Commit | Line | Data |
---|---|---|
cbb09f04 | 1 | /* Low-level bidirectional buffer/string-scanning functions for GNU Emacs. |
ba318903 | 2 | Copyright (C) 2000-2001, 2004-2005, 2009-2014 Free Software |
ab422c4d | 3 | 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 | 24 | |
aa5ccb01 EZ |
25 | Unlike the Reference Implementation and most other implementations, |
26 | this one is designed to be called once for every character in the | |
27 | buffer or string. That way, we can leave intact the design of the | |
28 | Emacs display engine, whereby an iterator object is used to | |
29 | traverse buffer or string text character by character, and generate | |
30 | the necessary data for displaying each character in 'struct glyph' | |
31 | objects. (See xdisp.c for the details of that iteration.) The | |
32 | functions on this file replace the original linear iteration in the | |
33 | logical order of the text with a non-linear iteration in the visual | |
34 | order, i.e. in the order characters should be shown on display. | |
b7b65b15 | 35 | |
4b292a22 | 36 | The main entry point is bidi_move_to_visually_next. Each time it |
b7b65b15 EZ |
37 | is called, it finds the next character in the visual order, and |
38 | returns its information in a special structure. The caller is then | |
39 | expected to process this character for display or any other | |
4b292a22 EZ |
40 | purposes, and call bidi_move_to_visually_next for the next |
41 | character. See the comments in bidi_move_to_visually_next for more | |
42 | details about its algorithm that finds the next visual-order | |
b7b65b15 EZ |
43 | character by resolving their levels on the fly. |
44 | ||
95e1afc6 | 45 | Two other entry points are bidi_paragraph_init and |
940afb59 EZ |
46 | bidi_mirror_char. The first determines the base direction of a |
47 | paragraph, while the second returns the mirrored version of its | |
48 | argument character. | |
49 | ||
95e1afc6 EZ |
50 | A few auxiliary entry points are used to initialize the bidi |
51 | iterator for iterating an object (buffer or string), push and pop | |
52 | the bidi iterator state, and save and restore the state of the bidi | |
53 | cache. | |
54 | ||
89d3374a EZ |
55 | If you want to understand the code, you will have to read it |
56 | together with the relevant portions of UAX#9. The comments include | |
57 | references to UAX#9 rules, for that very reason. | |
58 | ||
b7b65b15 EZ |
59 | A note about references to UAX#9 rules: if the reference says |
60 | something like "X9/Retaining", it means that you need to refer to | |
cd1181db | 61 | rule X9 and to its modifications described in the "Implementation |
aa5ccb01 EZ |
62 | Notes" section of UAX#9, under "Retaining Format Codes". |
63 | ||
64 | Here's the overview of the design of the reordering engine | |
65 | implemented by this file. | |
66 | ||
67 | Basic implementation structure | |
68 | ------------------------------ | |
69 | ||
70 | The sequential processing steps described by UAX#9 are implemented | |
71 | as recursive levels of processing, all of which examine the next | |
72 | character in the logical order. This hierarchy of processing looks | |
73 | as follows, from the innermost (deepest) to the outermost level, | |
74 | omitting some subroutines used by each level: | |
75 | ||
76 | bidi_fetch_char -- fetch next character | |
77 | bidi_resolve_explicit -- resolve explicit levels and directions | |
78 | bidi_resolve_weak -- resolve weak types | |
79 | bidi_resolve_neutral -- resolve neutral types | |
80 | bidi_level_of_next_char -- resolve implicit levels | |
81 | ||
82 | Each level calls the level below it, and works on the result | |
83 | returned by the lower level, including all of its sub-levels. | |
84 | ||
85 | Unlike all the levels below it, bidi_level_of_next_char can return | |
86 | the information about either the next or previous character in the | |
87 | logical order, depending on the current direction of scanning the | |
88 | buffer or string. For the next character, it calls all the levels | |
89 | below it; for the previous character, it uses the cache, described | |
90 | below. | |
91 | ||
92 | Thus, the result of calling bidi_level_of_next_char is the resolved | |
93 | level of the next or the previous character in the logical order. | |
94 | Based on this information, the function bidi_move_to_visually_next | |
95 | finds the next character in the visual order and updates the | |
96 | direction in which the buffer is scanned, either forward or | |
97 | backward, to find the next character to be displayed. (Text is | |
98 | scanned backwards when it needs to be reversed for display, i.e. if | |
99 | the visual order is the inverse of the logical order.) This | |
100 | implements the last, reordering steps of the UBA, by successively | |
101 | calling bidi_level_of_next_char until the character of the required | |
102 | embedding level is found; the scan direction is dynamically updated | |
103 | as a side effect. See the commentary before the 'while' loop in | |
104 | bidi_move_to_visually_next, for the details. | |
105 | ||
106 | Fetching characters | |
107 | ------------------- | |
108 | ||
109 | In a nutshell, fetching the next character boils down to calling | |
110 | STRING_CHAR_AND_LENGTH, passing it the address of a buffer or | |
111 | string position. See bidi_fetch_char. However, if the next | |
112 | character is "covered" by a display property of some kind, | |
113 | bidi_fetch_char returns the u+FFFC "object replacement character" | |
114 | that represents the entire run of text covered by the display | |
115 | property. (The ch_len and nchars members of 'struct bidi_it' | |
116 | reflect the length in bytes and characters of that text.) This is | |
117 | so we reorder text on both sides of the display property as | |
118 | appropriate for an image or embedded string. Similarly, text | |
119 | covered by a display spec of the form '(space ...)', is replaced | |
120 | with the u+2029 paragraph separator character, so such display | |
121 | specs produce the same effect as a TAB under UBA. Both these | |
122 | special characters are not actually displayed -- the display | |
123 | property is displayed instead -- but just used to compute the | |
124 | embedding level of the surrounding text so as to produce the | |
125 | required effect. | |
126 | ||
127 | Bidi iterator states | |
128 | -------------------- | |
129 | ||
130 | The UBA is highly context dependent in some of its parts, | |
131 | i.e. results of processing a character can generally depend on | |
132 | characters very far away. The UAX#9 description of the UBA | |
133 | prescribes a stateful processing of each character, whereby the | |
134 | results of this processing depend on various state variables, such | |
135 | as the current embedding level, level stack, and directional | |
136 | override status. In addition, the UAX#9 description includes many | |
137 | passages like this (from rule W2 in this case): | |
138 | ||
139 | Search backward from each instance of a European number until the | |
140 | first strong type (R, L, AL, or sos) is found. If an AL is found, | |
141 | change the type of the European number to Arabic number. | |
142 | ||
143 | To support this, we use a bidi iterator object, 'struct bidi_it', | |
144 | which is a sub-structure of 'struct it' used by xdisp.c (see | |
145 | dispextern.h for the definition of both of these structures). The | |
146 | bidi iterator holds the entire state of the iteration required by | |
147 | the UBA, and is updated as the text is traversed. In particular, | |
148 | the embedding level of the current character being resolved is | |
149 | recorded in the iterator state. To avoid costly searches backward | |
150 | in support of rules like W2 above, the necessary character types | |
151 | are also recorded in the iterator state as they are found during | |
152 | the forward scan, and then used when such rules need to be applied. | |
153 | (Forward scans cannot be avoided in this way; they need to be | |
154 | performed at least once, and the results recorded in the iterator | |
155 | state, to be reused until the forward scan oversteps the recorded | |
156 | position.) | |
157 | ||
158 | In this manner, the iterator state acts as a mini-cache of | |
159 | contextual information required for resolving the level of the | |
160 | current character by various UBA rules. | |
161 | ||
162 | Caching of bidi iterator states | |
163 | ------------------------------- | |
164 | ||
165 | As described above, the reordering engine uses the information | |
166 | recorded in the bidi iterator state in order to resolve the | |
167 | embedding level of the current character. When the reordering | |
168 | engine needs to process the next character in the logical order, it | |
169 | fetches it and applies to it all the UBA levels, updating the | |
170 | iterator state as it goes. But when the buffer or string is | |
171 | scanned backwards, i.e. in the reverse order of buffer/string | |
172 | positions, the scanned characters were already processed during the | |
173 | preceding forward scan (see bidi_find_other_level_edge). To avoid | |
174 | costly re-processing of characters that were already processed | |
175 | during the forward scan, the iterator states computed while | |
176 | scanning forward are cached. | |
177 | ||
178 | The cache is just a linear array of 'struct bidi_it' objects, which | |
179 | is dynamically allocated and reallocated as needed, since the size | |
180 | of the cache depends on the text being processed. We only need the | |
181 | cache while processing embedded levels higher than the base | |
182 | paragraph embedding level, because these higher levels require | |
183 | changes in scan direction. Therefore, as soon as we are back to | |
184 | the base embedding level, we can free the cache; see the calls to | |
185 | bidi_cache_reset and bidi_cache_shrink, for the conditions to do | |
186 | this. | |
187 | ||
188 | The cache maintains the index of the next unused cache slot -- this | |
189 | is where the next iterator state will be cached. The function | |
190 | bidi_cache_iterator_state saves an instance of the state in the | |
191 | cache and increments the unused slot index. The companion function | |
192 | bidi_cache_find looks up a cached state that corresponds to a given | |
193 | buffer/string position. All of the cached states must correspond | |
194 | 1:1 to the buffer or string region whose processing they reflect; | |
195 | bidi.c will abort if it finds cache slots that violate this 1:1 | |
196 | correspondence. | |
197 | ||
198 | When the parent iterator 'struct it' is pushed (see push_it in | |
199 | xdisp.c) to pause the current iteration and start iterating over a | |
200 | different object (e.g., a 'display' string that covers some buffer | |
201 | text), the bidi iterator cache needs to be "pushed" as well, so | |
202 | that a new empty cache could be used while iterating over the new | |
203 | object. Later, when the new object is exhausted, and xdisp.c calls | |
204 | pop_it, we need to "pop" the bidi cache as well and return to the | |
205 | original cache. See bidi_push_it and bidi_pop_it for how this is | |
206 | done. | |
207 | ||
208 | Some functions of the display engine save copies of 'struct it' in | |
209 | local variables, and restore them later. For examples, see | |
210 | pos_visible_p and move_it_in_display_line_to in xdisp.c, and | |
211 | window_scroll_pixel_based in window.c. When this happens, we need | |
212 | to save and restore the bidi cache as well, because conceptually | |
213 | the cache is part of the 'struct it' state, and needs to be in | |
214 | perfect sync with the portion of the buffer/string that is being | |
215 | processed. This saving and restoring of the cache state is handled | |
216 | by bidi_shelve_cache and bidi_unshelve_cache, and the helper macros | |
217 | SAVE_IT and RESTORE_IT defined on xdisp.c. | |
218 | ||
219 | Note that, because reordering is implemented below the level in | |
220 | xdisp.c that breaks glyphs into screen lines, we are violating | |
221 | paragraph 3.4 of UAX#9. which mandates that line breaking shall be | |
222 | done before reordering each screen line separately. However, | |
223 | following UAX#9 to the letter in this matter goes against the basic | |
224 | design of the Emacs display engine, and so we choose here this | |
225 | minor deviation from the UBA letter in preference to redesign of | |
226 | the display engine. The effect of this is only seen in continued | |
227 | lines that are broken into screen lines in the middle of a run | |
228 | whose direction is opposite to the paragraph's base direction. | |
229 | ||
230 | Important design and implementation note: when the code needs to | |
231 | scan far ahead, be sure to avoid such scans as much as possible | |
232 | when the buffer/string doesn't contain any RTL characters. Users | |
233 | of left-to-right scripts will never forgive you if you introduce | |
234 | some slow-down due to bidi in situations that don't involve any | |
235 | bidirectional text. See the large comment near the beginning of | |
236 | bidi_resolve_neutral, for one situation where such shortcut was | |
237 | necessary. */ | |
b7b65b15 | 238 | |
b7b65b15 | 239 | #include <config.h> |
b7b65b15 | 240 | #include <stdio.h> |
29e3d8d1 | 241 | |
b7b65b15 | 242 | #include "lisp.h" |
b7b65b15 | 243 | #include "character.h" |
e5560ff7 | 244 | #include "buffer.h" |
b7b65b15 | 245 | #include "dispextern.h" |
e30b79c1 | 246 | #include "region-cache.h" |
b7b65b15 | 247 | |
2cc21167 | 248 | static bool bidi_initialized = 0; |
b7b65b15 | 249 | |
cbc4fd20 | 250 | static Lisp_Object bidi_type_table, bidi_mirror_table; |
b7b65b15 | 251 | |
c521fe29 | 252 | #define BIDI_EOB (-1) |
b7b65b15 | 253 | |
b7b65b15 EZ |
254 | /* Data type for describing the bidirectional character categories. */ |
255 | typedef enum { | |
256 | UNKNOWN_BC, | |
257 | NEUTRAL, | |
258 | WEAK, | |
ad826124 EZ |
259 | STRONG, |
260 | EXPLICIT_FORMATTING | |
b7b65b15 EZ |
261 | } bidi_category_t; |
262 | ||
32413314 EZ |
263 | /* UAX#9 says to search only for L, AL, or R types of characters, and |
264 | ignore RLE, RLO, LRE, and LRO, when determining the base paragraph | |
265 | level. Yudit indeed ignores them. This variable is therefore set | |
2cc21167 PE |
266 | by default to ignore them, but clearing it will take them into |
267 | account. */ | |
268 | extern bool bidi_ignore_explicit_marks_for_paragraph_level EXTERNALLY_VISIBLE; | |
269 | bool bidi_ignore_explicit_marks_for_paragraph_level = 1; | |
b7b65b15 | 270 | |
372b7a95 | 271 | static Lisp_Object paragraph_start_re, paragraph_separate_re; |
6bff6497 | 272 | static Lisp_Object Qparagraph_start, Qparagraph_separate; |
b7b65b15 | 273 | |
cbb09f04 EZ |
274 | \f |
275 | /*********************************************************************** | |
276 | Utilities | |
277 | ***********************************************************************/ | |
b7b65b15 | 278 | |
6bff6497 EZ |
279 | /* Return the bidi type of a character CH, subject to the current |
280 | directional OVERRIDE. */ | |
0a99eee1 | 281 | static bidi_type_t |
6bff6497 | 282 | bidi_get_type (int ch, bidi_dir_t override) |
b7b65b15 | 283 | { |
6bff6497 EZ |
284 | bidi_type_t default_type; |
285 | ||
e7402cb2 EZ |
286 | if (ch == BIDI_EOB) |
287 | return NEUTRAL_B; | |
e342a24d | 288 | if (ch < 0 || ch > MAX_CHAR) |
1088b922 | 289 | emacs_abort (); |
6bff6497 EZ |
290 | |
291 | default_type = (bidi_type_t) XINT (CHAR_TABLE_REF (bidi_type_table, ch)); | |
bca633fb EZ |
292 | /* Every valid character code, even those that are unassigned by the |
293 | UCD, have some bidi-class property, according to | |
294 | DerivedBidiClass.txt file. Therefore, if we ever get UNKNOWN_BT | |
295 | (= zero) code from CHAR_TABLE_REF, that's a bug. */ | |
296 | if (default_type == UNKNOWN_BT) | |
1088b922 | 297 | emacs_abort (); |
6bff6497 | 298 | |
6bff6497 EZ |
299 | switch (default_type) |
300 | { | |
ad826124 | 301 | case WEAK_BN: |
6bff6497 EZ |
302 | case NEUTRAL_B: |
303 | case LRE: | |
304 | case LRO: | |
305 | case RLE: | |
306 | case RLO: | |
307 | case PDF: | |
308 | return default_type; | |
ad826124 EZ |
309 | /* FIXME: The isolate controls are treated as BN until we add |
310 | support for UBA v6.3. */ | |
311 | case LRI: | |
312 | case RLI: | |
313 | case FSI: | |
314 | case PDI: | |
315 | return WEAK_BN; | |
6bff6497 | 316 | default: |
ad826124 EZ |
317 | if (override == L2R) |
318 | return STRONG_L; | |
319 | else if (override == R2L) | |
320 | return STRONG_R; | |
321 | else | |
322 | return default_type; | |
6bff6497 | 323 | } |
b7b65b15 EZ |
324 | } |
325 | ||
0a99eee1 | 326 | static void |
2d6e4628 EZ |
327 | bidi_check_type (bidi_type_t type) |
328 | { | |
a54e2c05 | 329 | eassert (UNKNOWN_BT <= type && type <= NEUTRAL_ON); |
2d6e4628 EZ |
330 | } |
331 | ||
b7b65b15 | 332 | /* Given a bidi TYPE of a character, return its category. */ |
0a99eee1 | 333 | static bidi_category_t |
b7b65b15 EZ |
334 | bidi_get_category (bidi_type_t type) |
335 | { | |
336 | switch (type) | |
337 | { | |
338 | case UNKNOWN_BT: | |
339 | return UNKNOWN_BC; | |
340 | case STRONG_L: | |
341 | case STRONG_R: | |
342 | case STRONG_AL: | |
b7b65b15 | 343 | return STRONG; |
b7b65b15 EZ |
344 | case WEAK_EN: |
345 | case WEAK_ES: | |
346 | case WEAK_ET: | |
347 | case WEAK_AN: | |
348 | case WEAK_CS: | |
349 | case WEAK_NSM: | |
350 | case WEAK_BN: | |
ad826124 EZ |
351 | /* FIXME */ |
352 | case LRI: | |
353 | case RLI: | |
354 | case FSI: | |
355 | case PDI: | |
b7b65b15 EZ |
356 | return WEAK; |
357 | case NEUTRAL_B: | |
358 | case NEUTRAL_S: | |
359 | case NEUTRAL_WS: | |
360 | case NEUTRAL_ON: | |
361 | return NEUTRAL; | |
ad826124 EZ |
362 | case LRE: |
363 | case LRO: | |
364 | case RLE: | |
365 | case RLO: | |
366 | case PDF: | |
367 | #if 0 | |
368 | /* FIXME: This awaits implementation of isolate support. */ | |
369 | case LRI: | |
370 | case RLI: | |
371 | case FSI: | |
372 | case PDI: | |
373 | #endif | |
374 | return EXPLICIT_FORMATTING; | |
b7b65b15 | 375 | default: |
1088b922 | 376 | emacs_abort (); |
b7b65b15 EZ |
377 | } |
378 | } | |
379 | ||
a6e8d97c EZ |
380 | /* Return the mirrored character of C, if it has one. If C has no |
381 | mirrored counterpart, return C. | |
382 | Note: The conditions in UAX#9 clause L4 regarding the surrounding | |
383 | context must be tested by the caller. */ | |
b7b65b15 EZ |
384 | int |
385 | bidi_mirror_char (int c) | |
386 | { | |
cbc4fd20 EZ |
387 | Lisp_Object val; |
388 | ||
389 | if (c == BIDI_EOB) | |
390 | return c; | |
391 | if (c < 0 || c > MAX_CHAR) | |
1088b922 | 392 | emacs_abort (); |
b7b65b15 | 393 | |
cbc4fd20 EZ |
394 | val = CHAR_TABLE_REF (bidi_mirror_table, c); |
395 | if (INTEGERP (val)) | |
b7b65b15 | 396 | { |
da92a98c | 397 | int v; |
9e1a06fc | 398 | |
da92a98c PE |
399 | /* When debugging, check before assigning to V, so that the check |
400 | isn't broken by undefined behavior due to int overflow. */ | |
81899c91 | 401 | eassert (CHAR_VALID_P (XINT (val))); |
9e1a06fc | 402 | |
da92a98c PE |
403 | v = XINT (val); |
404 | ||
9e1a06fc EZ |
405 | /* Minimal test we must do in optimized builds, to prevent weird |
406 | crashes further down the road. */ | |
407 | if (v < 0 || v > MAX_CHAR) | |
1088b922 | 408 | emacs_abort (); |
9e1a06fc EZ |
409 | |
410 | return v; | |
b7b65b15 | 411 | } |
cbc4fd20 | 412 | |
b7b65b15 EZ |
413 | return c; |
414 | } | |
415 | ||
cbb09f04 EZ |
416 | /* Determine the start-of-run (sor) directional type given the two |
417 | embedding levels on either side of the run boundary. Also, update | |
418 | the saved info about previously seen characters, since that info is | |
419 | generally valid for a single level run. */ | |
0a99eee1 | 420 | static void |
cbb09f04 EZ |
421 | bidi_set_sor_type (struct bidi_it *bidi_it, int level_before, int level_after) |
422 | { | |
512a289d | 423 | int higher_level = (level_before > level_after ? level_before : level_after); |
cbb09f04 EZ |
424 | |
425 | /* The prev_was_pdf gork is required for when we have several PDFs | |
426 | in a row. In that case, we want to compute the sor type for the | |
427 | next level run only once: when we see the first PDF. That's | |
428 | because the sor type depends only on the higher of the two levels | |
429 | that we find on the two sides of the level boundary (see UAX#9, | |
430 | clause X10), and so we don't need to know the final embedding | |
431 | level to which we descend after processing all the PDFs. */ | |
432 | if (!bidi_it->prev_was_pdf || level_before < level_after) | |
433 | /* FIXME: should the default sor direction be user selectable? */ | |
512a289d | 434 | bidi_it->sor = ((higher_level & 1) != 0 ? R2L : L2R); |
cbb09f04 EZ |
435 | if (level_before > level_after) |
436 | bidi_it->prev_was_pdf = 1; | |
437 | ||
438 | bidi_it->prev.type = UNKNOWN_BT; | |
512a289d RS |
439 | bidi_it->last_strong.type = bidi_it->last_strong.type_after_w1 |
440 | = bidi_it->last_strong.orig_type = UNKNOWN_BT; | |
441 | bidi_it->prev_for_neutral.type = (bidi_it->sor == R2L ? STRONG_R : STRONG_L); | |
cbb09f04 EZ |
442 | bidi_it->prev_for_neutral.charpos = bidi_it->charpos; |
443 | bidi_it->prev_for_neutral.bytepos = bidi_it->bytepos; | |
512a289d RS |
444 | bidi_it->next_for_neutral.type = bidi_it->next_for_neutral.type_after_w1 |
445 | = bidi_it->next_for_neutral.orig_type = UNKNOWN_BT; | |
cbb09f04 EZ |
446 | bidi_it->ignore_bn_limit = -1; /* meaning it's unknown */ |
447 | } | |
448 | ||
449 | /* Push the current embedding level and override status; reset the | |
450 | current level to LEVEL and the current override status to OVERRIDE. */ | |
0a99eee1 | 451 | static void |
cbb09f04 EZ |
452 | bidi_push_embedding_level (struct bidi_it *bidi_it, |
453 | int level, bidi_dir_t override) | |
454 | { | |
455 | bidi_it->stack_idx++; | |
a54e2c05 | 456 | eassert (bidi_it->stack_idx < BIDI_MAXLEVEL); |
cbb09f04 EZ |
457 | bidi_it->level_stack[bidi_it->stack_idx].level = level; |
458 | bidi_it->level_stack[bidi_it->stack_idx].override = override; | |
459 | } | |
460 | ||
461 | /* Pop the embedding level and directional override status from the | |
462 | stack, and return the new level. */ | |
0a99eee1 | 463 | static int |
cbb09f04 EZ |
464 | bidi_pop_embedding_level (struct bidi_it *bidi_it) |
465 | { | |
466 | /* UAX#9 says to ignore invalid PDFs. */ | |
467 | if (bidi_it->stack_idx > 0) | |
468 | bidi_it->stack_idx--; | |
469 | return bidi_it->level_stack[bidi_it->stack_idx].level; | |
470 | } | |
471 | ||
472 | /* Record in SAVED_INFO the information about the current character. */ | |
0a99eee1 | 473 | static void |
cbb09f04 EZ |
474 | bidi_remember_char (struct bidi_saved_info *saved_info, |
475 | struct bidi_it *bidi_it) | |
476 | { | |
477 | saved_info->charpos = bidi_it->charpos; | |
478 | saved_info->bytepos = bidi_it->bytepos; | |
479 | saved_info->type = bidi_it->type; | |
480 | bidi_check_type (bidi_it->type); | |
481 | saved_info->type_after_w1 = bidi_it->type_after_w1; | |
482 | bidi_check_type (bidi_it->type_after_w1); | |
483 | saved_info->orig_type = bidi_it->orig_type; | |
484 | bidi_check_type (bidi_it->orig_type); | |
485 | } | |
486 | ||
b7b65b15 EZ |
487 | /* Copy the bidi iterator from FROM to TO. To save cycles, this only |
488 | copies the part of the level stack that is actually in use. */ | |
0a99eee1 | 489 | static void |
b7b65b15 EZ |
490 | bidi_copy_it (struct bidi_it *to, struct bidi_it *from) |
491 | { | |
05584c31 PE |
492 | /* Copy everything from the start through the active part of |
493 | the level stack. */ | |
494 | memcpy (to, from, | |
495 | (offsetof (struct bidi_it, level_stack[1]) | |
496 | + from->stack_idx * sizeof from->level_stack[0])); | |
b7b65b15 EZ |
497 | } |
498 | ||
cbb09f04 EZ |
499 | \f |
500 | /*********************************************************************** | |
501 | Caching the bidi iterator states | |
502 | ***********************************************************************/ | |
b7b65b15 | 503 | |
2fe72643 EZ |
504 | #define BIDI_CACHE_CHUNK 200 |
505 | static struct bidi_it *bidi_cache; | |
39e378da | 506 | static ptrdiff_t bidi_cache_size = 0; |
ad6042bb | 507 | enum { elsz = sizeof (struct bidi_it) }; |
39e378da PE |
508 | static ptrdiff_t bidi_cache_idx; /* next unused cache slot */ |
509 | static ptrdiff_t bidi_cache_last_idx; /* slot of last cache hit */ | |
510 | static ptrdiff_t bidi_cache_start = 0; /* start of cache for this | |
87e67904 EZ |
511 | "stack" level */ |
512 | ||
bc18e09d PE |
513 | /* 5-slot stack for saving the start of the previous level of the |
514 | cache. xdisp.c maintains a 5-slot stack for its iterator state, | |
515 | and we need the same size of our stack. */ | |
516 | static ptrdiff_t bidi_cache_start_stack[IT_STACK_SIZE]; | |
517 | static int bidi_cache_sp; | |
518 | ||
519 | /* Size of header used by bidi_shelve_cache. */ | |
520 | enum | |
521 | { | |
512a289d RS |
522 | bidi_shelve_header_size |
523 | = (sizeof (bidi_cache_idx) + sizeof (bidi_cache_start_stack) | |
524 | + sizeof (bidi_cache_sp) + sizeof (bidi_cache_start) | |
525 | + sizeof (bidi_cache_last_idx)) | |
bc18e09d PE |
526 | }; |
527 | ||
87e67904 EZ |
528 | /* Reset the cache state to the empty state. We only reset the part |
529 | of the cache relevant to iteration of the current object. Previous | |
530 | objects, which are pushed on the display iterator's stack, are left | |
531 | intact. This is called when the cached information is no more | |
532 | useful for the current iteration, e.g. when we were reseated to a | |
533 | new position on the same object. */ | |
0a99eee1 | 534 | static void |
b7b65b15 EZ |
535 | bidi_cache_reset (void) |
536 | { | |
87e67904 | 537 | bidi_cache_idx = bidi_cache_start; |
b7b65b15 EZ |
538 | bidi_cache_last_idx = -1; |
539 | } | |
540 | ||
87e67904 EZ |
541 | /* Shrink the cache to its minimal size. Called when we init the bidi |
542 | iterator for reordering a buffer or a string that does not come | |
543 | from display properties, because that means all the previously | |
544 | cached info is of no further use. */ | |
0a99eee1 | 545 | static void |
2fe72643 EZ |
546 | bidi_cache_shrink (void) |
547 | { | |
548 | if (bidi_cache_size > BIDI_CACHE_CHUNK) | |
549 | { | |
38182d90 | 550 | bidi_cache = xrealloc (bidi_cache, BIDI_CACHE_CHUNK * elsz); |
51f30bc5 | 551 | bidi_cache_size = BIDI_CACHE_CHUNK; |
2fe72643 EZ |
552 | } |
553 | bidi_cache_reset (); | |
554 | } | |
555 | ||
0a99eee1 | 556 | static void |
39e378da | 557 | bidi_cache_fetch_state (ptrdiff_t idx, struct bidi_it *bidi_it) |
b7b65b15 EZ |
558 | { |
559 | int current_scan_dir = bidi_it->scan_dir; | |
560 | ||
87e67904 | 561 | if (idx < bidi_cache_start || idx >= bidi_cache_idx) |
1088b922 | 562 | emacs_abort (); |
b7b65b15 EZ |
563 | |
564 | bidi_copy_it (bidi_it, &bidi_cache[idx]); | |
565 | bidi_it->scan_dir = current_scan_dir; | |
566 | bidi_cache_last_idx = idx; | |
567 | } | |
568 | ||
569 | /* Find a cached state with a given CHARPOS and resolved embedding | |
570 | level less or equal to LEVEL. if LEVEL is -1, disregard the | |
571 | resolved levels in cached states. DIR, if non-zero, means search | |
572 | in that direction from the last cache hit. */ | |
0a99eee1 | 573 | static ptrdiff_t |
d311d28c | 574 | bidi_cache_search (ptrdiff_t charpos, int level, int dir) |
b7b65b15 | 575 | { |
39e378da | 576 | ptrdiff_t i, i_start; |
b7b65b15 | 577 | |
6eec7596 | 578 | if (bidi_cache_idx > bidi_cache_start) |
b7b65b15 | 579 | { |
6eec7596 EZ |
580 | if (bidi_cache_last_idx == -1) |
581 | bidi_cache_last_idx = bidi_cache_idx - 1; | |
b7b65b15 | 582 | if (charpos < bidi_cache[bidi_cache_last_idx].charpos) |
182ce2d2 EZ |
583 | { |
584 | dir = -1; | |
585 | i_start = bidi_cache_last_idx - 1; | |
586 | } | |
587 | else if (charpos > (bidi_cache[bidi_cache_last_idx].charpos | |
588 | + bidi_cache[bidi_cache_last_idx].nchars - 1)) | |
589 | { | |
590 | dir = 1; | |
591 | i_start = bidi_cache_last_idx + 1; | |
592 | } | |
593 | else if (dir) | |
b7b65b15 EZ |
594 | i_start = bidi_cache_last_idx; |
595 | else | |
596 | { | |
597 | dir = -1; | |
598 | i_start = bidi_cache_idx - 1; | |
599 | } | |
600 | ||
601 | if (dir < 0) | |
602 | { | |
603 | /* Linear search for now; FIXME! */ | |
87e67904 | 604 | for (i = i_start; i >= bidi_cache_start; i--) |
182ce2d2 EZ |
605 | if (bidi_cache[i].charpos <= charpos |
606 | && charpos < bidi_cache[i].charpos + bidi_cache[i].nchars | |
b7b65b15 EZ |
607 | && (level == -1 || bidi_cache[i].resolved_level <= level)) |
608 | return i; | |
609 | } | |
610 | else | |
611 | { | |
612 | for (i = i_start; i < bidi_cache_idx; i++) | |
182ce2d2 EZ |
613 | if (bidi_cache[i].charpos <= charpos |
614 | && charpos < bidi_cache[i].charpos + bidi_cache[i].nchars | |
b7b65b15 EZ |
615 | && (level == -1 || bidi_cache[i].resolved_level <= level)) |
616 | return i; | |
617 | } | |
618 | } | |
619 | ||
620 | return -1; | |
621 | } | |
622 | ||
623 | /* Find a cached state where the resolved level changes to a value | |
624 | that is lower than LEVEL, and return its cache slot index. DIR is | |
625 | the direction to search, starting with the last used cache slot. | |
87e67904 | 626 | If DIR is zero, we search backwards from the last occupied cache |
2cc21167 | 627 | slot. BEFORE means return the index of the slot that |
87e67904 | 628 | is ``before'' the level change in the search direction. That is, |
b7b65b15 EZ |
629 | given the cached levels like this: |
630 | ||
631 | 1122333442211 | |
632 | AB C | |
633 | ||
634 | and assuming we are at the position cached at the slot marked with | |
635 | C, searching backwards (DIR = -1) for LEVEL = 2 will return the | |
636 | index of slot B or A, depending whether BEFORE is, respectively, | |
2cc21167 | 637 | true or false. */ |
39e378da | 638 | static ptrdiff_t |
2cc21167 | 639 | bidi_cache_find_level_change (int level, int dir, bool before) |
b7b65b15 EZ |
640 | { |
641 | if (bidi_cache_idx) | |
642 | { | |
39e378da | 643 | ptrdiff_t i = dir ? bidi_cache_last_idx : bidi_cache_idx - 1; |
b7b65b15 EZ |
644 | int incr = before ? 1 : 0; |
645 | ||
a54e2c05 | 646 | eassert (!dir || bidi_cache_last_idx >= 0); |
6eec7596 | 647 | |
b7b65b15 EZ |
648 | if (!dir) |
649 | dir = -1; | |
650 | else if (!incr) | |
651 | i += dir; | |
652 | ||
653 | if (dir < 0) | |
654 | { | |
87e67904 | 655 | while (i >= bidi_cache_start + incr) |
b7b65b15 EZ |
656 | { |
657 | if (bidi_cache[i - incr].resolved_level >= 0 | |
658 | && bidi_cache[i - incr].resolved_level < level) | |
659 | return i; | |
660 | i--; | |
661 | } | |
662 | } | |
663 | else | |
664 | { | |
665 | while (i < bidi_cache_idx - incr) | |
666 | { | |
667 | if (bidi_cache[i + incr].resolved_level >= 0 | |
668 | && bidi_cache[i + incr].resolved_level < level) | |
669 | return i; | |
670 | i++; | |
671 | } | |
672 | } | |
673 | } | |
674 | ||
675 | return -1; | |
676 | } | |
677 | ||
0a99eee1 | 678 | static void |
39e378da | 679 | bidi_cache_ensure_space (ptrdiff_t idx) |
58b9f433 EZ |
680 | { |
681 | /* Enlarge the cache as needed. */ | |
682 | if (idx >= bidi_cache_size) | |
683 | { | |
f0eb61e9 PE |
684 | /* The bidi cache cannot be larger than the largest Lisp string |
685 | or buffer. */ | |
512a289d RS |
686 | ptrdiff_t string_or_buffer_bound |
687 | = max (BUF_BYTES_MAX, STRING_BYTES_BOUND); | |
f0eb61e9 PE |
688 | |
689 | /* Also, it cannot be larger than what C can represent. */ | |
512a289d RS |
690 | ptrdiff_t c_bound |
691 | = (min (PTRDIFF_MAX, SIZE_MAX) - bidi_shelve_header_size) / elsz; | |
f0eb61e9 | 692 | |
512a289d RS |
693 | bidi_cache |
694 | = xpalloc (bidi_cache, &bidi_cache_size, | |
695 | max (BIDI_CACHE_CHUNK, idx - bidi_cache_size + 1), | |
696 | min (string_or_buffer_bound, c_bound), elsz); | |
58b9f433 EZ |
697 | } |
698 | } | |
699 | ||
0a99eee1 | 700 | static void |
2cc21167 | 701 | bidi_cache_iterator_state (struct bidi_it *bidi_it, bool resolved) |
b7b65b15 | 702 | { |
39e378da | 703 | ptrdiff_t idx; |
b7b65b15 EZ |
704 | |
705 | /* We should never cache on backward scans. */ | |
706 | if (bidi_it->scan_dir == -1) | |
1088b922 | 707 | emacs_abort (); |
b7b65b15 EZ |
708 | idx = bidi_cache_search (bidi_it->charpos, -1, 1); |
709 | ||
710 | if (idx < 0) | |
711 | { | |
712 | idx = bidi_cache_idx; | |
58b9f433 | 713 | bidi_cache_ensure_space (idx); |
bd924a5d EZ |
714 | /* Character positions should correspond to cache positions 1:1. |
715 | If we are outside the range of cached positions, the cache is | |
716 | useless and must be reset. */ | |
87e67904 | 717 | if (idx > bidi_cache_start && |
182ce2d2 EZ |
718 | (bidi_it->charpos > (bidi_cache[idx - 1].charpos |
719 | + bidi_cache[idx - 1].nchars) | |
a1344e7d | 720 | || bidi_it->charpos < bidi_cache[bidi_cache_start].charpos)) |
bd924a5d EZ |
721 | { |
722 | bidi_cache_reset (); | |
87e67904 | 723 | idx = bidi_cache_start; |
bd924a5d | 724 | } |
102ebb00 | 725 | if (bidi_it->nchars <= 0) |
1088b922 | 726 | emacs_abort (); |
b7b65b15 EZ |
727 | bidi_copy_it (&bidi_cache[idx], bidi_it); |
728 | if (!resolved) | |
729 | bidi_cache[idx].resolved_level = -1; | |
730 | } | |
731 | else | |
732 | { | |
733 | /* Copy only the members which could have changed, to avoid | |
734 | costly copying of the entire struct. */ | |
735 | bidi_cache[idx].type = bidi_it->type; | |
2d6e4628 | 736 | bidi_check_type (bidi_it->type); |
89d3374a EZ |
737 | bidi_cache[idx].type_after_w1 = bidi_it->type_after_w1; |
738 | bidi_check_type (bidi_it->type_after_w1); | |
b7b65b15 EZ |
739 | if (resolved) |
740 | bidi_cache[idx].resolved_level = bidi_it->resolved_level; | |
741 | else | |
742 | bidi_cache[idx].resolved_level = -1; | |
743 | bidi_cache[idx].invalid_levels = bidi_it->invalid_levels; | |
744 | bidi_cache[idx].invalid_rl_levels = bidi_it->invalid_rl_levels; | |
745 | bidi_cache[idx].next_for_neutral = bidi_it->next_for_neutral; | |
746 | bidi_cache[idx].next_for_ws = bidi_it->next_for_ws; | |
747 | bidi_cache[idx].ignore_bn_limit = bidi_it->ignore_bn_limit; | |
f67cdd7f | 748 | bidi_cache[idx].disp_pos = bidi_it->disp_pos; |
0c95fcf7 | 749 | bidi_cache[idx].disp_prop = bidi_it->disp_prop; |
b7b65b15 EZ |
750 | } |
751 | ||
752 | bidi_cache_last_idx = idx; | |
753 | if (idx >= bidi_cache_idx) | |
754 | bidi_cache_idx = idx + 1; | |
755 | } | |
756 | ||
0a99eee1 | 757 | static bidi_type_t |
d311d28c | 758 | bidi_cache_find (ptrdiff_t charpos, int level, struct bidi_it *bidi_it) |
b7b65b15 | 759 | { |
39e378da | 760 | ptrdiff_t i = bidi_cache_search (charpos, level, bidi_it->scan_dir); |
b7b65b15 | 761 | |
87e67904 | 762 | if (i >= bidi_cache_start) |
b7b65b15 EZ |
763 | { |
764 | bidi_dir_t current_scan_dir = bidi_it->scan_dir; | |
765 | ||
5009f85e | 766 | bidi_copy_it (bidi_it, &bidi_cache[i]); |
b7b65b15 | 767 | bidi_cache_last_idx = i; |
5a5fa834 | 768 | /* Don't let scan direction from the cached state override |
b7b65b15 EZ |
769 | the current scan direction. */ |
770 | bidi_it->scan_dir = current_scan_dir; | |
771 | return bidi_it->type; | |
772 | } | |
773 | ||
774 | return UNKNOWN_BT; | |
775 | } | |
776 | ||
0a99eee1 | 777 | static int |
b7b65b15 EZ |
778 | bidi_peek_at_next_level (struct bidi_it *bidi_it) |
779 | { | |
87e67904 | 780 | if (bidi_cache_idx == bidi_cache_start || bidi_cache_last_idx == -1) |
1088b922 | 781 | emacs_abort (); |
b7b65b15 EZ |
782 | return bidi_cache[bidi_cache_last_idx + bidi_it->scan_dir].resolved_level; |
783 | } | |
784 | ||
cbb09f04 | 785 | \f |
58b9f433 EZ |
786 | /*********************************************************************** |
787 | Pushing and popping the bidi iterator state | |
788 | ***********************************************************************/ | |
58b9f433 EZ |
789 | |
790 | /* Push the bidi iterator state in preparation for reordering a | |
791 | different object, e.g. display string found at certain buffer | |
7e2ad32c EZ |
792 | position. Pushing the bidi iterator boils down to saving its |
793 | entire state on the cache and starting a new cache "stacked" on top | |
794 | of the current cache. */ | |
58b9f433 EZ |
795 | void |
796 | bidi_push_it (struct bidi_it *bidi_it) | |
b7b65b15 | 797 | { |
58b9f433 EZ |
798 | /* Save the current iterator state in its entirety after the last |
799 | used cache slot. */ | |
800 | bidi_cache_ensure_space (bidi_cache_idx); | |
ae1d87e2 | 801 | bidi_cache[bidi_cache_idx++] = *bidi_it; |
be39f003 | 802 | |
58b9f433 | 803 | /* Push the current cache start onto the stack. */ |
a54e2c05 | 804 | eassert (bidi_cache_sp < IT_STACK_SIZE); |
58b9f433 | 805 | bidi_cache_start_stack[bidi_cache_sp++] = bidi_cache_start; |
be39f003 | 806 | |
58b9f433 EZ |
807 | /* Start a new level of cache, and make it empty. */ |
808 | bidi_cache_start = bidi_cache_idx; | |
809 | bidi_cache_last_idx = -1; | |
810 | } | |
811 | ||
812 | /* Restore the iterator state saved by bidi_push_it and return the | |
813 | cache to the corresponding state. */ | |
814 | void | |
815 | bidi_pop_it (struct bidi_it *bidi_it) | |
816 | { | |
817 | if (bidi_cache_start <= 0) | |
1088b922 | 818 | emacs_abort (); |
58b9f433 EZ |
819 | |
820 | /* Reset the next free cache slot index to what it was before the | |
821 | call to bidi_push_it. */ | |
822 | bidi_cache_idx = bidi_cache_start - 1; | |
823 | ||
824 | /* Restore the bidi iterator state saved in the cache. */ | |
ae1d87e2 | 825 | *bidi_it = bidi_cache[bidi_cache_idx]; |
58b9f433 EZ |
826 | |
827 | /* Pop the previous cache start from the stack. */ | |
828 | if (bidi_cache_sp <= 0) | |
1088b922 | 829 | emacs_abort (); |
58b9f433 EZ |
830 | bidi_cache_start = bidi_cache_start_stack[--bidi_cache_sp]; |
831 | ||
832 | /* Invalidate the last-used cache slot data. */ | |
833 | bidi_cache_last_idx = -1; | |
834 | } | |
835 | ||
ec7cc85b | 836 | static ptrdiff_t bidi_cache_total_alloc; |
35928349 | 837 | |
ed94e6d7 EZ |
838 | /* Stash away a copy of the cache and its control variables. */ |
839 | void * | |
840 | bidi_shelve_cache (void) | |
841 | { | |
842 | unsigned char *databuf; | |
458bfed3 | 843 | ptrdiff_t alloc; |
ed94e6d7 | 844 | |
35928349 | 845 | /* Empty cache. */ |
ed94e6d7 EZ |
846 | if (bidi_cache_idx == 0) |
847 | return NULL; | |
848 | ||
458bfed3 PE |
849 | alloc = (bidi_shelve_header_size |
850 | + bidi_cache_idx * sizeof (struct bidi_it)); | |
851 | databuf = xmalloc (alloc); | |
852 | bidi_cache_total_alloc += alloc; | |
35928349 | 853 | |
ed94e6d7 EZ |
854 | memcpy (databuf, &bidi_cache_idx, sizeof (bidi_cache_idx)); |
855 | memcpy (databuf + sizeof (bidi_cache_idx), | |
856 | bidi_cache, bidi_cache_idx * sizeof (struct bidi_it)); | |
857 | memcpy (databuf + sizeof (bidi_cache_idx) | |
858 | + bidi_cache_idx * sizeof (struct bidi_it), | |
859 | bidi_cache_start_stack, sizeof (bidi_cache_start_stack)); | |
860 | memcpy (databuf + sizeof (bidi_cache_idx) | |
861 | + bidi_cache_idx * sizeof (struct bidi_it) | |
862 | + sizeof (bidi_cache_start_stack), | |
863 | &bidi_cache_sp, sizeof (bidi_cache_sp)); | |
864 | memcpy (databuf + sizeof (bidi_cache_idx) | |
865 | + bidi_cache_idx * sizeof (struct bidi_it) | |
866 | + sizeof (bidi_cache_start_stack) + sizeof (bidi_cache_sp), | |
867 | &bidi_cache_start, sizeof (bidi_cache_start)); | |
868 | memcpy (databuf + sizeof (bidi_cache_idx) | |
869 | + bidi_cache_idx * sizeof (struct bidi_it) | |
870 | + sizeof (bidi_cache_start_stack) + sizeof (bidi_cache_sp) | |
871 | + sizeof (bidi_cache_start), | |
872 | &bidi_cache_last_idx, sizeof (bidi_cache_last_idx)); | |
873 | ||
874 | return databuf; | |
875 | } | |
876 | ||
d1410150 EZ |
877 | /* Restore the cache state from a copy stashed away by |
878 | bidi_shelve_cache, and free the buffer used to stash that copy. | |
2cc21167 | 879 | JUST_FREE means free the buffer, but don't restore the |
d1410150 EZ |
880 | cache; used when the corresponding iterator is discarded instead of |
881 | being restored. */ | |
ed94e6d7 | 882 | void |
2cc21167 | 883 | bidi_unshelve_cache (void *databuf, bool just_free) |
ed94e6d7 EZ |
884 | { |
885 | unsigned char *p = databuf; | |
886 | ||
887 | if (!p) | |
be39f003 | 888 | { |
d1410150 EZ |
889 | if (!just_free) |
890 | { | |
891 | /* A NULL pointer means an empty cache. */ | |
892 | bidi_cache_start = 0; | |
893 | bidi_cache_sp = 0; | |
894 | bidi_cache_reset (); | |
895 | } | |
be39f003 | 896 | } |
ed94e6d7 EZ |
897 | else |
898 | { | |
35928349 EZ |
899 | if (just_free) |
900 | { | |
901 | ptrdiff_t idx; | |
902 | ||
903 | memcpy (&idx, p, sizeof (bidi_cache_idx)); | |
512a289d RS |
904 | bidi_cache_total_alloc |
905 | -= bidi_shelve_header_size + idx * sizeof (struct bidi_it); | |
35928349 EZ |
906 | } |
907 | else | |
908 | { | |
909 | memcpy (&bidi_cache_idx, p, sizeof (bidi_cache_idx)); | |
910 | bidi_cache_ensure_space (bidi_cache_idx); | |
911 | memcpy (bidi_cache, p + sizeof (bidi_cache_idx), | |
912 | bidi_cache_idx * sizeof (struct bidi_it)); | |
913 | memcpy (bidi_cache_start_stack, | |
914 | p + sizeof (bidi_cache_idx) | |
915 | + bidi_cache_idx * sizeof (struct bidi_it), | |
916 | sizeof (bidi_cache_start_stack)); | |
917 | memcpy (&bidi_cache_sp, | |
918 | p + sizeof (bidi_cache_idx) | |
919 | + bidi_cache_idx * sizeof (struct bidi_it) | |
920 | + sizeof (bidi_cache_start_stack), | |
921 | sizeof (bidi_cache_sp)); | |
922 | memcpy (&bidi_cache_start, | |
923 | p + sizeof (bidi_cache_idx) | |
924 | + bidi_cache_idx * sizeof (struct bidi_it) | |
925 | + sizeof (bidi_cache_start_stack) + sizeof (bidi_cache_sp), | |
926 | sizeof (bidi_cache_start)); | |
927 | memcpy (&bidi_cache_last_idx, | |
928 | p + sizeof (bidi_cache_idx) | |
929 | + bidi_cache_idx * sizeof (struct bidi_it) | |
930 | + sizeof (bidi_cache_start_stack) + sizeof (bidi_cache_sp) | |
931 | + sizeof (bidi_cache_start), | |
932 | sizeof (bidi_cache_last_idx)); | |
512a289d RS |
933 | bidi_cache_total_alloc |
934 | -= (bidi_shelve_header_size | |
935 | + bidi_cache_idx * sizeof (struct bidi_it)); | |
35928349 | 936 | } |
ed94e6d7 EZ |
937 | |
938 | xfree (p); | |
939 | } | |
940 | } | |
b7b65b15 | 941 | |
58b9f433 | 942 | \f |
cbb09f04 EZ |
943 | /*********************************************************************** |
944 | Initialization | |
945 | ***********************************************************************/ | |
946 | static void | |
947 | bidi_initialize (void) | |
b7b65b15 | 948 | { |
474a8465 EZ |
949 | bidi_type_table = uniprop_table (intern ("bidi-class")); |
950 | if (NILP (bidi_type_table)) | |
1088b922 | 951 | emacs_abort (); |
cbb09f04 EZ |
952 | staticpro (&bidi_type_table); |
953 | ||
474a8465 EZ |
954 | bidi_mirror_table = uniprop_table (intern ("mirroring")); |
955 | if (NILP (bidi_mirror_table)) | |
1088b922 | 956 | emacs_abort (); |
cbb09f04 EZ |
957 | staticpro (&bidi_mirror_table); |
958 | ||
cbb09f04 EZ |
959 | Qparagraph_start = intern ("paragraph-start"); |
960 | staticpro (&Qparagraph_start); | |
961 | paragraph_start_re = Fsymbol_value (Qparagraph_start); | |
962 | if (!STRINGP (paragraph_start_re)) | |
963 | paragraph_start_re = build_string ("\f\\|[ \t]*$"); | |
964 | staticpro (¶graph_start_re); | |
965 | Qparagraph_separate = intern ("paragraph-separate"); | |
966 | staticpro (&Qparagraph_separate); | |
967 | paragraph_separate_re = Fsymbol_value (Qparagraph_separate); | |
968 | if (!STRINGP (paragraph_separate_re)) | |
969 | paragraph_separate_re = build_string ("[ \t\f]*$"); | |
970 | staticpro (¶graph_separate_re); | |
58b9f433 EZ |
971 | |
972 | bidi_cache_sp = 0; | |
ec7cc85b | 973 | bidi_cache_total_alloc = 0; |
58b9f433 | 974 | |
cbb09f04 | 975 | bidi_initialized = 1; |
b7b65b15 EZ |
976 | } |
977 | ||
cbb09f04 EZ |
978 | /* Do whatever UAX#9 clause X8 says should be done at paragraph's |
979 | end. */ | |
0a99eee1 | 980 | static void |
cbb09f04 | 981 | bidi_set_paragraph_end (struct bidi_it *bidi_it) |
b7b65b15 | 982 | { |
cbb09f04 EZ |
983 | bidi_it->invalid_levels = 0; |
984 | bidi_it->invalid_rl_levels = -1; | |
985 | bidi_it->stack_idx = 0; | |
986 | bidi_it->resolved_level = bidi_it->level_stack[0].level; | |
987 | } | |
b7b65b15 | 988 | |
cbb09f04 EZ |
989 | /* Initialize the bidi iterator from buffer/string position CHARPOS. */ |
990 | void | |
2cc21167 | 991 | bidi_init_it (ptrdiff_t charpos, ptrdiff_t bytepos, bool frame_window_p, |
cbb09f04 EZ |
992 | struct bidi_it *bidi_it) |
993 | { | |
994 | if (! bidi_initialized) | |
995 | bidi_initialize (); | |
996 | if (charpos >= 0) | |
997 | bidi_it->charpos = charpos; | |
998 | if (bytepos >= 0) | |
999 | bidi_it->bytepos = bytepos; | |
1000 | bidi_it->frame_window_p = frame_window_p; | |
1001 | bidi_it->nchars = -1; /* to be computed in bidi_resolve_explicit_1 */ | |
1002 | bidi_it->first_elt = 1; | |
1003 | bidi_set_paragraph_end (bidi_it); | |
1004 | bidi_it->new_paragraph = 1; | |
1005 | bidi_it->separator_limit = -1; | |
1006 | bidi_it->type = NEUTRAL_B; | |
1007 | bidi_it->type_after_w1 = NEUTRAL_B; | |
1008 | bidi_it->orig_type = NEUTRAL_B; | |
1009 | bidi_it->prev_was_pdf = 0; | |
512a289d RS |
1010 | bidi_it->prev.type = bidi_it->prev.type_after_w1 |
1011 | = bidi_it->prev.orig_type = UNKNOWN_BT; | |
1012 | bidi_it->last_strong.type = bidi_it->last_strong.type_after_w1 | |
1013 | = bidi_it->last_strong.orig_type = UNKNOWN_BT; | |
cbb09f04 | 1014 | bidi_it->next_for_neutral.charpos = -1; |
512a289d RS |
1015 | bidi_it->next_for_neutral.type |
1016 | = bidi_it->next_for_neutral.type_after_w1 | |
1017 | = bidi_it->next_for_neutral.orig_type = UNKNOWN_BT; | |
cbb09f04 | 1018 | bidi_it->prev_for_neutral.charpos = -1; |
512a289d RS |
1019 | bidi_it->prev_for_neutral.type |
1020 | = bidi_it->prev_for_neutral.type_after_w1 | |
1021 | = bidi_it->prev_for_neutral.orig_type = UNKNOWN_BT; | |
cbb09f04 EZ |
1022 | bidi_it->sor = L2R; /* FIXME: should it be user-selectable? */ |
1023 | bidi_it->disp_pos = -1; /* invalid/unknown */ | |
0c95fcf7 | 1024 | bidi_it->disp_prop = 0; |
cbb09f04 EZ |
1025 | /* We can only shrink the cache if we are at the bottom level of its |
1026 | "stack". */ | |
1027 | if (bidi_cache_start == 0) | |
1028 | bidi_cache_shrink (); | |
58b9f433 EZ |
1029 | else |
1030 | bidi_cache_reset (); | |
b7b65b15 EZ |
1031 | } |
1032 | ||
182ce2d2 | 1033 | /* Perform initializations for reordering a new line of bidi text. */ |
6bff6497 | 1034 | static void |
be39f003 EZ |
1035 | bidi_line_init (struct bidi_it *bidi_it) |
1036 | { | |
1037 | bidi_it->scan_dir = 1; /* FIXME: do we need to have control on this? */ | |
1038 | bidi_it->resolved_level = bidi_it->level_stack[0].level; | |
1039 | bidi_it->level_stack[0].override = NEUTRAL_DIR; /* X1 */ | |
1040 | bidi_it->invalid_levels = 0; | |
1041 | bidi_it->invalid_rl_levels = -1; | |
4787455f EZ |
1042 | /* Setting this to zero will force its recomputation the first time |
1043 | we need it for W5. */ | |
1044 | bidi_it->next_en_pos = 0; | |
7b5d6677 | 1045 | bidi_it->next_en_type = UNKNOWN_BT; |
be39f003 | 1046 | bidi_it->next_for_ws.type = UNKNOWN_BT; |
b44d9321 | 1047 | bidi_set_sor_type (bidi_it, |
512a289d | 1048 | (bidi_it->paragraph_dir == R2L ? 1 : 0), |
be39f003 EZ |
1049 | bidi_it->level_stack[0].level); /* X10 */ |
1050 | ||
1051 | bidi_cache_reset (); | |
1052 | } | |
1053 | ||
cbb09f04 EZ |
1054 | \f |
1055 | /*********************************************************************** | |
1056 | Fetching characters | |
1057 | ***********************************************************************/ | |
1058 | ||
f3014ef5 EZ |
1059 | /* Count bytes in string S between BEG/BEGBYTE and END. BEG and END |
1060 | are zero-based character positions in S, BEGBYTE is byte position | |
2cc21167 | 1061 | corresponding to BEG. UNIBYTE means S is a unibyte string. */ |
0a99eee1 | 1062 | static ptrdiff_t |
b7ad2f74 PE |
1063 | bidi_count_bytes (const unsigned char *s, ptrdiff_t beg, |
1064 | ptrdiff_t begbyte, ptrdiff_t end, bool unibyte) | |
87e67904 | 1065 | { |
d311d28c | 1066 | ptrdiff_t pos = beg; |
87e67904 EZ |
1067 | const unsigned char *p = s + begbyte, *start = p; |
1068 | ||
f3014ef5 EZ |
1069 | if (unibyte) |
1070 | p = s + end; | |
1071 | else | |
87e67904 | 1072 | { |
f3014ef5 | 1073 | if (!CHAR_HEAD_P (*p)) |
1088b922 | 1074 | emacs_abort (); |
f3014ef5 EZ |
1075 | |
1076 | while (pos < end) | |
1077 | { | |
1078 | p += BYTES_BY_CHAR_HEAD (*p); | |
1079 | pos++; | |
1080 | } | |
87e67904 EZ |
1081 | } |
1082 | ||
1083 | return p - start; | |
1084 | } | |
1085 | ||
05584c31 | 1086 | /* Fetch and return the character at byte position BYTEPOS. If S is |
87e67904 | 1087 | non-NULL, fetch the character from string S; otherwise fetch the |
2cc21167 | 1088 | character from the current buffer. UNIBYTE means S is a |
f3014ef5 | 1089 | unibyte string. */ |
0a99eee1 | 1090 | static int |
2cc21167 | 1091 | bidi_char_at_pos (ptrdiff_t bytepos, const unsigned char *s, bool unibyte) |
87e67904 EZ |
1092 | { |
1093 | if (s) | |
f3014ef5 | 1094 | { |
05584c31 | 1095 | s += bytepos; |
f3014ef5 | 1096 | if (unibyte) |
05584c31 | 1097 | return *s; |
f3014ef5 | 1098 | } |
87e67904 | 1099 | else |
05584c31 PE |
1100 | s = BYTE_POS_ADDR (bytepos); |
1101 | return STRING_CHAR (s); | |
87e67904 EZ |
1102 | } |
1103 | ||
36075a19 | 1104 | /* Fetch and return the character at CHARPOS/BYTEPOS. If that |
102ebb00 | 1105 | character is covered by a display string, treat the entire run of |
0c95fcf7 EZ |
1106 | covered characters as a single character, either u+2029 or u+FFFC, |
1107 | and return their combined length in CH_LEN and NCHARS. DISP_POS | |
1108 | specifies the character position of the next display string, or -1 | |
b75258b3 EZ |
1109 | if not yet computed. When the next character is at or beyond that |
1110 | position, the function updates DISP_POS with the position of the | |
2cc21167 | 1111 | next display string. *DISP_PROP non-zero means that there's really |
0c95fcf7 | 1112 | a display string at DISP_POS, as opposed to when we searched till |
2cc21167 | 1113 | DISP_POS without finding one. If *DISP_PROP is 2, it means the |
0c95fcf7 | 1114 | display spec is of the form `(space ...)', which is replaced with |
b75258b3 EZ |
1115 | u+2029 to handle it as a paragraph separator. STRING->s is the C |
1116 | string to iterate, or NULL if iterating over a buffer or a Lisp | |
1117 | string; in the latter case, STRING->lstring is the Lisp string. */ | |
0a99eee1 | 1118 | static int |
36075a19 | 1119 | bidi_fetch_char (ptrdiff_t charpos, ptrdiff_t bytepos, ptrdiff_t *disp_pos, |
0c95fcf7 | 1120 | int *disp_prop, struct bidi_string_data *string, |
5bf97bfc | 1121 | struct window *w, |
2cc21167 | 1122 | bool frame_window_p, ptrdiff_t *ch_len, ptrdiff_t *nchars) |
182ce2d2 EZ |
1123 | { |
1124 | int ch; | |
5895d7b9 | 1125 | ptrdiff_t endpos |
512a289d | 1126 | = (string->s || STRINGP (string->lstring)) ? string->schars : ZV; |
6db161be | 1127 | struct text_pos pos; |
e99a9b8b | 1128 | int len; |
182ce2d2 | 1129 | |
182ce2d2 | 1130 | /* If we got past the last known position of display string, compute |
87e67904 EZ |
1131 | the position of the next one. That position could be at CHARPOS. */ |
1132 | if (charpos < endpos && charpos > *disp_pos) | |
6db161be EZ |
1133 | { |
1134 | SET_TEXT_POS (pos, charpos, bytepos); | |
5bf97bfc | 1135 | *disp_pos = compute_display_string_pos (&pos, string, w, frame_window_p, |
0c95fcf7 | 1136 | disp_prop); |
6db161be | 1137 | } |
102ebb00 EZ |
1138 | |
1139 | /* Fetch the character at BYTEPOS. */ | |
87e67904 | 1140 | if (charpos >= endpos) |
182ce2d2 EZ |
1141 | { |
1142 | ch = BIDI_EOB; | |
1143 | *ch_len = 1; | |
1144 | *nchars = 1; | |
87e67904 | 1145 | *disp_pos = endpos; |
0c95fcf7 | 1146 | *disp_prop = 0; |
182ce2d2 | 1147 | } |
0c95fcf7 | 1148 | else if (charpos >= *disp_pos && *disp_prop) |
182ce2d2 | 1149 | { |
d311d28c | 1150 | ptrdiff_t disp_end_pos; |
7b600102 EZ |
1151 | |
1152 | /* We don't expect to find ourselves in the middle of a display | |
1153 | property. Hopefully, it will never be needed. */ | |
1154 | if (charpos > *disp_pos) | |
1088b922 | 1155 | emacs_abort (); |
0c95fcf7 EZ |
1156 | /* Text covered by `display' properties and overlays with |
1157 | display properties or display strings is handled as a single | |
1158 | character that represents the entire run of characters | |
1159 | covered by the display property. */ | |
1160 | if (*disp_prop == 2) | |
1161 | { | |
1162 | /* `(space ...)' display specs are handled as paragraph | |
1163 | separators for the purposes of the reordering; see UAX#9 | |
1164 | section 3 and clause HL1 in section 4.3 there. */ | |
1165 | ch = 0x2029; | |
1166 | } | |
1167 | else | |
1168 | { | |
1169 | /* All other display specs are handled as the Unicode Object | |
1170 | Replacement Character. */ | |
1171 | ch = 0xFFFC; | |
1172 | } | |
87e67904 | 1173 | disp_end_pos = compute_display_string_end (*disp_pos, string); |
fbcaa2f3 EZ |
1174 | if (disp_end_pos < 0) |
1175 | { | |
1176 | /* Somebody removed the display string from the buffer | |
1177 | behind our back. Recover by processing this buffer | |
1178 | position as if no display property were present there to | |
1179 | begin with. */ | |
1180 | *disp_prop = 0; | |
1181 | goto normal_char; | |
1182 | } | |
7b600102 | 1183 | *nchars = disp_end_pos - *disp_pos; |
f3014ef5 | 1184 | if (*nchars <= 0) |
1088b922 | 1185 | emacs_abort (); |
87e67904 EZ |
1186 | if (string->s) |
1187 | *ch_len = bidi_count_bytes (string->s, *disp_pos, bytepos, | |
f3014ef5 | 1188 | disp_end_pos, string->unibyte); |
9f257352 EZ |
1189 | else if (STRINGP (string->lstring)) |
1190 | *ch_len = bidi_count_bytes (SDATA (string->lstring), *disp_pos, | |
f3014ef5 | 1191 | bytepos, disp_end_pos, string->unibyte); |
87e67904 EZ |
1192 | else |
1193 | *ch_len = CHAR_TO_BYTE (disp_end_pos) - bytepos; | |
182ce2d2 | 1194 | } |
182ce2d2 EZ |
1195 | else |
1196 | { | |
fbcaa2f3 | 1197 | normal_char: |
87e67904 EZ |
1198 | if (string->s) |
1199 | { | |
87e67904 | 1200 | |
f3014ef5 EZ |
1201 | if (!string->unibyte) |
1202 | { | |
1203 | ch = STRING_CHAR_AND_LENGTH (string->s + bytepos, len); | |
1204 | *ch_len = len; | |
1205 | } | |
1206 | else | |
1207 | { | |
1208 | ch = UNIBYTE_TO_CHAR (string->s[bytepos]); | |
1209 | *ch_len = 1; | |
1210 | } | |
87e67904 | 1211 | } |
9f257352 EZ |
1212 | else if (STRINGP (string->lstring)) |
1213 | { | |
f3014ef5 EZ |
1214 | if (!string->unibyte) |
1215 | { | |
1216 | ch = STRING_CHAR_AND_LENGTH (SDATA (string->lstring) + bytepos, | |
1217 | len); | |
1218 | *ch_len = len; | |
1219 | } | |
1220 | else | |
1221 | { | |
1222 | ch = UNIBYTE_TO_CHAR (SREF (string->lstring, bytepos)); | |
1223 | *ch_len = 1; | |
1224 | } | |
9f257352 | 1225 | } |
87e67904 EZ |
1226 | else |
1227 | { | |
e99a9b8b EZ |
1228 | ch = STRING_CHAR_AND_LENGTH (BYTE_POS_ADDR (bytepos), len); |
1229 | *ch_len = len; | |
87e67904 | 1230 | } |
182ce2d2 EZ |
1231 | *nchars = 1; |
1232 | } | |
1233 | ||
1234 | /* If we just entered a run of characters covered by a display | |
1235 | string, compute the position of the next display string. */ | |
55439c61 | 1236 | if (charpos + *nchars <= endpos && charpos + *nchars > *disp_pos |
0c95fcf7 | 1237 | && *disp_prop) |
6db161be EZ |
1238 | { |
1239 | SET_TEXT_POS (pos, charpos + *nchars, bytepos + *ch_len); | |
5bf97bfc | 1240 | *disp_pos = compute_display_string_pos (&pos, string, w, frame_window_p, |
0c95fcf7 | 1241 | disp_prop); |
6db161be | 1242 | } |
182ce2d2 EZ |
1243 | |
1244 | return ch; | |
1245 | } | |
1246 | ||
cbb09f04 EZ |
1247 | \f |
1248 | /*********************************************************************** | |
1249 | Determining paragraph direction | |
1250 | ***********************************************************************/ | |
1251 | ||
1252 | /* Check if buffer position CHARPOS/BYTEPOS is the end of a paragraph. | |
1253 | Value is the non-negative length of the paragraph separator | |
1254 | following the buffer position, -1 if position is at the beginning | |
1255 | of a new paragraph, or -2 if position is neither at beginning nor | |
1256 | at end of a paragraph. */ | |
d311d28c PE |
1257 | static ptrdiff_t |
1258 | bidi_at_paragraph_end (ptrdiff_t charpos, ptrdiff_t bytepos) | |
cbb09f04 EZ |
1259 | { |
1260 | Lisp_Object sep_re; | |
1261 | Lisp_Object start_re; | |
d311d28c | 1262 | ptrdiff_t val; |
cbb09f04 EZ |
1263 | |
1264 | sep_re = paragraph_separate_re; | |
1265 | start_re = paragraph_start_re; | |
1266 | ||
1267 | val = fast_looking_at (sep_re, charpos, bytepos, ZV, ZV_BYTE, Qnil); | |
1268 | if (val < 0) | |
1269 | { | |
1270 | if (fast_looking_at (start_re, charpos, bytepos, ZV, ZV_BYTE, Qnil) >= 0) | |
1271 | val = -1; | |
1272 | else | |
1273 | val = -2; | |
1274 | } | |
1275 | ||
1276 | return val; | |
1277 | } | |
1278 | ||
e30b79c1 DA |
1279 | /* If the user has requested the long scans caching, make sure that |
1280 | BIDI cache is enabled. Otherwise, make sure it's disabled. */ | |
1281 | ||
1282 | static struct region_cache * | |
1283 | bidi_paragraph_cache_on_off (void) | |
1284 | { | |
c10e9ece EZ |
1285 | struct buffer *cache_buffer = current_buffer; |
1286 | bool indirect_p = false; | |
1287 | ||
1288 | /* For indirect buffers, make sure to use the cache of their base | |
1289 | buffer. */ | |
1290 | if (cache_buffer->base_buffer) | |
1291 | { | |
1292 | cache_buffer = cache_buffer->base_buffer; | |
1293 | indirect_p = true; | |
1294 | } | |
1295 | ||
1296 | /* Don't turn on or off the cache in the base buffer, if the value | |
1297 | of cache-long-scans of the base buffer is inconsistent with that. | |
1298 | This is because doing so will just make the cache pure overhead, | |
1299 | since if we turn it on via indirect buffer, it will be | |
1300 | immediately turned off by its base buffer. */ | |
e30b79c1 DA |
1301 | if (NILP (BVAR (current_buffer, cache_long_scans))) |
1302 | { | |
c10e9ece EZ |
1303 | if (!indirect_p |
1304 | || NILP (BVAR (cache_buffer, cache_long_scans))) | |
e30b79c1 | 1305 | { |
c10e9ece EZ |
1306 | if (cache_buffer->bidi_paragraph_cache) |
1307 | { | |
1308 | free_region_cache (cache_buffer->bidi_paragraph_cache); | |
1309 | cache_buffer->bidi_paragraph_cache = 0; | |
1310 | } | |
e30b79c1 DA |
1311 | } |
1312 | return NULL; | |
1313 | } | |
1314 | else | |
1315 | { | |
c10e9ece EZ |
1316 | if (!indirect_p |
1317 | || !NILP (BVAR (cache_buffer, cache_long_scans))) | |
1318 | { | |
1319 | if (!cache_buffer->bidi_paragraph_cache) | |
1320 | cache_buffer->bidi_paragraph_cache = new_region_cache (); | |
1321 | } | |
1322 | return cache_buffer->bidi_paragraph_cache; | |
e30b79c1 DA |
1323 | } |
1324 | } | |
1325 | ||
1137e8b8 EZ |
1326 | /* On my 2005-vintage machine, searching back for paragraph start |
1327 | takes ~1 ms per line. And bidi_paragraph_init is called 4 times | |
1328 | when user types C-p. The number below limits each call to | |
1329 | bidi_paragraph_init to about 10 ms. */ | |
1330 | #define MAX_PARAGRAPH_SEARCH 7500 | |
1331 | ||
be39f003 | 1332 | /* Find the beginning of this paragraph by looking back in the buffer. |
1137e8b8 EZ |
1333 | Value is the byte position of the paragraph's beginning, or |
1334 | BEGV_BYTE if paragraph_start_re is still not found after looking | |
1335 | back MAX_PARAGRAPH_SEARCH lines in the buffer. */ | |
d311d28c PE |
1336 | static ptrdiff_t |
1337 | bidi_find_paragraph_start (ptrdiff_t pos, ptrdiff_t pos_byte) | |
6bff6497 | 1338 | { |
372b7a95 | 1339 | Lisp_Object re = paragraph_start_re; |
d311d28c | 1340 | ptrdiff_t limit = ZV, limit_byte = ZV_BYTE; |
e30b79c1 DA |
1341 | struct region_cache *bpc = bidi_paragraph_cache_on_off (); |
1342 | ptrdiff_t n = 0, oldpos = pos, next; | |
c10e9ece EZ |
1343 | struct buffer *cache_buffer = current_buffer; |
1344 | ||
1345 | if (cache_buffer->base_buffer) | |
1346 | cache_buffer = cache_buffer->base_buffer; | |
6bff6497 | 1347 | |
6bff6497 | 1348 | while (pos_byte > BEGV_BYTE |
1137e8b8 | 1349 | && n++ < MAX_PARAGRAPH_SEARCH |
6bff6497 | 1350 | && fast_looking_at (re, pos, pos_byte, limit, limit_byte, Qnil) < 0) |
b5426561 DA |
1351 | { |
1352 | /* FIXME: What if the paragraph beginning is covered by a | |
1353 | display string? And what if a display string covering some | |
1354 | of the text over which we scan back includes | |
1355 | paragraph_start_re? */ | |
1356 | DEC_BOTH (pos, pos_byte); | |
c10e9ece | 1357 | if (bpc && region_cache_backward (cache_buffer, bpc, pos, &next)) |
e30b79c1 DA |
1358 | { |
1359 | pos = next, pos_byte = CHAR_TO_BYTE (pos); | |
1360 | break; | |
1361 | } | |
1362 | else | |
1363 | pos = find_newline_no_quit (pos, pos_byte, -1, &pos_byte); | |
b5426561 | 1364 | } |
1137e8b8 | 1365 | if (n >= MAX_PARAGRAPH_SEARCH) |
e30b79c1 DA |
1366 | pos = BEGV, pos_byte = BEGV_BYTE; |
1367 | if (bpc) | |
c10e9ece | 1368 | know_region_cache (cache_buffer, bpc, pos, oldpos); |
cd88d682 EZ |
1369 | /* Positions returned by the region cache are not limited to |
1370 | BEGV..ZV range, so we limit them here. */ | |
1371 | pos_byte = clip_to_bounds (BEGV_BYTE, pos_byte, ZV_BYTE); | |
be39f003 | 1372 | return pos_byte; |
6bff6497 EZ |
1373 | } |
1374 | ||
ce811ad9 EZ |
1375 | /* On a 3.4 GHz machine, searching forward for a strong directional |
1376 | character in a long paragraph full of weaks or neutrals takes about | |
1377 | 1 ms for each 20K characters. The number below limits each call to | |
1378 | bidi_paragraph_init to less than 10 ms even on slow machines. */ | |
1379 | #define MAX_STRONG_CHAR_SEARCH 100000 | |
1380 | ||
bea4f10c | 1381 | /* Determine the base direction, a.k.a. base embedding level, of the |
b44d9321 EZ |
1382 | paragraph we are about to iterate through. If DIR is either L2R or |
1383 | R2L, just use that. Otherwise, determine the paragraph direction | |
bea4f10c EZ |
1384 | from the first strong directional character of the paragraph. |
1385 | ||
2cc21167 | 1386 | NO_DEFAULT_P means don't default to L2R if the paragraph |
bea4f10c EZ |
1387 | has no strong directional characters and both DIR and |
1388 | bidi_it->paragraph_dir are NEUTRAL_DIR. In that case, search back | |
1389 | in the buffer until a paragraph is found with a strong character, | |
1390 | or until hitting BEGV. In the latter case, fall back to L2R. This | |
1391 | flag is used in current-bidi-paragraph-direction. | |
1392 | ||
1393 | Note that this function gives the paragraph separator the same | |
1394 | direction as the preceding paragraph, even though Emacs generally | |
9858f6c3 | 1395 | views the separator as not belonging to any paragraph. */ |
b7b65b15 | 1396 | void |
2cc21167 | 1397 | bidi_paragraph_init (bidi_dir_t dir, struct bidi_it *bidi_it, bool no_default_p) |
b7b65b15 | 1398 | { |
d311d28c | 1399 | ptrdiff_t bytepos = bidi_it->bytepos; |
2cc21167 | 1400 | bool string_p = bidi_it->string.s || STRINGP (bidi_it->string.lstring); |
d311d28c | 1401 | ptrdiff_t pstartbyte; |
87e67904 EZ |
1402 | /* Note that begbyte is a byte position, while end is a character |
1403 | position. Yes, this is ugly, but we are trying to avoid costly | |
1404 | calls to BYTE_TO_CHAR and its ilk. */ | |
d311d28c PE |
1405 | ptrdiff_t begbyte = string_p ? 0 : BEGV_BYTE; |
1406 | ptrdiff_t end = string_p ? bidi_it->string.schars : ZV; | |
e7402cb2 | 1407 | |
5e65aec0 | 1408 | /* Special case for an empty buffer. */ |
87e67904 | 1409 | if (bytepos == begbyte && bidi_it->charpos == end) |
5e65aec0 | 1410 | dir = L2R; |
9c82e145 | 1411 | /* We should never be called at EOB or before BEGV. */ |
87e67904 | 1412 | else if (bidi_it->charpos >= end || bytepos < begbyte) |
1088b922 | 1413 | emacs_abort (); |
9c82e145 | 1414 | |
be39f003 EZ |
1415 | if (dir == L2R) |
1416 | { | |
1417 | bidi_it->paragraph_dir = L2R; | |
1418 | bidi_it->new_paragraph = 0; | |
1419 | } | |
1420 | else if (dir == R2L) | |
1421 | { | |
1422 | bidi_it->paragraph_dir = R2L; | |
1423 | bidi_it->new_paragraph = 0; | |
1424 | } | |
b7b65b15 EZ |
1425 | else if (dir == NEUTRAL_DIR) /* P2 */ |
1426 | { | |
182ce2d2 | 1427 | int ch; |
d311d28c PE |
1428 | ptrdiff_t ch_len, nchars; |
1429 | ptrdiff_t pos, disp_pos = -1; | |
0c95fcf7 | 1430 | int disp_prop = 0; |
6bff6497 | 1431 | bidi_type_t type; |
9f257352 | 1432 | const unsigned char *s; |
be39f003 | 1433 | |
d20e1419 EZ |
1434 | if (!bidi_initialized) |
1435 | bidi_initialize (); | |
1436 | ||
be39f003 EZ |
1437 | /* If we are inside a paragraph separator, we are just waiting |
1438 | for the separator to be exhausted; use the previous paragraph | |
e5a2fec7 EZ |
1439 | direction. But don't do that if we have been just reseated, |
1440 | because we need to reinitialize below in that case. */ | |
1441 | if (!bidi_it->first_elt | |
1442 | && bidi_it->charpos < bidi_it->separator_limit) | |
be39f003 EZ |
1443 | return; |
1444 | ||
b44d9321 | 1445 | /* If we are on a newline, get past it to where the next |
5e65aec0 EZ |
1446 | paragraph might start. But don't do that at BEGV since then |
1447 | we are potentially in a new paragraph that doesn't yet | |
1448 | exist. */ | |
c143c213 | 1449 | pos = bidi_it->charpos; |
512a289d RS |
1450 | s = (STRINGP (bidi_it->string.lstring) |
1451 | ? SDATA (bidi_it->string.lstring) | |
1452 | : bidi_it->string.s); | |
f3014ef5 EZ |
1453 | if (bytepos > begbyte |
1454 | && bidi_char_at_pos (bytepos, s, bidi_it->string.unibyte) == '\n') | |
be39f003 | 1455 | { |
b44d9321 | 1456 | bytepos++; |
c143c213 | 1457 | pos++; |
be39f003 | 1458 | } |
b44d9321 EZ |
1459 | |
1460 | /* We are either at the beginning of a paragraph or in the | |
1461 | middle of it. Find where this paragraph starts. */ | |
87e67904 EZ |
1462 | if (string_p) |
1463 | { | |
1464 | /* We don't support changes of paragraph direction inside a | |
1465 | string. It is treated as a single paragraph. */ | |
1466 | pstartbyte = 0; | |
1467 | } | |
1468 | else | |
1469 | pstartbyte = bidi_find_paragraph_start (pos, bytepos); | |
be39f003 EZ |
1470 | bidi_it->separator_limit = -1; |
1471 | bidi_it->new_paragraph = 0; | |
bea4f10c | 1472 | |
2cc21167 PE |
1473 | /* The following loop is run more than once only if NO_DEFAULT_P, |
1474 | and only if we are iterating on a buffer. */ | |
bea4f10c | 1475 | do { |
ce811ad9 EZ |
1476 | ptrdiff_t pos1; |
1477 | ||
bea4f10c | 1478 | bytepos = pstartbyte; |
87e67904 EZ |
1479 | if (!string_p) |
1480 | pos = BYTE_TO_CHAR (bytepos); | |
36075a19 | 1481 | ch = bidi_fetch_char (pos, bytepos, &disp_pos, &disp_prop, |
5bf97bfc | 1482 | &bidi_it->string, bidi_it->w, |
87e67904 | 1483 | bidi_it->frame_window_p, &ch_len, &nchars); |
bea4f10c EZ |
1484 | type = bidi_get_type (ch, NEUTRAL_DIR); |
1485 | ||
ce811ad9 | 1486 | pos1 = pos; |
182ce2d2 | 1487 | for (pos += nchars, bytepos += ch_len; |
ce811ad9 EZ |
1488 | ((bidi_get_category (type) != STRONG) |
1489 | || (bidi_ignore_explicit_marks_for_paragraph_level | |
1490 | && (type == RLE || type == RLO | |
1491 | || type == LRE || type == LRO))) | |
1492 | /* Stop when searched too far into an abnormally large | |
1493 | paragraph full of weak or neutral characters. */ | |
1494 | && pos - pos1 < MAX_STRONG_CHAR_SEARCH; | |
bea4f10c EZ |
1495 | type = bidi_get_type (ch, NEUTRAL_DIR)) |
1496 | { | |
87e67904 | 1497 | if (pos >= end) |
bea4f10c EZ |
1498 | { |
1499 | /* Pretend there's a paragraph separator at end of | |
87e67904 | 1500 | buffer/string. */ |
bea4f10c EZ |
1501 | type = NEUTRAL_B; |
1502 | break; | |
1503 | } | |
0bb23927 EZ |
1504 | if (!string_p |
1505 | && type == NEUTRAL_B | |
1506 | && bidi_at_paragraph_end (pos, bytepos) >= -1) | |
cf99dcf8 | 1507 | break; |
102ebb00 | 1508 | /* Fetch next character and advance to get past it. */ |
36075a19 | 1509 | ch = bidi_fetch_char (pos, bytepos, &disp_pos, |
5bf97bfc | 1510 | &disp_prop, &bidi_it->string, bidi_it->w, |
fc6f18ce | 1511 | bidi_it->frame_window_p, &ch_len, &nchars); |
182ce2d2 EZ |
1512 | pos += nchars; |
1513 | bytepos += ch_len; | |
bea4f10c | 1514 | } |
32413314 EZ |
1515 | if ((type == STRONG_R || type == STRONG_AL) /* P3 */ |
1516 | || (!bidi_ignore_explicit_marks_for_paragraph_level | |
1517 | && (type == RLO || type == RLE))) | |
bea4f10c | 1518 | bidi_it->paragraph_dir = R2L; |
32413314 EZ |
1519 | else if (type == STRONG_L |
1520 | || (!bidi_ignore_explicit_marks_for_paragraph_level | |
1521 | && (type == LRO || type == LRE))) | |
bea4f10c | 1522 | bidi_it->paragraph_dir = L2R; |
87e67904 EZ |
1523 | if (!string_p |
1524 | && no_default_p && bidi_it->paragraph_dir == NEUTRAL_DIR) | |
bea4f10c EZ |
1525 | { |
1526 | /* If this paragraph is at BEGV, default to L2R. */ | |
1527 | if (pstartbyte == BEGV_BYTE) | |
1528 | bidi_it->paragraph_dir = L2R; /* P3 and HL1 */ | |
1529 | else | |
1530 | { | |
d311d28c PE |
1531 | ptrdiff_t prevpbyte = pstartbyte; |
1532 | ptrdiff_t p = BYTE_TO_CHAR (pstartbyte), pbyte = pstartbyte; | |
bea4f10c EZ |
1533 | |
1534 | /* Find the beginning of the previous paragraph, if any. */ | |
1535 | while (pbyte > BEGV_BYTE && prevpbyte >= pstartbyte) | |
1536 | { | |
182ce2d2 EZ |
1537 | /* FXIME: What if p is covered by a display |
1538 | string? See also a FIXME inside | |
1539 | bidi_find_paragraph_start. */ | |
36075a19 | 1540 | DEC_BOTH (p, pbyte); |
bea4f10c EZ |
1541 | prevpbyte = bidi_find_paragraph_start (p, pbyte); |
1542 | } | |
1543 | pstartbyte = prevpbyte; | |
1544 | } | |
1545 | } | |
87e67904 EZ |
1546 | } while (!string_p |
1547 | && no_default_p && bidi_it->paragraph_dir == NEUTRAL_DIR); | |
b7b65b15 | 1548 | } |
be39f003 | 1549 | else |
1088b922 | 1550 | emacs_abort (); |
b7b65b15 | 1551 | |
b44d9321 EZ |
1552 | /* Contrary to UAX#9 clause P3, we only default the paragraph |
1553 | direction to L2R if we have no previous usable paragraph | |
bea4f10c | 1554 | direction. This is allowed by the HL1 clause. */ |
d20e1419 | 1555 | if (bidi_it->paragraph_dir != L2R && bidi_it->paragraph_dir != R2L) |
bea4f10c | 1556 | bidi_it->paragraph_dir = L2R; /* P3 and HL1 ``higher-level protocols'' */ |
be39f003 | 1557 | if (bidi_it->paragraph_dir == R2L) |
b44d9321 | 1558 | bidi_it->level_stack[0].level = 1; |
be39f003 | 1559 | else |
b44d9321 | 1560 | bidi_it->level_stack[0].level = 0; |
be39f003 EZ |
1561 | |
1562 | bidi_line_init (bidi_it); | |
b7b65b15 EZ |
1563 | } |
1564 | ||
cbb09f04 EZ |
1565 | \f |
1566 | /*********************************************************************** | |
1567 | Resolving explicit and implicit levels. | |
58b9f433 | 1568 | The rest of this file constitutes the core of the UBA implementation. |
cbb09f04 | 1569 | ***********************************************************************/ |
b7b65b15 | 1570 | |
0a99eee1 | 1571 | static bool |
182ce2d2 | 1572 | bidi_explicit_dir_char (int ch) |
b7b65b15 | 1573 | { |
182ce2d2 EZ |
1574 | bidi_type_t ch_type; |
1575 | ||
1576 | if (!bidi_initialized) | |
1088b922 | 1577 | emacs_abort (); |
182ce2d2 EZ |
1578 | ch_type = (bidi_type_t) XINT (CHAR_TABLE_REF (bidi_type_table, ch)); |
1579 | return (ch_type == LRE || ch_type == LRO | |
1580 | || ch_type == RLE || ch_type == RLO | |
1581 | || ch_type == PDF); | |
b7b65b15 EZ |
1582 | } |
1583 | ||
1584 | /* A helper function for bidi_resolve_explicit. It advances to the | |
1585 | next character in logical order and determines the new embedding | |
1586 | level and directional override, but does not take into account | |
1587 | empty embeddings. */ | |
1588 | static int | |
1589 | bidi_resolve_explicit_1 (struct bidi_it *bidi_it) | |
1590 | { | |
1591 | int curchar; | |
1592 | bidi_type_t type; | |
1593 | int current_level; | |
1594 | int new_level; | |
1595 | bidi_dir_t override; | |
2cc21167 | 1596 | bool string_p = bidi_it->string.s || STRINGP (bidi_it->string.lstring); |
b7b65b15 | 1597 | |
182ce2d2 EZ |
1598 | /* If reseat()'ed, don't advance, so as to start iteration from the |
1599 | position where we were reseated. bidi_it->bytepos can be less | |
1600 | than BEGV_BYTE after reseat to BEGV. */ | |
87e67904 | 1601 | if (bidi_it->bytepos < (string_p ? 0 : BEGV_BYTE) |
9c82e145 | 1602 | || bidi_it->first_elt) |
e7402cb2 | 1603 | { |
9c82e145 | 1604 | bidi_it->first_elt = 0; |
87e67904 EZ |
1605 | if (string_p) |
1606 | { | |
512a289d RS |
1607 | const unsigned char *p |
1608 | = (STRINGP (bidi_it->string.lstring) | |
1609 | ? SDATA (bidi_it->string.lstring) | |
1610 | : bidi_it->string.s); | |
9f257352 | 1611 | |
87e67904 | 1612 | if (bidi_it->charpos < 0) |
05e193f1 EZ |
1613 | bidi_it->charpos = bidi_it->bytepos = 0; |
1614 | eassert (bidi_it->bytepos == bidi_count_bytes (p, 0, 0, | |
1615 | bidi_it->charpos, | |
1616 | bidi_it->string.unibyte)); | |
87e67904 EZ |
1617 | } |
1618 | else | |
1619 | { | |
1620 | if (bidi_it->charpos < BEGV) | |
05e193f1 EZ |
1621 | { |
1622 | bidi_it->charpos = BEGV; | |
1623 | bidi_it->bytepos = BEGV_BYTE; | |
1624 | } | |
1625 | eassert (bidi_it->bytepos == CHAR_TO_BYTE (bidi_it->charpos)); | |
87e67904 | 1626 | } |
e7402cb2 | 1627 | } |
87e67904 EZ |
1628 | /* Don't move at end of buffer/string. */ |
1629 | else if (bidi_it->charpos < (string_p ? bidi_it->string.schars : ZV)) | |
b7b65b15 | 1630 | { |
182ce2d2 EZ |
1631 | /* Advance to the next character, skipping characters covered by |
1632 | display strings (nchars > 1). */ | |
102ebb00 | 1633 | if (bidi_it->nchars <= 0) |
1088b922 | 1634 | emacs_abort (); |
182ce2d2 | 1635 | bidi_it->charpos += bidi_it->nchars; |
e7402cb2 | 1636 | if (bidi_it->ch_len == 0) |
1088b922 | 1637 | emacs_abort (); |
b7b65b15 EZ |
1638 | bidi_it->bytepos += bidi_it->ch_len; |
1639 | } | |
1640 | ||
1641 | current_level = bidi_it->level_stack[bidi_it->stack_idx].level; /* X1 */ | |
1642 | override = bidi_it->level_stack[bidi_it->stack_idx].override; | |
1643 | new_level = current_level; | |
1644 | ||
87e67904 | 1645 | if (bidi_it->charpos >= (string_p ? bidi_it->string.schars : ZV)) |
e7402cb2 EZ |
1646 | { |
1647 | curchar = BIDI_EOB; | |
1648 | bidi_it->ch_len = 1; | |
182ce2d2 | 1649 | bidi_it->nchars = 1; |
87e67904 | 1650 | bidi_it->disp_pos = (string_p ? bidi_it->string.schars : ZV); |
0c95fcf7 | 1651 | bidi_it->disp_prop = 0; |
e7402cb2 EZ |
1652 | } |
1653 | else | |
1654 | { | |
182ce2d2 EZ |
1655 | /* Fetch the character at BYTEPOS. If it is covered by a |
1656 | display string, treat the entire run of covered characters as | |
1657 | a single character u+FFFC. */ | |
36075a19 | 1658 | curchar = bidi_fetch_char (bidi_it->charpos, bidi_it->bytepos, |
0c95fcf7 | 1659 | &bidi_it->disp_pos, &bidi_it->disp_prop, |
5bf97bfc EZ |
1660 | &bidi_it->string, bidi_it->w, |
1661 | bidi_it->frame_window_p, | |
102ebb00 | 1662 | &bidi_it->ch_len, &bidi_it->nchars); |
e7402cb2 | 1663 | } |
b7b65b15 | 1664 | bidi_it->ch = curchar; |
b7b65b15 | 1665 | |
6bff6497 EZ |
1666 | /* Don't apply directional override here, as all the types we handle |
1667 | below will not be affected by the override anyway, and we need | |
1668 | the original type unaltered. The override will be applied in | |
1669 | bidi_resolve_weak. */ | |
1670 | type = bidi_get_type (curchar, NEUTRAL_DIR); | |
89d3374a EZ |
1671 | bidi_it->orig_type = type; |
1672 | bidi_check_type (bidi_it->orig_type); | |
b7b65b15 EZ |
1673 | |
1674 | if (type != PDF) | |
1675 | bidi_it->prev_was_pdf = 0; | |
1676 | ||
89d3374a | 1677 | bidi_it->type_after_w1 = UNKNOWN_BT; |
b7b65b15 EZ |
1678 | |
1679 | switch (type) | |
1680 | { | |
1681 | case RLE: /* X2 */ | |
1682 | case RLO: /* X4 */ | |
89d3374a EZ |
1683 | bidi_it->type_after_w1 = type; |
1684 | bidi_check_type (bidi_it->type_after_w1); | |
b7b65b15 | 1685 | type = WEAK_BN; /* X9/Retaining */ |
87e67904 | 1686 | if (bidi_it->ignore_bn_limit <= -1) |
b7b65b15 EZ |
1687 | { |
1688 | if (current_level <= BIDI_MAXLEVEL - 4) | |
1689 | { | |
1690 | /* Compute the least odd embedding level greater than | |
1691 | the current level. */ | |
1692 | new_level = ((current_level + 1) & ~1) + 1; | |
89d3374a | 1693 | if (bidi_it->type_after_w1 == RLE) |
b7b65b15 EZ |
1694 | override = NEUTRAL_DIR; |
1695 | else | |
1696 | override = R2L; | |
1697 | if (current_level == BIDI_MAXLEVEL - 4) | |
1698 | bidi_it->invalid_rl_levels = 0; | |
1699 | bidi_push_embedding_level (bidi_it, new_level, override); | |
1700 | } | |
1701 | else | |
1702 | { | |
1703 | bidi_it->invalid_levels++; | |
1704 | /* See the commentary about invalid_rl_levels below. */ | |
1705 | if (bidi_it->invalid_rl_levels < 0) | |
1706 | bidi_it->invalid_rl_levels = 0; | |
1707 | bidi_it->invalid_rl_levels++; | |
1708 | } | |
1709 | } | |
89d3374a | 1710 | else if (bidi_it->prev.type_after_w1 == WEAK_EN /* W5/Retaining */ |
7b5d6677 EZ |
1711 | || (bidi_it->next_en_pos > bidi_it->charpos |
1712 | && bidi_it->next_en_type == WEAK_EN)) | |
b7b65b15 EZ |
1713 | type = WEAK_EN; |
1714 | break; | |
1715 | case LRE: /* X3 */ | |
1716 | case LRO: /* X5 */ | |
89d3374a EZ |
1717 | bidi_it->type_after_w1 = type; |
1718 | bidi_check_type (bidi_it->type_after_w1); | |
b7b65b15 | 1719 | type = WEAK_BN; /* X9/Retaining */ |
87e67904 | 1720 | if (bidi_it->ignore_bn_limit <= -1) |
b7b65b15 EZ |
1721 | { |
1722 | if (current_level <= BIDI_MAXLEVEL - 5) | |
1723 | { | |
1724 | /* Compute the least even embedding level greater than | |
1725 | the current level. */ | |
1726 | new_level = ((current_level + 2) & ~1); | |
89d3374a | 1727 | if (bidi_it->type_after_w1 == LRE) |
b7b65b15 EZ |
1728 | override = NEUTRAL_DIR; |
1729 | else | |
1730 | override = L2R; | |
1731 | bidi_push_embedding_level (bidi_it, new_level, override); | |
1732 | } | |
1733 | else | |
1734 | { | |
1735 | bidi_it->invalid_levels++; | |
1736 | /* invalid_rl_levels counts invalid levels encountered | |
1737 | while the embedding level was already too high for | |
1738 | LRE/LRO, but not for RLE/RLO. That is because | |
1739 | there may be exactly one PDF which we should not | |
1740 | ignore even though invalid_levels is non-zero. | |
1741 | invalid_rl_levels helps to know what PDF is | |
1742 | that. */ | |
1743 | if (bidi_it->invalid_rl_levels >= 0) | |
1744 | bidi_it->invalid_rl_levels++; | |
1745 | } | |
1746 | } | |
89d3374a | 1747 | else if (bidi_it->prev.type_after_w1 == WEAK_EN /* W5/Retaining */ |
7b5d6677 EZ |
1748 | || (bidi_it->next_en_pos > bidi_it->charpos |
1749 | && bidi_it->next_en_type == WEAK_EN)) | |
b7b65b15 EZ |
1750 | type = WEAK_EN; |
1751 | break; | |
1752 | case PDF: /* X7 */ | |
89d3374a EZ |
1753 | bidi_it->type_after_w1 = type; |
1754 | bidi_check_type (bidi_it->type_after_w1); | |
b7b65b15 | 1755 | type = WEAK_BN; /* X9/Retaining */ |
87e67904 | 1756 | if (bidi_it->ignore_bn_limit <= -1) |
b7b65b15 EZ |
1757 | { |
1758 | if (!bidi_it->invalid_rl_levels) | |
1759 | { | |
1760 | new_level = bidi_pop_embedding_level (bidi_it); | |
1761 | bidi_it->invalid_rl_levels = -1; | |
1762 | if (bidi_it->invalid_levels) | |
1763 | bidi_it->invalid_levels--; | |
1764 | /* else nothing: UAX#9 says to ignore invalid PDFs */ | |
1765 | } | |
1766 | if (!bidi_it->invalid_levels) | |
1767 | new_level = bidi_pop_embedding_level (bidi_it); | |
1768 | else | |
1769 | { | |
1770 | bidi_it->invalid_levels--; | |
1771 | bidi_it->invalid_rl_levels--; | |
1772 | } | |
1773 | } | |
89d3374a | 1774 | else if (bidi_it->prev.type_after_w1 == WEAK_EN /* W5/Retaining */ |
7b5d6677 EZ |
1775 | || (bidi_it->next_en_pos > bidi_it->charpos |
1776 | && bidi_it->next_en_type == WEAK_EN)) | |
b7b65b15 EZ |
1777 | type = WEAK_EN; |
1778 | break; | |
1779 | default: | |
1780 | /* Nothing. */ | |
1781 | break; | |
1782 | } | |
1783 | ||
1784 | bidi_it->type = type; | |
2d6e4628 | 1785 | bidi_check_type (bidi_it->type); |
b7b65b15 EZ |
1786 | |
1787 | return new_level; | |
1788 | } | |
1789 | ||
1790 | /* Given an iterator state in BIDI_IT, advance one character position | |
182ce2d2 EZ |
1791 | in the buffer/string to the next character (in the logical order), |
1792 | resolve any explicit embeddings and directional overrides, and | |
1793 | return the embedding level of the character after resolving | |
1794 | explicit directives and ignoring empty embeddings. */ | |
b7b65b15 EZ |
1795 | static int |
1796 | bidi_resolve_explicit (struct bidi_it *bidi_it) | |
1797 | { | |
1798 | int prev_level = bidi_it->level_stack[bidi_it->stack_idx].level; | |
1799 | int new_level = bidi_resolve_explicit_1 (bidi_it); | |
d311d28c | 1800 | ptrdiff_t eob = bidi_it->string.s ? bidi_it->string.schars : ZV; |
512a289d RS |
1801 | const unsigned char *s |
1802 | = (STRINGP (bidi_it->string.lstring) | |
1803 | ? SDATA (bidi_it->string.lstring) | |
1804 | : bidi_it->string.s); | |
b7b65b15 EZ |
1805 | |
1806 | if (prev_level < new_level | |
1807 | && bidi_it->type == WEAK_BN | |
87e67904 EZ |
1808 | && bidi_it->ignore_bn_limit == -1 /* only if not already known */ |
1809 | && bidi_it->charpos < eob /* not already at EOB */ | |
1810 | && bidi_explicit_dir_char (bidi_char_at_pos (bidi_it->bytepos | |
f3014ef5 EZ |
1811 | + bidi_it->ch_len, s, |
1812 | bidi_it->string.unibyte))) | |
b7b65b15 EZ |
1813 | { |
1814 | /* Avoid pushing and popping embedding levels if the level run | |
1815 | is empty, as this breaks level runs where it shouldn't. | |
1816 | UAX#9 removes all the explicit embedding and override codes, | |
1817 | so empty embeddings disappear without a trace. We need to | |
1818 | behave as if we did the same. */ | |
1819 | struct bidi_it saved_it; | |
1820 | int level = prev_level; | |
1821 | ||
1822 | bidi_copy_it (&saved_it, bidi_it); | |
1823 | ||
87e67904 | 1824 | while (bidi_explicit_dir_char (bidi_char_at_pos (bidi_it->bytepos |
f3014ef5 EZ |
1825 | + bidi_it->ch_len, s, |
1826 | bidi_it->string.unibyte))) | |
b7b65b15 | 1827 | { |
182ce2d2 EZ |
1828 | /* This advances to the next character, skipping any |
1829 | characters covered by display strings. */ | |
b7b65b15 | 1830 | level = bidi_resolve_explicit_1 (bidi_it); |
9f257352 EZ |
1831 | /* If string.lstring was relocated inside bidi_resolve_explicit_1, |
1832 | a pointer to its data is no longer valid. */ | |
1833 | if (STRINGP (bidi_it->string.lstring)) | |
1834 | s = SDATA (bidi_it->string.lstring); | |
b7b65b15 EZ |
1835 | } |
1836 | ||
102ebb00 | 1837 | if (bidi_it->nchars <= 0) |
1088b922 | 1838 | emacs_abort (); |
b7b65b15 | 1839 | if (level == prev_level) /* empty embedding */ |
182ce2d2 | 1840 | saved_it.ignore_bn_limit = bidi_it->charpos + bidi_it->nchars; |
b7b65b15 | 1841 | else /* this embedding is non-empty */ |
87e67904 | 1842 | saved_it.ignore_bn_limit = -2; |
b7b65b15 EZ |
1843 | |
1844 | bidi_copy_it (bidi_it, &saved_it); | |
87e67904 | 1845 | if (bidi_it->ignore_bn_limit > -1) |
b7b65b15 EZ |
1846 | { |
1847 | /* We pushed a level, but we shouldn't have. Undo that. */ | |
1848 | if (!bidi_it->invalid_rl_levels) | |
1849 | { | |
1850 | new_level = bidi_pop_embedding_level (bidi_it); | |
1851 | bidi_it->invalid_rl_levels = -1; | |
1852 | if (bidi_it->invalid_levels) | |
1853 | bidi_it->invalid_levels--; | |
1854 | } | |
1855 | if (!bidi_it->invalid_levels) | |
1856 | new_level = bidi_pop_embedding_level (bidi_it); | |
1857 | else | |
1858 | { | |
1859 | bidi_it->invalid_levels--; | |
1860 | bidi_it->invalid_rl_levels--; | |
1861 | } | |
1862 | } | |
1863 | } | |
1864 | ||
b7b65b15 EZ |
1865 | if (bidi_it->type == NEUTRAL_B) /* X8 */ |
1866 | { | |
21fce5ab | 1867 | bidi_set_paragraph_end (bidi_it); |
6bff6497 EZ |
1868 | /* This is needed by bidi_resolve_weak below, and in L1. */ |
1869 | bidi_it->type_after_w1 = bidi_it->type; | |
89d3374a | 1870 | bidi_check_type (bidi_it->type_after_w1); |
b7b65b15 EZ |
1871 | } |
1872 | ||
1873 | return new_level; | |
1874 | } | |
1875 | ||
182ce2d2 EZ |
1876 | /* Advance in the buffer/string, resolve weak types and return the |
1877 | type of the next character after weak type resolution. */ | |
fd3998ff | 1878 | static bidi_type_t |
b7b65b15 EZ |
1879 | bidi_resolve_weak (struct bidi_it *bidi_it) |
1880 | { | |
1881 | bidi_type_t type; | |
1882 | bidi_dir_t override; | |
1883 | int prev_level = bidi_it->level_stack[bidi_it->stack_idx].level; | |
1884 | int new_level = bidi_resolve_explicit (bidi_it); | |
1885 | int next_char; | |
1886 | bidi_type_t type_of_next; | |
1887 | struct bidi_it saved_it; | |
5895d7b9 | 1888 | ptrdiff_t eob |
512a289d RS |
1889 | = ((STRINGP (bidi_it->string.lstring) || bidi_it->string.s) |
1890 | ? bidi_it->string.schars : ZV); | |
b7b65b15 EZ |
1891 | |
1892 | type = bidi_it->type; | |
1893 | override = bidi_it->level_stack[bidi_it->stack_idx].override; | |
1894 | ||
1895 | if (type == UNKNOWN_BT | |
1896 | || type == LRE | |
1897 | || type == LRO | |
1898 | || type == RLE | |
1899 | || type == RLO | |
1900 | || type == PDF) | |
1088b922 | 1901 | emacs_abort (); |
b7b65b15 EZ |
1902 | |
1903 | if (new_level != prev_level | |
1904 | || bidi_it->type == NEUTRAL_B) | |
1905 | { | |
1906 | /* We've got a new embedding level run, compute the directional | |
1907 | type of sor and initialize per-run variables (UAX#9, clause | |
1908 | X10). */ | |
1909 | bidi_set_sor_type (bidi_it, prev_level, new_level); | |
1910 | } | |
1911 | else if (type == NEUTRAL_S || type == NEUTRAL_WS | |
1912 | || type == WEAK_BN || type == STRONG_AL) | |
89d3374a EZ |
1913 | bidi_it->type_after_w1 = type; /* needed in L1 */ |
1914 | bidi_check_type (bidi_it->type_after_w1); | |
b7b65b15 EZ |
1915 | |
1916 | /* Level and directional override status are already recorded in | |
1917 | bidi_it, and do not need any change; see X6. */ | |
1918 | if (override == R2L) /* X6 */ | |
1919 | type = STRONG_R; | |
1920 | else if (override == L2R) | |
1921 | type = STRONG_L; | |
bc5a45f3 | 1922 | else |
b7b65b15 | 1923 | { |
bc5a45f3 | 1924 | if (type == WEAK_NSM) /* W1 */ |
b7b65b15 | 1925 | { |
bc5a45f3 | 1926 | /* Note that we don't need to consider the case where the |
5930fe97 EZ |
1927 | prev character has its type overridden by an RLO or LRO, |
1928 | because then either the type of this NSM would have been | |
1929 | also overridden, or the previous character is outside the | |
1930 | current level run, and thus not relevant to this NSM. | |
1931 | This is why NSM gets the type_after_w1 of the previous | |
1932 | character. */ | |
ebb5722e EZ |
1933 | if (bidi_it->prev.type_after_w1 != UNKNOWN_BT |
1934 | /* if type_after_w1 is NEUTRAL_B, this NSM is at sor */ | |
1935 | && bidi_it->prev.type_after_w1 != NEUTRAL_B) | |
5930fe97 | 1936 | type = bidi_it->prev.type_after_w1; |
bc5a45f3 EZ |
1937 | else if (bidi_it->sor == R2L) |
1938 | type = STRONG_R; | |
1939 | else if (bidi_it->sor == L2R) | |
1940 | type = STRONG_L; | |
1941 | else /* shouldn't happen! */ | |
1088b922 | 1942 | emacs_abort (); |
b7b65b15 | 1943 | } |
bc5a45f3 EZ |
1944 | if (type == WEAK_EN /* W2 */ |
1945 | && bidi_it->last_strong.type_after_w1 == STRONG_AL) | |
1946 | type = WEAK_AN; | |
1947 | else if (type == STRONG_AL) /* W3 */ | |
1948 | type = STRONG_R; | |
1949 | else if ((type == WEAK_ES /* W4 */ | |
1950 | && bidi_it->prev.type_after_w1 == WEAK_EN | |
1951 | && bidi_it->prev.orig_type == WEAK_EN) | |
1952 | || (type == WEAK_CS | |
1953 | && ((bidi_it->prev.type_after_w1 == WEAK_EN | |
1954 | && bidi_it->prev.orig_type == WEAK_EN) | |
1955 | || bidi_it->prev.type_after_w1 == WEAK_AN))) | |
b7b65b15 | 1956 | { |
512a289d RS |
1957 | const unsigned char *s |
1958 | = (STRINGP (bidi_it->string.lstring) | |
1959 | ? SDATA (bidi_it->string.lstring) | |
1960 | : bidi_it->string.s); | |
1961 | ||
1962 | next_char = (bidi_it->charpos + bidi_it->nchars >= eob | |
1963 | ? BIDI_EOB | |
1964 | : bidi_char_at_pos (bidi_it->bytepos + bidi_it->ch_len, | |
1965 | s, bidi_it->string.unibyte)); | |
6bff6497 | 1966 | type_of_next = bidi_get_type (next_char, override); |
b7b65b15 | 1967 | |
bc5a45f3 | 1968 | if (type_of_next == WEAK_BN |
b7b65b15 EZ |
1969 | || bidi_explicit_dir_char (next_char)) |
1970 | { | |
1971 | bidi_copy_it (&saved_it, bidi_it); | |
1972 | while (bidi_resolve_explicit (bidi_it) == new_level | |
bc5a45f3 | 1973 | && bidi_it->type == WEAK_BN) |
b7b65b15 EZ |
1974 | ; |
1975 | type_of_next = bidi_it->type; | |
b7b65b15 EZ |
1976 | bidi_copy_it (bidi_it, &saved_it); |
1977 | } | |
bc5a45f3 EZ |
1978 | |
1979 | /* If the next character is EN, but the last strong-type | |
1980 | character is AL, that next EN will be changed to AN when | |
1981 | we process it in W2 above. So in that case, this ES | |
1982 | should not be changed into EN. */ | |
1983 | if (type == WEAK_ES | |
1984 | && type_of_next == WEAK_EN | |
1985 | && bidi_it->last_strong.type_after_w1 != STRONG_AL) | |
1986 | type = WEAK_EN; | |
1987 | else if (type == WEAK_CS) | |
b7b65b15 | 1988 | { |
bc5a45f3 EZ |
1989 | if (bidi_it->prev.type_after_w1 == WEAK_AN |
1990 | && (type_of_next == WEAK_AN | |
1991 | /* If the next character is EN, but the last | |
1992 | strong-type character is AL, EN will be later | |
1993 | changed to AN when we process it in W2 above. | |
1994 | So in that case, this ES should not be | |
1995 | changed into EN. */ | |
1996 | || (type_of_next == WEAK_EN | |
1997 | && bidi_it->last_strong.type_after_w1 == STRONG_AL))) | |
1998 | type = WEAK_AN; | |
1999 | else if (bidi_it->prev.type_after_w1 == WEAK_EN | |
2000 | && type_of_next == WEAK_EN | |
2001 | && bidi_it->last_strong.type_after_w1 != STRONG_AL) | |
2002 | type = WEAK_EN; | |
2003 | } | |
2004 | } | |
2005 | else if (type == WEAK_ET /* W5: ET with EN before or after it */ | |
2006 | || type == WEAK_BN) /* W5/Retaining */ | |
2007 | { | |
7b5d6677 | 2008 | if (bidi_it->prev.type_after_w1 == WEAK_EN) /* ET/BN w/EN before it */ |
bc5a45f3 | 2009 | type = WEAK_EN; |
7b5d6677 EZ |
2010 | else if (bidi_it->next_en_pos > bidi_it->charpos |
2011 | && bidi_it->next_en_type != WEAK_BN) | |
2012 | { | |
2013 | if (bidi_it->next_en_type == WEAK_EN) /* ET/BN with EN after it */ | |
2014 | type = WEAK_EN; | |
2015 | } | |
2016 | else if (bidi_it->next_en_pos >=0) | |
bc5a45f3 | 2017 | { |
d311d28c | 2018 | ptrdiff_t en_pos = bidi_it->charpos + bidi_it->nchars; |
512a289d RS |
2019 | const unsigned char *s = (STRINGP (bidi_it->string.lstring) |
2020 | ? SDATA (bidi_it->string.lstring) | |
2021 | : bidi_it->string.s); | |
bc5a45f3 | 2022 | |
102ebb00 | 2023 | if (bidi_it->nchars <= 0) |
1088b922 | 2024 | emacs_abort (); |
512a289d RS |
2025 | next_char |
2026 | = (bidi_it->charpos + bidi_it->nchars >= eob | |
2027 | ? BIDI_EOB | |
2028 | : bidi_char_at_pos (bidi_it->bytepos + bidi_it->ch_len, s, | |
2029 | bidi_it->string.unibyte)); | |
bc5a45f3 EZ |
2030 | type_of_next = bidi_get_type (next_char, override); |
2031 | ||
2032 | if (type_of_next == WEAK_ET | |
2033 | || type_of_next == WEAK_BN | |
2034 | || bidi_explicit_dir_char (next_char)) | |
2035 | { | |
2036 | bidi_copy_it (&saved_it, bidi_it); | |
2037 | while (bidi_resolve_explicit (bidi_it) == new_level | |
2038 | && (bidi_it->type == WEAK_BN | |
2039 | || bidi_it->type == WEAK_ET)) | |
2040 | ; | |
2041 | type_of_next = bidi_it->type; | |
2042 | en_pos = bidi_it->charpos; | |
2043 | bidi_copy_it (bidi_it, &saved_it); | |
2044 | } | |
7b5d6677 EZ |
2045 | /* Remember this position, to speed up processing of the |
2046 | next ETs. */ | |
2047 | bidi_it->next_en_pos = en_pos; | |
bc5a45f3 | 2048 | if (type_of_next == WEAK_EN) |
b7b65b15 | 2049 | { |
bc5a45f3 EZ |
2050 | /* If the last strong character is AL, the EN we've |
2051 | found will become AN when we get to it (W2). */ | |
7b5d6677 EZ |
2052 | if (bidi_it->last_strong.type_after_w1 == STRONG_AL) |
2053 | type_of_next = WEAK_AN; | |
bc5a45f3 EZ |
2054 | else if (type == WEAK_BN) |
2055 | type = NEUTRAL_ON; /* W6/Retaining */ | |
7b5d6677 EZ |
2056 | else |
2057 | type = WEAK_EN; | |
b7b65b15 | 2058 | } |
4787455f EZ |
2059 | else if (type_of_next == NEUTRAL_B) |
2060 | /* Record the fact that there are no more ENs from | |
2061 | here to the end of paragraph, to avoid entering the | |
2062 | loop above ever again in this paragraph. */ | |
2063 | bidi_it->next_en_pos = -1; | |
7b5d6677 EZ |
2064 | /* Record the type of the character where we ended our search. */ |
2065 | bidi_it->next_en_type = type_of_next; | |
b7b65b15 EZ |
2066 | } |
2067 | } | |
2068 | } | |
2069 | ||
2070 | if (type == WEAK_ES || type == WEAK_ET || type == WEAK_CS /* W6 */ | |
89d3374a EZ |
2071 | || (type == WEAK_BN |
2072 | && (bidi_it->prev.type_after_w1 == WEAK_CS /* W6/Retaining */ | |
2073 | || bidi_it->prev.type_after_w1 == WEAK_ES | |
2074 | || bidi_it->prev.type_after_w1 == WEAK_ET))) | |
b7b65b15 EZ |
2075 | type = NEUTRAL_ON; |
2076 | ||
2077 | /* Store the type we've got so far, before we clobber it with strong | |
2078 | types in W7 and while resolving neutral types. But leave alone | |
2079 | the original types that were recorded above, because we will need | |
2080 | them for the L1 clause. */ | |
89d3374a EZ |
2081 | if (bidi_it->type_after_w1 == UNKNOWN_BT) |
2082 | bidi_it->type_after_w1 = type; | |
2083 | bidi_check_type (bidi_it->type_after_w1); | |
b7b65b15 EZ |
2084 | |
2085 | if (type == WEAK_EN) /* W7 */ | |
2086 | { | |
89d3374a | 2087 | if ((bidi_it->last_strong.type_after_w1 == STRONG_L) |
b7b65b15 EZ |
2088 | || (bidi_it->last_strong.type == UNKNOWN_BT && bidi_it->sor == L2R)) |
2089 | type = STRONG_L; | |
2090 | } | |
2091 | ||
2092 | bidi_it->type = type; | |
2d6e4628 | 2093 | bidi_check_type (bidi_it->type); |
b7b65b15 EZ |
2094 | return type; |
2095 | } | |
2096 | ||
cbb09f04 EZ |
2097 | /* Resolve the type of a neutral character according to the type of |
2098 | surrounding strong text and the current embedding level. */ | |
0a99eee1 | 2099 | static bidi_type_t |
cbb09f04 EZ |
2100 | bidi_resolve_neutral_1 (bidi_type_t prev_type, bidi_type_t next_type, int lev) |
2101 | { | |
2102 | /* N1: European and Arabic numbers are treated as though they were R. */ | |
2103 | if (next_type == WEAK_EN || next_type == WEAK_AN) | |
2104 | next_type = STRONG_R; | |
2105 | if (prev_type == WEAK_EN || prev_type == WEAK_AN) | |
2106 | prev_type = STRONG_R; | |
2107 | ||
2108 | if (next_type == prev_type) /* N1 */ | |
2109 | return next_type; | |
2110 | else if ((lev & 1) == 0) /* N2 */ | |
2111 | return STRONG_L; | |
2112 | else | |
2113 | return STRONG_R; | |
2114 | } | |
2115 | ||
fd3998ff | 2116 | static bidi_type_t |
b7b65b15 EZ |
2117 | bidi_resolve_neutral (struct bidi_it *bidi_it) |
2118 | { | |
2119 | int prev_level = bidi_it->level_stack[bidi_it->stack_idx].level; | |
2120 | bidi_type_t type = bidi_resolve_weak (bidi_it); | |
2121 | int current_level = bidi_it->level_stack[bidi_it->stack_idx].level; | |
2122 | ||
2123 | if (!(type == STRONG_R | |
2124 | || type == STRONG_L | |
2125 | || type == WEAK_BN | |
2126 | || type == WEAK_EN | |
2127 | || type == WEAK_AN | |
2128 | || type == NEUTRAL_B | |
2129 | || type == NEUTRAL_S | |
2130 | || type == NEUTRAL_WS | |
2131 | || type == NEUTRAL_ON)) | |
1088b922 | 2132 | emacs_abort (); |
b7b65b15 | 2133 | |
4787455f EZ |
2134 | if ((type != NEUTRAL_B /* Don't risk entering the long loop below if |
2135 | we are already at paragraph end. */ | |
2136 | && bidi_get_category (type) == NEUTRAL) | |
b7b65b15 EZ |
2137 | || (type == WEAK_BN && prev_level == current_level)) |
2138 | { | |
2139 | if (bidi_it->next_for_neutral.type != UNKNOWN_BT) | |
2140 | type = bidi_resolve_neutral_1 (bidi_it->prev_for_neutral.type, | |
2141 | bidi_it->next_for_neutral.type, | |
2142 | current_level); | |
4787455f EZ |
2143 | /* The next two "else if" clauses are shortcuts for the |
2144 | important special case when we have a long sequence of | |
2145 | neutral or WEAK_BN characters, such as whitespace or nulls or | |
2146 | other control characters, on the base embedding level of the | |
2147 | paragraph, and that sequence goes all the way to the end of | |
2148 | the paragraph and follows a character whose resolved | |
2149 | directionality is identical to the base embedding level. | |
2150 | (This is what happens in a buffer with plain L2R text that | |
2151 | happens to include long sequences of control characters.) By | |
2152 | virtue of N1, the result of examining this long sequence will | |
2153 | always be either STRONG_L or STRONG_R, depending on the base | |
2154 | embedding level. So we use this fact directly instead of | |
2155 | entering the expensive loop in the "else" clause. */ | |
2156 | else if (current_level == 0 | |
2157 | && bidi_it->prev_for_neutral.type == STRONG_L | |
2158 | && !bidi_explicit_dir_char (bidi_it->ch)) | |
2159 | type = bidi_resolve_neutral_1 (bidi_it->prev_for_neutral.type, | |
2160 | STRONG_L, current_level); | |
2161 | else if (/* current level is 1 */ | |
2162 | current_level == 1 | |
2163 | /* base embedding level is also 1 */ | |
2164 | && bidi_it->level_stack[0].level == 1 | |
2165 | /* previous character is one of those considered R for | |
2166 | the purposes of W5 */ | |
2167 | && (bidi_it->prev_for_neutral.type == STRONG_R | |
2168 | || bidi_it->prev_for_neutral.type == WEAK_EN | |
2169 | || bidi_it->prev_for_neutral.type == WEAK_AN) | |
2170 | && !bidi_explicit_dir_char (bidi_it->ch)) | |
2171 | type = bidi_resolve_neutral_1 (bidi_it->prev_for_neutral.type, | |
2172 | STRONG_R, current_level); | |
b7b65b15 EZ |
2173 | else |
2174 | { | |
2175 | /* Arrrgh!! The UAX#9 algorithm is too deeply entrenched in | |
2176 | the assumption of batch-style processing; see clauses W4, | |
2177 | W5, and especially N1, which require to look far forward | |
182ce2d2 EZ |
2178 | (as well as back) in the buffer/string. May the fleas of |
2179 | a thousand camels infest the armpits of those who design | |
b7b65b15 EZ |
2180 | supposedly general-purpose algorithms by looking at their |
2181 | own implementations, and fail to consider other possible | |
2182 | implementations! */ | |
2183 | struct bidi_it saved_it; | |
2184 | bidi_type_t next_type; | |
2185 | ||
2186 | if (bidi_it->scan_dir == -1) | |
1088b922 | 2187 | emacs_abort (); |
b7b65b15 EZ |
2188 | |
2189 | bidi_copy_it (&saved_it, bidi_it); | |
2190 | /* Scan the text forward until we find the first non-neutral | |
2191 | character, and then use that to resolve the neutral we | |
2192 | are dealing with now. We also cache the scanned iterator | |
2193 | states, to salvage some of the effort later. */ | |
2194 | bidi_cache_iterator_state (bidi_it, 0); | |
2195 | do { | |
2196 | /* Record the info about the previous character, so that | |
2197 | it will be cached below with this state. */ | |
89d3374a | 2198 | if (bidi_it->type_after_w1 != WEAK_BN /* W1/Retaining */ |
b7b65b15 EZ |
2199 | && bidi_it->type != WEAK_BN) |
2200 | bidi_remember_char (&bidi_it->prev, bidi_it); | |
2201 | type = bidi_resolve_weak (bidi_it); | |
2202 | /* Paragraph separators have their levels fully resolved | |
2203 | at this point, so cache them as resolved. */ | |
2204 | bidi_cache_iterator_state (bidi_it, type == NEUTRAL_B); | |
2205 | /* FIXME: implement L1 here, by testing for a newline and | |
2206 | resetting the level for any sequence of whitespace | |
2207 | characters adjacent to it. */ | |
2208 | } while (!(type == NEUTRAL_B | |
2209 | || (type != WEAK_BN | |
2210 | && bidi_get_category (type) != NEUTRAL) | |
2211 | /* This is all per level run, so stop when we | |
2212 | reach the end of this level run. */ | |
512a289d RS |
2213 | || (bidi_it->level_stack[bidi_it->stack_idx].level |
2214 | != current_level))); | |
b7b65b15 EZ |
2215 | |
2216 | bidi_remember_char (&saved_it.next_for_neutral, bidi_it); | |
2217 | ||
2218 | switch (type) | |
2219 | { | |
2220 | case STRONG_L: | |
2221 | case STRONG_R: | |
2222 | case STRONG_AL: | |
4787455f EZ |
2223 | /* Actually, STRONG_AL cannot happen here, because |
2224 | bidi_resolve_weak converts it to STRONG_R, per W3. */ | |
a54e2c05 | 2225 | eassert (type != STRONG_AL); |
b7b65b15 EZ |
2226 | next_type = type; |
2227 | break; | |
2228 | case WEAK_EN: | |
2229 | case WEAK_AN: | |
2230 | /* N1: ``European and Arabic numbers are treated as | |
2231 | though they were R.'' */ | |
2232 | next_type = STRONG_R; | |
b7b65b15 EZ |
2233 | break; |
2234 | case WEAK_BN: | |
713bfeaa | 2235 | case NEUTRAL_ON: /* W6/Retaining */ |
b7b65b15 | 2236 | if (!bidi_explicit_dir_char (bidi_it->ch)) |
1088b922 | 2237 | emacs_abort (); /* can't happen: BNs are skipped */ |
b7b65b15 EZ |
2238 | /* FALLTHROUGH */ |
2239 | case NEUTRAL_B: | |
2240 | /* Marched all the way to the end of this level run. | |
2241 | We need to use the eor type, whose information is | |
2242 | stored by bidi_set_sor_type in the prev_for_neutral | |
2243 | member. */ | |
2244 | if (saved_it.type != WEAK_BN | |
89d3374a | 2245 | || bidi_get_category (bidi_it->prev.type_after_w1) == NEUTRAL) |
4787455f | 2246 | next_type = bidi_it->prev_for_neutral.type; |
b7b65b15 EZ |
2247 | else |
2248 | { | |
2249 | /* This is a BN which does not adjoin neutrals. | |
2250 | Leave its type alone. */ | |
2251 | bidi_copy_it (bidi_it, &saved_it); | |
2252 | return bidi_it->type; | |
2253 | } | |
2254 | break; | |
2255 | default: | |
1088b922 | 2256 | emacs_abort (); |
b7b65b15 EZ |
2257 | } |
2258 | type = bidi_resolve_neutral_1 (saved_it.prev_for_neutral.type, | |
2259 | next_type, current_level); | |
4787455f | 2260 | saved_it.next_for_neutral.type = next_type; |
b7b65b15 | 2261 | saved_it.type = type; |
4787455f | 2262 | bidi_check_type (next_type); |
2d6e4628 | 2263 | bidi_check_type (type); |
b7b65b15 EZ |
2264 | bidi_copy_it (bidi_it, &saved_it); |
2265 | } | |
2266 | } | |
2267 | return type; | |
2268 | } | |
2269 | ||
2270 | /* Given an iterator state in BIDI_IT, advance one character position | |
182ce2d2 EZ |
2271 | in the buffer/string to the next character (in the logical order), |
2272 | resolve the bidi type of that next character, and return that | |
2273 | type. */ | |
fd3998ff | 2274 | static bidi_type_t |
b7b65b15 EZ |
2275 | bidi_type_of_next_char (struct bidi_it *bidi_it) |
2276 | { | |
2277 | bidi_type_t type; | |
2278 | ||
2279 | /* This should always be called during a forward scan. */ | |
2280 | if (bidi_it->scan_dir != 1) | |
1088b922 | 2281 | emacs_abort (); |
b7b65b15 EZ |
2282 | |
2283 | /* Reset the limit until which to ignore BNs if we step out of the | |
2284 | area where we found only empty levels. */ | |
87e67904 | 2285 | if ((bidi_it->ignore_bn_limit > -1 |
b7b65b15 | 2286 | && bidi_it->ignore_bn_limit <= bidi_it->charpos) |
87e67904 | 2287 | || (bidi_it->ignore_bn_limit == -2 |
b7b65b15 | 2288 | && !bidi_explicit_dir_char (bidi_it->ch))) |
87e67904 | 2289 | bidi_it->ignore_bn_limit = -1; |
b7b65b15 EZ |
2290 | |
2291 | type = bidi_resolve_neutral (bidi_it); | |
2292 | ||
2293 | return type; | |
2294 | } | |
2295 | ||
2296 | /* Given an iterator state BIDI_IT, advance one character position in | |
182ce2d2 EZ |
2297 | the buffer/string to the next character (in the current scan |
2298 | direction), resolve the embedding and implicit levels of that next | |
2299 | character, and return the resulting level. */ | |
fd3998ff | 2300 | static int |
b7b65b15 EZ |
2301 | bidi_level_of_next_char (struct bidi_it *bidi_it) |
2302 | { | |
2303 | bidi_type_t type; | |
2304 | int level, prev_level = -1; | |
2305 | struct bidi_saved_info next_for_neutral; | |
d311d28c | 2306 | ptrdiff_t next_char_pos = -2; |
b7b65b15 EZ |
2307 | |
2308 | if (bidi_it->scan_dir == 1) | |
2309 | { | |
5895d7b9 | 2310 | ptrdiff_t eob |
512a289d RS |
2311 | = ((bidi_it->string.s || STRINGP (bidi_it->string.lstring)) |
2312 | ? bidi_it->string.schars : ZV); | |
9f257352 | 2313 | |
b7b65b15 | 2314 | /* There's no sense in trying to advance if we hit end of text. */ |
9f257352 | 2315 | if (bidi_it->charpos >= eob) |
b7b65b15 EZ |
2316 | return bidi_it->resolved_level; |
2317 | ||
2318 | /* Record the info about the previous character. */ | |
89d3374a | 2319 | if (bidi_it->type_after_w1 != WEAK_BN /* W1/Retaining */ |
b7b65b15 EZ |
2320 | && bidi_it->type != WEAK_BN) |
2321 | bidi_remember_char (&bidi_it->prev, bidi_it); | |
89d3374a EZ |
2322 | if (bidi_it->type_after_w1 == STRONG_R |
2323 | || bidi_it->type_after_w1 == STRONG_L | |
2324 | || bidi_it->type_after_w1 == STRONG_AL) | |
b7b65b15 EZ |
2325 | bidi_remember_char (&bidi_it->last_strong, bidi_it); |
2326 | /* FIXME: it sounds like we don't need both prev and | |
2327 | prev_for_neutral members, but I'm leaving them both for now. */ | |
2328 | if (bidi_it->type == STRONG_R || bidi_it->type == STRONG_L | |
2329 | || bidi_it->type == WEAK_EN || bidi_it->type == WEAK_AN) | |
2330 | bidi_remember_char (&bidi_it->prev_for_neutral, bidi_it); | |
2331 | ||
2332 | /* If we overstepped the characters used for resolving neutrals | |
2333 | and whitespace, invalidate their info in the iterator. */ | |
2334 | if (bidi_it->charpos >= bidi_it->next_for_neutral.charpos) | |
2335 | bidi_it->next_for_neutral.type = UNKNOWN_BT; | |
2336 | if (bidi_it->next_en_pos >= 0 | |
2337 | && bidi_it->charpos >= bidi_it->next_en_pos) | |
7b5d6677 EZ |
2338 | { |
2339 | bidi_it->next_en_pos = 0; | |
2340 | bidi_it->next_en_type = UNKNOWN_BT; | |
2341 | } | |
b7b65b15 EZ |
2342 | if (bidi_it->next_for_ws.type != UNKNOWN_BT |
2343 | && bidi_it->charpos >= bidi_it->next_for_ws.charpos) | |
2344 | bidi_it->next_for_ws.type = UNKNOWN_BT; | |
2345 | ||
2346 | /* This must be taken before we fill the iterator with the info | |
2347 | about the next char. If we scan backwards, the iterator | |
2348 | state must be already cached, so there's no need to know the | |
2349 | embedding level of the previous character, since we will be | |
2350 | returning to our caller shortly. */ | |
2351 | prev_level = bidi_it->level_stack[bidi_it->stack_idx].level; | |
2352 | } | |
2353 | next_for_neutral = bidi_it->next_for_neutral; | |
2354 | ||
182ce2d2 EZ |
2355 | /* Perhaps the character we want is already cached. If it is, the |
2356 | call to bidi_cache_find below will return a type other than | |
2357 | UNKNOWN_BT. */ | |
87e67904 | 2358 | if (bidi_cache_idx > bidi_cache_start && !bidi_it->first_elt) |
102ebb00 | 2359 | { |
512a289d RS |
2360 | int bob = ((bidi_it->string.s || STRINGP (bidi_it->string.lstring)) |
2361 | ? 0 : 1); | |
102ebb00 EZ |
2362 | if (bidi_it->scan_dir > 0) |
2363 | { | |
2364 | if (bidi_it->nchars <= 0) | |
1088b922 | 2365 | emacs_abort (); |
102ebb00 EZ |
2366 | next_char_pos = bidi_it->charpos + bidi_it->nchars; |
2367 | } | |
9f257352 | 2368 | else if (bidi_it->charpos >= bob) |
bb269206 EZ |
2369 | /* Implementation note: we allow next_char_pos to be as low as |
2370 | 0 for buffers or -1 for strings, and that is okay because | |
2371 | that's the "position" of the sentinel iterator state we | |
2372 | cached at the beginning of the iteration. */ | |
102ebb00 | 2373 | next_char_pos = bidi_it->charpos - 1; |
578b494e | 2374 | if (next_char_pos >= bob - 1) |
87e67904 EZ |
2375 | type = bidi_cache_find (next_char_pos, -1, bidi_it); |
2376 | else | |
2377 | type = UNKNOWN_BT; | |
102ebb00 | 2378 | } |
182ce2d2 | 2379 | else |
102ebb00 | 2380 | type = UNKNOWN_BT; |
b7b65b15 EZ |
2381 | if (type != UNKNOWN_BT) |
2382 | { | |
2383 | /* Don't lose the information for resolving neutrals! The | |
2384 | cached states could have been cached before their | |
2385 | next_for_neutral member was computed. If we are on our way | |
2386 | forward, we can simply take the info from the previous | |
2387 | state. */ | |
2388 | if (bidi_it->scan_dir == 1 | |
2389 | && bidi_it->next_for_neutral.type == UNKNOWN_BT) | |
2390 | bidi_it->next_for_neutral = next_for_neutral; | |
2391 | ||
2392 | /* If resolved_level is -1, it means this state was cached | |
2393 | before it was completely resolved, so we cannot return | |
2394 | it. */ | |
2395 | if (bidi_it->resolved_level != -1) | |
2396 | return bidi_it->resolved_level; | |
2397 | } | |
2398 | if (bidi_it->scan_dir == -1) | |
2399 | /* If we are going backwards, the iterator state is already cached | |
2400 | from previous scans, and should be fully resolved. */ | |
1088b922 | 2401 | emacs_abort (); |
b7b65b15 EZ |
2402 | |
2403 | if (type == UNKNOWN_BT) | |
2404 | type = bidi_type_of_next_char (bidi_it); | |
2405 | ||
2406 | if (type == NEUTRAL_B) | |
2407 | return bidi_it->resolved_level; | |
2408 | ||
2409 | level = bidi_it->level_stack[bidi_it->stack_idx].level; | |
2410 | if ((bidi_get_category (type) == NEUTRAL /* && type != NEUTRAL_B */) | |
2411 | || (type == WEAK_BN && prev_level == level)) | |
2412 | { | |
2413 | if (bidi_it->next_for_neutral.type == UNKNOWN_BT) | |
1088b922 | 2414 | emacs_abort (); |
b7b65b15 EZ |
2415 | |
2416 | /* If the cached state shows a neutral character, it was not | |
2417 | resolved by bidi_resolve_neutral, so do it now. */ | |
2418 | type = bidi_resolve_neutral_1 (bidi_it->prev_for_neutral.type, | |
2419 | bidi_it->next_for_neutral.type, | |
2420 | level); | |
2421 | } | |
2422 | ||
2423 | if (!(type == STRONG_R | |
2424 | || type == STRONG_L | |
2425 | || type == WEAK_BN | |
2426 | || type == WEAK_EN | |
2427 | || type == WEAK_AN)) | |
1088b922 | 2428 | emacs_abort (); |
b7b65b15 | 2429 | bidi_it->type = type; |
2d6e4628 | 2430 | bidi_check_type (bidi_it->type); |
b7b65b15 EZ |
2431 | |
2432 | /* For L1 below, we need to know, for each WS character, whether | |
0d327994 | 2433 | it belongs to a sequence of WS characters preceding a newline |
b7b65b15 | 2434 | or a TAB or a paragraph separator. */ |
89d3374a | 2435 | if (bidi_it->orig_type == NEUTRAL_WS |
b7b65b15 EZ |
2436 | && bidi_it->next_for_ws.type == UNKNOWN_BT) |
2437 | { | |
2438 | int ch; | |
d311d28c PE |
2439 | ptrdiff_t clen = bidi_it->ch_len; |
2440 | ptrdiff_t bpos = bidi_it->bytepos; | |
2441 | ptrdiff_t cpos = bidi_it->charpos; | |
2442 | ptrdiff_t disp_pos = bidi_it->disp_pos; | |
2443 | ptrdiff_t nc = bidi_it->nchars; | |
87e67904 | 2444 | struct bidi_string_data bs = bidi_it->string; |
b7b65b15 | 2445 | bidi_type_t chtype; |
2cc21167 | 2446 | bool fwp = bidi_it->frame_window_p; |
0c95fcf7 | 2447 | int dpp = bidi_it->disp_prop; |
b7b65b15 | 2448 | |
102ebb00 | 2449 | if (bidi_it->nchars <= 0) |
1088b922 | 2450 | emacs_abort (); |
b7b65b15 | 2451 | do { |
36075a19 | 2452 | ch = bidi_fetch_char (cpos += nc, bpos += clen, &disp_pos, &dpp, &bs, |
5bf97bfc | 2453 | bidi_it->w, fwp, &clen, &nc); |
79beb178 | 2454 | if (ch == '\n' || ch == BIDI_EOB) |
b7b65b15 EZ |
2455 | chtype = NEUTRAL_B; |
2456 | else | |
6bff6497 | 2457 | chtype = bidi_get_type (ch, NEUTRAL_DIR); |
b7b65b15 EZ |
2458 | } while (chtype == NEUTRAL_WS || chtype == WEAK_BN |
2459 | || bidi_explicit_dir_char (ch)); /* L1/Retaining */ | |
2460 | bidi_it->next_for_ws.type = chtype; | |
2d6e4628 | 2461 | bidi_check_type (bidi_it->next_for_ws.type); |
b7b65b15 EZ |
2462 | bidi_it->next_for_ws.charpos = cpos; |
2463 | bidi_it->next_for_ws.bytepos = bpos; | |
2464 | } | |
2465 | ||
2466 | /* Resolve implicit levels, with a twist: PDFs get the embedding | |
55622b93 | 2467 | level of the embedding they terminate. See below for the |
b7b65b15 | 2468 | reason. */ |
89d3374a | 2469 | if (bidi_it->orig_type == PDF |
b7b65b15 EZ |
2470 | /* Don't do this if this formatting code didn't change the |
2471 | embedding level due to invalid or empty embeddings. */ | |
2472 | && prev_level != level) | |
2473 | { | |
2474 | /* Don't look in UAX#9 for the reason for this: it's our own | |
2475 | private quirk. The reason is that we want the formatting | |
2476 | codes to be delivered so that they bracket the text of their | |
2477 | embedding. For example, given the text | |
2478 | ||
2479 | {RLO}teST{PDF} | |
2480 | ||
2481 | we want it to be displayed as | |
2482 | ||
0d68907d | 2483 | {PDF}STet{RLO} |
b7b65b15 EZ |
2484 | |
2485 | not as | |
2486 | ||
2487 | STet{RLO}{PDF} | |
2488 | ||
2489 | which will result because we bump up the embedding level as | |
2490 | soon as we see the RLO and pop it as soon as we see the PDF, | |
2491 | so RLO itself has the same embedding level as "teST", and | |
2492 | thus would be normally delivered last, just before the PDF. | |
2493 | The switch below fiddles with the level of PDF so that this | |
2494 | ugly side effect does not happen. | |
2495 | ||
2496 | (This is, of course, only important if the formatting codes | |
e7402cb2 EZ |
2497 | are actually displayed, but Emacs does need to display them |
2498 | if the user wants to.) */ | |
b7b65b15 EZ |
2499 | level = prev_level; |
2500 | } | |
89d3374a EZ |
2501 | else if (bidi_it->orig_type == NEUTRAL_B /* L1 */ |
2502 | || bidi_it->orig_type == NEUTRAL_S | |
e7402cb2 | 2503 | || bidi_it->ch == '\n' || bidi_it->ch == BIDI_EOB |
89d3374a | 2504 | || (bidi_it->orig_type == NEUTRAL_WS |
b7b65b15 EZ |
2505 | && (bidi_it->next_for_ws.type == NEUTRAL_B |
2506 | || bidi_it->next_for_ws.type == NEUTRAL_S))) | |
2507 | level = bidi_it->level_stack[0].level; | |
2508 | else if ((level & 1) == 0) /* I1 */ | |
2509 | { | |
2510 | if (type == STRONG_R) | |
2511 | level++; | |
2512 | else if (type == WEAK_EN || type == WEAK_AN) | |
2513 | level += 2; | |
2514 | } | |
2515 | else /* I2 */ | |
2516 | { | |
2517 | if (type == STRONG_L || type == WEAK_EN || type == WEAK_AN) | |
2518 | level++; | |
2519 | } | |
2520 | ||
2521 | bidi_it->resolved_level = level; | |
2522 | return level; | |
2523 | } | |
2524 | ||
2cc21167 PE |
2525 | /* Move to the other edge of a level given by LEVEL. If END_FLAG, |
2526 | we are at the end of a level, and we need to prepare to | |
b7b65b15 EZ |
2527 | resume the scan of the lower level. |
2528 | ||
2529 | If this level's other edge is cached, we simply jump to it, filling | |
2530 | the iterator structure with the iterator state on the other edge. | |
182ce2d2 EZ |
2531 | Otherwise, we walk the buffer or string until we come back to the |
2532 | same level as LEVEL. | |
b7b65b15 EZ |
2533 | |
2534 | Note: we are not talking here about a ``level run'' in the UAX#9 | |
2535 | sense of the term, but rather about a ``level'' which includes | |
2536 | all the levels higher than it. In other words, given the levels | |
2537 | like this: | |
2538 | ||
2539 | 11111112222222333333334443343222222111111112223322111 | |
2540 | A B C | |
2541 | ||
2542 | and assuming we are at point A scanning left to right, this | |
2543 | function moves to point C, whereas the UAX#9 ``level 2 run'' ends | |
2544 | at point B. */ | |
2545 | static void | |
2cc21167 | 2546 | bidi_find_other_level_edge (struct bidi_it *bidi_it, int level, bool end_flag) |
b7b65b15 EZ |
2547 | { |
2548 | int dir = end_flag ? -bidi_it->scan_dir : bidi_it->scan_dir; | |
39e378da | 2549 | ptrdiff_t idx; |
b7b65b15 EZ |
2550 | |
2551 | /* Try the cache first. */ | |
87e67904 EZ |
2552 | if ((idx = bidi_cache_find_level_change (level, dir, end_flag)) |
2553 | >= bidi_cache_start) | |
b7b65b15 EZ |
2554 | bidi_cache_fetch_state (idx, bidi_it); |
2555 | else | |
2556 | { | |
2557 | int new_level; | |
2558 | ||
1088b922 | 2559 | /* If we are at end of level, its edges must be cached. */ |
b7b65b15 | 2560 | if (end_flag) |
1088b922 | 2561 | emacs_abort (); |
b7b65b15 EZ |
2562 | |
2563 | bidi_cache_iterator_state (bidi_it, 1); | |
2564 | do { | |
2565 | new_level = bidi_level_of_next_char (bidi_it); | |
2566 | bidi_cache_iterator_state (bidi_it, 1); | |
2567 | } while (new_level >= level); | |
2568 | } | |
2569 | } | |
2570 | ||
2571 | void | |
4b292a22 | 2572 | bidi_move_to_visually_next (struct bidi_it *bidi_it) |
b7b65b15 EZ |
2573 | { |
2574 | int old_level, new_level, next_level; | |
9c82e145 | 2575 | struct bidi_it sentinel; |
acb28818 | 2576 | struct gcpro gcpro1; |
b7b65b15 | 2577 | |
87e67904 | 2578 | if (bidi_it->charpos < 0 || bidi_it->bytepos < 0) |
1088b922 | 2579 | emacs_abort (); |
b7b65b15 EZ |
2580 | |
2581 | if (bidi_it->scan_dir == 0) | |
2582 | { | |
2583 | bidi_it->scan_dir = 1; /* default to logical order */ | |
2584 | } | |
2585 | ||
acb28818 | 2586 | /* The code below can call eval, and thus cause GC. If we are |
7e2ad32c | 2587 | iterating a Lisp string, make sure it won't be GCed. */ |
acb28818 EZ |
2588 | if (STRINGP (bidi_it->string.lstring)) |
2589 | GCPRO1 (bidi_it->string.lstring); | |
2590 | ||
be39f003 | 2591 | /* If we just passed a newline, initialize for the next line. */ |
1137e8b8 EZ |
2592 | if (!bidi_it->first_elt |
2593 | && (bidi_it->ch == '\n' || bidi_it->ch == BIDI_EOB)) | |
be39f003 EZ |
2594 | bidi_line_init (bidi_it); |
2595 | ||
6dcfd253 EZ |
2596 | /* Prepare the sentinel iterator state, and cache it. When we bump |
2597 | into it, scanning backwards, we'll know that the last non-base | |
2598 | level is exhausted. */ | |
87e67904 | 2599 | if (bidi_cache_idx == bidi_cache_start) |
9c82e145 EZ |
2600 | { |
2601 | bidi_copy_it (&sentinel, bidi_it); | |
2602 | if (bidi_it->first_elt) | |
2603 | { | |
2604 | sentinel.charpos--; /* cached charpos needs to be monotonic */ | |
2605 | sentinel.bytepos--; | |
2606 | sentinel.ch = '\n'; /* doesn't matter, but why not? */ | |
2607 | sentinel.ch_len = 1; | |
182ce2d2 | 2608 | sentinel.nchars = 1; |
9c82e145 | 2609 | } |
6dcfd253 | 2610 | bidi_cache_iterator_state (&sentinel, 1); |
9c82e145 | 2611 | } |
b7b65b15 EZ |
2612 | |
2613 | old_level = bidi_it->resolved_level; | |
2614 | new_level = bidi_level_of_next_char (bidi_it); | |
b7b65b15 EZ |
2615 | |
2616 | /* Reordering of resolved levels (clause L2) is implemented by | |
2617 | jumping to the other edge of the level and flipping direction of | |
c0546589 | 2618 | scanning the text whenever we find a level change. */ |
b7b65b15 EZ |
2619 | if (new_level != old_level) |
2620 | { | |
2cc21167 | 2621 | bool ascending = new_level > old_level; |
b7b65b15 EZ |
2622 | int level_to_search = ascending ? old_level + 1 : old_level; |
2623 | int incr = ascending ? 1 : -1; | |
2624 | int expected_next_level = old_level + incr; | |
2625 | ||
b7b65b15 EZ |
2626 | /* Jump (or walk) to the other edge of this level. */ |
2627 | bidi_find_other_level_edge (bidi_it, level_to_search, !ascending); | |
2628 | /* Switch scan direction and peek at the next character in the | |
2629 | new direction. */ | |
2630 | bidi_it->scan_dir = -bidi_it->scan_dir; | |
2631 | ||
2632 | /* The following loop handles the case where the resolved level | |
2633 | jumps by more than one. This is typical for numbers inside a | |
2634 | run of text with left-to-right embedding direction, but can | |
2635 | also happen in other situations. In those cases the decision | |
2636 | where to continue after a level change, and in what direction, | |
2637 | is tricky. For example, given a text like below: | |
2638 | ||
2639 | abcdefgh | |
2640 | 11336622 | |
2641 | ||
2642 | (where the numbers below the text show the resolved levels), | |
2643 | the result of reordering according to UAX#9 should be this: | |
2644 | ||
2645 | efdcghba | |
2646 | ||
2647 | This is implemented by the loop below which flips direction | |
2648 | and jumps to the other edge of the level each time it finds | |
2649 | the new level not to be the expected one. The expected level | |
2650 | is always one more or one less than the previous one. */ | |
2651 | next_level = bidi_peek_at_next_level (bidi_it); | |
2652 | while (next_level != expected_next_level) | |
2653 | { | |
713bfeaa EZ |
2654 | /* If next_level is -1, it means we have an unresolved level |
2655 | in the cache, which at this point should not happen. If | |
2656 | it does, we will infloop. */ | |
2657 | eassert (next_level >= 0); | |
b7b65b15 EZ |
2658 | expected_next_level += incr; |
2659 | level_to_search += incr; | |
2660 | bidi_find_other_level_edge (bidi_it, level_to_search, !ascending); | |
2661 | bidi_it->scan_dir = -bidi_it->scan_dir; | |
2662 | next_level = bidi_peek_at_next_level (bidi_it); | |
2663 | } | |
2664 | ||
2665 | /* Finally, deliver the next character in the new direction. */ | |
2666 | next_level = bidi_level_of_next_char (bidi_it); | |
2667 | } | |
2668 | ||
b44d9321 EZ |
2669 | /* Take note when we have just processed the newline that precedes |
2670 | the end of the paragraph. The next time we are about to be | |
2671 | called, set_iterator_to_next will automatically reinit the | |
2672 | paragraph direction, if needed. We do this at the newline before | |
2673 | the paragraph separator, because the next character might not be | |
2674 | the first character of the next paragraph, due to the bidi | |
c0546589 EZ |
2675 | reordering, whereas we _must_ know the paragraph base direction |
2676 | _before_ we process the paragraph's text, since the base | |
2677 | direction affects the reordering. */ | |
1137e8b8 EZ |
2678 | if (bidi_it->scan_dir == 1 |
2679 | && (bidi_it->ch == '\n' || bidi_it->ch == BIDI_EOB)) | |
be39f003 | 2680 | { |
87e67904 EZ |
2681 | /* The paragraph direction of the entire string, once |
2682 | determined, is in effect for the entire string. Setting the | |
2683 | separator limit to the end of the string prevents | |
2684 | bidi_paragraph_init from being called automatically on this | |
2685 | string. */ | |
9f257352 | 2686 | if (bidi_it->string.s || STRINGP (bidi_it->string.lstring)) |
87e67904 EZ |
2687 | bidi_it->separator_limit = bidi_it->string.schars; |
2688 | else if (bidi_it->bytepos < ZV_BYTE) | |
be39f003 | 2689 | { |
5895d7b9 | 2690 | ptrdiff_t sep_len |
512a289d RS |
2691 | = bidi_at_paragraph_end (bidi_it->charpos + bidi_it->nchars, |
2692 | bidi_it->bytepos + bidi_it->ch_len); | |
87e67904 | 2693 | if (bidi_it->nchars <= 0) |
1088b922 | 2694 | emacs_abort (); |
87e67904 EZ |
2695 | if (sep_len >= 0) |
2696 | { | |
2697 | bidi_it->new_paragraph = 1; | |
2698 | /* Record the buffer position of the last character of the | |
2699 | paragraph separator. */ | |
512a289d RS |
2700 | bidi_it->separator_limit |
2701 | = bidi_it->charpos + bidi_it->nchars + sep_len; | |
87e67904 | 2702 | } |
be39f003 EZ |
2703 | } |
2704 | } | |
6bff6497 | 2705 | |
87e67904 | 2706 | if (bidi_it->scan_dir == 1 && bidi_cache_idx > bidi_cache_start) |
b7b65b15 EZ |
2707 | { |
2708 | /* If we are at paragraph's base embedding level and beyond the | |
2709 | last cached position, the cache's job is done and we can | |
2710 | discard it. */ | |
2711 | if (bidi_it->resolved_level == bidi_it->level_stack[0].level | |
182ce2d2 EZ |
2712 | && bidi_it->charpos > (bidi_cache[bidi_cache_idx - 1].charpos |
2713 | + bidi_cache[bidi_cache_idx - 1].nchars - 1)) | |
b7b65b15 EZ |
2714 | bidi_cache_reset (); |
2715 | /* But as long as we are caching during forward scan, we must | |
2716 | cache each state, or else the cache integrity will be | |
2717 | compromised: it assumes cached states correspond to buffer | |
2718 | positions 1:1. */ | |
2719 | else | |
2720 | bidi_cache_iterator_state (bidi_it, 1); | |
2721 | } | |
acb28818 EZ |
2722 | |
2723 | if (STRINGP (bidi_it->string.lstring)) | |
2724 | UNGCPRO; | |
b7b65b15 EZ |
2725 | } |
2726 | ||
2727 | /* This is meant to be called from within the debugger, whenever you | |
2728 | wish to examine the cache contents. */ | |
e78aecca | 2729 | void bidi_dump_cached_states (void) EXTERNALLY_VISIBLE; |
b7b65b15 EZ |
2730 | void |
2731 | bidi_dump_cached_states (void) | |
2732 | { | |
39e378da | 2733 | ptrdiff_t i; |
b7b65b15 EZ |
2734 | int ndigits = 1; |
2735 | ||
2736 | if (bidi_cache_idx == 0) | |
2737 | { | |
2738 | fprintf (stderr, "The cache is empty.\n"); | |
2739 | return; | |
2740 | } | |
df9733bf | 2741 | fprintf (stderr, "Total of %"pD"d state%s in cache:\n", |
b7b65b15 EZ |
2742 | bidi_cache_idx, bidi_cache_idx == 1 ? "" : "s"); |
2743 | ||
2744 | for (i = bidi_cache[bidi_cache_idx - 1].charpos; i > 0; i /= 10) | |
2745 | ndigits++; | |
2746 | fputs ("ch ", stderr); | |
2747 | for (i = 0; i < bidi_cache_idx; i++) | |
2748 | fprintf (stderr, "%*c", ndigits, bidi_cache[i].ch); | |
2749 | fputs ("\n", stderr); | |
2750 | fputs ("lvl ", stderr); | |
2751 | for (i = 0; i < bidi_cache_idx; i++) | |
2752 | fprintf (stderr, "%*d", ndigits, bidi_cache[i].resolved_level); | |
2753 | fputs ("\n", stderr); | |
2754 | fputs ("pos ", stderr); | |
2755 | for (i = 0; i < bidi_cache_idx; i++) | |
7c85f529 | 2756 | fprintf (stderr, "%*"pD"d", ndigits, bidi_cache[i].charpos); |
b7b65b15 EZ |
2757 | fputs ("\n", stderr); |
2758 | } |