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 | ||
32413314 EZ |
82 | /* UAX#9 says to search only for L, AL, or R types of characters, and |
83 | ignore RLE, RLO, LRE, and LRO, when determining the base paragraph | |
84 | level. Yudit indeed ignores them. This variable is therefore set | |
85 | by default to ignore them, but setting it to zero will take them | |
86 | into account. */ | |
e78aecca | 87 | extern int bidi_ignore_explicit_marks_for_paragraph_level EXTERNALLY_VISIBLE; |
b7b65b15 EZ |
88 | int bidi_ignore_explicit_marks_for_paragraph_level = 1; |
89 | ||
372b7a95 | 90 | static Lisp_Object paragraph_start_re, paragraph_separate_re; |
6bff6497 | 91 | static Lisp_Object Qparagraph_start, Qparagraph_separate; |
b7b65b15 | 92 | |
cbb09f04 EZ |
93 | \f |
94 | /*********************************************************************** | |
95 | Utilities | |
96 | ***********************************************************************/ | |
b7b65b15 | 97 | |
6bff6497 EZ |
98 | /* Return the bidi type of a character CH, subject to the current |
99 | directional OVERRIDE. */ | |
55d4c1b2 | 100 | static inline bidi_type_t |
6bff6497 | 101 | bidi_get_type (int ch, bidi_dir_t override) |
b7b65b15 | 102 | { |
6bff6497 EZ |
103 | bidi_type_t default_type; |
104 | ||
e7402cb2 EZ |
105 | if (ch == BIDI_EOB) |
106 | return NEUTRAL_B; | |
e342a24d EZ |
107 | if (ch < 0 || ch > MAX_CHAR) |
108 | abort (); | |
6bff6497 EZ |
109 | |
110 | default_type = (bidi_type_t) XINT (CHAR_TABLE_REF (bidi_type_table, ch)); | |
bca633fb EZ |
111 | /* Every valid character code, even those that are unassigned by the |
112 | UCD, have some bidi-class property, according to | |
113 | DerivedBidiClass.txt file. Therefore, if we ever get UNKNOWN_BT | |
114 | (= zero) code from CHAR_TABLE_REF, that's a bug. */ | |
115 | if (default_type == UNKNOWN_BT) | |
116 | abort (); | |
6bff6497 EZ |
117 | |
118 | if (override == NEUTRAL_DIR) | |
119 | return default_type; | |
120 | ||
121 | switch (default_type) | |
122 | { | |
123 | /* Although UAX#9 does not tell, it doesn't make sense to | |
124 | override NEUTRAL_B and LRM/RLM characters. */ | |
125 | case NEUTRAL_B: | |
126 | case LRE: | |
127 | case LRO: | |
128 | case RLE: | |
129 | case RLO: | |
130 | case PDF: | |
131 | return default_type; | |
132 | default: | |
133 | switch (ch) | |
134 | { | |
135 | case LRM_CHAR: | |
136 | case RLM_CHAR: | |
137 | return default_type; | |
138 | default: | |
139 | if (override == L2R) /* X6 */ | |
140 | return STRONG_L; | |
141 | else if (override == R2L) | |
142 | return STRONG_R; | |
143 | else | |
144 | abort (); /* can't happen: handled above */ | |
145 | } | |
146 | } | |
b7b65b15 EZ |
147 | } |
148 | ||
f67cdd7f | 149 | static inline void |
2d6e4628 EZ |
150 | bidi_check_type (bidi_type_t type) |
151 | { | |
f67cdd7f | 152 | xassert (UNKNOWN_BT <= type && type <= NEUTRAL_ON); |
2d6e4628 EZ |
153 | } |
154 | ||
b7b65b15 | 155 | /* Given a bidi TYPE of a character, return its category. */ |
55d4c1b2 | 156 | static inline bidi_category_t |
b7b65b15 EZ |
157 | bidi_get_category (bidi_type_t type) |
158 | { | |
159 | switch (type) | |
160 | { | |
161 | case UNKNOWN_BT: | |
162 | return UNKNOWN_BC; | |
163 | case STRONG_L: | |
164 | case STRONG_R: | |
165 | case STRONG_AL: | |
166 | case LRE: | |
167 | case LRO: | |
168 | case RLE: | |
169 | case RLO: | |
170 | return STRONG; | |
171 | case PDF: /* ??? really?? */ | |
172 | case WEAK_EN: | |
173 | case WEAK_ES: | |
174 | case WEAK_ET: | |
175 | case WEAK_AN: | |
176 | case WEAK_CS: | |
177 | case WEAK_NSM: | |
178 | case WEAK_BN: | |
179 | return WEAK; | |
180 | case NEUTRAL_B: | |
181 | case NEUTRAL_S: | |
182 | case NEUTRAL_WS: | |
183 | case NEUTRAL_ON: | |
184 | return NEUTRAL; | |
185 | default: | |
186 | abort (); | |
187 | } | |
188 | } | |
189 | ||
a6e8d97c EZ |
190 | /* Return the mirrored character of C, if it has one. If C has no |
191 | mirrored counterpart, return C. | |
192 | Note: The conditions in UAX#9 clause L4 regarding the surrounding | |
193 | context must be tested by the caller. */ | |
b7b65b15 EZ |
194 | int |
195 | bidi_mirror_char (int c) | |
196 | { | |
cbc4fd20 EZ |
197 | Lisp_Object val; |
198 | ||
199 | if (c == BIDI_EOB) | |
200 | return c; | |
201 | if (c < 0 || c > MAX_CHAR) | |
202 | abort (); | |
b7b65b15 | 203 | |
cbc4fd20 EZ |
204 | val = CHAR_TABLE_REF (bidi_mirror_table, c); |
205 | if (INTEGERP (val)) | |
b7b65b15 | 206 | { |
cbc4fd20 | 207 | int v = XINT (val); |
b7b65b15 | 208 | |
cbc4fd20 EZ |
209 | if (v < 0 || v > MAX_CHAR) |
210 | abort (); | |
211 | ||
212 | return v; | |
b7b65b15 | 213 | } |
cbc4fd20 | 214 | |
b7b65b15 EZ |
215 | return c; |
216 | } | |
217 | ||
cbb09f04 EZ |
218 | /* Determine the start-of-run (sor) directional type given the two |
219 | embedding levels on either side of the run boundary. Also, update | |
220 | the saved info about previously seen characters, since that info is | |
221 | generally valid for a single level run. */ | |
58b9f433 | 222 | static inline void |
cbb09f04 EZ |
223 | bidi_set_sor_type (struct bidi_it *bidi_it, int level_before, int level_after) |
224 | { | |
225 | int higher_level = level_before > level_after ? level_before : level_after; | |
226 | ||
227 | /* The prev_was_pdf gork is required for when we have several PDFs | |
228 | in a row. In that case, we want to compute the sor type for the | |
229 | next level run only once: when we see the first PDF. That's | |
230 | because the sor type depends only on the higher of the two levels | |
231 | that we find on the two sides of the level boundary (see UAX#9, | |
232 | clause X10), and so we don't need to know the final embedding | |
233 | level to which we descend after processing all the PDFs. */ | |
234 | if (!bidi_it->prev_was_pdf || level_before < level_after) | |
235 | /* FIXME: should the default sor direction be user selectable? */ | |
236 | bidi_it->sor = (higher_level & 1) != 0 ? R2L : L2R; | |
237 | if (level_before > level_after) | |
238 | bidi_it->prev_was_pdf = 1; | |
239 | ||
240 | bidi_it->prev.type = UNKNOWN_BT; | |
241 | bidi_it->last_strong.type = bidi_it->last_strong.type_after_w1 = | |
242 | bidi_it->last_strong.orig_type = UNKNOWN_BT; | |
243 | bidi_it->prev_for_neutral.type = bidi_it->sor == R2L ? STRONG_R : STRONG_L; | |
244 | bidi_it->prev_for_neutral.charpos = bidi_it->charpos; | |
245 | bidi_it->prev_for_neutral.bytepos = bidi_it->bytepos; | |
246 | bidi_it->next_for_neutral.type = bidi_it->next_for_neutral.type_after_w1 = | |
247 | bidi_it->next_for_neutral.orig_type = UNKNOWN_BT; | |
248 | bidi_it->ignore_bn_limit = -1; /* meaning it's unknown */ | |
249 | } | |
250 | ||
251 | /* Push the current embedding level and override status; reset the | |
252 | current level to LEVEL and the current override status to OVERRIDE. */ | |
58b9f433 | 253 | static inline void |
cbb09f04 EZ |
254 | bidi_push_embedding_level (struct bidi_it *bidi_it, |
255 | int level, bidi_dir_t override) | |
256 | { | |
257 | bidi_it->stack_idx++; | |
7e2ad32c | 258 | xassert (bidi_it->stack_idx < BIDI_MAXLEVEL); |
cbb09f04 EZ |
259 | bidi_it->level_stack[bidi_it->stack_idx].level = level; |
260 | bidi_it->level_stack[bidi_it->stack_idx].override = override; | |
261 | } | |
262 | ||
263 | /* Pop the embedding level and directional override status from the | |
264 | stack, and return the new level. */ | |
58b9f433 | 265 | static inline int |
cbb09f04 EZ |
266 | bidi_pop_embedding_level (struct bidi_it *bidi_it) |
267 | { | |
268 | /* UAX#9 says to ignore invalid PDFs. */ | |
269 | if (bidi_it->stack_idx > 0) | |
270 | bidi_it->stack_idx--; | |
271 | return bidi_it->level_stack[bidi_it->stack_idx].level; | |
272 | } | |
273 | ||
274 | /* Record in SAVED_INFO the information about the current character. */ | |
58b9f433 | 275 | static inline void |
cbb09f04 EZ |
276 | bidi_remember_char (struct bidi_saved_info *saved_info, |
277 | struct bidi_it *bidi_it) | |
278 | { | |
279 | saved_info->charpos = bidi_it->charpos; | |
280 | saved_info->bytepos = bidi_it->bytepos; | |
281 | saved_info->type = bidi_it->type; | |
282 | bidi_check_type (bidi_it->type); | |
283 | saved_info->type_after_w1 = bidi_it->type_after_w1; | |
284 | bidi_check_type (bidi_it->type_after_w1); | |
285 | saved_info->orig_type = bidi_it->orig_type; | |
286 | bidi_check_type (bidi_it->orig_type); | |
287 | } | |
288 | ||
b7b65b15 EZ |
289 | /* Copy the bidi iterator from FROM to TO. To save cycles, this only |
290 | copies the part of the level stack that is actually in use. */ | |
55d4c1b2 | 291 | static inline void |
b7b65b15 EZ |
292 | bidi_copy_it (struct bidi_it *to, struct bidi_it *from) |
293 | { | |
294 | int i; | |
295 | ||
e69a9370 | 296 | /* Copy everything except the level stack and beyond. */ |
182ce2d2 | 297 | memcpy (to, from, offsetof (struct bidi_it, level_stack[0])); |
b7b65b15 EZ |
298 | |
299 | /* Copy the active part of the level stack. */ | |
300 | to->level_stack[0] = from->level_stack[0]; /* level zero is always in use */ | |
301 | for (i = 1; i <= from->stack_idx; i++) | |
302 | to->level_stack[i] = from->level_stack[i]; | |
303 | } | |
304 | ||
cbb09f04 EZ |
305 | \f |
306 | /*********************************************************************** | |
307 | Caching the bidi iterator states | |
308 | ***********************************************************************/ | |
b7b65b15 | 309 | |
2fe72643 EZ |
310 | #define BIDI_CACHE_CHUNK 200 |
311 | static struct bidi_it *bidi_cache; | |
39e378da | 312 | static ptrdiff_t bidi_cache_size = 0; |
ad6042bb | 313 | enum { elsz = sizeof (struct bidi_it) }; |
39e378da PE |
314 | static ptrdiff_t bidi_cache_idx; /* next unused cache slot */ |
315 | static ptrdiff_t bidi_cache_last_idx; /* slot of last cache hit */ | |
316 | static ptrdiff_t bidi_cache_start = 0; /* start of cache for this | |
87e67904 EZ |
317 | "stack" level */ |
318 | ||
319 | /* Reset the cache state to the empty state. We only reset the part | |
320 | of the cache relevant to iteration of the current object. Previous | |
321 | objects, which are pushed on the display iterator's stack, are left | |
322 | intact. This is called when the cached information is no more | |
323 | useful for the current iteration, e.g. when we were reseated to a | |
324 | new position on the same object. */ | |
55d4c1b2 | 325 | static inline void |
b7b65b15 EZ |
326 | bidi_cache_reset (void) |
327 | { | |
87e67904 | 328 | bidi_cache_idx = bidi_cache_start; |
b7b65b15 EZ |
329 | bidi_cache_last_idx = -1; |
330 | } | |
331 | ||
87e67904 EZ |
332 | /* Shrink the cache to its minimal size. Called when we init the bidi |
333 | iterator for reordering a buffer or a string that does not come | |
334 | from display properties, because that means all the previously | |
335 | cached info is of no further use. */ | |
55d4c1b2 | 336 | static inline void |
2fe72643 EZ |
337 | bidi_cache_shrink (void) |
338 | { | |
339 | if (bidi_cache_size > BIDI_CACHE_CHUNK) | |
340 | { | |
0416466c EZ |
341 | bidi_cache_size = BIDI_CACHE_CHUNK; |
342 | bidi_cache = | |
343 | (struct bidi_it *) xrealloc (bidi_cache, bidi_cache_size * elsz); | |
2fe72643 EZ |
344 | } |
345 | bidi_cache_reset (); | |
346 | } | |
347 | ||
55d4c1b2 | 348 | static inline void |
39e378da | 349 | bidi_cache_fetch_state (ptrdiff_t idx, struct bidi_it *bidi_it) |
b7b65b15 EZ |
350 | { |
351 | int current_scan_dir = bidi_it->scan_dir; | |
352 | ||
87e67904 | 353 | if (idx < bidi_cache_start || idx >= bidi_cache_idx) |
b7b65b15 EZ |
354 | abort (); |
355 | ||
356 | bidi_copy_it (bidi_it, &bidi_cache[idx]); | |
357 | bidi_it->scan_dir = current_scan_dir; | |
358 | bidi_cache_last_idx = idx; | |
359 | } | |
360 | ||
361 | /* Find a cached state with a given CHARPOS and resolved embedding | |
362 | level less or equal to LEVEL. if LEVEL is -1, disregard the | |
363 | resolved levels in cached states. DIR, if non-zero, means search | |
364 | in that direction from the last cache hit. */ | |
39e378da | 365 | static inline ptrdiff_t |
61bfec98 | 366 | bidi_cache_search (EMACS_INT charpos, int level, int dir) |
b7b65b15 | 367 | { |
39e378da | 368 | ptrdiff_t i, i_start; |
b7b65b15 | 369 | |
6eec7596 | 370 | if (bidi_cache_idx > bidi_cache_start) |
b7b65b15 | 371 | { |
6eec7596 EZ |
372 | if (bidi_cache_last_idx == -1) |
373 | bidi_cache_last_idx = bidi_cache_idx - 1; | |
b7b65b15 | 374 | if (charpos < bidi_cache[bidi_cache_last_idx].charpos) |
182ce2d2 EZ |
375 | { |
376 | dir = -1; | |
377 | i_start = bidi_cache_last_idx - 1; | |
378 | } | |
379 | else if (charpos > (bidi_cache[bidi_cache_last_idx].charpos | |
380 | + bidi_cache[bidi_cache_last_idx].nchars - 1)) | |
381 | { | |
382 | dir = 1; | |
383 | i_start = bidi_cache_last_idx + 1; | |
384 | } | |
385 | else if (dir) | |
b7b65b15 EZ |
386 | i_start = bidi_cache_last_idx; |
387 | else | |
388 | { | |
389 | dir = -1; | |
390 | i_start = bidi_cache_idx - 1; | |
391 | } | |
392 | ||
393 | if (dir < 0) | |
394 | { | |
395 | /* Linear search for now; FIXME! */ | |
87e67904 | 396 | for (i = i_start; i >= bidi_cache_start; i--) |
182ce2d2 EZ |
397 | if (bidi_cache[i].charpos <= charpos |
398 | && charpos < bidi_cache[i].charpos + bidi_cache[i].nchars | |
b7b65b15 EZ |
399 | && (level == -1 || bidi_cache[i].resolved_level <= level)) |
400 | return i; | |
401 | } | |
402 | else | |
403 | { | |
404 | for (i = i_start; i < bidi_cache_idx; i++) | |
182ce2d2 EZ |
405 | if (bidi_cache[i].charpos <= charpos |
406 | && charpos < bidi_cache[i].charpos + bidi_cache[i].nchars | |
b7b65b15 EZ |
407 | && (level == -1 || bidi_cache[i].resolved_level <= level)) |
408 | return i; | |
409 | } | |
410 | } | |
411 | ||
412 | return -1; | |
413 | } | |
414 | ||
415 | /* Find a cached state where the resolved level changes to a value | |
416 | that is lower than LEVEL, and return its cache slot index. DIR is | |
417 | the direction to search, starting with the last used cache slot. | |
87e67904 EZ |
418 | If DIR is zero, we search backwards from the last occupied cache |
419 | slot. BEFORE, if non-zero, means return the index of the slot that | |
420 | is ``before'' the level change in the search direction. That is, | |
b7b65b15 EZ |
421 | given the cached levels like this: |
422 | ||
423 | 1122333442211 | |
424 | AB C | |
425 | ||
426 | and assuming we are at the position cached at the slot marked with | |
427 | C, searching backwards (DIR = -1) for LEVEL = 2 will return the | |
428 | index of slot B or A, depending whether BEFORE is, respectively, | |
429 | non-zero or zero. */ | |
39e378da | 430 | static ptrdiff_t |
b7b65b15 EZ |
431 | bidi_cache_find_level_change (int level, int dir, int before) |
432 | { | |
433 | if (bidi_cache_idx) | |
434 | { | |
39e378da | 435 | ptrdiff_t i = dir ? bidi_cache_last_idx : bidi_cache_idx - 1; |
b7b65b15 EZ |
436 | int incr = before ? 1 : 0; |
437 | ||
6eec7596 EZ |
438 | xassert (!dir || bidi_cache_last_idx >= 0); |
439 | ||
b7b65b15 EZ |
440 | if (!dir) |
441 | dir = -1; | |
442 | else if (!incr) | |
443 | i += dir; | |
444 | ||
445 | if (dir < 0) | |
446 | { | |
87e67904 | 447 | while (i >= bidi_cache_start + incr) |
b7b65b15 EZ |
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 | else | |
456 | { | |
457 | while (i < bidi_cache_idx - incr) | |
458 | { | |
459 | if (bidi_cache[i + incr].resolved_level >= 0 | |
460 | && bidi_cache[i + incr].resolved_level < level) | |
461 | return i; | |
462 | i++; | |
463 | } | |
464 | } | |
465 | } | |
466 | ||
467 | return -1; | |
468 | } | |
469 | ||
58b9f433 | 470 | static inline void |
39e378da | 471 | bidi_cache_ensure_space (ptrdiff_t idx) |
58b9f433 EZ |
472 | { |
473 | /* Enlarge the cache as needed. */ | |
474 | if (idx >= bidi_cache_size) | |
475 | { | |
39e378da | 476 | ptrdiff_t new_size; |
f0eb61e9 PE |
477 | |
478 | /* The bidi cache cannot be larger than the largest Lisp string | |
479 | or buffer. */ | |
480 | ptrdiff_t string_or_buffer_bound = | |
481 | max (BUF_BYTES_MAX, STRING_BYTES_BOUND); | |
482 | ||
483 | /* Also, it cannot be larger than what C can represent. */ | |
484 | ptrdiff_t c_bound = min (PTRDIFF_MAX, SIZE_MAX) / elsz; | |
485 | ||
486 | if (min (string_or_buffer_bound, c_bound) <= idx) | |
39e378da PE |
487 | memory_full (SIZE_MAX); |
488 | new_size = idx - idx % BIDI_CACHE_CHUNK + BIDI_CACHE_CHUNK; | |
489 | bidi_cache = (struct bidi_it *) xrealloc (bidi_cache, new_size * elsz); | |
490 | bidi_cache_size = new_size; | |
58b9f433 EZ |
491 | } |
492 | } | |
493 | ||
55d4c1b2 | 494 | static inline void |
b7b65b15 EZ |
495 | bidi_cache_iterator_state (struct bidi_it *bidi_it, int resolved) |
496 | { | |
39e378da | 497 | ptrdiff_t idx; |
b7b65b15 EZ |
498 | |
499 | /* We should never cache on backward scans. */ | |
500 | if (bidi_it->scan_dir == -1) | |
501 | abort (); | |
502 | idx = bidi_cache_search (bidi_it->charpos, -1, 1); | |
503 | ||
504 | if (idx < 0) | |
505 | { | |
506 | idx = bidi_cache_idx; | |
58b9f433 | 507 | bidi_cache_ensure_space (idx); |
bd924a5d EZ |
508 | /* Character positions should correspond to cache positions 1:1. |
509 | If we are outside the range of cached positions, the cache is | |
510 | useless and must be reset. */ | |
87e67904 | 511 | if (idx > bidi_cache_start && |
182ce2d2 EZ |
512 | (bidi_it->charpos > (bidi_cache[idx - 1].charpos |
513 | + bidi_cache[idx - 1].nchars) | |
a1344e7d | 514 | || bidi_it->charpos < bidi_cache[bidi_cache_start].charpos)) |
bd924a5d EZ |
515 | { |
516 | bidi_cache_reset (); | |
87e67904 | 517 | idx = bidi_cache_start; |
bd924a5d | 518 | } |
102ebb00 EZ |
519 | if (bidi_it->nchars <= 0) |
520 | abort (); | |
b7b65b15 EZ |
521 | bidi_copy_it (&bidi_cache[idx], bidi_it); |
522 | if (!resolved) | |
523 | bidi_cache[idx].resolved_level = -1; | |
524 | } | |
525 | else | |
526 | { | |
527 | /* Copy only the members which could have changed, to avoid | |
528 | costly copying of the entire struct. */ | |
529 | bidi_cache[idx].type = bidi_it->type; | |
2d6e4628 | 530 | bidi_check_type (bidi_it->type); |
89d3374a EZ |
531 | bidi_cache[idx].type_after_w1 = bidi_it->type_after_w1; |
532 | bidi_check_type (bidi_it->type_after_w1); | |
b7b65b15 EZ |
533 | if (resolved) |
534 | bidi_cache[idx].resolved_level = bidi_it->resolved_level; | |
535 | else | |
536 | bidi_cache[idx].resolved_level = -1; | |
537 | bidi_cache[idx].invalid_levels = bidi_it->invalid_levels; | |
538 | bidi_cache[idx].invalid_rl_levels = bidi_it->invalid_rl_levels; | |
539 | bidi_cache[idx].next_for_neutral = bidi_it->next_for_neutral; | |
540 | bidi_cache[idx].next_for_ws = bidi_it->next_for_ws; | |
541 | bidi_cache[idx].ignore_bn_limit = bidi_it->ignore_bn_limit; | |
f67cdd7f | 542 | bidi_cache[idx].disp_pos = bidi_it->disp_pos; |
0c95fcf7 | 543 | bidi_cache[idx].disp_prop = bidi_it->disp_prop; |
b7b65b15 EZ |
544 | } |
545 | ||
546 | bidi_cache_last_idx = idx; | |
547 | if (idx >= bidi_cache_idx) | |
548 | bidi_cache_idx = idx + 1; | |
549 | } | |
550 | ||
55d4c1b2 | 551 | static inline bidi_type_t |
61bfec98 | 552 | bidi_cache_find (EMACS_INT charpos, int level, struct bidi_it *bidi_it) |
b7b65b15 | 553 | { |
39e378da | 554 | ptrdiff_t i = bidi_cache_search (charpos, level, bidi_it->scan_dir); |
b7b65b15 | 555 | |
87e67904 | 556 | if (i >= bidi_cache_start) |
b7b65b15 EZ |
557 | { |
558 | bidi_dir_t current_scan_dir = bidi_it->scan_dir; | |
559 | ||
5009f85e | 560 | bidi_copy_it (bidi_it, &bidi_cache[i]); |
b7b65b15 EZ |
561 | bidi_cache_last_idx = i; |
562 | /* Don't let scan direction from from the cached state override | |
563 | the current scan direction. */ | |
564 | bidi_it->scan_dir = current_scan_dir; | |
565 | return bidi_it->type; | |
566 | } | |
567 | ||
568 | return UNKNOWN_BT; | |
569 | } | |
570 | ||
55d4c1b2 | 571 | static inline int |
b7b65b15 EZ |
572 | bidi_peek_at_next_level (struct bidi_it *bidi_it) |
573 | { | |
87e67904 | 574 | if (bidi_cache_idx == bidi_cache_start || bidi_cache_last_idx == -1) |
b7b65b15 EZ |
575 | abort (); |
576 | return bidi_cache[bidi_cache_last_idx + bidi_it->scan_dir].resolved_level; | |
577 | } | |
578 | ||
cbb09f04 | 579 | \f |
58b9f433 EZ |
580 | /*********************************************************************** |
581 | Pushing and popping the bidi iterator state | |
582 | ***********************************************************************/ | |
0c22566f EZ |
583 | /* 5-slot stack for saving the start of the previous level of the |
584 | cache. xdisp.c maintains a 5-slot stack for its iterator state, | |
585 | and we need the same size of our stack. */ | |
39e378da | 586 | static ptrdiff_t bidi_cache_start_stack[IT_STACK_SIZE]; |
58b9f433 EZ |
587 | static int bidi_cache_sp; |
588 | ||
589 | /* Push the bidi iterator state in preparation for reordering a | |
590 | different object, e.g. display string found at certain buffer | |
7e2ad32c EZ |
591 | position. Pushing the bidi iterator boils down to saving its |
592 | entire state on the cache and starting a new cache "stacked" on top | |
593 | of the current cache. */ | |
58b9f433 EZ |
594 | void |
595 | bidi_push_it (struct bidi_it *bidi_it) | |
b7b65b15 | 596 | { |
58b9f433 EZ |
597 | /* Save the current iterator state in its entirety after the last |
598 | used cache slot. */ | |
599 | bidi_cache_ensure_space (bidi_cache_idx); | |
600 | memcpy (&bidi_cache[bidi_cache_idx++], bidi_it, sizeof (struct bidi_it)); | |
be39f003 | 601 | |
58b9f433 | 602 | /* Push the current cache start onto the stack. */ |
7e2ad32c | 603 | xassert (bidi_cache_sp < IT_STACK_SIZE); |
58b9f433 | 604 | bidi_cache_start_stack[bidi_cache_sp++] = bidi_cache_start; |
be39f003 | 605 | |
58b9f433 EZ |
606 | /* Start a new level of cache, and make it empty. */ |
607 | bidi_cache_start = bidi_cache_idx; | |
608 | bidi_cache_last_idx = -1; | |
609 | } | |
610 | ||
611 | /* Restore the iterator state saved by bidi_push_it and return the | |
612 | cache to the corresponding state. */ | |
613 | void | |
614 | bidi_pop_it (struct bidi_it *bidi_it) | |
615 | { | |
616 | if (bidi_cache_start <= 0) | |
617 | abort (); | |
618 | ||
619 | /* Reset the next free cache slot index to what it was before the | |
620 | call to bidi_push_it. */ | |
621 | bidi_cache_idx = bidi_cache_start - 1; | |
622 | ||
623 | /* Restore the bidi iterator state saved in the cache. */ | |
624 | memcpy (bidi_it, &bidi_cache[bidi_cache_idx], sizeof (struct bidi_it)); | |
625 | ||
626 | /* Pop the previous cache start from the stack. */ | |
627 | if (bidi_cache_sp <= 0) | |
628 | abort (); | |
629 | bidi_cache_start = bidi_cache_start_stack[--bidi_cache_sp]; | |
630 | ||
631 | /* Invalidate the last-used cache slot data. */ | |
632 | bidi_cache_last_idx = -1; | |
633 | } | |
634 | ||
ec7cc85b | 635 | static ptrdiff_t bidi_cache_total_alloc; |
35928349 | 636 | |
ed94e6d7 EZ |
637 | /* Stash away a copy of the cache and its control variables. */ |
638 | void * | |
639 | bidi_shelve_cache (void) | |
640 | { | |
641 | unsigned char *databuf; | |
642 | ||
35928349 | 643 | /* Empty cache. */ |
ed94e6d7 EZ |
644 | if (bidi_cache_idx == 0) |
645 | return NULL; | |
646 | ||
647 | databuf = xmalloc (sizeof (bidi_cache_idx) | |
648 | + bidi_cache_idx * sizeof (struct bidi_it) | |
649 | + sizeof (bidi_cache_start_stack) | |
650 | + sizeof (bidi_cache_sp) + sizeof (bidi_cache_start) | |
651 | + sizeof (bidi_cache_last_idx)); | |
35928349 EZ |
652 | bidi_cache_total_alloc += |
653 | sizeof (bidi_cache_idx) + bidi_cache_idx * sizeof (struct bidi_it) | |
654 | + sizeof (bidi_cache_start_stack) | |
655 | + sizeof (bidi_cache_sp) + sizeof (bidi_cache_start) | |
656 | + sizeof (bidi_cache_last_idx); | |
657 | ||
ed94e6d7 EZ |
658 | memcpy (databuf, &bidi_cache_idx, sizeof (bidi_cache_idx)); |
659 | memcpy (databuf + sizeof (bidi_cache_idx), | |
660 | bidi_cache, bidi_cache_idx * sizeof (struct bidi_it)); | |
661 | memcpy (databuf + sizeof (bidi_cache_idx) | |
662 | + bidi_cache_idx * sizeof (struct bidi_it), | |
663 | bidi_cache_start_stack, sizeof (bidi_cache_start_stack)); | |
664 | memcpy (databuf + sizeof (bidi_cache_idx) | |
665 | + bidi_cache_idx * sizeof (struct bidi_it) | |
666 | + sizeof (bidi_cache_start_stack), | |
667 | &bidi_cache_sp, sizeof (bidi_cache_sp)); | |
668 | memcpy (databuf + sizeof (bidi_cache_idx) | |
669 | + bidi_cache_idx * sizeof (struct bidi_it) | |
670 | + sizeof (bidi_cache_start_stack) + sizeof (bidi_cache_sp), | |
671 | &bidi_cache_start, sizeof (bidi_cache_start)); | |
672 | memcpy (databuf + sizeof (bidi_cache_idx) | |
673 | + bidi_cache_idx * sizeof (struct bidi_it) | |
674 | + sizeof (bidi_cache_start_stack) + sizeof (bidi_cache_sp) | |
675 | + sizeof (bidi_cache_start), | |
676 | &bidi_cache_last_idx, sizeof (bidi_cache_last_idx)); | |
677 | ||
678 | return databuf; | |
679 | } | |
680 | ||
d1410150 EZ |
681 | /* Restore the cache state from a copy stashed away by |
682 | bidi_shelve_cache, and free the buffer used to stash that copy. | |
683 | JUST_FREE non-zero means free the buffer, but don't restore the | |
684 | cache; used when the corresponding iterator is discarded instead of | |
685 | being restored. */ | |
ed94e6d7 | 686 | void |
35928349 | 687 | bidi_unshelve_cache (void *databuf, int just_free) |
ed94e6d7 EZ |
688 | { |
689 | unsigned char *p = databuf; | |
690 | ||
691 | if (!p) | |
be39f003 | 692 | { |
d1410150 EZ |
693 | if (!just_free) |
694 | { | |
695 | /* A NULL pointer means an empty cache. */ | |
696 | bidi_cache_start = 0; | |
697 | bidi_cache_sp = 0; | |
698 | bidi_cache_reset (); | |
699 | } | |
be39f003 | 700 | } |
ed94e6d7 EZ |
701 | else |
702 | { | |
35928349 EZ |
703 | if (just_free) |
704 | { | |
705 | ptrdiff_t idx; | |
706 | ||
707 | memcpy (&idx, p, sizeof (bidi_cache_idx)); | |
708 | bidi_cache_total_alloc -= | |
709 | sizeof (bidi_cache_idx) + idx * sizeof (struct bidi_it) | |
710 | + sizeof (bidi_cache_start_stack) + sizeof (bidi_cache_sp) | |
711 | + sizeof (bidi_cache_start) + sizeof (bidi_cache_last_idx); | |
712 | } | |
713 | else | |
714 | { | |
715 | memcpy (&bidi_cache_idx, p, sizeof (bidi_cache_idx)); | |
716 | bidi_cache_ensure_space (bidi_cache_idx); | |
717 | memcpy (bidi_cache, p + sizeof (bidi_cache_idx), | |
718 | bidi_cache_idx * sizeof (struct bidi_it)); | |
719 | memcpy (bidi_cache_start_stack, | |
720 | p + sizeof (bidi_cache_idx) | |
721 | + bidi_cache_idx * sizeof (struct bidi_it), | |
722 | sizeof (bidi_cache_start_stack)); | |
723 | memcpy (&bidi_cache_sp, | |
724 | p + sizeof (bidi_cache_idx) | |
725 | + bidi_cache_idx * sizeof (struct bidi_it) | |
726 | + sizeof (bidi_cache_start_stack), | |
727 | sizeof (bidi_cache_sp)); | |
728 | memcpy (&bidi_cache_start, | |
729 | p + sizeof (bidi_cache_idx) | |
730 | + bidi_cache_idx * sizeof (struct bidi_it) | |
731 | + sizeof (bidi_cache_start_stack) + sizeof (bidi_cache_sp), | |
732 | sizeof (bidi_cache_start)); | |
733 | memcpy (&bidi_cache_last_idx, | |
734 | p + sizeof (bidi_cache_idx) | |
735 | + bidi_cache_idx * sizeof (struct bidi_it) | |
736 | + sizeof (bidi_cache_start_stack) + sizeof (bidi_cache_sp) | |
737 | + sizeof (bidi_cache_start), | |
738 | sizeof (bidi_cache_last_idx)); | |
739 | bidi_cache_total_alloc -= | |
740 | sizeof (bidi_cache_idx) + bidi_cache_idx * sizeof (struct bidi_it) | |
741 | + sizeof (bidi_cache_start_stack) + sizeof (bidi_cache_sp) | |
742 | + sizeof (bidi_cache_start) + sizeof (bidi_cache_last_idx); | |
743 | } | |
ed94e6d7 EZ |
744 | |
745 | xfree (p); | |
746 | } | |
747 | } | |
b7b65b15 | 748 | |
58b9f433 | 749 | \f |
cbb09f04 EZ |
750 | /*********************************************************************** |
751 | Initialization | |
752 | ***********************************************************************/ | |
753 | static void | |
754 | bidi_initialize (void) | |
b7b65b15 | 755 | { |
474a8465 EZ |
756 | bidi_type_table = uniprop_table (intern ("bidi-class")); |
757 | if (NILP (bidi_type_table)) | |
758 | abort (); | |
cbb09f04 EZ |
759 | staticpro (&bidi_type_table); |
760 | ||
474a8465 EZ |
761 | bidi_mirror_table = uniprop_table (intern ("mirroring")); |
762 | if (NILP (bidi_mirror_table)) | |
763 | abort (); | |
cbb09f04 EZ |
764 | staticpro (&bidi_mirror_table); |
765 | ||
cbb09f04 EZ |
766 | Qparagraph_start = intern ("paragraph-start"); |
767 | staticpro (&Qparagraph_start); | |
768 | paragraph_start_re = Fsymbol_value (Qparagraph_start); | |
769 | if (!STRINGP (paragraph_start_re)) | |
770 | paragraph_start_re = build_string ("\f\\|[ \t]*$"); | |
771 | staticpro (¶graph_start_re); | |
772 | Qparagraph_separate = intern ("paragraph-separate"); | |
773 | staticpro (&Qparagraph_separate); | |
774 | paragraph_separate_re = Fsymbol_value (Qparagraph_separate); | |
775 | if (!STRINGP (paragraph_separate_re)) | |
776 | paragraph_separate_re = build_string ("[ \t\f]*$"); | |
777 | staticpro (¶graph_separate_re); | |
58b9f433 EZ |
778 | |
779 | bidi_cache_sp = 0; | |
ec7cc85b | 780 | bidi_cache_total_alloc = 0; |
58b9f433 | 781 | |
cbb09f04 | 782 | bidi_initialized = 1; |
b7b65b15 EZ |
783 | } |
784 | ||
cbb09f04 EZ |
785 | /* Do whatever UAX#9 clause X8 says should be done at paragraph's |
786 | end. */ | |
55d4c1b2 | 787 | static inline void |
cbb09f04 | 788 | bidi_set_paragraph_end (struct bidi_it *bidi_it) |
b7b65b15 | 789 | { |
cbb09f04 EZ |
790 | bidi_it->invalid_levels = 0; |
791 | bidi_it->invalid_rl_levels = -1; | |
792 | bidi_it->stack_idx = 0; | |
793 | bidi_it->resolved_level = bidi_it->level_stack[0].level; | |
794 | } | |
b7b65b15 | 795 | |
cbb09f04 EZ |
796 | /* Initialize the bidi iterator from buffer/string position CHARPOS. */ |
797 | void | |
798 | bidi_init_it (EMACS_INT charpos, EMACS_INT bytepos, int frame_window_p, | |
799 | struct bidi_it *bidi_it) | |
800 | { | |
801 | if (! bidi_initialized) | |
802 | bidi_initialize (); | |
803 | if (charpos >= 0) | |
804 | bidi_it->charpos = charpos; | |
805 | if (bytepos >= 0) | |
806 | bidi_it->bytepos = bytepos; | |
807 | bidi_it->frame_window_p = frame_window_p; | |
808 | bidi_it->nchars = -1; /* to be computed in bidi_resolve_explicit_1 */ | |
809 | bidi_it->first_elt = 1; | |
810 | bidi_set_paragraph_end (bidi_it); | |
811 | bidi_it->new_paragraph = 1; | |
812 | bidi_it->separator_limit = -1; | |
813 | bidi_it->type = NEUTRAL_B; | |
814 | bidi_it->type_after_w1 = NEUTRAL_B; | |
815 | bidi_it->orig_type = NEUTRAL_B; | |
816 | bidi_it->prev_was_pdf = 0; | |
817 | bidi_it->prev.type = bidi_it->prev.type_after_w1 = | |
818 | bidi_it->prev.orig_type = UNKNOWN_BT; | |
89d3374a EZ |
819 | bidi_it->last_strong.type = bidi_it->last_strong.type_after_w1 = |
820 | bidi_it->last_strong.orig_type = UNKNOWN_BT; | |
cbb09f04 EZ |
821 | bidi_it->next_for_neutral.charpos = -1; |
822 | bidi_it->next_for_neutral.type = | |
823 | bidi_it->next_for_neutral.type_after_w1 = | |
89d3374a | 824 | bidi_it->next_for_neutral.orig_type = UNKNOWN_BT; |
cbb09f04 EZ |
825 | bidi_it->prev_for_neutral.charpos = -1; |
826 | bidi_it->prev_for_neutral.type = | |
827 | bidi_it->prev_for_neutral.type_after_w1 = | |
828 | bidi_it->prev_for_neutral.orig_type = UNKNOWN_BT; | |
829 | bidi_it->sor = L2R; /* FIXME: should it be user-selectable? */ | |
830 | bidi_it->disp_pos = -1; /* invalid/unknown */ | |
0c95fcf7 | 831 | bidi_it->disp_prop = 0; |
cbb09f04 EZ |
832 | /* We can only shrink the cache if we are at the bottom level of its |
833 | "stack". */ | |
834 | if (bidi_cache_start == 0) | |
835 | bidi_cache_shrink (); | |
58b9f433 EZ |
836 | else |
837 | bidi_cache_reset (); | |
b7b65b15 EZ |
838 | } |
839 | ||
182ce2d2 | 840 | /* Perform initializations for reordering a new line of bidi text. */ |
6bff6497 | 841 | static void |
be39f003 EZ |
842 | bidi_line_init (struct bidi_it *bidi_it) |
843 | { | |
844 | bidi_it->scan_dir = 1; /* FIXME: do we need to have control on this? */ | |
845 | bidi_it->resolved_level = bidi_it->level_stack[0].level; | |
846 | bidi_it->level_stack[0].override = NEUTRAL_DIR; /* X1 */ | |
847 | bidi_it->invalid_levels = 0; | |
848 | bidi_it->invalid_rl_levels = -1; | |
849 | bidi_it->next_en_pos = -1; | |
850 | bidi_it->next_for_ws.type = UNKNOWN_BT; | |
b44d9321 EZ |
851 | bidi_set_sor_type (bidi_it, |
852 | bidi_it->paragraph_dir == R2L ? 1 : 0, | |
be39f003 EZ |
853 | bidi_it->level_stack[0].level); /* X10 */ |
854 | ||
855 | bidi_cache_reset (); | |
856 | } | |
857 | ||
cbb09f04 EZ |
858 | \f |
859 | /*********************************************************************** | |
860 | Fetching characters | |
861 | ***********************************************************************/ | |
862 | ||
f3014ef5 EZ |
863 | /* Count bytes in string S between BEG/BEGBYTE and END. BEG and END |
864 | are zero-based character positions in S, BEGBYTE is byte position | |
865 | corresponding to BEG. UNIBYTE, if non-zero, means S is a unibyte | |
866 | string. */ | |
87e67904 EZ |
867 | static inline EMACS_INT |
868 | bidi_count_bytes (const unsigned char *s, const EMACS_INT beg, | |
f3014ef5 | 869 | const EMACS_INT begbyte, const EMACS_INT end, int unibyte) |
87e67904 EZ |
870 | { |
871 | EMACS_INT pos = beg; | |
872 | const unsigned char *p = s + begbyte, *start = p; | |
873 | ||
f3014ef5 EZ |
874 | if (unibyte) |
875 | p = s + end; | |
876 | else | |
87e67904 | 877 | { |
f3014ef5 EZ |
878 | if (!CHAR_HEAD_P (*p)) |
879 | abort (); | |
880 | ||
881 | while (pos < end) | |
882 | { | |
883 | p += BYTES_BY_CHAR_HEAD (*p); | |
884 | pos++; | |
885 | } | |
87e67904 EZ |
886 | } |
887 | ||
888 | return p - start; | |
889 | } | |
890 | ||
891 | /* Fetch and returns the character at byte position BYTEPOS. If S is | |
892 | non-NULL, fetch the character from string S; otherwise fetch the | |
f3014ef5 EZ |
893 | character from the current buffer. UNIBYTE non-zero means S is a |
894 | unibyte string. */ | |
87e67904 | 895 | static inline int |
f3014ef5 | 896 | bidi_char_at_pos (EMACS_INT bytepos, const unsigned char *s, int unibyte) |
87e67904 EZ |
897 | { |
898 | if (s) | |
f3014ef5 EZ |
899 | { |
900 | if (unibyte) | |
901 | return s[bytepos]; | |
902 | else | |
903 | return STRING_CHAR (s + bytepos); | |
904 | } | |
87e67904 EZ |
905 | else |
906 | return FETCH_MULTIBYTE_CHAR (bytepos); | |
907 | } | |
908 | ||
102ebb00 EZ |
909 | /* Fetch and return the character at BYTEPOS/CHARPOS. If that |
910 | character is covered by a display string, treat the entire run of | |
0c95fcf7 EZ |
911 | covered characters as a single character, either u+2029 or u+FFFC, |
912 | and return their combined length in CH_LEN and NCHARS. DISP_POS | |
913 | specifies the character position of the next display string, or -1 | |
914 | if not yet computed. DISP_PROP non-zero means that there's really | |
915 | a display string at DISP_POS, as opposed to when we searched till | |
916 | DISP_POS without finding one. If DISP_PROP is 2, it means the | |
917 | display spec is of the form `(space ...)', which is replaced with | |
918 | u+2029 to handle it as a paragraph separator. When the next | |
919 | character is at or beyond that position, the function updates | |
920 | DISP_POS with the position of the next display string. STRING->s | |
921 | is the C string to iterate, or NULL if iterating over a buffer or a | |
922 | Lisp string; in the latter case, STRING->lstring is the Lisp | |
923 | string. */ | |
fec2107c | 924 | static inline int |
102ebb00 | 925 | bidi_fetch_char (EMACS_INT bytepos, EMACS_INT charpos, EMACS_INT *disp_pos, |
0c95fcf7 | 926 | int *disp_prop, struct bidi_string_data *string, |
fc6f18ce | 927 | int frame_window_p, EMACS_INT *ch_len, EMACS_INT *nchars) |
182ce2d2 EZ |
928 | { |
929 | int ch; | |
9f257352 EZ |
930 | EMACS_INT endpos = |
931 | (string->s || STRINGP (string->lstring)) ? string->schars : ZV; | |
6db161be | 932 | struct text_pos pos; |
182ce2d2 | 933 | |
182ce2d2 | 934 | /* If we got past the last known position of display string, compute |
87e67904 EZ |
935 | the position of the next one. That position could be at CHARPOS. */ |
936 | if (charpos < endpos && charpos > *disp_pos) | |
6db161be EZ |
937 | { |
938 | SET_TEXT_POS (pos, charpos, bytepos); | |
55439c61 | 939 | *disp_pos = compute_display_string_pos (&pos, string, frame_window_p, |
0c95fcf7 | 940 | disp_prop); |
6db161be | 941 | } |
102ebb00 EZ |
942 | |
943 | /* Fetch the character at BYTEPOS. */ | |
87e67904 | 944 | if (charpos >= endpos) |
182ce2d2 EZ |
945 | { |
946 | ch = BIDI_EOB; | |
947 | *ch_len = 1; | |
948 | *nchars = 1; | |
87e67904 | 949 | *disp_pos = endpos; |
0c95fcf7 | 950 | *disp_prop = 0; |
182ce2d2 | 951 | } |
0c95fcf7 | 952 | else if (charpos >= *disp_pos && *disp_prop) |
182ce2d2 | 953 | { |
7b600102 EZ |
954 | EMACS_INT disp_end_pos; |
955 | ||
956 | /* We don't expect to find ourselves in the middle of a display | |
957 | property. Hopefully, it will never be needed. */ | |
958 | if (charpos > *disp_pos) | |
959 | abort (); | |
0c95fcf7 EZ |
960 | /* Text covered by `display' properties and overlays with |
961 | display properties or display strings is handled as a single | |
962 | character that represents the entire run of characters | |
963 | covered by the display property. */ | |
964 | if (*disp_prop == 2) | |
965 | { | |
966 | /* `(space ...)' display specs are handled as paragraph | |
967 | separators for the purposes of the reordering; see UAX#9 | |
968 | section 3 and clause HL1 in section 4.3 there. */ | |
969 | ch = 0x2029; | |
970 | } | |
971 | else | |
972 | { | |
973 | /* All other display specs are handled as the Unicode Object | |
974 | Replacement Character. */ | |
975 | ch = 0xFFFC; | |
976 | } | |
87e67904 | 977 | disp_end_pos = compute_display_string_end (*disp_pos, string); |
7b600102 | 978 | *nchars = disp_end_pos - *disp_pos; |
f3014ef5 EZ |
979 | if (*nchars <= 0) |
980 | abort (); | |
87e67904 EZ |
981 | if (string->s) |
982 | *ch_len = bidi_count_bytes (string->s, *disp_pos, bytepos, | |
f3014ef5 | 983 | disp_end_pos, string->unibyte); |
9f257352 EZ |
984 | else if (STRINGP (string->lstring)) |
985 | *ch_len = bidi_count_bytes (SDATA (string->lstring), *disp_pos, | |
f3014ef5 | 986 | bytepos, disp_end_pos, string->unibyte); |
87e67904 EZ |
987 | else |
988 | *ch_len = CHAR_TO_BYTE (disp_end_pos) - bytepos; | |
182ce2d2 | 989 | } |
182ce2d2 EZ |
990 | else |
991 | { | |
87e67904 EZ |
992 | if (string->s) |
993 | { | |
66ff7ddf | 994 | int len; |
87e67904 | 995 | |
f3014ef5 EZ |
996 | if (!string->unibyte) |
997 | { | |
998 | ch = STRING_CHAR_AND_LENGTH (string->s + bytepos, len); | |
999 | *ch_len = len; | |
1000 | } | |
1001 | else | |
1002 | { | |
1003 | ch = UNIBYTE_TO_CHAR (string->s[bytepos]); | |
1004 | *ch_len = 1; | |
1005 | } | |
87e67904 | 1006 | } |
9f257352 EZ |
1007 | else if (STRINGP (string->lstring)) |
1008 | { | |
66ff7ddf | 1009 | int len; |
9f257352 | 1010 | |
f3014ef5 EZ |
1011 | if (!string->unibyte) |
1012 | { | |
1013 | ch = STRING_CHAR_AND_LENGTH (SDATA (string->lstring) + bytepos, | |
1014 | len); | |
1015 | *ch_len = len; | |
1016 | } | |
1017 | else | |
1018 | { | |
1019 | ch = UNIBYTE_TO_CHAR (SREF (string->lstring, bytepos)); | |
1020 | *ch_len = 1; | |
1021 | } | |
9f257352 | 1022 | } |
87e67904 EZ |
1023 | else |
1024 | { | |
1025 | ch = FETCH_MULTIBYTE_CHAR (bytepos); | |
1026 | *ch_len = CHAR_BYTES (ch); | |
1027 | } | |
182ce2d2 EZ |
1028 | *nchars = 1; |
1029 | } | |
1030 | ||
1031 | /* If we just entered a run of characters covered by a display | |
1032 | string, compute the position of the next display string. */ | |
55439c61 | 1033 | if (charpos + *nchars <= endpos && charpos + *nchars > *disp_pos |
0c95fcf7 | 1034 | && *disp_prop) |
6db161be EZ |
1035 | { |
1036 | SET_TEXT_POS (pos, charpos + *nchars, bytepos + *ch_len); | |
55439c61 | 1037 | *disp_pos = compute_display_string_pos (&pos, string, frame_window_p, |
0c95fcf7 | 1038 | disp_prop); |
6db161be | 1039 | } |
182ce2d2 EZ |
1040 | |
1041 | return ch; | |
1042 | } | |
1043 | ||
cbb09f04 EZ |
1044 | \f |
1045 | /*********************************************************************** | |
1046 | Determining paragraph direction | |
1047 | ***********************************************************************/ | |
1048 | ||
1049 | /* Check if buffer position CHARPOS/BYTEPOS is the end of a paragraph. | |
1050 | Value is the non-negative length of the paragraph separator | |
1051 | following the buffer position, -1 if position is at the beginning | |
1052 | of a new paragraph, or -2 if position is neither at beginning nor | |
1053 | at end of a paragraph. */ | |
1054 | static EMACS_INT | |
1055 | bidi_at_paragraph_end (EMACS_INT charpos, EMACS_INT bytepos) | |
1056 | { | |
1057 | Lisp_Object sep_re; | |
1058 | Lisp_Object start_re; | |
1059 | EMACS_INT val; | |
1060 | ||
1061 | sep_re = paragraph_separate_re; | |
1062 | start_re = paragraph_start_re; | |
1063 | ||
1064 | val = fast_looking_at (sep_re, charpos, bytepos, ZV, ZV_BYTE, Qnil); | |
1065 | if (val < 0) | |
1066 | { | |
1067 | if (fast_looking_at (start_re, charpos, bytepos, ZV, ZV_BYTE, Qnil) >= 0) | |
1068 | val = -1; | |
1069 | else | |
1070 | val = -2; | |
1071 | } | |
1072 | ||
1073 | return val; | |
1074 | } | |
1075 | ||
be39f003 EZ |
1076 | /* Find the beginning of this paragraph by looking back in the buffer. |
1077 | Value is the byte position of the paragraph's beginning. */ | |
1078 | static EMACS_INT | |
b44d9321 | 1079 | bidi_find_paragraph_start (EMACS_INT pos, EMACS_INT pos_byte) |
6bff6497 | 1080 | { |
372b7a95 | 1081 | Lisp_Object re = paragraph_start_re; |
6bff6497 EZ |
1082 | EMACS_INT limit = ZV, limit_byte = ZV_BYTE; |
1083 | ||
6bff6497 EZ |
1084 | while (pos_byte > BEGV_BYTE |
1085 | && fast_looking_at (re, pos, pos_byte, limit, limit_byte, Qnil) < 0) | |
1086 | { | |
182ce2d2 EZ |
1087 | /* FIXME: What if the paragraph beginning is covered by a |
1088 | display string? And what if a display string covering some | |
1089 | of the text over which we scan back includes | |
1090 | paragraph_start_re? */ | |
be39f003 EZ |
1091 | pos = find_next_newline_no_quit (pos - 1, -1); |
1092 | pos_byte = CHAR_TO_BYTE (pos); | |
6bff6497 | 1093 | } |
be39f003 | 1094 | return pos_byte; |
6bff6497 EZ |
1095 | } |
1096 | ||
bea4f10c | 1097 | /* Determine the base direction, a.k.a. base embedding level, of the |
b44d9321 EZ |
1098 | paragraph we are about to iterate through. If DIR is either L2R or |
1099 | R2L, just use that. Otherwise, determine the paragraph direction | |
bea4f10c EZ |
1100 | from the first strong directional character of the paragraph. |
1101 | ||
182ce2d2 | 1102 | NO_DEFAULT_P non-zero means don't default to L2R if the paragraph |
bea4f10c EZ |
1103 | has no strong directional characters and both DIR and |
1104 | bidi_it->paragraph_dir are NEUTRAL_DIR. In that case, search back | |
1105 | in the buffer until a paragraph is found with a strong character, | |
1106 | or until hitting BEGV. In the latter case, fall back to L2R. This | |
1107 | flag is used in current-bidi-paragraph-direction. | |
1108 | ||
1109 | Note that this function gives the paragraph separator the same | |
1110 | direction as the preceding paragraph, even though Emacs generally | |
1111 | views the separartor as not belonging to any paragraph. */ | |
b7b65b15 | 1112 | void |
bea4f10c | 1113 | bidi_paragraph_init (bidi_dir_t dir, struct bidi_it *bidi_it, int no_default_p) |
b7b65b15 | 1114 | { |
6bff6497 | 1115 | EMACS_INT bytepos = bidi_it->bytepos; |
9f257352 | 1116 | int string_p = bidi_it->string.s != NULL || STRINGP (bidi_it->string.lstring); |
bea4f10c | 1117 | EMACS_INT pstartbyte; |
87e67904 EZ |
1118 | /* Note that begbyte is a byte position, while end is a character |
1119 | position. Yes, this is ugly, but we are trying to avoid costly | |
1120 | calls to BYTE_TO_CHAR and its ilk. */ | |
1121 | EMACS_INT begbyte = string_p ? 0 : BEGV_BYTE; | |
1122 | EMACS_INT end = string_p ? bidi_it->string.schars : ZV; | |
e7402cb2 | 1123 | |
5e65aec0 | 1124 | /* Special case for an empty buffer. */ |
87e67904 | 1125 | if (bytepos == begbyte && bidi_it->charpos == end) |
5e65aec0 | 1126 | dir = L2R; |
9c82e145 | 1127 | /* We should never be called at EOB or before BEGV. */ |
87e67904 | 1128 | else if (bidi_it->charpos >= end || bytepos < begbyte) |
9c82e145 EZ |
1129 | abort (); |
1130 | ||
be39f003 EZ |
1131 | if (dir == L2R) |
1132 | { | |
1133 | bidi_it->paragraph_dir = L2R; | |
1134 | bidi_it->new_paragraph = 0; | |
1135 | } | |
1136 | else if (dir == R2L) | |
1137 | { | |
1138 | bidi_it->paragraph_dir = R2L; | |
1139 | bidi_it->new_paragraph = 0; | |
1140 | } | |
b7b65b15 EZ |
1141 | else if (dir == NEUTRAL_DIR) /* P2 */ |
1142 | { | |
182ce2d2 EZ |
1143 | int ch; |
1144 | EMACS_INT ch_len, nchars; | |
1145 | EMACS_INT pos, disp_pos = -1; | |
0c95fcf7 | 1146 | int disp_prop = 0; |
6bff6497 | 1147 | bidi_type_t type; |
9f257352 | 1148 | const unsigned char *s; |
be39f003 | 1149 | |
d20e1419 EZ |
1150 | if (!bidi_initialized) |
1151 | bidi_initialize (); | |
1152 | ||
be39f003 EZ |
1153 | /* If we are inside a paragraph separator, we are just waiting |
1154 | for the separator to be exhausted; use the previous paragraph | |
e5a2fec7 EZ |
1155 | direction. But don't do that if we have been just reseated, |
1156 | because we need to reinitialize below in that case. */ | |
1157 | if (!bidi_it->first_elt | |
1158 | && bidi_it->charpos < bidi_it->separator_limit) | |
be39f003 EZ |
1159 | return; |
1160 | ||
b44d9321 | 1161 | /* If we are on a newline, get past it to where the next |
5e65aec0 EZ |
1162 | paragraph might start. But don't do that at BEGV since then |
1163 | we are potentially in a new paragraph that doesn't yet | |
1164 | exist. */ | |
c143c213 | 1165 | pos = bidi_it->charpos; |
9f257352 EZ |
1166 | s = STRINGP (bidi_it->string.lstring) ? |
1167 | SDATA (bidi_it->string.lstring) : bidi_it->string.s; | |
f3014ef5 EZ |
1168 | if (bytepos > begbyte |
1169 | && bidi_char_at_pos (bytepos, s, bidi_it->string.unibyte) == '\n') | |
be39f003 | 1170 | { |
b44d9321 | 1171 | bytepos++; |
c143c213 | 1172 | pos++; |
be39f003 | 1173 | } |
b44d9321 EZ |
1174 | |
1175 | /* We are either at the beginning of a paragraph or in the | |
1176 | middle of it. Find where this paragraph starts. */ | |
87e67904 EZ |
1177 | if (string_p) |
1178 | { | |
1179 | /* We don't support changes of paragraph direction inside a | |
1180 | string. It is treated as a single paragraph. */ | |
1181 | pstartbyte = 0; | |
1182 | } | |
1183 | else | |
1184 | pstartbyte = bidi_find_paragraph_start (pos, bytepos); | |
be39f003 EZ |
1185 | bidi_it->separator_limit = -1; |
1186 | bidi_it->new_paragraph = 0; | |
bea4f10c EZ |
1187 | |
1188 | /* The following loop is run more than once only if NO_DEFAULT_P | |
87e67904 | 1189 | is non-zero, and only if we are iterating on a buffer. */ |
bea4f10c EZ |
1190 | do { |
1191 | bytepos = pstartbyte; | |
87e67904 EZ |
1192 | if (!string_p) |
1193 | pos = BYTE_TO_CHAR (bytepos); | |
0c95fcf7 | 1194 | ch = bidi_fetch_char (bytepos, pos, &disp_pos, &disp_prop, |
55439c61 | 1195 | &bidi_it->string, |
87e67904 | 1196 | bidi_it->frame_window_p, &ch_len, &nchars); |
bea4f10c EZ |
1197 | type = bidi_get_type (ch, NEUTRAL_DIR); |
1198 | ||
182ce2d2 | 1199 | for (pos += nchars, bytepos += ch_len; |
bea4f10c EZ |
1200 | (bidi_get_category (type) != STRONG) |
1201 | || (bidi_ignore_explicit_marks_for_paragraph_level | |
1202 | && (type == RLE || type == RLO | |
1203 | || type == LRE || type == LRO)); | |
1204 | type = bidi_get_type (ch, NEUTRAL_DIR)) | |
1205 | { | |
87e67904 | 1206 | if (pos >= end) |
bea4f10c EZ |
1207 | { |
1208 | /* Pretend there's a paragraph separator at end of | |
87e67904 | 1209 | buffer/string. */ |
bea4f10c EZ |
1210 | type = NEUTRAL_B; |
1211 | break; | |
1212 | } | |
0bb23927 EZ |
1213 | if (!string_p |
1214 | && type == NEUTRAL_B | |
1215 | && bidi_at_paragraph_end (pos, bytepos) >= -1) | |
cf99dcf8 | 1216 | break; |
102ebb00 | 1217 | /* Fetch next character and advance to get past it. */ |
55439c61 | 1218 | ch = bidi_fetch_char (bytepos, pos, &disp_pos, |
0c95fcf7 | 1219 | &disp_prop, &bidi_it->string, |
fc6f18ce | 1220 | bidi_it->frame_window_p, &ch_len, &nchars); |
182ce2d2 EZ |
1221 | pos += nchars; |
1222 | bytepos += ch_len; | |
bea4f10c | 1223 | } |
32413314 EZ |
1224 | if ((type == STRONG_R || type == STRONG_AL) /* P3 */ |
1225 | || (!bidi_ignore_explicit_marks_for_paragraph_level | |
1226 | && (type == RLO || type == RLE))) | |
bea4f10c | 1227 | bidi_it->paragraph_dir = R2L; |
32413314 EZ |
1228 | else if (type == STRONG_L |
1229 | || (!bidi_ignore_explicit_marks_for_paragraph_level | |
1230 | && (type == LRO || type == LRE))) | |
bea4f10c | 1231 | bidi_it->paragraph_dir = L2R; |
87e67904 EZ |
1232 | if (!string_p |
1233 | && no_default_p && bidi_it->paragraph_dir == NEUTRAL_DIR) | |
bea4f10c EZ |
1234 | { |
1235 | /* If this paragraph is at BEGV, default to L2R. */ | |
1236 | if (pstartbyte == BEGV_BYTE) | |
1237 | bidi_it->paragraph_dir = L2R; /* P3 and HL1 */ | |
1238 | else | |
1239 | { | |
1240 | EMACS_INT prevpbyte = pstartbyte; | |
1241 | EMACS_INT p = BYTE_TO_CHAR (pstartbyte), pbyte = pstartbyte; | |
1242 | ||
1243 | /* Find the beginning of the previous paragraph, if any. */ | |
1244 | while (pbyte > BEGV_BYTE && prevpbyte >= pstartbyte) | |
1245 | { | |
182ce2d2 EZ |
1246 | /* FXIME: What if p is covered by a display |
1247 | string? See also a FIXME inside | |
1248 | bidi_find_paragraph_start. */ | |
bea4f10c EZ |
1249 | p--; |
1250 | pbyte = CHAR_TO_BYTE (p); | |
1251 | prevpbyte = bidi_find_paragraph_start (p, pbyte); | |
1252 | } | |
1253 | pstartbyte = prevpbyte; | |
1254 | } | |
1255 | } | |
87e67904 EZ |
1256 | } while (!string_p |
1257 | && no_default_p && bidi_it->paragraph_dir == NEUTRAL_DIR); | |
b7b65b15 | 1258 | } |
be39f003 EZ |
1259 | else |
1260 | abort (); | |
b7b65b15 | 1261 | |
b44d9321 EZ |
1262 | /* Contrary to UAX#9 clause P3, we only default the paragraph |
1263 | direction to L2R if we have no previous usable paragraph | |
bea4f10c | 1264 | direction. This is allowed by the HL1 clause. */ |
d20e1419 | 1265 | if (bidi_it->paragraph_dir != L2R && bidi_it->paragraph_dir != R2L) |
bea4f10c | 1266 | bidi_it->paragraph_dir = L2R; /* P3 and HL1 ``higher-level protocols'' */ |
be39f003 | 1267 | if (bidi_it->paragraph_dir == R2L) |
b44d9321 | 1268 | bidi_it->level_stack[0].level = 1; |
be39f003 | 1269 | else |
b44d9321 | 1270 | bidi_it->level_stack[0].level = 0; |
be39f003 EZ |
1271 | |
1272 | bidi_line_init (bidi_it); | |
b7b65b15 EZ |
1273 | } |
1274 | ||
cbb09f04 EZ |
1275 | \f |
1276 | /*********************************************************************** | |
1277 | Resolving explicit and implicit levels. | |
58b9f433 | 1278 | The rest of this file constitutes the core of the UBA implementation. |
cbb09f04 | 1279 | ***********************************************************************/ |
b7b65b15 | 1280 | |
55d4c1b2 | 1281 | static inline int |
182ce2d2 | 1282 | bidi_explicit_dir_char (int ch) |
b7b65b15 | 1283 | { |
182ce2d2 EZ |
1284 | bidi_type_t ch_type; |
1285 | ||
1286 | if (!bidi_initialized) | |
1287 | abort (); | |
1288 | ch_type = (bidi_type_t) XINT (CHAR_TABLE_REF (bidi_type_table, ch)); | |
1289 | return (ch_type == LRE || ch_type == LRO | |
1290 | || ch_type == RLE || ch_type == RLO | |
1291 | || ch_type == PDF); | |
b7b65b15 EZ |
1292 | } |
1293 | ||
1294 | /* A helper function for bidi_resolve_explicit. It advances to the | |
1295 | next character in logical order and determines the new embedding | |
1296 | level and directional override, but does not take into account | |
1297 | empty embeddings. */ | |
1298 | static int | |
1299 | bidi_resolve_explicit_1 (struct bidi_it *bidi_it) | |
1300 | { | |
1301 | int curchar; | |
1302 | bidi_type_t type; | |
1303 | int current_level; | |
1304 | int new_level; | |
1305 | bidi_dir_t override; | |
9f257352 | 1306 | int string_p = bidi_it->string.s != NULL || STRINGP (bidi_it->string.lstring); |
b7b65b15 | 1307 | |
182ce2d2 EZ |
1308 | /* If reseat()'ed, don't advance, so as to start iteration from the |
1309 | position where we were reseated. bidi_it->bytepos can be less | |
1310 | than BEGV_BYTE after reseat to BEGV. */ | |
87e67904 | 1311 | if (bidi_it->bytepos < (string_p ? 0 : BEGV_BYTE) |
9c82e145 | 1312 | || bidi_it->first_elt) |
e7402cb2 | 1313 | { |
9c82e145 | 1314 | bidi_it->first_elt = 0; |
87e67904 EZ |
1315 | if (string_p) |
1316 | { | |
9f257352 EZ |
1317 | const unsigned char *p = |
1318 | STRINGP (bidi_it->string.lstring) | |
1319 | ? SDATA (bidi_it->string.lstring) : bidi_it->string.s; | |
1320 | ||
87e67904 EZ |
1321 | if (bidi_it->charpos < 0) |
1322 | bidi_it->charpos = 0; | |
f3014ef5 EZ |
1323 | bidi_it->bytepos = bidi_count_bytes (p, 0, 0, bidi_it->charpos, |
1324 | bidi_it->string.unibyte); | |
87e67904 EZ |
1325 | } |
1326 | else | |
1327 | { | |
1328 | if (bidi_it->charpos < BEGV) | |
1329 | bidi_it->charpos = BEGV; | |
1330 | bidi_it->bytepos = CHAR_TO_BYTE (bidi_it->charpos); | |
1331 | } | |
e7402cb2 | 1332 | } |
87e67904 EZ |
1333 | /* Don't move at end of buffer/string. */ |
1334 | else if (bidi_it->charpos < (string_p ? bidi_it->string.schars : ZV)) | |
b7b65b15 | 1335 | { |
182ce2d2 EZ |
1336 | /* Advance to the next character, skipping characters covered by |
1337 | display strings (nchars > 1). */ | |
102ebb00 EZ |
1338 | if (bidi_it->nchars <= 0) |
1339 | abort (); | |
182ce2d2 | 1340 | bidi_it->charpos += bidi_it->nchars; |
e7402cb2 EZ |
1341 | if (bidi_it->ch_len == 0) |
1342 | abort (); | |
b7b65b15 EZ |
1343 | bidi_it->bytepos += bidi_it->ch_len; |
1344 | } | |
1345 | ||
1346 | current_level = bidi_it->level_stack[bidi_it->stack_idx].level; /* X1 */ | |
1347 | override = bidi_it->level_stack[bidi_it->stack_idx].override; | |
1348 | new_level = current_level; | |
1349 | ||
87e67904 | 1350 | if (bidi_it->charpos >= (string_p ? bidi_it->string.schars : ZV)) |
e7402cb2 EZ |
1351 | { |
1352 | curchar = BIDI_EOB; | |
1353 | bidi_it->ch_len = 1; | |
182ce2d2 | 1354 | bidi_it->nchars = 1; |
87e67904 | 1355 | bidi_it->disp_pos = (string_p ? bidi_it->string.schars : ZV); |
0c95fcf7 | 1356 | bidi_it->disp_prop = 0; |
e7402cb2 EZ |
1357 | } |
1358 | else | |
1359 | { | |
182ce2d2 EZ |
1360 | /* Fetch the character at BYTEPOS. If it is covered by a |
1361 | display string, treat the entire run of covered characters as | |
1362 | a single character u+FFFC. */ | |
102ebb00 | 1363 | curchar = bidi_fetch_char (bidi_it->bytepos, bidi_it->charpos, |
0c95fcf7 | 1364 | &bidi_it->disp_pos, &bidi_it->disp_prop, |
55439c61 | 1365 | &bidi_it->string, bidi_it->frame_window_p, |
102ebb00 | 1366 | &bidi_it->ch_len, &bidi_it->nchars); |
e7402cb2 | 1367 | } |
b7b65b15 | 1368 | bidi_it->ch = curchar; |
b7b65b15 | 1369 | |
6bff6497 EZ |
1370 | /* Don't apply directional override here, as all the types we handle |
1371 | below will not be affected by the override anyway, and we need | |
1372 | the original type unaltered. The override will be applied in | |
1373 | bidi_resolve_weak. */ | |
1374 | type = bidi_get_type (curchar, NEUTRAL_DIR); | |
89d3374a EZ |
1375 | bidi_it->orig_type = type; |
1376 | bidi_check_type (bidi_it->orig_type); | |
b7b65b15 EZ |
1377 | |
1378 | if (type != PDF) | |
1379 | bidi_it->prev_was_pdf = 0; | |
1380 | ||
89d3374a | 1381 | bidi_it->type_after_w1 = UNKNOWN_BT; |
b7b65b15 EZ |
1382 | |
1383 | switch (type) | |
1384 | { | |
1385 | case RLE: /* X2 */ | |
1386 | case RLO: /* X4 */ | |
89d3374a EZ |
1387 | bidi_it->type_after_w1 = type; |
1388 | bidi_check_type (bidi_it->type_after_w1); | |
b7b65b15 | 1389 | type = WEAK_BN; /* X9/Retaining */ |
87e67904 | 1390 | if (bidi_it->ignore_bn_limit <= -1) |
b7b65b15 EZ |
1391 | { |
1392 | if (current_level <= BIDI_MAXLEVEL - 4) | |
1393 | { | |
1394 | /* Compute the least odd embedding level greater than | |
1395 | the current level. */ | |
1396 | new_level = ((current_level + 1) & ~1) + 1; | |
89d3374a | 1397 | if (bidi_it->type_after_w1 == RLE) |
b7b65b15 EZ |
1398 | override = NEUTRAL_DIR; |
1399 | else | |
1400 | override = R2L; | |
1401 | if (current_level == BIDI_MAXLEVEL - 4) | |
1402 | bidi_it->invalid_rl_levels = 0; | |
1403 | bidi_push_embedding_level (bidi_it, new_level, override); | |
1404 | } | |
1405 | else | |
1406 | { | |
1407 | bidi_it->invalid_levels++; | |
1408 | /* See the commentary about invalid_rl_levels below. */ | |
1409 | if (bidi_it->invalid_rl_levels < 0) | |
1410 | bidi_it->invalid_rl_levels = 0; | |
1411 | bidi_it->invalid_rl_levels++; | |
1412 | } | |
1413 | } | |
89d3374a | 1414 | else if (bidi_it->prev.type_after_w1 == WEAK_EN /* W5/Retaining */ |
b7b65b15 EZ |
1415 | || bidi_it->next_en_pos > bidi_it->charpos) |
1416 | type = WEAK_EN; | |
1417 | break; | |
1418 | case LRE: /* X3 */ | |
1419 | case LRO: /* X5 */ | |
89d3374a EZ |
1420 | bidi_it->type_after_w1 = type; |
1421 | bidi_check_type (bidi_it->type_after_w1); | |
b7b65b15 | 1422 | type = WEAK_BN; /* X9/Retaining */ |
87e67904 | 1423 | if (bidi_it->ignore_bn_limit <= -1) |
b7b65b15 EZ |
1424 | { |
1425 | if (current_level <= BIDI_MAXLEVEL - 5) | |
1426 | { | |
1427 | /* Compute the least even embedding level greater than | |
1428 | the current level. */ | |
1429 | new_level = ((current_level + 2) & ~1); | |
89d3374a | 1430 | if (bidi_it->type_after_w1 == LRE) |
b7b65b15 EZ |
1431 | override = NEUTRAL_DIR; |
1432 | else | |
1433 | override = L2R; | |
1434 | bidi_push_embedding_level (bidi_it, new_level, override); | |
1435 | } | |
1436 | else | |
1437 | { | |
1438 | bidi_it->invalid_levels++; | |
1439 | /* invalid_rl_levels counts invalid levels encountered | |
1440 | while the embedding level was already too high for | |
1441 | LRE/LRO, but not for RLE/RLO. That is because | |
1442 | there may be exactly one PDF which we should not | |
1443 | ignore even though invalid_levels is non-zero. | |
1444 | invalid_rl_levels helps to know what PDF is | |
1445 | that. */ | |
1446 | if (bidi_it->invalid_rl_levels >= 0) | |
1447 | bidi_it->invalid_rl_levels++; | |
1448 | } | |
1449 | } | |
89d3374a | 1450 | else if (bidi_it->prev.type_after_w1 == WEAK_EN /* W5/Retaining */ |
b7b65b15 EZ |
1451 | || bidi_it->next_en_pos > bidi_it->charpos) |
1452 | type = WEAK_EN; | |
1453 | break; | |
1454 | case PDF: /* X7 */ | |
89d3374a EZ |
1455 | bidi_it->type_after_w1 = type; |
1456 | bidi_check_type (bidi_it->type_after_w1); | |
b7b65b15 | 1457 | type = WEAK_BN; /* X9/Retaining */ |
87e67904 | 1458 | if (bidi_it->ignore_bn_limit <= -1) |
b7b65b15 EZ |
1459 | { |
1460 | if (!bidi_it->invalid_rl_levels) | |
1461 | { | |
1462 | new_level = bidi_pop_embedding_level (bidi_it); | |
1463 | bidi_it->invalid_rl_levels = -1; | |
1464 | if (bidi_it->invalid_levels) | |
1465 | bidi_it->invalid_levels--; | |
1466 | /* else nothing: UAX#9 says to ignore invalid PDFs */ | |
1467 | } | |
1468 | if (!bidi_it->invalid_levels) | |
1469 | new_level = bidi_pop_embedding_level (bidi_it); | |
1470 | else | |
1471 | { | |
1472 | bidi_it->invalid_levels--; | |
1473 | bidi_it->invalid_rl_levels--; | |
1474 | } | |
1475 | } | |
89d3374a | 1476 | else if (bidi_it->prev.type_after_w1 == WEAK_EN /* W5/Retaining */ |
b7b65b15 EZ |
1477 | || bidi_it->next_en_pos > bidi_it->charpos) |
1478 | type = WEAK_EN; | |
1479 | break; | |
1480 | default: | |
1481 | /* Nothing. */ | |
1482 | break; | |
1483 | } | |
1484 | ||
1485 | bidi_it->type = type; | |
2d6e4628 | 1486 | bidi_check_type (bidi_it->type); |
b7b65b15 EZ |
1487 | |
1488 | return new_level; | |
1489 | } | |
1490 | ||
1491 | /* Given an iterator state in BIDI_IT, advance one character position | |
182ce2d2 EZ |
1492 | in the buffer/string to the next character (in the logical order), |
1493 | resolve any explicit embeddings and directional overrides, and | |
1494 | return the embedding level of the character after resolving | |
1495 | explicit directives and ignoring empty embeddings. */ | |
b7b65b15 EZ |
1496 | static int |
1497 | bidi_resolve_explicit (struct bidi_it *bidi_it) | |
1498 | { | |
1499 | int prev_level = bidi_it->level_stack[bidi_it->stack_idx].level; | |
1500 | int new_level = bidi_resolve_explicit_1 (bidi_it); | |
87e67904 | 1501 | EMACS_INT eob = bidi_it->string.s ? bidi_it->string.schars : ZV; |
9f257352 EZ |
1502 | const unsigned char *s = STRINGP (bidi_it->string.lstring) |
1503 | ? SDATA (bidi_it->string.lstring) : bidi_it->string.s; | |
b7b65b15 EZ |
1504 | |
1505 | if (prev_level < new_level | |
1506 | && bidi_it->type == WEAK_BN | |
87e67904 EZ |
1507 | && bidi_it->ignore_bn_limit == -1 /* only if not already known */ |
1508 | && bidi_it->charpos < eob /* not already at EOB */ | |
1509 | && bidi_explicit_dir_char (bidi_char_at_pos (bidi_it->bytepos | |
f3014ef5 EZ |
1510 | + bidi_it->ch_len, s, |
1511 | bidi_it->string.unibyte))) | |
b7b65b15 EZ |
1512 | { |
1513 | /* Avoid pushing and popping embedding levels if the level run | |
1514 | is empty, as this breaks level runs where it shouldn't. | |
1515 | UAX#9 removes all the explicit embedding and override codes, | |
1516 | so empty embeddings disappear without a trace. We need to | |
1517 | behave as if we did the same. */ | |
1518 | struct bidi_it saved_it; | |
1519 | int level = prev_level; | |
1520 | ||
1521 | bidi_copy_it (&saved_it, bidi_it); | |
1522 | ||
87e67904 | 1523 | while (bidi_explicit_dir_char (bidi_char_at_pos (bidi_it->bytepos |
f3014ef5 EZ |
1524 | + bidi_it->ch_len, s, |
1525 | bidi_it->string.unibyte))) | |
b7b65b15 | 1526 | { |
182ce2d2 EZ |
1527 | /* This advances to the next character, skipping any |
1528 | characters covered by display strings. */ | |
b7b65b15 | 1529 | level = bidi_resolve_explicit_1 (bidi_it); |
9f257352 EZ |
1530 | /* If string.lstring was relocated inside bidi_resolve_explicit_1, |
1531 | a pointer to its data is no longer valid. */ | |
1532 | if (STRINGP (bidi_it->string.lstring)) | |
1533 | s = SDATA (bidi_it->string.lstring); | |
b7b65b15 EZ |
1534 | } |
1535 | ||
102ebb00 EZ |
1536 | if (bidi_it->nchars <= 0) |
1537 | abort (); | |
b7b65b15 | 1538 | if (level == prev_level) /* empty embedding */ |
182ce2d2 | 1539 | saved_it.ignore_bn_limit = bidi_it->charpos + bidi_it->nchars; |
b7b65b15 | 1540 | else /* this embedding is non-empty */ |
87e67904 | 1541 | saved_it.ignore_bn_limit = -2; |
b7b65b15 EZ |
1542 | |
1543 | bidi_copy_it (bidi_it, &saved_it); | |
87e67904 | 1544 | if (bidi_it->ignore_bn_limit > -1) |
b7b65b15 EZ |
1545 | { |
1546 | /* We pushed a level, but we shouldn't have. Undo that. */ | |
1547 | if (!bidi_it->invalid_rl_levels) | |
1548 | { | |
1549 | new_level = bidi_pop_embedding_level (bidi_it); | |
1550 | bidi_it->invalid_rl_levels = -1; | |
1551 | if (bidi_it->invalid_levels) | |
1552 | bidi_it->invalid_levels--; | |
1553 | } | |
1554 | if (!bidi_it->invalid_levels) | |
1555 | new_level = bidi_pop_embedding_level (bidi_it); | |
1556 | else | |
1557 | { | |
1558 | bidi_it->invalid_levels--; | |
1559 | bidi_it->invalid_rl_levels--; | |
1560 | } | |
1561 | } | |
1562 | } | |
1563 | ||
b7b65b15 EZ |
1564 | if (bidi_it->type == NEUTRAL_B) /* X8 */ |
1565 | { | |
21fce5ab | 1566 | bidi_set_paragraph_end (bidi_it); |
6bff6497 EZ |
1567 | /* This is needed by bidi_resolve_weak below, and in L1. */ |
1568 | bidi_it->type_after_w1 = bidi_it->type; | |
89d3374a | 1569 | bidi_check_type (bidi_it->type_after_w1); |
b7b65b15 EZ |
1570 | } |
1571 | ||
1572 | return new_level; | |
1573 | } | |
1574 | ||
182ce2d2 EZ |
1575 | /* Advance in the buffer/string, resolve weak types and return the |
1576 | type of the next character after weak type resolution. */ | |
fd3998ff | 1577 | static bidi_type_t |
b7b65b15 EZ |
1578 | bidi_resolve_weak (struct bidi_it *bidi_it) |
1579 | { | |
1580 | bidi_type_t type; | |
1581 | bidi_dir_t override; | |
1582 | int prev_level = bidi_it->level_stack[bidi_it->stack_idx].level; | |
1583 | int new_level = bidi_resolve_explicit (bidi_it); | |
1584 | int next_char; | |
1585 | bidi_type_t type_of_next; | |
1586 | struct bidi_it saved_it; | |
9f257352 EZ |
1587 | EMACS_INT eob = |
1588 | (STRINGP (bidi_it->string.lstring) || bidi_it->string.s) | |
1589 | ? bidi_it->string.schars : ZV; | |
b7b65b15 EZ |
1590 | |
1591 | type = bidi_it->type; | |
1592 | override = bidi_it->level_stack[bidi_it->stack_idx].override; | |
1593 | ||
1594 | if (type == UNKNOWN_BT | |
1595 | || type == LRE | |
1596 | || type == LRO | |
1597 | || type == RLE | |
1598 | || type == RLO | |
1599 | || type == PDF) | |
1600 | abort (); | |
1601 | ||
1602 | if (new_level != prev_level | |
1603 | || bidi_it->type == NEUTRAL_B) | |
1604 | { | |
1605 | /* We've got a new embedding level run, compute the directional | |
1606 | type of sor and initialize per-run variables (UAX#9, clause | |
1607 | X10). */ | |
1608 | bidi_set_sor_type (bidi_it, prev_level, new_level); | |
1609 | } | |
1610 | else if (type == NEUTRAL_S || type == NEUTRAL_WS | |
1611 | || type == WEAK_BN || type == STRONG_AL) | |
89d3374a EZ |
1612 | bidi_it->type_after_w1 = type; /* needed in L1 */ |
1613 | bidi_check_type (bidi_it->type_after_w1); | |
b7b65b15 EZ |
1614 | |
1615 | /* Level and directional override status are already recorded in | |
1616 | bidi_it, and do not need any change; see X6. */ | |
1617 | if (override == R2L) /* X6 */ | |
1618 | type = STRONG_R; | |
1619 | else if (override == L2R) | |
1620 | type = STRONG_L; | |
bc5a45f3 | 1621 | else |
b7b65b15 | 1622 | { |
bc5a45f3 | 1623 | if (type == WEAK_NSM) /* W1 */ |
b7b65b15 | 1624 | { |
bc5a45f3 | 1625 | /* Note that we don't need to consider the case where the |
5930fe97 EZ |
1626 | prev character has its type overridden by an RLO or LRO, |
1627 | because then either the type of this NSM would have been | |
1628 | also overridden, or the previous character is outside the | |
1629 | current level run, and thus not relevant to this NSM. | |
1630 | This is why NSM gets the type_after_w1 of the previous | |
1631 | character. */ | |
ebb5722e EZ |
1632 | if (bidi_it->prev.type_after_w1 != UNKNOWN_BT |
1633 | /* if type_after_w1 is NEUTRAL_B, this NSM is at sor */ | |
1634 | && bidi_it->prev.type_after_w1 != NEUTRAL_B) | |
5930fe97 | 1635 | type = bidi_it->prev.type_after_w1; |
bc5a45f3 EZ |
1636 | else if (bidi_it->sor == R2L) |
1637 | type = STRONG_R; | |
1638 | else if (bidi_it->sor == L2R) | |
1639 | type = STRONG_L; | |
1640 | else /* shouldn't happen! */ | |
1641 | abort (); | |
b7b65b15 | 1642 | } |
bc5a45f3 EZ |
1643 | if (type == WEAK_EN /* W2 */ |
1644 | && bidi_it->last_strong.type_after_w1 == STRONG_AL) | |
1645 | type = WEAK_AN; | |
1646 | else if (type == STRONG_AL) /* W3 */ | |
1647 | type = STRONG_R; | |
1648 | else if ((type == WEAK_ES /* W4 */ | |
1649 | && bidi_it->prev.type_after_w1 == WEAK_EN | |
1650 | && bidi_it->prev.orig_type == WEAK_EN) | |
1651 | || (type == WEAK_CS | |
1652 | && ((bidi_it->prev.type_after_w1 == WEAK_EN | |
1653 | && bidi_it->prev.orig_type == WEAK_EN) | |
1654 | || bidi_it->prev.type_after_w1 == WEAK_AN))) | |
b7b65b15 | 1655 | { |
9f257352 EZ |
1656 | const unsigned char *s = |
1657 | STRINGP (bidi_it->string.lstring) | |
1658 | ? SDATA (bidi_it->string.lstring) : bidi_it->string.s; | |
1659 | ||
e7402cb2 | 1660 | next_char = |
87e67904 EZ |
1661 | bidi_it->charpos + bidi_it->nchars >= eob |
1662 | ? BIDI_EOB | |
f3014ef5 EZ |
1663 | : bidi_char_at_pos (bidi_it->bytepos + bidi_it->ch_len, s, |
1664 | bidi_it->string.unibyte); | |
6bff6497 | 1665 | type_of_next = bidi_get_type (next_char, override); |
b7b65b15 | 1666 | |
bc5a45f3 | 1667 | if (type_of_next == WEAK_BN |
b7b65b15 EZ |
1668 | || bidi_explicit_dir_char (next_char)) |
1669 | { | |
1670 | bidi_copy_it (&saved_it, bidi_it); | |
1671 | while (bidi_resolve_explicit (bidi_it) == new_level | |
bc5a45f3 | 1672 | && bidi_it->type == WEAK_BN) |
b7b65b15 EZ |
1673 | ; |
1674 | type_of_next = bidi_it->type; | |
b7b65b15 EZ |
1675 | bidi_copy_it (bidi_it, &saved_it); |
1676 | } | |
bc5a45f3 EZ |
1677 | |
1678 | /* If the next character is EN, but the last strong-type | |
1679 | character is AL, that next EN will be changed to AN when | |
1680 | we process it in W2 above. So in that case, this ES | |
1681 | should not be changed into EN. */ | |
1682 | if (type == WEAK_ES | |
1683 | && type_of_next == WEAK_EN | |
1684 | && bidi_it->last_strong.type_after_w1 != STRONG_AL) | |
1685 | type = WEAK_EN; | |
1686 | else if (type == WEAK_CS) | |
b7b65b15 | 1687 | { |
bc5a45f3 EZ |
1688 | if (bidi_it->prev.type_after_w1 == WEAK_AN |
1689 | && (type_of_next == WEAK_AN | |
1690 | /* If the next character is EN, but the last | |
1691 | strong-type character is AL, EN will be later | |
1692 | changed to AN when we process it in W2 above. | |
1693 | So in that case, this ES should not be | |
1694 | changed into EN. */ | |
1695 | || (type_of_next == WEAK_EN | |
1696 | && bidi_it->last_strong.type_after_w1 == STRONG_AL))) | |
1697 | type = WEAK_AN; | |
1698 | else if (bidi_it->prev.type_after_w1 == WEAK_EN | |
1699 | && type_of_next == WEAK_EN | |
1700 | && bidi_it->last_strong.type_after_w1 != STRONG_AL) | |
1701 | type = WEAK_EN; | |
1702 | } | |
1703 | } | |
1704 | else if (type == WEAK_ET /* W5: ET with EN before or after it */ | |
1705 | || type == WEAK_BN) /* W5/Retaining */ | |
1706 | { | |
1707 | if (bidi_it->prev.type_after_w1 == WEAK_EN /* ET/BN w/EN before it */ | |
1708 | || bidi_it->next_en_pos > bidi_it->charpos) | |
1709 | type = WEAK_EN; | |
1710 | else /* W5: ET/BN with EN after it. */ | |
1711 | { | |
182ce2d2 | 1712 | EMACS_INT en_pos = bidi_it->charpos + bidi_it->nchars; |
9f257352 EZ |
1713 | const unsigned char *s = |
1714 | STRINGP (bidi_it->string.lstring) | |
1715 | ? SDATA (bidi_it->string.lstring) : bidi_it->string.s; | |
bc5a45f3 | 1716 | |
102ebb00 EZ |
1717 | if (bidi_it->nchars <= 0) |
1718 | abort (); | |
bc5a45f3 | 1719 | next_char = |
87e67904 EZ |
1720 | bidi_it->charpos + bidi_it->nchars >= eob |
1721 | ? BIDI_EOB | |
f3014ef5 EZ |
1722 | : bidi_char_at_pos (bidi_it->bytepos + bidi_it->ch_len, s, |
1723 | bidi_it->string.unibyte); | |
bc5a45f3 EZ |
1724 | type_of_next = bidi_get_type (next_char, override); |
1725 | ||
1726 | if (type_of_next == WEAK_ET | |
1727 | || type_of_next == WEAK_BN | |
1728 | || bidi_explicit_dir_char (next_char)) | |
1729 | { | |
1730 | bidi_copy_it (&saved_it, bidi_it); | |
1731 | while (bidi_resolve_explicit (bidi_it) == new_level | |
1732 | && (bidi_it->type == WEAK_BN | |
1733 | || bidi_it->type == WEAK_ET)) | |
1734 | ; | |
1735 | type_of_next = bidi_it->type; | |
1736 | en_pos = bidi_it->charpos; | |
1737 | bidi_copy_it (bidi_it, &saved_it); | |
1738 | } | |
1739 | if (type_of_next == WEAK_EN) | |
b7b65b15 | 1740 | { |
bc5a45f3 EZ |
1741 | /* If the last strong character is AL, the EN we've |
1742 | found will become AN when we get to it (W2). */ | |
1743 | if (bidi_it->last_strong.type_after_w1 != STRONG_AL) | |
1744 | { | |
1745 | type = WEAK_EN; | |
1746 | /* Remember this EN position, to speed up processing | |
1747 | of the next ETs. */ | |
1748 | bidi_it->next_en_pos = en_pos; | |
1749 | } | |
1750 | else if (type == WEAK_BN) | |
1751 | type = NEUTRAL_ON; /* W6/Retaining */ | |
b7b65b15 | 1752 | } |
b7b65b15 EZ |
1753 | } |
1754 | } | |
1755 | } | |
1756 | ||
1757 | if (type == WEAK_ES || type == WEAK_ET || type == WEAK_CS /* W6 */ | |
89d3374a EZ |
1758 | || (type == WEAK_BN |
1759 | && (bidi_it->prev.type_after_w1 == WEAK_CS /* W6/Retaining */ | |
1760 | || bidi_it->prev.type_after_w1 == WEAK_ES | |
1761 | || bidi_it->prev.type_after_w1 == WEAK_ET))) | |
b7b65b15 EZ |
1762 | type = NEUTRAL_ON; |
1763 | ||
1764 | /* Store the type we've got so far, before we clobber it with strong | |
1765 | types in W7 and while resolving neutral types. But leave alone | |
1766 | the original types that were recorded above, because we will need | |
1767 | them for the L1 clause. */ | |
89d3374a EZ |
1768 | if (bidi_it->type_after_w1 == UNKNOWN_BT) |
1769 | bidi_it->type_after_w1 = type; | |
1770 | bidi_check_type (bidi_it->type_after_w1); | |
b7b65b15 EZ |
1771 | |
1772 | if (type == WEAK_EN) /* W7 */ | |
1773 | { | |
89d3374a | 1774 | if ((bidi_it->last_strong.type_after_w1 == STRONG_L) |
b7b65b15 EZ |
1775 | || (bidi_it->last_strong.type == UNKNOWN_BT && bidi_it->sor == L2R)) |
1776 | type = STRONG_L; | |
1777 | } | |
1778 | ||
1779 | bidi_it->type = type; | |
2d6e4628 | 1780 | bidi_check_type (bidi_it->type); |
b7b65b15 EZ |
1781 | return type; |
1782 | } | |
1783 | ||
cbb09f04 EZ |
1784 | /* Resolve the type of a neutral character according to the type of |
1785 | surrounding strong text and the current embedding level. */ | |
58b9f433 | 1786 | static inline bidi_type_t |
cbb09f04 EZ |
1787 | bidi_resolve_neutral_1 (bidi_type_t prev_type, bidi_type_t next_type, int lev) |
1788 | { | |
1789 | /* N1: European and Arabic numbers are treated as though they were R. */ | |
1790 | if (next_type == WEAK_EN || next_type == WEAK_AN) | |
1791 | next_type = STRONG_R; | |
1792 | if (prev_type == WEAK_EN || prev_type == WEAK_AN) | |
1793 | prev_type = STRONG_R; | |
1794 | ||
1795 | if (next_type == prev_type) /* N1 */ | |
1796 | return next_type; | |
1797 | else if ((lev & 1) == 0) /* N2 */ | |
1798 | return STRONG_L; | |
1799 | else | |
1800 | return STRONG_R; | |
1801 | } | |
1802 | ||
fd3998ff | 1803 | static bidi_type_t |
b7b65b15 EZ |
1804 | bidi_resolve_neutral (struct bidi_it *bidi_it) |
1805 | { | |
1806 | int prev_level = bidi_it->level_stack[bidi_it->stack_idx].level; | |
1807 | bidi_type_t type = bidi_resolve_weak (bidi_it); | |
1808 | int current_level = bidi_it->level_stack[bidi_it->stack_idx].level; | |
1809 | ||
1810 | if (!(type == STRONG_R | |
1811 | || type == STRONG_L | |
1812 | || type == WEAK_BN | |
1813 | || type == WEAK_EN | |
1814 | || type == WEAK_AN | |
1815 | || type == NEUTRAL_B | |
1816 | || type == NEUTRAL_S | |
1817 | || type == NEUTRAL_WS | |
1818 | || type == NEUTRAL_ON)) | |
1819 | abort (); | |
1820 | ||
1821 | if (bidi_get_category (type) == NEUTRAL | |
1822 | || (type == WEAK_BN && prev_level == current_level)) | |
1823 | { | |
1824 | if (bidi_it->next_for_neutral.type != UNKNOWN_BT) | |
1825 | type = bidi_resolve_neutral_1 (bidi_it->prev_for_neutral.type, | |
1826 | bidi_it->next_for_neutral.type, | |
1827 | current_level); | |
1828 | else | |
1829 | { | |
1830 | /* Arrrgh!! The UAX#9 algorithm is too deeply entrenched in | |
1831 | the assumption of batch-style processing; see clauses W4, | |
1832 | W5, and especially N1, which require to look far forward | |
182ce2d2 EZ |
1833 | (as well as back) in the buffer/string. May the fleas of |
1834 | a thousand camels infest the armpits of those who design | |
b7b65b15 EZ |
1835 | supposedly general-purpose algorithms by looking at their |
1836 | own implementations, and fail to consider other possible | |
1837 | implementations! */ | |
1838 | struct bidi_it saved_it; | |
1839 | bidi_type_t next_type; | |
1840 | ||
1841 | if (bidi_it->scan_dir == -1) | |
1842 | abort (); | |
1843 | ||
1844 | bidi_copy_it (&saved_it, bidi_it); | |
1845 | /* Scan the text forward until we find the first non-neutral | |
1846 | character, and then use that to resolve the neutral we | |
1847 | are dealing with now. We also cache the scanned iterator | |
1848 | states, to salvage some of the effort later. */ | |
1849 | bidi_cache_iterator_state (bidi_it, 0); | |
1850 | do { | |
1851 | /* Record the info about the previous character, so that | |
1852 | it will be cached below with this state. */ | |
89d3374a | 1853 | if (bidi_it->type_after_w1 != WEAK_BN /* W1/Retaining */ |
b7b65b15 EZ |
1854 | && bidi_it->type != WEAK_BN) |
1855 | bidi_remember_char (&bidi_it->prev, bidi_it); | |
1856 | type = bidi_resolve_weak (bidi_it); | |
1857 | /* Paragraph separators have their levels fully resolved | |
1858 | at this point, so cache them as resolved. */ | |
1859 | bidi_cache_iterator_state (bidi_it, type == NEUTRAL_B); | |
1860 | /* FIXME: implement L1 here, by testing for a newline and | |
1861 | resetting the level for any sequence of whitespace | |
1862 | characters adjacent to it. */ | |
1863 | } while (!(type == NEUTRAL_B | |
1864 | || (type != WEAK_BN | |
1865 | && bidi_get_category (type) != NEUTRAL) | |
1866 | /* This is all per level run, so stop when we | |
1867 | reach the end of this level run. */ | |
1868 | || bidi_it->level_stack[bidi_it->stack_idx].level != | |
1869 | current_level)); | |
1870 | ||
1871 | bidi_remember_char (&saved_it.next_for_neutral, bidi_it); | |
1872 | ||
1873 | switch (type) | |
1874 | { | |
1875 | case STRONG_L: | |
1876 | case STRONG_R: | |
1877 | case STRONG_AL: | |
1878 | next_type = type; | |
1879 | break; | |
1880 | case WEAK_EN: | |
1881 | case WEAK_AN: | |
1882 | /* N1: ``European and Arabic numbers are treated as | |
1883 | though they were R.'' */ | |
1884 | next_type = STRONG_R; | |
1885 | saved_it.next_for_neutral.type = STRONG_R; | |
1886 | break; | |
1887 | case WEAK_BN: | |
1888 | if (!bidi_explicit_dir_char (bidi_it->ch)) | |
1889 | abort (); /* can't happen: BNs are skipped */ | |
1890 | /* FALLTHROUGH */ | |
1891 | case NEUTRAL_B: | |
1892 | /* Marched all the way to the end of this level run. | |
1893 | We need to use the eor type, whose information is | |
1894 | stored by bidi_set_sor_type in the prev_for_neutral | |
1895 | member. */ | |
1896 | if (saved_it.type != WEAK_BN | |
89d3374a | 1897 | || bidi_get_category (bidi_it->prev.type_after_w1) == NEUTRAL) |
b7b65b15 EZ |
1898 | { |
1899 | next_type = bidi_it->prev_for_neutral.type; | |
1900 | saved_it.next_for_neutral.type = next_type; | |
2d6e4628 | 1901 | bidi_check_type (next_type); |
b7b65b15 EZ |
1902 | } |
1903 | else | |
1904 | { | |
1905 | /* This is a BN which does not adjoin neutrals. | |
1906 | Leave its type alone. */ | |
1907 | bidi_copy_it (bidi_it, &saved_it); | |
1908 | return bidi_it->type; | |
1909 | } | |
1910 | break; | |
1911 | default: | |
1912 | abort (); | |
1913 | } | |
1914 | type = bidi_resolve_neutral_1 (saved_it.prev_for_neutral.type, | |
1915 | next_type, current_level); | |
1916 | saved_it.type = type; | |
2d6e4628 | 1917 | bidi_check_type (type); |
b7b65b15 EZ |
1918 | bidi_copy_it (bidi_it, &saved_it); |
1919 | } | |
1920 | } | |
1921 | return type; | |
1922 | } | |
1923 | ||
1924 | /* Given an iterator state in BIDI_IT, advance one character position | |
182ce2d2 EZ |
1925 | in the buffer/string to the next character (in the logical order), |
1926 | resolve the bidi type of that next character, and return that | |
1927 | type. */ | |
fd3998ff | 1928 | static bidi_type_t |
b7b65b15 EZ |
1929 | bidi_type_of_next_char (struct bidi_it *bidi_it) |
1930 | { | |
1931 | bidi_type_t type; | |
1932 | ||
1933 | /* This should always be called during a forward scan. */ | |
1934 | if (bidi_it->scan_dir != 1) | |
1935 | abort (); | |
1936 | ||
1937 | /* Reset the limit until which to ignore BNs if we step out of the | |
1938 | area where we found only empty levels. */ | |
87e67904 | 1939 | if ((bidi_it->ignore_bn_limit > -1 |
b7b65b15 | 1940 | && bidi_it->ignore_bn_limit <= bidi_it->charpos) |
87e67904 | 1941 | || (bidi_it->ignore_bn_limit == -2 |
b7b65b15 | 1942 | && !bidi_explicit_dir_char (bidi_it->ch))) |
87e67904 | 1943 | bidi_it->ignore_bn_limit = -1; |
b7b65b15 EZ |
1944 | |
1945 | type = bidi_resolve_neutral (bidi_it); | |
1946 | ||
1947 | return type; | |
1948 | } | |
1949 | ||
1950 | /* Given an iterator state BIDI_IT, advance one character position in | |
182ce2d2 EZ |
1951 | the buffer/string to the next character (in the current scan |
1952 | direction), resolve the embedding and implicit levels of that next | |
1953 | character, and return the resulting level. */ | |
fd3998ff | 1954 | static int |
b7b65b15 EZ |
1955 | bidi_level_of_next_char (struct bidi_it *bidi_it) |
1956 | { | |
1957 | bidi_type_t type; | |
1958 | int level, prev_level = -1; | |
1959 | struct bidi_saved_info next_for_neutral; | |
578b494e | 1960 | EMACS_INT next_char_pos = -2; |
b7b65b15 EZ |
1961 | |
1962 | if (bidi_it->scan_dir == 1) | |
1963 | { | |
9f257352 EZ |
1964 | EMACS_INT eob = |
1965 | (bidi_it->string.s || STRINGP (bidi_it->string.lstring)) | |
1966 | ? bidi_it->string.schars : ZV; | |
1967 | ||
b7b65b15 | 1968 | /* There's no sense in trying to advance if we hit end of text. */ |
9f257352 | 1969 | if (bidi_it->charpos >= eob) |
b7b65b15 EZ |
1970 | return bidi_it->resolved_level; |
1971 | ||
1972 | /* Record the info about the previous character. */ | |
89d3374a | 1973 | if (bidi_it->type_after_w1 != WEAK_BN /* W1/Retaining */ |
b7b65b15 EZ |
1974 | && bidi_it->type != WEAK_BN) |
1975 | bidi_remember_char (&bidi_it->prev, bidi_it); | |
89d3374a EZ |
1976 | if (bidi_it->type_after_w1 == STRONG_R |
1977 | || bidi_it->type_after_w1 == STRONG_L | |
1978 | || bidi_it->type_after_w1 == STRONG_AL) | |
b7b65b15 EZ |
1979 | bidi_remember_char (&bidi_it->last_strong, bidi_it); |
1980 | /* FIXME: it sounds like we don't need both prev and | |
1981 | prev_for_neutral members, but I'm leaving them both for now. */ | |
1982 | if (bidi_it->type == STRONG_R || bidi_it->type == STRONG_L | |
1983 | || bidi_it->type == WEAK_EN || bidi_it->type == WEAK_AN) | |
1984 | bidi_remember_char (&bidi_it->prev_for_neutral, bidi_it); | |
1985 | ||
1986 | /* If we overstepped the characters used for resolving neutrals | |
1987 | and whitespace, invalidate their info in the iterator. */ | |
1988 | if (bidi_it->charpos >= bidi_it->next_for_neutral.charpos) | |
1989 | bidi_it->next_for_neutral.type = UNKNOWN_BT; | |
1990 | if (bidi_it->next_en_pos >= 0 | |
1991 | && bidi_it->charpos >= bidi_it->next_en_pos) | |
1992 | bidi_it->next_en_pos = -1; | |
1993 | if (bidi_it->next_for_ws.type != UNKNOWN_BT | |
1994 | && bidi_it->charpos >= bidi_it->next_for_ws.charpos) | |
1995 | bidi_it->next_for_ws.type = UNKNOWN_BT; | |
1996 | ||
1997 | /* This must be taken before we fill the iterator with the info | |
1998 | about the next char. If we scan backwards, the iterator | |
1999 | state must be already cached, so there's no need to know the | |
2000 | embedding level of the previous character, since we will be | |
2001 | returning to our caller shortly. */ | |
2002 | prev_level = bidi_it->level_stack[bidi_it->stack_idx].level; | |
2003 | } | |
2004 | next_for_neutral = bidi_it->next_for_neutral; | |
2005 | ||
182ce2d2 EZ |
2006 | /* Perhaps the character we want is already cached. If it is, the |
2007 | call to bidi_cache_find below will return a type other than | |
2008 | UNKNOWN_BT. */ | |
87e67904 | 2009 | if (bidi_cache_idx > bidi_cache_start && !bidi_it->first_elt) |
102ebb00 | 2010 | { |
9f257352 EZ |
2011 | int bob = |
2012 | (bidi_it->string.s || STRINGP (bidi_it->string.lstring)) ? 0 : 1; | |
2013 | ||
102ebb00 EZ |
2014 | if (bidi_it->scan_dir > 0) |
2015 | { | |
2016 | if (bidi_it->nchars <= 0) | |
2017 | abort (); | |
2018 | next_char_pos = bidi_it->charpos + bidi_it->nchars; | |
2019 | } | |
9f257352 | 2020 | else if (bidi_it->charpos >= bob) |
bb269206 EZ |
2021 | /* Implementation note: we allow next_char_pos to be as low as |
2022 | 0 for buffers or -1 for strings, and that is okay because | |
2023 | that's the "position" of the sentinel iterator state we | |
2024 | cached at the beginning of the iteration. */ | |
102ebb00 | 2025 | next_char_pos = bidi_it->charpos - 1; |
578b494e | 2026 | if (next_char_pos >= bob - 1) |
87e67904 EZ |
2027 | type = bidi_cache_find (next_char_pos, -1, bidi_it); |
2028 | else | |
2029 | type = UNKNOWN_BT; | |
102ebb00 | 2030 | } |
182ce2d2 | 2031 | else |
102ebb00 | 2032 | type = UNKNOWN_BT; |
b7b65b15 EZ |
2033 | if (type != UNKNOWN_BT) |
2034 | { | |
2035 | /* Don't lose the information for resolving neutrals! The | |
2036 | cached states could have been cached before their | |
2037 | next_for_neutral member was computed. If we are on our way | |
2038 | forward, we can simply take the info from the previous | |
2039 | state. */ | |
2040 | if (bidi_it->scan_dir == 1 | |
2041 | && bidi_it->next_for_neutral.type == UNKNOWN_BT) | |
2042 | bidi_it->next_for_neutral = next_for_neutral; | |
2043 | ||
2044 | /* If resolved_level is -1, it means this state was cached | |
2045 | before it was completely resolved, so we cannot return | |
2046 | it. */ | |
2047 | if (bidi_it->resolved_level != -1) | |
2048 | return bidi_it->resolved_level; | |
2049 | } | |
2050 | if (bidi_it->scan_dir == -1) | |
2051 | /* If we are going backwards, the iterator state is already cached | |
2052 | from previous scans, and should be fully resolved. */ | |
2053 | abort (); | |
2054 | ||
2055 | if (type == UNKNOWN_BT) | |
2056 | type = bidi_type_of_next_char (bidi_it); | |
2057 | ||
2058 | if (type == NEUTRAL_B) | |
2059 | return bidi_it->resolved_level; | |
2060 | ||
2061 | level = bidi_it->level_stack[bidi_it->stack_idx].level; | |
2062 | if ((bidi_get_category (type) == NEUTRAL /* && type != NEUTRAL_B */) | |
2063 | || (type == WEAK_BN && prev_level == level)) | |
2064 | { | |
2065 | if (bidi_it->next_for_neutral.type == UNKNOWN_BT) | |
2066 | abort (); | |
2067 | ||
2068 | /* If the cached state shows a neutral character, it was not | |
2069 | resolved by bidi_resolve_neutral, so do it now. */ | |
2070 | type = bidi_resolve_neutral_1 (bidi_it->prev_for_neutral.type, | |
2071 | bidi_it->next_for_neutral.type, | |
2072 | level); | |
2073 | } | |
2074 | ||
2075 | if (!(type == STRONG_R | |
2076 | || type == STRONG_L | |
2077 | || type == WEAK_BN | |
2078 | || type == WEAK_EN | |
2079 | || type == WEAK_AN)) | |
2080 | abort (); | |
2081 | bidi_it->type = type; | |
2d6e4628 | 2082 | bidi_check_type (bidi_it->type); |
b7b65b15 EZ |
2083 | |
2084 | /* For L1 below, we need to know, for each WS character, whether | |
0d327994 | 2085 | it belongs to a sequence of WS characters preceding a newline |
b7b65b15 | 2086 | or a TAB or a paragraph separator. */ |
89d3374a | 2087 | if (bidi_it->orig_type == NEUTRAL_WS |
b7b65b15 EZ |
2088 | && bidi_it->next_for_ws.type == UNKNOWN_BT) |
2089 | { | |
2090 | int ch; | |
8264569d | 2091 | EMACS_INT clen = bidi_it->ch_len; |
6bff6497 EZ |
2092 | EMACS_INT bpos = bidi_it->bytepos; |
2093 | EMACS_INT cpos = bidi_it->charpos; | |
182ce2d2 | 2094 | EMACS_INT disp_pos = bidi_it->disp_pos; |
102ebb00 | 2095 | EMACS_INT nc = bidi_it->nchars; |
87e67904 | 2096 | struct bidi_string_data bs = bidi_it->string; |
b7b65b15 | 2097 | bidi_type_t chtype; |
fc6f18ce | 2098 | int fwp = bidi_it->frame_window_p; |
0c95fcf7 | 2099 | int dpp = bidi_it->disp_prop; |
b7b65b15 | 2100 | |
102ebb00 EZ |
2101 | if (bidi_it->nchars <= 0) |
2102 | abort (); | |
b7b65b15 | 2103 | do { |
55439c61 EZ |
2104 | ch = bidi_fetch_char (bpos += clen, cpos += nc, &disp_pos, &dpp, &bs, |
2105 | fwp, &clen, &nc); | |
e7402cb2 | 2106 | if (ch == '\n' || ch == BIDI_EOB /* || ch == LINESEP_CHAR */) |
b7b65b15 EZ |
2107 | chtype = NEUTRAL_B; |
2108 | else | |
6bff6497 | 2109 | chtype = bidi_get_type (ch, NEUTRAL_DIR); |
b7b65b15 EZ |
2110 | } while (chtype == NEUTRAL_WS || chtype == WEAK_BN |
2111 | || bidi_explicit_dir_char (ch)); /* L1/Retaining */ | |
2112 | bidi_it->next_for_ws.type = chtype; | |
2d6e4628 | 2113 | bidi_check_type (bidi_it->next_for_ws.type); |
b7b65b15 EZ |
2114 | bidi_it->next_for_ws.charpos = cpos; |
2115 | bidi_it->next_for_ws.bytepos = bpos; | |
2116 | } | |
2117 | ||
2118 | /* Resolve implicit levels, with a twist: PDFs get the embedding | |
2119 | level of the enbedding they terminate. See below for the | |
2120 | reason. */ | |
89d3374a | 2121 | if (bidi_it->orig_type == PDF |
b7b65b15 EZ |
2122 | /* Don't do this if this formatting code didn't change the |
2123 | embedding level due to invalid or empty embeddings. */ | |
2124 | && prev_level != level) | |
2125 | { | |
2126 | /* Don't look in UAX#9 for the reason for this: it's our own | |
2127 | private quirk. The reason is that we want the formatting | |
2128 | codes to be delivered so that they bracket the text of their | |
2129 | embedding. For example, given the text | |
2130 | ||
2131 | {RLO}teST{PDF} | |
2132 | ||
2133 | we want it to be displayed as | |
2134 | ||
0d68907d | 2135 | {PDF}STet{RLO} |
b7b65b15 EZ |
2136 | |
2137 | not as | |
2138 | ||
2139 | STet{RLO}{PDF} | |
2140 | ||
2141 | which will result because we bump up the embedding level as | |
2142 | soon as we see the RLO and pop it as soon as we see the PDF, | |
2143 | so RLO itself has the same embedding level as "teST", and | |
2144 | thus would be normally delivered last, just before the PDF. | |
2145 | The switch below fiddles with the level of PDF so that this | |
2146 | ugly side effect does not happen. | |
2147 | ||
2148 | (This is, of course, only important if the formatting codes | |
e7402cb2 EZ |
2149 | are actually displayed, but Emacs does need to display them |
2150 | if the user wants to.) */ | |
b7b65b15 EZ |
2151 | level = prev_level; |
2152 | } | |
89d3374a EZ |
2153 | else if (bidi_it->orig_type == NEUTRAL_B /* L1 */ |
2154 | || bidi_it->orig_type == NEUTRAL_S | |
e7402cb2 EZ |
2155 | || bidi_it->ch == '\n' || bidi_it->ch == BIDI_EOB |
2156 | /* || bidi_it->ch == LINESEP_CHAR */ | |
89d3374a | 2157 | || (bidi_it->orig_type == NEUTRAL_WS |
b7b65b15 EZ |
2158 | && (bidi_it->next_for_ws.type == NEUTRAL_B |
2159 | || bidi_it->next_for_ws.type == NEUTRAL_S))) | |
2160 | level = bidi_it->level_stack[0].level; | |
2161 | else if ((level & 1) == 0) /* I1 */ | |
2162 | { | |
2163 | if (type == STRONG_R) | |
2164 | level++; | |
2165 | else if (type == WEAK_EN || type == WEAK_AN) | |
2166 | level += 2; | |
2167 | } | |
2168 | else /* I2 */ | |
2169 | { | |
2170 | if (type == STRONG_L || type == WEAK_EN || type == WEAK_AN) | |
2171 | level++; | |
2172 | } | |
2173 | ||
2174 | bidi_it->resolved_level = level; | |
2175 | return level; | |
2176 | } | |
2177 | ||
2178 | /* Move to the other edge of a level given by LEVEL. If END_FLAG is | |
2179 | non-zero, we are at the end of a level, and we need to prepare to | |
2180 | resume the scan of the lower level. | |
2181 | ||
2182 | If this level's other edge is cached, we simply jump to it, filling | |
2183 | the iterator structure with the iterator state on the other edge. | |
182ce2d2 EZ |
2184 | Otherwise, we walk the buffer or string until we come back to the |
2185 | same level as LEVEL. | |
b7b65b15 EZ |
2186 | |
2187 | Note: we are not talking here about a ``level run'' in the UAX#9 | |
2188 | sense of the term, but rather about a ``level'' which includes | |
2189 | all the levels higher than it. In other words, given the levels | |
2190 | like this: | |
2191 | ||
2192 | 11111112222222333333334443343222222111111112223322111 | |
2193 | A B C | |
2194 | ||
2195 | and assuming we are at point A scanning left to right, this | |
2196 | function moves to point C, whereas the UAX#9 ``level 2 run'' ends | |
2197 | at point B. */ | |
2198 | static void | |
2199 | bidi_find_other_level_edge (struct bidi_it *bidi_it, int level, int end_flag) | |
2200 | { | |
2201 | int dir = end_flag ? -bidi_it->scan_dir : bidi_it->scan_dir; | |
39e378da | 2202 | ptrdiff_t idx; |
b7b65b15 EZ |
2203 | |
2204 | /* Try the cache first. */ | |
87e67904 EZ |
2205 | if ((idx = bidi_cache_find_level_change (level, dir, end_flag)) |
2206 | >= bidi_cache_start) | |
b7b65b15 EZ |
2207 | bidi_cache_fetch_state (idx, bidi_it); |
2208 | else | |
2209 | { | |
2210 | int new_level; | |
2211 | ||
2212 | if (end_flag) | |
2213 | abort (); /* if we are at end of level, its edges must be cached */ | |
2214 | ||
2215 | bidi_cache_iterator_state (bidi_it, 1); | |
2216 | do { | |
2217 | new_level = bidi_level_of_next_char (bidi_it); | |
2218 | bidi_cache_iterator_state (bidi_it, 1); | |
2219 | } while (new_level >= level); | |
2220 | } | |
2221 | } | |
2222 | ||
2223 | void | |
4b292a22 | 2224 | bidi_move_to_visually_next (struct bidi_it *bidi_it) |
b7b65b15 EZ |
2225 | { |
2226 | int old_level, new_level, next_level; | |
9c82e145 | 2227 | struct bidi_it sentinel; |
acb28818 | 2228 | struct gcpro gcpro1; |
b7b65b15 | 2229 | |
87e67904 EZ |
2230 | if (bidi_it->charpos < 0 || bidi_it->bytepos < 0) |
2231 | abort (); | |
b7b65b15 EZ |
2232 | |
2233 | if (bidi_it->scan_dir == 0) | |
2234 | { | |
2235 | bidi_it->scan_dir = 1; /* default to logical order */ | |
2236 | } | |
2237 | ||
acb28818 | 2238 | /* The code below can call eval, and thus cause GC. If we are |
7e2ad32c | 2239 | iterating a Lisp string, make sure it won't be GCed. */ |
acb28818 EZ |
2240 | if (STRINGP (bidi_it->string.lstring)) |
2241 | GCPRO1 (bidi_it->string.lstring); | |
2242 | ||
be39f003 EZ |
2243 | /* If we just passed a newline, initialize for the next line. */ |
2244 | if (!bidi_it->first_elt && bidi_it->orig_type == NEUTRAL_B) | |
2245 | bidi_line_init (bidi_it); | |
2246 | ||
6dcfd253 EZ |
2247 | /* Prepare the sentinel iterator state, and cache it. When we bump |
2248 | into it, scanning backwards, we'll know that the last non-base | |
2249 | level is exhausted. */ | |
87e67904 | 2250 | if (bidi_cache_idx == bidi_cache_start) |
9c82e145 EZ |
2251 | { |
2252 | bidi_copy_it (&sentinel, bidi_it); | |
2253 | if (bidi_it->first_elt) | |
2254 | { | |
2255 | sentinel.charpos--; /* cached charpos needs to be monotonic */ | |
2256 | sentinel.bytepos--; | |
2257 | sentinel.ch = '\n'; /* doesn't matter, but why not? */ | |
2258 | sentinel.ch_len = 1; | |
182ce2d2 | 2259 | sentinel.nchars = 1; |
9c82e145 | 2260 | } |
6dcfd253 | 2261 | bidi_cache_iterator_state (&sentinel, 1); |
9c82e145 | 2262 | } |
b7b65b15 EZ |
2263 | |
2264 | old_level = bidi_it->resolved_level; | |
2265 | new_level = bidi_level_of_next_char (bidi_it); | |
b7b65b15 EZ |
2266 | |
2267 | /* Reordering of resolved levels (clause L2) is implemented by | |
2268 | jumping to the other edge of the level and flipping direction of | |
c0546589 | 2269 | scanning the text whenever we find a level change. */ |
b7b65b15 EZ |
2270 | if (new_level != old_level) |
2271 | { | |
2272 | int ascending = new_level > old_level; | |
2273 | int level_to_search = ascending ? old_level + 1 : old_level; | |
2274 | int incr = ascending ? 1 : -1; | |
2275 | int expected_next_level = old_level + incr; | |
2276 | ||
b7b65b15 EZ |
2277 | /* Jump (or walk) to the other edge of this level. */ |
2278 | bidi_find_other_level_edge (bidi_it, level_to_search, !ascending); | |
2279 | /* Switch scan direction and peek at the next character in the | |
2280 | new direction. */ | |
2281 | bidi_it->scan_dir = -bidi_it->scan_dir; | |
2282 | ||
2283 | /* The following loop handles the case where the resolved level | |
2284 | jumps by more than one. This is typical for numbers inside a | |
2285 | run of text with left-to-right embedding direction, but can | |
2286 | also happen in other situations. In those cases the decision | |
2287 | where to continue after a level change, and in what direction, | |
2288 | is tricky. For example, given a text like below: | |
2289 | ||
2290 | abcdefgh | |
2291 | 11336622 | |
2292 | ||
2293 | (where the numbers below the text show the resolved levels), | |
2294 | the result of reordering according to UAX#9 should be this: | |
2295 | ||
2296 | efdcghba | |
2297 | ||
2298 | This is implemented by the loop below which flips direction | |
2299 | and jumps to the other edge of the level each time it finds | |
2300 | the new level not to be the expected one. The expected level | |
2301 | is always one more or one less than the previous one. */ | |
2302 | next_level = bidi_peek_at_next_level (bidi_it); | |
2303 | while (next_level != expected_next_level) | |
2304 | { | |
2305 | expected_next_level += incr; | |
2306 | level_to_search += incr; | |
2307 | bidi_find_other_level_edge (bidi_it, level_to_search, !ascending); | |
2308 | bidi_it->scan_dir = -bidi_it->scan_dir; | |
2309 | next_level = bidi_peek_at_next_level (bidi_it); | |
2310 | } | |
2311 | ||
2312 | /* Finally, deliver the next character in the new direction. */ | |
2313 | next_level = bidi_level_of_next_char (bidi_it); | |
2314 | } | |
2315 | ||
b44d9321 EZ |
2316 | /* Take note when we have just processed the newline that precedes |
2317 | the end of the paragraph. The next time we are about to be | |
2318 | called, set_iterator_to_next will automatically reinit the | |
2319 | paragraph direction, if needed. We do this at the newline before | |
2320 | the paragraph separator, because the next character might not be | |
2321 | the first character of the next paragraph, due to the bidi | |
c0546589 EZ |
2322 | reordering, whereas we _must_ know the paragraph base direction |
2323 | _before_ we process the paragraph's text, since the base | |
2324 | direction affects the reordering. */ | |
87e67904 | 2325 | if (bidi_it->scan_dir == 1 && bidi_it->orig_type == NEUTRAL_B) |
be39f003 | 2326 | { |
87e67904 EZ |
2327 | /* The paragraph direction of the entire string, once |
2328 | determined, is in effect for the entire string. Setting the | |
2329 | separator limit to the end of the string prevents | |
2330 | bidi_paragraph_init from being called automatically on this | |
2331 | string. */ | |
9f257352 | 2332 | if (bidi_it->string.s || STRINGP (bidi_it->string.lstring)) |
87e67904 EZ |
2333 | bidi_it->separator_limit = bidi_it->string.schars; |
2334 | else if (bidi_it->bytepos < ZV_BYTE) | |
be39f003 | 2335 | { |
87e67904 EZ |
2336 | EMACS_INT sep_len = |
2337 | bidi_at_paragraph_end (bidi_it->charpos + bidi_it->nchars, | |
2338 | bidi_it->bytepos + bidi_it->ch_len); | |
2339 | if (bidi_it->nchars <= 0) | |
2340 | abort (); | |
2341 | if (sep_len >= 0) | |
2342 | { | |
2343 | bidi_it->new_paragraph = 1; | |
2344 | /* Record the buffer position of the last character of the | |
2345 | paragraph separator. */ | |
2346 | bidi_it->separator_limit = | |
2347 | bidi_it->charpos + bidi_it->nchars + sep_len; | |
2348 | } | |
be39f003 EZ |
2349 | } |
2350 | } | |
6bff6497 | 2351 | |
87e67904 | 2352 | if (bidi_it->scan_dir == 1 && bidi_cache_idx > bidi_cache_start) |
b7b65b15 EZ |
2353 | { |
2354 | /* If we are at paragraph's base embedding level and beyond the | |
2355 | last cached position, the cache's job is done and we can | |
2356 | discard it. */ | |
2357 | if (bidi_it->resolved_level == bidi_it->level_stack[0].level | |
182ce2d2 EZ |
2358 | && bidi_it->charpos > (bidi_cache[bidi_cache_idx - 1].charpos |
2359 | + bidi_cache[bidi_cache_idx - 1].nchars - 1)) | |
b7b65b15 EZ |
2360 | bidi_cache_reset (); |
2361 | /* But as long as we are caching during forward scan, we must | |
2362 | cache each state, or else the cache integrity will be | |
2363 | compromised: it assumes cached states correspond to buffer | |
2364 | positions 1:1. */ | |
2365 | else | |
2366 | bidi_cache_iterator_state (bidi_it, 1); | |
2367 | } | |
acb28818 EZ |
2368 | |
2369 | if (STRINGP (bidi_it->string.lstring)) | |
2370 | UNGCPRO; | |
b7b65b15 EZ |
2371 | } |
2372 | ||
2373 | /* This is meant to be called from within the debugger, whenever you | |
2374 | wish to examine the cache contents. */ | |
e78aecca | 2375 | void bidi_dump_cached_states (void) EXTERNALLY_VISIBLE; |
b7b65b15 EZ |
2376 | void |
2377 | bidi_dump_cached_states (void) | |
2378 | { | |
39e378da | 2379 | ptrdiff_t i; |
b7b65b15 EZ |
2380 | int ndigits = 1; |
2381 | ||
2382 | if (bidi_cache_idx == 0) | |
2383 | { | |
2384 | fprintf (stderr, "The cache is empty.\n"); | |
2385 | return; | |
2386 | } | |
8265d3bb | 2387 | fprintf (stderr, "Total of %"pD"d state%s in cache:\n", |
b7b65b15 EZ |
2388 | bidi_cache_idx, bidi_cache_idx == 1 ? "" : "s"); |
2389 | ||
2390 | for (i = bidi_cache[bidi_cache_idx - 1].charpos; i > 0; i /= 10) | |
2391 | ndigits++; | |
2392 | fputs ("ch ", stderr); | |
2393 | for (i = 0; i < bidi_cache_idx; i++) | |
2394 | fprintf (stderr, "%*c", ndigits, bidi_cache[i].ch); | |
2395 | fputs ("\n", stderr); | |
2396 | fputs ("lvl ", stderr); | |
2397 | for (i = 0; i < bidi_cache_idx; i++) | |
2398 | fprintf (stderr, "%*d", ndigits, bidi_cache[i].resolved_level); | |
2399 | fputs ("\n", stderr); | |
2400 | fputs ("pos ", stderr); | |
2401 | for (i = 0; i < bidi_cache_idx; i++) | |
c2982e87 | 2402 | fprintf (stderr, "%*"pI"d", ndigits, bidi_cache[i].charpos); |
b7b65b15 EZ |
2403 | fputs ("\n", stderr); |
2404 | } |