1 /* Low-level bidirectional buffer-scanning functions for GNU Emacs.
2 Copyright (C) 2000, 2001, 2004, 2005, 2009, 2010
3 Free Software Foundation, Inc.
5 This file is part of GNU Emacs.
7 GNU Emacs is free software: you can redistribute it and/or modify
8 it under the terms of the GNU General Public License as published by
9 the Free Software Foundation, either version 3 of the License, or
10 (at your option) any later version.
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.
17 You should have received a copy of the GNU General Public License
18 along with GNU Emacs. If not, see <http://www.gnu.org/licenses/>. */
20 /* Written by Eli Zaretskii <eliz@gnu.org>.
22 A sequential implementation of the Unicode Bidirectional algorithm,
23 as per UAX#9, a part of the Unicode Standard.
25 Unlike the reference and most other implementations, this one is
26 designed to be called once for every character in the buffer or
29 The main entry point is bidi_move_to_visually_next. Each time it
30 is called, it finds the next character in the visual order, and
31 returns its information in a special structure. The caller is then
32 expected to process this character for display or any other
33 purposes, and call bidi_move_to_visually_next for the next
34 character. See the comments in bidi_move_to_visually_next for more
35 details about its algorithm that finds the next visual-order
36 character by resolving their levels on the fly.
38 The two other entry points are bidi_paragraph_init and
39 bidi_mirror_char. The first determines the base direction of a
40 paragraph, while the second returns the mirrored version of its
43 If you want to understand the code, you will have to read it
44 together with the relevant portions of UAX#9. The comments include
45 references to UAX#9 rules, for that very reason.
47 A note about references to UAX#9 rules: if the reference says
48 something like "X9/Retaining", it means that you need to refer to
49 rule X9 and to its modifications decribed in the "Implementation
50 Notes" section of UAX#9, under "Retaining Format Codes". */
66 #include "character.h"
67 #include "dispextern.h"
69 static int bidi_initialized
= 0;
71 static Lisp_Object bidi_type_table
;
73 /* FIXME: Remove these when bidi_explicit_dir_char uses a lookup table. */
74 #define LRM_CHAR 0x200E
75 #define RLM_CHAR 0x200F
76 #define LRE_CHAR 0x202A
77 #define RLE_CHAR 0x202B
78 #define PDF_CHAR 0x202C
79 #define LRO_CHAR 0x202D
80 #define RLO_CHAR 0x202E
83 #define BIDI_BOB -2 /* FIXME: Is this needed? */
85 /* Local data structures. (Look in dispextern.h for the rest.) */
87 /* What we need to know about the current paragraph. */
88 struct bidi_paragraph_info
{
89 int start_bytepos
; /* byte position where it begins */
90 int end_bytepos
; /* byte position where it ends */
91 int embedding_level
; /* its basic embedding level */
92 bidi_dir_t base_dir
; /* its base direction */
95 /* Data type for describing the bidirectional character categories. */
103 int bidi_ignore_explicit_marks_for_paragraph_level
= 1;
105 static Lisp_Object paragraph_start_re
, paragraph_separate_re
;
106 static Lisp_Object Qparagraph_start
, Qparagraph_separate
;
111 /* FIXME: This should come from the Unicode Database. */
116 { { 0x0000, 0x0008, WEAK_BN
},
117 { 0x0009, 0x0000, NEUTRAL_S
},
118 { 0x000A, 0x0000, NEUTRAL_B
},
119 { 0x000B, 0x0000, NEUTRAL_S
},
120 { 0x000C, 0x0000, NEUTRAL_WS
},
121 { 0x000D, 0x0000, NEUTRAL_B
},
122 { 0x000E, 0x001B, WEAK_BN
},
123 { 0x001C, 0x001E, NEUTRAL_B
},
124 { 0x001F, 0x0000, NEUTRAL_S
},
125 { 0x0020, 0x0000, NEUTRAL_WS
},
126 { 0x0021, 0x0022, NEUTRAL_ON
},
127 { 0x0023, 0x0025, WEAK_ET
},
128 { 0x0026, 0x002A, NEUTRAL_ON
},
129 { 0x002B, 0x0000, WEAK_ES
},
130 { 0x002C, 0x0000, WEAK_CS
},
131 { 0x002D, 0x0000, WEAK_ES
},
132 { 0x002E, 0x002F, WEAK_CS
},
133 { 0x0030, 0x0039, WEAK_EN
},
134 { 0x003A, 0x0000, WEAK_CS
},
135 { 0x003B, 0x0040, NEUTRAL_ON
},
136 { 0x005B, 0x0060, NEUTRAL_ON
},
137 { 0x007B, 0x007E, NEUTRAL_ON
},
138 { 0x007F, 0x0084, WEAK_BN
},
139 { 0x0085, 0x0000, NEUTRAL_B
},
140 { 0x0086, 0x009F, WEAK_BN
},
141 { 0x00A0, 0x0000, WEAK_CS
},
142 { 0x00A1, 0x0000, NEUTRAL_ON
},
143 { 0x00A2, 0x00A5, WEAK_ET
},
144 { 0x00A6, 0x00A9, NEUTRAL_ON
},
145 { 0x00AB, 0x00AC, NEUTRAL_ON
},
146 { 0x00AD, 0x0000, WEAK_BN
},
147 { 0x00AE, 0x00Af, NEUTRAL_ON
},
148 { 0x00B0, 0x00B1, WEAK_ET
},
149 { 0x00B2, 0x00B3, WEAK_EN
},
150 { 0x00B4, 0x0000, NEUTRAL_ON
},
151 { 0x00B6, 0x00B8, NEUTRAL_ON
},
152 { 0x00B9, 0x0000, WEAK_EN
},
153 { 0x00BB, 0x00BF, NEUTRAL_ON
},
154 { 0x00D7, 0x0000, NEUTRAL_ON
},
155 { 0x00F7, 0x0000, NEUTRAL_ON
},
156 { 0x02B9, 0x02BA, NEUTRAL_ON
},
157 { 0x02C2, 0x02CF, NEUTRAL_ON
},
158 { 0x02D2, 0x02DF, NEUTRAL_ON
},
159 { 0x02E5, 0x02ED, NEUTRAL_ON
},
160 { 0x0300, 0x036F, WEAK_NSM
},
161 { 0x0374, 0x0375, NEUTRAL_ON
},
162 { 0x037E, 0x0385, NEUTRAL_ON
},
163 { 0x0387, 0x0000, NEUTRAL_ON
},
164 { 0x03F6, 0x0000, NEUTRAL_ON
},
165 { 0x0483, 0x0489, WEAK_NSM
},
166 { 0x058A, 0x0000, NEUTRAL_ON
},
167 { 0x0591, 0x05BD, WEAK_NSM
},
168 { 0x05BE, 0x0000, STRONG_R
},
169 { 0x05BF, 0x0000, WEAK_NSM
},
170 { 0x05C0, 0x0000, STRONG_R
},
171 { 0x05C1, 0x05C2, WEAK_NSM
},
172 { 0x05C3, 0x0000, STRONG_R
},
173 { 0x05C4, 0x05C5, WEAK_NSM
},
174 { 0x05C6, 0x0000, STRONG_R
},
175 { 0x05C7, 0x0000, WEAK_NSM
},
176 { 0x05D0, 0x05F4, STRONG_R
},
177 { 0x060C, 0x0000, WEAK_CS
},
178 { 0x061B, 0x064A, STRONG_AL
},
179 { 0x064B, 0x0655, WEAK_NSM
},
180 { 0x0660, 0x0669, WEAK_AN
},
181 { 0x066A, 0x0000, WEAK_ET
},
182 { 0x066B, 0x066C, WEAK_AN
},
183 { 0x066D, 0x066F, STRONG_AL
},
184 { 0x0670, 0x0000, WEAK_NSM
},
185 { 0x0671, 0x06D5, STRONG_AL
},
186 { 0x06D6, 0x06DC, WEAK_NSM
},
187 { 0x06DD, 0x0000, STRONG_AL
},
188 { 0x06DE, 0x06E4, WEAK_NSM
},
189 { 0x06E5, 0x06E6, STRONG_AL
},
190 { 0x06E7, 0x06E8, WEAK_NSM
},
191 { 0x06E9, 0x0000, NEUTRAL_ON
},
192 { 0x06EA, 0x06ED, WEAK_NSM
},
193 { 0x06F0, 0x06F9, WEAK_EN
},
194 { 0x06FA, 0x070D, STRONG_AL
},
195 { 0x070F, 0x0000, WEAK_BN
},
196 { 0x0710, 0x0000, STRONG_AL
},
197 { 0x0711, 0x0000, WEAK_NSM
},
198 { 0x0712, 0x072C, STRONG_AL
},
199 { 0x0730, 0x074A, WEAK_NSM
},
200 { 0x0780, 0x07A5, STRONG_AL
},
201 { 0x07A6, 0x07B0, WEAK_NSM
},
202 { 0x07B1, 0x0000, STRONG_AL
},
203 { 0x0901, 0x0902, WEAK_NSM
},
204 { 0x093C, 0x0000, WEAK_NSM
},
205 { 0x0941, 0x0948, WEAK_NSM
},
206 { 0x094D, 0x0000, WEAK_NSM
},
207 { 0x0951, 0x0954, WEAK_NSM
},
208 { 0x0962, 0x0963, WEAK_NSM
},
209 { 0x0981, 0x0000, WEAK_NSM
},
210 { 0x09BC, 0x0000, WEAK_NSM
},
211 { 0x09C1, 0x09C4, WEAK_NSM
},
212 { 0x09CD, 0x0000, WEAK_NSM
},
213 { 0x09E2, 0x09E3, WEAK_NSM
},
214 { 0x09F2, 0x09F3, WEAK_ET
},
215 { 0x0A02, 0x0000, WEAK_NSM
},
216 { 0x0A3C, 0x0000, WEAK_NSM
},
217 { 0x0A41, 0x0A4D, WEAK_NSM
},
218 { 0x0A70, 0x0A71, WEAK_NSM
},
219 { 0x0A81, 0x0A82, WEAK_NSM
},
220 { 0x0ABC, 0x0000, WEAK_NSM
},
221 { 0x0AC1, 0x0AC8, WEAK_NSM
},
222 { 0x0ACD, 0x0000, WEAK_NSM
},
223 { 0x0B01, 0x0000, WEAK_NSM
},
224 { 0x0B3C, 0x0000, WEAK_NSM
},
225 { 0x0B3F, 0x0000, WEAK_NSM
},
226 { 0x0B41, 0x0B43, WEAK_NSM
},
227 { 0x0B4D, 0x0B56, WEAK_NSM
},
228 { 0x0B82, 0x0000, WEAK_NSM
},
229 { 0x0BC0, 0x0000, WEAK_NSM
},
230 { 0x0BCD, 0x0000, WEAK_NSM
},
231 { 0x0C3E, 0x0C40, WEAK_NSM
},
232 { 0x0C46, 0x0C56, WEAK_NSM
},
233 { 0x0CBF, 0x0000, WEAK_NSM
},
234 { 0x0CC6, 0x0000, WEAK_NSM
},
235 { 0x0CCC, 0x0CCD, WEAK_NSM
},
236 { 0x0D41, 0x0D43, WEAK_NSM
},
237 { 0x0D4D, 0x0000, WEAK_NSM
},
238 { 0x0DCA, 0x0000, WEAK_NSM
},
239 { 0x0DD2, 0x0DD6, WEAK_NSM
},
240 { 0x0E31, 0x0000, WEAK_NSM
},
241 { 0x0E34, 0x0E3A, WEAK_NSM
},
242 { 0x0E3F, 0x0000, WEAK_ET
},
243 { 0x0E47, 0x0E4E, WEAK_NSM
},
244 { 0x0EB1, 0x0000, WEAK_NSM
},
245 { 0x0EB4, 0x0EBC, WEAK_NSM
},
246 { 0x0EC8, 0x0ECD, WEAK_NSM
},
247 { 0x0F18, 0x0F19, WEAK_NSM
},
248 { 0x0F35, 0x0000, WEAK_NSM
},
249 { 0x0F37, 0x0000, WEAK_NSM
},
250 { 0x0F39, 0x0000, WEAK_NSM
},
251 { 0x0F3A, 0x0F3D, NEUTRAL_ON
},
252 { 0x0F71, 0x0F7E, WEAK_NSM
},
253 { 0x0F80, 0x0F84, WEAK_NSM
},
254 { 0x0F86, 0x0F87, WEAK_NSM
},
255 { 0x0F90, 0x0FBC, WEAK_NSM
},
256 { 0x0FC6, 0x0000, WEAK_NSM
},
257 { 0x102D, 0x1030, WEAK_NSM
},
258 { 0x1032, 0x1037, WEAK_NSM
},
259 { 0x1039, 0x0000, WEAK_NSM
},
260 { 0x1058, 0x1059, WEAK_NSM
},
261 { 0x1680, 0x0000, NEUTRAL_WS
},
262 { 0x169B, 0x169C, NEUTRAL_ON
},
263 { 0x1712, 0x1714, WEAK_NSM
},
264 { 0x1732, 0x1734, WEAK_NSM
},
265 { 0x1752, 0x1753, WEAK_NSM
},
266 { 0x1772, 0x1773, WEAK_NSM
},
267 { 0x17B7, 0x17BD, WEAK_NSM
},
268 { 0x17C6, 0x0000, WEAK_NSM
},
269 { 0x17C9, 0x17D3, WEAK_NSM
},
270 { 0x17DB, 0x0000, WEAK_ET
},
271 { 0x1800, 0x180A, NEUTRAL_ON
},
272 { 0x180B, 0x180D, WEAK_NSM
},
273 { 0x180E, 0x0000, WEAK_BN
},
274 { 0x18A9, 0x0000, WEAK_NSM
},
275 { 0x1FBD, 0x0000, NEUTRAL_ON
},
276 { 0x1FBF, 0x1FC1, NEUTRAL_ON
},
277 { 0x1FCD, 0x1FCF, NEUTRAL_ON
},
278 { 0x1FDD, 0x1FDF, NEUTRAL_ON
},
279 { 0x1FED, 0x1FEF, NEUTRAL_ON
},
280 { 0x1FFD, 0x1FFE, NEUTRAL_ON
},
281 { 0x2000, 0x200A, NEUTRAL_WS
},
282 { 0x200B, 0x200D, WEAK_BN
},
283 { 0x200F, 0x0000, STRONG_R
},
284 { 0x2010, 0x2027, NEUTRAL_ON
},
285 { 0x2028, 0x0000, NEUTRAL_WS
},
286 { 0x2029, 0x0000, NEUTRAL_B
},
287 { 0x202A, 0x0000, LRE
},
288 { 0x202B, 0x0000, RLE
},
289 { 0x202C, 0x0000, PDF
},
290 { 0x202D, 0x0000, LRO
},
291 { 0x202E, 0x0000, RLO
},
292 { 0x202F, 0x0000, NEUTRAL_WS
},
293 { 0x2030, 0x2034, WEAK_ET
},
294 { 0x2035, 0x2057, NEUTRAL_ON
},
295 { 0x205F, 0x0000, NEUTRAL_WS
},
296 { 0x2060, 0x206F, WEAK_BN
},
297 { 0x2070, 0x0000, WEAK_EN
},
298 { 0x2074, 0x2079, WEAK_EN
},
299 { 0x207A, 0x207B, WEAK_ET
},
300 { 0x207C, 0x207E, NEUTRAL_ON
},
301 { 0x2080, 0x2089, WEAK_EN
},
302 { 0x208A, 0x208B, WEAK_ET
},
303 { 0x208C, 0x208E, NEUTRAL_ON
},
304 { 0x20A0, 0x20B1, WEAK_ET
},
305 { 0x20D0, 0x20EA, WEAK_NSM
},
306 { 0x2100, 0x2101, NEUTRAL_ON
},
307 { 0x2103, 0x2106, NEUTRAL_ON
},
308 { 0x2108, 0x2109, NEUTRAL_ON
},
309 { 0x2114, 0x0000, NEUTRAL_ON
},
310 { 0x2116, 0x2118, NEUTRAL_ON
},
311 { 0x211E, 0x2123, NEUTRAL_ON
},
312 { 0x2125, 0x0000, NEUTRAL_ON
},
313 { 0x2127, 0x0000, NEUTRAL_ON
},
314 { 0x2129, 0x0000, NEUTRAL_ON
},
315 { 0x212E, 0x0000, WEAK_ET
},
316 { 0x2132, 0x0000, NEUTRAL_ON
},
317 { 0x213A, 0x0000, NEUTRAL_ON
},
318 { 0x2140, 0x2144, NEUTRAL_ON
},
319 { 0x214A, 0x215F, NEUTRAL_ON
},
320 { 0x2190, 0x2211, NEUTRAL_ON
},
321 { 0x2212, 0x2213, WEAK_ET
},
322 { 0x2214, 0x2335, NEUTRAL_ON
},
323 { 0x237B, 0x2394, NEUTRAL_ON
},
324 { 0x2396, 0x244A, NEUTRAL_ON
},
325 { 0x2460, 0x249B, WEAK_EN
},
326 { 0x24EA, 0x0000, WEAK_EN
},
327 { 0x24EB, 0x2FFB, NEUTRAL_ON
},
328 { 0x3000, 0x0000, NEUTRAL_WS
},
329 { 0x3001, 0x3004, NEUTRAL_ON
},
330 { 0x3008, 0x3020, NEUTRAL_ON
},
331 { 0x302A, 0x302F, WEAK_NSM
},
332 { 0x3030, 0x0000, NEUTRAL_ON
},
333 { 0x3036, 0x3037, NEUTRAL_ON
},
334 { 0x303D, 0x303F, NEUTRAL_ON
},
335 { 0x3099, 0x309A, WEAK_NSM
},
336 { 0x309B, 0x309C, NEUTRAL_ON
},
337 { 0x30A0, 0x0000, NEUTRAL_ON
},
338 { 0x30FB, 0x0000, NEUTRAL_ON
},
339 { 0x3251, 0x325F, NEUTRAL_ON
},
340 { 0x32B1, 0x32BF, NEUTRAL_ON
},
341 { 0xA490, 0xA4C6, NEUTRAL_ON
},
342 { 0xFB1D, 0x0000, STRONG_R
},
343 { 0xFB1E, 0x0000, WEAK_NSM
},
344 { 0xFB1F, 0xFB28, STRONG_R
},
345 { 0xFB29, 0x0000, WEAK_ET
},
346 { 0xFB2A, 0xFB4F, STRONG_R
},
347 { 0xFB50, 0xFD3D, STRONG_AL
},
348 { 0xFD3E, 0xFD3F, NEUTRAL_ON
},
349 { 0xFD50, 0xFDFC, STRONG_AL
},
350 { 0xFE00, 0xFE23, WEAK_NSM
},
351 { 0xFE30, 0xFE4F, NEUTRAL_ON
},
352 { 0xFE50, 0x0000, WEAK_CS
},
353 { 0xFE51, 0x0000, NEUTRAL_ON
},
354 { 0xFE52, 0x0000, WEAK_CS
},
355 { 0xFE54, 0x0000, NEUTRAL_ON
},
356 { 0xFE55, 0x0000, WEAK_CS
},
357 { 0xFE56, 0xFE5E, NEUTRAL_ON
},
358 { 0xFE5F, 0x0000, WEAK_ET
},
359 { 0xFE60, 0xFE61, NEUTRAL_ON
},
360 { 0xFE62, 0xFE63, WEAK_ET
},
361 { 0xFE64, 0xFE68, NEUTRAL_ON
},
362 { 0xFE69, 0xFE6A, WEAK_ET
},
363 { 0xFE6B, 0x0000, NEUTRAL_ON
},
364 { 0xFE70, 0xFEFC, STRONG_AL
},
365 { 0xFEFF, 0x0000, WEAK_BN
},
366 { 0xFF01, 0xFF02, NEUTRAL_ON
},
367 { 0xFF03, 0xFF05, WEAK_ET
},
368 { 0xFF06, 0xFF0A, NEUTRAL_ON
},
369 { 0xFF0B, 0x0000, WEAK_ET
},
370 { 0xFF0C, 0x0000, WEAK_CS
},
371 { 0xFF0D, 0x0000, WEAK_ET
},
372 { 0xFF0E, 0x0000, WEAK_CS
},
373 { 0xFF0F, 0x0000, WEAK_ES
},
374 { 0xFF10, 0xFF19, WEAK_EN
},
375 { 0xFF1A, 0x0000, WEAK_CS
},
376 { 0xFF1B, 0xFF20, NEUTRAL_ON
},
377 { 0xFF3B, 0xFF40, NEUTRAL_ON
},
378 { 0xFF5B, 0xFF65, NEUTRAL_ON
},
379 { 0xFFE0, 0xFFE1, WEAK_ET
},
380 { 0xFFE2, 0xFFE4, NEUTRAL_ON
},
381 { 0xFFE5, 0xFFE6, WEAK_ET
},
382 { 0xFFE8, 0xFFEE, NEUTRAL_ON
},
383 { 0xFFF9, 0xFFFB, WEAK_BN
},
384 { 0xFFFC, 0xFFFD, NEUTRAL_ON
},
385 { 0x1D167, 0x1D169, WEAK_NSM
},
386 { 0x1D173, 0x1D17A, WEAK_BN
},
387 { 0x1D17B, 0x1D182, WEAK_NSM
},
388 { 0x1D185, 0x1D18B, WEAK_NSM
},
389 { 0x1D1AA, 0x1D1AD, WEAK_NSM
},
390 { 0x1D7CE, 0x1D7FF, WEAK_EN
},
391 { 0xE0001, 0xE007F, WEAK_BN
} };
394 bidi_type_table
= Fmake_char_table (Qnil
, make_number (STRONG_L
));
395 staticpro (&bidi_type_table
);
397 for (i
= 0; i
< sizeof bidi_type
/ sizeof bidi_type
[0]; i
++)
398 char_table_set_range (bidi_type_table
, bidi_type
[i
].from
,
399 bidi_type
[i
].to
? bidi_type
[i
].to
: bidi_type
[i
].from
,
400 make_number (bidi_type
[i
].type
));
402 Qparagraph_start
= intern ("paragraph-start");
403 staticpro (&Qparagraph_start
);
404 paragraph_start_re
= Fsymbol_value (Qparagraph_start
);
405 if (!STRINGP (paragraph_start_re
))
406 paragraph_start_re
= build_string ("\f\\|[ \t]*$");
407 staticpro (¶graph_start_re
);
408 Qparagraph_separate
= intern ("paragraph-separate");
409 staticpro (&Qparagraph_separate
);
410 paragraph_separate_re
= Fsymbol_value (Qparagraph_separate
);
411 if (!STRINGP (paragraph_separate_re
))
412 paragraph_separate_re
= build_string ("[ \t\f]*$");
413 staticpro (¶graph_separate_re
);
414 bidi_initialized
= 1;
417 /* Return the bidi type of a character CH, subject to the current
418 directional OVERRIDE. */
419 static INLINE bidi_type_t
420 bidi_get_type (int ch
, bidi_dir_t override
)
422 bidi_type_t default_type
;
426 if (ch
< 0 || ch
> MAX_CHAR
)
429 default_type
= (bidi_type_t
) XINT (CHAR_TABLE_REF (bidi_type_table
, ch
));
431 if (override
== NEUTRAL_DIR
)
434 switch (default_type
)
436 /* Although UAX#9 does not tell, it doesn't make sense to
437 override NEUTRAL_B and LRM/RLM characters. */
452 if (override
== L2R
) /* X6 */
454 else if (override
== R2L
)
457 abort (); /* can't happen: handled above */
463 bidi_check_type (bidi_type_t type
)
465 if (type
< UNKNOWN_BT
|| type
> NEUTRAL_ON
)
469 /* Given a bidi TYPE of a character, return its category. */
470 static INLINE bidi_category_t
471 bidi_get_category (bidi_type_t type
)
485 case PDF
: /* ??? really?? */
504 /* Return the mirrored character of C, if any.
506 Note: The conditions in UAX#9 clause L4 must be tested by the
508 /* FIXME: exceedingly temporary! Should consult the Unicode database
509 of character properties. */
511 bidi_mirror_char (int c
)
513 static const char mirrored_pairs
[] = "()<>[]{}";
514 const char *p
= c
> 0 && c
< 128 ? strchr (mirrored_pairs
, c
) : NULL
;
518 size_t i
= p
- mirrored_pairs
;
520 return mirrored_pairs
[(i
^ 1)];
525 /* Copy the bidi iterator from FROM to TO. To save cycles, this only
526 copies the part of the level stack that is actually in use. */
528 bidi_copy_it (struct bidi_it
*to
, struct bidi_it
*from
)
532 /* Copy everything except the level stack and beyond. */
533 memcpy (to
, from
, ((size_t)&((struct bidi_it
*)0)->level_stack
[0]));
535 /* Copy the active part of the level stack. */
536 to
->level_stack
[0] = from
->level_stack
[0]; /* level zero is always in use */
537 for (i
= 1; i
<= from
->stack_idx
; i
++)
538 to
->level_stack
[i
] = from
->level_stack
[i
];
541 /* Caching the bidi iterator states. */
543 #define BIDI_CACHE_CHUNK 200
544 static struct bidi_it
*bidi_cache
;
545 static size_t bidi_cache_size
= 0;
546 static int bidi_cache_idx
; /* next unused cache slot */
547 static int bidi_cache_last_idx
; /* slot of last cache hit */
550 bidi_cache_reset (void)
553 bidi_cache_last_idx
= -1;
557 bidi_cache_shrink (void)
559 if (bidi_cache_size
> BIDI_CACHE_CHUNK
)
561 bidi_cache_size
= BIDI_CACHE_CHUNK
* sizeof (struct bidi_it
);
562 bidi_cache
= (struct bidi_it
*) xrealloc (bidi_cache
, bidi_cache_size
);
568 bidi_cache_fetch_state (int idx
, struct bidi_it
*bidi_it
)
570 int current_scan_dir
= bidi_it
->scan_dir
;
572 if (idx
< 0 || idx
>= bidi_cache_idx
)
575 bidi_copy_it (bidi_it
, &bidi_cache
[idx
]);
576 bidi_it
->scan_dir
= current_scan_dir
;
577 bidi_cache_last_idx
= idx
;
580 /* Find a cached state with a given CHARPOS and resolved embedding
581 level less or equal to LEVEL. if LEVEL is -1, disregard the
582 resolved levels in cached states. DIR, if non-zero, means search
583 in that direction from the last cache hit. */
585 bidi_cache_search (int charpos
, int level
, int dir
)
591 if (charpos
< bidi_cache
[bidi_cache_last_idx
].charpos
)
593 else if (charpos
> bidi_cache
[bidi_cache_last_idx
].charpos
)
596 i_start
= bidi_cache_last_idx
;
600 i_start
= bidi_cache_idx
- 1;
605 /* Linear search for now; FIXME! */
606 for (i
= i_start
; i
>= 0; i
--)
607 if (bidi_cache
[i
].charpos
== charpos
608 && (level
== -1 || bidi_cache
[i
].resolved_level
<= level
))
613 for (i
= i_start
; i
< bidi_cache_idx
; i
++)
614 if (bidi_cache
[i
].charpos
== charpos
615 && (level
== -1 || bidi_cache
[i
].resolved_level
<= level
))
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.
626 BEFORE, if non-zero, means return the index of the slot that is
627 ``before'' the level change in the search direction. That is,
628 given the cached levels like this:
633 and assuming we are at the position cached at the slot marked with
634 C, searching backwards (DIR = -1) for LEVEL = 2 will return the
635 index of slot B or A, depending whether BEFORE is, respectively,
638 bidi_cache_find_level_change (int level
, int dir
, int before
)
642 int i
= dir
? bidi_cache_last_idx
: bidi_cache_idx
- 1;
643 int incr
= before
? 1 : 0;
654 if (bidi_cache
[i
- incr
].resolved_level
>= 0
655 && bidi_cache
[i
- incr
].resolved_level
< level
)
662 while (i
< bidi_cache_idx
- incr
)
664 if (bidi_cache
[i
+ incr
].resolved_level
>= 0
665 && bidi_cache
[i
+ incr
].resolved_level
< level
)
676 bidi_cache_iterator_state (struct bidi_it
*bidi_it
, int resolved
)
680 /* We should never cache on backward scans. */
681 if (bidi_it
->scan_dir
== -1)
683 idx
= bidi_cache_search (bidi_it
->charpos
, -1, 1);
687 idx
= bidi_cache_idx
;
688 /* Enlarge the cache as needed. */
689 if (idx
>= bidi_cache_size
)
691 bidi_cache_size
+= BIDI_CACHE_CHUNK
* sizeof (struct bidi_it
);
693 (struct bidi_it
*) xrealloc (bidi_cache
, bidi_cache_size
);
695 /* Character positions should correspond to cache positions 1:1.
696 If we are outside the range of cached positions, the cache is
697 useless and must be reset. */
699 (bidi_it
->charpos
> bidi_cache
[idx
- 1].charpos
+ 1
700 || bidi_it
->charpos
< bidi_cache
[0].charpos
))
705 bidi_copy_it (&bidi_cache
[idx
], bidi_it
);
707 bidi_cache
[idx
].resolved_level
= -1;
708 bidi_cache
[idx
].new_paragraph
= 0;
712 /* Copy only the members which could have changed, to avoid
713 costly copying of the entire struct. */
714 bidi_cache
[idx
].type
= bidi_it
->type
;
715 bidi_check_type (bidi_it
->type
);
716 bidi_cache
[idx
].type_after_w1
= bidi_it
->type_after_w1
;
717 bidi_check_type (bidi_it
->type_after_w1
);
719 bidi_cache
[idx
].resolved_level
= bidi_it
->resolved_level
;
721 bidi_cache
[idx
].resolved_level
= -1;
722 bidi_cache
[idx
].invalid_levels
= bidi_it
->invalid_levels
;
723 bidi_cache
[idx
].invalid_rl_levels
= bidi_it
->invalid_rl_levels
;
724 bidi_cache
[idx
].next_for_neutral
= bidi_it
->next_for_neutral
;
725 bidi_cache
[idx
].next_for_ws
= bidi_it
->next_for_ws
;
726 bidi_cache
[idx
].ignore_bn_limit
= bidi_it
->ignore_bn_limit
;
729 bidi_cache_last_idx
= idx
;
730 if (idx
>= bidi_cache_idx
)
731 bidi_cache_idx
= idx
+ 1;
734 static INLINE bidi_type_t
735 bidi_cache_find (int charpos
, int level
, struct bidi_it
*bidi_it
)
737 int i
= bidi_cache_search (charpos
, level
, bidi_it
->scan_dir
);
741 bidi_dir_t current_scan_dir
= bidi_it
->scan_dir
;
743 bidi_copy_it (bidi_it
, &bidi_cache
[i
]);
744 bidi_cache_last_idx
= i
;
745 /* Don't let scan direction from from the cached state override
746 the current scan direction. */
747 bidi_it
->scan_dir
= current_scan_dir
;
748 return bidi_it
->type
;
755 bidi_peek_at_next_level (struct bidi_it
*bidi_it
)
757 if (bidi_cache_idx
== 0 || bidi_cache_last_idx
== -1)
759 return bidi_cache
[bidi_cache_last_idx
+ bidi_it
->scan_dir
].resolved_level
;
762 /* Check if buffer position CHARPOS/BYTEPOS is the end of a paragraph.
763 Value is the non-negative length of the paragraph separator
764 following the buffer position, -1 if position is at the beginning
765 of a new paragraph, or -2 if position is neither at beginning nor
766 at end of a paragraph. */
768 bidi_at_paragraph_end (EMACS_INT charpos
, EMACS_INT bytepos
)
770 /* FIXME: Why Fbuffer_local_value rather than just Fsymbol_value? */
772 Lisp_Object start_re
;
775 sep_re
= paragraph_separate_re
;
776 start_re
= paragraph_start_re
;
778 val
= fast_looking_at (sep_re
, charpos
, bytepos
, ZV
, ZV_BYTE
, Qnil
);
781 if (fast_looking_at (start_re
, charpos
, bytepos
, ZV
, ZV_BYTE
, Qnil
) >= 0)
790 /* Determine the start-of-run (sor) directional type given the two
791 embedding levels on either side of the run boundary. Also, update
792 the saved info about previously seen characters, since that info is
793 generally valid for a single level run. */
795 bidi_set_sor_type (struct bidi_it
*bidi_it
, int level_before
, int level_after
)
797 int higher_level
= level_before
> level_after
? level_before
: level_after
;
799 /* The prev_was_pdf gork is required for when we have several PDFs
800 in a row. In that case, we want to compute the sor type for the
801 next level run only once: when we see the first PDF. That's
802 because the sor type depends only on the higher of the two levels
803 that we find on the two sides of the level boundary (see UAX#9,
804 clause X10), and so we don't need to know the final embedding
805 level to which we descend after processing all the PDFs. */
806 if (!bidi_it
->prev_was_pdf
|| level_before
< level_after
)
807 /* FIXME: should the default sor direction be user selectable? */
808 bidi_it
->sor
= (higher_level
& 1) != 0 ? R2L
: L2R
;
809 if (level_before
> level_after
)
810 bidi_it
->prev_was_pdf
= 1;
812 bidi_it
->prev
.type
= UNKNOWN_BT
;
813 bidi_it
->last_strong
.type
= bidi_it
->last_strong
.type_after_w1
=
814 bidi_it
->last_strong
.orig_type
= UNKNOWN_BT
;
815 bidi_it
->prev_for_neutral
.type
= bidi_it
->sor
== R2L
? STRONG_R
: STRONG_L
;
816 bidi_it
->prev_for_neutral
.charpos
= bidi_it
->charpos
;
817 bidi_it
->prev_for_neutral
.bytepos
= bidi_it
->bytepos
;
818 bidi_it
->next_for_neutral
.type
= bidi_it
->next_for_neutral
.type_after_w1
=
819 bidi_it
->next_for_neutral
.orig_type
= UNKNOWN_BT
;
820 bidi_it
->ignore_bn_limit
= 0; /* meaning it's unknown */
824 bidi_line_init (struct bidi_it
*bidi_it
)
826 bidi_it
->scan_dir
= 1; /* FIXME: do we need to have control on this? */
827 bidi_it
->resolved_level
= bidi_it
->level_stack
[0].level
;
828 bidi_it
->level_stack
[0].override
= NEUTRAL_DIR
; /* X1 */
829 bidi_it
->invalid_levels
= 0;
830 bidi_it
->invalid_rl_levels
= -1;
831 bidi_it
->next_en_pos
= -1;
832 bidi_it
->next_for_ws
.type
= UNKNOWN_BT
;
833 bidi_set_sor_type (bidi_it
,
834 bidi_it
->paragraph_dir
== R2L
? 1 : 0,
835 bidi_it
->level_stack
[0].level
); /* X10 */
840 /* Find the beginning of this paragraph by looking back in the buffer.
841 Value is the byte position of the paragraph's beginning. */
843 bidi_find_paragraph_start (EMACS_INT pos
, EMACS_INT pos_byte
)
845 Lisp_Object re
= paragraph_start_re
;
846 EMACS_INT limit
= ZV
, limit_byte
= ZV_BYTE
;
848 while (pos_byte
> BEGV_BYTE
849 && fast_looking_at (re
, pos
, pos_byte
, limit
, limit_byte
, Qnil
) < 0)
851 pos
= find_next_newline_no_quit (pos
- 1, -1);
852 pos_byte
= CHAR_TO_BYTE (pos
);
857 /* Determine the direction, a.k.a. base embedding level, of the
858 paragraph we are about to iterate through. If DIR is either L2R or
859 R2L, just use that. Otherwise, determine the paragraph direction
860 from the first strong character of the paragraph.
862 Note that this gives the paragraph separator the same direction as
863 the preceding paragraph, even though Emacs generally views the
864 separartor as not belonging to any paragraph. */
866 bidi_paragraph_init (bidi_dir_t dir
, struct bidi_it
*bidi_it
)
868 EMACS_INT bytepos
= bidi_it
->bytepos
;
870 /* Special case for an empty buffer. */
871 if (bytepos
== BEGV_BYTE
&& bytepos
== ZV_BYTE
)
873 /* We should never be called at EOB or before BEGV. */
874 else if (bytepos
>= ZV_BYTE
|| bytepos
< BEGV_BYTE
)
879 bidi_it
->paragraph_dir
= L2R
;
880 bidi_it
->new_paragraph
= 0;
884 bidi_it
->paragraph_dir
= R2L
;
885 bidi_it
->new_paragraph
= 0;
887 else if (dir
== NEUTRAL_DIR
) /* P2 */
893 if (!bidi_initialized
)
896 /* If we are inside a paragraph separator, we are just waiting
897 for the separator to be exhausted; use the previous paragraph
898 direction. But don't do that if we have been just reseated,
899 because we need to reinitialize below in that case. */
900 if (!bidi_it
->first_elt
901 && bidi_it
->charpos
< bidi_it
->separator_limit
)
904 /* If we are on a newline, get past it to where the next
905 paragraph might start. But don't do that at BEGV since then
906 we are potentially in a new paragraph that doesn't yet
908 pos
= bidi_it
->charpos
;
909 if (bytepos
> BEGV_BYTE
&& FETCH_CHAR (bytepos
) == '\n')
915 /* We are either at the beginning of a paragraph or in the
916 middle of it. Find where this paragraph starts. */
917 bytepos
= bidi_find_paragraph_start (pos
, bytepos
);
919 bidi_it
->separator_limit
= -1;
920 bidi_it
->new_paragraph
= 0;
921 ch
= FETCH_CHAR (bytepos
);
922 ch_len
= CHAR_BYTES (ch
);
923 pos
= BYTE_TO_CHAR (bytepos
);
924 type
= bidi_get_type (ch
, NEUTRAL_DIR
);
926 for (pos
++, bytepos
+= ch_len
;
927 /* NOTE: UAX#9 says to search only for L, AL, or R types of
928 characters, and ignore RLE, RLO, LRE, and LRO. However,
929 I'm not sure it makes sense to omit those 4; should try
930 with and without that to see the effect. */
931 (bidi_get_category (type
) != STRONG
)
932 || (bidi_ignore_explicit_marks_for_paragraph_level
933 && (type
== RLE
|| type
== RLO
934 || type
== LRE
|| type
== LRO
));
935 type
= bidi_get_type (ch
, NEUTRAL_DIR
))
937 if (type
== NEUTRAL_B
&& bidi_at_paragraph_end (pos
, bytepos
) >= -1)
939 if (bytepos
>= ZV_BYTE
)
941 /* Pretend there's a paragraph separator at end of buffer. */
945 FETCH_CHAR_ADVANCE (ch
, pos
, bytepos
);
947 if (type
== STRONG_R
|| type
== STRONG_AL
) /* P3 */
948 bidi_it
->paragraph_dir
= R2L
;
949 else if (type
== STRONG_L
)
950 bidi_it
->paragraph_dir
= L2R
;
955 /* Contrary to UAX#9 clause P3, we only default the paragraph
956 direction to L2R if we have no previous usable paragraph
958 if (bidi_it
->paragraph_dir
!= L2R
&& bidi_it
->paragraph_dir
!= R2L
)
959 bidi_it
->paragraph_dir
= L2R
; /* P3 and ``higher protocols'' */
960 if (bidi_it
->paragraph_dir
== R2L
)
961 bidi_it
->level_stack
[0].level
= 1;
963 bidi_it
->level_stack
[0].level
= 0;
965 bidi_line_init (bidi_it
);
968 /* Do whatever UAX#9 clause X8 says should be done at paragraph's
971 bidi_set_paragraph_end (struct bidi_it
*bidi_it
)
973 bidi_it
->invalid_levels
= 0;
974 bidi_it
->invalid_rl_levels
= -1;
975 bidi_it
->stack_idx
= 0;
976 bidi_it
->resolved_level
= bidi_it
->level_stack
[0].level
;
979 /* Initialize the bidi iterator from buffer position CHARPOS. */
981 bidi_init_it (EMACS_INT charpos
, EMACS_INT bytepos
, struct bidi_it
*bidi_it
)
983 if (! bidi_initialized
)
985 bidi_it
->charpos
= charpos
;
986 bidi_it
->bytepos
= bytepos
;
987 bidi_it
->first_elt
= 1;
988 bidi_set_paragraph_end (bidi_it
);
989 bidi_it
->new_paragraph
= 1;
990 bidi_it
->separator_limit
= -1;
991 bidi_it
->type
= NEUTRAL_B
;
992 bidi_it
->type_after_w1
= NEUTRAL_B
;
993 bidi_it
->orig_type
= NEUTRAL_B
;
994 bidi_it
->prev_was_pdf
= 0;
995 bidi_it
->prev
.type
= bidi_it
->prev
.type_after_w1
=
996 bidi_it
->prev
.orig_type
= UNKNOWN_BT
;
997 bidi_it
->last_strong
.type
= bidi_it
->last_strong
.type_after_w1
=
998 bidi_it
->last_strong
.orig_type
= UNKNOWN_BT
;
999 bidi_it
->next_for_neutral
.charpos
= -1;
1000 bidi_it
->next_for_neutral
.type
=
1001 bidi_it
->next_for_neutral
.type_after_w1
=
1002 bidi_it
->next_for_neutral
.orig_type
= UNKNOWN_BT
;
1003 bidi_it
->prev_for_neutral
.charpos
= -1;
1004 bidi_it
->prev_for_neutral
.type
=
1005 bidi_it
->prev_for_neutral
.type_after_w1
=
1006 bidi_it
->prev_for_neutral
.orig_type
= UNKNOWN_BT
;
1007 bidi_it
->sor
= L2R
; /* FIXME: should it be user-selectable? */
1008 bidi_cache_shrink ();
1011 /* Push the current embedding level and override status; reset the
1012 current level to LEVEL and the current override status to OVERRIDE. */
1014 bidi_push_embedding_level (struct bidi_it
*bidi_it
,
1015 int level
, bidi_dir_t override
)
1017 bidi_it
->stack_idx
++;
1018 if (bidi_it
->stack_idx
>= BIDI_MAXLEVEL
)
1020 bidi_it
->level_stack
[bidi_it
->stack_idx
].level
= level
;
1021 bidi_it
->level_stack
[bidi_it
->stack_idx
].override
= override
;
1024 /* Pop the embedding level and directional override status from the
1025 stack, and return the new level. */
1027 bidi_pop_embedding_level (struct bidi_it
*bidi_it
)
1029 /* UAX#9 says to ignore invalid PDFs. */
1030 if (bidi_it
->stack_idx
> 0)
1031 bidi_it
->stack_idx
--;
1032 return bidi_it
->level_stack
[bidi_it
->stack_idx
].level
;
1035 /* Record in SAVED_INFO the information about the current character. */
1037 bidi_remember_char (struct bidi_saved_info
*saved_info
,
1038 struct bidi_it
*bidi_it
)
1040 saved_info
->charpos
= bidi_it
->charpos
;
1041 saved_info
->bytepos
= bidi_it
->bytepos
;
1042 saved_info
->type
= bidi_it
->type
;
1043 bidi_check_type (bidi_it
->type
);
1044 saved_info
->type_after_w1
= bidi_it
->type_after_w1
;
1045 bidi_check_type (bidi_it
->type_after_w1
);
1046 saved_info
->orig_type
= bidi_it
->orig_type
;
1047 bidi_check_type (bidi_it
->orig_type
);
1050 /* Resolve the type of a neutral character according to the type of
1051 surrounding strong text and the current embedding level. */
1052 static INLINE bidi_type_t
1053 bidi_resolve_neutral_1 (bidi_type_t prev_type
, bidi_type_t next_type
, int lev
)
1055 /* N1: European and Arabic numbers are treated as though they were R. */
1056 if (next_type
== WEAK_EN
|| next_type
== WEAK_AN
)
1057 next_type
= STRONG_R
;
1058 if (prev_type
== WEAK_EN
|| prev_type
== WEAK_AN
)
1059 prev_type
= STRONG_R
;
1061 if (next_type
== prev_type
) /* N1 */
1063 else if ((lev
& 1) == 0) /* N2 */
1070 bidi_explicit_dir_char (int c
)
1072 /* FIXME: this should be replaced with a lookup table with suitable
1073 bits set, like standard C ctype macros do. */
1074 return (c
== LRE_CHAR
|| c
== LRO_CHAR
1075 || c
== RLE_CHAR
|| c
== RLO_CHAR
|| c
== PDF_CHAR
);
1078 /* A helper function for bidi_resolve_explicit. It advances to the
1079 next character in logical order and determines the new embedding
1080 level and directional override, but does not take into account
1081 empty embeddings. */
1083 bidi_resolve_explicit_1 (struct bidi_it
*bidi_it
)
1089 bidi_dir_t override
;
1091 if (bidi_it
->bytepos
< BEGV_BYTE
/* after reseat to BEGV? */
1092 || bidi_it
->first_elt
)
1094 bidi_it
->first_elt
= 0;
1095 if (bidi_it
->charpos
< BEGV
)
1096 bidi_it
->charpos
= BEGV
;
1097 bidi_it
->bytepos
= CHAR_TO_BYTE (bidi_it
->charpos
);
1099 else if (bidi_it
->bytepos
< ZV_BYTE
) /* don't move at ZV */
1102 if (bidi_it
->ch_len
== 0)
1104 bidi_it
->bytepos
+= bidi_it
->ch_len
;
1107 current_level
= bidi_it
->level_stack
[bidi_it
->stack_idx
].level
; /* X1 */
1108 override
= bidi_it
->level_stack
[bidi_it
->stack_idx
].override
;
1109 new_level
= current_level
;
1111 /* in case it is a unibyte character (not yet implemented) */
1112 /* _fetch_multibyte_char_len = 1; */
1113 if (bidi_it
->bytepos
>= ZV_BYTE
)
1116 bidi_it
->ch_len
= 1;
1120 curchar
= FETCH_CHAR (bidi_it
->bytepos
);
1121 bidi_it
->ch_len
= CHAR_BYTES (curchar
);
1123 bidi_it
->ch
= curchar
;
1125 /* Don't apply directional override here, as all the types we handle
1126 below will not be affected by the override anyway, and we need
1127 the original type unaltered. The override will be applied in
1128 bidi_resolve_weak. */
1129 type
= bidi_get_type (curchar
, NEUTRAL_DIR
);
1130 bidi_it
->orig_type
= type
;
1131 bidi_check_type (bidi_it
->orig_type
);
1134 bidi_it
->prev_was_pdf
= 0;
1136 bidi_it
->type_after_w1
= UNKNOWN_BT
;
1142 bidi_it
->type_after_w1
= type
;
1143 bidi_check_type (bidi_it
->type_after_w1
);
1144 type
= WEAK_BN
; /* X9/Retaining */
1145 if (bidi_it
->ignore_bn_limit
<= 0)
1147 if (current_level
<= BIDI_MAXLEVEL
- 4)
1149 /* Compute the least odd embedding level greater than
1150 the current level. */
1151 new_level
= ((current_level
+ 1) & ~1) + 1;
1152 if (bidi_it
->type_after_w1
== RLE
)
1153 override
= NEUTRAL_DIR
;
1156 if (current_level
== BIDI_MAXLEVEL
- 4)
1157 bidi_it
->invalid_rl_levels
= 0;
1158 bidi_push_embedding_level (bidi_it
, new_level
, override
);
1162 bidi_it
->invalid_levels
++;
1163 /* See the commentary about invalid_rl_levels below. */
1164 if (bidi_it
->invalid_rl_levels
< 0)
1165 bidi_it
->invalid_rl_levels
= 0;
1166 bidi_it
->invalid_rl_levels
++;
1169 else if (bidi_it
->prev
.type_after_w1
== WEAK_EN
/* W5/Retaining */
1170 || bidi_it
->next_en_pos
> bidi_it
->charpos
)
1175 bidi_it
->type_after_w1
= type
;
1176 bidi_check_type (bidi_it
->type_after_w1
);
1177 type
= WEAK_BN
; /* X9/Retaining */
1178 if (bidi_it
->ignore_bn_limit
<= 0)
1180 if (current_level
<= BIDI_MAXLEVEL
- 5)
1182 /* Compute the least even embedding level greater than
1183 the current level. */
1184 new_level
= ((current_level
+ 2) & ~1);
1185 if (bidi_it
->type_after_w1
== LRE
)
1186 override
= NEUTRAL_DIR
;
1189 bidi_push_embedding_level (bidi_it
, new_level
, override
);
1193 bidi_it
->invalid_levels
++;
1194 /* invalid_rl_levels counts invalid levels encountered
1195 while the embedding level was already too high for
1196 LRE/LRO, but not for RLE/RLO. That is because
1197 there may be exactly one PDF which we should not
1198 ignore even though invalid_levels is non-zero.
1199 invalid_rl_levels helps to know what PDF is
1201 if (bidi_it
->invalid_rl_levels
>= 0)
1202 bidi_it
->invalid_rl_levels
++;
1205 else if (bidi_it
->prev
.type_after_w1
== WEAK_EN
/* W5/Retaining */
1206 || bidi_it
->next_en_pos
> bidi_it
->charpos
)
1210 bidi_it
->type_after_w1
= type
;
1211 bidi_check_type (bidi_it
->type_after_w1
);
1212 type
= WEAK_BN
; /* X9/Retaining */
1213 if (bidi_it
->ignore_bn_limit
<= 0)
1215 if (!bidi_it
->invalid_rl_levels
)
1217 new_level
= bidi_pop_embedding_level (bidi_it
);
1218 bidi_it
->invalid_rl_levels
= -1;
1219 if (bidi_it
->invalid_levels
)
1220 bidi_it
->invalid_levels
--;
1221 /* else nothing: UAX#9 says to ignore invalid PDFs */
1223 if (!bidi_it
->invalid_levels
)
1224 new_level
= bidi_pop_embedding_level (bidi_it
);
1227 bidi_it
->invalid_levels
--;
1228 bidi_it
->invalid_rl_levels
--;
1231 else if (bidi_it
->prev
.type_after_w1
== WEAK_EN
/* W5/Retaining */
1232 || bidi_it
->next_en_pos
> bidi_it
->charpos
)
1240 bidi_it
->type
= type
;
1241 bidi_check_type (bidi_it
->type
);
1246 /* Given an iterator state in BIDI_IT, advance one character position
1247 in the buffer to the next character (in the logical order), resolve
1248 any explicit embeddings and directional overrides, and return the
1249 embedding level of the character after resolving explicit
1250 directives and ignoring empty embeddings. */
1252 bidi_resolve_explicit (struct bidi_it
*bidi_it
)
1254 int prev_level
= bidi_it
->level_stack
[bidi_it
->stack_idx
].level
;
1255 int new_level
= bidi_resolve_explicit_1 (bidi_it
);
1257 if (prev_level
< new_level
1258 && bidi_it
->type
== WEAK_BN
1259 && bidi_it
->ignore_bn_limit
== 0 /* only if not already known */
1260 && bidi_it
->bytepos
< ZV_BYTE
/* not already at EOB */
1261 && bidi_explicit_dir_char (FETCH_CHAR (bidi_it
->bytepos
1262 + bidi_it
->ch_len
)))
1264 /* Avoid pushing and popping embedding levels if the level run
1265 is empty, as this breaks level runs where it shouldn't.
1266 UAX#9 removes all the explicit embedding and override codes,
1267 so empty embeddings disappear without a trace. We need to
1268 behave as if we did the same. */
1269 struct bidi_it saved_it
;
1270 int level
= prev_level
;
1272 bidi_copy_it (&saved_it
, bidi_it
);
1274 while (bidi_explicit_dir_char (FETCH_CHAR (bidi_it
->bytepos
1275 + bidi_it
->ch_len
)))
1277 level
= bidi_resolve_explicit_1 (bidi_it
);
1280 if (level
== prev_level
) /* empty embedding */
1281 saved_it
.ignore_bn_limit
= bidi_it
->charpos
+ 1;
1282 else /* this embedding is non-empty */
1283 saved_it
.ignore_bn_limit
= -1;
1285 bidi_copy_it (bidi_it
, &saved_it
);
1286 if (bidi_it
->ignore_bn_limit
> 0)
1288 /* We pushed a level, but we shouldn't have. Undo that. */
1289 if (!bidi_it
->invalid_rl_levels
)
1291 new_level
= bidi_pop_embedding_level (bidi_it
);
1292 bidi_it
->invalid_rl_levels
= -1;
1293 if (bidi_it
->invalid_levels
)
1294 bidi_it
->invalid_levels
--;
1296 if (!bidi_it
->invalid_levels
)
1297 new_level
= bidi_pop_embedding_level (bidi_it
);
1300 bidi_it
->invalid_levels
--;
1301 bidi_it
->invalid_rl_levels
--;
1306 if (bidi_it
->type
== NEUTRAL_B
) /* X8 */
1308 bidi_set_paragraph_end (bidi_it
);
1309 /* This is needed by bidi_resolve_weak below, and in L1. */
1310 bidi_it
->type_after_w1
= bidi_it
->type
;
1311 bidi_check_type (bidi_it
->type_after_w1
);
1317 /* Advance in the buffer, resolve weak types and return the type of
1318 the next character after weak type resolution. */
1320 bidi_resolve_weak (struct bidi_it
*bidi_it
)
1323 bidi_dir_t override
;
1324 int prev_level
= bidi_it
->level_stack
[bidi_it
->stack_idx
].level
;
1325 int new_level
= bidi_resolve_explicit (bidi_it
);
1327 bidi_type_t type_of_next
;
1328 struct bidi_it saved_it
;
1330 type
= bidi_it
->type
;
1331 override
= bidi_it
->level_stack
[bidi_it
->stack_idx
].override
;
1333 if (type
== UNKNOWN_BT
1341 if (new_level
!= prev_level
1342 || bidi_it
->type
== NEUTRAL_B
)
1344 /* We've got a new embedding level run, compute the directional
1345 type of sor and initialize per-run variables (UAX#9, clause
1347 bidi_set_sor_type (bidi_it
, prev_level
, new_level
);
1349 else if (type
== NEUTRAL_S
|| type
== NEUTRAL_WS
1350 || type
== WEAK_BN
|| type
== STRONG_AL
)
1351 bidi_it
->type_after_w1
= type
; /* needed in L1 */
1352 bidi_check_type (bidi_it
->type_after_w1
);
1354 /* Level and directional override status are already recorded in
1355 bidi_it, and do not need any change; see X6. */
1356 if (override
== R2L
) /* X6 */
1358 else if (override
== L2R
)
1362 if (type
== WEAK_NSM
) /* W1 */
1364 /* Note that we don't need to consider the case where the
1365 prev character has its type overridden by an RLO or LRO,
1366 because then either the type of this NSM would have been
1367 also overridden, or the previous character is outside the
1368 current level run, and thus not relevant to this NSM.
1369 This is why NSM gets the type_after_w1 of the previous
1371 if (bidi_it
->prev
.type_after_w1
!= UNKNOWN_BT
1372 /* if type_after_w1 is NEUTRAL_B, this NSM is at sor */
1373 && bidi_it
->prev
.type_after_w1
!= NEUTRAL_B
)
1374 type
= bidi_it
->prev
.type_after_w1
;
1375 else if (bidi_it
->sor
== R2L
)
1377 else if (bidi_it
->sor
== L2R
)
1379 else /* shouldn't happen! */
1382 if (type
== WEAK_EN
/* W2 */
1383 && bidi_it
->last_strong
.type_after_w1
== STRONG_AL
)
1385 else if (type
== STRONG_AL
) /* W3 */
1387 else if ((type
== WEAK_ES
/* W4 */
1388 && bidi_it
->prev
.type_after_w1
== WEAK_EN
1389 && bidi_it
->prev
.orig_type
== WEAK_EN
)
1391 && ((bidi_it
->prev
.type_after_w1
== WEAK_EN
1392 && bidi_it
->prev
.orig_type
== WEAK_EN
)
1393 || bidi_it
->prev
.type_after_w1
== WEAK_AN
)))
1396 bidi_it
->bytepos
+ bidi_it
->ch_len
>= ZV_BYTE
1397 ? BIDI_EOB
: FETCH_CHAR (bidi_it
->bytepos
+ bidi_it
->ch_len
);
1398 type_of_next
= bidi_get_type (next_char
, override
);
1400 if (type_of_next
== WEAK_BN
1401 || bidi_explicit_dir_char (next_char
))
1403 bidi_copy_it (&saved_it
, bidi_it
);
1404 while (bidi_resolve_explicit (bidi_it
) == new_level
1405 && bidi_it
->type
== WEAK_BN
)
1407 type_of_next
= bidi_it
->type
;
1408 bidi_copy_it (bidi_it
, &saved_it
);
1411 /* If the next character is EN, but the last strong-type
1412 character is AL, that next EN will be changed to AN when
1413 we process it in W2 above. So in that case, this ES
1414 should not be changed into EN. */
1416 && type_of_next
== WEAK_EN
1417 && bidi_it
->last_strong
.type_after_w1
!= STRONG_AL
)
1419 else if (type
== WEAK_CS
)
1421 if (bidi_it
->prev
.type_after_w1
== WEAK_AN
1422 && (type_of_next
== WEAK_AN
1423 /* If the next character is EN, but the last
1424 strong-type character is AL, EN will be later
1425 changed to AN when we process it in W2 above.
1426 So in that case, this ES should not be
1428 || (type_of_next
== WEAK_EN
1429 && bidi_it
->last_strong
.type_after_w1
== STRONG_AL
)))
1431 else if (bidi_it
->prev
.type_after_w1
== WEAK_EN
1432 && type_of_next
== WEAK_EN
1433 && bidi_it
->last_strong
.type_after_w1
!= STRONG_AL
)
1437 else if (type
== WEAK_ET
/* W5: ET with EN before or after it */
1438 || type
== WEAK_BN
) /* W5/Retaining */
1440 if (bidi_it
->prev
.type_after_w1
== WEAK_EN
/* ET/BN w/EN before it */
1441 || bidi_it
->next_en_pos
> bidi_it
->charpos
)
1443 else /* W5: ET/BN with EN after it. */
1445 EMACS_INT en_pos
= bidi_it
->charpos
+ 1;
1448 bidi_it
->bytepos
+ bidi_it
->ch_len
>= ZV_BYTE
1449 ? BIDI_EOB
: FETCH_CHAR (bidi_it
->bytepos
+ bidi_it
->ch_len
);
1450 type_of_next
= bidi_get_type (next_char
, override
);
1452 if (type_of_next
== WEAK_ET
1453 || type_of_next
== WEAK_BN
1454 || bidi_explicit_dir_char (next_char
))
1456 bidi_copy_it (&saved_it
, bidi_it
);
1457 while (bidi_resolve_explicit (bidi_it
) == new_level
1458 && (bidi_it
->type
== WEAK_BN
1459 || bidi_it
->type
== WEAK_ET
))
1461 type_of_next
= bidi_it
->type
;
1462 en_pos
= bidi_it
->charpos
;
1463 bidi_copy_it (bidi_it
, &saved_it
);
1465 if (type_of_next
== WEAK_EN
)
1467 /* If the last strong character is AL, the EN we've
1468 found will become AN when we get to it (W2). */
1469 if (bidi_it
->last_strong
.type_after_w1
!= STRONG_AL
)
1472 /* Remember this EN position, to speed up processing
1474 bidi_it
->next_en_pos
= en_pos
;
1476 else if (type
== WEAK_BN
)
1477 type
= NEUTRAL_ON
; /* W6/Retaining */
1483 if (type
== WEAK_ES
|| type
== WEAK_ET
|| type
== WEAK_CS
/* W6 */
1485 && (bidi_it
->prev
.type_after_w1
== WEAK_CS
/* W6/Retaining */
1486 || bidi_it
->prev
.type_after_w1
== WEAK_ES
1487 || bidi_it
->prev
.type_after_w1
== WEAK_ET
)))
1490 /* Store the type we've got so far, before we clobber it with strong
1491 types in W7 and while resolving neutral types. But leave alone
1492 the original types that were recorded above, because we will need
1493 them for the L1 clause. */
1494 if (bidi_it
->type_after_w1
== UNKNOWN_BT
)
1495 bidi_it
->type_after_w1
= type
;
1496 bidi_check_type (bidi_it
->type_after_w1
);
1498 if (type
== WEAK_EN
) /* W7 */
1500 if ((bidi_it
->last_strong
.type_after_w1
== STRONG_L
)
1501 || (bidi_it
->last_strong
.type
== UNKNOWN_BT
&& bidi_it
->sor
== L2R
))
1505 bidi_it
->type
= type
;
1506 bidi_check_type (bidi_it
->type
);
1511 bidi_resolve_neutral (struct bidi_it
*bidi_it
)
1513 int prev_level
= bidi_it
->level_stack
[bidi_it
->stack_idx
].level
;
1514 bidi_type_t type
= bidi_resolve_weak (bidi_it
);
1515 int current_level
= bidi_it
->level_stack
[bidi_it
->stack_idx
].level
;
1517 if (!(type
== STRONG_R
1522 || type
== NEUTRAL_B
1523 || type
== NEUTRAL_S
1524 || type
== NEUTRAL_WS
1525 || type
== NEUTRAL_ON
))
1528 if (bidi_get_category (type
) == NEUTRAL
1529 || (type
== WEAK_BN
&& prev_level
== current_level
))
1531 if (bidi_it
->next_for_neutral
.type
!= UNKNOWN_BT
)
1532 type
= bidi_resolve_neutral_1 (bidi_it
->prev_for_neutral
.type
,
1533 bidi_it
->next_for_neutral
.type
,
1537 /* Arrrgh!! The UAX#9 algorithm is too deeply entrenched in
1538 the assumption of batch-style processing; see clauses W4,
1539 W5, and especially N1, which require to look far forward
1540 (as well as back) in the buffer. May the fleas of a
1541 thousand camels infest the armpits of those who design
1542 supposedly general-purpose algorithms by looking at their
1543 own implementations, and fail to consider other possible
1545 struct bidi_it saved_it
;
1546 bidi_type_t next_type
;
1548 if (bidi_it
->scan_dir
== -1)
1551 bidi_copy_it (&saved_it
, bidi_it
);
1552 /* Scan the text forward until we find the first non-neutral
1553 character, and then use that to resolve the neutral we
1554 are dealing with now. We also cache the scanned iterator
1555 states, to salvage some of the effort later. */
1556 bidi_cache_iterator_state (bidi_it
, 0);
1558 /* Record the info about the previous character, so that
1559 it will be cached below with this state. */
1560 if (bidi_it
->type_after_w1
!= WEAK_BN
/* W1/Retaining */
1561 && bidi_it
->type
!= WEAK_BN
)
1562 bidi_remember_char (&bidi_it
->prev
, bidi_it
);
1563 type
= bidi_resolve_weak (bidi_it
);
1564 /* Paragraph separators have their levels fully resolved
1565 at this point, so cache them as resolved. */
1566 bidi_cache_iterator_state (bidi_it
, type
== NEUTRAL_B
);
1567 /* FIXME: implement L1 here, by testing for a newline and
1568 resetting the level for any sequence of whitespace
1569 characters adjacent to it. */
1570 } while (!(type
== NEUTRAL_B
1572 && bidi_get_category (type
) != NEUTRAL
)
1573 /* This is all per level run, so stop when we
1574 reach the end of this level run. */
1575 || bidi_it
->level_stack
[bidi_it
->stack_idx
].level
!=
1578 bidi_remember_char (&saved_it
.next_for_neutral
, bidi_it
);
1589 /* N1: ``European and Arabic numbers are treated as
1590 though they were R.'' */
1591 next_type
= STRONG_R
;
1592 saved_it
.next_for_neutral
.type
= STRONG_R
;
1595 if (!bidi_explicit_dir_char (bidi_it
->ch
))
1596 abort (); /* can't happen: BNs are skipped */
1599 /* Marched all the way to the end of this level run.
1600 We need to use the eor type, whose information is
1601 stored by bidi_set_sor_type in the prev_for_neutral
1603 if (saved_it
.type
!= WEAK_BN
1604 || bidi_get_category (bidi_it
->prev
.type_after_w1
) == NEUTRAL
)
1606 next_type
= bidi_it
->prev_for_neutral
.type
;
1607 saved_it
.next_for_neutral
.type
= next_type
;
1608 bidi_check_type (next_type
);
1612 /* This is a BN which does not adjoin neutrals.
1613 Leave its type alone. */
1614 bidi_copy_it (bidi_it
, &saved_it
);
1615 return bidi_it
->type
;
1621 type
= bidi_resolve_neutral_1 (saved_it
.prev_for_neutral
.type
,
1622 next_type
, current_level
);
1623 saved_it
.type
= type
;
1624 bidi_check_type (type
);
1625 bidi_copy_it (bidi_it
, &saved_it
);
1631 /* Given an iterator state in BIDI_IT, advance one character position
1632 in the buffer to the next character (in the logical order), resolve
1633 the bidi type of that next character, and return that type. */
1635 bidi_type_of_next_char (struct bidi_it
*bidi_it
)
1639 /* This should always be called during a forward scan. */
1640 if (bidi_it
->scan_dir
!= 1)
1643 /* Reset the limit until which to ignore BNs if we step out of the
1644 area where we found only empty levels. */
1645 if ((bidi_it
->ignore_bn_limit
> 0
1646 && bidi_it
->ignore_bn_limit
<= bidi_it
->charpos
)
1647 || (bidi_it
->ignore_bn_limit
== -1
1648 && !bidi_explicit_dir_char (bidi_it
->ch
)))
1649 bidi_it
->ignore_bn_limit
= 0;
1651 type
= bidi_resolve_neutral (bidi_it
);
1656 /* Given an iterator state BIDI_IT, advance one character position in
1657 the buffer to the next character (in the logical order), resolve
1658 the embedding and implicit levels of that next character, and
1659 return the resulting level. */
1661 bidi_level_of_next_char (struct bidi_it
*bidi_it
)
1664 int level
, prev_level
= -1;
1665 struct bidi_saved_info next_for_neutral
;
1667 if (bidi_it
->scan_dir
== 1)
1669 /* There's no sense in trying to advance if we hit end of text. */
1670 if (bidi_it
->bytepos
>= ZV_BYTE
)
1671 return bidi_it
->resolved_level
;
1673 /* Record the info about the previous character. */
1674 if (bidi_it
->type_after_w1
!= WEAK_BN
/* W1/Retaining */
1675 && bidi_it
->type
!= WEAK_BN
)
1676 bidi_remember_char (&bidi_it
->prev
, bidi_it
);
1677 if (bidi_it
->type_after_w1
== STRONG_R
1678 || bidi_it
->type_after_w1
== STRONG_L
1679 || bidi_it
->type_after_w1
== STRONG_AL
)
1680 bidi_remember_char (&bidi_it
->last_strong
, bidi_it
);
1681 /* FIXME: it sounds like we don't need both prev and
1682 prev_for_neutral members, but I'm leaving them both for now. */
1683 if (bidi_it
->type
== STRONG_R
|| bidi_it
->type
== STRONG_L
1684 || bidi_it
->type
== WEAK_EN
|| bidi_it
->type
== WEAK_AN
)
1685 bidi_remember_char (&bidi_it
->prev_for_neutral
, bidi_it
);
1687 /* If we overstepped the characters used for resolving neutrals
1688 and whitespace, invalidate their info in the iterator. */
1689 if (bidi_it
->charpos
>= bidi_it
->next_for_neutral
.charpos
)
1690 bidi_it
->next_for_neutral
.type
= UNKNOWN_BT
;
1691 if (bidi_it
->next_en_pos
>= 0
1692 && bidi_it
->charpos
>= bidi_it
->next_en_pos
)
1693 bidi_it
->next_en_pos
= -1;
1694 if (bidi_it
->next_for_ws
.type
!= UNKNOWN_BT
1695 && bidi_it
->charpos
>= bidi_it
->next_for_ws
.charpos
)
1696 bidi_it
->next_for_ws
.type
= UNKNOWN_BT
;
1698 /* This must be taken before we fill the iterator with the info
1699 about the next char. If we scan backwards, the iterator
1700 state must be already cached, so there's no need to know the
1701 embedding level of the previous character, since we will be
1702 returning to our caller shortly. */
1703 prev_level
= bidi_it
->level_stack
[bidi_it
->stack_idx
].level
;
1705 next_for_neutral
= bidi_it
->next_for_neutral
;
1707 /* Perhaps it is already cached. */
1708 type
= bidi_cache_find (bidi_it
->charpos
+ bidi_it
->scan_dir
, -1, bidi_it
);
1709 if (type
!= UNKNOWN_BT
)
1711 /* Don't lose the information for resolving neutrals! The
1712 cached states could have been cached before their
1713 next_for_neutral member was computed. If we are on our way
1714 forward, we can simply take the info from the previous
1716 if (bidi_it
->scan_dir
== 1
1717 && bidi_it
->next_for_neutral
.type
== UNKNOWN_BT
)
1718 bidi_it
->next_for_neutral
= next_for_neutral
;
1720 /* If resolved_level is -1, it means this state was cached
1721 before it was completely resolved, so we cannot return
1723 if (bidi_it
->resolved_level
!= -1)
1724 return bidi_it
->resolved_level
;
1726 if (bidi_it
->scan_dir
== -1)
1727 /* If we are going backwards, the iterator state is already cached
1728 from previous scans, and should be fully resolved. */
1731 if (type
== UNKNOWN_BT
)
1732 type
= bidi_type_of_next_char (bidi_it
);
1734 if (type
== NEUTRAL_B
)
1735 return bidi_it
->resolved_level
;
1737 level
= bidi_it
->level_stack
[bidi_it
->stack_idx
].level
;
1738 if ((bidi_get_category (type
) == NEUTRAL
/* && type != NEUTRAL_B */)
1739 || (type
== WEAK_BN
&& prev_level
== level
))
1741 if (bidi_it
->next_for_neutral
.type
== UNKNOWN_BT
)
1744 /* If the cached state shows a neutral character, it was not
1745 resolved by bidi_resolve_neutral, so do it now. */
1746 type
= bidi_resolve_neutral_1 (bidi_it
->prev_for_neutral
.type
,
1747 bidi_it
->next_for_neutral
.type
,
1751 if (!(type
== STRONG_R
1755 || type
== WEAK_AN
))
1757 bidi_it
->type
= type
;
1758 bidi_check_type (bidi_it
->type
);
1760 /* For L1 below, we need to know, for each WS character, whether
1761 it belongs to a sequence of WS characters preceeding a newline
1762 or a TAB or a paragraph separator. */
1763 if (bidi_it
->orig_type
== NEUTRAL_WS
1764 && bidi_it
->next_for_ws
.type
== UNKNOWN_BT
)
1767 int clen
= bidi_it
->ch_len
;
1768 EMACS_INT bpos
= bidi_it
->bytepos
;
1769 EMACS_INT cpos
= bidi_it
->charpos
;
1773 /*_fetch_multibyte_char_len = 1;*/
1774 ch
= bpos
+ clen
>= ZV_BYTE
? BIDI_EOB
: FETCH_CHAR (bpos
+ clen
);
1777 clen
= (ch
== BIDI_EOB
? 1 : CHAR_BYTES (ch
));
1778 if (ch
== '\n' || ch
== BIDI_EOB
/* || ch == LINESEP_CHAR */)
1781 chtype
= bidi_get_type (ch
, NEUTRAL_DIR
);
1782 } while (chtype
== NEUTRAL_WS
|| chtype
== WEAK_BN
1783 || bidi_explicit_dir_char (ch
)); /* L1/Retaining */
1784 bidi_it
->next_for_ws
.type
= chtype
;
1785 bidi_check_type (bidi_it
->next_for_ws
.type
);
1786 bidi_it
->next_for_ws
.charpos
= cpos
;
1787 bidi_it
->next_for_ws
.bytepos
= bpos
;
1790 /* Resolve implicit levels, with a twist: PDFs get the embedding
1791 level of the enbedding they terminate. See below for the
1793 if (bidi_it
->orig_type
== PDF
1794 /* Don't do this if this formatting code didn't change the
1795 embedding level due to invalid or empty embeddings. */
1796 && prev_level
!= level
)
1798 /* Don't look in UAX#9 for the reason for this: it's our own
1799 private quirk. The reason is that we want the formatting
1800 codes to be delivered so that they bracket the text of their
1801 embedding. For example, given the text
1805 we want it to be displayed as
1813 which will result because we bump up the embedding level as
1814 soon as we see the RLO and pop it as soon as we see the PDF,
1815 so RLO itself has the same embedding level as "teST", and
1816 thus would be normally delivered last, just before the PDF.
1817 The switch below fiddles with the level of PDF so that this
1818 ugly side effect does not happen.
1820 (This is, of course, only important if the formatting codes
1821 are actually displayed, but Emacs does need to display them
1822 if the user wants to.) */
1825 else if (bidi_it
->orig_type
== NEUTRAL_B
/* L1 */
1826 || bidi_it
->orig_type
== NEUTRAL_S
1827 || bidi_it
->ch
== '\n' || bidi_it
->ch
== BIDI_EOB
1828 /* || bidi_it->ch == LINESEP_CHAR */
1829 || (bidi_it
->orig_type
== NEUTRAL_WS
1830 && (bidi_it
->next_for_ws
.type
== NEUTRAL_B
1831 || bidi_it
->next_for_ws
.type
== NEUTRAL_S
)))
1832 level
= bidi_it
->level_stack
[0].level
;
1833 else if ((level
& 1) == 0) /* I1 */
1835 if (type
== STRONG_R
)
1837 else if (type
== WEAK_EN
|| type
== WEAK_AN
)
1842 if (type
== STRONG_L
|| type
== WEAK_EN
|| type
== WEAK_AN
)
1846 bidi_it
->resolved_level
= level
;
1850 /* Move to the other edge of a level given by LEVEL. If END_FLAG is
1851 non-zero, we are at the end of a level, and we need to prepare to
1852 resume the scan of the lower level.
1854 If this level's other edge is cached, we simply jump to it, filling
1855 the iterator structure with the iterator state on the other edge.
1856 Otherwise, we walk the buffer until we come back to the same level
1859 Note: we are not talking here about a ``level run'' in the UAX#9
1860 sense of the term, but rather about a ``level'' which includes
1861 all the levels higher than it. In other words, given the levels
1864 11111112222222333333334443343222222111111112223322111
1867 and assuming we are at point A scanning left to right, this
1868 function moves to point C, whereas the UAX#9 ``level 2 run'' ends
1871 bidi_find_other_level_edge (struct bidi_it
*bidi_it
, int level
, int end_flag
)
1873 int dir
= end_flag
? -bidi_it
->scan_dir
: bidi_it
->scan_dir
;
1876 /* Try the cache first. */
1877 if ((idx
= bidi_cache_find_level_change (level
, dir
, end_flag
)) >= 0)
1878 bidi_cache_fetch_state (idx
, bidi_it
);
1884 abort (); /* if we are at end of level, its edges must be cached */
1886 bidi_cache_iterator_state (bidi_it
, 1);
1888 new_level
= bidi_level_of_next_char (bidi_it
);
1889 bidi_cache_iterator_state (bidi_it
, 1);
1890 } while (new_level
>= level
);
1895 bidi_move_to_visually_next (struct bidi_it
*bidi_it
)
1897 int old_level
, new_level
, next_level
;
1898 struct bidi_it sentinel
;
1900 if (bidi_it
->scan_dir
== 0)
1902 bidi_it
->scan_dir
= 1; /* default to logical order */
1905 /* If we just passed a newline, initialize for the next line. */
1906 if (!bidi_it
->first_elt
&& bidi_it
->orig_type
== NEUTRAL_B
)
1907 bidi_line_init (bidi_it
);
1909 /* Prepare the sentinel iterator state. */
1910 if (bidi_cache_idx
== 0)
1912 bidi_copy_it (&sentinel
, bidi_it
);
1913 if (bidi_it
->first_elt
)
1915 sentinel
.charpos
--; /* cached charpos needs to be monotonic */
1917 sentinel
.ch
= '\n'; /* doesn't matter, but why not? */
1918 sentinel
.ch_len
= 1;
1922 old_level
= bidi_it
->resolved_level
;
1923 new_level
= bidi_level_of_next_char (bidi_it
);
1925 /* Reordering of resolved levels (clause L2) is implemented by
1926 jumping to the other edge of the level and flipping direction of
1927 scanning the text whenever we find a level change. */
1928 if (new_level
!= old_level
)
1930 int ascending
= new_level
> old_level
;
1931 int level_to_search
= ascending
? old_level
+ 1 : old_level
;
1932 int incr
= ascending
? 1 : -1;
1933 int expected_next_level
= old_level
+ incr
;
1935 /* If we don't have anything cached yet, we need to cache the
1936 sentinel state, since we'll need it to record where to jump
1937 when the last non-base level is exhausted. */
1938 if (bidi_cache_idx
== 0)
1939 bidi_cache_iterator_state (&sentinel
, 1);
1940 /* Jump (or walk) to the other edge of this level. */
1941 bidi_find_other_level_edge (bidi_it
, level_to_search
, !ascending
);
1942 /* Switch scan direction and peek at the next character in the
1944 bidi_it
->scan_dir
= -bidi_it
->scan_dir
;
1946 /* The following loop handles the case where the resolved level
1947 jumps by more than one. This is typical for numbers inside a
1948 run of text with left-to-right embedding direction, but can
1949 also happen in other situations. In those cases the decision
1950 where to continue after a level change, and in what direction,
1951 is tricky. For example, given a text like below:
1956 (where the numbers below the text show the resolved levels),
1957 the result of reordering according to UAX#9 should be this:
1961 This is implemented by the loop below which flips direction
1962 and jumps to the other edge of the level each time it finds
1963 the new level not to be the expected one. The expected level
1964 is always one more or one less than the previous one. */
1965 next_level
= bidi_peek_at_next_level (bidi_it
);
1966 while (next_level
!= expected_next_level
)
1968 expected_next_level
+= incr
;
1969 level_to_search
+= incr
;
1970 bidi_find_other_level_edge (bidi_it
, level_to_search
, !ascending
);
1971 bidi_it
->scan_dir
= -bidi_it
->scan_dir
;
1972 next_level
= bidi_peek_at_next_level (bidi_it
);
1975 /* Finally, deliver the next character in the new direction. */
1976 next_level
= bidi_level_of_next_char (bidi_it
);
1979 /* Take note when we have just processed the newline that precedes
1980 the end of the paragraph. The next time we are about to be
1981 called, set_iterator_to_next will automatically reinit the
1982 paragraph direction, if needed. We do this at the newline before
1983 the paragraph separator, because the next character might not be
1984 the first character of the next paragraph, due to the bidi
1985 reordering, whereas we _must_ know the paragraph base direction
1986 _before_ we process the paragraph's text, since the base
1987 direction affects the reordering. */
1988 if (bidi_it
->scan_dir
== 1
1989 && bidi_it
->orig_type
== NEUTRAL_B
1990 && bidi_it
->bytepos
< ZV_BYTE
)
1993 bidi_at_paragraph_end (bidi_it
->charpos
+ 1,
1994 bidi_it
->bytepos
+ bidi_it
->ch_len
);
1997 bidi_it
->new_paragraph
= 1;
1998 /* Record the buffer position of the last character of the
1999 paragraph separator. */
2000 bidi_it
->separator_limit
= bidi_it
->charpos
+ 1 + sep_len
;
2004 if (bidi_it
->scan_dir
== 1 && bidi_cache_idx
)
2006 /* If we are at paragraph's base embedding level and beyond the
2007 last cached position, the cache's job is done and we can
2009 if (bidi_it
->resolved_level
== bidi_it
->level_stack
[0].level
2010 && bidi_it
->charpos
> bidi_cache
[bidi_cache_idx
- 1].charpos
)
2011 bidi_cache_reset ();
2012 /* But as long as we are caching during forward scan, we must
2013 cache each state, or else the cache integrity will be
2014 compromised: it assumes cached states correspond to buffer
2017 bidi_cache_iterator_state (bidi_it
, 1);
2021 /* This is meant to be called from within the debugger, whenever you
2022 wish to examine the cache contents. */
2024 bidi_dump_cached_states (void)
2029 if (bidi_cache_idx
== 0)
2031 fprintf (stderr
, "The cache is empty.\n");
2034 fprintf (stderr
, "Total of %d state%s in cache:\n",
2035 bidi_cache_idx
, bidi_cache_idx
== 1 ? "" : "s");
2037 for (i
= bidi_cache
[bidi_cache_idx
- 1].charpos
; i
> 0; i
/= 10)
2039 fputs ("ch ", stderr
);
2040 for (i
= 0; i
< bidi_cache_idx
; i
++)
2041 fprintf (stderr
, "%*c", ndigits
, bidi_cache
[i
].ch
);
2042 fputs ("\n", stderr
);
2043 fputs ("lvl ", stderr
);
2044 for (i
= 0; i
< bidi_cache_idx
; i
++)
2045 fprintf (stderr
, "%*d", ndigits
, bidi_cache
[i
].resolved_level
);
2046 fputs ("\n", stderr
);
2047 fputs ("pos ", stderr
);
2048 for (i
= 0; i
< bidi_cache_idx
; i
++)
2049 fprintf (stderr
, "%*d", ndigits
, bidi_cache
[i
].charpos
);
2050 fputs ("\n", stderr
);