
1902 | ? bidi_it->string.schars : ZV; | |
1903 | ||
b7b65b15 | 1904 | /* There's no sense in trying to advance if we hit end of text. */ |
9f257352 | 1905 | if (bidi_it->charpos >= eob) |
b7b65b15 EZ |
1906 | return bidi_it->resolved_level; |
1907 | ||
1908 | /* Record the info about the previous character. */ | |
89d3374a | 1909 | if (bidi_it->type_after_w1 != WEAK_BN /* W1/Retaining */ |
b7b65b15 EZ |
1910 | && bidi_it->type != WEAK_BN) |
1911 | bidi_remember_char (&bidi_it->prev, bidi_it); | |
89d3374a EZ |
1912 | if (bidi_it->type_after_w1 == STRONG_R |
1913 | || bidi_it->type_after_w1 == STRONG_L | |
1914 | || bidi_it->type_after_w1 == STRONG_AL) | |
b7b65b15 EZ |
1915 | bidi_remember_char (&bidi_it->last_strong, bidi_it); |
1916 | /* FIXME: it sounds like we don't need both prev and | |
1917 | prev_for_neutral members, but I'm leaving them both for now. */ | |
1918 | if (bidi_it->type == STRONG_R || bidi_it->type == STRONG_L | |
1919 | || bidi_it->type == WEAK_EN || bidi_it->type == WEAK_AN) | |
1920 | bidi_remember_char (&bidi_it->prev_for_neutral, bidi_it); | |
1921 | ||
1922 | /* If we overstepped the characters used for resolving neutrals | |
1923 | and whitespace, invalidate their info in the iterator. */ | |
1924 | if (bidi_it->charpos >= bidi_it->next_for_neutral.charpos) | |
1925 | bidi_it->next_for_neutral.type = UNKNOWN_BT; | |
1926 | if (bidi_it->next_en_pos >= 0 | |
1927 | && bidi_it->charpos >= bidi_it->next_en_pos) | |
1928 | bidi_it->next_en_pos = -1; | |
1929 | if (bidi_it->next_for_ws.type != UNKNOWN_BT | |
1930 | && bidi_it->charpos >= bidi_it->next_for_ws.charpos) | |
1931 | bidi_it->next_for_ws.type = UNKNOWN_BT; | |
1932 | ||
1933 | /* This must be taken before we fill the iterator with the info | |
1934 | about the next char. If we scan backwards, the iterator | |
1935 | state must be already cached, so there's no need to know the | |
1936 | embedding level of the previous character, since we will be | |
1937 | returning to our caller shortly. */ | |
1938 | prev_level = bidi_it->level_stack[bidi_it->stack_idx].level; | |
1939 | } | |
1940 | next_for_neutral = bidi_it->next_for_neutral; | |
1941 | ||
182ce2d2 EZ |
1942 | /* Perhaps the character we want is already cached. If it is, the |
1943 | call to bidi_cache_find below will return a type other than | |
1944 | UNKNOWN_BT. */ | |
87e67904 | 1945 | if (bidi_cache_idx > bidi_cache_start && !bidi_it->first_elt) |
102ebb00 | 1946 | { |
9f257352 EZ |
1947 | int bob = |
1948 | (bidi_it->string.s || STRINGP (bidi_it->string.lstring)) ? 0 : 1; | |
1949 | ||
102ebb00 EZ |
1950 | if (bidi_it->scan_dir > 0) |
1951 | { | |
1952 | if (bidi_it->nchars <= 0) | |
1953 | abort (); | |
1954 | next_char_pos = bidi_it->charpos + bidi_it->nchars; | |
1955 | } | |
9f257352 | 1956 | else if (bidi_it->charpos >= bob) |
bb269206 EZ |
1957 | /* Implementation note: we allow next_char_pos to be as low as |
1958 | 0 for buffers or -1 for strings, and that is okay because | |
1959 | that's the "position" of the sentinel iterator state we | |
1960 | cached at the beginning of the iteration. */ | |
102ebb00 | 1961 | next_char_pos = bidi_it->charpos - 1; |
578b494e | 1962 | if (next_char_pos >= bob - 1) |
87e67904 EZ |
1963 | type = bidi_cache_find (next_char_pos, -1, bidi_it); |
1964 | else | |
1965 | type = UNKNOWN_BT; | |
102ebb00 | 1966 | } |
182ce2d2 | 1967 | else |
102ebb00 | 1968 | type = UNKNOWN_BT; |
b7b65b15 EZ |
1969 | if (type != UNKNOWN_BT) |
1970 | { | |
1971 | /* Don't lose the information for resolving neutrals! The | |
1972 | cached states could have been cached before their | |
1973 | next_for_neutral member was computed. If we are on our way | |
1974 | forward, we can simply take the info from the previous | |
1975 | state. */ | |
1976 | if (bidi_it->scan_dir == 1 | |
1977 | && bidi_it->next_for_neutral.type == UNKNOWN_BT) | |
1978 | bidi_it->next_for_neutral = next_for_neutral; | |
1979 | ||
1980 | /* If resolved_level is -1, it means this state was cached | |
1981 | before it was completely resolved, so we cannot return | |
1982 | it. */ | |
1983 | if (bidi_it->resolved_level != -1) | |
1984 | return bidi_it->resolved_level; | |
1985 | } | |
1986 | if (bidi_it->scan_dir == -1) | |
1987 | /* If we are going backwards, the iterator state is already cached | |
1988 | from previous scans, and should be fully resolved. */ | |
1989 | abort (); | |
1990 | ||
1991 | if (type == UNKNOWN_BT) | |
1992 | type = bidi_type_of_next_char (bidi_it); | |
1993 | ||
1994 | if (type == NEUTRAL_B) | |
1995 | return bidi_it->resolved_level; | |
1996 | ||
1997 | level = bidi_it->level_stack[bidi_it->stack_idx].level; | |
1998 | if ((bidi_get_category (type) == NEUTRAL /* && type != NEUTRAL_B */) | |
1999 | || (type == WEAK_BN && prev_level == level)) | |
2000 | { | |
2001 | if (bidi_it->next_for_neutral.type == UNKNOWN_BT) | |
2002 | abort (); | |
2003 | ||
2004 | /* If the cached state shows a neutral character, it was not | |
2005 | resolved by bidi_resolve_neutral, so do it now. */ | |
2006 | type = bidi_resolve_neutral_1 (bidi_it->prev_for_neutral.type, | |
2007 | bidi_it->next_for_neutral.type, | |
2008 | level); | |
2009 | } | |
2010 | ||
2011 | if (!(type == STRONG_R | |
2012 | || type == STRONG_L | |
2013 | || type == WEAK_BN | |
2014 | || type == WEAK_EN | |
2015 | || type == WEAK_AN)) | |
2016 | abort (); | |
2017 | bidi_it->type = type; | |
2d6e4628 | 2018 | bidi_check_type (bidi_it->type); |
b7b65b15 EZ |
2019 | |
2020 | /* For L1 below, we need to know, for each WS character, whether | |
0d327994 | 2021 | it belongs to a sequence of WS characters preceding a newline |
b7b65b15 | 2022 | or a TAB or a paragraph separator. */ |
89d3374a | 2023 | if (bidi_it->orig_type == NEUTRAL_WS |
b7b65b15 EZ |
2024 | && bidi_it->next_for_ws.type == UNKNOWN_BT) |
2025 | { | |
2026 | int ch; | |
8264569d | 2027 | EMACS_INT clen = bidi_it->ch_len; |
6bff6497 EZ |
2028 | EMACS_INT bpos = bidi_it->bytepos; |
2029 | EMACS_INT cpos = bidi_it->charpos; | |
182ce2d2 | 2030 | EMACS_INT disp_pos = bidi_it->disp_pos; |
102ebb00 | 2031 | EMACS_INT nc = bidi_it->nchars; |
87e67904 | 2032 | struct bidi_string_data bs = bidi_it->string; |
b7b65b15 | 2033 | bidi_type_t chtype; |
fc6f18ce | 2034 | int fwp = bidi_it->frame_window_p; |
b7b65b15 | 2035 | |
102ebb00 EZ |
2036 | if (bidi_it->nchars <= 0) |
2037 | abort (); | |
b7b65b15 | 2038 | do { |
87e67904 | 2039 | ch = bidi_fetch_char (bpos += clen, cpos += nc, &disp_pos, &bs, fwp, |
fc6f18ce | 2040 | &clen, &nc); |
e7402cb2 | 2041 | if (ch == '\n' || ch == BIDI_EOB /* || ch == LINESEP_CHAR */) |
b7b65b15 EZ |
2042 | chtype = NEUTRAL_B; |
2043 | else | |
6bff6497 | 2044 | chtype = bidi_get_type (ch, NEUTRAL_DIR); |
b7b65b15 EZ |
2045 | } while (chtype == NEUTRAL_WS || chtype == WEAK_BN |
2046 | || bidi_explicit_dir_char (ch)); /* L1/Retaining */ | |
2047 | bidi_it->next_for_ws.type = chtype; | |
2d6e4628 | 2048 | bidi_check_type (bidi_it->next_for_ws.type); |
b7b65b15 EZ |
2049 | bidi_it->next_for_ws.charpos = cpos; |
2050 | bidi_it->next_for_ws.bytepos = bpos; | |
2051 | } | |
2052 | ||
2053 | /* Resolve implicit levels, with a twist: PDFs get the embedding | |
2054 | level of the enbedding they terminate. See below for the | |
2055 | reason. */ | |
89d3374a | 2056 | if (bidi_it->orig_type == PDF |
b7b65b15 EZ |
2057 | /* Don't do this if this formatting code didn't change the |
2058 | embedding level due to invalid or empty embeddings. */ | |
2059 | && prev_level != level) | |
2060 | { | |
2061 | /* Don't look in UAX#9 for the reason for this: it's our own | |
2062 | private quirk. The reason is that we want the formatting | |
2063 | codes to be delivered so that they bracket the text of their | |
2064 | embedding. For example, given the text | |
2065 | ||
2066 | {RLO}teST{PDF} | |
2067 | ||
2068 | we want it to be displayed as | |
2069 | ||
0d68907d | 2070 | {PDF}STet{RLO} |
b7b65b15 EZ |
2071 | |
2072 | not as | |
2073 | ||
2074 | STet{RLO}{PDF} | |
2075 | ||
2076 | which will result because we bump up the embedding level as | |
2077 | soon as we see the RLO and pop it as soon as we see the PDF, | |
2078 | so RLO itself has the same embedding level as "teST", and | |
2079 | thus would be normally delivered last, just before the PDF. | |
2080 | The switch below fiddles with the level of PDF so that this | |
2081 | ugly side effect does not happen. | |
2082 | ||
2083 | (This is, of course, only important if the formatting codes | |
e7402cb2 EZ |
2084 | are actually displayed, but Emacs does need to display them |
2085 | if the user wants to.) */ | |
b7b65b15 EZ |
2086 | level = prev_level; |
2087 | } | |
89d3374a EZ |
2088 | else if (bidi_it->orig_type == NEUTRAL_B /* L1 */ |
2089 | || bidi_it->orig_type == NEUTRAL_S | |
e7402cb2 EZ |
2090 | || bidi_it->ch == '\n' || bidi_it->ch == BIDI_EOB |
2091 | /* || bidi_it->ch == LINESEP_CHAR */ | |
89d3374a | 2092 | || (bidi_it->orig_type == NEUTRAL_WS |
b7b65b15 EZ |
2093 | && (bidi_it->next_for_ws.type == NEUTRAL_B |
2094 | || bidi_it->next_for_ws.type == NEUTRAL_S))) | |
2095 | level = bidi_it->level_stack[0].level; | |
2096 | else if ((level & 1) == 0) /* I1 */ | |
2097 | { | |
2098 | if (type == STRONG_R) | |
2099 | level++; | |
2100 | else if (type == WEAK_EN || type == WEAK_AN) | |
2101 | level += 2; | |
2102 | } | |
2103 | else /* I2 */ | |
2104 | { | |
2105 | if (type == STRONG_L || type == WEAK_EN || type == WEAK_AN) | |
2106 | level++; | |
2107 | } | |
2108 | ||
2109 | bidi_it->resolved_level = level; | |
2110 | return level; | |
2111 | } | |
2112 | ||
2113 | /* Move to the other edge of a level given by LEVEL. If END_FLAG is | |
2114 | non-zero, we are at the end of a level, and we need to prepare to | |
2115 | resume the scan of the lower level. | |
2116 | ||
2117 | If this level's other edge is cached, we simply jump to it, filling | |
2118 | the iterator structure with the iterator state on the other edge. | |
182ce2d2 EZ |
2119 | Otherwise, we walk the buffer or string until we come back to the |
2120 | same level as LEVEL. | |
b7b65b15 EZ |
2121 | |
2122 | Note: we are not talking here about a ``level run'' in the UAX#9 | |
2123 | sense of the term, but rather about a ``level'' which includes | |
2124 | all the levels higher than it. In other words, given the levels | |
2125 | like this: | |
2126 | ||
2127 | 11111112222222333333334443343222222111111112223322111 | |
2128 | A B C | |
2129 | ||
2130 | and assuming we are at point A scanning left to right, this | |
2131 | function moves to point C, whereas the UAX#9 ``level 2 run'' ends | |
2132 | at point B. */ | |
2133 | static void | |
2134 | bidi_find_other_level_edge (struct bidi_it *bidi_it, int level, int end_flag) | |
2135 | { | |
2136 | int dir = end_flag ? -bidi_it->scan_dir : bidi_it->scan_dir; | |
39e378da | 2137 | ptrdiff_t idx; |
b7b65b15 EZ |
2138 | |
2139 | /* Try the cache first. */ | |
87e67904 EZ |
2140 | if ((idx = bidi_cache_find_level_change (level, dir, end_flag)) |
2141 | >= bidi_cache_start) | |
b7b65b15 EZ |
2142 | bidi_cache_fetch_state (idx, bidi_it); |
2143 | else | |
2144 | { | |
2145 | int new_level; | |
2146 | ||
2147 | if (end_flag) | |
2148 | abort (); /* if we are at end of level, its edges must be cached */ | |
2149 | ||
2150 | bidi_cache_iterator_state (bidi_it, 1); | |
2151 | do { | |
2152 | new_level = bidi_level_of_next_char (bidi_it); | |
2153 | bidi_cache_iterator_state (bidi_it, 1); | |
2154 | } while (new_level >= level); | |
2155 | } | |
2156 | } | |
2157 | ||
2158 | void | |
4b292a22 | 2159 | bidi_move_to_visually_next (struct bidi_it *bidi_it) |
b7b65b15 EZ |
2160 | { |
2161 | int old_level, new_level, next_level; | |
9c82e145 | 2162 | struct bidi_it sentinel; |
acb28818 | 2163 | struct gcpro gcpro1; |
b7b65b15 | 2164 | |
87e67904 EZ |
2165 | if (bidi_it->charpos < 0 || bidi_it->bytepos < 0) |
2166 | abort (); | |
b7b65b15 EZ |
2167 | |
2168 | if (bidi_it->scan_dir == 0) | |
2169 | { | |
2170 | bidi_it->scan_dir = 1; /* default to logical order */ | |
2171 | } | |
2172 | ||
acb28818 | 2173 | /* The code below can call eval, and thus cause GC. If we are |
7e2ad32c | 2174 | iterating a Lisp string, make sure it won't be GCed. */ |
acb28818 EZ |
2175 | if (STRINGP (bidi_it->string.lstring)) |
2176 | GCPRO1 (bidi_it->string.lstring); | |
2177 | ||
be39f003 EZ |
2178 | /* If we just passed a newline, initialize for the next line. */ |
2179 | if (!bidi_it->first_elt && bidi_it->orig_type == NEUTRAL_B) | |
2180 | bidi_line_init (bidi_it); | |
2181 | ||
6dcfd253 EZ |
2182 | /* Prepare the sentinel iterator state, and cache it. When we bump |
2183 | into it, scanning backwards, we'll know that the last non-base | |
2184 | level is exhausted. */ | |
87e67904 | 2185 | if (bidi_cache_idx == bidi_cache_start) |
9c82e145 EZ |
2186 | { |
2187 | bidi_copy_it (&sentinel, bidi_it); | |
2188 | if (bidi_it->first_elt) | |
2189 | { | |
2190 | sentinel.charpos--; /* cached charpos needs to be monotonic */ | |
2191 | sentinel.bytepos--; | |
2192 | sentinel.ch = '\n'; /* doesn't matter, but why not? */ | |
2193 | sentinel.ch_len = 1; | |
182ce2d2 | 2194 | sentinel.nchars = 1; |
9c82e145 | 2195 | } |
6dcfd253 | 2196 | bidi_cache_iterator_state (&sentinel, 1); |
9c82e145 | 2197 | } |
b7b65b15 EZ |
2198 | |
2199 | old_level = bidi_it->resolved_level; | |
2200 | new_level = bidi_level_of_next_char (bidi_it); | |
b7b65b15 EZ |
2201 | |
2202 | /* Reordering of resolved levels (clause L2) is implemented by | |
2203 | jumping to the other edge of the level and flipping direction of | |
c0546589 | 2204 | scanning the text whenever we find a level change. */ |
b7b65b15 EZ |
2205 | if (new_level != old_level) |
2206 | { | |
2207 | int ascending = new_level > old_level; | |
2208 | int level_to_search = ascending ? old_level + 1 : old_level; | |
2209 | int incr = ascending ? 1 : -1; | |
2210 | int expected_next_level = old_level + incr; | |
2211 | ||
b7b65b15 EZ |
2212 | /* Jump (or walk) to the other edge of this level. */ |
2213 | bidi_find_other_level_edge (bidi_it, level_to_search, !ascending); | |
2214 | /* Switch scan direction and peek at the next character in the | |
2215 | new direction. */ | |
2216 | bidi_it->scan_dir = -bidi_it->scan_dir; | |
2217 | ||
2218 | /* The following loop handles the case where the resolved level | |
2219 | jumps by more than one. This is typical for numbers inside a | |
2220 | run of text with left-to-right embedding direction, but can | |
2221 | also happen in other situations. In those cases the decision | |
2222 | where to continue after a level change, and in what direction, | |
2223 | is tricky. For example, given a text like below: | |
2224 | ||
2225 | abcdefgh | |
2226 | 11336622 | |
2227 | ||
2228 | (where the numbers below the text show the resolved levels), | |
2229 | the result of reordering according to UAX#9 should be this: | |
2230 | ||
2231 | efdcghba | |
2232 | ||
2233 | This is implemented by the loop below which flips direction | |
2234 | and jumps to the other edge of the level each time it finds | |
2235 | the new level not to be the expected one. The expected level | |
2236 | is always one more or one less than the previous one. */ | |
2237 | next_level = bidi_peek_at_next_level (bidi_it); | |
2238 | while (next_level != expected_next_level) | |
2239 | { | |
2240 | expected_next_level += incr; | |
2241 | level_to_search += incr; | |
2242 | bidi_find_other_level_edge (bidi_it, level_to_search, !ascending); | |
2243 | bidi_it->scan_dir = -bidi_it->scan_dir; | |
2244 | next_level = bidi_peek_at_next_level (bidi_it); | |
2245 | } | |
2246 | ||
2247 | /* Finally, deliver the next character in the new direction. */ | |
2248 | next_level = bidi_level_of_next_char (bidi_it); | |
2249 | } | |
2250 | ||
b44d9321 EZ |
2251 | /* Take note when we have just processed the newline that precedes |
2252 | the end of the paragraph. The next time we are about to be | |
2253 | called, set_iterator_to_next will automatically reinit the | |
2254 | paragraph direction, if needed. We do this at the newline before | |
2255 | the paragraph separator, because the next character might not be | |
2256 | the first character of the next paragraph, due to the bidi | |
c0546589 EZ |
2257 | reordering, whereas we _must_ know the paragraph base direction |
2258 | _before_ we process the paragraph's text, since the base | |
2259 | direction affects the reordering. */ | |
87e67904 | 2260 | if (bidi_it->scan_dir == 1 && bidi_it->orig_type == NEUTRAL_B) |
be39f003 | 2261 | { |
87e67904 EZ |
2262 | /* The paragraph direction of the entire string, once |
2263 | determined, is in effect for the entire string. Setting the | |
2264 | separator limit to the end of the string prevents | |
2265 | bidi_paragraph_init from being called automatically on this | |
2266 | string. */ | |
9f257352 | 2267 | if (bidi_it->string.s || STRINGP (bidi_it->string.lstring)) |
87e67904 EZ |
2268 | bidi_it->separator_limit = bidi_it->string.schars; |
2269 | else if (bidi_it->bytepos < ZV_BYTE) | |
be39f003 | 2270 | { |
87e67904 EZ |
2271 | EMACS_INT sep_len = |
2272 | bidi_at_paragraph_end (bidi_it->charpos + bidi_it->nchars, | |
2273 | bidi_it->bytepos + bidi_it->ch_len); | |
2274 | if (bidi_it->nchars <= 0) | |
2275 | abort (); | |
2276 | if (sep_len >= 0) | |
2277 | { | |
2278 | bidi_it->new_paragraph = 1; | |
2279 | /* Record the buffer position of the last character of the | |
2280 | paragraph separator. */ | |
2281 | bidi_it->separator_limit = | |
2282 | bidi_it->charpos + bidi_it->nchars + sep_len; | |
2283 | } | |
be39f003 EZ |
2284 | } |
2285 | } | |
6bff6497 | 2286 | |
87e67904 | 2287 | if (bidi_it->scan_dir == 1 && bidi_cache_idx > bidi_cache_start) |
b7b65b15 EZ |
2288 | { |
2289 | /* If we are at paragraph's base embedding level and beyond the | |
2290 | last cached position, the cache's job is done and we can | |
2291 | discard it. */ | |
2292 | if (bidi_it->resolved_level == bidi_it->level_stack[0].level | |
182ce2d2 EZ |
2293 | && bidi_it->charpos > (bidi_cache[bidi_cache_idx - 1].charpos |
2294 | + bidi_cache[bidi_cache_idx - 1].nchars - 1)) | |
b7b65b15 EZ |
2295 | bidi_cache_reset (); |
2296 | /* But as long as we are caching during forward scan, we must | |
2297 | cache each state, or else the cache integrity will be | |
2298 | compromised: it assumes cached states correspond to buffer | |
2299 | positions 1:1. */ | |
2300 | else | |
2301 | bidi_cache_iterator_state (bidi_it, 1); | |
2302 | } | |
acb28818 EZ |
2303 | |
2304 | if (STRINGP (bidi_it->string.lstring)) | |
2305 | UNGCPRO; | |
b7b65b15 EZ |
2306 | } |
2307 | ||
2308 | /* This is meant to be called from within the debugger, whenever you | |
2309 | wish to examine the cache contents. */ | |
e78aecca | 2310 | void bidi_dump_cached_states (void) EXTERNALLY_VISIBLE; |
b7b65b15 EZ |
2311 | void |
2312 | bidi_dump_cached_states (void) | |
2313 | { | |
39e378da | 2314 | ptrdiff_t i; |
b7b65b15 EZ |
2315 | int ndigits = 1; |
2316 | ||
2317 | if (bidi_cache_idx == 0) | |
2318 | { | |
2319 | fprintf (stderr, "The cache is empty.\n"); | |
2320 | return; | |
2321 | } | |
df9733bf | 2322 | fprintf (stderr, "Total of %"pD"d state%s in cache:\n", |
b7b65b15 EZ |
2323 | bidi_cache_idx, bidi_cache_idx == 1 ? "" : "s"); |
2324 | ||
2325 | for (i = bidi_cache[bidi_cache_idx - 1].charpos; i > 0; i /= 10) | |
2326 | ndigits++; | |
2327 | fputs ("ch ", stderr); | |
2328 | for (i = 0; i < bidi_cache_idx; i++) | |
2329 | fprintf (stderr, "%*c", ndigits, bidi_cache[i].ch); | |
2330 | fputs ("\n", stderr); | |
2331 | fputs ("lvl ", stderr); | |
2332 | for (i = 0; i < bidi_cache_idx; i++) | |
2333 | fprintf (stderr, "%*d", ndigits, bidi_cache[i].resolved_level); | |
2334 | fputs ("\n", stderr); | |
2335 | fputs ("pos ", stderr); | |
2336 | for (i = 0; i < bidi_cache_idx; i++) | |
c2982e87 | 2337 | fprintf (stderr, "%*"pI"d", ndigits, bidi_cache[i].charpos); |
b7b65b15 EZ |
2338 | fputs ("\n", stderr); |
2339 | } |