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 | ||
bc18e09d PE |
319 | /* 5-slot stack for saving the start of the previous level of the |
320 | cache. xdisp.c maintains a 5-slot stack for its iterator state, | |
321 | and we need the same size of our stack. */ | |
322 | static ptrdiff_t bidi_cache_start_stack[IT_STACK_SIZE]; | |
323 | static int bidi_cache_sp; | |
324 | ||
325 | /* Size of header used by bidi_shelve_cache. */ | |
326 | enum | |
327 | { | |
328 | bidi_shelve_header_size = | |
329 | (sizeof (bidi_cache_idx) + sizeof (bidi_cache_start_stack) | |
330 | + sizeof (bidi_cache_sp) + sizeof (bidi_cache_start) | |
331 | + sizeof (bidi_cache_last_idx)) | |
332 | }; | |
333 | ||
87e67904 EZ |
334 | /* Reset the cache state to the empty state. We only reset the part |
335 | of the cache relevant to iteration of the current object. Previous | |
336 | objects, which are pushed on the display iterator's stack, are left | |
337 | intact. This is called when the cached information is no more | |
338 | useful for the current iteration, e.g. when we were reseated to a | |
339 | new position on the same object. */ | |
55d4c1b2 | 340 | static inline void |
b7b65b15 EZ |
341 | bidi_cache_reset (void) |
342 | { | |
87e67904 | 343 | bidi_cache_idx = bidi_cache_start; |
b7b65b15 EZ |
344 | bidi_cache_last_idx = -1; |
345 | } | |
346 | ||
87e67904 EZ |
347 | /* Shrink the cache to its minimal size. Called when we init the bidi |
348 | iterator for reordering a buffer or a string that does not come | |
349 | from display properties, because that means all the previously | |
350 | cached info is of no further use. */ | |
55d4c1b2 | 351 | static inline void |
2fe72643 EZ |
352 | bidi_cache_shrink (void) |
353 | { | |
354 | if (bidi_cache_size > BIDI_CACHE_CHUNK) | |
355 | { | |
0416466c | 356 | bidi_cache = |
51f30bc5 PE |
357 | (struct bidi_it *) xrealloc (bidi_cache, BIDI_CACHE_CHUNK * elsz); |
358 | bidi_cache_size = BIDI_CACHE_CHUNK; | |
2fe72643 EZ |
359 | } |
360 | bidi_cache_reset (); | |
361 | } | |
362 | ||
55d4c1b2 | 363 | static inline void |
39e378da | 364 | bidi_cache_fetch_state (ptrdiff_t idx, struct bidi_it *bidi_it) |
b7b65b15 EZ |
365 | { |
366 | int current_scan_dir = bidi_it->scan_dir; | |
367 | ||
87e67904 | 368 | if (idx < bidi_cache_start || idx >= bidi_cache_idx) |
b7b65b15 EZ |
369 | abort (); |
370 | ||
371 | bidi_copy_it (bidi_it, &bidi_cache[idx]); | |
372 | bidi_it->scan_dir = current_scan_dir; | |
373 | bidi_cache_last_idx = idx; | |
374 | } | |
375 | ||
376 | /* Find a cached state with a given CHARPOS and resolved embedding | |
377 | level less or equal to LEVEL. if LEVEL is -1, disregard the | |
378 | resolved levels in cached states. DIR, if non-zero, means search | |
379 | in that direction from the last cache hit. */ | |
39e378da | 380 | static inline ptrdiff_t |
61bfec98 | 381 | bidi_cache_search (EMACS_INT charpos, int level, int dir) |
b7b65b15 | 382 | { |
39e378da | 383 | ptrdiff_t i, i_start; |
b7b65b15 | 384 | |
6eec7596 | 385 | if (bidi_cache_idx > bidi_cache_start) |
b7b65b15 | 386 | { |
6eec7596 EZ |
387 | if (bidi_cache_last_idx == -1) |
388 | bidi_cache_last_idx = bidi_cache_idx - 1; | |
b7b65b15 | 389 | if (charpos < bidi_cache[bidi_cache_last_idx].charpos) |
182ce2d2 EZ |
390 | { |
391 | dir = -1; | |
392 | i_start = bidi_cache_last_idx - 1; | |
393 | } | |
394 | else if (charpos > (bidi_cache[bidi_cache_last_idx].charpos | |
395 | + bidi_cache[bidi_cache_last_idx].nchars - 1)) | |
396 | { | |
397 | dir = 1; | |
398 | i_start = bidi_cache_last_idx + 1; | |
399 | } | |
400 | else if (dir) | |
b7b65b15 EZ |
401 | i_start = bidi_cache_last_idx; |
402 | else | |
403 | { | |
404 | dir = -1; | |
405 | i_start = bidi_cache_idx - 1; | |
406 | } | |
407 | ||
408 | if (dir < 0) | |
409 | { | |
410 | /* Linear search for now; FIXME! */ | |
87e67904 | 411 | for (i = i_start; i >= bidi_cache_start; i--) |
182ce2d2 EZ |
412 | if (bidi_cache[i].charpos <= charpos |
413 | && charpos < bidi_cache[i].charpos + bidi_cache[i].nchars | |
b7b65b15 EZ |
414 | && (level == -1 || bidi_cache[i].resolved_level <= level)) |
415 | return i; | |
416 | } | |
417 | else | |
418 | { | |
419 | for (i = i_start; i < bidi_cache_idx; i++) | |
182ce2d2 EZ |
420 | if (bidi_cache[i].charpos <= charpos |
421 | && charpos < bidi_cache[i].charpos + bidi_cache[i].nchars | |
b7b65b15 EZ |
422 | && (level == -1 || bidi_cache[i].resolved_level <= level)) |
423 | return i; | |
424 | } | |
425 | } | |
426 | ||
427 | return -1; | |
428 | } | |
429 | ||
430 | /* Find a cached state where the resolved level changes to a value | |
431 | that is lower than LEVEL, and return its cache slot index. DIR is | |
432 | the direction to search, starting with the last used cache slot. | |
87e67904 EZ |
433 | If DIR is zero, we search backwards from the last occupied cache |
434 | slot. BEFORE, if non-zero, means return the index of the slot that | |
435 | is ``before'' the level change in the search direction. That is, | |
b7b65b15 EZ |
436 | given the cached levels like this: |
437 | ||
438 | 1122333442211 | |
439 | AB C | |
440 | ||
441 | and assuming we are at the position cached at the slot marked with | |
442 | C, searching backwards (DIR = -1) for LEVEL = 2 will return the | |
443 | index of slot B or A, depending whether BEFORE is, respectively, | |
444 | non-zero or zero. */ | |
39e378da | 445 | static ptrdiff_t |
b7b65b15 EZ |
446 | bidi_cache_find_level_change (int level, int dir, int before) |
447 | { | |
448 | if (bidi_cache_idx) | |
449 | { | |
39e378da | 450 | ptrdiff_t i = dir ? bidi_cache_last_idx : bidi_cache_idx - 1; |
b7b65b15 EZ |
451 | int incr = before ? 1 : 0; |
452 | ||
6eec7596 EZ |
453 | xassert (!dir || bidi_cache_last_idx >= 0); |
454 | ||
b7b65b15 EZ |
455 | if (!dir) |
456 | dir = -1; | |
457 | else if (!incr) | |
458 | i += dir; | |
459 | ||
460 | if (dir < 0) | |
461 | { | |
87e67904 | 462 | while (i >= bidi_cache_start + incr) |
b7b65b15 EZ |
463 | { |
464 | if (bidi_cache[i - incr].resolved_level >= 0 | |
465 | && bidi_cache[i - incr].resolved_level < level) | |
466 | return i; | |
467 | i--; | |
468 | } | |
469 | } | |
470 | else | |
471 | { | |
472 | while (i < bidi_cache_idx - incr) | |
473 | { | |
474 | if (bidi_cache[i + incr].resolved_level >= 0 | |
475 | && bidi_cache[i + incr].resolved_level < level) | |
476 | return i; | |
477 | i++; | |
478 | } | |
479 | } | |
480 | } | |
481 | ||
482 | return -1; | |
483 | } | |
484 | ||
58b9f433 | 485 | static inline void |
39e378da | 486 | bidi_cache_ensure_space (ptrdiff_t idx) |
58b9f433 EZ |
487 | { |
488 | /* Enlarge the cache as needed. */ | |
489 | if (idx >= bidi_cache_size) | |
490 | { | |
f0eb61e9 PE |
491 | /* The bidi cache cannot be larger than the largest Lisp string |
492 | or buffer. */ | |
493 | ptrdiff_t string_or_buffer_bound = | |
494 | max (BUF_BYTES_MAX, STRING_BYTES_BOUND); | |
495 | ||
496 | /* Also, it cannot be larger than what C can represent. */ | |
bc18e09d PE |
497 | ptrdiff_t c_bound = |
498 | (min (PTRDIFF_MAX, SIZE_MAX) - bidi_shelve_header_size) / elsz; | |
f0eb61e9 | 499 | |
0065d054 PE |
500 | bidi_cache = |
501 | xpalloc (bidi_cache, &bidi_cache_size, | |
502 | max (BIDI_CACHE_CHUNK, idx - bidi_cache_size + 1), | |
503 | min (string_or_buffer_bound, c_bound), elsz); | |
58b9f433 EZ |
504 | } |
505 | } | |
506 | ||
55d4c1b2 | 507 | static inline void |
b7b65b15 EZ |
508 | bidi_cache_iterator_state (struct bidi_it *bidi_it, int resolved) |
509 | { | |
39e378da | 510 | ptrdiff_t idx; |
b7b65b15 EZ |
511 | |
512 | /* We should never cache on backward scans. */ | |
513 | if (bidi_it->scan_dir == -1) | |
514 | abort (); | |
515 | idx = bidi_cache_search (bidi_it->charpos, -1, 1); | |
516 | ||
517 | if (idx < 0) | |
518 | { | |
519 | idx = bidi_cache_idx; | |
58b9f433 | 520 | bidi_cache_ensure_space (idx); |
bd924a5d EZ |
521 | /* Character positions should correspond to cache positions 1:1. |
522 | If we are outside the range of cached positions, the cache is | |
523 | useless and must be reset. */ | |
87e67904 | 524 | if (idx > bidi_cache_start && |
182ce2d2 EZ |
525 | (bidi_it->charpos > (bidi_cache[idx - 1].charpos |
526 | + bidi_cache[idx - 1].nchars) | |
a1344e7d | 527 | || bidi_it->charpos < bidi_cache[bidi_cache_start].charpos)) |
bd924a5d EZ |
528 | { |
529 | bidi_cache_reset (); | |
87e67904 | 530 | idx = bidi_cache_start; |
bd924a5d | 531 | } |
102ebb00 EZ |
532 | if (bidi_it->nchars <= 0) |
533 | abort (); | |
b7b65b15 EZ |
534 | bidi_copy_it (&bidi_cache[idx], bidi_it); |
535 | if (!resolved) | |
536 | bidi_cache[idx].resolved_level = -1; | |
537 | } | |
538 | else | |
539 | { | |
540 | /* Copy only the members which could have changed, to avoid | |
541 | costly copying of the entire struct. */ | |
542 | bidi_cache[idx].type = bidi_it->type; | |
2d6e4628 | 543 | bidi_check_type (bidi_it->type); |
89d3374a EZ |
544 | bidi_cache[idx].type_after_w1 = bidi_it->type_after_w1; |
545 | bidi_check_type (bidi_it->type_after_w1); | |
b7b65b15 EZ |
546 | if (resolved) |
547 | bidi_cache[idx].resolved_level = bidi_it->resolved_level; | |
548 | else | |
549 | bidi_cache[idx].resolved_level = -1; | |
550 | bidi_cache[idx].invalid_levels = bidi_it->invalid_levels; | |
551 | bidi_cache[idx].invalid_rl_levels = bidi_it->invalid_rl_levels; | |
552 | bidi_cache[idx].next_for_neutral = bidi_it->next_for_neutral; | |
553 | bidi_cache[idx].next_for_ws = bidi_it->next_for_ws; | |
554 | bidi_cache[idx].ignore_bn_limit = bidi_it->ignore_bn_limit; | |
f67cdd7f | 555 | bidi_cache[idx].disp_pos = bidi_it->disp_pos; |
0c95fcf7 | 556 | bidi_cache[idx].disp_prop = bidi_it->disp_prop; |
b7b65b15 EZ |
557 | } |
558 | ||
559 | bidi_cache_last_idx = idx; | |
560 | if (idx >= bidi_cache_idx) | |
561 | bidi_cache_idx = idx + 1; | |
562 | } | |
563 | ||
55d4c1b2 | 564 | static inline bidi_type_t |
61bfec98 | 565 | bidi_cache_find (EMACS_INT charpos, int level, struct bidi_it *bidi_it) |
b7b65b15 | 566 | { |
39e378da | 567 | ptrdiff_t i = bidi_cache_search (charpos, level, bidi_it->scan_dir); |
b7b65b15 | 568 | |
87e67904 | 569 | if (i >= bidi_cache_start) |
b7b65b15 EZ |
570 | { |
571 | bidi_dir_t current_scan_dir = bidi_it->scan_dir; | |
572 | ||
5009f85e | 573 | bidi_copy_it (bidi_it, &bidi_cache[i]); |
b7b65b15 EZ |
574 | bidi_cache_last_idx = i; |
575 | /* Don't let scan direction from from the cached state override | |
576 | the current scan direction. */ | |
577 | bidi_it->scan_dir = current_scan_dir; | |
578 | return bidi_it->type; | |
579 | } | |
580 | ||
581 | return UNKNOWN_BT; | |
582 | } | |
583 | ||
55d4c1b2 | 584 | static inline int |
b7b65b15 EZ |
585 | bidi_peek_at_next_level (struct bidi_it *bidi_it) |
586 | { | |
87e67904 | 587 | if (bidi_cache_idx == bidi_cache_start || bidi_cache_last_idx == -1) |
b7b65b15 EZ |
588 | abort (); |
589 | return bidi_cache[bidi_cache_last_idx + bidi_it->scan_dir].resolved_level; | |
590 | } | |
591 | ||
cbb09f04 | 592 | \f |
58b9f433 EZ |
593 | /*********************************************************************** |
594 | Pushing and popping the bidi iterator state | |
595 | ***********************************************************************/ | |
58b9f433 EZ |
596 | |
597 | /* Push the bidi iterator state in preparation for reordering a | |
598 | different object, e.g. display string found at certain buffer | |
7e2ad32c EZ |
599 | position. Pushing the bidi iterator boils down to saving its |
600 | entire state on the cache and starting a new cache "stacked" on top | |
601 | of the current cache. */ | |
58b9f433 EZ |
602 | void |
603 | bidi_push_it (struct bidi_it *bidi_it) | |
b7b65b15 | 604 | { |
58b9f433 EZ |
605 | /* Save the current iterator state in its entirety after the last |
606 | used cache slot. */ | |
607 | bidi_cache_ensure_space (bidi_cache_idx); | |
608 | memcpy (&bidi_cache[bidi_cache_idx++], bidi_it, sizeof (struct bidi_it)); | |
be39f003 | 609 | |
58b9f433 | 610 | /* Push the current cache start onto the stack. */ |
7e2ad32c | 611 | xassert (bidi_cache_sp < IT_STACK_SIZE); |
58b9f433 | 612 | bidi_cache_start_stack[bidi_cache_sp++] = bidi_cache_start; |
be39f003 | 613 | |
58b9f433 EZ |
614 | /* Start a new level of cache, and make it empty. */ |
615 | bidi_cache_start = bidi_cache_idx; | |
616 | bidi_cache_last_idx = -1; | |
617 | } | |
618 | ||
619 | /* Restore the iterator state saved by bidi_push_it and return the | |
620 | cache to the corresponding state. */ | |
621 | void | |
622 | bidi_pop_it (struct bidi_it *bidi_it) | |
623 | { | |
624 | if (bidi_cache_start <= 0) | |
625 | abort (); | |
626 | ||
627 | /* Reset the next free cache slot index to what it was before the | |
628 | call to bidi_push_it. */ | |
629 | bidi_cache_idx = bidi_cache_start - 1; | |
630 | ||
631 | /* Restore the bidi iterator state saved in the cache. */ | |
632 | memcpy (bidi_it, &bidi_cache[bidi_cache_idx], sizeof (struct bidi_it)); | |
633 | ||
634 | /* Pop the previous cache start from the stack. */ | |
635 | if (bidi_cache_sp <= 0) | |
636 | abort (); | |
637 | bidi_cache_start = bidi_cache_start_stack[--bidi_cache_sp]; | |
638 | ||
639 | /* Invalidate the last-used cache slot data. */ | |
640 | bidi_cache_last_idx = -1; | |
641 | } | |
642 | ||
ec7cc85b | 643 | static ptrdiff_t bidi_cache_total_alloc; |
35928349 | 644 | |
ed94e6d7 EZ |
645 | /* Stash away a copy of the cache and its control variables. */ |
646 | void * | |
647 | bidi_shelve_cache (void) | |
648 | { | |
649 | unsigned char *databuf; | |
458bfed3 | 650 | ptrdiff_t alloc; |
ed94e6d7 | 651 | |
35928349 | 652 | /* Empty cache. */ |
ed94e6d7 EZ |
653 | if (bidi_cache_idx == 0) |
654 | return NULL; | |
655 | ||
458bfed3 PE |
656 | alloc = (bidi_shelve_header_size |
657 | + bidi_cache_idx * sizeof (struct bidi_it)); | |
658 | databuf = xmalloc (alloc); | |
659 | bidi_cache_total_alloc += alloc; | |
35928349 | 660 | |
ed94e6d7 EZ |
661 | memcpy (databuf, &bidi_cache_idx, sizeof (bidi_cache_idx)); |
662 | memcpy (databuf + sizeof (bidi_cache_idx), | |
663 | bidi_cache, bidi_cache_idx * sizeof (struct bidi_it)); | |
664 | memcpy (databuf + sizeof (bidi_cache_idx) | |
665 | + bidi_cache_idx * sizeof (struct bidi_it), | |
666 | bidi_cache_start_stack, sizeof (bidi_cache_start_stack)); | |
667 | memcpy (databuf + sizeof (bidi_cache_idx) | |
668 | + bidi_cache_idx * sizeof (struct bidi_it) | |
669 | + sizeof (bidi_cache_start_stack), | |
670 | &bidi_cache_sp, sizeof (bidi_cache_sp)); | |
671 | memcpy (databuf + sizeof (bidi_cache_idx) | |
672 | + bidi_cache_idx * sizeof (struct bidi_it) | |
673 | + sizeof (bidi_cache_start_stack) + sizeof (bidi_cache_sp), | |
674 | &bidi_cache_start, sizeof (bidi_cache_start)); | |
675 | memcpy (databuf + sizeof (bidi_cache_idx) | |
676 | + bidi_cache_idx * sizeof (struct bidi_it) | |
677 | + sizeof (bidi_cache_start_stack) + sizeof (bidi_cache_sp) | |
678 | + sizeof (bidi_cache_start), | |
679 | &bidi_cache_last_idx, sizeof (bidi_cache_last_idx)); | |
680 | ||
681 | return databuf; | |
682 | } | |
683 | ||
d1410150 EZ |
684 | /* Restore the cache state from a copy stashed away by |
685 | bidi_shelve_cache, and free the buffer used to stash that copy. | |
686 | JUST_FREE non-zero means free the buffer, but don't restore the | |
687 | cache; used when the corresponding iterator is discarded instead of | |
688 | being restored. */ | |
ed94e6d7 | 689 | void |
35928349 | 690 | bidi_unshelve_cache (void *databuf, int just_free) |
ed94e6d7 EZ |
691 | { |
692 | unsigned char *p = databuf; | |
693 | ||
694 | if (!p) | |
be39f003 | 695 | { |
d1410150 EZ |
696 | if (!just_free) |
697 | { | |
698 | /* A NULL pointer means an empty cache. */ | |
699 | bidi_cache_start = 0; | |
700 | bidi_cache_sp = 0; | |
701 | bidi_cache_reset (); | |
702 | } | |
be39f003 | 703 | } |
ed94e6d7 EZ |
704 | else |
705 | { | |
35928349 EZ |
706 | if (just_free) |
707 | { | |
708 | ptrdiff_t idx; | |
709 | ||
710 | memcpy (&idx, p, sizeof (bidi_cache_idx)); | |
711 | bidi_cache_total_alloc -= | |
458bfed3 | 712 | bidi_shelve_header_size + idx * sizeof (struct bidi_it); |
35928349 EZ |
713 | } |
714 | else | |
715 | { | |
716 | memcpy (&bidi_cache_idx, p, sizeof (bidi_cache_idx)); | |
717 | bidi_cache_ensure_space (bidi_cache_idx); | |
718 | memcpy (bidi_cache, p + sizeof (bidi_cache_idx), | |
719 | bidi_cache_idx * sizeof (struct bidi_it)); | |
720 | memcpy (bidi_cache_start_stack, | |
721 | p + sizeof (bidi_cache_idx) | |
722 | + bidi_cache_idx * sizeof (struct bidi_it), | |
723 | sizeof (bidi_cache_start_stack)); | |
724 | memcpy (&bidi_cache_sp, | |
725 | p + sizeof (bidi_cache_idx) | |
726 | + bidi_cache_idx * sizeof (struct bidi_it) | |
727 | + sizeof (bidi_cache_start_stack), | |
728 | sizeof (bidi_cache_sp)); | |
729 | memcpy (&bidi_cache_start, | |
730 | p + sizeof (bidi_cache_idx) | |
731 | + bidi_cache_idx * sizeof (struct bidi_it) | |
732 | + sizeof (bidi_cache_start_stack) + sizeof (bidi_cache_sp), | |
733 | sizeof (bidi_cache_start)); | |
734 | memcpy (&bidi_cache_last_idx, | |
735 | p + sizeof (bidi_cache_idx) | |
736 | + bidi_cache_idx * sizeof (struct bidi_it) | |
737 | + sizeof (bidi_cache_start_stack) + sizeof (bidi_cache_sp) | |
738 | + sizeof (bidi_cache_start), | |
739 | sizeof (bidi_cache_last_idx)); | |
740 | bidi_cache_total_alloc -= | |
458bfed3 | 741 | bidi_shelve_header_size + bidi_cache_idx * sizeof (struct bidi_it); |
35928349 | 742 | } |
ed94e6d7 EZ |
743 | |
744 | xfree (p); | |
745 | } | |
746 | } | |
b7b65b15 | 747 | |
58b9f433 | 748 | \f |
cbb09f04 EZ |
749 | /*********************************************************************** |
750 | Initialization | |
751 | ***********************************************************************/ | |
752 | static void | |
753 | bidi_initialize (void) | |
b7b65b15 | 754 | { |
474a8465 EZ |
755 | bidi_type_table = uniprop_table (intern ("bidi-class")); |
756 | if (NILP (bidi_type_table)) | |
757 | abort (); | |
cbb09f04 EZ |
758 | staticpro (&bidi_type_table); |
759 | ||
474a8465 EZ |
760 | bidi_mirror_table = uniprop_table (intern ("mirroring")); |
761 | if (NILP (bidi_mirror_table)) | |
762 | abort (); | |
cbb09f04 EZ |
763 | staticpro (&bidi_mirror_table); |
764 | ||
cbb09f04 EZ |
765 | Qparagraph_start = intern ("paragraph-start"); |
766 | staticpro (&Qparagraph_start); | |
767 | paragraph_start_re = Fsymbol_value (Qparagraph_start); | |
768 | if (!STRINGP (paragraph_start_re)) | |
769 | paragraph_start_re = build_string ("\f\\|[ \t]*$"); | |
770 | staticpro (¶graph_start_re); | |
771 | Qparagraph_separate = intern ("paragraph-separate"); | |
772 | staticpro (&Qparagraph_separate); | |
773 | paragraph_separate_re = Fsymbol_value (Qparagraph_separate); | |
774 | if (!STRINGP (paragraph_separate_re)) | |
775 | paragraph_separate_re = build_string ("[ \t\f]*$"); | |
776 | staticpro (¶graph_separate_re); | |
58b9f433 EZ |
777 | |
778 | bidi_cache_sp = 0; | |
ec7cc85b | 779 | bidi_cache_total_alloc = 0; |
58b9f433 | 780 | |
cbb09f04 | 781 | bidi_initialized = 1; |
b7b65b15 EZ |
782 | } |
783 | ||
cbb09f04 EZ |
784 | /* Do whatever UAX#9 clause X8 says should be done at paragraph's |
785 | end. */ | |
55d4c1b2 | 786 | static inline void |
cbb09f04 | 787 | bidi_set_paragraph_end (struct bidi_it *bidi_it) |
b7b65b15 | 788 | { |
cbb09f04 EZ |
789 | bidi_it->invalid_levels = 0; |
790 | bidi_it->invalid_rl_levels = -1; | |
791 | bidi_it->stack_idx = 0; | |
792 | bidi_it->resolved_level = bidi_it->level_stack[0].level; | |
793 | } | |
b7b65b15 | 794 | |
cbb09f04 EZ |
795 | /* Initialize the bidi iterator from buffer/string position CHARPOS. */ |
796 | void | |
797 | bidi_init_it (EMACS_INT charpos, EMACS_INT bytepos, int frame_window_p, | |
798 | struct bidi_it *bidi_it) | |
799 | { | |
800 | if (! bidi_initialized) | |
801 | bidi_initialize (); | |
802 | if (charpos >= 0) | |
803 | bidi_it->charpos = charpos; | |
804 | if (bytepos >= 0) | |
805 | bidi_it->bytepos = bytepos; | |
806 | bidi_it->frame_window_p = frame_window_p; | |
807 | bidi_it->nchars = -1; /* to be computed in bidi_resolve_explicit_1 */ | |
808 | bidi_it->first_elt = 1; | |
809 | bidi_set_paragraph_end (bidi_it); | |
810 | bidi_it->new_paragraph = 1; | |
811 | bidi_it->separator_limit = -1; | |
812 | bidi_it->type = NEUTRAL_B; | |
813 | bidi_it->type_after_w1 = NEUTRAL_B; | |
814 | bidi_it->orig_type = NEUTRAL_B; | |
815 | bidi_it->prev_was_pdf = 0; | |
816 | bidi_it->prev.type = bidi_it->prev.type_after_w1 = | |
817 | bidi_it->prev.orig_type = UNKNOWN_BT; | |
89d3374a EZ |
818 | bidi_it->last_strong.type = bidi_it->last_strong.type_after_w1 = |
819 | bidi_it->last_strong.orig_type = UNKNOWN_BT; | |
cbb09f04 EZ |
820 | bidi_it->next_for_neutral.charpos = -1; |
821 | bidi_it->next_for_neutral.type = | |
822 | bidi_it->next_for_neutral.type_after_w1 = | |
89d3374a | 823 | bidi_it->next_for_neutral.orig_type = UNKNOWN_BT; |
cbb09f04 EZ |
824 | bidi_it->prev_for_neutral.charpos = -1; |
825 | bidi_it->prev_for_neutral.type = | |
826 | bidi_it->prev_for_neutral.type_after_w1 = | |
827 | bidi_it->prev_for_neutral.orig_type = UNKNOWN_BT; | |
828 | bidi_it->sor = L2R; /* FIXME: should it be user-selectable? */ | |
829 | bidi_it->disp_pos = -1; /* invalid/unknown */ | |
0c95fcf7 | 830 | bidi_it->disp_prop = 0; |
cbb09f04 EZ |
831 | /* We can only shrink the cache if we are at the bottom level of its |
832 | "stack". */ | |
833 | if (bidi_cache_start == 0) | |
834 | bidi_cache_shrink (); | |
58b9f433 EZ |
835 | else |
836 | bidi_cache_reset (); | |
b7b65b15 EZ |
837 | } |
838 | ||
182ce2d2 | 839 | /* Perform initializations for reordering a new line of bidi text. */ |
6bff6497 | 840 | static void |
be39f003 EZ |
841 | bidi_line_init (struct bidi_it *bidi_it) |
842 | { | |
843 | bidi_it->scan_dir = 1; /* FIXME: do we need to have control on this? */ | |
844 | bidi_it->resolved_level = bidi_it->level_stack[0].level; | |
845 | bidi_it->level_stack[0].override = NEUTRAL_DIR; /* X1 */ | |
846 | bidi_it->invalid_levels = 0; | |
847 | bidi_it->invalid_rl_levels = -1; | |
848 | bidi_it->next_en_pos = -1; | |
849 | bidi_it->next_for_ws.type = UNKNOWN_BT; | |
b44d9321 EZ |
850 | bidi_set_sor_type (bidi_it, |
851 | bidi_it->paragraph_dir == R2L ? 1 : 0, | |
be39f003 EZ |
852 | bidi_it->level_stack[0].level); /* X10 */ |
853 | ||
854 | bidi_cache_reset (); | |
855 | } | |
856 | ||
cbb09f04 EZ |
857 | \f |
858 | /*********************************************************************** | |
859 | Fetching characters | |
860 | ***********************************************************************/ | |
861 | ||
f3014ef5 EZ |
862 | /* Count bytes in string S between BEG/BEGBYTE and END. BEG and END |
863 | are zero-based character positions in S, BEGBYTE is byte position | |
864 | corresponding to BEG. UNIBYTE, if non-zero, means S is a unibyte | |
865 | string. */ | |
87e67904 EZ |
866 | static inline EMACS_INT |
867 | bidi_count_bytes (const unsigned char *s, const EMACS_INT beg, | |
f3014ef5 | 868 | const EMACS_INT begbyte, const EMACS_INT end, int unibyte) |
87e67904 EZ |
869 | { |
870 | EMACS_INT pos = beg; | |
871 | const unsigned char *p = s + begbyte, *start = p; | |
872 | ||
f3014ef5 EZ |
873 | if (unibyte) |
874 | p = s + end; | |
875 | else | |
87e67904 | 876 | { |
f3014ef5 EZ |
877 | if (!CHAR_HEAD_P (*p)) |
878 | abort (); | |
879 | ||
880 | while (pos < end) | |
881 | { | |
882 | p += BYTES_BY_CHAR_HEAD (*p); | |
883 | pos++; | |
884 | } | |
87e67904 EZ |
885 | } |
886 | ||
887 | return p - start; | |
888 | } | |
889 | ||
890 | /* Fetch and returns the character at byte position BYTEPOS. If S is | |
891 | non-NULL, fetch the character from string S; otherwise fetch the | |
f3014ef5 EZ |
892 | character from the current buffer. UNIBYTE non-zero means S is a |
893 | unibyte string. */ | |
87e67904 | 894 | static inline int |
f3014ef5 | 895 | bidi_char_at_pos (EMACS_INT bytepos, const unsigned char *s, int unibyte) |
87e67904 EZ |
896 | { |
897 | if (s) | |
f3014ef5 EZ |
898 | { |
899 | if (unibyte) | |
900 | return s[bytepos]; | |
901 | else | |
902 | return STRING_CHAR (s + bytepos); | |
903 | } | |
87e67904 EZ |
904 | else |
905 | return FETCH_MULTIBYTE_CHAR (bytepos); | |
906 | } | |
907 | ||
102ebb00 EZ |
908 | /* Fetch and return the character at BYTEPOS/CHARPOS. If that |
909 | character is covered by a display string, treat the entire run of | |
0c95fcf7 EZ |
910 | covered characters as a single character, either u+2029 or u+FFFC, |
911 | and return their combined length in CH_LEN and NCHARS. DISP_POS | |
912 | specifies the character position of the next display string, or -1 | |
913 | if not yet computed. DISP_PROP non-zero means that there's really | |
914 | a display string at DISP_POS, as opposed to when we searched till | |
915 | DISP_POS without finding one. If DISP_PROP is 2, it means the | |
916 | display spec is of the form `(space ...)', which is replaced with | |
917 | u+2029 to handle it as a paragraph separator. When the next | |
918 | character is at or beyond that position, the function updates | |
919 | DISP_POS with the position of the next display string. STRING->s | |
920 | is the C string to iterate, or NULL if iterating over a buffer or a | |
921 | Lisp string; in the latter case, STRING->lstring is the Lisp | |
922 | string. */ | |
fec2107c | 923 | static inline int |
102ebb00 | 924 | bidi_fetch_char (EMACS_INT bytepos, EMACS_INT charpos, EMACS_INT *disp_pos, |
0c95fcf7 | 925 | int *disp_prop, struct bidi_string_data *string, |
fc6f18ce | 926 | int frame_window_p, EMACS_INT *ch_len, EMACS_INT *nchars) |
182ce2d2 EZ |
927 | { |
928 | int ch; | |
9f257352 EZ |
929 | EMACS_INT endpos = |
930 | (string->s || STRINGP (string->lstring)) ? string->schars : ZV; | |
6db161be | 931 | struct text_pos pos; |
182ce2d2 | 932 | |
182ce2d2 | 933 | /* If we got past the last known position of display string, compute |
87e67904 EZ |
934 | the position of the next one. That position could be at CHARPOS. */ |
935 | if (charpos < endpos && charpos > *disp_pos) | |
6db161be EZ |
936 | { |
937 | SET_TEXT_POS (pos, charpos, bytepos); | |
55439c61 | 938 | *disp_pos = compute_display_string_pos (&pos, string, frame_window_p, |
0c95fcf7 | 939 | disp_prop); |
6db161be | 940 | } |
102ebb00 EZ |
941 | |
942 | /* Fetch the character at BYTEPOS. */ | |
87e67904 | 943 | if (charpos >= endpos) |
182ce2d2 EZ |
944 | { |
945 | ch = BIDI_EOB; | |
946 | *ch_len = 1; | |
947 | *nchars = 1; | |
87e67904 | 948 | *disp_pos = endpos; |
0c95fcf7 | 949 | *disp_prop = 0; |
182ce2d2 | 950 | } |
0c95fcf7 | 951 | else if (charpos >= *disp_pos && *disp_prop) |
182ce2d2 | 952 | { |
7b600102 EZ |
953 | EMACS_INT disp_end_pos; |
954 | ||
955 | /* We don't expect to find ourselves in the middle of a display | |
956 | property. Hopefully, it will never be needed. */ | |
957 | if (charpos > *disp_pos) | |
958 | abort (); | |
0c95fcf7 EZ |
959 | /* Text covered by `display' properties and overlays with |
960 | display properties or display strings is handled as a single | |
961 | character that represents the entire run of characters | |
962 | covered by the display property. */ | |
963 | if (*disp_prop == 2) | |
964 | { | |
965 | /* `(space ...)' display specs are handled as paragraph | |
966 | separators for the purposes of the reordering; see UAX#9 | |
967 | section 3 and clause HL1 in section 4.3 there. */ | |
968 | ch = 0x2029; | |
969 | } | |
970 | else | |
971 | { | |
972 | /* All other display specs are handled as the Unicode Object | |
973 | Replacement Character. */ | |
974 | ch = 0xFFFC; | |
975 | } | |
87e67904 | 976 | disp_end_pos = compute_display_string_end (*disp_pos, string); |
7b600102 | 977 | *nchars = disp_end_pos - *disp_pos; |
f3014ef5 EZ |
978 | if (*nchars <= 0) |
979 | abort (); | |
87e67904 EZ |
980 | if (string->s) |
981 | *ch_len = bidi_count_bytes (string->s, *disp_pos, bytepos, | |
f3014ef5 | 982 | disp_end_pos, string->unibyte); |
9f257352 EZ |
983 | else if (STRINGP (string->lstring)) |
984 | *ch_len = bidi_count_bytes (SDATA (string->lstring), *disp_pos, | |
f3014ef5 | 985 | bytepos, disp_end_pos, string->unibyte); |
87e67904 EZ |
986 | else |
987 | *ch_len = CHAR_TO_BYTE (disp_end_pos) - bytepos; | |
182ce2d2 | 988 | } |
182ce2d2 EZ |
989 | else |
990 | { | |
87e67904 EZ |
991 | if (string->s) |
992 | { | |
66ff7ddf | 993 | int len; |
87e67904 | 994 | |
f3014ef5 EZ |
995 | if (!string->unibyte) |
996 | { | |
997 | ch = STRING_CHAR_AND_LENGTH (string->s + bytepos, len); | |
998 | *ch_len = len; | |
999 | } | |
1000 | else | |
1001 | { | |
1002 | ch = UNIBYTE_TO_CHAR (string->s[bytepos]); | |
1003 | *ch_len = 1; | |
1004 | } | |
87e67904 | 1005 | } |
9f257352 EZ |
1006 | else if (STRINGP (string->lstring)) |
1007 | { | |
66ff7ddf | 1008 | int len; |
9f257352 | 1009 | |
f3014ef5 EZ |
1010 | if (!string->unibyte) |
1011 | { | |
1012 | ch = STRING_CHAR_AND_LENGTH (SDATA (string->lstring) + bytepos, | |
1013 | len); | |
1014 | *ch_len = len; | |
1015 | } | |
1016 | else | |
1017 | { | |
1018 | ch = UNIBYTE_TO_CHAR (SREF (string->lstring, bytepos)); | |
1019 | *ch_len = 1; | |
1020 | } | |
9f257352 | 1021 | } |
87e67904 EZ |
1022 | else |
1023 | { | |
1024 | ch = FETCH_MULTIBYTE_CHAR (bytepos); | |
1025 | *ch_len = CHAR_BYTES (ch); | |
1026 | } | |
182ce2d2 EZ |
1027 | *nchars = 1; |
1028 | } | |
1029 | ||
1030 | /* If we just entered a run of characters covered by a display | |
1031 | string, compute the position of the next display string. */ | |
55439c61 | 1032 | if (charpos + *nchars <= endpos && charpos + *nchars > *disp_pos |
0c95fcf7 | 1033 | && *disp_prop) |
6db161be EZ |
1034 | { |
1035 | SET_TEXT_POS (pos, charpos + *nchars, bytepos + *ch_len); | |
55439c61 | 1036 | *disp_pos = compute_display_string_pos (&pos, string, frame_window_p, |
0c95fcf7 | 1037 | disp_prop); |
6db161be | 1038 | } |
182ce2d2 EZ |
1039 | |
1040 | return ch; | |
1041 | } | |
1042 | ||
cbb09f04 EZ |
1043 | \f |
1044 | /*********************************************************************** | |
1045 | Determining paragraph direction | |
1046 | ***********************************************************************/ | |
1047 | ||
1048 | /* Check if buffer position CHARPOS/BYTEPOS is the end of a paragraph. | |
1049 | Value is the non-negative length of the paragraph separator | |
1050 | following the buffer position, -1 if position is at the beginning | |
1051 | of a new paragraph, or -2 if position is neither at beginning nor | |
1052 | at end of a paragraph. */ | |
1053 | static EMACS_INT | |
1054 | bidi_at_paragraph_end (EMACS_INT charpos, EMACS_INT bytepos) | |
1055 | { | |
1056 | Lisp_Object sep_re; | |
1057 | Lisp_Object start_re; | |
1058 | EMACS_INT val; | |
1059 | ||
1060 | sep_re = paragraph_separate_re; | |
1061 | start_re = paragraph_start_re; | |
1062 | ||
1063 | val = fast_looking_at (sep_re, charpos, bytepos, ZV, ZV_BYTE, Qnil); | |
1064 | if (val < 0) | |
1065 | { | |
1066 | if (fast_looking_at (start_re, charpos, bytepos, ZV, ZV_BYTE, Qnil) >= 0) | |
1067 | val = -1; | |
1068 | else | |
1069 | val = -2; | |
1070 | } | |
1071 | ||
1072 | return val; | |
1073 | } | |
1074 | ||
be39f003 EZ |
1075 | /* Find the beginning of this paragraph by looking back in the buffer. |
1076 | Value is the byte position of the paragraph's beginning. */ | |
1077 | static EMACS_INT | |
b44d9321 | 1078 | bidi_find_paragraph_start (EMACS_INT pos, EMACS_INT pos_byte) |
6bff6497 | 1079 | { |
372b7a95 | 1080 | Lisp_Object re = paragraph_start_re; |
6bff6497 EZ |
1081 | EMACS_INT limit = ZV, limit_byte = ZV_BYTE; |
1082 | ||
6bff6497 EZ |
1083 | while (pos_byte > BEGV_BYTE |
1084 | && fast_looking_at (re, pos, pos_byte, limit, limit_byte, Qnil) < 0) | |
1085 | { | |
182ce2d2 EZ |
1086 | /* FIXME: What if the paragraph beginning is covered by a |
1087 | display string? And what if a display string covering some | |
1088 | of the text over which we scan back includes | |
1089 | paragraph_start_re? */ | |
be39f003 EZ |
1090 | pos = find_next_newline_no_quit (pos - 1, -1); |
1091 | pos_byte = CHAR_TO_BYTE (pos); | |
6bff6497 | 1092 | } |
be39f003 | 1093 | return pos_byte; |
6bff6497 EZ |
1094 | } |
1095 | ||
bea4f10c | 1096 | /* Determine the base direction, a.k.a. base embedding level, of the |
b44d9321 EZ |
1097 | paragraph we are about to iterate through. If DIR is either L2R or |
1098 | R2L, just use that. Otherwise, determine the paragraph direction | |
bea4f10c EZ |
1099 | from the first strong directional character of the paragraph. |
1100 | ||
182ce2d2 | 1101 | NO_DEFAULT_P non-zero means don't default to L2R if the paragraph |
bea4f10c EZ |
1102 | has no strong directional characters and both DIR and |
1103 | bidi_it->paragraph_dir are NEUTRAL_DIR. In that case, search back | |
1104 | in the buffer until a paragraph is found with a strong character, | |
1105 | or until hitting BEGV. In the latter case, fall back to L2R. This | |
1106 | flag is used in current-bidi-paragraph-direction. | |
1107 | ||
1108 | Note that this function gives the paragraph separator the same | |
1109 | direction as the preceding paragraph, even though Emacs generally | |
1110 | views the separartor as not belonging to any paragraph. */ | |
b7b65b15 | 1111 | void |
bea4f10c | 1112 | bidi_paragraph_init (bidi_dir_t dir, struct bidi_it *bidi_it, int no_default_p) |
b7b65b15 | 1113 | { |
6bff6497 | 1114 | EMACS_INT bytepos = bidi_it->bytepos; |
9f257352 | 1115 | int string_p = bidi_it->string.s != NULL || STRINGP (bidi_it->string.lstring); |
bea4f10c | 1116 | EMACS_INT pstartbyte; |
87e67904 EZ |
1117 | /* Note that begbyte is a byte position, while end is a character |
1118 | position. Yes, this is ugly, but we are trying to avoid costly | |
1119 | calls to BYTE_TO_CHAR and its ilk. */ | |
1120 | EMACS_INT begbyte = string_p ? 0 : BEGV_BYTE; | |
1121 | EMACS_INT end = string_p ? bidi_it->string.schars : ZV; | |
e7402cb2 | 1122 | |
5e65aec0 | 1123 | /* Special case for an empty buffer. */ |
87e67904 | 1124 | if (bytepos == begbyte && bidi_it->charpos == end) |
5e65aec0 | 1125 | dir = L2R; |
9c82e145 | 1126 | /* We should never be called at EOB or before BEGV. */ |
87e67904 | 1127 | else if (bidi_it->charpos >= end || bytepos < begbyte) |
9c82e145 EZ |
1128 | abort (); |
1129 | ||
be39f003 EZ |
1130 | if (dir == L2R) |
1131 | { | |
1132 | bidi_it->paragraph_dir = L2R; | |
1133 | bidi_it->new_paragraph = 0; | |
1134 | } | |
1135 | else if (dir == R2L) | |
1136 | { | |
1137 | bidi_it->paragraph_dir = R2L; | |
1138 | bidi_it->new_paragraph = 0; | |
1139 | } | |
b7b65b15 EZ |
1140 | else if (dir == NEUTRAL_DIR) /* P2 */ |
1141 | { | |
182ce2d2 EZ |
1142 | int ch; |
1143 | EMACS_INT ch_len, nchars; | |
1144 | EMACS_INT pos, disp_pos = -1; | |
0c95fcf7 | 1145 | int disp_prop = 0; |
6bff6497 | 1146 | bidi_type_t type; |
9f257352 | 1147 | const unsigned char *s; |
be39f003 | 1148 | |
d20e1419 EZ |
1149 | if (!bidi_initialized) |
1150 | bidi_initialize (); | |
1151 | ||
be39f003 EZ |
1152 | /* If we are inside a paragraph separator, we are just waiting |
1153 | for the separator to be exhausted; use the previous paragraph | |
e5a2fec7 EZ |
1154 | direction. But don't do that if we have been just reseated, |
1155 | because we need to reinitialize below in that case. */ | |
1156 | if (!bidi_it->first_elt | |
1157 | && bidi_it->charpos < bidi_it->separator_limit) | |
be39f003 EZ |
1158 | return; |
1159 | ||
b44d9321 | 1160 | /* If we are on a newline, get past it to where the next |
5e65aec0 EZ |
1161 | paragraph might start. But don't do that at BEGV since then |
1162 | we are potentially in a new paragraph that doesn't yet | |
1163 | exist. */ | |
c143c213 | 1164 | pos = bidi_it->charpos; |
9f257352 EZ |
1165 | s = STRINGP (bidi_it->string.lstring) ? |
1166 | SDATA (bidi_it->string.lstring) : bidi_it->string.s; | |
f3014ef5 EZ |
1167 | if (bytepos > begbyte |
1168 | && bidi_char_at_pos (bytepos, s, bidi_it->string.unibyte) == '\n') | |
be39f003 | 1169 | { |
b44d9321 | 1170 | bytepos++; |
c143c213 | 1171 | pos++; |
be39f003 | 1172 | } |
b44d9321 EZ |
1173 | |
1174 | /* We are either at the beginning of a paragraph or in the | |
1175 | middle of it. Find where this paragraph starts. */ | |
87e67904 EZ |
1176 | if (string_p) |
1177 | { | |
1178 | /* We don't support changes of paragraph direction inside a | |
1179 | string. It is treated as a single paragraph. */ | |
1180 | pstartbyte = 0; | |
1181 | } | |
1182 | else | |
1183 | pstartbyte = bidi_find_paragraph_start (pos, bytepos); | |
be39f003 EZ |
1184 | bidi_it->separator_limit = -1; |
1185 | bidi_it->new_paragraph = 0; | |
bea4f10c EZ |
1186 | |
1187 | /* The following loop is run more than once only if NO_DEFAULT_P | |
87e67904 | 1188 | is non-zero, and only if we are iterating on a buffer. */ |
bea4f10c EZ |
1189 | do { |
1190 | bytepos = pstartbyte; | |
87e67904 EZ |
1191 | if (!string_p) |
1192 | pos = BYTE_TO_CHAR (bytepos); | |
0c95fcf7 | 1193 | ch = bidi_fetch_char (bytepos, pos, &disp_pos, &disp_prop, |
55439c61 | 1194 | &bidi_it->string, |
87e67904 | 1195 | bidi_it->frame_window_p, &ch_len, &nchars); |
bea4f10c EZ |
1196 | type = bidi_get_type (ch, NEUTRAL_DIR); |
1197 | ||
182ce2d2 | 1198 | for (pos += nchars, bytepos += ch_len; |
bea4f10c EZ |
1199 | (bidi_get_category (type) != STRONG) |
1200 | || (bidi_ignore_explicit_marks_for_paragraph_level | |
1201 | && (type == RLE || type == RLO | |
1202 | || type == LRE || type == LRO)); | |
1203 | type = bidi_get_type (ch, NEUTRAL_DIR)) | |
1204 | { | |
87e67904 | 1205 | if (pos >= end) |
bea4f10c EZ |
1206 | { |
1207 | /* Pretend there's a paragraph separator at end of | |
87e67904 | 1208 | buffer/string. */ |
bea4f10c EZ |
1209 | type = NEUTRAL_B; |
1210 | break; | |
1211 | } | |
0bb23927 EZ |
1212 | if (!string_p |
1213 | && type == NEUTRAL_B | |
1214 | && bidi_at_paragraph_end (pos, bytepos) >= -1) | |
cf99dcf8 | 1215 | break; |
102ebb00 | 1216 | /* Fetch next character and advance to get past it. */ |
55439c61 | 1217 | ch = bidi_fetch_char (bytepos, pos, &disp_pos, |
0c95fcf7 | 1218 | &disp_prop, &bidi_it->string, |
fc6f18ce | 1219 | bidi_it->frame_window_p, &ch_len, &nchars); |
182ce2d2 EZ |
1220 | pos += nchars; |
1221 | bytepos += ch_len; | |
bea4f10c | 1222 | } |
32413314 EZ |
1223 | if ((type == STRONG_R || type == STRONG_AL) /* P3 */ |
1224 | || (!bidi_ignore_explicit_marks_for_paragraph_level | |
1225 | && (type == RLO || type == RLE))) | |
bea4f10c | 1226 | bidi_it->paragraph_dir = R2L; |
32413314 EZ |
1227 | else if (type == STRONG_L |
1228 | || (!bidi_ignore_explicit_marks_for_paragraph_level | |
1229 | && (type == LRO || type == LRE))) | |
bea4f10c | 1230 | bidi_it->paragraph_dir = L2R; |
87e67904 EZ |
1231 | if (!string_p |
1232 | && no_default_p && bidi_it->paragraph_dir == NEUTRAL_DIR) | |
bea4f10c EZ |
1233 | { |
1234 | /* If this paragraph is at BEGV, default to L2R. */ | |
1235 | if (pstartbyte == BEGV_BYTE) | |
1236 | bidi_it->paragraph_dir = L2R; /* P3 and HL1 */ | |
1237 | else | |
1238 | { | |
1239 | EMACS_INT prevpbyte = pstartbyte; | |
1240 | EMACS_INT p = BYTE_TO_CHAR (pstartbyte), pbyte = pstartbyte; | |
1241 | ||
1242 | /* Find the beginning of the previous paragraph, if any. */ | |
1243 | while (pbyte > BEGV_BYTE && prevpbyte >= pstartbyte) | |
1244 | { | |
182ce2d2 EZ |
1245 | /* FXIME: What if p is covered by a display |
1246 | string? See also a FIXME inside | |
1247 | bidi_find_paragraph_start. */ | |
bea4f10c EZ |
1248 | p--; |
1249 | pbyte = CHAR_TO_BYTE (p); | |
1250 | prevpbyte = bidi_find_paragraph_start (p, pbyte); | |
1251 | } | |
1252 | pstartbyte = prevpbyte; | |
1253 | } | |
1254 | } | |
87e67904 EZ |
1255 | } while (!string_p |
1256 | && no_default_p && bidi_it->paragraph_dir == NEUTRAL_DIR); | |
b7b65b15 | 1257 | } |
be39f003 EZ |
1258 | else |
1259 | abort (); | |
b7b65b15 | 1260 | |
b44d9321 EZ |
1261 | /* Contrary to UAX#9 clause P3, we only default the paragraph |
1262 | direction to L2R if we have no previous usable paragraph | |
bea4f10c | 1263 | direction. This is allowed by the HL1 clause. */ |
d20e1419 | 1264 | if (bidi_it->paragraph_dir != L2R && bidi_it->paragraph_dir != R2L) |
bea4f10c | 1265 | bidi_it->paragraph_dir = L2R; /* P3 and HL1 ``higher-level protocols'' */ |
be39f003 | 1266 | if (bidi_it->paragraph_dir == R2L) |
b44d9321 | 1267 | bidi_it->level_stack[0].level = 1; |
be39f003 | 1268 | else |
b44d9321 | 1269 | bidi_it->level_stack[0].level = 0; |
be39f003 EZ |
1270 | |
1271 | bidi_line_init (bidi_it); | |
b7b65b15 EZ |
1272 | } |
1273 | ||
cbb09f04 EZ |
1274 | \f |
1275 | /*********************************************************************** | |
1276 | Resolving explicit and implicit levels. | |
58b9f433 | 1277 | The rest of this file constitutes the core of the UBA implementation. |
cbb09f04 | 1278 | ***********************************************************************/ |
b7b65b15 | 1279 | |
55d4c1b2 | 1280 | static inline int |
182ce2d2 | 1281 | bidi_explicit_dir_char (int ch) |
b7b65b15 | 1282 | { |
182ce2d2 EZ |
1283 | bidi_type_t ch_type; |
1284 | ||
1285 | if (!bidi_initialized) | |
1286 | abort (); | |
1287 | ch_type = (bidi_type_t) XINT (CHAR_TABLE_REF (bidi_type_table, ch)); | |
1288 | return (ch_type == LRE || ch_type == LRO | |
1289 | || ch_type == RLE || ch_type == RLO | |
1290 | || ch_type == PDF); | |
b7b65b15 EZ |
1291 | } |
1292 | ||
1293 | /* A helper function for bidi_resolve_explicit. It advances to the | |
1294 | next character in logical order and determines the new embedding | |
1295 | level and directional override, but does not take into account | |
1296 | empty embeddings. */ | |
1297 | static int | |
1298 | bidi_resolve_explicit_1 (struct bidi_it *bidi_it) | |
1299 | { | |
1300 | int curchar; | |
1301 | bidi_type_t type; | |
1302 | int current_level; | |
1303 | int new_level; | |
1304 | bidi_dir_t override; | |
9f257352 | 1305 | int string_p = bidi_it->string.s != NULL || STRINGP (bidi_it->string.lstring); |
b7b65b15 | 1306 | |
182ce2d2 EZ |
1307 | /* If reseat()'ed, don't advance, so as to start iteration from the |
1308 | position where we were reseated. bidi_it->bytepos can be less | |
1309 | than BEGV_BYTE after reseat to BEGV. */ | |
87e67904 | 1310 | if (bidi_it->bytepos < (string_p ? 0 : BEGV_BYTE) |
9c82e145 | 1311 | || bidi_it->first_elt) |
e7402cb2 | 1312 | { |
9c82e145 | 1313 | bidi_it->first_elt = 0; |
87e67904 EZ |
1314 | if (string_p) |
1315 | { | |
9f257352 EZ |
1316 | const unsigned char *p = |
1317 | STRINGP (bidi_it->string.lstring) | |
1318 | ? SDATA (bidi_it->string.lstring) : bidi_it->string.s; | |
1319 | ||
87e67904 EZ |
1320 | if (bidi_it->charpos < 0) |
1321 | bidi_it->charpos = 0; | |
f3014ef5 EZ |
1322 | bidi_it->bytepos = bidi_count_bytes (p, 0, 0, bidi_it->charpos, |
1323 | bidi_it->string.unibyte); | |
87e67904 EZ |
1324 | } |
1325 | else | |
1326 | { | |
1327 | if (bidi_it->charpos < BEGV) | |
1328 | bidi_it->charpos = BEGV; | |
1329 | bidi_it->bytepos = CHAR_TO_BYTE (bidi_it->charpos); | |
1330 | } | |
e7402cb2 | 1331 | } |
87e67904 EZ |
1332 | /* Don't move at end of buffer/string. */ |
1333 | else if (bidi_it->charpos < (string_p ? bidi_it->string.schars : ZV)) | |
b7b65b15 | 1334 | { |
182ce2d2 EZ |
1335 | /* Advance to the next character, skipping characters covered by |
1336 | display strings (nchars > 1). */ | |
102ebb00 EZ |
1337 | if (bidi_it->nchars <= 0) |
1338 | abort (); | |
182ce2d2 | 1339 | bidi_it->charpos += bidi_it->nchars; |
e7402cb2 EZ |
1340 | if (bidi_it->ch_len == 0) |
1341 | abort (); | |
b7b65b15 EZ |
1342 | bidi_it->bytepos += bidi_it->ch_len; |
1343 | } | |
1344 | ||
1345 | current_level = bidi_it->level_stack[bidi_it->stack_idx].level; /* X1 */ | |
1346 | override = bidi_it->level_stack[bidi_it->stack_idx].override; | |
1347 | new_level = current_level; | |
1348 | ||
87e67904 | 1349 | if (bidi_it->charpos >= (string_p ? bidi_it->string.schars : ZV)) |
e7402cb2 EZ |
1350 | { |
1351 | curchar = BIDI_EOB; | |
1352 | bidi_it->ch_len = 1; | |
182ce2d2 | 1353 | bidi_it->nchars = 1; |
87e67904 | 1354 | bidi_it->disp_pos = (string_p ? bidi_it->string.schars : ZV); |
0c95fcf7 | 1355 | bidi_it->disp_prop = 0; |
e7402cb2 EZ |
1356 | } |
1357 | else | |
1358 | { | |
182ce2d2 EZ |
1359 | /* Fetch the character at BYTEPOS. If it is covered by a |
1360 | display string, treat the entire run of covered characters as | |
1361 | a single character u+FFFC. */ | |
102ebb00 | 1362 | curchar = bidi_fetch_char (bidi_it->bytepos, bidi_it->charpos, |
0c95fcf7 | 1363 | &bidi_it->disp_pos, &bidi_it->disp_prop, |
55439c61 | 1364 | &bidi_it->string, bidi_it->frame_window_p, |
102ebb00 | 1365 | &bidi_it->ch_len, &bidi_it->nchars); |
e7402cb2 | 1366 | } |
b7b65b15 | 1367 | bidi_it->ch = curchar; |
b7b65b15 | 1368 | |
6bff6497 EZ |
1369 | /* Don't apply directional override here, as all the types we handle |
1370 | below will not be affected by the override anyway, and we need | |
1371 | the original type unaltered. The override will be applied in | |
1372 | bidi_resolve_weak. */ | |
1373 | type = bidi_get_type (curchar, NEUTRAL_DIR); | |
89d3374a EZ |
1374 | bidi_it->orig_type = type; |
1375 | bidi_check_type (bidi_it->orig_type); | |
b7b65b15 EZ |
1376 | |
1377 | if (type != PDF) | |
1378 | bidi_it->prev_was_pdf = 0; | |
1379 | ||
89d3374a | 1380 | bidi_it->type_after_w1 = UNKNOWN_BT; |
b7b65b15 EZ |
1381 | |
1382 | switch (type) | |
1383 | { | |
1384 | case RLE: /* X2 */ | |
1385 | case RLO: /* X4 */ | |
89d3374a EZ |
1386 | bidi_it->type_after_w1 = type; |
1387 | bidi_check_type (bidi_it->type_after_w1); | |
b7b65b15 | 1388 | type = WEAK_BN; /* X9/Retaining */ |
87e67904 | 1389 | if (bidi_it->ignore_bn_limit <= -1) |
b7b65b15 EZ |
1390 | { |
1391 | if (current_level <= BIDI_MAXLEVEL - 4) | |
1392 | { | |
1393 | /* Compute the least odd embedding level greater than | |
1394 | the current level. */ | |
1395 | new_level = ((current_level + 1) & ~1) + 1; | |
89d3374a | 1396 | if (bidi_it->type_after_w1 == RLE) |
b7b65b15 EZ |
1397 | override = NEUTRAL_DIR; |
1398 | else | |
1399 | override = R2L; | |
1400 | if (current_level == BIDI_MAXLEVEL - 4) | |
1401 | bidi_it->invalid_rl_levels = 0; | |
1402 | bidi_push_embedding_level (bidi_it, new_level, override); | |
1403 | } | |
1404 | else | |
1405 | { | |
1406 | bidi_it->invalid_levels++; | |
1407 | /* See the commentary about invalid_rl_levels below. */ | |
1408 | if (bidi_it->invalid_rl_levels < 0) | |
1409 | bidi_it->invalid_rl_levels = 0; | |
1410 | bidi_it->invalid_rl_levels++; | |
1411 | } | |
1412 | } | |
89d3374a | 1413 | else if (bidi_it->prev.type_after_w1 == WEAK_EN /* W5/Retaining */ |
b7b65b15 EZ |
1414 | || bidi_it->next_en_pos > bidi_it->charpos) |
1415 | type = WEAK_EN; | |
1416 | break; | |
1417 | case LRE: /* X3 */ | |
1418 | case LRO: /* X5 */ | |
89d3374a EZ |
1419 | bidi_it->type_after_w1 = type; |
1420 | bidi_check_type (bidi_it->type_after_w1); | |
b7b65b15 | 1421 | type = WEAK_BN; /* X9/Retaining */ |
87e67904 | 1422 | if (bidi_it->ignore_bn_limit <= -1) |
b7b65b15 EZ |
1423 | { |
1424 | if (current_level <= BIDI_MAXLEVEL - 5) | |
1425 | { | |
1426 | /* Compute the least even embedding level greater than | |
1427 | the current level. */ | |
1428 | new_level = ((current_level + 2) & ~1); | |
89d3374a | 1429 | if (bidi_it->type_after_w1 == LRE) |
b7b65b15 EZ |
1430 | override = NEUTRAL_DIR; |
1431 | else | |
1432 | override = L2R; | |
1433 | bidi_push_embedding_level (bidi_it, new_level, override); | |
1434 | } | |
1435 | else | |
1436 | { | |
1437 | bidi_it->invalid_levels++; | |
1438 | /* invalid_rl_levels counts invalid levels encountered | |
1439 | while the embedding level was already too high for | |
1440 | LRE/LRO, but not for RLE/RLO. That is because | |
1441 | there may be exactly one PDF which we should not | |
1442 | ignore even though invalid_levels is non-zero. | |
1443 | invalid_rl_levels helps to know what PDF is | |
1444 | that. */ | |
1445 | if (bidi_it->invalid_rl_levels >= 0) | |
1446 | bidi_it->invalid_rl_levels++; | |
1447 | } | |
1448 | } | |
89d3374a | 1449 | else if (bidi_it->prev.type_after_w1 == WEAK_EN /* W5/Retaining */ |
b7b65b15 EZ |
1450 | || bidi_it->next_en_pos > bidi_it->charpos) |
1451 | type = WEAK_EN; | |
1452 | break; | |
1453 | case PDF: /* X7 */ | |
89d3374a EZ |
1454 | bidi_it->type_after_w1 = type; |
1455 | bidi_check_type (bidi_it->type_after_w1); | |
b7b65b15 | 1456 | type = WEAK_BN; /* X9/Retaining */ |
87e67904 | 1457 | if (bidi_it->ignore_bn_limit <= -1) |
b7b65b15 EZ |
1458 | { |
1459 | if (!bidi_it->invalid_rl_levels) | |
1460 | { | |
1461 | new_level = bidi_pop_embedding_level (bidi_it); | |
1462 | bidi_it->invalid_rl_levels = -1; | |
1463 | if (bidi_it->invalid_levels) | |
1464 | bidi_it->invalid_levels--; | |
1465 | /* else nothing: UAX#9 says to ignore invalid PDFs */ | |
1466 | } | |
1467 | if (!bidi_it->invalid_levels) | |
1468 | new_level = bidi_pop_embedding_level (bidi_it); | |
1469 | else | |
1470 | { | |
1471 | bidi_it->invalid_levels--; | |
1472 | bidi_it->invalid_rl_levels--; | |
1473 | } | |
1474 | } | |
89d3374a | 1475 | else if (bidi_it->prev.type_after_w1 == WEAK_EN /* W5/Retaining */ |
b7b65b15 EZ |
1476 | || bidi_it->next_en_pos > bidi_it->charpos) |
1477 | type = WEAK_EN; | |
1478 | break; | |
1479 | default: | |
1480 | /* Nothing. */ | |
1481 | break; | |
1482 | } | |
1483 | ||
1484 | bidi_it->type = type; | |
2d6e4628 | 1485 | bidi_check_type (bidi_it->type); |
b7b65b15 EZ |
1486 | |
1487 | return new_level; | |
1488 | } | |
1489 | ||
1490 | /* Given an iterator state in BIDI_IT, advance one character position | |
182ce2d2 EZ |
1491 | in the buffer/string to the next character (in the logical order), |
1492 | resolve any explicit embeddings and directional overrides, and | |
1493 | return the embedding level of the character after resolving | |
1494 | explicit directives and ignoring empty embeddings. */ | |
b7b65b15 EZ |
1495 | static int |
1496 | bidi_resolve_explicit (struct bidi_it *bidi_it) | |
1497 | { | |
1498 | int prev_level = bidi_it->level_stack[bidi_it->stack_idx].level; | |
1499 | int new_level = bidi_resolve_explicit_1 (bidi_it); | |
87e67904 | 1500 | EMACS_INT eob = bidi_it->string.s ? bidi_it->string.schars : ZV; |
9f257352 EZ |
1501 | const unsigned char *s = STRINGP (bidi_it->string.lstring) |
1502 | ? SDATA (bidi_it->string.lstring) : bidi_it->string.s; | |
b7b65b15 EZ |
1503 | |
1504 | if (prev_level < new_level | |
1505 | && bidi_it->type == WEAK_BN | |
87e67904 EZ |
1506 | && bidi_it->ignore_bn_limit == -1 /* only if not already known */ |
1507 | && bidi_it->charpos < eob /* not already at EOB */ | |
1508 | && bidi_explicit_dir_char (bidi_char_at_pos (bidi_it->bytepos | |
f3014ef5 EZ |
1509 | + bidi_it->ch_len, s, |
1510 | bidi_it->string.unibyte))) | |
b7b65b15 EZ |
1511 | { |
1512 | /* Avoid pushing and popping embedding levels if the level run | |
1513 | is empty, as this breaks level runs where it shouldn't. | |
1514 | UAX#9 removes all the explicit embedding and override codes, | |
1515 | so empty embeddings disappear without a trace. We need to | |
1516 | behave as if we did the same. */ | |
1517 | struct bidi_it saved_it; | |
1518 | int level = prev_level; | |
1519 | ||
1520 | bidi_copy_it (&saved_it, bidi_it); | |
1521 | ||
87e67904 | 1522 | while (bidi_explicit_dir_char (bidi_char_at_pos (bidi_it->bytepos |
f3014ef5 EZ |
1523 | + bidi_it->ch_len, s, |
1524 | bidi_it->string.unibyte))) | |
b7b65b15 | 1525 | { |
182ce2d2 EZ |
1526 | /* This advances to the next character, skipping any |
1527 | characters covered by display strings. */ | |
b7b65b15 | 1528 | level = bidi_resolve_explicit_1 (bidi_it); |
9f257352 EZ |
1529 | /* If string.lstring was relocated inside bidi_resolve_explicit_1, |
1530 | a pointer to its data is no longer valid. */ | |
1531 | if (STRINGP (bidi_it->string.lstring)) | |
1532 | s = SDATA (bidi_it->string.lstring); | |
b7b65b15 EZ |
1533 | } |
1534 | ||
102ebb00 EZ |
1535 | if (bidi_it->nchars <= 0) |
1536 | abort (); | |
b7b65b15 | 1537 | if (level == prev_level) /* empty embedding */ |
182ce2d2 | 1538 | saved_it.ignore_bn_limit = bidi_it->charpos + bidi_it->nchars; |
b7b65b15 | 1539 | else /* this embedding is non-empty */ |
87e67904 | 1540 | saved_it.ignore_bn_limit = -2; |
b7b65b15 EZ |
1541 | |
1542 | bidi_copy_it (bidi_it, &saved_it); | |
87e67904 | 1543 | if (bidi_it->ignore_bn_limit > -1) |
b7b65b15 EZ |
1544 | { |
1545 | /* We pushed a level, but we shouldn't have. Undo that. */ | |
1546 | if (!bidi_it->invalid_rl_levels) | |
1547 | { | |
1548 | new_level = bidi_pop_embedding_level (bidi_it); | |
1549 | bidi_it->invalid_rl_levels = -1; | |
1550 | if (bidi_it->invalid_levels) | |
1551 | bidi_it->invalid_levels--; | |
1552 | } | |
1553 | if (!bidi_it->invalid_levels) | |
1554 | new_level = bidi_pop_embedding_level (bidi_it); | |
1555 | else | |
1556 | { | |
1557 | bidi_it->invalid_levels--; | |
1558 | bidi_it->invalid_rl_levels--; | |
1559 | } | |
1560 | } | |
1561 | } | |
1562 | ||
b7b65b15 EZ |
1563 | if (bidi_it->type == NEUTRAL_B) /* X8 */ |
1564 | { | |
21fce5ab | 1565 | bidi_set_paragraph_end (bidi_it); |
6bff6497 EZ |
1566 | /* This is needed by bidi_resolve_weak below, and in L1. */ |
1567 | bidi_it->type_after_w1 = bidi_it->type; | |
89d3374a | 1568 | bidi_check_type (bidi_it->type_after_w1); |
b7b65b15 EZ |
1569 | } |
1570 | ||
1571 | return new_level; | |
1572 | } | |
1573 | ||
182ce2d2 EZ |
1574 | /* Advance in the buffer/string, resolve weak types and return the |
1575 | type of the next character after weak type resolution. */ | |
fd3998ff | 1576 | static bidi_type_t |
b7b65b15 EZ |
1577 | bidi_resolve_weak (struct bidi_it *bidi_it) |
1578 | { | |
1579 | bidi_type_t type; | |
1580 | bidi_dir_t override; | |
1581 | int prev_level = bidi_it->level_stack[bidi_it->stack_idx].level; | |
1582 | int new_level = bidi_resolve_explicit (bidi_it); | |
1583 | int next_char; | |
1584 | bidi_type_t type_of_next; | |
1585 | struct bidi_it saved_it; | |
9f257352 EZ |
1586 | EMACS_INT eob = |
1587 | (STRINGP (bidi_it->string.lstring) || bidi_it->string.s) | |
1588 | ? bidi_it->string.schars : ZV; | |
b7b65b15 EZ |
1589 | |
1590 | type = bidi_it->type; | |
1591 | override = bidi_it->level_stack[bidi_it->stack_idx].override; | |
1592 | ||
1593 | if (type == UNKNOWN_BT | |
1594 | || type == LRE | |
1595 | || type == LRO | |
1596 | || type == RLE | |
1597 | || type == RLO | |
1598 | || type == PDF) | |
1599 | abort (); | |
1600 | ||
1601 | if (new_level != prev_level | |
1602 | || bidi_it->type == NEUTRAL_B) | |
1603 | { | |
1604 | /* We've got a new embedding level run, compute the directional | |
1605 | type of sor and initialize per-run variables (UAX#9, clause | |
1606 | X10). */ | |
1607 | bidi_set_sor_type (bidi_it, prev_level, new_level); | |
1608 | } | |
1609 | else if (type == NEUTRAL_S || type == NEUTRAL_WS | |
1610 | || type == WEAK_BN || type == STRONG_AL) | |
89d3374a EZ |
1611 | bidi_it->type_after_w1 = type; /* needed in L1 */ |
1612 | bidi_check_type (bidi_it->type_after_w1); | |
b7b65b15 EZ |
1613 | |
1614 | /* Level and directional override status are already recorded in | |
1615 | bidi_it, and do not need any change; see X6. */ | |
1616 | if (override == R2L) /* X6 */ | |
1617 | type = STRONG_R; | |
1618 | else if (override == L2R) | |
1619 | type = STRONG_L; | |
bc5a45f3 | 1620 | else |
b7b65b15 | 1621 | { |
bc5a45f3 | 1622 | if (type == WEAK_NSM) /* W1 */ |
b7b65b15 | 1623 | { |
bc5a45f3 | 1624 | /* Note that we don't need to consider the case where the |
5930fe97 EZ |
1625 | prev character has its type overridden by an RLO or LRO, |
1626 | because then either the type of this NSM would have been | |
1627 | also overridden, or the previous character is outside the | |
1628 | current level run, and thus not relevant to this NSM. | |
1629 | This is why NSM gets the type_after_w1 of the previous | |
1630 | character. */ | |
ebb5722e EZ |
1631 | if (bidi_it->prev.type_after_w1 != UNKNOWN_BT |
1632 | /* if type_after_w1 is NEUTRAL_B, this NSM is at sor */ | |
1633 | && bidi_it->prev.type_after_w1 != NEUTRAL_B) | |
5930fe97 | 1634 | type = bidi_it->prev.type_after_w1; |
bc5a45f3 EZ |
1635 | else if (bidi_it->sor == R2L) |
1636 | type = STRONG_R; | |
1637 | else if (bidi_it->sor == L2R) | |
1638 | type = STRONG_L; | |
1639 | else /* shouldn't happen! */ | |
1640 | abort (); | |
b7b65b15 | 1641 | } |
bc5a45f3 EZ |
1642 | if (type == WEAK_EN /* W2 */ |
1643 | && bidi_it->last_strong.type_after_w1 == STRONG_AL) | |
1644 | type = WEAK_AN; | |
1645 | else if (type == STRONG_AL) /* W3 */ | |
1646 | type = STRONG_R; | |
1647 | else if ((type == WEAK_ES /* W4 */ | |
1648 | && bidi_it->prev.type_after_w1 == WEAK_EN | |
1649 | && bidi_it->prev.orig_type == WEAK_EN) | |
1650 | || (type == WEAK_CS | |
1651 | && ((bidi_it->prev.type_after_w1 == WEAK_EN | |
1652 | && bidi_it->prev.orig_type == WEAK_EN) | |
1653 | || bidi_it->prev.type_after_w1 == WEAK_AN))) | |
b7b65b15 | 1654 | { |
9f257352 EZ |
1655 | const unsigned char *s = |
1656 | STRINGP (bidi_it->string.lstring) | |
1657 | ? SDATA (bidi_it->string.lstring) : bidi_it->string.s; | |
1658 | ||
e7402cb2 | 1659 | next_char = |
87e67904 EZ |
1660 | bidi_it->charpos + bidi_it->nchars >= eob |
1661 | ? BIDI_EOB | |
f3014ef5 EZ |
1662 | : bidi_char_at_pos (bidi_it->bytepos + bidi_it->ch_len, s, |
1663 | bidi_it->string.unibyte); | |
6bff6497 | 1664 | type_of_next = bidi_get_type (next_char, override); |
b7b65b15 | 1665 | |
bc5a45f3 | 1666 | if (type_of_next == WEAK_BN |
b7b65b15 EZ |
1667 | || bidi_explicit_dir_char (next_char)) |
1668 | { | |
1669 | bidi_copy_it (&saved_it, bidi_it); | |
1670 | while (bidi_resolve_explicit (bidi_it) == new_level | |
bc5a45f3 | 1671 | && bidi_it->type == WEAK_BN) |
b7b65b15 EZ |
1672 | ; |
1673 | type_of_next = bidi_it->type; | |
b7b65b15 EZ |
1674 | bidi_copy_it (bidi_it, &saved_it); |
1675 | } | |
bc5a45f3 EZ |
1676 | |
1677 | /* If the next character is EN, but the last strong-type | |
1678 | character is AL, that next EN will be changed to AN when | |
1679 | we process it in W2 above. So in that case, this ES | |
1680 | should not be changed into EN. */ | |
1681 | if (type == WEAK_ES | |
1682 | && type_of_next == WEAK_EN | |
1683 | && bidi_it->last_strong.type_after_w1 != STRONG_AL) | |
1684 | type = WEAK_EN; | |
1685 | else if (type == WEAK_CS) | |
b7b65b15 | 1686 | { |
bc5a45f3 EZ |
1687 | if (bidi_it->prev.type_after_w1 == WEAK_AN |
1688 | && (type_of_next == WEAK_AN | |
1689 | /* If the next character is EN, but the last | |
1690 | strong-type character is AL, EN will be later | |
1691 | changed to AN when we process it in W2 above. | |
1692 | So in that case, this ES should not be | |
1693 | changed into EN. */ | |
1694 | || (type_of_next == WEAK_EN | |
1695 | && bidi_it->last_strong.type_after_w1 == STRONG_AL))) | |
1696 | type = WEAK_AN; | |
1697 | else if (bidi_it->prev.type_after_w1 == WEAK_EN | |
1698 | && type_of_next == WEAK_EN | |
1699 | && bidi_it->last_strong.type_after_w1 != STRONG_AL) | |
1700 | type = WEAK_EN; | |
1701 | } | |
1702 | } | |
1703 | else if (type == WEAK_ET /* W5: ET with EN before or after it */ | |
1704 | || type == WEAK_BN) /* W5/Retaining */ | |
1705 | { | |
1706 | if (bidi_it->prev.type_after_w1 == WEAK_EN /* ET/BN w/EN before it */ | |
1707 | || bidi_it->next_en_pos > bidi_it->charpos) | |
1708 | type = WEAK_EN; | |
1709 | else /* W5: ET/BN with EN after it. */ | |
1710 | { | |
182ce2d2 | 1711 | EMACS_INT en_pos = bidi_it->charpos + bidi_it->nchars; |
9f257352 EZ |
1712 | const unsigned char *s = |
1713 | STRINGP (bidi_it->string.lstring) | |
1714 | ? SDATA (bidi_it->string.lstring) : bidi_it->string.s; | |
bc5a45f3 | 1715 | |
102ebb00 EZ |
1716 | if (bidi_it->nchars <= 0) |
1717 | abort (); | |
bc5a45f3 | 1718 | next_char = |
87e67904 EZ |
1719 | bidi_it->charpos + bidi_it->nchars >= eob |
1720 | ? BIDI_EOB | |
f3014ef5 EZ |
1721 | : bidi_char_at_pos (bidi_it->bytepos + bidi_it->ch_len, s, |
1722 | bidi_it->string.unibyte); | |
bc5a45f3 EZ |
1723 | type_of_next = bidi_get_type (next_char, override); |
1724 | ||
1725 | if (type_of_next == WEAK_ET | |
1726 | || type_of_next == WEAK_BN | |
1727 | || bidi_explicit_dir_char (next_char)) | |
1728 | { | |
1729 | bidi_copy_it (&saved_it, bidi_it); | |
1730 | while (bidi_resolve_explicit (bidi_it) == new_level | |
1731 | && (bidi_it->type == WEAK_BN | |
1732 | || bidi_it->type == WEAK_ET)) | |
1733 | ; | |
1734 | type_of_next = bidi_it->type; | |
1735 | en_pos = bidi_it->charpos; | |
1736 | bidi_copy_it (bidi_it, &saved_it); | |
1737 | } | |
1738 | if (type_of_next == WEAK_EN) | |
b7b65b15 | 1739 | { |
bc5a45f3 EZ |
1740 | /* If the last strong character is AL, the EN we've |
1741 | found will become AN when we get to it (W2). */ | |
1742 | if (bidi_it->last_strong.type_after_w1 != STRONG_AL) | |
1743 | { | |
1744 | type = WEAK_EN; | |
1745 | /* Remember this EN position, to speed up processing | |
1746 | of the next ETs. */ | |
1747 | bidi_it->next_en_pos = en_pos; | |
1748 | } | |
1749 | else if (type == WEAK_BN) | |
1750 | type = NEUTRAL_ON; /* W6/Retaining */ | |
b7b65b15 | 1751 | } |
b7b65b15 EZ |
1752 | } |
1753 | } | |
1754 | } | |
1755 | ||
1756 | if (type == WEAK_ES || type == WEAK_ET || type == WEAK_CS /* W6 */ | |
89d3374a EZ |
1757 | || (type == WEAK_BN |
1758 | && (bidi_it->prev.type_after_w1 == WEAK_CS /* W6/Retaining */ | |
1759 | || bidi_it->prev.type_after_w1 == WEAK_ES | |
1760 | || bidi_it->prev.type_after_w1 == WEAK_ET))) | |
b7b65b15 EZ |
1761 | type = NEUTRAL_ON; |
1762 | ||
1763 | /* Store the type we've got so far, before we clobber it with strong | |
1764 | types in W7 and while resolving neutral types. But leave alone | |
1765 | the original types that were recorded above, because we will need | |
1766 | them for the L1 clause. */ | |
89d3374a EZ |
1767 | if (bidi_it->type_after_w1 == UNKNOWN_BT) |
1768 | bidi_it->type_after_w1 = type; | |
1769 | bidi_check_type (bidi_it->type_after_w1); | |
b7b65b15 EZ |
1770 | |
1771 | if (type == WEAK_EN) /* W7 */ | |
1772 | { | |
89d3374a | 1773 | if ((bidi_it->last_strong.type_after_w1 == STRONG_L) |
b7b65b15 EZ |
1774 | || (bidi_it->last_strong.type == UNKNOWN_BT && bidi_it->sor == L2R)) |
1775 | type = STRONG_L; | |
1776 | } | |
1777 | ||
1778 | bidi_it->type = type; | |
2d6e4628 | 1779 | bidi_check_type (bidi_it->type); |
b7b65b15 EZ |
1780 | return type; |
1781 | } | |
1782 | ||
cbb09f04 EZ |
1783 | /* Resolve the type of a neutral character according to the type of |
1784 | surrounding strong text and the current embedding level. */ | |
58b9f433 | 1785 | static inline bidi_type_t |
cbb09f04 EZ |
1786 | bidi_resolve_neutral_1 (bidi_type_t prev_type, bidi_type_t next_type, int lev) |
1787 | { | |
1788 | /* N1: European and Arabic numbers are treated as though they were R. */ | |
1789 | if (next_type == WEAK_EN || next_type == WEAK_AN) | |
1790 | next_type = STRONG_R; | |
1791 | if (prev_type == WEAK_EN || prev_type == WEAK_AN) | |
1792 | prev_type = STRONG_R; | |
1793 | ||
1794 | if (next_type == prev_type) /* N1 */ | |
1795 | return next_type; | |
1796 | else if ((lev & 1) == 0) /* N2 */ | |
1797 | return STRONG_L; | |
1798 | else | |
1799 | return STRONG_R; | |
1800 | } | |
1801 | ||
fd3998ff | 1802 | static bidi_type_t |
b7b65b15 EZ |
1803 | bidi_resolve_neutral (struct bidi_it *bidi_it) |
1804 | { | |
1805 | int prev_level = bidi_it->level_stack[bidi_it->stack_idx].level; | |
1806 | bidi_type_t type = bidi_resolve_weak (bidi_it); | |
1807 | int current_level = bidi_it->level_stack[bidi_it->stack_idx].level; | |
1808 | ||
1809 | if (!(type == STRONG_R | |
1810 | || type == STRONG_L | |
1811 | || type == WEAK_BN | |
1812 | || type == WEAK_EN | |
1813 | || type == WEAK_AN | |
1814 | || type == NEUTRAL_B | |
1815 | || type == NEUTRAL_S | |
1816 | || type == NEUTRAL_WS | |
1817 | || type == NEUTRAL_ON)) | |
1818 | abort (); | |
1819 | ||
1820 | if (bidi_get_category (type) == NEUTRAL | |
1821 | || (type == WEAK_BN && prev_level == current_level)) | |
1822 | { | |
1823 | if (bidi_it->next_for_neutral.type != UNKNOWN_BT) | |
1824 | type = bidi_resolve_neutral_1 (bidi_it->prev_for_neutral.type, | |
1825 | bidi_it->next_for_neutral.type, | |
1826 | current_level); | |
1827 | else | |
1828 | { | |
1829 | /* Arrrgh!! The UAX#9 algorithm is too deeply entrenched in | |
1830 | the assumption of batch-style processing; see clauses W4, | |
1831 | W5, and especially N1, which require to look far forward | |
182ce2d2 EZ |
1832 | (as well as back) in the buffer/string. May the fleas of |
1833 | a thousand camels infest the armpits of those who design | |
b7b65b15 EZ |
1834 | supposedly general-purpose algorithms by looking at their |
1835 | own implementations, and fail to consider other possible | |
1836 | implementations! */ | |
1837 | struct bidi_it saved_it; | |
1838 | bidi_type_t next_type; | |
1839 | ||
1840 | if (bidi_it->scan_dir == -1) | |
1841 | abort (); | |
1842 | ||
1843 | bidi_copy_it (&saved_it, bidi_it); | |
1844 | /* Scan the text forward until we find the first non-neutral | |
1845 | character, and then use that to resolve the neutral we | |
1846 | are dealing with now. We also cache the scanned iterator | |
1847 | states, to salvage some of the effort later. */ | |
1848 | bidi_cache_iterator_state (bidi_it, 0); | |
1849 | do { | |
1850 | /* Record the info about the previous character, so that | |
1851 | it will be cached below with this state. */ | |
89d3374a | 1852 | if (bidi_it->type_after_w1 != WEAK_BN /* W1/Retaining */ |
b7b65b15 EZ |
1853 | && bidi_it->type != WEAK_BN) |
1854 | bidi_remember_char (&bidi_it->prev, bidi_it); | |
1855 | type = bidi_resolve_weak (bidi_it); | |
1856 | /* Paragraph separators have their levels fully resolved | |
1857 | at this point, so cache them as resolved. */ | |
1858 | bidi_cache_iterator_state (bidi_it, type == NEUTRAL_B); | |
1859 | /* FIXME: implement L1 here, by testing for a newline and | |
1860 | resetting the level for any sequence of whitespace | |
1861 | characters adjacent to it. */ | |
1862 | } while (!(type == NEUTRAL_B | |
1863 | || (type != WEAK_BN | |
1864 | && bidi_get_category (type) != NEUTRAL) | |
1865 | /* This is all per level run, so stop when we | |
1866 | reach the end of this level run. */ | |
1867 | || bidi_it->level_stack[bidi_it->stack_idx].level != | |
1868 | current_level)); | |
1869 | ||
1870 | bidi_remember_char (&saved_it.next_for_neutral, bidi_it); | |
1871 | ||
1872 | switch (type) | |
1873 | { | |
1874 | case STRONG_L: | |
1875 | case STRONG_R: | |
1876 | case STRONG_AL: | |
1877 | next_type = type; | |
1878 | break; | |
1879 | case WEAK_EN: | |
1880 | case WEAK_AN: | |
1881 | /* N1: ``European and Arabic numbers are treated as | |
1882 | though they were R.'' */ | |
1883 | next_type = STRONG_R; | |
1884 | saved_it.next_for_neutral.type = STRONG_R; | |
1885 | break; | |
1886 | case WEAK_BN: | |
1887 | if (!bidi_explicit_dir_char (bidi_it->ch)) | |
1888 | abort (); /* can't happen: BNs are skipped */ | |
1889 | /* FALLTHROUGH */ | |
1890 | case NEUTRAL_B: | |
1891 | /* Marched all the way to the end of this level run. | |
1892 | We need to use the eor type, whose information is | |
1893 | stored by bidi_set_sor_type in the prev_for_neutral | |
1894 | member. */ | |
1895 | if (saved_it.type != WEAK_BN | |
89d3374a | 1896 | || bidi_get_category (bidi_it->prev.type_after_w1) == NEUTRAL) |
b7b65b15 EZ |
1897 | { |
1898 | next_type = bidi_it->prev_for_neutral.type; | |
1899 | saved_it.next_for_neutral.type = next_type; | |
2d6e4628 | 1900 | bidi_check_type (next_type); |
b7b65b15 EZ |
1901 | } |
1902 | else | |
1903 | { | |
1904 | /* This is a BN which does not adjoin neutrals. | |
1905 | Leave its type alone. */ | |
1906 | bidi_copy_it (bidi_it, &saved_it); | |
1907 | return bidi_it->type; | |
1908 | } | |
1909 | break; | |
1910 | default: | |
1911 | abort (); | |
1912 | } | |
1913 | type = bidi_resolve_neutral_1 (saved_it.prev_for_neutral.type, | |
1914 | next_type, current_level); | |
1915 | saved_it.type = type; | |
2d6e4628 | 1916 | bidi_check_type (type); |
b7b65b15 EZ |
1917 | bidi_copy_it (bidi_it, &saved_it); |
1918 | } | |
1919 | } | |
1920 | return type; | |
1921 | } | |
1922 | ||
1923 | /* Given an iterator state in BIDI_IT, advance one character position | |
182ce2d2 EZ |
1924 | in the buffer/string to the next character (in the logical order), |
1925 | resolve the bidi type of that next character, and return that | |
1926 | type. */ | |
fd3998ff | 1927 | static bidi_type_t |
b7b65b15 EZ |
1928 | bidi_type_of_next_char (struct bidi_it *bidi_it) |
1929 | { | |
1930 | bidi_type_t type; | |
1931 | ||
1932 | /* This should always be called during a forward scan. */ | |
1933 | if (bidi_it->scan_dir != 1) | |
1934 | abort (); | |
1935 | ||
1936 | /* Reset the limit until which to ignore BNs if we step out of the | |
1937 | area where we found only empty levels. */ | |
87e67904 | 1938 | if ((bidi_it->ignore_bn_limit > -1 |
b7b65b15 | 1939 | && bidi_it->ignore_bn_limit <= bidi_it->charpos) |
87e67904 | 1940 | || (bidi_it->ignore_bn_limit == -2 |
b7b65b15 | 1941 | && !bidi_explicit_dir_char (bidi_it->ch))) |
87e67904 | 1942 | bidi_it->ignore_bn_limit = -1; |
b7b65b15 EZ |
1943 | |
1944 | type = bidi_resolve_neutral (bidi_it); | |
1945 | ||
1946 | return type; | |
1947 | } | |
1948 | ||
1949 | /* Given an iterator state BIDI_IT, advance one character position in | |
182ce2d2 EZ |
1950 | the buffer/string to the next character (in the current scan |
1951 | direction), resolve the embedding and implicit levels of that next | |
1952 | character, and return the resulting level. */ | |
fd3998ff | 1953 | static int |
b7b65b15 EZ |
1954 | bidi_level_of_next_char (struct bidi_it *bidi_it) |
1955 | { | |
1956 | bidi_type_t type; | |
1957 | int level, prev_level = -1; | |
1958 | struct bidi_saved_info next_for_neutral; | |
578b494e | 1959 | EMACS_INT next_char_pos = -2; |
b7b65b15 EZ |
1960 | |
1961 | if (bidi_it->scan_dir == 1) | |
1962 | { | |
9f257352 EZ |
1963 | EMACS_INT eob = |
1964 | (bidi_it->string.s || STRINGP (bidi_it->string.lstring)) | |
1965 | ? bidi_it->string.schars : ZV; | |
1966 | ||
b7b65b15 | 1967 | /* There's no sense in trying to advance if we hit end of text. */ |
9f257352 | 1968 | if (bidi_it->charpos >= eob) |
b7b65b15 EZ |
1969 | return bidi_it->resolved_level; |
1970 | ||
1971 | /* Record the info about the previous character. */ | |
89d3374a | 1972 | if (bidi_it->type_after_w1 != WEAK_BN /* W1/Retaining */ |
b7b65b15 EZ |
1973 | && bidi_it->type != WEAK_BN) |
1974 | bidi_remember_char (&bidi_it->prev, bidi_it); | |
89d3374a EZ |
1975 | if (bidi_it->type_after_w1 == STRONG_R |
1976 | || bidi_it->type_after_w1 == STRONG_L | |
1977 | || bidi_it->type_after_w1 == STRONG_AL) | |
b7b65b15 EZ |
1978 | bidi_remember_char (&bidi_it->last_strong, bidi_it); |
1979 | /* FIXME: it sounds like we don't need both prev and | |
1980 | prev_for_neutral members, but I'm leaving them both for now. */ | |
1981 | if (bidi_it->type == STRONG_R || bidi_it->type == STRONG_L | |
1982 | || bidi_it->type == WEAK_EN || bidi_it->type == WEAK_AN) | |
1983 | bidi_remember_char (&bidi_it->prev_for_neutral, bidi_it); | |
1984 | ||
1985 | /* If we overstepped the characters used for resolving neutrals | |
1986 | and whitespace, invalidate their info in the iterator. */ | |
1987 | if (bidi_it->charpos >= bidi_it->next_for_neutral.charpos) | |
1988 | bidi_it->next_for_neutral.type = UNKNOWN_BT; | |
1989 | if (bidi_it->next_en_pos >= 0 | |
1990 | && bidi_it->charpos >= bidi_it->next_en_pos) | |
1991 | bidi_it->next_en_pos = -1; | |
1992 | if (bidi_it->next_for_ws.type != UNKNOWN_BT | |
1993 | && bidi_it->charpos >= bidi_it->next_for_ws.charpos) | |
1994 | bidi_it->next_for_ws.type = UNKNOWN_BT; | |
1995 | ||
1996 | /* This must be taken before we fill the iterator with the info | |
1997 | about the next char. If we scan backwards, the iterator | |
1998 | state must be already cached, so there's no need to know the | |
1999 | embedding level of the previous character, since we will be | |
2000 | returning to our caller shortly. */ | |
2001 | prev_level = bidi_it->level_stack[bidi_it->stack_idx].level; | |
2002 | } | |
2003 | next_for_neutral = bidi_it->next_for_neutral; | |
2004 | ||
182ce2d2 EZ |
2005 | /* Perhaps the character we want is already cached. If it is, the |
2006 | call to bidi_cache_find below will return a type other than | |
2007 | UNKNOWN_BT. */ | |
87e67904 | 2008 | if (bidi_cache_idx > bidi_cache_start && !bidi_it->first_elt) |
102ebb00 | 2009 | { |
9f257352 EZ |
2010 | int bob = |
2011 | (bidi_it->string.s || STRINGP (bidi_it->string.lstring)) ? 0 : 1; | |
2012 | ||
102ebb00 EZ |
2013 | if (bidi_it->scan_dir > 0) |
2014 | { | |
2015 | if (bidi_it->nchars <= 0) | |
2016 | abort (); | |
2017 | next_char_pos = bidi_it->charpos + bidi_it->nchars; | |
2018 | } | |
9f257352 | 2019 | else if (bidi_it->charpos >= bob) |
bb269206 EZ |
2020 | /* Implementation note: we allow next_char_pos to be as low as |
2021 | 0 for buffers or -1 for strings, and that is okay because | |
2022 | that's the "position" of the sentinel iterator state we | |
2023 | cached at the beginning of the iteration. */ | |
102ebb00 | 2024 | next_char_pos = bidi_it->charpos - 1; |
578b494e | 2025 | if (next_char_pos >= bob - 1) |
87e67904 EZ |
2026 | type = bidi_cache_find (next_char_pos, -1, bidi_it); |
2027 | else | |
2028 | type = UNKNOWN_BT; | |
102ebb00 | 2029 | } |
182ce2d2 | 2030 | else |
102ebb00 | 2031 | type = UNKNOWN_BT; |
b7b65b15 EZ |
2032 | if (type != UNKNOWN_BT) |
2033 | { | |
2034 | /* Don't lose the information for resolving neutrals! The | |
2035 | cached states could have been cached before their | |
2036 | next_for_neutral member was computed. If we are on our way | |
2037 | forward, we can simply take the info from the previous | |
2038 | state. */ | |
2039 | if (bidi_it->scan_dir == 1 | |
2040 | && bidi_it->next_for_neutral.type == UNKNOWN_BT) | |
2041 | bidi_it->next_for_neutral = next_for_neutral; | |
2042 | ||
2043 | /* If resolved_level is -1, it means this state was cached | |
2044 | before it was completely resolved, so we cannot return | |
2045 | it. */ | |
2046 | if (bidi_it->resolved_level != -1) | |
2047 | return bidi_it->resolved_level; | |
2048 | } | |
2049 | if (bidi_it->scan_dir == -1) | |
2050 | /* If we are going backwards, the iterator state is already cached | |
2051 | from previous scans, and should be fully resolved. */ | |
2052 | abort (); | |
2053 | ||
2054 | if (type == UNKNOWN_BT) | |
2055 | type = bidi_type_of_next_char (bidi_it); | |
2056 | ||
2057 | if (type == NEUTRAL_B) | |
2058 | return bidi_it->resolved_level; | |
2059 | ||
2060 | level = bidi_it->level_stack[bidi_it->stack_idx].level; | |
2061 | if ((bidi_get_category (type) == NEUTRAL /* && type != NEUTRAL_B */) | |
2062 | || (type == WEAK_BN && prev_level == level)) | |
2063 | { | |
2064 | if (bidi_it->next_for_neutral.type == UNKNOWN_BT) | |
2065 | abort (); | |
2066 | ||
2067 | /* If the cached state shows a neutral character, it was not | |
2068 | resolved by bidi_resolve_neutral, so do it now. */ | |
2069 | type = bidi_resolve_neutral_1 (bidi_it->prev_for_neutral.type, | |
2070 | bidi_it->next_for_neutral.type, | |
2071 | level); | |
2072 | } | |
2073 | ||
2074 | if (!(type == STRONG_R | |
2075 | || type == STRONG_L | |
2076 | || type == WEAK_BN | |
2077 | || type == WEAK_EN | |
2078 | || type == WEAK_AN)) | |
2079 | abort (); | |
2080 | bidi_it->type = type; | |
2d6e4628 | 2081 | bidi_check_type (bidi_it->type); |
b7b65b15 EZ |
2082 | |
2083 | /* For L1 below, we need to know, for each WS character, whether | |
0d327994 | 2084 | it belongs to a sequence of WS characters preceding a newline |
b7b65b15 | 2085 | or a TAB or a paragraph separator. */ |
89d3374a | 2086 | if (bidi_it->orig_type == NEUTRAL_WS |
b7b65b15 EZ |
2087 | && bidi_it->next_for_ws.type == UNKNOWN_BT) |
2088 | { | |
2089 | int ch; | |
8264569d | 2090 | EMACS_INT clen = bidi_it->ch_len; |
6bff6497 EZ |
2091 | EMACS_INT bpos = bidi_it->bytepos; |
2092 | EMACS_INT cpos = bidi_it->charpos; | |
182ce2d2 | 2093 | EMACS_INT disp_pos = bidi_it->disp_pos; |
102ebb00 | 2094 | EMACS_INT nc = bidi_it->nchars; |
87e67904 | 2095 | struct bidi_string_data bs = bidi_it->string; |
b7b65b15 | 2096 | bidi_type_t chtype; |
fc6f18ce | 2097 | int fwp = bidi_it->frame_window_p; |
0c95fcf7 | 2098 | int dpp = bidi_it->disp_prop; |
b7b65b15 | 2099 | |
102ebb00 EZ |
2100 | if (bidi_it->nchars <= 0) |
2101 | abort (); | |
b7b65b15 | 2102 | do { |
55439c61 EZ |
2103 | ch = bidi_fetch_char (bpos += clen, cpos += nc, &disp_pos, &dpp, &bs, |
2104 | fwp, &clen, &nc); | |
e7402cb2 | 2105 | if (ch == '\n' || ch == BIDI_EOB /* || ch == LINESEP_CHAR */) |
b7b65b15 EZ |
2106 | chtype = NEUTRAL_B; |
2107 | else | |
6bff6497 | 2108 | chtype = bidi_get_type (ch, NEUTRAL_DIR); |
b7b65b15 EZ |
2109 | } while (chtype == NEUTRAL_WS || chtype == WEAK_BN |
2110 | || bidi_explicit_dir_char (ch)); /* L1/Retaining */ | |
2111 | bidi_it->next_for_ws.type = chtype; | |
2d6e4628 | 2112 | bidi_check_type (bidi_it->next_for_ws.type); |
b7b65b15 EZ |
2113 | bidi_it->next_for_ws.charpos = cpos; |
2114 | bidi_it->next_for_ws.bytepos = bpos; | |
2115 | } | |
2116 | ||
2117 | /* Resolve implicit levels, with a twist: PDFs get the embedding | |
2118 | level of the enbedding they terminate. See below for the | |
2119 | reason. */ | |
89d3374a | 2120 | if (bidi_it->orig_type == PDF |
b7b65b15 EZ |
2121 | /* Don't do this if this formatting code didn't change the |
2122 | embedding level due to invalid or empty embeddings. */ | |
2123 | && prev_level != level) | |
2124 | { | |
2125 | /* Don't look in UAX#9 for the reason for this: it's our own | |
2126 | private quirk. The reason is that we want the formatting | |
2127 | codes to be delivered so that they bracket the text of their | |
2128 | embedding. For example, given the text | |
2129 | ||
2130 | {RLO}teST{PDF} | |
2131 | ||
2132 | we want it to be displayed as | |
2133 | ||
0d68907d | 2134 | {PDF}STet{RLO} |
b7b65b15 EZ |
2135 | |
2136 | not as | |
2137 | ||
2138 | STet{RLO}{PDF} | |
2139 | ||
2140 | which will result because we bump up the embedding level as | |
2141 | soon as we see the RLO and pop it as soon as we see the PDF, | |
2142 | so RLO itself has the same embedding level as "teST", and | |
2143 | thus would be normally delivered last, just before the PDF. | |
2144 | The switch below fiddles with the level of PDF so that this | |
2145 | ugly side effect does not happen. | |
2146 | ||
2147 | (This is, of course, only important if the formatting codes | |
e7402cb2 EZ |
2148 | are actually displayed, but Emacs does need to display them |
2149 | if the user wants to.) */ | |
b7b65b15 EZ |
2150 | level = prev_level; |
2151 | } | |
89d3374a EZ |
2152 | else if (bidi_it->orig_type == NEUTRAL_B /* L1 */ |
2153 | || bidi_it->orig_type == NEUTRAL_S | |
e7402cb2 EZ |
2154 | || bidi_it->ch == '\n' || bidi_it->ch == BIDI_EOB |
2155 | /* || bidi_it->ch == LINESEP_CHAR */ | |
89d3374a | 2156 | || (bidi_it->orig_type == NEUTRAL_WS |
b7b65b15 EZ |
2157 | && (bidi_it->next_for_ws.type == NEUTRAL_B |
2158 | || bidi_it->next_for_ws.type == NEUTRAL_S))) | |
2159 | level = bidi_it->level_stack[0].level; | |
2160 | else if ((level & 1) == 0) /* I1 */ | |
2161 | { | |
2162 | if (type == STRONG_R) | |
2163 | level++; | |
2164 | else if (type == WEAK_EN || type == WEAK_AN) | |
2165 | level += 2; | |
2166 | } | |
2167 | else /* I2 */ | |
2168 | { | |
2169 | if (type == STRONG_L || type == WEAK_EN || type == WEAK_AN) | |
2170 | level++; | |
2171 | } | |
2172 | ||
2173 | bidi_it->resolved_level = level; | |
2174 | return level; | |
2175 | } | |
2176 | ||
2177 | /* Move to the other edge of a level given by LEVEL. If END_FLAG is | |
2178 | non-zero, we are at the end of a level, and we need to prepare to | |
2179 | resume the scan of the lower level. | |
2180 | ||
2181 | If this level's other edge is cached, we simply jump to it, filling | |
2182 | the iterator structure with the iterator state on the other edge. | |
182ce2d2 EZ |
2183 | Otherwise, we walk the buffer or string until we come back to the |
2184 | same level as LEVEL. | |
b7b65b15 EZ |
2185 | |
2186 | Note: we are not talking here about a ``level run'' in the UAX#9 | |
2187 | sense of the term, but rather about a ``level'' which includes | |
2188 | all the levels higher than it. In other words, given the levels | |
2189 | like this: | |
2190 | ||
2191 | 11111112222222333333334443343222222111111112223322111 | |
2192 | A B C | |
2193 | ||
2194 | and assuming we are at point A scanning left to right, this | |
2195 | function moves to point C, whereas the UAX#9 ``level 2 run'' ends | |
2196 | at point B. */ | |
2197 | static void | |
2198 | bidi_find_other_level_edge (struct bidi_it *bidi_it, int level, int end_flag) | |
2199 | { | |
2200 | int dir = end_flag ? -bidi_it->scan_dir : bidi_it->scan_dir; | |
39e378da | 2201 | ptrdiff_t idx; |
b7b65b15 EZ |
2202 | |
2203 | /* Try the cache first. */ | |
87e67904 EZ |
2204 | if ((idx = bidi_cache_find_level_change (level, dir, end_flag)) |
2205 | >= bidi_cache_start) | |
b7b65b15 EZ |
2206 | bidi_cache_fetch_state (idx, bidi_it); |
2207 | else | |
2208 | { | |
2209 | int new_level; | |
2210 | ||
2211 | if (end_flag) | |
2212 | abort (); /* if we are at end of level, its edges must be cached */ | |
2213 | ||
2214 | bidi_cache_iterator_state (bidi_it, 1); | |
2215 | do { | |
2216 | new_level = bidi_level_of_next_char (bidi_it); | |
2217 | bidi_cache_iterator_state (bidi_it, 1); | |
2218 | } while (new_level >= level); | |
2219 | } | |
2220 | } | |
2221 | ||
2222 | void | |
4b292a22 | 2223 | bidi_move_to_visually_next (struct bidi_it *bidi_it) |
b7b65b15 EZ |
2224 | { |
2225 | int old_level, new_level, next_level; | |
9c82e145 | 2226 | struct bidi_it sentinel; |
acb28818 | 2227 | struct gcpro gcpro1; |
b7b65b15 | 2228 | |
87e67904 EZ |
2229 | if (bidi_it->charpos < 0 || bidi_it->bytepos < 0) |
2230 | abort (); | |
b7b65b15 EZ |
2231 | |
2232 | if (bidi_it->scan_dir == 0) | |
2233 | { | |
2234 | bidi_it->scan_dir = 1; /* default to logical order */ | |
2235 | } | |
2236 | ||
acb28818 | 2237 | /* The code below can call eval, and thus cause GC. If we are |
7e2ad32c | 2238 | iterating a Lisp string, make sure it won't be GCed. */ |
acb28818 EZ |
2239 | if (STRINGP (bidi_it->string.lstring)) |
2240 | GCPRO1 (bidi_it->string.lstring); | |
2241 | ||
be39f003 EZ |
2242 | /* If we just passed a newline, initialize for the next line. */ |
2243 | if (!bidi_it->first_elt && bidi_it->orig_type == NEUTRAL_B) | |
2244 | bidi_line_init (bidi_it); | |
2245 | ||
6dcfd253 EZ |
2246 | /* Prepare the sentinel iterator state, and cache it. When we bump |
2247 | into it, scanning backwards, we'll know that the last non-base | |
2248 | level is exhausted. */ | |
87e67904 | 2249 | if (bidi_cache_idx == bidi_cache_start) |
9c82e145 EZ |
2250 | { |
2251 | bidi_copy_it (&sentinel, bidi_it); | |
2252 | if (bidi_it->first_elt) | |
2253 | { | |
2254 | sentinel.charpos--; /* cached charpos needs to be monotonic */ | |
2255 | sentinel.bytepos--; | |
2256 | sentinel.ch = '\n'; /* doesn't matter, but why not? */ | |
2257 | sentinel.ch_len = 1; | |
182ce2d2 | 2258 | sentinel.nchars = 1; |
9c82e145 | 2259 | } |
6dcfd253 | 2260 | bidi_cache_iterator_state (&sentinel, 1); |
9c82e145 | 2261 | } |
b7b65b15 EZ |
2262 | |
2263 | old_level = bidi_it->resolved_level; | |
2264 | new_level = bidi_level_of_next_char (bidi_it); | |
b7b65b15 EZ |
2265 | |
2266 | /* Reordering of resolved levels (clause L2) is implemented by | |
2267 | jumping to the other edge of the level and flipping direction of | |
c0546589 | 2268 | scanning the text whenever we find a level change. */ |
b7b65b15 EZ |
2269 | if (new_level != old_level) |
2270 | { | |
2271 | int ascending = new_level > old_level; | |
2272 | int level_to_search = ascending ? old_level + 1 : old_level; | |
2273 | int incr = ascending ? 1 : -1; | |
2274 | int expected_next_level = old_level + incr; | |
2275 | ||
b7b65b15 EZ |
2276 | /* Jump (or walk) to the other edge of this level. */ |
2277 | bidi_find_other_level_edge (bidi_it, level_to_search, !ascending); | |
2278 | /* Switch scan direction and peek at the next character in the | |
2279 | new direction. */ | |
2280 | bidi_it->scan_dir = -bidi_it->scan_dir; | |
2281 | ||
2282 | /* The following loop handles the case where the resolved level | |
2283 | jumps by more than one. This is typical for numbers inside a | |
2284 | run of text with left-to-right embedding direction, but can | |
2285 | also happen in other situations. In those cases the decision | |
2286 | where to continue after a level change, and in what direction, | |
2287 | is tricky. For example, given a text like below: | |
2288 | ||
2289 | abcdefgh | |
2290 | 11336622 | |
2291 | ||
2292 | (where the numbers below the text show the resolved levels), | |
2293 | the result of reordering according to UAX#9 should be this: | |
2294 | ||
2295 | efdcghba | |
2296 | ||
2297 | This is implemented by the loop below which flips direction | |
2298 | and jumps to the other edge of the level each time it finds | |
2299 | the new level not to be the expected one. The expected level | |
2300 | is always one more or one less than the previous one. */ | |
2301 | next_level = bidi_peek_at_next_level (bidi_it); | |
2302 | while (next_level != expected_next_level) | |
2303 | { | |
2304 | expected_next_level += incr; | |
2305 | level_to_search += incr; | |
2306 | bidi_find_other_level_edge (bidi_it, level_to_search, !ascending); | |
2307 | bidi_it->scan_dir = -bidi_it->scan_dir; | |
2308 | next_level = bidi_peek_at_next_level (bidi_it); | |
2309 | } | |
2310 | ||
2311 | /* Finally, deliver the next character in the new direction. */ | |
2312 | next_level = bidi_level_of_next_char (bidi_it); | |
2313 | } | |
2314 | ||
b44d9321 EZ |
2315 | /* Take note when we have just processed the newline that precedes |
2316 | the end of the paragraph. The next time we are about to be | |
2317 | called, set_iterator_to_next will automatically reinit the | |
2318 | paragraph direction, if needed. We do this at the newline before | |
2319 | the paragraph separator, because the next character might not be | |
2320 | the first character of the next paragraph, due to the bidi | |
c0546589 EZ |
2321 | reordering, whereas we _must_ know the paragraph base direction |
2322 | _before_ we process the paragraph's text, since the base | |
2323 | direction affects the reordering. */ | |
87e67904 | 2324 | if (bidi_it->scan_dir == 1 && bidi_it->orig_type == NEUTRAL_B) |
be39f003 | 2325 | { |
87e67904 EZ |
2326 | /* The paragraph direction of the entire string, once |
2327 | determined, is in effect for the entire string. Setting the | |
2328 | separator limit to the end of the string prevents | |
2329 | bidi_paragraph_init from being called automatically on this | |
2330 | string. */ | |
9f257352 | 2331 | if (bidi_it->string.s || STRINGP (bidi_it->string.lstring)) |
87e67904 EZ |
2332 | bidi_it->separator_limit = bidi_it->string.schars; |
2333 | else if (bidi_it->bytepos < ZV_BYTE) | |
be39f003 | 2334 | { |
87e67904 EZ |
2335 | EMACS_INT sep_len = |
2336 | bidi_at_paragraph_end (bidi_it->charpos + bidi_it->nchars, | |
2337 | bidi_it->bytepos + bidi_it->ch_len); | |
2338 | if (bidi_it->nchars <= 0) | |
2339 | abort (); | |
2340 | if (sep_len >= 0) | |
2341 | { | |
2342 | bidi_it->new_paragraph = 1; | |
2343 | /* Record the buffer position of the last character of the | |
2344 | paragraph separator. */ | |
2345 | bidi_it->separator_limit = | |
2346 | bidi_it->charpos + bidi_it->nchars + sep_len; | |
2347 | } | |
be39f003 EZ |
2348 | } |
2349 | } | |
6bff6497 | 2350 | |
87e67904 | 2351 | if (bidi_it->scan_dir == 1 && bidi_cache_idx > bidi_cache_start) |
b7b65b15 EZ |
2352 | { |
2353 | /* If we are at paragraph's base embedding level and beyond the | |
2354 | last cached position, the cache's job is done and we can | |
2355 | discard it. */ | |
2356 | if (bidi_it->resolved_level == bidi_it->level_stack[0].level | |
182ce2d2 EZ |
2357 | && bidi_it->charpos > (bidi_cache[bidi_cache_idx - 1].charpos |
2358 | + bidi_cache[bidi_cache_idx - 1].nchars - 1)) | |
b7b65b15 EZ |
2359 | bidi_cache_reset (); |
2360 | /* But as long as we are caching during forward scan, we must | |
2361 | cache each state, or else the cache integrity will be | |
2362 | compromised: it assumes cached states correspond to buffer | |
2363 | positions 1:1. */ | |
2364 | else | |
2365 | bidi_cache_iterator_state (bidi_it, 1); | |
2366 | } | |
acb28818 EZ |
2367 | |
2368 | if (STRINGP (bidi_it->string.lstring)) | |
2369 | UNGCPRO; | |
b7b65b15 EZ |
2370 | } |
2371 | ||
2372 | /* This is meant to be called from within the debugger, whenever you | |
2373 | wish to examine the cache contents. */ | |
e78aecca | 2374 | void bidi_dump_cached_states (void) EXTERNALLY_VISIBLE; |
b7b65b15 EZ |
2375 | void |
2376 | bidi_dump_cached_states (void) | |
2377 | { | |
39e378da | 2378 | ptrdiff_t i; |
b7b65b15 EZ |
2379 | int ndigits = 1; |
2380 | ||
2381 | if (bidi_cache_idx == 0) | |
2382 | { | |
2383 | fprintf (stderr, "The cache is empty.\n"); | |
2384 | return; | |
2385 | } | |
df9733bf | 2386 | fprintf (stderr, "Total of %"pD"d state%s in cache:\n", |
b7b65b15 EZ |
2387 | bidi_cache_idx, bidi_cache_idx == 1 ? "" : "s"); |
2388 | ||
2389 | for (i = bidi_cache[bidi_cache_idx - 1].charpos; i > 0; i /= 10) | |
2390 | ndigits++; | |
2391 | fputs ("ch ", stderr); | |
2392 | for (i = 0; i < bidi_cache_idx; i++) | |
2393 | fprintf (stderr, "%*c", ndigits, bidi_cache[i].ch); | |
2394 | fputs ("\n", stderr); | |
2395 | fputs ("lvl ", stderr); | |
2396 | for (i = 0; i < bidi_cache_idx; i++) | |
2397 | fprintf (stderr, "%*d", ndigits, bidi_cache[i].resolved_level); | |
2398 | fputs ("\n", stderr); | |
2399 | fputs ("pos ", stderr); | |
2400 | for (i = 0; i < bidi_cache_idx; i++) | |
c2982e87 | 2401 | fprintf (stderr, "%*"pI"d", ndigits, bidi_cache[i].charpos); |
b7b65b15 EZ |
2402 | fputs ("\n", stderr); |
2403 | } |