1 /* Low-level bidirectional buffer/string-scanning functions for GNU Emacs.
2 Copyright (C) 2000-2001, 2004-2005, 2009-2011
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 (UBA) 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 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 A few auxiliary entry points are used to initialize the bidi
44 iterator for iterating an object (buffer or string), push and pop
45 the bidi iterator state, and save and restore the state of the bidi
48 If you want to understand the code, you will have to read it
49 together with the relevant portions of UAX#9. The comments include
50 references to UAX#9 rules, for that very reason.
52 A note about references to UAX#9 rules: if the reference says
53 something like "X9/Retaining", it means that you need to refer to
54 rule X9 and to its modifications decribed in the "Implementation
55 Notes" section of UAX#9, under "Retaining Format Codes". */
63 #include "character.h"
64 #include "dispextern.h"
66 static int bidi_initialized
= 0;
68 static Lisp_Object bidi_type_table
, bidi_mirror_table
;
70 #define LRM_CHAR 0x200E
71 #define RLM_CHAR 0x200F
74 /* Data type for describing the bidirectional character categories. */
82 /* UAX#9 says to search only for L, AL, or R types of characters, and
83 ignore RLE, RLO, LRE, and LRO, when determining the base paragraph
84 level. Yudit indeed ignores them. This variable is therefore set
85 by default to ignore them, but setting it to zero will take them
87 extern int bidi_ignore_explicit_marks_for_paragraph_level EXTERNALLY_VISIBLE
;
88 int bidi_ignore_explicit_marks_for_paragraph_level
= 1;
90 static Lisp_Object paragraph_start_re
, paragraph_separate_re
;
91 static Lisp_Object Qparagraph_start
, Qparagraph_separate
;
94 /***********************************************************************
96 ***********************************************************************/
98 /* Return the bidi type of a character CH, subject to the current
99 directional OVERRIDE. */
100 static inline bidi_type_t
101 bidi_get_type (int ch
, bidi_dir_t override
)
103 bidi_type_t default_type
;
107 if (ch
< 0 || ch
> MAX_CHAR
)
110 default_type
= (bidi_type_t
) XINT (CHAR_TABLE_REF (bidi_type_table
, ch
));
111 /* Every valid character code, even those that are unassigned by the
112 UCD, have some bidi-class property, according to
113 DerivedBidiClass.txt file. Therefore, if we ever get UNKNOWN_BT
114 (= zero) code from CHAR_TABLE_REF, that's a bug. */
115 if (default_type
== UNKNOWN_BT
)
118 if (override
== NEUTRAL_DIR
)
121 switch (default_type
)
123 /* Although UAX#9 does not tell, it doesn't make sense to
124 override NEUTRAL_B and LRM/RLM characters. */
139 if (override
== L2R
) /* X6 */
141 else if (override
== R2L
)
144 abort (); /* can't happen: handled above */
150 bidi_check_type (bidi_type_t type
)
152 xassert (UNKNOWN_BT
<= type
&& type
<= NEUTRAL_ON
);
155 /* Given a bidi TYPE of a character, return its category. */
156 static inline bidi_category_t
157 bidi_get_category (bidi_type_t type
)
171 case PDF
: /* ??? really?? */
190 /* Return the mirrored character of C, if it has one. If C has no
191 mirrored counterpart, return C.
192 Note: The conditions in UAX#9 clause L4 regarding the surrounding
193 context must be tested by the caller. */
195 bidi_mirror_char (int c
)
201 if (c
< 0 || c
> MAX_CHAR
)
204 val
= CHAR_TABLE_REF (bidi_mirror_table
, c
);
209 if (v
< 0 || v
> MAX_CHAR
)
218 /* Determine the start-of-run (sor) directional type given the two
219 embedding levels on either side of the run boundary. Also, update
220 the saved info about previously seen characters, since that info is
221 generally valid for a single level run. */
223 bidi_set_sor_type (struct bidi_it
*bidi_it
, int level_before
, int level_after
)
225 int higher_level
= level_before
> level_after
? level_before
: level_after
;
227 /* The prev_was_pdf gork is required for when we have several PDFs
228 in a row. In that case, we want to compute the sor type for the
229 next level run only once: when we see the first PDF. That's
230 because the sor type depends only on the higher of the two levels
231 that we find on the two sides of the level boundary (see UAX#9,
232 clause X10), and so we don't need to know the final embedding
233 level to which we descend after processing all the PDFs. */
234 if (!bidi_it
->prev_was_pdf
|| level_before
< level_after
)
235 /* FIXME: should the default sor direction be user selectable? */
236 bidi_it
->sor
= (higher_level
& 1) != 0 ? R2L
: L2R
;
237 if (level_before
> level_after
)
238 bidi_it
->prev_was_pdf
= 1;
240 bidi_it
->prev
.type
= UNKNOWN_BT
;
241 bidi_it
->last_strong
.type
= bidi_it
->last_strong
.type_after_w1
=
242 bidi_it
->last_strong
.orig_type
= UNKNOWN_BT
;
243 bidi_it
->prev_for_neutral
.type
= bidi_it
->sor
== R2L
? STRONG_R
: STRONG_L
;
244 bidi_it
->prev_for_neutral
.charpos
= bidi_it
->charpos
;
245 bidi_it
->prev_for_neutral
.bytepos
= bidi_it
->bytepos
;
246 bidi_it
->next_for_neutral
.type
= bidi_it
->next_for_neutral
.type_after_w1
=
247 bidi_it
->next_for_neutral
.orig_type
= UNKNOWN_BT
;
248 bidi_it
->ignore_bn_limit
= -1; /* meaning it's unknown */
251 /* Push the current embedding level and override status; reset the
252 current level to LEVEL and the current override status to OVERRIDE. */
254 bidi_push_embedding_level (struct bidi_it
*bidi_it
,
255 int level
, bidi_dir_t override
)
257 bidi_it
->stack_idx
++;
258 xassert (bidi_it
->stack_idx
< BIDI_MAXLEVEL
);
259 bidi_it
->level_stack
[bidi_it
->stack_idx
].level
= level
;
260 bidi_it
->level_stack
[bidi_it
->stack_idx
].override
= override
;
263 /* Pop the embedding level and directional override status from the
264 stack, and return the new level. */
266 bidi_pop_embedding_level (struct bidi_it
*bidi_it
)
268 /* UAX#9 says to ignore invalid PDFs. */
269 if (bidi_it
->stack_idx
> 0)
270 bidi_it
->stack_idx
--;
271 return bidi_it
->level_stack
[bidi_it
->stack_idx
].level
;
274 /* Record in SAVED_INFO the information about the current character. */
276 bidi_remember_char (struct bidi_saved_info
*saved_info
,
277 struct bidi_it
*bidi_it
)
279 saved_info
->charpos
= bidi_it
->charpos
;
280 saved_info
->bytepos
= bidi_it
->bytepos
;
281 saved_info
->type
= bidi_it
->type
;
282 bidi_check_type (bidi_it
->type
);
283 saved_info
->type_after_w1
= bidi_it
->type_after_w1
;
284 bidi_check_type (bidi_it
->type_after_w1
);
285 saved_info
->orig_type
= bidi_it
->orig_type
;
286 bidi_check_type (bidi_it
->orig_type
);
289 /* Copy the bidi iterator from FROM to TO. To save cycles, this only
290 copies the part of the level stack that is actually in use. */
292 bidi_copy_it (struct bidi_it
*to
, struct bidi_it
*from
)
296 /* Copy everything except the level stack and beyond. */
297 memcpy (to
, from
, offsetof (struct bidi_it
, level_stack
[0]));
299 /* Copy the active part of the level stack. */
300 to
->level_stack
[0] = from
->level_stack
[0]; /* level zero is always in use */
301 for (i
= 1; i
<= from
->stack_idx
; i
++)
302 to
->level_stack
[i
] = from
->level_stack
[i
];
306 /***********************************************************************
307 Caching the bidi iterator states
308 ***********************************************************************/
310 #define BIDI_CACHE_CHUNK 200
311 static struct bidi_it
*bidi_cache
;
312 static ptrdiff_t bidi_cache_size
= 0;
313 enum { elsz
= sizeof (struct bidi_it
) };
314 static ptrdiff_t bidi_cache_idx
; /* next unused cache slot */
315 static ptrdiff_t bidi_cache_last_idx
; /* slot of last cache hit */
316 static ptrdiff_t bidi_cache_start
= 0; /* start of cache for this
319 /* Reset the cache state to the empty state. We only reset the part
320 of the cache relevant to iteration of the current object. Previous
321 objects, which are pushed on the display iterator's stack, are left
322 intact. This is called when the cached information is no more
323 useful for the current iteration, e.g. when we were reseated to a
324 new position on the same object. */
326 bidi_cache_reset (void)
328 bidi_cache_idx
= bidi_cache_start
;
329 bidi_cache_last_idx
= -1;
332 /* Shrink the cache to its minimal size. Called when we init the bidi
333 iterator for reordering a buffer or a string that does not come
334 from display properties, because that means all the previously
335 cached info is of no further use. */
337 bidi_cache_shrink (void)
339 if (bidi_cache_size
> BIDI_CACHE_CHUNK
)
341 bidi_cache_size
= BIDI_CACHE_CHUNK
;
343 (struct bidi_it
*) xrealloc (bidi_cache
, bidi_cache_size
* elsz
);
349 bidi_cache_fetch_state (ptrdiff_t idx
, struct bidi_it
*bidi_it
)
351 int current_scan_dir
= bidi_it
->scan_dir
;
353 if (idx
< bidi_cache_start
|| idx
>= bidi_cache_idx
)
356 bidi_copy_it (bidi_it
, &bidi_cache
[idx
]);
357 bidi_it
->scan_dir
= current_scan_dir
;
358 bidi_cache_last_idx
= idx
;
361 /* Find a cached state with a given CHARPOS and resolved embedding
362 level less or equal to LEVEL. if LEVEL is -1, disregard the
363 resolved levels in cached states. DIR, if non-zero, means search
364 in that direction from the last cache hit. */
365 static inline ptrdiff_t
366 bidi_cache_search (EMACS_INT charpos
, int level
, int dir
)
368 ptrdiff_t i
, i_start
;
370 if (bidi_cache_idx
> bidi_cache_start
)
372 if (bidi_cache_last_idx
== -1)
373 bidi_cache_last_idx
= bidi_cache_idx
- 1;
374 if (charpos
< bidi_cache
[bidi_cache_last_idx
].charpos
)
377 i_start
= bidi_cache_last_idx
- 1;
379 else if (charpos
> (bidi_cache
[bidi_cache_last_idx
].charpos
380 + bidi_cache
[bidi_cache_last_idx
].nchars
- 1))
383 i_start
= bidi_cache_last_idx
+ 1;
386 i_start
= bidi_cache_last_idx
;
390 i_start
= bidi_cache_idx
- 1;
395 /* Linear search for now; FIXME! */
396 for (i
= i_start
; i
>= bidi_cache_start
; i
--)
397 if (bidi_cache
[i
].charpos
<= charpos
398 && charpos
< bidi_cache
[i
].charpos
+ bidi_cache
[i
].nchars
399 && (level
== -1 || bidi_cache
[i
].resolved_level
<= level
))
404 for (i
= i_start
; i
< bidi_cache_idx
; i
++)
405 if (bidi_cache
[i
].charpos
<= charpos
406 && charpos
< bidi_cache
[i
].charpos
+ bidi_cache
[i
].nchars
407 && (level
== -1 || bidi_cache
[i
].resolved_level
<= level
))
415 /* Find a cached state where the resolved level changes to a value
416 that is lower than LEVEL, and return its cache slot index. DIR is
417 the direction to search, starting with the last used cache slot.
418 If DIR is zero, we search backwards from the last occupied cache
419 slot. BEFORE, if non-zero, means return the index of the slot that
420 is ``before'' the level change in the search direction. That is,
421 given the cached levels like this:
426 and assuming we are at the position cached at the slot marked with
427 C, searching backwards (DIR = -1) for LEVEL = 2 will return the
428 index of slot B or A, depending whether BEFORE is, respectively,
431 bidi_cache_find_level_change (int level
, int dir
, int before
)
435 ptrdiff_t i
= dir
? bidi_cache_last_idx
: bidi_cache_idx
- 1;
436 int incr
= before
? 1 : 0;
438 xassert (!dir
|| bidi_cache_last_idx
>= 0);
447 while (i
>= bidi_cache_start
+ incr
)
449 if (bidi_cache
[i
- incr
].resolved_level
>= 0
450 && bidi_cache
[i
- incr
].resolved_level
< level
)
457 while (i
< bidi_cache_idx
- incr
)
459 if (bidi_cache
[i
+ incr
].resolved_level
>= 0
460 && bidi_cache
[i
+ incr
].resolved_level
< level
)
471 bidi_cache_ensure_space (ptrdiff_t idx
)
473 /* Enlarge the cache as needed. */
474 if (idx
>= bidi_cache_size
)
478 /* The bidi cache cannot be larger than the largest Lisp string
480 ptrdiff_t string_or_buffer_bound
=
481 max (BUF_BYTES_MAX
, STRING_BYTES_BOUND
);
483 /* Also, it cannot be larger than what C can represent. */
484 ptrdiff_t c_bound
= min (PTRDIFF_MAX
, SIZE_MAX
) / elsz
;
486 if (min (string_or_buffer_bound
, c_bound
) <= idx
)
487 memory_full (SIZE_MAX
);
488 new_size
= idx
- idx
% BIDI_CACHE_CHUNK
+ BIDI_CACHE_CHUNK
;
489 bidi_cache
= (struct bidi_it
*) xrealloc (bidi_cache
, new_size
* elsz
);
490 bidi_cache_size
= new_size
;
495 bidi_cache_iterator_state (struct bidi_it
*bidi_it
, int resolved
)
499 /* We should never cache on backward scans. */
500 if (bidi_it
->scan_dir
== -1)
502 idx
= bidi_cache_search (bidi_it
->charpos
, -1, 1);
506 idx
= bidi_cache_idx
;
507 bidi_cache_ensure_space (idx
);
508 /* Character positions should correspond to cache positions 1:1.
509 If we are outside the range of cached positions, the cache is
510 useless and must be reset. */
511 if (idx
> bidi_cache_start
&&
512 (bidi_it
->charpos
> (bidi_cache
[idx
- 1].charpos
513 + bidi_cache
[idx
- 1].nchars
)
514 || bidi_it
->charpos
< bidi_cache
[bidi_cache_start
].charpos
))
517 idx
= bidi_cache_start
;
519 if (bidi_it
->nchars
<= 0)
521 bidi_copy_it (&bidi_cache
[idx
], bidi_it
);
523 bidi_cache
[idx
].resolved_level
= -1;
527 /* Copy only the members which could have changed, to avoid
528 costly copying of the entire struct. */
529 bidi_cache
[idx
].type
= bidi_it
->type
;
530 bidi_check_type (bidi_it
->type
);
531 bidi_cache
[idx
].type_after_w1
= bidi_it
->type_after_w1
;
532 bidi_check_type (bidi_it
->type_after_w1
);
534 bidi_cache
[idx
].resolved_level
= bidi_it
->resolved_level
;
536 bidi_cache
[idx
].resolved_level
= -1;
537 bidi_cache
[idx
].invalid_levels
= bidi_it
->invalid_levels
;
538 bidi_cache
[idx
].invalid_rl_levels
= bidi_it
->invalid_rl_levels
;
539 bidi_cache
[idx
].next_for_neutral
= bidi_it
->next_for_neutral
;
540 bidi_cache
[idx
].next_for_ws
= bidi_it
->next_for_ws
;
541 bidi_cache
[idx
].ignore_bn_limit
= bidi_it
->ignore_bn_limit
;
542 bidi_cache
[idx
].disp_pos
= bidi_it
->disp_pos
;
543 bidi_cache
[idx
].disp_prop
= bidi_it
->disp_prop
;
546 bidi_cache_last_idx
= idx
;
547 if (idx
>= bidi_cache_idx
)
548 bidi_cache_idx
= idx
+ 1;
551 static inline bidi_type_t
552 bidi_cache_find (EMACS_INT charpos
, int level
, struct bidi_it
*bidi_it
)
554 ptrdiff_t i
= bidi_cache_search (charpos
, level
, bidi_it
->scan_dir
);
556 if (i
>= bidi_cache_start
)
558 bidi_dir_t current_scan_dir
= bidi_it
->scan_dir
;
560 bidi_copy_it (bidi_it
, &bidi_cache
[i
]);
561 bidi_cache_last_idx
= i
;
562 /* Don't let scan direction from from the cached state override
563 the current scan direction. */
564 bidi_it
->scan_dir
= current_scan_dir
;
565 return bidi_it
->type
;
572 bidi_peek_at_next_level (struct bidi_it
*bidi_it
)
574 if (bidi_cache_idx
== bidi_cache_start
|| bidi_cache_last_idx
== -1)
576 return bidi_cache
[bidi_cache_last_idx
+ bidi_it
->scan_dir
].resolved_level
;
580 /***********************************************************************
581 Pushing and popping the bidi iterator state
582 ***********************************************************************/
583 /* 5-slot stack for saving the start of the previous level of the
584 cache. xdisp.c maintains a 5-slot stack for its iterator state,
585 and we need the same size of our stack. */
586 static ptrdiff_t bidi_cache_start_stack
[IT_STACK_SIZE
];
587 static int bidi_cache_sp
;
589 /* Push the bidi iterator state in preparation for reordering a
590 different object, e.g. display string found at certain buffer
591 position. Pushing the bidi iterator boils down to saving its
592 entire state on the cache and starting a new cache "stacked" on top
593 of the current cache. */
595 bidi_push_it (struct bidi_it
*bidi_it
)
597 /* Save the current iterator state in its entirety after the last
599 bidi_cache_ensure_space (bidi_cache_idx
);
600 memcpy (&bidi_cache
[bidi_cache_idx
++], bidi_it
, sizeof (struct bidi_it
));
602 /* Push the current cache start onto the stack. */
603 xassert (bidi_cache_sp
< IT_STACK_SIZE
);
604 bidi_cache_start_stack
[bidi_cache_sp
++] = bidi_cache_start
;
606 /* Start a new level of cache, and make it empty. */
607 bidi_cache_start
= bidi_cache_idx
;
608 bidi_cache_last_idx
= -1;
611 /* Restore the iterator state saved by bidi_push_it and return the
612 cache to the corresponding state. */
614 bidi_pop_it (struct bidi_it
*bidi_it
)
616 if (bidi_cache_start
<= 0)
619 /* Reset the next free cache slot index to what it was before the
620 call to bidi_push_it. */
621 bidi_cache_idx
= bidi_cache_start
- 1;
623 /* Restore the bidi iterator state saved in the cache. */
624 memcpy (bidi_it
, &bidi_cache
[bidi_cache_idx
], sizeof (struct bidi_it
));
626 /* Pop the previous cache start from the stack. */
627 if (bidi_cache_sp
<= 0)
629 bidi_cache_start
= bidi_cache_start_stack
[--bidi_cache_sp
];
631 /* Invalidate the last-used cache slot data. */
632 bidi_cache_last_idx
= -1;
635 static ptrdiff_t bidi_cache_total_alloc
;
637 /* Stash away a copy of the cache and its control variables. */
639 bidi_shelve_cache (void)
641 unsigned char *databuf
;
644 if (bidi_cache_idx
== 0)
647 databuf
= xmalloc (sizeof (bidi_cache_idx
)
648 + bidi_cache_idx
* sizeof (struct bidi_it
)
649 + sizeof (bidi_cache_start_stack
)
650 + sizeof (bidi_cache_sp
) + sizeof (bidi_cache_start
)
651 + sizeof (bidi_cache_last_idx
));
652 bidi_cache_total_alloc
+=
653 sizeof (bidi_cache_idx
) + bidi_cache_idx
* sizeof (struct bidi_it
)
654 + sizeof (bidi_cache_start_stack
)
655 + sizeof (bidi_cache_sp
) + sizeof (bidi_cache_start
)
656 + sizeof (bidi_cache_last_idx
);
658 memcpy (databuf
, &bidi_cache_idx
, sizeof (bidi_cache_idx
));
659 memcpy (databuf
+ sizeof (bidi_cache_idx
),
660 bidi_cache
, bidi_cache_idx
* sizeof (struct bidi_it
));
661 memcpy (databuf
+ sizeof (bidi_cache_idx
)
662 + bidi_cache_idx
* sizeof (struct bidi_it
),
663 bidi_cache_start_stack
, sizeof (bidi_cache_start_stack
));
664 memcpy (databuf
+ sizeof (bidi_cache_idx
)
665 + bidi_cache_idx
* sizeof (struct bidi_it
)
666 + sizeof (bidi_cache_start_stack
),
667 &bidi_cache_sp
, sizeof (bidi_cache_sp
));
668 memcpy (databuf
+ sizeof (bidi_cache_idx
)
669 + bidi_cache_idx
* sizeof (struct bidi_it
)
670 + sizeof (bidi_cache_start_stack
) + sizeof (bidi_cache_sp
),
671 &bidi_cache_start
, sizeof (bidi_cache_start
));
672 memcpy (databuf
+ sizeof (bidi_cache_idx
)
673 + bidi_cache_idx
* sizeof (struct bidi_it
)
674 + sizeof (bidi_cache_start_stack
) + sizeof (bidi_cache_sp
)
675 + sizeof (bidi_cache_start
),
676 &bidi_cache_last_idx
, sizeof (bidi_cache_last_idx
));
681 /* Restore the cache state from a copy stashed away by
682 bidi_shelve_cache, and free the buffer used to stash that copy.
683 JUST_FREE non-zero means free the buffer, but don't restore the
684 cache; used when the corresponding iterator is discarded instead of
687 bidi_unshelve_cache (void *databuf
, int just_free
)
689 unsigned char *p
= databuf
;
695 /* A NULL pointer means an empty cache. */
696 bidi_cache_start
= 0;
707 memcpy (&idx
, p
, sizeof (bidi_cache_idx
));
708 bidi_cache_total_alloc
-=
709 sizeof (bidi_cache_idx
) + idx
* sizeof (struct bidi_it
)
710 + sizeof (bidi_cache_start_stack
) + sizeof (bidi_cache_sp
)
711 + sizeof (bidi_cache_start
) + sizeof (bidi_cache_last_idx
);
715 memcpy (&bidi_cache_idx
, p
, sizeof (bidi_cache_idx
));
716 bidi_cache_ensure_space (bidi_cache_idx
);
717 memcpy (bidi_cache
, p
+ sizeof (bidi_cache_idx
),
718 bidi_cache_idx
* sizeof (struct bidi_it
));
719 memcpy (bidi_cache_start_stack
,
720 p
+ sizeof (bidi_cache_idx
)
721 + bidi_cache_idx
* sizeof (struct bidi_it
),
722 sizeof (bidi_cache_start_stack
));
723 memcpy (&bidi_cache_sp
,
724 p
+ sizeof (bidi_cache_idx
)
725 + bidi_cache_idx
* sizeof (struct bidi_it
)
726 + sizeof (bidi_cache_start_stack
),
727 sizeof (bidi_cache_sp
));
728 memcpy (&bidi_cache_start
,
729 p
+ sizeof (bidi_cache_idx
)
730 + bidi_cache_idx
* sizeof (struct bidi_it
)
731 + sizeof (bidi_cache_start_stack
) + sizeof (bidi_cache_sp
),
732 sizeof (bidi_cache_start
));
733 memcpy (&bidi_cache_last_idx
,
734 p
+ sizeof (bidi_cache_idx
)
735 + bidi_cache_idx
* sizeof (struct bidi_it
)
736 + sizeof (bidi_cache_start_stack
) + sizeof (bidi_cache_sp
)
737 + sizeof (bidi_cache_start
),
738 sizeof (bidi_cache_last_idx
));
739 bidi_cache_total_alloc
-=
740 sizeof (bidi_cache_idx
) + bidi_cache_idx
* sizeof (struct bidi_it
)
741 + sizeof (bidi_cache_start_stack
) + sizeof (bidi_cache_sp
)
742 + sizeof (bidi_cache_start
) + sizeof (bidi_cache_last_idx
);
750 /***********************************************************************
752 ***********************************************************************/
754 bidi_initialize (void)
756 bidi_type_table
= uniprop_table (intern ("bidi-class"));
757 if (NILP (bidi_type_table
))
759 staticpro (&bidi_type_table
);
761 bidi_mirror_table
= uniprop_table (intern ("mirroring"));
762 if (NILP (bidi_mirror_table
))
764 staticpro (&bidi_mirror_table
);
766 Qparagraph_start
= intern ("paragraph-start");
767 staticpro (&Qparagraph_start
);
768 paragraph_start_re
= Fsymbol_value (Qparagraph_start
);
769 if (!STRINGP (paragraph_start_re
))
770 paragraph_start_re
= build_string ("\f\\|[ \t]*$");
771 staticpro (¶graph_start_re
);
772 Qparagraph_separate
= intern ("paragraph-separate");
773 staticpro (&Qparagraph_separate
);
774 paragraph_separate_re
= Fsymbol_value (Qparagraph_separate
);
775 if (!STRINGP (paragraph_separate_re
))
776 paragraph_separate_re
= build_string ("[ \t\f]*$");
777 staticpro (¶graph_separate_re
);
780 bidi_cache_total_alloc
= 0;
782 bidi_initialized
= 1;
785 /* Do whatever UAX#9 clause X8 says should be done at paragraph's
788 bidi_set_paragraph_end (struct bidi_it
*bidi_it
)
790 bidi_it
->invalid_levels
= 0;
791 bidi_it
->invalid_rl_levels
= -1;
792 bidi_it
->stack_idx
= 0;
793 bidi_it
->resolved_level
= bidi_it
->level_stack
[0].level
;
796 /* Initialize the bidi iterator from buffer/string position CHARPOS. */
798 bidi_init_it (EMACS_INT charpos
, EMACS_INT bytepos
, int frame_window_p
,
799 struct bidi_it
*bidi_it
)
801 if (! bidi_initialized
)
804 bidi_it
->charpos
= charpos
;
806 bidi_it
->bytepos
= bytepos
;
807 bidi_it
->frame_window_p
= frame_window_p
;
808 bidi_it
->nchars
= -1; /* to be computed in bidi_resolve_explicit_1 */
809 bidi_it
->first_elt
= 1;
810 bidi_set_paragraph_end (bidi_it
);
811 bidi_it
->new_paragraph
= 1;
812 bidi_it
->separator_limit
= -1;
813 bidi_it
->type
= NEUTRAL_B
;
814 bidi_it
->type_after_w1
= NEUTRAL_B
;
815 bidi_it
->orig_type
= NEUTRAL_B
;
816 bidi_it
->prev_was_pdf
= 0;
817 bidi_it
->prev
.type
= bidi_it
->prev
.type_after_w1
=
818 bidi_it
->prev
.orig_type
= UNKNOWN_BT
;
819 bidi_it
->last_strong
.type
= bidi_it
->last_strong
.type_after_w1
=
820 bidi_it
->last_strong
.orig_type
= UNKNOWN_BT
;
821 bidi_it
->next_for_neutral
.charpos
= -1;
822 bidi_it
->next_for_neutral
.type
=
823 bidi_it
->next_for_neutral
.type_after_w1
=
824 bidi_it
->next_for_neutral
.orig_type
= UNKNOWN_BT
;
825 bidi_it
->prev_for_neutral
.charpos
= -1;
826 bidi_it
->prev_for_neutral
.type
=
827 bidi_it
->prev_for_neutral
.type_after_w1
=
828 bidi_it
->prev_for_neutral
.orig_type
= UNKNOWN_BT
;
829 bidi_it
->sor
= L2R
; /* FIXME: should it be user-selectable? */
830 bidi_it
->disp_pos
= -1; /* invalid/unknown */
831 bidi_it
->disp_prop
= 0;
832 /* We can only shrink the cache if we are at the bottom level of its
834 if (bidi_cache_start
== 0)
835 bidi_cache_shrink ();
840 /* Perform initializations for reordering a new line of bidi text. */
842 bidi_line_init (struct bidi_it
*bidi_it
)
844 bidi_it
->scan_dir
= 1; /* FIXME: do we need to have control on this? */
845 bidi_it
->resolved_level
= bidi_it
->level_stack
[0].level
;
846 bidi_it
->level_stack
[0].override
= NEUTRAL_DIR
; /* X1 */
847 bidi_it
->invalid_levels
= 0;
848 bidi_it
->invalid_rl_levels
= -1;
849 bidi_it
->next_en_pos
= -1;
850 bidi_it
->next_for_ws
.type
= UNKNOWN_BT
;
851 bidi_set_sor_type (bidi_it
,
852 bidi_it
->paragraph_dir
== R2L
? 1 : 0,
853 bidi_it
->level_stack
[0].level
); /* X10 */
859 /***********************************************************************
861 ***********************************************************************/
863 /* Count bytes in string S between BEG/BEGBYTE and END. BEG and END
864 are zero-based character positions in S, BEGBYTE is byte position
865 corresponding to BEG. UNIBYTE, if non-zero, means S is a unibyte
867 static inline EMACS_INT
868 bidi_count_bytes (const unsigned char *s
, const EMACS_INT beg
,
869 const EMACS_INT begbyte
, const EMACS_INT end
, int unibyte
)
872 const unsigned char *p
= s
+ begbyte
, *start
= p
;
878 if (!CHAR_HEAD_P (*p
))
883 p
+= BYTES_BY_CHAR_HEAD (*p
);
891 /* Fetch and returns the character at byte position BYTEPOS. If S is
892 non-NULL, fetch the character from string S; otherwise fetch the
893 character from the current buffer. UNIBYTE non-zero means S is a
896 bidi_char_at_pos (EMACS_INT bytepos
, const unsigned char *s
, int unibyte
)
903 return STRING_CHAR (s
+ bytepos
);
906 return FETCH_MULTIBYTE_CHAR (bytepos
);
909 /* Fetch and return the character at BYTEPOS/CHARPOS. If that
910 character is covered by a display string, treat the entire run of
911 covered characters as a single character, either u+2029 or u+FFFC,
912 and return their combined length in CH_LEN and NCHARS. DISP_POS
913 specifies the character position of the next display string, or -1
914 if not yet computed. DISP_PROP non-zero means that there's really
915 a display string at DISP_POS, as opposed to when we searched till
916 DISP_POS without finding one. If DISP_PROP is 2, it means the
917 display spec is of the form `(space ...)', which is replaced with
918 u+2029 to handle it as a paragraph separator. When the next
919 character is at or beyond that position, the function updates
920 DISP_POS with the position of the next display string. STRING->s
921 is the C string to iterate, or NULL if iterating over a buffer or a
922 Lisp string; in the latter case, STRING->lstring is the Lisp
925 bidi_fetch_char (EMACS_INT bytepos
, EMACS_INT charpos
, EMACS_INT
*disp_pos
,
926 int *disp_prop
, struct bidi_string_data
*string
,
927 int frame_window_p
, EMACS_INT
*ch_len
, EMACS_INT
*nchars
)
931 (string
->s
|| STRINGP (string
->lstring
)) ? string
->schars
: ZV
;
934 /* If we got past the last known position of display string, compute
935 the position of the next one. That position could be at CHARPOS. */
936 if (charpos
< endpos
&& charpos
> *disp_pos
)
938 SET_TEXT_POS (pos
, charpos
, bytepos
);
939 *disp_pos
= compute_display_string_pos (&pos
, string
, frame_window_p
,
943 /* Fetch the character at BYTEPOS. */
944 if (charpos
>= endpos
)
952 else if (charpos
>= *disp_pos
&& *disp_prop
)
954 EMACS_INT disp_end_pos
;
956 /* We don't expect to find ourselves in the middle of a display
957 property. Hopefully, it will never be needed. */
958 if (charpos
> *disp_pos
)
960 /* Text covered by `display' properties and overlays with
961 display properties or display strings is handled as a single
962 character that represents the entire run of characters
963 covered by the display property. */
966 /* `(space ...)' display specs are handled as paragraph
967 separators for the purposes of the reordering; see UAX#9
968 section 3 and clause HL1 in section 4.3 there. */
973 /* All other display specs are handled as the Unicode Object
974 Replacement Character. */
977 disp_end_pos
= compute_display_string_end (*disp_pos
, string
);
978 *nchars
= disp_end_pos
- *disp_pos
;
982 *ch_len
= bidi_count_bytes (string
->s
, *disp_pos
, bytepos
,
983 disp_end_pos
, string
->unibyte
);
984 else if (STRINGP (string
->lstring
))
985 *ch_len
= bidi_count_bytes (SDATA (string
->lstring
), *disp_pos
,
986 bytepos
, disp_end_pos
, string
->unibyte
);
988 *ch_len
= CHAR_TO_BYTE (disp_end_pos
) - bytepos
;
996 if (!string
->unibyte
)
998 ch
= STRING_CHAR_AND_LENGTH (string
->s
+ bytepos
, len
);
1003 ch
= UNIBYTE_TO_CHAR (string
->s
[bytepos
]);
1007 else if (STRINGP (string
->lstring
))
1011 if (!string
->unibyte
)
1013 ch
= STRING_CHAR_AND_LENGTH (SDATA (string
->lstring
) + bytepos
,
1019 ch
= UNIBYTE_TO_CHAR (SREF (string
->lstring
, bytepos
));
1025 ch
= FETCH_MULTIBYTE_CHAR (bytepos
);
1026 *ch_len
= CHAR_BYTES (ch
);
1031 /* If we just entered a run of characters covered by a display
1032 string, compute the position of the next display string. */
1033 if (charpos
+ *nchars
<= endpos
&& charpos
+ *nchars
> *disp_pos
1036 SET_TEXT_POS (pos
, charpos
+ *nchars
, bytepos
+ *ch_len
);
1037 *disp_pos
= compute_display_string_pos (&pos
, string
, frame_window_p
,
1045 /***********************************************************************
1046 Determining paragraph direction
1047 ***********************************************************************/
1049 /* Check if buffer position CHARPOS/BYTEPOS is the end of a paragraph.
1050 Value is the non-negative length of the paragraph separator
1051 following the buffer position, -1 if position is at the beginning
1052 of a new paragraph, or -2 if position is neither at beginning nor
1053 at end of a paragraph. */
1055 bidi_at_paragraph_end (EMACS_INT charpos
, EMACS_INT bytepos
)
1058 Lisp_Object start_re
;
1061 sep_re
= paragraph_separate_re
;
1062 start_re
= paragraph_start_re
;
1064 val
= fast_looking_at (sep_re
, charpos
, bytepos
, ZV
, ZV_BYTE
, Qnil
);
1067 if (fast_looking_at (start_re
, charpos
, bytepos
, ZV
, ZV_BYTE
, Qnil
) >= 0)
1076 /* Find the beginning of this paragraph by looking back in the buffer.
1077 Value is the byte position of the paragraph's beginning. */
1079 bidi_find_paragraph_start (EMACS_INT pos
, EMACS_INT pos_byte
)
1081 Lisp_Object re
= paragraph_start_re
;
1082 EMACS_INT limit
= ZV
, limit_byte
= ZV_BYTE
;
1084 while (pos_byte
> BEGV_BYTE
1085 && fast_looking_at (re
, pos
, pos_byte
, limit
, limit_byte
, Qnil
) < 0)
1087 /* FIXME: What if the paragraph beginning is covered by a
1088 display string? And what if a display string covering some
1089 of the text over which we scan back includes
1090 paragraph_start_re? */
1091 pos
= find_next_newline_no_quit (pos
- 1, -1);
1092 pos_byte
= CHAR_TO_BYTE (pos
);
1097 /* Determine the base direction, a.k.a. base embedding level, of the
1098 paragraph we are about to iterate through. If DIR is either L2R or
1099 R2L, just use that. Otherwise, determine the paragraph direction
1100 from the first strong directional character of the paragraph.
1102 NO_DEFAULT_P non-zero means don't default to L2R if the paragraph
1103 has no strong directional characters and both DIR and
1104 bidi_it->paragraph_dir are NEUTRAL_DIR. In that case, search back
1105 in the buffer until a paragraph is found with a strong character,
1106 or until hitting BEGV. In the latter case, fall back to L2R. This
1107 flag is used in current-bidi-paragraph-direction.
1109 Note that this function gives the paragraph separator the same
1110 direction as the preceding paragraph, even though Emacs generally
1111 views the separartor as not belonging to any paragraph. */
1113 bidi_paragraph_init (bidi_dir_t dir
, struct bidi_it
*bidi_it
, int no_default_p
)
1115 EMACS_INT bytepos
= bidi_it
->bytepos
;
1116 int string_p
= bidi_it
->string
.s
!= NULL
|| STRINGP (bidi_it
->string
.lstring
);
1117 EMACS_INT pstartbyte
;
1118 /* Note that begbyte is a byte position, while end is a character
1119 position. Yes, this is ugly, but we are trying to avoid costly
1120 calls to BYTE_TO_CHAR and its ilk. */
1121 EMACS_INT begbyte
= string_p
? 0 : BEGV_BYTE
;
1122 EMACS_INT end
= string_p
? bidi_it
->string
.schars
: ZV
;
1124 /* Special case for an empty buffer. */
1125 if (bytepos
== begbyte
&& bidi_it
->charpos
== end
)
1127 /* We should never be called at EOB or before BEGV. */
1128 else if (bidi_it
->charpos
>= end
|| bytepos
< begbyte
)
1133 bidi_it
->paragraph_dir
= L2R
;
1134 bidi_it
->new_paragraph
= 0;
1136 else if (dir
== R2L
)
1138 bidi_it
->paragraph_dir
= R2L
;
1139 bidi_it
->new_paragraph
= 0;
1141 else if (dir
== NEUTRAL_DIR
) /* P2 */
1144 EMACS_INT ch_len
, nchars
;
1145 EMACS_INT pos
, disp_pos
= -1;
1148 const unsigned char *s
;
1150 if (!bidi_initialized
)
1153 /* If we are inside a paragraph separator, we are just waiting
1154 for the separator to be exhausted; use the previous paragraph
1155 direction. But don't do that if we have been just reseated,
1156 because we need to reinitialize below in that case. */
1157 if (!bidi_it
->first_elt
1158 && bidi_it
->charpos
< bidi_it
->separator_limit
)
1161 /* If we are on a newline, get past it to where the next
1162 paragraph might start. But don't do that at BEGV since then
1163 we are potentially in a new paragraph that doesn't yet
1165 pos
= bidi_it
->charpos
;
1166 s
= STRINGP (bidi_it
->string
.lstring
) ?
1167 SDATA (bidi_it
->string
.lstring
) : bidi_it
->string
.s
;
1168 if (bytepos
> begbyte
1169 && bidi_char_at_pos (bytepos
, s
, bidi_it
->string
.unibyte
) == '\n')
1175 /* We are either at the beginning of a paragraph or in the
1176 middle of it. Find where this paragraph starts. */
1179 /* We don't support changes of paragraph direction inside a
1180 string. It is treated as a single paragraph. */
1184 pstartbyte
= bidi_find_paragraph_start (pos
, bytepos
);
1185 bidi_it
->separator_limit
= -1;
1186 bidi_it
->new_paragraph
= 0;
1188 /* The following loop is run more than once only if NO_DEFAULT_P
1189 is non-zero, and only if we are iterating on a buffer. */
1191 bytepos
= pstartbyte
;
1193 pos
= BYTE_TO_CHAR (bytepos
);
1194 ch
= bidi_fetch_char (bytepos
, pos
, &disp_pos
, &disp_prop
,
1196 bidi_it
->frame_window_p
, &ch_len
, &nchars
);
1197 type
= bidi_get_type (ch
, NEUTRAL_DIR
);
1199 for (pos
+= nchars
, bytepos
+= ch_len
;
1200 (bidi_get_category (type
) != STRONG
)
1201 || (bidi_ignore_explicit_marks_for_paragraph_level
1202 && (type
== RLE
|| type
== RLO
1203 || type
== LRE
|| type
== LRO
));
1204 type
= bidi_get_type (ch
, NEUTRAL_DIR
))
1208 /* Pretend there's a paragraph separator at end of
1214 && type
== NEUTRAL_B
1215 && bidi_at_paragraph_end (pos
, bytepos
) >= -1)
1217 /* Fetch next character and advance to get past it. */
1218 ch
= bidi_fetch_char (bytepos
, pos
, &disp_pos
,
1219 &disp_prop
, &bidi_it
->string
,
1220 bidi_it
->frame_window_p
, &ch_len
, &nchars
);
1224 if ((type
== STRONG_R
|| type
== STRONG_AL
) /* P3 */
1225 || (!bidi_ignore_explicit_marks_for_paragraph_level
1226 && (type
== RLO
|| type
== RLE
)))
1227 bidi_it
->paragraph_dir
= R2L
;
1228 else if (type
== STRONG_L
1229 || (!bidi_ignore_explicit_marks_for_paragraph_level
1230 && (type
== LRO
|| type
== LRE
)))
1231 bidi_it
->paragraph_dir
= L2R
;
1233 && no_default_p
&& bidi_it
->paragraph_dir
== NEUTRAL_DIR
)
1235 /* If this paragraph is at BEGV, default to L2R. */
1236 if (pstartbyte
== BEGV_BYTE
)
1237 bidi_it
->paragraph_dir
= L2R
; /* P3 and HL1 */
1240 EMACS_INT prevpbyte
= pstartbyte
;
1241 EMACS_INT p
= BYTE_TO_CHAR (pstartbyte
), pbyte
= pstartbyte
;
1243 /* Find the beginning of the previous paragraph, if any. */
1244 while (pbyte
> BEGV_BYTE
&& prevpbyte
>= pstartbyte
)
1246 /* FXIME: What if p is covered by a display
1247 string? See also a FIXME inside
1248 bidi_find_paragraph_start. */
1250 pbyte
= CHAR_TO_BYTE (p
);
1251 prevpbyte
= bidi_find_paragraph_start (p
, pbyte
);
1253 pstartbyte
= prevpbyte
;
1257 && no_default_p
&& bidi_it
->paragraph_dir
== NEUTRAL_DIR
);
1262 /* Contrary to UAX#9 clause P3, we only default the paragraph
1263 direction to L2R if we have no previous usable paragraph
1264 direction. This is allowed by the HL1 clause. */
1265 if (bidi_it
->paragraph_dir
!= L2R
&& bidi_it
->paragraph_dir
!= R2L
)
1266 bidi_it
->paragraph_dir
= L2R
; /* P3 and HL1 ``higher-level protocols'' */
1267 if (bidi_it
->paragraph_dir
== R2L
)
1268 bidi_it
->level_stack
[0].level
= 1;
1270 bidi_it
->level_stack
[0].level
= 0;
1272 bidi_line_init (bidi_it
);
1276 /***********************************************************************
1277 Resolving explicit and implicit levels.
1278 The rest of this file constitutes the core of the UBA implementation.
1279 ***********************************************************************/
1282 bidi_explicit_dir_char (int ch
)
1284 bidi_type_t ch_type
;
1286 if (!bidi_initialized
)
1288 ch_type
= (bidi_type_t
) XINT (CHAR_TABLE_REF (bidi_type_table
, ch
));
1289 return (ch_type
== LRE
|| ch_type
== LRO
1290 || ch_type
== RLE
|| ch_type
== RLO
1294 /* A helper function for bidi_resolve_explicit. It advances to the
1295 next character in logical order and determines the new embedding
1296 level and directional override, but does not take into account
1297 empty embeddings. */
1299 bidi_resolve_explicit_1 (struct bidi_it
*bidi_it
)
1305 bidi_dir_t override
;
1306 int string_p
= bidi_it
->string
.s
!= NULL
|| STRINGP (bidi_it
->string
.lstring
);
1308 /* If reseat()'ed, don't advance, so as to start iteration from the
1309 position where we were reseated. bidi_it->bytepos can be less
1310 than BEGV_BYTE after reseat to BEGV. */
1311 if (bidi_it
->bytepos
< (string_p
? 0 : BEGV_BYTE
)
1312 || bidi_it
->first_elt
)
1314 bidi_it
->first_elt
= 0;
1317 const unsigned char *p
=
1318 STRINGP (bidi_it
->string
.lstring
)
1319 ? SDATA (bidi_it
->string
.lstring
) : bidi_it
->string
.s
;
1321 if (bidi_it
->charpos
< 0)
1322 bidi_it
->charpos
= 0;
1323 bidi_it
->bytepos
= bidi_count_bytes (p
, 0, 0, bidi_it
->charpos
,
1324 bidi_it
->string
.unibyte
);
1328 if (bidi_it
->charpos
< BEGV
)
1329 bidi_it
->charpos
= BEGV
;
1330 bidi_it
->bytepos
= CHAR_TO_BYTE (bidi_it
->charpos
);
1333 /* Don't move at end of buffer/string. */
1334 else if (bidi_it
->charpos
< (string_p
? bidi_it
->string
.schars
: ZV
))
1336 /* Advance to the next character, skipping characters covered by
1337 display strings (nchars > 1). */
1338 if (bidi_it
->nchars
<= 0)
1340 bidi_it
->charpos
+= bidi_it
->nchars
;
1341 if (bidi_it
->ch_len
== 0)
1343 bidi_it
->bytepos
+= bidi_it
->ch_len
;
1346 current_level
= bidi_it
->level_stack
[bidi_it
->stack_idx
].level
; /* X1 */
1347 override
= bidi_it
->level_stack
[bidi_it
->stack_idx
].override
;
1348 new_level
= current_level
;
1350 if (bidi_it
->charpos
>= (string_p
? bidi_it
->string
.schars
: ZV
))
1353 bidi_it
->ch_len
= 1;
1354 bidi_it
->nchars
= 1;
1355 bidi_it
->disp_pos
= (string_p
? bidi_it
->string
.schars
: ZV
);
1356 bidi_it
->disp_prop
= 0;
1360 /* Fetch the character at BYTEPOS. If it is covered by a
1361 display string, treat the entire run of covered characters as
1362 a single character u+FFFC. */
1363 curchar
= bidi_fetch_char (bidi_it
->bytepos
, bidi_it
->charpos
,
1364 &bidi_it
->disp_pos
, &bidi_it
->disp_prop
,
1365 &bidi_it
->string
, bidi_it
->frame_window_p
,
1366 &bidi_it
->ch_len
, &bidi_it
->nchars
);
1368 bidi_it
->ch
= curchar
;
1370 /* Don't apply directional override here, as all the types we handle
1371 below will not be affected by the override anyway, and we need
1372 the original type unaltered. The override will be applied in
1373 bidi_resolve_weak. */
1374 type
= bidi_get_type (curchar
, NEUTRAL_DIR
);
1375 bidi_it
->orig_type
= type
;
1376 bidi_check_type (bidi_it
->orig_type
);
1379 bidi_it
->prev_was_pdf
= 0;
1381 bidi_it
->type_after_w1
= UNKNOWN_BT
;
1387 bidi_it
->type_after_w1
= type
;
1388 bidi_check_type (bidi_it
->type_after_w1
);
1389 type
= WEAK_BN
; /* X9/Retaining */
1390 if (bidi_it
->ignore_bn_limit
<= -1)
1392 if (current_level
<= BIDI_MAXLEVEL
- 4)
1394 /* Compute the least odd embedding level greater than
1395 the current level. */
1396 new_level
= ((current_level
+ 1) & ~1) + 1;
1397 if (bidi_it
->type_after_w1
== RLE
)
1398 override
= NEUTRAL_DIR
;
1401 if (current_level
== BIDI_MAXLEVEL
- 4)
1402 bidi_it
->invalid_rl_levels
= 0;
1403 bidi_push_embedding_level (bidi_it
, new_level
, override
);
1407 bidi_it
->invalid_levels
++;
1408 /* See the commentary about invalid_rl_levels below. */
1409 if (bidi_it
->invalid_rl_levels
< 0)
1410 bidi_it
->invalid_rl_levels
= 0;
1411 bidi_it
->invalid_rl_levels
++;
1414 else if (bidi_it
->prev
.type_after_w1
== WEAK_EN
/* W5/Retaining */
1415 || bidi_it
->next_en_pos
> bidi_it
->charpos
)
1420 bidi_it
->type_after_w1
= type
;
1421 bidi_check_type (bidi_it
->type_after_w1
);
1422 type
= WEAK_BN
; /* X9/Retaining */
1423 if (bidi_it
->ignore_bn_limit
<= -1)
1425 if (current_level
<= BIDI_MAXLEVEL
- 5)
1427 /* Compute the least even embedding level greater than
1428 the current level. */
1429 new_level
= ((current_level
+ 2) & ~1);
1430 if (bidi_it
->type_after_w1
== LRE
)
1431 override
= NEUTRAL_DIR
;
1434 bidi_push_embedding_level (bidi_it
, new_level
, override
);
1438 bidi_it
->invalid_levels
++;
1439 /* invalid_rl_levels counts invalid levels encountered
1440 while the embedding level was already too high for
1441 LRE/LRO, but not for RLE/RLO. That is because
1442 there may be exactly one PDF which we should not
1443 ignore even though invalid_levels is non-zero.
1444 invalid_rl_levels helps to know what PDF is
1446 if (bidi_it
->invalid_rl_levels
>= 0)
1447 bidi_it
->invalid_rl_levels
++;
1450 else if (bidi_it
->prev
.type_after_w1
== WEAK_EN
/* W5/Retaining */
1451 || bidi_it
->next_en_pos
> bidi_it
->charpos
)
1455 bidi_it
->type_after_w1
= type
;
1456 bidi_check_type (bidi_it
->type_after_w1
);
1457 type
= WEAK_BN
; /* X9/Retaining */
1458 if (bidi_it
->ignore_bn_limit
<= -1)
1460 if (!bidi_it
->invalid_rl_levels
)
1462 new_level
= bidi_pop_embedding_level (bidi_it
);
1463 bidi_it
->invalid_rl_levels
= -1;
1464 if (bidi_it
->invalid_levels
)
1465 bidi_it
->invalid_levels
--;
1466 /* else nothing: UAX#9 says to ignore invalid PDFs */
1468 if (!bidi_it
->invalid_levels
)
1469 new_level
= bidi_pop_embedding_level (bidi_it
);
1472 bidi_it
->invalid_levels
--;
1473 bidi_it
->invalid_rl_levels
--;
1476 else if (bidi_it
->prev
.type_after_w1
== WEAK_EN
/* W5/Retaining */
1477 || bidi_it
->next_en_pos
> bidi_it
->charpos
)
1485 bidi_it
->type
= type
;
1486 bidi_check_type (bidi_it
->type
);
1491 /* Given an iterator state in BIDI_IT, advance one character position
1492 in the buffer/string to the next character (in the logical order),
1493 resolve any explicit embeddings and directional overrides, and
1494 return the embedding level of the character after resolving
1495 explicit directives and ignoring empty embeddings. */
1497 bidi_resolve_explicit (struct bidi_it
*bidi_it
)
1499 int prev_level
= bidi_it
->level_stack
[bidi_it
->stack_idx
].level
;
1500 int new_level
= bidi_resolve_explicit_1 (bidi_it
);
1501 EMACS_INT eob
= bidi_it
->string
.s
? bidi_it
->string
.schars
: ZV
;
1502 const unsigned char *s
= STRINGP (bidi_it
->string
.lstring
)
1503 ? SDATA (bidi_it
->string
.lstring
) : bidi_it
->string
.s
;
1505 if (prev_level
< new_level
1506 && bidi_it
->type
== WEAK_BN
1507 && bidi_it
->ignore_bn_limit
== -1 /* only if not already known */
1508 && bidi_it
->charpos
< eob
/* not already at EOB */
1509 && bidi_explicit_dir_char (bidi_char_at_pos (bidi_it
->bytepos
1510 + bidi_it
->ch_len
, s
,
1511 bidi_it
->string
.unibyte
)))
1513 /* Avoid pushing and popping embedding levels if the level run
1514 is empty, as this breaks level runs where it shouldn't.
1515 UAX#9 removes all the explicit embedding and override codes,
1516 so empty embeddings disappear without a trace. We need to
1517 behave as if we did the same. */
1518 struct bidi_it saved_it
;
1519 int level
= prev_level
;
1521 bidi_copy_it (&saved_it
, bidi_it
);
1523 while (bidi_explicit_dir_char (bidi_char_at_pos (bidi_it
->bytepos
1524 + bidi_it
->ch_len
, s
,
1525 bidi_it
->string
.unibyte
)))
1527 /* This advances to the next character, skipping any
1528 characters covered by display strings. */
1529 level
= bidi_resolve_explicit_1 (bidi_it
);
1530 /* If string.lstring was relocated inside bidi_resolve_explicit_1,
1531 a pointer to its data is no longer valid. */
1532 if (STRINGP (bidi_it
->string
.lstring
))
1533 s
= SDATA (bidi_it
->string
.lstring
);
1536 if (bidi_it
->nchars
<= 0)
1538 if (level
== prev_level
) /* empty embedding */
1539 saved_it
.ignore_bn_limit
= bidi_it
->charpos
+ bidi_it
->nchars
;
1540 else /* this embedding is non-empty */
1541 saved_it
.ignore_bn_limit
= -2;
1543 bidi_copy_it (bidi_it
, &saved_it
);
1544 if (bidi_it
->ignore_bn_limit
> -1)
1546 /* We pushed a level, but we shouldn't have. Undo that. */
1547 if (!bidi_it
->invalid_rl_levels
)
1549 new_level
= bidi_pop_embedding_level (bidi_it
);
1550 bidi_it
->invalid_rl_levels
= -1;
1551 if (bidi_it
->invalid_levels
)
1552 bidi_it
->invalid_levels
--;
1554 if (!bidi_it
->invalid_levels
)
1555 new_level
= bidi_pop_embedding_level (bidi_it
);
1558 bidi_it
->invalid_levels
--;
1559 bidi_it
->invalid_rl_levels
--;
1564 if (bidi_it
->type
== NEUTRAL_B
) /* X8 */
1566 bidi_set_paragraph_end (bidi_it
);
1567 /* This is needed by bidi_resolve_weak below, and in L1. */
1568 bidi_it
->type_after_w1
= bidi_it
->type
;
1569 bidi_check_type (bidi_it
->type_after_w1
);
1575 /* Advance in the buffer/string, resolve weak types and return the
1576 type of the next character after weak type resolution. */
1578 bidi_resolve_weak (struct bidi_it
*bidi_it
)
1581 bidi_dir_t override
;
1582 int prev_level
= bidi_it
->level_stack
[bidi_it
->stack_idx
].level
;
1583 int new_level
= bidi_resolve_explicit (bidi_it
);
1585 bidi_type_t type_of_next
;
1586 struct bidi_it saved_it
;
1588 (STRINGP (bidi_it
->string
.lstring
) || bidi_it
->string
.s
)
1589 ? bidi_it
->string
.schars
: ZV
;
1591 type
= bidi_it
->type
;
1592 override
= bidi_it
->level_stack
[bidi_it
->stack_idx
].override
;
1594 if (type
== UNKNOWN_BT
1602 if (new_level
!= prev_level
1603 || bidi_it
->type
== NEUTRAL_B
)
1605 /* We've got a new embedding level run, compute the directional
1606 type of sor and initialize per-run variables (UAX#9, clause
1608 bidi_set_sor_type (bidi_it
, prev_level
, new_level
);
1610 else if (type
== NEUTRAL_S
|| type
== NEUTRAL_WS
1611 || type
== WEAK_BN
|| type
== STRONG_AL
)
1612 bidi_it
->type_after_w1
= type
; /* needed in L1 */
1613 bidi_check_type (bidi_it
->type_after_w1
);
1615 /* Level and directional override status are already recorded in
1616 bidi_it, and do not need any change; see X6. */
1617 if (override
== R2L
) /* X6 */
1619 else if (override
== L2R
)
1623 if (type
== WEAK_NSM
) /* W1 */
1625 /* Note that we don't need to consider the case where the
1626 prev character has its type overridden by an RLO or LRO,
1627 because then either the type of this NSM would have been
1628 also overridden, or the previous character is outside the
1629 current level run, and thus not relevant to this NSM.
1630 This is why NSM gets the type_after_w1 of the previous
1632 if (bidi_it
->prev
.type_after_w1
!= UNKNOWN_BT
1633 /* if type_after_w1 is NEUTRAL_B, this NSM is at sor */
1634 && bidi_it
->prev
.type_after_w1
!= NEUTRAL_B
)
1635 type
= bidi_it
->prev
.type_after_w1
;
1636 else if (bidi_it
->sor
== R2L
)
1638 else if (bidi_it
->sor
== L2R
)
1640 else /* shouldn't happen! */
1643 if (type
== WEAK_EN
/* W2 */
1644 && bidi_it
->last_strong
.type_after_w1
== STRONG_AL
)
1646 else if (type
== STRONG_AL
) /* W3 */
1648 else if ((type
== WEAK_ES
/* W4 */
1649 && bidi_it
->prev
.type_after_w1
== WEAK_EN
1650 && bidi_it
->prev
.orig_type
== WEAK_EN
)
1652 && ((bidi_it
->prev
.type_after_w1
== WEAK_EN
1653 && bidi_it
->prev
.orig_type
== WEAK_EN
)
1654 || bidi_it
->prev
.type_after_w1
== WEAK_AN
)))
1656 const unsigned char *s
=
1657 STRINGP (bidi_it
->string
.lstring
)
1658 ? SDATA (bidi_it
->string
.lstring
) : bidi_it
->string
.s
;
1661 bidi_it
->charpos
+ bidi_it
->nchars
>= eob
1663 : bidi_char_at_pos (bidi_it
->bytepos
+ bidi_it
->ch_len
, s
,
1664 bidi_it
->string
.unibyte
);
1665 type_of_next
= bidi_get_type (next_char
, override
);
1667 if (type_of_next
== WEAK_BN
1668 || bidi_explicit_dir_char (next_char
))
1670 bidi_copy_it (&saved_it
, bidi_it
);
1671 while (bidi_resolve_explicit (bidi_it
) == new_level
1672 && bidi_it
->type
== WEAK_BN
)
1674 type_of_next
= bidi_it
->type
;
1675 bidi_copy_it (bidi_it
, &saved_it
);
1678 /* If the next character is EN, but the last strong-type
1679 character is AL, that next EN will be changed to AN when
1680 we process it in W2 above. So in that case, this ES
1681 should not be changed into EN. */
1683 && type_of_next
== WEAK_EN
1684 && bidi_it
->last_strong
.type_after_w1
!= STRONG_AL
)
1686 else if (type
== WEAK_CS
)
1688 if (bidi_it
->prev
.type_after_w1
== WEAK_AN
1689 && (type_of_next
== WEAK_AN
1690 /* If the next character is EN, but the last
1691 strong-type character is AL, EN will be later
1692 changed to AN when we process it in W2 above.
1693 So in that case, this ES should not be
1695 || (type_of_next
== WEAK_EN
1696 && bidi_it
->last_strong
.type_after_w1
== STRONG_AL
)))
1698 else if (bidi_it
->prev
.type_after_w1
== WEAK_EN
1699 && type_of_next
== WEAK_EN
1700 && bidi_it
->last_strong
.type_after_w1
!= STRONG_AL
)
1704 else if (type
== WEAK_ET
/* W5: ET with EN before or after it */
1705 || type
== WEAK_BN
) /* W5/Retaining */
1707 if (bidi_it
->prev
.type_after_w1
== WEAK_EN
/* ET/BN w/EN before it */
1708 || bidi_it
->next_en_pos
> bidi_it
->charpos
)
1710 else /* W5: ET/BN with EN after it. */
1712 EMACS_INT en_pos
= bidi_it
->charpos
+ bidi_it
->nchars
;
1713 const unsigned char *s
=
1714 STRINGP (bidi_it
->string
.lstring
)
1715 ? SDATA (bidi_it
->string
.lstring
) : bidi_it
->string
.s
;
1717 if (bidi_it
->nchars
<= 0)
1720 bidi_it
->charpos
+ bidi_it
->nchars
>= eob
1722 : bidi_char_at_pos (bidi_it
->bytepos
+ bidi_it
->ch_len
, s
,
1723 bidi_it
->string
.unibyte
);
1724 type_of_next
= bidi_get_type (next_char
, override
);
1726 if (type_of_next
== WEAK_ET
1727 || type_of_next
== WEAK_BN
1728 || bidi_explicit_dir_char (next_char
))
1730 bidi_copy_it (&saved_it
, bidi_it
);
1731 while (bidi_resolve_explicit (bidi_it
) == new_level
1732 && (bidi_it
->type
== WEAK_BN
1733 || bidi_it
->type
== WEAK_ET
))
1735 type_of_next
= bidi_it
->type
;
1736 en_pos
= bidi_it
->charpos
;
1737 bidi_copy_it (bidi_it
, &saved_it
);
1739 if (type_of_next
== WEAK_EN
)
1741 /* If the last strong character is AL, the EN we've
1742 found will become AN when we get to it (W2). */
1743 if (bidi_it
->last_strong
.type_after_w1
!= STRONG_AL
)
1746 /* Remember this EN position, to speed up processing
1748 bidi_it
->next_en_pos
= en_pos
;
1750 else if (type
== WEAK_BN
)
1751 type
= NEUTRAL_ON
; /* W6/Retaining */
1757 if (type
== WEAK_ES
|| type
== WEAK_ET
|| type
== WEAK_CS
/* W6 */
1759 && (bidi_it
->prev
.type_after_w1
== WEAK_CS
/* W6/Retaining */
1760 || bidi_it
->prev
.type_after_w1
== WEAK_ES
1761 || bidi_it
->prev
.type_after_w1
== WEAK_ET
)))
1764 /* Store the type we've got so far, before we clobber it with strong
1765 types in W7 and while resolving neutral types. But leave alone
1766 the original types that were recorded above, because we will need
1767 them for the L1 clause. */
1768 if (bidi_it
->type_after_w1
== UNKNOWN_BT
)
1769 bidi_it
->type_after_w1
= type
;
1770 bidi_check_type (bidi_it
->type_after_w1
);
1772 if (type
== WEAK_EN
) /* W7 */
1774 if ((bidi_it
->last_strong
.type_after_w1
== STRONG_L
)
1775 || (bidi_it
->last_strong
.type
== UNKNOWN_BT
&& bidi_it
->sor
== L2R
))
1779 bidi_it
->type
= type
;
1780 bidi_check_type (bidi_it
->type
);
1784 /* Resolve the type of a neutral character according to the type of
1785 surrounding strong text and the current embedding level. */
1786 static inline bidi_type_t
1787 bidi_resolve_neutral_1 (bidi_type_t prev_type
, bidi_type_t next_type
, int lev
)
1789 /* N1: European and Arabic numbers are treated as though they were R. */
1790 if (next_type
== WEAK_EN
|| next_type
== WEAK_AN
)
1791 next_type
= STRONG_R
;
1792 if (prev_type
== WEAK_EN
|| prev_type
== WEAK_AN
)
1793 prev_type
= STRONG_R
;
1795 if (next_type
== prev_type
) /* N1 */
1797 else if ((lev
& 1) == 0) /* N2 */
1804 bidi_resolve_neutral (struct bidi_it
*bidi_it
)
1806 int prev_level
= bidi_it
->level_stack
[bidi_it
->stack_idx
].level
;
1807 bidi_type_t type
= bidi_resolve_weak (bidi_it
);
1808 int current_level
= bidi_it
->level_stack
[bidi_it
->stack_idx
].level
;
1810 if (!(type
== STRONG_R
1815 || type
== NEUTRAL_B
1816 || type
== NEUTRAL_S
1817 || type
== NEUTRAL_WS
1818 || type
== NEUTRAL_ON
))
1821 if (bidi_get_category (type
) == NEUTRAL
1822 || (type
== WEAK_BN
&& prev_level
== current_level
))
1824 if (bidi_it
->next_for_neutral
.type
!= UNKNOWN_BT
)
1825 type
= bidi_resolve_neutral_1 (bidi_it
->prev_for_neutral
.type
,
1826 bidi_it
->next_for_neutral
.type
,
1830 /* Arrrgh!! The UAX#9 algorithm is too deeply entrenched in
1831 the assumption of batch-style processing; see clauses W4,
1832 W5, and especially N1, which require to look far forward
1833 (as well as back) in the buffer/string. May the fleas of
1834 a thousand camels infest the armpits of those who design
1835 supposedly general-purpose algorithms by looking at their
1836 own implementations, and fail to consider other possible
1838 struct bidi_it saved_it
;
1839 bidi_type_t next_type
;
1841 if (bidi_it
->scan_dir
== -1)
1844 bidi_copy_it (&saved_it
, bidi_it
);
1845 /* Scan the text forward until we find the first non-neutral
1846 character, and then use that to resolve the neutral we
1847 are dealing with now. We also cache the scanned iterator
1848 states, to salvage some of the effort later. */
1849 bidi_cache_iterator_state (bidi_it
, 0);
1851 /* Record the info about the previous character, so that
1852 it will be cached below with this state. */
1853 if (bidi_it
->type_after_w1
!= WEAK_BN
/* W1/Retaining */
1854 && bidi_it
->type
!= WEAK_BN
)
1855 bidi_remember_char (&bidi_it
->prev
, bidi_it
);
1856 type
= bidi_resolve_weak (bidi_it
);
1857 /* Paragraph separators have their levels fully resolved
1858 at this point, so cache them as resolved. */
1859 bidi_cache_iterator_state (bidi_it
, type
== NEUTRAL_B
);
1860 /* FIXME: implement L1 here, by testing for a newline and
1861 resetting the level for any sequence of whitespace
1862 characters adjacent to it. */
1863 } while (!(type
== NEUTRAL_B
1865 && bidi_get_category (type
) != NEUTRAL
)
1866 /* This is all per level run, so stop when we
1867 reach the end of this level run. */
1868 || bidi_it
->level_stack
[bidi_it
->stack_idx
].level
!=
1871 bidi_remember_char (&saved_it
.next_for_neutral
, bidi_it
);
1882 /* N1: ``European and Arabic numbers are treated as
1883 though they were R.'' */
1884 next_type
= STRONG_R
;
1885 saved_it
.next_for_neutral
.type
= STRONG_R
;
1888 if (!bidi_explicit_dir_char (bidi_it
->ch
))
1889 abort (); /* can't happen: BNs are skipped */
1892 /* Marched all the way to the end of this level run.
1893 We need to use the eor type, whose information is
1894 stored by bidi_set_sor_type in the prev_for_neutral
1896 if (saved_it
.type
!= WEAK_BN
1897 || bidi_get_category (bidi_it
->prev
.type_after_w1
) == NEUTRAL
)
1899 next_type
= bidi_it
->prev_for_neutral
.type
;
1900 saved_it
.next_for_neutral
.type
= next_type
;
1901 bidi_check_type (next_type
);
1905 /* This is a BN which does not adjoin neutrals.
1906 Leave its type alone. */
1907 bidi_copy_it (bidi_it
, &saved_it
);
1908 return bidi_it
->type
;
1914 type
= bidi_resolve_neutral_1 (saved_it
.prev_for_neutral
.type
,
1915 next_type
, current_level
);
1916 saved_it
.type
= type
;
1917 bidi_check_type (type
);
1918 bidi_copy_it (bidi_it
, &saved_it
);
1924 /* Given an iterator state in BIDI_IT, advance one character position
1925 in the buffer/string to the next character (in the logical order),
1926 resolve the bidi type of that next character, and return that
1929 bidi_type_of_next_char (struct bidi_it
*bidi_it
)
1933 /* This should always be called during a forward scan. */
1934 if (bidi_it
->scan_dir
!= 1)
1937 /* Reset the limit until which to ignore BNs if we step out of the
1938 area where we found only empty levels. */
1939 if ((bidi_it
->ignore_bn_limit
> -1
1940 && bidi_it
->ignore_bn_limit
<= bidi_it
->charpos
)
1941 || (bidi_it
->ignore_bn_limit
== -2
1942 && !bidi_explicit_dir_char (bidi_it
->ch
)))
1943 bidi_it
->ignore_bn_limit
= -1;
1945 type
= bidi_resolve_neutral (bidi_it
);
1950 /* Given an iterator state BIDI_IT, advance one character position in
1951 the buffer/string to the next character (in the current scan
1952 direction), resolve the embedding and implicit levels of that next
1953 character, and return the resulting level. */
1955 bidi_level_of_next_char (struct bidi_it
*bidi_it
)
1958 int level
, prev_level
= -1;
1959 struct bidi_saved_info next_for_neutral
;
1960 EMACS_INT next_char_pos
= -2;
1962 if (bidi_it
->scan_dir
== 1)
1965 (bidi_it
->string
.s
|| STRINGP (bidi_it
->string
.lstring
))
1966 ? bidi_it
->string
.schars
: ZV
;
1968 /* There's no sense in trying to advance if we hit end of text. */
1969 if (bidi_it
->charpos
>= eob
)
1970 return bidi_it
->resolved_level
;
1972 /* Record the info about the previous character. */
1973 if (bidi_it
->type_after_w1
!= WEAK_BN
/* W1/Retaining */
1974 && bidi_it
->type
!= WEAK_BN
)
1975 bidi_remember_char (&bidi_it
->prev
, bidi_it
);
1976 if (bidi_it
->type_after_w1
== STRONG_R
1977 || bidi_it
->type_after_w1
== STRONG_L
1978 || bidi_it
->type_after_w1
== STRONG_AL
)
1979 bidi_remember_char (&bidi_it
->last_strong
, bidi_it
);
1980 /* FIXME: it sounds like we don't need both prev and
1981 prev_for_neutral members, but I'm leaving them both for now. */
1982 if (bidi_it
->type
== STRONG_R
|| bidi_it
->type
== STRONG_L
1983 || bidi_it
->type
== WEAK_EN
|| bidi_it
->type
== WEAK_AN
)
1984 bidi_remember_char (&bidi_it
->prev_for_neutral
, bidi_it
);
1986 /* If we overstepped the characters used for resolving neutrals
1987 and whitespace, invalidate their info in the iterator. */
1988 if (bidi_it
->charpos
>= bidi_it
->next_for_neutral
.charpos
)
1989 bidi_it
->next_for_neutral
.type
= UNKNOWN_BT
;
1990 if (bidi_it
->next_en_pos
>= 0
1991 && bidi_it
->charpos
>= bidi_it
->next_en_pos
)
1992 bidi_it
->next_en_pos
= -1;
1993 if (bidi_it
->next_for_ws
.type
!= UNKNOWN_BT
1994 && bidi_it
->charpos
>= bidi_it
->next_for_ws
.charpos
)
1995 bidi_it
->next_for_ws
.type
= UNKNOWN_BT
;
1997 /* This must be taken before we fill the iterator with the info
1998 about the next char. If we scan backwards, the iterator
1999 state must be already cached, so there's no need to know the
2000 embedding level of the previous character, since we will be
2001 returning to our caller shortly. */
2002 prev_level
= bidi_it
->level_stack
[bidi_it
->stack_idx
].level
;
2004 next_for_neutral
= bidi_it
->next_for_neutral
;
2006 /* Perhaps the character we want is already cached. If it is, the
2007 call to bidi_cache_find below will return a type other than
2009 if (bidi_cache_idx
> bidi_cache_start
&& !bidi_it
->first_elt
)
2012 (bidi_it
->string
.s
|| STRINGP (bidi_it
->string
.lstring
)) ? 0 : 1;
2014 if (bidi_it
->scan_dir
> 0)
2016 if (bidi_it
->nchars
<= 0)
2018 next_char_pos
= bidi_it
->charpos
+ bidi_it
->nchars
;
2020 else if (bidi_it
->charpos
>= bob
)
2021 /* Implementation note: we allow next_char_pos to be as low as
2022 0 for buffers or -1 for strings, and that is okay because
2023 that's the "position" of the sentinel iterator state we
2024 cached at the beginning of the iteration. */
2025 next_char_pos
= bidi_it
->charpos
- 1;
2026 if (next_char_pos
>= bob
- 1)
2027 type
= bidi_cache_find (next_char_pos
, -1, bidi_it
);
2033 if (type
!= UNKNOWN_BT
)
2035 /* Don't lose the information for resolving neutrals! The
2036 cached states could have been cached before their
2037 next_for_neutral member was computed. If we are on our way
2038 forward, we can simply take the info from the previous
2040 if (bidi_it
->scan_dir
== 1
2041 && bidi_it
->next_for_neutral
.type
== UNKNOWN_BT
)
2042 bidi_it
->next_for_neutral
= next_for_neutral
;
2044 /* If resolved_level is -1, it means this state was cached
2045 before it was completely resolved, so we cannot return
2047 if (bidi_it
->resolved_level
!= -1)
2048 return bidi_it
->resolved_level
;
2050 if (bidi_it
->scan_dir
== -1)
2051 /* If we are going backwards, the iterator state is already cached
2052 from previous scans, and should be fully resolved. */
2055 if (type
== UNKNOWN_BT
)
2056 type
= bidi_type_of_next_char (bidi_it
);
2058 if (type
== NEUTRAL_B
)
2059 return bidi_it
->resolved_level
;
2061 level
= bidi_it
->level_stack
[bidi_it
->stack_idx
].level
;
2062 if ((bidi_get_category (type
) == NEUTRAL
/* && type != NEUTRAL_B */)
2063 || (type
== WEAK_BN
&& prev_level
== level
))
2065 if (bidi_it
->next_for_neutral
.type
== UNKNOWN_BT
)
2068 /* If the cached state shows a neutral character, it was not
2069 resolved by bidi_resolve_neutral, so do it now. */
2070 type
= bidi_resolve_neutral_1 (bidi_it
->prev_for_neutral
.type
,
2071 bidi_it
->next_for_neutral
.type
,
2075 if (!(type
== STRONG_R
2079 || type
== WEAK_AN
))
2081 bidi_it
->type
= type
;
2082 bidi_check_type (bidi_it
->type
);
2084 /* For L1 below, we need to know, for each WS character, whether
2085 it belongs to a sequence of WS characters preceding a newline
2086 or a TAB or a paragraph separator. */
2087 if (bidi_it
->orig_type
== NEUTRAL_WS
2088 && bidi_it
->next_for_ws
.type
== UNKNOWN_BT
)
2091 EMACS_INT clen
= bidi_it
->ch_len
;
2092 EMACS_INT bpos
= bidi_it
->bytepos
;
2093 EMACS_INT cpos
= bidi_it
->charpos
;
2094 EMACS_INT disp_pos
= bidi_it
->disp_pos
;
2095 EMACS_INT nc
= bidi_it
->nchars
;
2096 struct bidi_string_data bs
= bidi_it
->string
;
2098 int fwp
= bidi_it
->frame_window_p
;
2099 int dpp
= bidi_it
->disp_prop
;
2101 if (bidi_it
->nchars
<= 0)
2104 ch
= bidi_fetch_char (bpos
+= clen
, cpos
+= nc
, &disp_pos
, &dpp
, &bs
,
2106 if (ch
== '\n' || ch
== BIDI_EOB
/* || ch == LINESEP_CHAR */)
2109 chtype
= bidi_get_type (ch
, NEUTRAL_DIR
);
2110 } while (chtype
== NEUTRAL_WS
|| chtype
== WEAK_BN
2111 || bidi_explicit_dir_char (ch
)); /* L1/Retaining */
2112 bidi_it
->next_for_ws
.type
= chtype
;
2113 bidi_check_type (bidi_it
->next_for_ws
.type
);
2114 bidi_it
->next_for_ws
.charpos
= cpos
;
2115 bidi_it
->next_for_ws
.bytepos
= bpos
;
2118 /* Resolve implicit levels, with a twist: PDFs get the embedding
2119 level of the enbedding they terminate. See below for the
2121 if (bidi_it
->orig_type
== PDF
2122 /* Don't do this if this formatting code didn't change the
2123 embedding level due to invalid or empty embeddings. */
2124 && prev_level
!= level
)
2126 /* Don't look in UAX#9 for the reason for this: it's our own
2127 private quirk. The reason is that we want the formatting
2128 codes to be delivered so that they bracket the text of their
2129 embedding. For example, given the text
2133 we want it to be displayed as
2141 which will result because we bump up the embedding level as
2142 soon as we see the RLO and pop it as soon as we see the PDF,
2143 so RLO itself has the same embedding level as "teST", and
2144 thus would be normally delivered last, just before the PDF.
2145 The switch below fiddles with the level of PDF so that this
2146 ugly side effect does not happen.
2148 (This is, of course, only important if the formatting codes
2149 are actually displayed, but Emacs does need to display them
2150 if the user wants to.) */
2153 else if (bidi_it
->orig_type
== NEUTRAL_B
/* L1 */
2154 || bidi_it
->orig_type
== NEUTRAL_S
2155 || bidi_it
->ch
== '\n' || bidi_it
->ch
== BIDI_EOB
2156 /* || bidi_it->ch == LINESEP_CHAR */
2157 || (bidi_it
->orig_type
== NEUTRAL_WS
2158 && (bidi_it
->next_for_ws
.type
== NEUTRAL_B
2159 || bidi_it
->next_for_ws
.type
== NEUTRAL_S
)))
2160 level
= bidi_it
->level_stack
[0].level
;
2161 else if ((level
& 1) == 0) /* I1 */
2163 if (type
== STRONG_R
)
2165 else if (type
== WEAK_EN
|| type
== WEAK_AN
)
2170 if (type
== STRONG_L
|| type
== WEAK_EN
|| type
== WEAK_AN
)
2174 bidi_it
->resolved_level
= level
;
2178 /* Move to the other edge of a level given by LEVEL. If END_FLAG is
2179 non-zero, we are at the end of a level, and we need to prepare to
2180 resume the scan of the lower level.
2182 If this level's other edge is cached, we simply jump to it, filling
2183 the iterator structure with the iterator state on the other edge.
2184 Otherwise, we walk the buffer or string until we come back to the
2185 same level as LEVEL.
2187 Note: we are not talking here about a ``level run'' in the UAX#9
2188 sense of the term, but rather about a ``level'' which includes
2189 all the levels higher than it. In other words, given the levels
2192 11111112222222333333334443343222222111111112223322111
2195 and assuming we are at point A scanning left to right, this
2196 function moves to point C, whereas the UAX#9 ``level 2 run'' ends
2199 bidi_find_other_level_edge (struct bidi_it
*bidi_it
, int level
, int end_flag
)
2201 int dir
= end_flag
? -bidi_it
->scan_dir
: bidi_it
->scan_dir
;
2204 /* Try the cache first. */
2205 if ((idx
= bidi_cache_find_level_change (level
, dir
, end_flag
))
2206 >= bidi_cache_start
)
2207 bidi_cache_fetch_state (idx
, bidi_it
);
2213 abort (); /* if we are at end of level, its edges must be cached */
2215 bidi_cache_iterator_state (bidi_it
, 1);
2217 new_level
= bidi_level_of_next_char (bidi_it
);
2218 bidi_cache_iterator_state (bidi_it
, 1);
2219 } while (new_level
>= level
);
2224 bidi_move_to_visually_next (struct bidi_it
*bidi_it
)
2226 int old_level
, new_level
, next_level
;
2227 struct bidi_it sentinel
;
2228 struct gcpro gcpro1
;
2230 if (bidi_it
->charpos
< 0 || bidi_it
->bytepos
< 0)
2233 if (bidi_it
->scan_dir
== 0)
2235 bidi_it
->scan_dir
= 1; /* default to logical order */
2238 /* The code below can call eval, and thus cause GC. If we are
2239 iterating a Lisp string, make sure it won't be GCed. */
2240 if (STRINGP (bidi_it
->string
.lstring
))
2241 GCPRO1 (bidi_it
->string
.lstring
);
2243 /* If we just passed a newline, initialize for the next line. */
2244 if (!bidi_it
->first_elt
&& bidi_it
->orig_type
== NEUTRAL_B
)
2245 bidi_line_init (bidi_it
);
2247 /* Prepare the sentinel iterator state, and cache it. When we bump
2248 into it, scanning backwards, we'll know that the last non-base
2249 level is exhausted. */
2250 if (bidi_cache_idx
== bidi_cache_start
)
2252 bidi_copy_it (&sentinel
, bidi_it
);
2253 if (bidi_it
->first_elt
)
2255 sentinel
.charpos
--; /* cached charpos needs to be monotonic */
2257 sentinel
.ch
= '\n'; /* doesn't matter, but why not? */
2258 sentinel
.ch_len
= 1;
2259 sentinel
.nchars
= 1;
2261 bidi_cache_iterator_state (&sentinel
, 1);
2264 old_level
= bidi_it
->resolved_level
;
2265 new_level
= bidi_level_of_next_char (bidi_it
);
2267 /* Reordering of resolved levels (clause L2) is implemented by
2268 jumping to the other edge of the level and flipping direction of
2269 scanning the text whenever we find a level change. */
2270 if (new_level
!= old_level
)
2272 int ascending
= new_level
> old_level
;
2273 int level_to_search
= ascending
? old_level
+ 1 : old_level
;
2274 int incr
= ascending
? 1 : -1;
2275 int expected_next_level
= old_level
+ incr
;
2277 /* Jump (or walk) to the other edge of this level. */
2278 bidi_find_other_level_edge (bidi_it
, level_to_search
, !ascending
);
2279 /* Switch scan direction and peek at the next character in the
2281 bidi_it
->scan_dir
= -bidi_it
->scan_dir
;
2283 /* The following loop handles the case where the resolved level
2284 jumps by more than one. This is typical for numbers inside a
2285 run of text with left-to-right embedding direction, but can
2286 also happen in other situations. In those cases the decision
2287 where to continue after a level change, and in what direction,
2288 is tricky. For example, given a text like below:
2293 (where the numbers below the text show the resolved levels),
2294 the result of reordering according to UAX#9 should be this:
2298 This is implemented by the loop below which flips direction
2299 and jumps to the other edge of the level each time it finds
2300 the new level not to be the expected one. The expected level
2301 is always one more or one less than the previous one. */
2302 next_level
= bidi_peek_at_next_level (bidi_it
);
2303 while (next_level
!= expected_next_level
)
2305 expected_next_level
+= incr
;
2306 level_to_search
+= incr
;
2307 bidi_find_other_level_edge (bidi_it
, level_to_search
, !ascending
);
2308 bidi_it
->scan_dir
= -bidi_it
->scan_dir
;
2309 next_level
= bidi_peek_at_next_level (bidi_it
);
2312 /* Finally, deliver the next character in the new direction. */
2313 next_level
= bidi_level_of_next_char (bidi_it
);
2316 /* Take note when we have just processed the newline that precedes
2317 the end of the paragraph. The next time we are about to be
2318 called, set_iterator_to_next will automatically reinit the
2319 paragraph direction, if needed. We do this at the newline before
2320 the paragraph separator, because the next character might not be
2321 the first character of the next paragraph, due to the bidi
2322 reordering, whereas we _must_ know the paragraph base direction
2323 _before_ we process the paragraph's text, since the base
2324 direction affects the reordering. */
2325 if (bidi_it
->scan_dir
== 1 && bidi_it
->orig_type
== NEUTRAL_B
)
2327 /* The paragraph direction of the entire string, once
2328 determined, is in effect for the entire string. Setting the
2329 separator limit to the end of the string prevents
2330 bidi_paragraph_init from being called automatically on this
2332 if (bidi_it
->string
.s
|| STRINGP (bidi_it
->string
.lstring
))
2333 bidi_it
->separator_limit
= bidi_it
->string
.schars
;
2334 else if (bidi_it
->bytepos
< ZV_BYTE
)
2337 bidi_at_paragraph_end (bidi_it
->charpos
+ bidi_it
->nchars
,
2338 bidi_it
->bytepos
+ bidi_it
->ch_len
);
2339 if (bidi_it
->nchars
<= 0)
2343 bidi_it
->new_paragraph
= 1;
2344 /* Record the buffer position of the last character of the
2345 paragraph separator. */
2346 bidi_it
->separator_limit
=
2347 bidi_it
->charpos
+ bidi_it
->nchars
+ sep_len
;
2352 if (bidi_it
->scan_dir
== 1 && bidi_cache_idx
> bidi_cache_start
)
2354 /* If we are at paragraph's base embedding level and beyond the
2355 last cached position, the cache's job is done and we can
2357 if (bidi_it
->resolved_level
== bidi_it
->level_stack
[0].level
2358 && bidi_it
->charpos
> (bidi_cache
[bidi_cache_idx
- 1].charpos
2359 + bidi_cache
[bidi_cache_idx
- 1].nchars
- 1))
2360 bidi_cache_reset ();
2361 /* But as long as we are caching during forward scan, we must
2362 cache each state, or else the cache integrity will be
2363 compromised: it assumes cached states correspond to buffer
2366 bidi_cache_iterator_state (bidi_it
, 1);
2369 if (STRINGP (bidi_it
->string
.lstring
))
2373 /* This is meant to be called from within the debugger, whenever you
2374 wish to examine the cache contents. */
2375 void bidi_dump_cached_states (void) EXTERNALLY_VISIBLE
;
2377 bidi_dump_cached_states (void)
2382 if (bidi_cache_idx
== 0)
2384 fprintf (stderr
, "The cache is empty.\n");
2387 fprintf (stderr
, "Total of %"pD
"d state%s in cache:\n",
2388 bidi_cache_idx
, bidi_cache_idx
== 1 ? "" : "s");
2390 for (i
= bidi_cache
[bidi_cache_idx
- 1].charpos
; i
> 0; i
/= 10)
2392 fputs ("ch ", stderr
);
2393 for (i
= 0; i
< bidi_cache_idx
; i
++)
2394 fprintf (stderr
, "%*c", ndigits
, bidi_cache
[i
].ch
);
2395 fputs ("\n", stderr
);
2396 fputs ("lvl ", stderr
);
2397 for (i
= 0; i
< bidi_cache_idx
; i
++)
2398 fprintf (stderr
, "%*d", ndigits
, bidi_cache
[i
].resolved_level
);
2399 fputs ("\n", stderr
);
2400 fputs ("pos ", stderr
);
2401 for (i
= 0; i
< bidi_cache_idx
; i
++)
2402 fprintf (stderr
, "%*"pI
"d", ndigits
, bidi_cache
[i
].charpos
);
2403 fputs ("\n", stderr
);