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 /* 5-slot stack for saving the start of the previous level of the
320 cache. xdisp.c maintains a 5-slot stack for its iterator state,
321 and we need the same size of our stack. */
322 static ptrdiff_t bidi_cache_start_stack
[IT_STACK_SIZE
];
323 static int bidi_cache_sp
;
325 /* Size of header used by bidi_shelve_cache. */
328 bidi_shelve_header_size
=
329 (sizeof (bidi_cache_idx
) + sizeof (bidi_cache_start_stack
)
330 + sizeof (bidi_cache_sp
) + sizeof (bidi_cache_start
)
331 + sizeof (bidi_cache_last_idx
))
334 /* Reset the cache state to the empty state. We only reset the part
335 of the cache relevant to iteration of the current object. Previous
336 objects, which are pushed on the display iterator's stack, are left
337 intact. This is called when the cached information is no more
338 useful for the current iteration, e.g. when we were reseated to a
339 new position on the same object. */
341 bidi_cache_reset (void)
343 bidi_cache_idx
= bidi_cache_start
;
344 bidi_cache_last_idx
= -1;
347 /* Shrink the cache to its minimal size. Called when we init the bidi
348 iterator for reordering a buffer or a string that does not come
349 from display properties, because that means all the previously
350 cached info is of no further use. */
352 bidi_cache_shrink (void)
354 if (bidi_cache_size
> BIDI_CACHE_CHUNK
)
357 (struct bidi_it
*) xrealloc (bidi_cache
, BIDI_CACHE_CHUNK
* elsz
);
358 bidi_cache_size
= BIDI_CACHE_CHUNK
;
364 bidi_cache_fetch_state (ptrdiff_t idx
, struct bidi_it
*bidi_it
)
366 int current_scan_dir
= bidi_it
->scan_dir
;
368 if (idx
< bidi_cache_start
|| idx
>= bidi_cache_idx
)
371 bidi_copy_it (bidi_it
, &bidi_cache
[idx
]);
372 bidi_it
->scan_dir
= current_scan_dir
;
373 bidi_cache_last_idx
= idx
;
376 /* Find a cached state with a given CHARPOS and resolved embedding
377 level less or equal to LEVEL. if LEVEL is -1, disregard the
378 resolved levels in cached states. DIR, if non-zero, means search
379 in that direction from the last cache hit. */
380 static inline ptrdiff_t
381 bidi_cache_search (EMACS_INT charpos
, int level
, int dir
)
383 ptrdiff_t i
, i_start
;
385 if (bidi_cache_idx
> bidi_cache_start
)
387 if (bidi_cache_last_idx
== -1)
388 bidi_cache_last_idx
= bidi_cache_idx
- 1;
389 if (charpos
< bidi_cache
[bidi_cache_last_idx
].charpos
)
392 i_start
= bidi_cache_last_idx
- 1;
394 else if (charpos
> (bidi_cache
[bidi_cache_last_idx
].charpos
395 + bidi_cache
[bidi_cache_last_idx
].nchars
- 1))
398 i_start
= bidi_cache_last_idx
+ 1;
401 i_start
= bidi_cache_last_idx
;
405 i_start
= bidi_cache_idx
- 1;
410 /* Linear search for now; FIXME! */
411 for (i
= i_start
; i
>= bidi_cache_start
; i
--)
412 if (bidi_cache
[i
].charpos
<= charpos
413 && charpos
< bidi_cache
[i
].charpos
+ bidi_cache
[i
].nchars
414 && (level
== -1 || bidi_cache
[i
].resolved_level
<= level
))
419 for (i
= i_start
; i
< bidi_cache_idx
; i
++)
420 if (bidi_cache
[i
].charpos
<= charpos
421 && charpos
< bidi_cache
[i
].charpos
+ bidi_cache
[i
].nchars
422 && (level
== -1 || bidi_cache
[i
].resolved_level
<= level
))
430 /* Find a cached state where the resolved level changes to a value
431 that is lower than LEVEL, and return its cache slot index. DIR is
432 the direction to search, starting with the last used cache slot.
433 If DIR is zero, we search backwards from the last occupied cache
434 slot. BEFORE, if non-zero, means return the index of the slot that
435 is ``before'' the level change in the search direction. That is,
436 given the cached levels like this:
441 and assuming we are at the position cached at the slot marked with
442 C, searching backwards (DIR = -1) for LEVEL = 2 will return the
443 index of slot B or A, depending whether BEFORE is, respectively,
446 bidi_cache_find_level_change (int level
, int dir
, int before
)
450 ptrdiff_t i
= dir
? bidi_cache_last_idx
: bidi_cache_idx
- 1;
451 int incr
= before
? 1 : 0;
453 xassert (!dir
|| bidi_cache_last_idx
>= 0);
462 while (i
>= bidi_cache_start
+ incr
)
464 if (bidi_cache
[i
- incr
].resolved_level
>= 0
465 && bidi_cache
[i
- incr
].resolved_level
< level
)
472 while (i
< bidi_cache_idx
- incr
)
474 if (bidi_cache
[i
+ incr
].resolved_level
>= 0
475 && bidi_cache
[i
+ incr
].resolved_level
< level
)
486 bidi_cache_ensure_space (ptrdiff_t idx
)
488 /* Enlarge the cache as needed. */
489 if (idx
>= bidi_cache_size
)
491 /* The bidi cache cannot be larger than the largest Lisp string
493 ptrdiff_t string_or_buffer_bound
=
494 max (BUF_BYTES_MAX
, STRING_BYTES_BOUND
);
496 /* Also, it cannot be larger than what C can represent. */
498 (min (PTRDIFF_MAX
, SIZE_MAX
) - bidi_shelve_header_size
) / elsz
;
501 xpalloc (bidi_cache
, &bidi_cache_size
,
502 max (BIDI_CACHE_CHUNK
, idx
- bidi_cache_size
+ 1),
503 min (string_or_buffer_bound
, c_bound
), elsz
);
508 bidi_cache_iterator_state (struct bidi_it
*bidi_it
, int resolved
)
512 /* We should never cache on backward scans. */
513 if (bidi_it
->scan_dir
== -1)
515 idx
= bidi_cache_search (bidi_it
->charpos
, -1, 1);
519 idx
= bidi_cache_idx
;
520 bidi_cache_ensure_space (idx
);
521 /* Character positions should correspond to cache positions 1:1.
522 If we are outside the range of cached positions, the cache is
523 useless and must be reset. */
524 if (idx
> bidi_cache_start
&&
525 (bidi_it
->charpos
> (bidi_cache
[idx
- 1].charpos
526 + bidi_cache
[idx
- 1].nchars
)
527 || bidi_it
->charpos
< bidi_cache
[bidi_cache_start
].charpos
))
530 idx
= bidi_cache_start
;
532 if (bidi_it
->nchars
<= 0)
534 bidi_copy_it (&bidi_cache
[idx
], bidi_it
);
536 bidi_cache
[idx
].resolved_level
= -1;
540 /* Copy only the members which could have changed, to avoid
541 costly copying of the entire struct. */
542 bidi_cache
[idx
].type
= bidi_it
->type
;
543 bidi_check_type (bidi_it
->type
);
544 bidi_cache
[idx
].type_after_w1
= bidi_it
->type_after_w1
;
545 bidi_check_type (bidi_it
->type_after_w1
);
547 bidi_cache
[idx
].resolved_level
= bidi_it
->resolved_level
;
549 bidi_cache
[idx
].resolved_level
= -1;
550 bidi_cache
[idx
].invalid_levels
= bidi_it
->invalid_levels
;
551 bidi_cache
[idx
].invalid_rl_levels
= bidi_it
->invalid_rl_levels
;
552 bidi_cache
[idx
].next_for_neutral
= bidi_it
->next_for_neutral
;
553 bidi_cache
[idx
].next_for_ws
= bidi_it
->next_for_ws
;
554 bidi_cache
[idx
].ignore_bn_limit
= bidi_it
->ignore_bn_limit
;
555 bidi_cache
[idx
].disp_pos
= bidi_it
->disp_pos
;
556 bidi_cache
[idx
].disp_prop
= bidi_it
->disp_prop
;
559 bidi_cache_last_idx
= idx
;
560 if (idx
>= bidi_cache_idx
)
561 bidi_cache_idx
= idx
+ 1;
564 static inline bidi_type_t
565 bidi_cache_find (EMACS_INT charpos
, int level
, struct bidi_it
*bidi_it
)
567 ptrdiff_t i
= bidi_cache_search (charpos
, level
, bidi_it
->scan_dir
);
569 if (i
>= bidi_cache_start
)
571 bidi_dir_t current_scan_dir
= bidi_it
->scan_dir
;
573 bidi_copy_it (bidi_it
, &bidi_cache
[i
]);
574 bidi_cache_last_idx
= i
;
575 /* Don't let scan direction from from the cached state override
576 the current scan direction. */
577 bidi_it
->scan_dir
= current_scan_dir
;
578 return bidi_it
->type
;
585 bidi_peek_at_next_level (struct bidi_it
*bidi_it
)
587 if (bidi_cache_idx
== bidi_cache_start
|| bidi_cache_last_idx
== -1)
589 return bidi_cache
[bidi_cache_last_idx
+ bidi_it
->scan_dir
].resolved_level
;
593 /***********************************************************************
594 Pushing and popping the bidi iterator state
595 ***********************************************************************/
597 /* Push the bidi iterator state in preparation for reordering a
598 different object, e.g. display string found at certain buffer
599 position. Pushing the bidi iterator boils down to saving its
600 entire state on the cache and starting a new cache "stacked" on top
601 of the current cache. */
603 bidi_push_it (struct bidi_it
*bidi_it
)
605 /* Save the current iterator state in its entirety after the last
607 bidi_cache_ensure_space (bidi_cache_idx
);
608 memcpy (&bidi_cache
[bidi_cache_idx
++], bidi_it
, sizeof (struct bidi_it
));
610 /* Push the current cache start onto the stack. */
611 xassert (bidi_cache_sp
< IT_STACK_SIZE
);
612 bidi_cache_start_stack
[bidi_cache_sp
++] = bidi_cache_start
;
614 /* Start a new level of cache, and make it empty. */
615 bidi_cache_start
= bidi_cache_idx
;
616 bidi_cache_last_idx
= -1;
619 /* Restore the iterator state saved by bidi_push_it and return the
620 cache to the corresponding state. */
622 bidi_pop_it (struct bidi_it
*bidi_it
)
624 if (bidi_cache_start
<= 0)
627 /* Reset the next free cache slot index to what it was before the
628 call to bidi_push_it. */
629 bidi_cache_idx
= bidi_cache_start
- 1;
631 /* Restore the bidi iterator state saved in the cache. */
632 memcpy (bidi_it
, &bidi_cache
[bidi_cache_idx
], sizeof (struct bidi_it
));
634 /* Pop the previous cache start from the stack. */
635 if (bidi_cache_sp
<= 0)
637 bidi_cache_start
= bidi_cache_start_stack
[--bidi_cache_sp
];
639 /* Invalidate the last-used cache slot data. */
640 bidi_cache_last_idx
= -1;
643 static ptrdiff_t bidi_cache_total_alloc
;
645 /* Stash away a copy of the cache and its control variables. */
647 bidi_shelve_cache (void)
649 unsigned char *databuf
;
653 if (bidi_cache_idx
== 0)
656 alloc
= (bidi_shelve_header_size
657 + bidi_cache_idx
* sizeof (struct bidi_it
));
658 databuf
= xmalloc (alloc
);
659 bidi_cache_total_alloc
+= alloc
;
661 memcpy (databuf
, &bidi_cache_idx
, sizeof (bidi_cache_idx
));
662 memcpy (databuf
+ sizeof (bidi_cache_idx
),
663 bidi_cache
, bidi_cache_idx
* sizeof (struct bidi_it
));
664 memcpy (databuf
+ sizeof (bidi_cache_idx
)
665 + bidi_cache_idx
* sizeof (struct bidi_it
),
666 bidi_cache_start_stack
, sizeof (bidi_cache_start_stack
));
667 memcpy (databuf
+ sizeof (bidi_cache_idx
)
668 + bidi_cache_idx
* sizeof (struct bidi_it
)
669 + sizeof (bidi_cache_start_stack
),
670 &bidi_cache_sp
, sizeof (bidi_cache_sp
));
671 memcpy (databuf
+ sizeof (bidi_cache_idx
)
672 + bidi_cache_idx
* sizeof (struct bidi_it
)
673 + sizeof (bidi_cache_start_stack
) + sizeof (bidi_cache_sp
),
674 &bidi_cache_start
, sizeof (bidi_cache_start
));
675 memcpy (databuf
+ sizeof (bidi_cache_idx
)
676 + bidi_cache_idx
* sizeof (struct bidi_it
)
677 + sizeof (bidi_cache_start_stack
) + sizeof (bidi_cache_sp
)
678 + sizeof (bidi_cache_start
),
679 &bidi_cache_last_idx
, sizeof (bidi_cache_last_idx
));
684 /* Restore the cache state from a copy stashed away by
685 bidi_shelve_cache, and free the buffer used to stash that copy.
686 JUST_FREE non-zero means free the buffer, but don't restore the
687 cache; used when the corresponding iterator is discarded instead of
690 bidi_unshelve_cache (void *databuf
, int just_free
)
692 unsigned char *p
= databuf
;
698 /* A NULL pointer means an empty cache. */
699 bidi_cache_start
= 0;
710 memcpy (&idx
, p
, sizeof (bidi_cache_idx
));
711 bidi_cache_total_alloc
-=
712 bidi_shelve_header_size
+ idx
* sizeof (struct bidi_it
);
716 memcpy (&bidi_cache_idx
, p
, sizeof (bidi_cache_idx
));
717 bidi_cache_ensure_space (bidi_cache_idx
);
718 memcpy (bidi_cache
, p
+ sizeof (bidi_cache_idx
),
719 bidi_cache_idx
* sizeof (struct bidi_it
));
720 memcpy (bidi_cache_start_stack
,
721 p
+ sizeof (bidi_cache_idx
)
722 + bidi_cache_idx
* sizeof (struct bidi_it
),
723 sizeof (bidi_cache_start_stack
));
724 memcpy (&bidi_cache_sp
,
725 p
+ sizeof (bidi_cache_idx
)
726 + bidi_cache_idx
* sizeof (struct bidi_it
)
727 + sizeof (bidi_cache_start_stack
),
728 sizeof (bidi_cache_sp
));
729 memcpy (&bidi_cache_start
,
730 p
+ sizeof (bidi_cache_idx
)
731 + bidi_cache_idx
* sizeof (struct bidi_it
)
732 + sizeof (bidi_cache_start_stack
) + sizeof (bidi_cache_sp
),
733 sizeof (bidi_cache_start
));
734 memcpy (&bidi_cache_last_idx
,
735 p
+ sizeof (bidi_cache_idx
)
736 + bidi_cache_idx
* sizeof (struct bidi_it
)
737 + sizeof (bidi_cache_start_stack
) + sizeof (bidi_cache_sp
)
738 + sizeof (bidi_cache_start
),
739 sizeof (bidi_cache_last_idx
));
740 bidi_cache_total_alloc
-=
741 bidi_shelve_header_size
+ bidi_cache_idx
* sizeof (struct bidi_it
);
749 /***********************************************************************
751 ***********************************************************************/
753 bidi_initialize (void)
755 bidi_type_table
= uniprop_table (intern ("bidi-class"));
756 if (NILP (bidi_type_table
))
758 staticpro (&bidi_type_table
);
760 bidi_mirror_table
= uniprop_table (intern ("mirroring"));
761 if (NILP (bidi_mirror_table
))
763 staticpro (&bidi_mirror_table
);
765 Qparagraph_start
= intern ("paragraph-start");
766 staticpro (&Qparagraph_start
);
767 paragraph_start_re
= Fsymbol_value (Qparagraph_start
);
768 if (!STRINGP (paragraph_start_re
))
769 paragraph_start_re
= build_string ("\f\\|[ \t]*$");
770 staticpro (¶graph_start_re
);
771 Qparagraph_separate
= intern ("paragraph-separate");
772 staticpro (&Qparagraph_separate
);
773 paragraph_separate_re
= Fsymbol_value (Qparagraph_separate
);
774 if (!STRINGP (paragraph_separate_re
))
775 paragraph_separate_re
= build_string ("[ \t\f]*$");
776 staticpro (¶graph_separate_re
);
779 bidi_cache_total_alloc
= 0;
781 bidi_initialized
= 1;
784 /* Do whatever UAX#9 clause X8 says should be done at paragraph's
787 bidi_set_paragraph_end (struct bidi_it
*bidi_it
)
789 bidi_it
->invalid_levels
= 0;
790 bidi_it
->invalid_rl_levels
= -1;
791 bidi_it
->stack_idx
= 0;
792 bidi_it
->resolved_level
= bidi_it
->level_stack
[0].level
;
795 /* Initialize the bidi iterator from buffer/string position CHARPOS. */
797 bidi_init_it (EMACS_INT charpos
, EMACS_INT bytepos
, int frame_window_p
,
798 struct bidi_it
*bidi_it
)
800 if (! bidi_initialized
)
803 bidi_it
->charpos
= charpos
;
805 bidi_it
->bytepos
= bytepos
;
806 bidi_it
->frame_window_p
= frame_window_p
;
807 bidi_it
->nchars
= -1; /* to be computed in bidi_resolve_explicit_1 */
808 bidi_it
->first_elt
= 1;
809 bidi_set_paragraph_end (bidi_it
);
810 bidi_it
->new_paragraph
= 1;
811 bidi_it
->separator_limit
= -1;
812 bidi_it
->type
= NEUTRAL_B
;
813 bidi_it
->type_after_w1
= NEUTRAL_B
;
814 bidi_it
->orig_type
= NEUTRAL_B
;
815 bidi_it
->prev_was_pdf
= 0;
816 bidi_it
->prev
.type
= bidi_it
->prev
.type_after_w1
=
817 bidi_it
->prev
.orig_type
= UNKNOWN_BT
;
818 bidi_it
->last_strong
.type
= bidi_it
->last_strong
.type_after_w1
=
819 bidi_it
->last_strong
.orig_type
= UNKNOWN_BT
;
820 bidi_it
->next_for_neutral
.charpos
= -1;
821 bidi_it
->next_for_neutral
.type
=
822 bidi_it
->next_for_neutral
.type_after_w1
=
823 bidi_it
->next_for_neutral
.orig_type
= UNKNOWN_BT
;
824 bidi_it
->prev_for_neutral
.charpos
= -1;
825 bidi_it
->prev_for_neutral
.type
=
826 bidi_it
->prev_for_neutral
.type_after_w1
=
827 bidi_it
->prev_for_neutral
.orig_type
= UNKNOWN_BT
;
828 bidi_it
->sor
= L2R
; /* FIXME: should it be user-selectable? */
829 bidi_it
->disp_pos
= -1; /* invalid/unknown */
830 bidi_it
->disp_prop
= 0;
831 /* We can only shrink the cache if we are at the bottom level of its
833 if (bidi_cache_start
== 0)
834 bidi_cache_shrink ();
839 /* Perform initializations for reordering a new line of bidi text. */
841 bidi_line_init (struct bidi_it
*bidi_it
)
843 bidi_it
->scan_dir
= 1; /* FIXME: do we need to have control on this? */
844 bidi_it
->resolved_level
= bidi_it
->level_stack
[0].level
;
845 bidi_it
->level_stack
[0].override
= NEUTRAL_DIR
; /* X1 */
846 bidi_it
->invalid_levels
= 0;
847 bidi_it
->invalid_rl_levels
= -1;
848 bidi_it
->next_en_pos
= -1;
849 bidi_it
->next_for_ws
.type
= UNKNOWN_BT
;
850 bidi_set_sor_type (bidi_it
,
851 bidi_it
->paragraph_dir
== R2L
? 1 : 0,
852 bidi_it
->level_stack
[0].level
); /* X10 */
858 /***********************************************************************
860 ***********************************************************************/
862 /* Count bytes in string S between BEG/BEGBYTE and END. BEG and END
863 are zero-based character positions in S, BEGBYTE is byte position
864 corresponding to BEG. UNIBYTE, if non-zero, means S is a unibyte
866 static inline EMACS_INT
867 bidi_count_bytes (const unsigned char *s
, const EMACS_INT beg
,
868 const EMACS_INT begbyte
, const EMACS_INT end
, int unibyte
)
871 const unsigned char *p
= s
+ begbyte
, *start
= p
;
877 if (!CHAR_HEAD_P (*p
))
882 p
+= BYTES_BY_CHAR_HEAD (*p
);
890 /* Fetch and returns the character at byte position BYTEPOS. If S is
891 non-NULL, fetch the character from string S; otherwise fetch the
892 character from the current buffer. UNIBYTE non-zero means S is a
895 bidi_char_at_pos (EMACS_INT bytepos
, const unsigned char *s
, int unibyte
)
902 return STRING_CHAR (s
+ bytepos
);
905 return FETCH_MULTIBYTE_CHAR (bytepos
);
908 /* Fetch and return the character at BYTEPOS/CHARPOS. If that
909 character is covered by a display string, treat the entire run of
910 covered characters as a single character, either u+2029 or u+FFFC,
911 and return their combined length in CH_LEN and NCHARS. DISP_POS
912 specifies the character position of the next display string, or -1
913 if not yet computed. When the next character is at or beyond that
914 position, the function updates DISP_POS with the position of the
915 next display string. DISP_PROP non-zero means that there's really
916 a display string at DISP_POS, as opposed to when we searched till
917 DISP_POS without finding one. If DISP_PROP is 2, it means the
918 display spec is of the form `(space ...)', which is replaced with
919 u+2029 to handle it as a paragraph separator. STRING->s is the C
920 string to iterate, or NULL if iterating over a buffer or a Lisp
921 string; in the latter case, STRING->lstring is the Lisp string. */
923 bidi_fetch_char (EMACS_INT bytepos
, EMACS_INT charpos
, EMACS_INT
*disp_pos
,
924 int *disp_prop
, struct bidi_string_data
*string
,
925 int frame_window_p
, EMACS_INT
*ch_len
, EMACS_INT
*nchars
)
929 (string
->s
|| STRINGP (string
->lstring
)) ? string
->schars
: ZV
;
932 /* If we got past the last known position of display string, compute
933 the position of the next one. That position could be at CHARPOS. */
934 if (charpos
< endpos
&& charpos
> *disp_pos
)
936 SET_TEXT_POS (pos
, charpos
, bytepos
);
937 *disp_pos
= compute_display_string_pos (&pos
, string
, frame_window_p
,
941 /* Fetch the character at BYTEPOS. */
942 if (charpos
>= endpos
)
950 else if (charpos
>= *disp_pos
&& *disp_prop
)
952 EMACS_INT disp_end_pos
;
954 /* We don't expect to find ourselves in the middle of a display
955 property. Hopefully, it will never be needed. */
956 if (charpos
> *disp_pos
)
958 /* Text covered by `display' properties and overlays with
959 display properties or display strings is handled as a single
960 character that represents the entire run of characters
961 covered by the display property. */
964 /* `(space ...)' display specs are handled as paragraph
965 separators for the purposes of the reordering; see UAX#9
966 section 3 and clause HL1 in section 4.3 there. */
971 /* All other display specs are handled as the Unicode Object
972 Replacement Character. */
975 disp_end_pos
= compute_display_string_end (*disp_pos
, string
);
976 *nchars
= disp_end_pos
- *disp_pos
;
980 *ch_len
= bidi_count_bytes (string
->s
, *disp_pos
, bytepos
,
981 disp_end_pos
, string
->unibyte
);
982 else if (STRINGP (string
->lstring
))
983 *ch_len
= bidi_count_bytes (SDATA (string
->lstring
), *disp_pos
,
984 bytepos
, disp_end_pos
, string
->unibyte
);
986 *ch_len
= CHAR_TO_BYTE (disp_end_pos
) - bytepos
;
994 if (!string
->unibyte
)
996 ch
= STRING_CHAR_AND_LENGTH (string
->s
+ bytepos
, len
);
1001 ch
= UNIBYTE_TO_CHAR (string
->s
[bytepos
]);
1005 else if (STRINGP (string
->lstring
))
1009 if (!string
->unibyte
)
1011 ch
= STRING_CHAR_AND_LENGTH (SDATA (string
->lstring
) + bytepos
,
1017 ch
= UNIBYTE_TO_CHAR (SREF (string
->lstring
, bytepos
));
1023 ch
= FETCH_MULTIBYTE_CHAR (bytepos
);
1024 *ch_len
= CHAR_BYTES (ch
);
1029 /* If we just entered a run of characters covered by a display
1030 string, compute the position of the next display string. */
1031 if (charpos
+ *nchars
<= endpos
&& charpos
+ *nchars
> *disp_pos
1034 SET_TEXT_POS (pos
, charpos
+ *nchars
, bytepos
+ *ch_len
);
1035 *disp_pos
= compute_display_string_pos (&pos
, string
, frame_window_p
,
1043 /***********************************************************************
1044 Determining paragraph direction
1045 ***********************************************************************/
1047 /* Check if buffer position CHARPOS/BYTEPOS is the end of a paragraph.
1048 Value is the non-negative length of the paragraph separator
1049 following the buffer position, -1 if position is at the beginning
1050 of a new paragraph, or -2 if position is neither at beginning nor
1051 at end of a paragraph. */
1053 bidi_at_paragraph_end (EMACS_INT charpos
, EMACS_INT bytepos
)
1056 Lisp_Object start_re
;
1059 sep_re
= paragraph_separate_re
;
1060 start_re
= paragraph_start_re
;
1062 val
= fast_looking_at (sep_re
, charpos
, bytepos
, ZV
, ZV_BYTE
, Qnil
);
1065 if (fast_looking_at (start_re
, charpos
, bytepos
, ZV
, ZV_BYTE
, Qnil
) >= 0)
1074 /* Find the beginning of this paragraph by looking back in the buffer.
1075 Value is the byte position of the paragraph's beginning. */
1077 bidi_find_paragraph_start (EMACS_INT pos
, EMACS_INT pos_byte
)
1079 Lisp_Object re
= paragraph_start_re
;
1080 EMACS_INT limit
= ZV
, limit_byte
= ZV_BYTE
;
1082 while (pos_byte
> BEGV_BYTE
1083 && fast_looking_at (re
, pos
, pos_byte
, limit
, limit_byte
, Qnil
) < 0)
1085 /* FIXME: What if the paragraph beginning is covered by a
1086 display string? And what if a display string covering some
1087 of the text over which we scan back includes
1088 paragraph_start_re? */
1089 pos
= find_next_newline_no_quit (pos
- 1, -1);
1090 pos_byte
= CHAR_TO_BYTE (pos
);
1095 /* Determine the base direction, a.k.a. base embedding level, of the
1096 paragraph we are about to iterate through. If DIR is either L2R or
1097 R2L, just use that. Otherwise, determine the paragraph direction
1098 from the first strong directional character of the paragraph.
1100 NO_DEFAULT_P non-zero means don't default to L2R if the paragraph
1101 has no strong directional characters and both DIR and
1102 bidi_it->paragraph_dir are NEUTRAL_DIR. In that case, search back
1103 in the buffer until a paragraph is found with a strong character,
1104 or until hitting BEGV. In the latter case, fall back to L2R. This
1105 flag is used in current-bidi-paragraph-direction.
1107 Note that this function gives the paragraph separator the same
1108 direction as the preceding paragraph, even though Emacs generally
1109 views the separartor as not belonging to any paragraph. */
1111 bidi_paragraph_init (bidi_dir_t dir
, struct bidi_it
*bidi_it
, int no_default_p
)
1113 EMACS_INT bytepos
= bidi_it
->bytepos
;
1114 int string_p
= bidi_it
->string
.s
!= NULL
|| STRINGP (bidi_it
->string
.lstring
);
1115 EMACS_INT pstartbyte
;
1116 /* Note that begbyte is a byte position, while end is a character
1117 position. Yes, this is ugly, but we are trying to avoid costly
1118 calls to BYTE_TO_CHAR and its ilk. */
1119 EMACS_INT begbyte
= string_p
? 0 : BEGV_BYTE
;
1120 EMACS_INT end
= string_p
? bidi_it
->string
.schars
: ZV
;
1122 /* Special case for an empty buffer. */
1123 if (bytepos
== begbyte
&& bidi_it
->charpos
== end
)
1125 /* We should never be called at EOB or before BEGV. */
1126 else if (bidi_it
->charpos
>= end
|| bytepos
< begbyte
)
1131 bidi_it
->paragraph_dir
= L2R
;
1132 bidi_it
->new_paragraph
= 0;
1134 else if (dir
== R2L
)
1136 bidi_it
->paragraph_dir
= R2L
;
1137 bidi_it
->new_paragraph
= 0;
1139 else if (dir
== NEUTRAL_DIR
) /* P2 */
1142 EMACS_INT ch_len
, nchars
;
1143 EMACS_INT pos
, disp_pos
= -1;
1146 const unsigned char *s
;
1148 if (!bidi_initialized
)
1151 /* If we are inside a paragraph separator, we are just waiting
1152 for the separator to be exhausted; use the previous paragraph
1153 direction. But don't do that if we have been just reseated,
1154 because we need to reinitialize below in that case. */
1155 if (!bidi_it
->first_elt
1156 && bidi_it
->charpos
< bidi_it
->separator_limit
)
1159 /* If we are on a newline, get past it to where the next
1160 paragraph might start. But don't do that at BEGV since then
1161 we are potentially in a new paragraph that doesn't yet
1163 pos
= bidi_it
->charpos
;
1164 s
= STRINGP (bidi_it
->string
.lstring
) ?
1165 SDATA (bidi_it
->string
.lstring
) : bidi_it
->string
.s
;
1166 if (bytepos
> begbyte
1167 && bidi_char_at_pos (bytepos
, s
, bidi_it
->string
.unibyte
) == '\n')
1173 /* We are either at the beginning of a paragraph or in the
1174 middle of it. Find where this paragraph starts. */
1177 /* We don't support changes of paragraph direction inside a
1178 string. It is treated as a single paragraph. */
1182 pstartbyte
= bidi_find_paragraph_start (pos
, bytepos
);
1183 bidi_it
->separator_limit
= -1;
1184 bidi_it
->new_paragraph
= 0;
1186 /* The following loop is run more than once only if NO_DEFAULT_P
1187 is non-zero, and only if we are iterating on a buffer. */
1189 bytepos
= pstartbyte
;
1191 pos
= BYTE_TO_CHAR (bytepos
);
1192 ch
= bidi_fetch_char (bytepos
, pos
, &disp_pos
, &disp_prop
,
1194 bidi_it
->frame_window_p
, &ch_len
, &nchars
);
1195 type
= bidi_get_type (ch
, NEUTRAL_DIR
);
1197 for (pos
+= nchars
, bytepos
+= ch_len
;
1198 (bidi_get_category (type
) != STRONG
)
1199 || (bidi_ignore_explicit_marks_for_paragraph_level
1200 && (type
== RLE
|| type
== RLO
1201 || type
== LRE
|| type
== LRO
));
1202 type
= bidi_get_type (ch
, NEUTRAL_DIR
))
1206 /* Pretend there's a paragraph separator at end of
1212 && type
== NEUTRAL_B
1213 && bidi_at_paragraph_end (pos
, bytepos
) >= -1)
1215 /* Fetch next character and advance to get past it. */
1216 ch
= bidi_fetch_char (bytepos
, pos
, &disp_pos
,
1217 &disp_prop
, &bidi_it
->string
,
1218 bidi_it
->frame_window_p
, &ch_len
, &nchars
);
1222 if ((type
== STRONG_R
|| type
== STRONG_AL
) /* P3 */
1223 || (!bidi_ignore_explicit_marks_for_paragraph_level
1224 && (type
== RLO
|| type
== RLE
)))
1225 bidi_it
->paragraph_dir
= R2L
;
1226 else if (type
== STRONG_L
1227 || (!bidi_ignore_explicit_marks_for_paragraph_level
1228 && (type
== LRO
|| type
== LRE
)))
1229 bidi_it
->paragraph_dir
= L2R
;
1231 && no_default_p
&& bidi_it
->paragraph_dir
== NEUTRAL_DIR
)
1233 /* If this paragraph is at BEGV, default to L2R. */
1234 if (pstartbyte
== BEGV_BYTE
)
1235 bidi_it
->paragraph_dir
= L2R
; /* P3 and HL1 */
1238 EMACS_INT prevpbyte
= pstartbyte
;
1239 EMACS_INT p
= BYTE_TO_CHAR (pstartbyte
), pbyte
= pstartbyte
;
1241 /* Find the beginning of the previous paragraph, if any. */
1242 while (pbyte
> BEGV_BYTE
&& prevpbyte
>= pstartbyte
)
1244 /* FXIME: What if p is covered by a display
1245 string? See also a FIXME inside
1246 bidi_find_paragraph_start. */
1248 pbyte
= CHAR_TO_BYTE (p
);
1249 prevpbyte
= bidi_find_paragraph_start (p
, pbyte
);
1251 pstartbyte
= prevpbyte
;
1255 && no_default_p
&& bidi_it
->paragraph_dir
== NEUTRAL_DIR
);
1260 /* Contrary to UAX#9 clause P3, we only default the paragraph
1261 direction to L2R if we have no previous usable paragraph
1262 direction. This is allowed by the HL1 clause. */
1263 if (bidi_it
->paragraph_dir
!= L2R
&& bidi_it
->paragraph_dir
!= R2L
)
1264 bidi_it
->paragraph_dir
= L2R
; /* P3 and HL1 ``higher-level protocols'' */
1265 if (bidi_it
->paragraph_dir
== R2L
)
1266 bidi_it
->level_stack
[0].level
= 1;
1268 bidi_it
->level_stack
[0].level
= 0;
1270 bidi_line_init (bidi_it
);
1274 /***********************************************************************
1275 Resolving explicit and implicit levels.
1276 The rest of this file constitutes the core of the UBA implementation.
1277 ***********************************************************************/
1280 bidi_explicit_dir_char (int ch
)
1282 bidi_type_t ch_type
;
1284 if (!bidi_initialized
)
1286 ch_type
= (bidi_type_t
) XINT (CHAR_TABLE_REF (bidi_type_table
, ch
));
1287 return (ch_type
== LRE
|| ch_type
== LRO
1288 || ch_type
== RLE
|| ch_type
== RLO
1292 /* A helper function for bidi_resolve_explicit. It advances to the
1293 next character in logical order and determines the new embedding
1294 level and directional override, but does not take into account
1295 empty embeddings. */
1297 bidi_resolve_explicit_1 (struct bidi_it
*bidi_it
)
1303 bidi_dir_t override
;
1304 int string_p
= bidi_it
->string
.s
!= NULL
|| STRINGP (bidi_it
->string
.lstring
);
1306 /* If reseat()'ed, don't advance, so as to start iteration from the
1307 position where we were reseated. bidi_it->bytepos can be less
1308 than BEGV_BYTE after reseat to BEGV. */
1309 if (bidi_it
->bytepos
< (string_p
? 0 : BEGV_BYTE
)
1310 || bidi_it
->first_elt
)
1312 bidi_it
->first_elt
= 0;
1315 const unsigned char *p
=
1316 STRINGP (bidi_it
->string
.lstring
)
1317 ? SDATA (bidi_it
->string
.lstring
) : bidi_it
->string
.s
;
1319 if (bidi_it
->charpos
< 0)
1320 bidi_it
->charpos
= 0;
1321 bidi_it
->bytepos
= bidi_count_bytes (p
, 0, 0, bidi_it
->charpos
,
1322 bidi_it
->string
.unibyte
);
1326 if (bidi_it
->charpos
< BEGV
)
1327 bidi_it
->charpos
= BEGV
;
1328 bidi_it
->bytepos
= CHAR_TO_BYTE (bidi_it
->charpos
);
1331 /* Don't move at end of buffer/string. */
1332 else if (bidi_it
->charpos
< (string_p
? bidi_it
->string
.schars
: ZV
))
1334 /* Advance to the next character, skipping characters covered by
1335 display strings (nchars > 1). */
1336 if (bidi_it
->nchars
<= 0)
1338 bidi_it
->charpos
+= bidi_it
->nchars
;
1339 if (bidi_it
->ch_len
== 0)
1341 bidi_it
->bytepos
+= bidi_it
->ch_len
;
1344 current_level
= bidi_it
->level_stack
[bidi_it
->stack_idx
].level
; /* X1 */
1345 override
= bidi_it
->level_stack
[bidi_it
->stack_idx
].override
;
1346 new_level
= current_level
;
1348 if (bidi_it
->charpos
>= (string_p
? bidi_it
->string
.schars
: ZV
))
1351 bidi_it
->ch_len
= 1;
1352 bidi_it
->nchars
= 1;
1353 bidi_it
->disp_pos
= (string_p
? bidi_it
->string
.schars
: ZV
);
1354 bidi_it
->disp_prop
= 0;
1358 /* Fetch the character at BYTEPOS. If it is covered by a
1359 display string, treat the entire run of covered characters as
1360 a single character u+FFFC. */
1361 curchar
= bidi_fetch_char (bidi_it
->bytepos
, bidi_it
->charpos
,
1362 &bidi_it
->disp_pos
, &bidi_it
->disp_prop
,
1363 &bidi_it
->string
, bidi_it
->frame_window_p
,
1364 &bidi_it
->ch_len
, &bidi_it
->nchars
);
1366 bidi_it
->ch
= curchar
;
1368 /* Don't apply directional override here, as all the types we handle
1369 below will not be affected by the override anyway, and we need
1370 the original type unaltered. The override will be applied in
1371 bidi_resolve_weak. */
1372 type
= bidi_get_type (curchar
, NEUTRAL_DIR
);
1373 bidi_it
->orig_type
= type
;
1374 bidi_check_type (bidi_it
->orig_type
);
1377 bidi_it
->prev_was_pdf
= 0;
1379 bidi_it
->type_after_w1
= UNKNOWN_BT
;
1385 bidi_it
->type_after_w1
= type
;
1386 bidi_check_type (bidi_it
->type_after_w1
);
1387 type
= WEAK_BN
; /* X9/Retaining */
1388 if (bidi_it
->ignore_bn_limit
<= -1)
1390 if (current_level
<= BIDI_MAXLEVEL
- 4)
1392 /* Compute the least odd embedding level greater than
1393 the current level. */
1394 new_level
= ((current_level
+ 1) & ~1) + 1;
1395 if (bidi_it
->type_after_w1
== RLE
)
1396 override
= NEUTRAL_DIR
;
1399 if (current_level
== BIDI_MAXLEVEL
- 4)
1400 bidi_it
->invalid_rl_levels
= 0;
1401 bidi_push_embedding_level (bidi_it
, new_level
, override
);
1405 bidi_it
->invalid_levels
++;
1406 /* See the commentary about invalid_rl_levels below. */
1407 if (bidi_it
->invalid_rl_levels
< 0)
1408 bidi_it
->invalid_rl_levels
= 0;
1409 bidi_it
->invalid_rl_levels
++;
1412 else if (bidi_it
->prev
.type_after_w1
== WEAK_EN
/* W5/Retaining */
1413 || bidi_it
->next_en_pos
> bidi_it
->charpos
)
1418 bidi_it
->type_after_w1
= type
;
1419 bidi_check_type (bidi_it
->type_after_w1
);
1420 type
= WEAK_BN
; /* X9/Retaining */
1421 if (bidi_it
->ignore_bn_limit
<= -1)
1423 if (current_level
<= BIDI_MAXLEVEL
- 5)
1425 /* Compute the least even embedding level greater than
1426 the current level. */
1427 new_level
= ((current_level
+ 2) & ~1);
1428 if (bidi_it
->type_after_w1
== LRE
)
1429 override
= NEUTRAL_DIR
;
1432 bidi_push_embedding_level (bidi_it
, new_level
, override
);
1436 bidi_it
->invalid_levels
++;
1437 /* invalid_rl_levels counts invalid levels encountered
1438 while the embedding level was already too high for
1439 LRE/LRO, but not for RLE/RLO. That is because
1440 there may be exactly one PDF which we should not
1441 ignore even though invalid_levels is non-zero.
1442 invalid_rl_levels helps to know what PDF is
1444 if (bidi_it
->invalid_rl_levels
>= 0)
1445 bidi_it
->invalid_rl_levels
++;
1448 else if (bidi_it
->prev
.type_after_w1
== WEAK_EN
/* W5/Retaining */
1449 || bidi_it
->next_en_pos
> bidi_it
->charpos
)
1453 bidi_it
->type_after_w1
= type
;
1454 bidi_check_type (bidi_it
->type_after_w1
);
1455 type
= WEAK_BN
; /* X9/Retaining */
1456 if (bidi_it
->ignore_bn_limit
<= -1)
1458 if (!bidi_it
->invalid_rl_levels
)
1460 new_level
= bidi_pop_embedding_level (bidi_it
);
1461 bidi_it
->invalid_rl_levels
= -1;
1462 if (bidi_it
->invalid_levels
)
1463 bidi_it
->invalid_levels
--;
1464 /* else nothing: UAX#9 says to ignore invalid PDFs */
1466 if (!bidi_it
->invalid_levels
)
1467 new_level
= bidi_pop_embedding_level (bidi_it
);
1470 bidi_it
->invalid_levels
--;
1471 bidi_it
->invalid_rl_levels
--;
1474 else if (bidi_it
->prev
.type_after_w1
== WEAK_EN
/* W5/Retaining */
1475 || bidi_it
->next_en_pos
> bidi_it
->charpos
)
1483 bidi_it
->type
= type
;
1484 bidi_check_type (bidi_it
->type
);
1489 /* Given an iterator state in BIDI_IT, advance one character position
1490 in the buffer/string to the next character (in the logical order),
1491 resolve any explicit embeddings and directional overrides, and
1492 return the embedding level of the character after resolving
1493 explicit directives and ignoring empty embeddings. */
1495 bidi_resolve_explicit (struct bidi_it
*bidi_it
)
1497 int prev_level
= bidi_it
->level_stack
[bidi_it
->stack_idx
].level
;
1498 int new_level
= bidi_resolve_explicit_1 (bidi_it
);
1499 EMACS_INT eob
= bidi_it
->string
.s
? bidi_it
->string
.schars
: ZV
;
1500 const unsigned char *s
= STRINGP (bidi_it
->string
.lstring
)
1501 ? SDATA (bidi_it
->string
.lstring
) : bidi_it
->string
.s
;
1503 if (prev_level
< new_level
1504 && bidi_it
->type
== WEAK_BN
1505 && bidi_it
->ignore_bn_limit
== -1 /* only if not already known */
1506 && bidi_it
->charpos
< eob
/* not already at EOB */
1507 && bidi_explicit_dir_char (bidi_char_at_pos (bidi_it
->bytepos
1508 + bidi_it
->ch_len
, s
,
1509 bidi_it
->string
.unibyte
)))
1511 /* Avoid pushing and popping embedding levels if the level run
1512 is empty, as this breaks level runs where it shouldn't.
1513 UAX#9 removes all the explicit embedding and override codes,
1514 so empty embeddings disappear without a trace. We need to
1515 behave as if we did the same. */
1516 struct bidi_it saved_it
;
1517 int level
= prev_level
;
1519 bidi_copy_it (&saved_it
, bidi_it
);
1521 while (bidi_explicit_dir_char (bidi_char_at_pos (bidi_it
->bytepos
1522 + bidi_it
->ch_len
, s
,
1523 bidi_it
->string
.unibyte
)))
1525 /* This advances to the next character, skipping any
1526 characters covered by display strings. */
1527 level
= bidi_resolve_explicit_1 (bidi_it
);
1528 /* If string.lstring was relocated inside bidi_resolve_explicit_1,
1529 a pointer to its data is no longer valid. */
1530 if (STRINGP (bidi_it
->string
.lstring
))
1531 s
= SDATA (bidi_it
->string
.lstring
);
1534 if (bidi_it
->nchars
<= 0)
1536 if (level
== prev_level
) /* empty embedding */
1537 saved_it
.ignore_bn_limit
= bidi_it
->charpos
+ bidi_it
->nchars
;
1538 else /* this embedding is non-empty */
1539 saved_it
.ignore_bn_limit
= -2;
1541 bidi_copy_it (bidi_it
, &saved_it
);
1542 if (bidi_it
->ignore_bn_limit
> -1)
1544 /* We pushed a level, but we shouldn't have. Undo that. */
1545 if (!bidi_it
->invalid_rl_levels
)
1547 new_level
= bidi_pop_embedding_level (bidi_it
);
1548 bidi_it
->invalid_rl_levels
= -1;
1549 if (bidi_it
->invalid_levels
)
1550 bidi_it
->invalid_levels
--;
1552 if (!bidi_it
->invalid_levels
)
1553 new_level
= bidi_pop_embedding_level (bidi_it
);
1556 bidi_it
->invalid_levels
--;
1557 bidi_it
->invalid_rl_levels
--;
1562 if (bidi_it
->type
== NEUTRAL_B
) /* X8 */
1564 bidi_set_paragraph_end (bidi_it
);
1565 /* This is needed by bidi_resolve_weak below, and in L1. */
1566 bidi_it
->type_after_w1
= bidi_it
->type
;
1567 bidi_check_type (bidi_it
->type_after_w1
);
1573 /* Advance in the buffer/string, resolve weak types and return the
1574 type of the next character after weak type resolution. */
1576 bidi_resolve_weak (struct bidi_it
*bidi_it
)
1579 bidi_dir_t override
;
1580 int prev_level
= bidi_it
->level_stack
[bidi_it
->stack_idx
].level
;
1581 int new_level
= bidi_resolve_explicit (bidi_it
);
1583 bidi_type_t type_of_next
;
1584 struct bidi_it saved_it
;
1586 (STRINGP (bidi_it
->string
.lstring
) || bidi_it
->string
.s
)
1587 ? bidi_it
->string
.schars
: ZV
;
1589 type
= bidi_it
->type
;
1590 override
= bidi_it
->level_stack
[bidi_it
->stack_idx
].override
;
1592 if (type
== UNKNOWN_BT
1600 if (new_level
!= prev_level
1601 || bidi_it
->type
== NEUTRAL_B
)
1603 /* We've got a new embedding level run, compute the directional
1604 type of sor and initialize per-run variables (UAX#9, clause
1606 bidi_set_sor_type (bidi_it
, prev_level
, new_level
);
1608 else if (type
== NEUTRAL_S
|| type
== NEUTRAL_WS
1609 || type
== WEAK_BN
|| type
== STRONG_AL
)
1610 bidi_it
->type_after_w1
= type
; /* needed in L1 */
1611 bidi_check_type (bidi_it
->type_after_w1
);
1613 /* Level and directional override status are already recorded in
1614 bidi_it, and do not need any change; see X6. */
1615 if (override
== R2L
) /* X6 */
1617 else if (override
== L2R
)
1621 if (type
== WEAK_NSM
) /* W1 */
1623 /* Note that we don't need to consider the case where the
1624 prev character has its type overridden by an RLO or LRO,
1625 because then either the type of this NSM would have been
1626 also overridden, or the previous character is outside the
1627 current level run, and thus not relevant to this NSM.
1628 This is why NSM gets the type_after_w1 of the previous
1630 if (bidi_it
->prev
.type_after_w1
!= UNKNOWN_BT
1631 /* if type_after_w1 is NEUTRAL_B, this NSM is at sor */
1632 && bidi_it
->prev
.type_after_w1
!= NEUTRAL_B
)
1633 type
= bidi_it
->prev
.type_after_w1
;
1634 else if (bidi_it
->sor
== R2L
)
1636 else if (bidi_it
->sor
== L2R
)
1638 else /* shouldn't happen! */
1641 if (type
== WEAK_EN
/* W2 */
1642 && bidi_it
->last_strong
.type_after_w1
== STRONG_AL
)
1644 else if (type
== STRONG_AL
) /* W3 */
1646 else if ((type
== WEAK_ES
/* W4 */
1647 && bidi_it
->prev
.type_after_w1
== WEAK_EN
1648 && bidi_it
->prev
.orig_type
== WEAK_EN
)
1650 && ((bidi_it
->prev
.type_after_w1
== WEAK_EN
1651 && bidi_it
->prev
.orig_type
== WEAK_EN
)
1652 || bidi_it
->prev
.type_after_w1
== WEAK_AN
)))
1654 const unsigned char *s
=
1655 STRINGP (bidi_it
->string
.lstring
)
1656 ? SDATA (bidi_it
->string
.lstring
) : bidi_it
->string
.s
;
1659 bidi_it
->charpos
+ bidi_it
->nchars
>= eob
1661 : bidi_char_at_pos (bidi_it
->bytepos
+ bidi_it
->ch_len
, s
,
1662 bidi_it
->string
.unibyte
);
1663 type_of_next
= bidi_get_type (next_char
, override
);
1665 if (type_of_next
== WEAK_BN
1666 || bidi_explicit_dir_char (next_char
))
1668 bidi_copy_it (&saved_it
, bidi_it
);
1669 while (bidi_resolve_explicit (bidi_it
) == new_level
1670 && bidi_it
->type
== WEAK_BN
)
1672 type_of_next
= bidi_it
->type
;
1673 bidi_copy_it (bidi_it
, &saved_it
);
1676 /* If the next character is EN, but the last strong-type
1677 character is AL, that next EN will be changed to AN when
1678 we process it in W2 above. So in that case, this ES
1679 should not be changed into EN. */
1681 && type_of_next
== WEAK_EN
1682 && bidi_it
->last_strong
.type_after_w1
!= STRONG_AL
)
1684 else if (type
== WEAK_CS
)
1686 if (bidi_it
->prev
.type_after_w1
== WEAK_AN
1687 && (type_of_next
== WEAK_AN
1688 /* If the next character is EN, but the last
1689 strong-type character is AL, EN will be later
1690 changed to AN when we process it in W2 above.
1691 So in that case, this ES should not be
1693 || (type_of_next
== WEAK_EN
1694 && bidi_it
->last_strong
.type_after_w1
== STRONG_AL
)))
1696 else if (bidi_it
->prev
.type_after_w1
== WEAK_EN
1697 && type_of_next
== WEAK_EN
1698 && bidi_it
->last_strong
.type_after_w1
!= STRONG_AL
)
1702 else if (type
== WEAK_ET
/* W5: ET with EN before or after it */
1703 || type
== WEAK_BN
) /* W5/Retaining */
1705 if (bidi_it
->prev
.type_after_w1
== WEAK_EN
/* ET/BN w/EN before it */
1706 || bidi_it
->next_en_pos
> bidi_it
->charpos
)
1708 else /* W5: ET/BN with EN after it. */
1710 EMACS_INT en_pos
= bidi_it
->charpos
+ bidi_it
->nchars
;
1711 const unsigned char *s
=
1712 STRINGP (bidi_it
->string
.lstring
)
1713 ? SDATA (bidi_it
->string
.lstring
) : bidi_it
->string
.s
;
1715 if (bidi_it
->nchars
<= 0)
1718 bidi_it
->charpos
+ bidi_it
->nchars
>= eob
1720 : bidi_char_at_pos (bidi_it
->bytepos
+ bidi_it
->ch_len
, s
,
1721 bidi_it
->string
.unibyte
);
1722 type_of_next
= bidi_get_type (next_char
, override
);
1724 if (type_of_next
== WEAK_ET
1725 || type_of_next
== WEAK_BN
1726 || bidi_explicit_dir_char (next_char
))
1728 bidi_copy_it (&saved_it
, bidi_it
);
1729 while (bidi_resolve_explicit (bidi_it
) == new_level
1730 && (bidi_it
->type
== WEAK_BN
1731 || bidi_it
->type
== WEAK_ET
))
1733 type_of_next
= bidi_it
->type
;
1734 en_pos
= bidi_it
->charpos
;
1735 bidi_copy_it (bidi_it
, &saved_it
);
1737 if (type_of_next
== WEAK_EN
)
1739 /* If the last strong character is AL, the EN we've
1740 found will become AN when we get to it (W2). */
1741 if (bidi_it
->last_strong
.type_after_w1
!= STRONG_AL
)
1744 /* Remember this EN position, to speed up processing
1746 bidi_it
->next_en_pos
= en_pos
;
1748 else if (type
== WEAK_BN
)
1749 type
= NEUTRAL_ON
; /* W6/Retaining */
1755 if (type
== WEAK_ES
|| type
== WEAK_ET
|| type
== WEAK_CS
/* W6 */
1757 && (bidi_it
->prev
.type_after_w1
== WEAK_CS
/* W6/Retaining */
1758 || bidi_it
->prev
.type_after_w1
== WEAK_ES
1759 || bidi_it
->prev
.type_after_w1
== WEAK_ET
)))
1762 /* Store the type we've got so far, before we clobber it with strong
1763 types in W7 and while resolving neutral types. But leave alone
1764 the original types that were recorded above, because we will need
1765 them for the L1 clause. */
1766 if (bidi_it
->type_after_w1
== UNKNOWN_BT
)
1767 bidi_it
->type_after_w1
= type
;
1768 bidi_check_type (bidi_it
->type_after_w1
);
1770 if (type
== WEAK_EN
) /* W7 */
1772 if ((bidi_it
->last_strong
.type_after_w1
== STRONG_L
)
1773 || (bidi_it
->last_strong
.type
== UNKNOWN_BT
&& bidi_it
->sor
== L2R
))
1777 bidi_it
->type
= type
;
1778 bidi_check_type (bidi_it
->type
);
1782 /* Resolve the type of a neutral character according to the type of
1783 surrounding strong text and the current embedding level. */
1784 static inline bidi_type_t
1785 bidi_resolve_neutral_1 (bidi_type_t prev_type
, bidi_type_t next_type
, int lev
)
1787 /* N1: European and Arabic numbers are treated as though they were R. */
1788 if (next_type
== WEAK_EN
|| next_type
== WEAK_AN
)
1789 next_type
= STRONG_R
;
1790 if (prev_type
== WEAK_EN
|| prev_type
== WEAK_AN
)
1791 prev_type
= STRONG_R
;
1793 if (next_type
== prev_type
) /* N1 */
1795 else if ((lev
& 1) == 0) /* N2 */
1802 bidi_resolve_neutral (struct bidi_it
*bidi_it
)
1804 int prev_level
= bidi_it
->level_stack
[bidi_it
->stack_idx
].level
;
1805 bidi_type_t type
= bidi_resolve_weak (bidi_it
);
1806 int current_level
= bidi_it
->level_stack
[bidi_it
->stack_idx
].level
;
1808 if (!(type
== STRONG_R
1813 || type
== NEUTRAL_B
1814 || type
== NEUTRAL_S
1815 || type
== NEUTRAL_WS
1816 || type
== NEUTRAL_ON
))
1819 if (bidi_get_category (type
) == NEUTRAL
1820 || (type
== WEAK_BN
&& prev_level
== current_level
))
1822 if (bidi_it
->next_for_neutral
.type
!= UNKNOWN_BT
)
1823 type
= bidi_resolve_neutral_1 (bidi_it
->prev_for_neutral
.type
,
1824 bidi_it
->next_for_neutral
.type
,
1828 /* Arrrgh!! The UAX#9 algorithm is too deeply entrenched in
1829 the assumption of batch-style processing; see clauses W4,
1830 W5, and especially N1, which require to look far forward
1831 (as well as back) in the buffer/string. May the fleas of
1832 a thousand camels infest the armpits of those who design
1833 supposedly general-purpose algorithms by looking at their
1834 own implementations, and fail to consider other possible
1836 struct bidi_it saved_it
;
1837 bidi_type_t next_type
;
1839 if (bidi_it
->scan_dir
== -1)
1842 bidi_copy_it (&saved_it
, bidi_it
);
1843 /* Scan the text forward until we find the first non-neutral
1844 character, and then use that to resolve the neutral we
1845 are dealing with now. We also cache the scanned iterator
1846 states, to salvage some of the effort later. */
1847 bidi_cache_iterator_state (bidi_it
, 0);
1849 /* Record the info about the previous character, so that
1850 it will be cached below with this state. */
1851 if (bidi_it
->type_after_w1
!= WEAK_BN
/* W1/Retaining */
1852 && bidi_it
->type
!= WEAK_BN
)
1853 bidi_remember_char (&bidi_it
->prev
, bidi_it
);
1854 type
= bidi_resolve_weak (bidi_it
);
1855 /* Paragraph separators have their levels fully resolved
1856 at this point, so cache them as resolved. */
1857 bidi_cache_iterator_state (bidi_it
, type
== NEUTRAL_B
);
1858 /* FIXME: implement L1 here, by testing for a newline and
1859 resetting the level for any sequence of whitespace
1860 characters adjacent to it. */
1861 } while (!(type
== NEUTRAL_B
1863 && bidi_get_category (type
) != NEUTRAL
)
1864 /* This is all per level run, so stop when we
1865 reach the end of this level run. */
1866 || bidi_it
->level_stack
[bidi_it
->stack_idx
].level
!=
1869 bidi_remember_char (&saved_it
.next_for_neutral
, bidi_it
);
1880 /* N1: ``European and Arabic numbers are treated as
1881 though they were R.'' */
1882 next_type
= STRONG_R
;
1883 saved_it
.next_for_neutral
.type
= STRONG_R
;
1886 if (!bidi_explicit_dir_char (bidi_it
->ch
))
1887 abort (); /* can't happen: BNs are skipped */
1890 /* Marched all the way to the end of this level run.
1891 We need to use the eor type, whose information is
1892 stored by bidi_set_sor_type in the prev_for_neutral
1894 if (saved_it
.type
!= WEAK_BN
1895 || bidi_get_category (bidi_it
->prev
.type_after_w1
) == NEUTRAL
)
1897 next_type
= bidi_it
->prev_for_neutral
.type
;
1898 saved_it
.next_for_neutral
.type
= next_type
;
1899 bidi_check_type (next_type
);
1903 /* This is a BN which does not adjoin neutrals.
1904 Leave its type alone. */
1905 bidi_copy_it (bidi_it
, &saved_it
);
1906 return bidi_it
->type
;
1912 type
= bidi_resolve_neutral_1 (saved_it
.prev_for_neutral
.type
,
1913 next_type
, current_level
);
1914 saved_it
.type
= type
;
1915 bidi_check_type (type
);
1916 bidi_copy_it (bidi_it
, &saved_it
);
1922 /* Given an iterator state in BIDI_IT, advance one character position
1923 in the buffer/string to the next character (in the logical order),
1924 resolve the bidi type of that next character, and return that
1927 bidi_type_of_next_char (struct bidi_it
*bidi_it
)
1931 /* This should always be called during a forward scan. */
1932 if (bidi_it
->scan_dir
!= 1)
1935 /* Reset the limit until which to ignore BNs if we step out of the
1936 area where we found only empty levels. */
1937 if ((bidi_it
->ignore_bn_limit
> -1
1938 && bidi_it
->ignore_bn_limit
<= bidi_it
->charpos
)
1939 || (bidi_it
->ignore_bn_limit
== -2
1940 && !bidi_explicit_dir_char (bidi_it
->ch
)))
1941 bidi_it
->ignore_bn_limit
= -1;
1943 type
= bidi_resolve_neutral (bidi_it
);
1948 /* Given an iterator state BIDI_IT, advance one character position in
1949 the buffer/string to the next character (in the current scan
1950 direction), resolve the embedding and implicit levels of that next
1951 character, and return the resulting level. */
1953 bidi_level_of_next_char (struct bidi_it
*bidi_it
)
1956 int level
, prev_level
= -1;
1957 struct bidi_saved_info next_for_neutral
;
1958 EMACS_INT next_char_pos
= -2;
1960 if (bidi_it
->scan_dir
== 1)
1963 (bidi_it
->string
.s
|| STRINGP (bidi_it
->string
.lstring
))
1964 ? bidi_it
->string
.schars
: ZV
;
1966 /* There's no sense in trying to advance if we hit end of text. */
1967 if (bidi_it
->charpos
>= eob
)
1968 return bidi_it
->resolved_level
;
1970 /* Record the info about the previous character. */
1971 if (bidi_it
->type_after_w1
!= WEAK_BN
/* W1/Retaining */
1972 && bidi_it
->type
!= WEAK_BN
)
1973 bidi_remember_char (&bidi_it
->prev
, bidi_it
);
1974 if (bidi_it
->type_after_w1
== STRONG_R
1975 || bidi_it
->type_after_w1
== STRONG_L
1976 || bidi_it
->type_after_w1
== STRONG_AL
)
1977 bidi_remember_char (&bidi_it
->last_strong
, bidi_it
);
1978 /* FIXME: it sounds like we don't need both prev and
1979 prev_for_neutral members, but I'm leaving them both for now. */
1980 if (bidi_it
->type
== STRONG_R
|| bidi_it
->type
== STRONG_L
1981 || bidi_it
->type
== WEAK_EN
|| bidi_it
->type
== WEAK_AN
)
1982 bidi_remember_char (&bidi_it
->prev_for_neutral
, bidi_it
);
1984 /* If we overstepped the characters used for resolving neutrals
1985 and whitespace, invalidate their info in the iterator. */
1986 if (bidi_it
->charpos
>= bidi_it
->next_for_neutral
.charpos
)
1987 bidi_it
->next_for_neutral
.type
= UNKNOWN_BT
;
1988 if (bidi_it
->next_en_pos
>= 0
1989 && bidi_it
->charpos
>= bidi_it
->next_en_pos
)
1990 bidi_it
->next_en_pos
= -1;
1991 if (bidi_it
->next_for_ws
.type
!= UNKNOWN_BT
1992 && bidi_it
->charpos
>= bidi_it
->next_for_ws
.charpos
)
1993 bidi_it
->next_for_ws
.type
= UNKNOWN_BT
;
1995 /* This must be taken before we fill the iterator with the info
1996 about the next char. If we scan backwards, the iterator
1997 state must be already cached, so there's no need to know the
1998 embedding level of the previous character, since we will be
1999 returning to our caller shortly. */
2000 prev_level
= bidi_it
->level_stack
[bidi_it
->stack_idx
].level
;
2002 next_for_neutral
= bidi_it
->next_for_neutral
;
2004 /* Perhaps the character we want is already cached. If it is, the
2005 call to bidi_cache_find below will return a type other than
2007 if (bidi_cache_idx
> bidi_cache_start
&& !bidi_it
->first_elt
)
2010 (bidi_it
->string
.s
|| STRINGP (bidi_it
->string
.lstring
)) ? 0 : 1;
2012 if (bidi_it
->scan_dir
> 0)
2014 if (bidi_it
->nchars
<= 0)
2016 next_char_pos
= bidi_it
->charpos
+ bidi_it
->nchars
;
2018 else if (bidi_it
->charpos
>= bob
)
2019 /* Implementation note: we allow next_char_pos to be as low as
2020 0 for buffers or -1 for strings, and that is okay because
2021 that's the "position" of the sentinel iterator state we
2022 cached at the beginning of the iteration. */
2023 next_char_pos
= bidi_it
->charpos
- 1;
2024 if (next_char_pos
>= bob
- 1)
2025 type
= bidi_cache_find (next_char_pos
, -1, bidi_it
);
2031 if (type
!= UNKNOWN_BT
)
2033 /* Don't lose the information for resolving neutrals! The
2034 cached states could have been cached before their
2035 next_for_neutral member was computed. If we are on our way
2036 forward, we can simply take the info from the previous
2038 if (bidi_it
->scan_dir
== 1
2039 && bidi_it
->next_for_neutral
.type
== UNKNOWN_BT
)
2040 bidi_it
->next_for_neutral
= next_for_neutral
;
2042 /* If resolved_level is -1, it means this state was cached
2043 before it was completely resolved, so we cannot return
2045 if (bidi_it
->resolved_level
!= -1)
2046 return bidi_it
->resolved_level
;
2048 if (bidi_it
->scan_dir
== -1)
2049 /* If we are going backwards, the iterator state is already cached
2050 from previous scans, and should be fully resolved. */
2053 if (type
== UNKNOWN_BT
)
2054 type
= bidi_type_of_next_char (bidi_it
);
2056 if (type
== NEUTRAL_B
)
2057 return bidi_it
->resolved_level
;
2059 level
= bidi_it
->level_stack
[bidi_it
->stack_idx
].level
;
2060 if ((bidi_get_category (type
) == NEUTRAL
/* && type != NEUTRAL_B */)
2061 || (type
== WEAK_BN
&& prev_level
== level
))
2063 if (bidi_it
->next_for_neutral
.type
== UNKNOWN_BT
)
2066 /* If the cached state shows a neutral character, it was not
2067 resolved by bidi_resolve_neutral, so do it now. */
2068 type
= bidi_resolve_neutral_1 (bidi_it
->prev_for_neutral
.type
,
2069 bidi_it
->next_for_neutral
.type
,
2073 if (!(type
== STRONG_R
2077 || type
== WEAK_AN
))
2079 bidi_it
->type
= type
;
2080 bidi_check_type (bidi_it
->type
);
2082 /* For L1 below, we need to know, for each WS character, whether
2083 it belongs to a sequence of WS characters preceding a newline
2084 or a TAB or a paragraph separator. */
2085 if (bidi_it
->orig_type
== NEUTRAL_WS
2086 && bidi_it
->next_for_ws
.type
== UNKNOWN_BT
)
2089 EMACS_INT clen
= bidi_it
->ch_len
;
2090 EMACS_INT bpos
= bidi_it
->bytepos
;
2091 EMACS_INT cpos
= bidi_it
->charpos
;
2092 EMACS_INT disp_pos
= bidi_it
->disp_pos
;
2093 EMACS_INT nc
= bidi_it
->nchars
;
2094 struct bidi_string_data bs
= bidi_it
->string
;
2096 int fwp
= bidi_it
->frame_window_p
;
2097 int dpp
= bidi_it
->disp_prop
;
2099 if (bidi_it
->nchars
<= 0)
2102 ch
= bidi_fetch_char (bpos
+= clen
, cpos
+= nc
, &disp_pos
, &dpp
, &bs
,
2104 if (ch
== '\n' || ch
== BIDI_EOB
/* || ch == LINESEP_CHAR */)
2107 chtype
= bidi_get_type (ch
, NEUTRAL_DIR
);
2108 } while (chtype
== NEUTRAL_WS
|| chtype
== WEAK_BN
2109 || bidi_explicit_dir_char (ch
)); /* L1/Retaining */
2110 bidi_it
->next_for_ws
.type
= chtype
;
2111 bidi_check_type (bidi_it
->next_for_ws
.type
);
2112 bidi_it
->next_for_ws
.charpos
= cpos
;
2113 bidi_it
->next_for_ws
.bytepos
= bpos
;
2116 /* Resolve implicit levels, with a twist: PDFs get the embedding
2117 level of the enbedding they terminate. See below for the
2119 if (bidi_it
->orig_type
== PDF
2120 /* Don't do this if this formatting code didn't change the
2121 embedding level due to invalid or empty embeddings. */
2122 && prev_level
!= level
)
2124 /* Don't look in UAX#9 for the reason for this: it's our own
2125 private quirk. The reason is that we want the formatting
2126 codes to be delivered so that they bracket the text of their
2127 embedding. For example, given the text
2131 we want it to be displayed as
2139 which will result because we bump up the embedding level as
2140 soon as we see the RLO and pop it as soon as we see the PDF,
2141 so RLO itself has the same embedding level as "teST", and
2142 thus would be normally delivered last, just before the PDF.
2143 The switch below fiddles with the level of PDF so that this
2144 ugly side effect does not happen.
2146 (This is, of course, only important if the formatting codes
2147 are actually displayed, but Emacs does need to display them
2148 if the user wants to.) */
2151 else if (bidi_it
->orig_type
== NEUTRAL_B
/* L1 */
2152 || bidi_it
->orig_type
== NEUTRAL_S
2153 || bidi_it
->ch
== '\n' || bidi_it
->ch
== BIDI_EOB
2154 /* || bidi_it->ch == LINESEP_CHAR */
2155 || (bidi_it
->orig_type
== NEUTRAL_WS
2156 && (bidi_it
->next_for_ws
.type
== NEUTRAL_B
2157 || bidi_it
->next_for_ws
.type
== NEUTRAL_S
)))
2158 level
= bidi_it
->level_stack
[0].level
;
2159 else if ((level
& 1) == 0) /* I1 */
2161 if (type
== STRONG_R
)
2163 else if (type
== WEAK_EN
|| type
== WEAK_AN
)
2168 if (type
== STRONG_L
|| type
== WEAK_EN
|| type
== WEAK_AN
)
2172 bidi_it
->resolved_level
= level
;
2176 /* Move to the other edge of a level given by LEVEL. If END_FLAG is
2177 non-zero, we are at the end of a level, and we need to prepare to
2178 resume the scan of the lower level.
2180 If this level's other edge is cached, we simply jump to it, filling
2181 the iterator structure with the iterator state on the other edge.
2182 Otherwise, we walk the buffer or string until we come back to the
2183 same level as LEVEL.
2185 Note: we are not talking here about a ``level run'' in the UAX#9
2186 sense of the term, but rather about a ``level'' which includes
2187 all the levels higher than it. In other words, given the levels
2190 11111112222222333333334443343222222111111112223322111
2193 and assuming we are at point A scanning left to right, this
2194 function moves to point C, whereas the UAX#9 ``level 2 run'' ends
2197 bidi_find_other_level_edge (struct bidi_it
*bidi_it
, int level
, int end_flag
)
2199 int dir
= end_flag
? -bidi_it
->scan_dir
: bidi_it
->scan_dir
;
2202 /* Try the cache first. */
2203 if ((idx
= bidi_cache_find_level_change (level
, dir
, end_flag
))
2204 >= bidi_cache_start
)
2205 bidi_cache_fetch_state (idx
, bidi_it
);
2211 abort (); /* if we are at end of level, its edges must be cached */
2213 bidi_cache_iterator_state (bidi_it
, 1);
2215 new_level
= bidi_level_of_next_char (bidi_it
);
2216 bidi_cache_iterator_state (bidi_it
, 1);
2217 } while (new_level
>= level
);
2222 bidi_move_to_visually_next (struct bidi_it
*bidi_it
)
2224 int old_level
, new_level
, next_level
;
2225 struct bidi_it sentinel
;
2226 struct gcpro gcpro1
;
2228 if (bidi_it
->charpos
< 0 || bidi_it
->bytepos
< 0)
2231 if (bidi_it
->scan_dir
== 0)
2233 bidi_it
->scan_dir
= 1; /* default to logical order */
2236 /* The code below can call eval, and thus cause GC. If we are
2237 iterating a Lisp string, make sure it won't be GCed. */
2238 if (STRINGP (bidi_it
->string
.lstring
))
2239 GCPRO1 (bidi_it
->string
.lstring
);
2241 /* If we just passed a newline, initialize for the next line. */
2242 if (!bidi_it
->first_elt
&& bidi_it
->orig_type
== NEUTRAL_B
)
2243 bidi_line_init (bidi_it
);
2245 /* Prepare the sentinel iterator state, and cache it. When we bump
2246 into it, scanning backwards, we'll know that the last non-base
2247 level is exhausted. */
2248 if (bidi_cache_idx
== bidi_cache_start
)
2250 bidi_copy_it (&sentinel
, bidi_it
);
2251 if (bidi_it
->first_elt
)
2253 sentinel
.charpos
--; /* cached charpos needs to be monotonic */
2255 sentinel
.ch
= '\n'; /* doesn't matter, but why not? */
2256 sentinel
.ch_len
= 1;
2257 sentinel
.nchars
= 1;
2259 bidi_cache_iterator_state (&sentinel
, 1);
2262 old_level
= bidi_it
->resolved_level
;
2263 new_level
= bidi_level_of_next_char (bidi_it
);
2265 /* Reordering of resolved levels (clause L2) is implemented by
2266 jumping to the other edge of the level and flipping direction of
2267 scanning the text whenever we find a level change. */
2268 if (new_level
!= old_level
)
2270 int ascending
= new_level
> old_level
;
2271 int level_to_search
= ascending
? old_level
+ 1 : old_level
;
2272 int incr
= ascending
? 1 : -1;
2273 int expected_next_level
= old_level
+ incr
;
2275 /* Jump (or walk) to the other edge of this level. */
2276 bidi_find_other_level_edge (bidi_it
, level_to_search
, !ascending
);
2277 /* Switch scan direction and peek at the next character in the
2279 bidi_it
->scan_dir
= -bidi_it
->scan_dir
;
2281 /* The following loop handles the case where the resolved level
2282 jumps by more than one. This is typical for numbers inside a
2283 run of text with left-to-right embedding direction, but can
2284 also happen in other situations. In those cases the decision
2285 where to continue after a level change, and in what direction,
2286 is tricky. For example, given a text like below:
2291 (where the numbers below the text show the resolved levels),
2292 the result of reordering according to UAX#9 should be this:
2296 This is implemented by the loop below which flips direction
2297 and jumps to the other edge of the level each time it finds
2298 the new level not to be the expected one. The expected level
2299 is always one more or one less than the previous one. */
2300 next_level
= bidi_peek_at_next_level (bidi_it
);
2301 while (next_level
!= expected_next_level
)
2303 expected_next_level
+= incr
;
2304 level_to_search
+= incr
;
2305 bidi_find_other_level_edge (bidi_it
, level_to_search
, !ascending
);
2306 bidi_it
->scan_dir
= -bidi_it
->scan_dir
;
2307 next_level
= bidi_peek_at_next_level (bidi_it
);
2310 /* Finally, deliver the next character in the new direction. */
2311 next_level
= bidi_level_of_next_char (bidi_it
);
2314 /* Take note when we have just processed the newline that precedes
2315 the end of the paragraph. The next time we are about to be
2316 called, set_iterator_to_next will automatically reinit the
2317 paragraph direction, if needed. We do this at the newline before
2318 the paragraph separator, because the next character might not be
2319 the first character of the next paragraph, due to the bidi
2320 reordering, whereas we _must_ know the paragraph base direction
2321 _before_ we process the paragraph's text, since the base
2322 direction affects the reordering. */
2323 if (bidi_it
->scan_dir
== 1 && bidi_it
->orig_type
== NEUTRAL_B
)
2325 /* The paragraph direction of the entire string, once
2326 determined, is in effect for the entire string. Setting the
2327 separator limit to the end of the string prevents
2328 bidi_paragraph_init from being called automatically on this
2330 if (bidi_it
->string
.s
|| STRINGP (bidi_it
->string
.lstring
))
2331 bidi_it
->separator_limit
= bidi_it
->string
.schars
;
2332 else if (bidi_it
->bytepos
< ZV_BYTE
)
2335 bidi_at_paragraph_end (bidi_it
->charpos
+ bidi_it
->nchars
,
2336 bidi_it
->bytepos
+ bidi_it
->ch_len
);
2337 if (bidi_it
->nchars
<= 0)
2341 bidi_it
->new_paragraph
= 1;
2342 /* Record the buffer position of the last character of the
2343 paragraph separator. */
2344 bidi_it
->separator_limit
=
2345 bidi_it
->charpos
+ bidi_it
->nchars
+ sep_len
;
2350 if (bidi_it
->scan_dir
== 1 && bidi_cache_idx
> bidi_cache_start
)
2352 /* If we are at paragraph's base embedding level and beyond the
2353 last cached position, the cache's job is done and we can
2355 if (bidi_it
->resolved_level
== bidi_it
->level_stack
[0].level
2356 && bidi_it
->charpos
> (bidi_cache
[bidi_cache_idx
- 1].charpos
2357 + bidi_cache
[bidi_cache_idx
- 1].nchars
- 1))
2358 bidi_cache_reset ();
2359 /* But as long as we are caching during forward scan, we must
2360 cache each state, or else the cache integrity will be
2361 compromised: it assumes cached states correspond to buffer
2364 bidi_cache_iterator_state (bidi_it
, 1);
2367 if (STRINGP (bidi_it
->string
.lstring
))
2371 /* This is meant to be called from within the debugger, whenever you
2372 wish to examine the cache contents. */
2373 void bidi_dump_cached_states (void) EXTERNALLY_VISIBLE
;
2375 bidi_dump_cached_states (void)
2380 if (bidi_cache_idx
== 0)
2382 fprintf (stderr
, "The cache is empty.\n");
2385 fprintf (stderr
, "Total of %"pD
"d state%s in cache:\n",
2386 bidi_cache_idx
, bidi_cache_idx
== 1 ? "" : "s");
2388 for (i
= bidi_cache
[bidi_cache_idx
- 1].charpos
; i
> 0; i
/= 10)
2390 fputs ("ch ", stderr
);
2391 for (i
= 0; i
< bidi_cache_idx
; i
++)
2392 fprintf (stderr
, "%*c", ndigits
, bidi_cache
[i
].ch
);
2393 fputs ("\n", stderr
);
2394 fputs ("lvl ", stderr
);
2395 for (i
= 0; i
< bidi_cache_idx
; i
++)
2396 fprintf (stderr
, "%*d", ndigits
, bidi_cache
[i
].resolved_level
);
2397 fputs ("\n", stderr
);
2398 fputs ("pos ", stderr
);
2399 for (i
= 0; i
< bidi_cache_idx
; i
++)
2400 fprintf (stderr
, "%*"pI
"d", ndigits
, bidi_cache
[i
].charpos
);
2401 fputs ("\n", stderr
);