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 extern int bidi_ignore_explicit_marks_for_paragraph_level EXTERNALLY_VISIBLE
;
83 int bidi_ignore_explicit_marks_for_paragraph_level
= 1;
85 static Lisp_Object paragraph_start_re
, paragraph_separate_re
;
86 static Lisp_Object Qparagraph_start
, Qparagraph_separate
;
89 /***********************************************************************
91 ***********************************************************************/
93 /* Return the bidi type of a character CH, subject to the current
94 directional OVERRIDE. */
95 static inline bidi_type_t
96 bidi_get_type (int ch
, bidi_dir_t override
)
98 bidi_type_t default_type
;
102 if (ch
< 0 || ch
> MAX_CHAR
)
105 default_type
= (bidi_type_t
) XINT (CHAR_TABLE_REF (bidi_type_table
, ch
));
107 if (override
== NEUTRAL_DIR
)
110 switch (default_type
)
112 /* Although UAX#9 does not tell, it doesn't make sense to
113 override NEUTRAL_B and LRM/RLM characters. */
128 if (override
== L2R
) /* X6 */
130 else if (override
== R2L
)
133 abort (); /* can't happen: handled above */
139 bidi_check_type (bidi_type_t type
)
141 if (type
< UNKNOWN_BT
|| type
> NEUTRAL_ON
)
145 /* Given a bidi TYPE of a character, return its category. */
146 static inline bidi_category_t
147 bidi_get_category (bidi_type_t type
)
161 case PDF
: /* ??? really?? */
180 /* Return the mirrored character of C, if it has one. If C has no
181 mirrored counterpart, return C.
182 Note: The conditions in UAX#9 clause L4 regarding the surrounding
183 context must be tested by the caller. */
185 bidi_mirror_char (int c
)
191 if (c
< 0 || c
> MAX_CHAR
)
194 val
= CHAR_TABLE_REF (bidi_mirror_table
, c
);
199 if (v
< 0 || v
> MAX_CHAR
)
208 /* Determine the start-of-run (sor) directional type given the two
209 embedding levels on either side of the run boundary. Also, update
210 the saved info about previously seen characters, since that info is
211 generally valid for a single level run. */
213 bidi_set_sor_type (struct bidi_it
*bidi_it
, int level_before
, int level_after
)
215 int higher_level
= level_before
> level_after
? level_before
: level_after
;
217 /* The prev_was_pdf gork is required for when we have several PDFs
218 in a row. In that case, we want to compute the sor type for the
219 next level run only once: when we see the first PDF. That's
220 because the sor type depends only on the higher of the two levels
221 that we find on the two sides of the level boundary (see UAX#9,
222 clause X10), and so we don't need to know the final embedding
223 level to which we descend after processing all the PDFs. */
224 if (!bidi_it
->prev_was_pdf
|| level_before
< level_after
)
225 /* FIXME: should the default sor direction be user selectable? */
226 bidi_it
->sor
= (higher_level
& 1) != 0 ? R2L
: L2R
;
227 if (level_before
> level_after
)
228 bidi_it
->prev_was_pdf
= 1;
230 bidi_it
->prev
.type
= UNKNOWN_BT
;
231 bidi_it
->last_strong
.type
= bidi_it
->last_strong
.type_after_w1
=
232 bidi_it
->last_strong
.orig_type
= UNKNOWN_BT
;
233 bidi_it
->prev_for_neutral
.type
= bidi_it
->sor
== R2L
? STRONG_R
: STRONG_L
;
234 bidi_it
->prev_for_neutral
.charpos
= bidi_it
->charpos
;
235 bidi_it
->prev_for_neutral
.bytepos
= bidi_it
->bytepos
;
236 bidi_it
->next_for_neutral
.type
= bidi_it
->next_for_neutral
.type_after_w1
=
237 bidi_it
->next_for_neutral
.orig_type
= UNKNOWN_BT
;
238 bidi_it
->ignore_bn_limit
= -1; /* meaning it's unknown */
241 /* Push the current embedding level and override status; reset the
242 current level to LEVEL and the current override status to OVERRIDE. */
244 bidi_push_embedding_level (struct bidi_it
*bidi_it
,
245 int level
, bidi_dir_t override
)
247 bidi_it
->stack_idx
++;
248 xassert (bidi_it
->stack_idx
< BIDI_MAXLEVEL
);
249 bidi_it
->level_stack
[bidi_it
->stack_idx
].level
= level
;
250 bidi_it
->level_stack
[bidi_it
->stack_idx
].override
= override
;
253 /* Pop the embedding level and directional override status from the
254 stack, and return the new level. */
256 bidi_pop_embedding_level (struct bidi_it
*bidi_it
)
258 /* UAX#9 says to ignore invalid PDFs. */
259 if (bidi_it
->stack_idx
> 0)
260 bidi_it
->stack_idx
--;
261 return bidi_it
->level_stack
[bidi_it
->stack_idx
].level
;
264 /* Record in SAVED_INFO the information about the current character. */
266 bidi_remember_char (struct bidi_saved_info
*saved_info
,
267 struct bidi_it
*bidi_it
)
269 saved_info
->charpos
= bidi_it
->charpos
;
270 saved_info
->bytepos
= bidi_it
->bytepos
;
271 saved_info
->type
= bidi_it
->type
;
272 bidi_check_type (bidi_it
->type
);
273 saved_info
->type_after_w1
= bidi_it
->type_after_w1
;
274 bidi_check_type (bidi_it
->type_after_w1
);
275 saved_info
->orig_type
= bidi_it
->orig_type
;
276 bidi_check_type (bidi_it
->orig_type
);
279 /* Copy the bidi iterator from FROM to TO. To save cycles, this only
280 copies the part of the level stack that is actually in use. */
282 bidi_copy_it (struct bidi_it
*to
, struct bidi_it
*from
)
286 /* Copy everything except the level stack and beyond. */
287 memcpy (to
, from
, offsetof (struct bidi_it
, level_stack
[0]));
289 /* Copy the active part of the level stack. */
290 to
->level_stack
[0] = from
->level_stack
[0]; /* level zero is always in use */
291 for (i
= 1; i
<= from
->stack_idx
; i
++)
292 to
->level_stack
[i
] = from
->level_stack
[i
];
296 /***********************************************************************
297 Caching the bidi iterator states
298 ***********************************************************************/
300 #define BIDI_CACHE_CHUNK 200
301 static struct bidi_it
*bidi_cache
;
302 static ptrdiff_t bidi_cache_size
= 0;
303 enum { elsz
= sizeof (struct bidi_it
) };
304 static ptrdiff_t bidi_cache_idx
; /* next unused cache slot */
305 static ptrdiff_t bidi_cache_last_idx
; /* slot of last cache hit */
306 static ptrdiff_t bidi_cache_start
= 0; /* start of cache for this
309 /* Reset the cache state to the empty state. We only reset the part
310 of the cache relevant to iteration of the current object. Previous
311 objects, which are pushed on the display iterator's stack, are left
312 intact. This is called when the cached information is no more
313 useful for the current iteration, e.g. when we were reseated to a
314 new position on the same object. */
316 bidi_cache_reset (void)
318 bidi_cache_idx
= bidi_cache_start
;
319 bidi_cache_last_idx
= -1;
322 /* Shrink the cache to its minimal size. Called when we init the bidi
323 iterator for reordering a buffer or a string that does not come
324 from display properties, because that means all the previously
325 cached info is of no further use. */
327 bidi_cache_shrink (void)
329 if (bidi_cache_size
> BIDI_CACHE_CHUNK
)
331 bidi_cache_size
= BIDI_CACHE_CHUNK
;
333 (struct bidi_it
*) xrealloc (bidi_cache
, bidi_cache_size
* elsz
);
339 bidi_cache_fetch_state (ptrdiff_t idx
, struct bidi_it
*bidi_it
)
341 int current_scan_dir
= bidi_it
->scan_dir
;
343 if (idx
< bidi_cache_start
|| idx
>= bidi_cache_idx
)
346 bidi_copy_it (bidi_it
, &bidi_cache
[idx
]);
347 bidi_it
->scan_dir
= current_scan_dir
;
348 bidi_cache_last_idx
= idx
;
351 /* Find a cached state with a given CHARPOS and resolved embedding
352 level less or equal to LEVEL. if LEVEL is -1, disregard the
353 resolved levels in cached states. DIR, if non-zero, means search
354 in that direction from the last cache hit. */
355 static inline ptrdiff_t
356 bidi_cache_search (EMACS_INT charpos
, int level
, int dir
)
358 ptrdiff_t i
, i_start
;
360 if (bidi_cache_idx
> bidi_cache_start
)
362 if (bidi_cache_last_idx
== -1)
363 bidi_cache_last_idx
= bidi_cache_idx
- 1;
364 if (charpos
< bidi_cache
[bidi_cache_last_idx
].charpos
)
367 i_start
= bidi_cache_last_idx
- 1;
369 else if (charpos
> (bidi_cache
[bidi_cache_last_idx
].charpos
370 + bidi_cache
[bidi_cache_last_idx
].nchars
- 1))
373 i_start
= bidi_cache_last_idx
+ 1;
376 i_start
= bidi_cache_last_idx
;
380 i_start
= bidi_cache_idx
- 1;
385 /* Linear search for now; FIXME! */
386 for (i
= i_start
; i
>= bidi_cache_start
; i
--)
387 if (bidi_cache
[i
].charpos
<= charpos
388 && charpos
< bidi_cache
[i
].charpos
+ bidi_cache
[i
].nchars
389 && (level
== -1 || bidi_cache
[i
].resolved_level
<= level
))
394 for (i
= i_start
; i
< bidi_cache_idx
; i
++)
395 if (bidi_cache
[i
].charpos
<= charpos
396 && charpos
< bidi_cache
[i
].charpos
+ bidi_cache
[i
].nchars
397 && (level
== -1 || bidi_cache
[i
].resolved_level
<= level
))
405 /* Find a cached state where the resolved level changes to a value
406 that is lower than LEVEL, and return its cache slot index. DIR is
407 the direction to search, starting with the last used cache slot.
408 If DIR is zero, we search backwards from the last occupied cache
409 slot. BEFORE, if non-zero, means return the index of the slot that
410 is ``before'' the level change in the search direction. That is,
411 given the cached levels like this:
416 and assuming we are at the position cached at the slot marked with
417 C, searching backwards (DIR = -1) for LEVEL = 2 will return the
418 index of slot B or A, depending whether BEFORE is, respectively,
421 bidi_cache_find_level_change (int level
, int dir
, int before
)
425 ptrdiff_t i
= dir
? bidi_cache_last_idx
: bidi_cache_idx
- 1;
426 int incr
= before
? 1 : 0;
428 xassert (!dir
|| bidi_cache_last_idx
>= 0);
437 while (i
>= bidi_cache_start
+ incr
)
439 if (bidi_cache
[i
- incr
].resolved_level
>= 0
440 && bidi_cache
[i
- incr
].resolved_level
< level
)
447 while (i
< bidi_cache_idx
- incr
)
449 if (bidi_cache
[i
+ incr
].resolved_level
>= 0
450 && bidi_cache
[i
+ incr
].resolved_level
< level
)
461 bidi_cache_ensure_space (ptrdiff_t idx
)
463 /* Enlarge the cache as needed. */
464 if (idx
>= bidi_cache_size
)
468 /* The bidi cache cannot be larger than the largest Lisp string
470 ptrdiff_t string_or_buffer_bound
=
471 max (BUF_BYTES_MAX
, STRING_BYTES_BOUND
);
473 /* Also, it cannot be larger than what C can represent. */
474 ptrdiff_t c_bound
= min (PTRDIFF_MAX
, SIZE_MAX
) / elsz
;
476 if (min (string_or_buffer_bound
, c_bound
) <= idx
)
477 memory_full (SIZE_MAX
);
478 new_size
= idx
- idx
% BIDI_CACHE_CHUNK
+ BIDI_CACHE_CHUNK
;
479 bidi_cache
= (struct bidi_it
*) xrealloc (bidi_cache
, new_size
* elsz
);
480 bidi_cache_size
= new_size
;
485 bidi_cache_iterator_state (struct bidi_it
*bidi_it
, int resolved
)
489 /* We should never cache on backward scans. */
490 if (bidi_it
->scan_dir
== -1)
492 idx
= bidi_cache_search (bidi_it
->charpos
, -1, 1);
496 idx
= bidi_cache_idx
;
497 bidi_cache_ensure_space (idx
);
498 /* Character positions should correspond to cache positions 1:1.
499 If we are outside the range of cached positions, the cache is
500 useless and must be reset. */
501 if (idx
> bidi_cache_start
&&
502 (bidi_it
->charpos
> (bidi_cache
[idx
- 1].charpos
503 + bidi_cache
[idx
- 1].nchars
)
504 || bidi_it
->charpos
< bidi_cache
[bidi_cache_start
].charpos
))
507 idx
= bidi_cache_start
;
509 if (bidi_it
->nchars
<= 0)
511 bidi_copy_it (&bidi_cache
[idx
], bidi_it
);
513 bidi_cache
[idx
].resolved_level
= -1;
517 /* Copy only the members which could have changed, to avoid
518 costly copying of the entire struct. */
519 bidi_cache
[idx
].type
= bidi_it
->type
;
520 bidi_check_type (bidi_it
->type
);
521 bidi_cache
[idx
].type_after_w1
= bidi_it
->type_after_w1
;
522 bidi_check_type (bidi_it
->type_after_w1
);
524 bidi_cache
[idx
].resolved_level
= bidi_it
->resolved_level
;
526 bidi_cache
[idx
].resolved_level
= -1;
527 bidi_cache
[idx
].invalid_levels
= bidi_it
->invalid_levels
;
528 bidi_cache
[idx
].invalid_rl_levels
= bidi_it
->invalid_rl_levels
;
529 bidi_cache
[idx
].next_for_neutral
= bidi_it
->next_for_neutral
;
530 bidi_cache
[idx
].next_for_ws
= bidi_it
->next_for_ws
;
531 bidi_cache
[idx
].ignore_bn_limit
= bidi_it
->ignore_bn_limit
;
534 bidi_cache_last_idx
= idx
;
535 if (idx
>= bidi_cache_idx
)
536 bidi_cache_idx
= idx
+ 1;
539 static inline bidi_type_t
540 bidi_cache_find (EMACS_INT charpos
, int level
, struct bidi_it
*bidi_it
)
542 ptrdiff_t i
= bidi_cache_search (charpos
, level
, bidi_it
->scan_dir
);
544 if (i
>= bidi_cache_start
)
546 bidi_dir_t current_scan_dir
= bidi_it
->scan_dir
;
548 bidi_copy_it (bidi_it
, &bidi_cache
[i
]);
549 bidi_cache_last_idx
= i
;
550 /* Don't let scan direction from from the cached state override
551 the current scan direction. */
552 bidi_it
->scan_dir
= current_scan_dir
;
553 return bidi_it
->type
;
560 bidi_peek_at_next_level (struct bidi_it
*bidi_it
)
562 if (bidi_cache_idx
== bidi_cache_start
|| bidi_cache_last_idx
== -1)
564 return bidi_cache
[bidi_cache_last_idx
+ bidi_it
->scan_dir
].resolved_level
;
568 /***********************************************************************
569 Pushing and popping the bidi iterator state
570 ***********************************************************************/
571 /* 5-slot stack for saving the start of the previous level of the
572 cache. xdisp.c maintains a 5-slot stack for its iterator state,
573 and we need the same size of our stack. */
574 static ptrdiff_t bidi_cache_start_stack
[IT_STACK_SIZE
];
575 static int bidi_cache_sp
;
577 /* Push the bidi iterator state in preparation for reordering a
578 different object, e.g. display string found at certain buffer
579 position. Pushing the bidi iterator boils down to saving its
580 entire state on the cache and starting a new cache "stacked" on top
581 of the current cache. */
583 bidi_push_it (struct bidi_it
*bidi_it
)
585 /* Save the current iterator state in its entirety after the last
587 bidi_cache_ensure_space (bidi_cache_idx
);
588 memcpy (&bidi_cache
[bidi_cache_idx
++], bidi_it
, sizeof (struct bidi_it
));
590 /* Push the current cache start onto the stack. */
591 xassert (bidi_cache_sp
< IT_STACK_SIZE
);
592 bidi_cache_start_stack
[bidi_cache_sp
++] = bidi_cache_start
;
594 /* Start a new level of cache, and make it empty. */
595 bidi_cache_start
= bidi_cache_idx
;
596 bidi_cache_last_idx
= -1;
599 /* Restore the iterator state saved by bidi_push_it and return the
600 cache to the corresponding state. */
602 bidi_pop_it (struct bidi_it
*bidi_it
)
604 if (bidi_cache_start
<= 0)
607 /* Reset the next free cache slot index to what it was before the
608 call to bidi_push_it. */
609 bidi_cache_idx
= bidi_cache_start
- 1;
611 /* Restore the bidi iterator state saved in the cache. */
612 memcpy (bidi_it
, &bidi_cache
[bidi_cache_idx
], sizeof (struct bidi_it
));
614 /* Pop the previous cache start from the stack. */
615 if (bidi_cache_sp
<= 0)
617 bidi_cache_start
= bidi_cache_start_stack
[--bidi_cache_sp
];
619 /* Invalidate the last-used cache slot data. */
620 bidi_cache_last_idx
= -1;
623 static ptrdiff_t bidi_cache_total_alloc
;
625 /* Stash away a copy of the cache and its control variables. */
627 bidi_shelve_cache (void)
629 unsigned char *databuf
;
632 if (bidi_cache_idx
== 0)
635 databuf
= xmalloc (sizeof (bidi_cache_idx
)
636 + bidi_cache_idx
* sizeof (struct bidi_it
)
637 + sizeof (bidi_cache_start_stack
)
638 + sizeof (bidi_cache_sp
) + sizeof (bidi_cache_start
)
639 + sizeof (bidi_cache_last_idx
));
640 bidi_cache_total_alloc
+=
641 sizeof (bidi_cache_idx
) + bidi_cache_idx
* sizeof (struct bidi_it
)
642 + sizeof (bidi_cache_start_stack
)
643 + sizeof (bidi_cache_sp
) + sizeof (bidi_cache_start
)
644 + sizeof (bidi_cache_last_idx
);
646 memcpy (databuf
, &bidi_cache_idx
, sizeof (bidi_cache_idx
));
647 memcpy (databuf
+ sizeof (bidi_cache_idx
),
648 bidi_cache
, bidi_cache_idx
* sizeof (struct bidi_it
));
649 memcpy (databuf
+ sizeof (bidi_cache_idx
)
650 + bidi_cache_idx
* sizeof (struct bidi_it
),
651 bidi_cache_start_stack
, sizeof (bidi_cache_start_stack
));
652 memcpy (databuf
+ sizeof (bidi_cache_idx
)
653 + bidi_cache_idx
* sizeof (struct bidi_it
)
654 + sizeof (bidi_cache_start_stack
),
655 &bidi_cache_sp
, sizeof (bidi_cache_sp
));
656 memcpy (databuf
+ sizeof (bidi_cache_idx
)
657 + bidi_cache_idx
* sizeof (struct bidi_it
)
658 + sizeof (bidi_cache_start_stack
) + sizeof (bidi_cache_sp
),
659 &bidi_cache_start
, sizeof (bidi_cache_start
));
660 memcpy (databuf
+ sizeof (bidi_cache_idx
)
661 + bidi_cache_idx
* sizeof (struct bidi_it
)
662 + sizeof (bidi_cache_start_stack
) + sizeof (bidi_cache_sp
)
663 + sizeof (bidi_cache_start
),
664 &bidi_cache_last_idx
, sizeof (bidi_cache_last_idx
));
669 /* Restore the cache state from a copy stashed away by
670 bidi_shelve_cache, and free the buffer used to stash that copy.
671 JUST_FREE non-zero means free the buffer, but don't restore the
672 cache; used when the corresponding iterator is discarded instead of
675 bidi_unshelve_cache (void *databuf
, int just_free
)
677 unsigned char *p
= databuf
;
683 /* A NULL pointer means an empty cache. */
684 bidi_cache_start
= 0;
695 memcpy (&idx
, p
, sizeof (bidi_cache_idx
));
696 bidi_cache_total_alloc
-=
697 sizeof (bidi_cache_idx
) + idx
* sizeof (struct bidi_it
)
698 + sizeof (bidi_cache_start_stack
) + sizeof (bidi_cache_sp
)
699 + sizeof (bidi_cache_start
) + sizeof (bidi_cache_last_idx
);
703 memcpy (&bidi_cache_idx
, p
, sizeof (bidi_cache_idx
));
704 bidi_cache_ensure_space (bidi_cache_idx
);
705 memcpy (bidi_cache
, p
+ sizeof (bidi_cache_idx
),
706 bidi_cache_idx
* sizeof (struct bidi_it
));
707 memcpy (bidi_cache_start_stack
,
708 p
+ sizeof (bidi_cache_idx
)
709 + bidi_cache_idx
* sizeof (struct bidi_it
),
710 sizeof (bidi_cache_start_stack
));
711 memcpy (&bidi_cache_sp
,
712 p
+ sizeof (bidi_cache_idx
)
713 + bidi_cache_idx
* sizeof (struct bidi_it
)
714 + sizeof (bidi_cache_start_stack
),
715 sizeof (bidi_cache_sp
));
716 memcpy (&bidi_cache_start
,
717 p
+ sizeof (bidi_cache_idx
)
718 + bidi_cache_idx
* sizeof (struct bidi_it
)
719 + sizeof (bidi_cache_start_stack
) + sizeof (bidi_cache_sp
),
720 sizeof (bidi_cache_start
));
721 memcpy (&bidi_cache_last_idx
,
722 p
+ sizeof (bidi_cache_idx
)
723 + bidi_cache_idx
* sizeof (struct bidi_it
)
724 + sizeof (bidi_cache_start_stack
) + sizeof (bidi_cache_sp
)
725 + sizeof (bidi_cache_start
),
726 sizeof (bidi_cache_last_idx
));
727 bidi_cache_total_alloc
-=
728 sizeof (bidi_cache_idx
) + bidi_cache_idx
* sizeof (struct bidi_it
)
729 + sizeof (bidi_cache_start_stack
) + sizeof (bidi_cache_sp
)
730 + sizeof (bidi_cache_start
) + sizeof (bidi_cache_last_idx
);
738 /***********************************************************************
740 ***********************************************************************/
742 bidi_initialize (void)
745 #include "biditype.h"
746 #include "bidimirror.h"
750 bidi_type_table
= Fmake_char_table (Qnil
, make_number (STRONG_L
));
751 staticpro (&bidi_type_table
);
753 for (i
= 0; i
< sizeof bidi_type
/ sizeof bidi_type
[0]; i
++)
754 char_table_set_range (bidi_type_table
, bidi_type
[i
].from
, bidi_type
[i
].to
,
755 make_number (bidi_type
[i
].type
));
757 bidi_mirror_table
= Fmake_char_table (Qnil
, Qnil
);
758 staticpro (&bidi_mirror_table
);
760 for (i
= 0; i
< sizeof bidi_mirror
/ sizeof bidi_mirror
[0]; i
++)
761 char_table_set (bidi_mirror_table
, bidi_mirror
[i
].from
,
762 make_number (bidi_mirror
[i
].to
));
764 Qparagraph_start
= intern ("paragraph-start");
765 staticpro (&Qparagraph_start
);
766 paragraph_start_re
= Fsymbol_value (Qparagraph_start
);
767 if (!STRINGP (paragraph_start_re
))
768 paragraph_start_re
= build_string ("\f\\|[ \t]*$");
769 staticpro (¶graph_start_re
);
770 Qparagraph_separate
= intern ("paragraph-separate");
771 staticpro (&Qparagraph_separate
);
772 paragraph_separate_re
= Fsymbol_value (Qparagraph_separate
);
773 if (!STRINGP (paragraph_separate_re
))
774 paragraph_separate_re
= build_string ("[ \t\f]*$");
775 staticpro (¶graph_separate_re
);
778 bidi_cache_total_alloc
= 0;
780 bidi_initialized
= 1;
783 /* Do whatever UAX#9 clause X8 says should be done at paragraph's
786 bidi_set_paragraph_end (struct bidi_it
*bidi_it
)
788 bidi_it
->invalid_levels
= 0;
789 bidi_it
->invalid_rl_levels
= -1;
790 bidi_it
->stack_idx
= 0;
791 bidi_it
->resolved_level
= bidi_it
->level_stack
[0].level
;
794 /* Initialize the bidi iterator from buffer/string position CHARPOS. */
796 bidi_init_it (EMACS_INT charpos
, EMACS_INT bytepos
, int frame_window_p
,
797 struct bidi_it
*bidi_it
)
799 if (! bidi_initialized
)
802 bidi_it
->charpos
= charpos
;
804 bidi_it
->bytepos
= bytepos
;
805 bidi_it
->frame_window_p
= frame_window_p
;
806 bidi_it
->nchars
= -1; /* to be computed in bidi_resolve_explicit_1 */
807 bidi_it
->first_elt
= 1;
808 bidi_set_paragraph_end (bidi_it
);
809 bidi_it
->new_paragraph
= 1;
810 bidi_it
->separator_limit
= -1;
811 bidi_it
->type
= NEUTRAL_B
;
812 bidi_it
->type_after_w1
= NEUTRAL_B
;
813 bidi_it
->orig_type
= NEUTRAL_B
;
814 bidi_it
->prev_was_pdf
= 0;
815 bidi_it
->prev
.type
= bidi_it
->prev
.type_after_w1
=
816 bidi_it
->prev
.orig_type
= UNKNOWN_BT
;
817 bidi_it
->last_strong
.type
= bidi_it
->last_strong
.type_after_w1
=
818 bidi_it
->last_strong
.orig_type
= UNKNOWN_BT
;
819 bidi_it
->next_for_neutral
.charpos
= -1;
820 bidi_it
->next_for_neutral
.type
=
821 bidi_it
->next_for_neutral
.type_after_w1
=
822 bidi_it
->next_for_neutral
.orig_type
= UNKNOWN_BT
;
823 bidi_it
->prev_for_neutral
.charpos
= -1;
824 bidi_it
->prev_for_neutral
.type
=
825 bidi_it
->prev_for_neutral
.type_after_w1
=
826 bidi_it
->prev_for_neutral
.orig_type
= UNKNOWN_BT
;
827 bidi_it
->sor
= L2R
; /* FIXME: should it be user-selectable? */
828 bidi_it
->disp_pos
= -1; /* invalid/unknown */
829 bidi_it
->disp_prop_p
= 0;
830 /* We can only shrink the cache if we are at the bottom level of its
832 if (bidi_cache_start
== 0)
833 bidi_cache_shrink ();
838 /* Perform initializations for reordering a new line of bidi text. */
840 bidi_line_init (struct bidi_it
*bidi_it
)
842 bidi_it
->scan_dir
= 1; /* FIXME: do we need to have control on this? */
843 bidi_it
->resolved_level
= bidi_it
->level_stack
[0].level
;
844 bidi_it
->level_stack
[0].override
= NEUTRAL_DIR
; /* X1 */
845 bidi_it
->invalid_levels
= 0;
846 bidi_it
->invalid_rl_levels
= -1;
847 bidi_it
->next_en_pos
= -1;
848 bidi_it
->next_for_ws
.type
= UNKNOWN_BT
;
849 bidi_set_sor_type (bidi_it
,
850 bidi_it
->paragraph_dir
== R2L
? 1 : 0,
851 bidi_it
->level_stack
[0].level
); /* X10 */
857 /***********************************************************************
859 ***********************************************************************/
861 /* Count bytes in string S between BEG/BEGBYTE and END. BEG and END
862 are zero-based character positions in S, BEGBYTE is byte position
863 corresponding to BEG. UNIBYTE, if non-zero, means S is a unibyte
865 static inline EMACS_INT
866 bidi_count_bytes (const unsigned char *s
, const EMACS_INT beg
,
867 const EMACS_INT begbyte
, const EMACS_INT end
, int unibyte
)
870 const unsigned char *p
= s
+ begbyte
, *start
= p
;
876 if (!CHAR_HEAD_P (*p
))
881 p
+= BYTES_BY_CHAR_HEAD (*p
);
889 /* Fetch and returns the character at byte position BYTEPOS. If S is
890 non-NULL, fetch the character from string S; otherwise fetch the
891 character from the current buffer. UNIBYTE non-zero means S is a
894 bidi_char_at_pos (EMACS_INT bytepos
, const unsigned char *s
, int unibyte
)
901 return STRING_CHAR (s
+ bytepos
);
904 return FETCH_MULTIBYTE_CHAR (bytepos
);
907 /* Fetch and return the character at BYTEPOS/CHARPOS. If that
908 character is covered by a display string, treat the entire run of
909 covered characters as a single character u+FFFC, and return their
910 combined length in CH_LEN and NCHARS. DISP_POS specifies the
911 character position of the next display string, or -1 if not yet
912 computed. DISP_PROP_P non-zero means that there's really a display
913 string at DISP_POS, as opposed to when we searched till DISP_POS
914 without findingone. When the next character is at or beyond that
915 position, the function updates DISP_POS with the position of the
916 next display string. STRING->s is the C string to iterate, or NULL
917 if iterating over a buffer or a Lisp string; in the latter case,
918 STRING->lstring is the Lisp string. */
920 bidi_fetch_char (EMACS_INT bytepos
, EMACS_INT charpos
, EMACS_INT
*disp_pos
,
921 int *disp_prop_p
, struct bidi_string_data
*string
,
922 int frame_window_p
, EMACS_INT
*ch_len
, EMACS_INT
*nchars
)
926 (string
->s
|| STRINGP (string
->lstring
)) ? string
->schars
: ZV
;
929 /* If we got past the last known position of display string, compute
930 the position of the next one. That position could be at CHARPOS. */
931 if (charpos
< endpos
&& charpos
> *disp_pos
)
933 SET_TEXT_POS (pos
, charpos
, bytepos
);
934 *disp_pos
= compute_display_string_pos (&pos
, string
, frame_window_p
,
938 /* Fetch the character at BYTEPOS. */
939 if (charpos
>= endpos
)
947 else if (charpos
>= *disp_pos
&& *disp_prop_p
)
949 EMACS_INT disp_end_pos
;
951 /* We don't expect to find ourselves in the middle of a display
952 property. Hopefully, it will never be needed. */
953 if (charpos
> *disp_pos
)
955 /* Return the Unicode Object Replacement Character to represent
956 the entire run of characters covered by the display string. */
958 disp_end_pos
= compute_display_string_end (*disp_pos
, string
);
959 *nchars
= disp_end_pos
- *disp_pos
;
963 *ch_len
= bidi_count_bytes (string
->s
, *disp_pos
, bytepos
,
964 disp_end_pos
, string
->unibyte
);
965 else if (STRINGP (string
->lstring
))
966 *ch_len
= bidi_count_bytes (SDATA (string
->lstring
), *disp_pos
,
967 bytepos
, disp_end_pos
, string
->unibyte
);
969 *ch_len
= CHAR_TO_BYTE (disp_end_pos
) - bytepos
;
977 if (!string
->unibyte
)
979 ch
= STRING_CHAR_AND_LENGTH (string
->s
+ bytepos
, len
);
984 ch
= UNIBYTE_TO_CHAR (string
->s
[bytepos
]);
988 else if (STRINGP (string
->lstring
))
992 if (!string
->unibyte
)
994 ch
= STRING_CHAR_AND_LENGTH (SDATA (string
->lstring
) + bytepos
,
1000 ch
= UNIBYTE_TO_CHAR (SREF (string
->lstring
, bytepos
));
1006 ch
= FETCH_MULTIBYTE_CHAR (bytepos
);
1007 *ch_len
= CHAR_BYTES (ch
);
1012 /* If we just entered a run of characters covered by a display
1013 string, compute the position of the next display string. */
1014 if (charpos
+ *nchars
<= endpos
&& charpos
+ *nchars
> *disp_pos
1017 SET_TEXT_POS (pos
, charpos
+ *nchars
, bytepos
+ *ch_len
);
1018 *disp_pos
= compute_display_string_pos (&pos
, string
, frame_window_p
,
1026 /***********************************************************************
1027 Determining paragraph direction
1028 ***********************************************************************/
1030 /* Check if buffer position CHARPOS/BYTEPOS is the end of a paragraph.
1031 Value is the non-negative length of the paragraph separator
1032 following the buffer position, -1 if position is at the beginning
1033 of a new paragraph, or -2 if position is neither at beginning nor
1034 at end of a paragraph. */
1036 bidi_at_paragraph_end (EMACS_INT charpos
, EMACS_INT bytepos
)
1039 Lisp_Object start_re
;
1042 sep_re
= paragraph_separate_re
;
1043 start_re
= paragraph_start_re
;
1045 val
= fast_looking_at (sep_re
, charpos
, bytepos
, ZV
, ZV_BYTE
, Qnil
);
1048 if (fast_looking_at (start_re
, charpos
, bytepos
, ZV
, ZV_BYTE
, Qnil
) >= 0)
1057 /* Find the beginning of this paragraph by looking back in the buffer.
1058 Value is the byte position of the paragraph's beginning. */
1060 bidi_find_paragraph_start (EMACS_INT pos
, EMACS_INT pos_byte
)
1062 Lisp_Object re
= paragraph_start_re
;
1063 EMACS_INT limit
= ZV
, limit_byte
= ZV_BYTE
;
1065 while (pos_byte
> BEGV_BYTE
1066 && fast_looking_at (re
, pos
, pos_byte
, limit
, limit_byte
, Qnil
) < 0)
1068 /* FIXME: What if the paragraph beginning is covered by a
1069 display string? And what if a display string covering some
1070 of the text over which we scan back includes
1071 paragraph_start_re? */
1072 pos
= find_next_newline_no_quit (pos
- 1, -1);
1073 pos_byte
= CHAR_TO_BYTE (pos
);
1078 /* Determine the base direction, a.k.a. base embedding level, of the
1079 paragraph we are about to iterate through. If DIR is either L2R or
1080 R2L, just use that. Otherwise, determine the paragraph direction
1081 from the first strong directional character of the paragraph.
1083 NO_DEFAULT_P non-zero means don't default to L2R if the paragraph
1084 has no strong directional characters and both DIR and
1085 bidi_it->paragraph_dir are NEUTRAL_DIR. In that case, search back
1086 in the buffer until a paragraph is found with a strong character,
1087 or until hitting BEGV. In the latter case, fall back to L2R. This
1088 flag is used in current-bidi-paragraph-direction.
1090 Note that this function gives the paragraph separator the same
1091 direction as the preceding paragraph, even though Emacs generally
1092 views the separartor as not belonging to any paragraph. */
1094 bidi_paragraph_init (bidi_dir_t dir
, struct bidi_it
*bidi_it
, int no_default_p
)
1096 EMACS_INT bytepos
= bidi_it
->bytepos
;
1097 int string_p
= bidi_it
->string
.s
!= NULL
|| STRINGP (bidi_it
->string
.lstring
);
1098 EMACS_INT pstartbyte
;
1099 /* Note that begbyte is a byte position, while end is a character
1100 position. Yes, this is ugly, but we are trying to avoid costly
1101 calls to BYTE_TO_CHAR and its ilk. */
1102 EMACS_INT begbyte
= string_p
? 0 : BEGV_BYTE
;
1103 EMACS_INT end
= string_p
? bidi_it
->string
.schars
: ZV
;
1105 /* Special case for an empty buffer. */
1106 if (bytepos
== begbyte
&& bidi_it
->charpos
== end
)
1108 /* We should never be called at EOB or before BEGV. */
1109 else if (bidi_it
->charpos
>= end
|| bytepos
< begbyte
)
1114 bidi_it
->paragraph_dir
= L2R
;
1115 bidi_it
->new_paragraph
= 0;
1117 else if (dir
== R2L
)
1119 bidi_it
->paragraph_dir
= R2L
;
1120 bidi_it
->new_paragraph
= 0;
1122 else if (dir
== NEUTRAL_DIR
) /* P2 */
1125 EMACS_INT ch_len
, nchars
;
1126 EMACS_INT pos
, disp_pos
= -1;
1127 int disp_prop_p
= 0;
1129 const unsigned char *s
;
1131 if (!bidi_initialized
)
1134 /* If we are inside a paragraph separator, we are just waiting
1135 for the separator to be exhausted; use the previous paragraph
1136 direction. But don't do that if we have been just reseated,
1137 because we need to reinitialize below in that case. */
1138 if (!bidi_it
->first_elt
1139 && bidi_it
->charpos
< bidi_it
->separator_limit
)
1142 /* If we are on a newline, get past it to where the next
1143 paragraph might start. But don't do that at BEGV since then
1144 we are potentially in a new paragraph that doesn't yet
1146 pos
= bidi_it
->charpos
;
1147 s
= STRINGP (bidi_it
->string
.lstring
) ?
1148 SDATA (bidi_it
->string
.lstring
) : bidi_it
->string
.s
;
1149 if (bytepos
> begbyte
1150 && bidi_char_at_pos (bytepos
, s
, bidi_it
->string
.unibyte
) == '\n')
1156 /* We are either at the beginning of a paragraph or in the
1157 middle of it. Find where this paragraph starts. */
1160 /* We don't support changes of paragraph direction inside a
1161 string. It is treated as a single paragraph. */
1165 pstartbyte
= bidi_find_paragraph_start (pos
, bytepos
);
1166 bidi_it
->separator_limit
= -1;
1167 bidi_it
->new_paragraph
= 0;
1169 /* The following loop is run more than once only if NO_DEFAULT_P
1170 is non-zero, and only if we are iterating on a buffer. */
1172 bytepos
= pstartbyte
;
1174 pos
= BYTE_TO_CHAR (bytepos
);
1175 ch
= bidi_fetch_char (bytepos
, pos
, &disp_pos
, &disp_prop_p
,
1177 bidi_it
->frame_window_p
, &ch_len
, &nchars
);
1178 type
= bidi_get_type (ch
, NEUTRAL_DIR
);
1180 for (pos
+= nchars
, bytepos
+= ch_len
;
1181 /* NOTE: UAX#9 says to search only for L, AL, or R types
1182 of characters, and ignore RLE, RLO, LRE, and LRO.
1183 However, I'm not sure it makes sense to omit those 4;
1184 should try with and without that to see the effect. */
1185 (bidi_get_category (type
) != STRONG
)
1186 || (bidi_ignore_explicit_marks_for_paragraph_level
1187 && (type
== RLE
|| type
== RLO
1188 || type
== LRE
|| type
== LRO
));
1189 type
= bidi_get_type (ch
, NEUTRAL_DIR
))
1193 /* Pretend there's a paragraph separator at end of
1199 && type
== NEUTRAL_B
1200 && bidi_at_paragraph_end (pos
, bytepos
) >= -1)
1202 /* Fetch next character and advance to get past it. */
1203 ch
= bidi_fetch_char (bytepos
, pos
, &disp_pos
,
1204 &disp_prop_p
, &bidi_it
->string
,
1205 bidi_it
->frame_window_p
, &ch_len
, &nchars
);
1209 if (type
== STRONG_R
|| type
== STRONG_AL
) /* P3 */
1210 bidi_it
->paragraph_dir
= R2L
;
1211 else if (type
== STRONG_L
)
1212 bidi_it
->paragraph_dir
= L2R
;
1214 && no_default_p
&& bidi_it
->paragraph_dir
== NEUTRAL_DIR
)
1216 /* If this paragraph is at BEGV, default to L2R. */
1217 if (pstartbyte
== BEGV_BYTE
)
1218 bidi_it
->paragraph_dir
= L2R
; /* P3 and HL1 */
1221 EMACS_INT prevpbyte
= pstartbyte
;
1222 EMACS_INT p
= BYTE_TO_CHAR (pstartbyte
), pbyte
= pstartbyte
;
1224 /* Find the beginning of the previous paragraph, if any. */
1225 while (pbyte
> BEGV_BYTE
&& prevpbyte
>= pstartbyte
)
1227 /* FXIME: What if p is covered by a display
1228 string? See also a FIXME inside
1229 bidi_find_paragraph_start. */
1231 pbyte
= CHAR_TO_BYTE (p
);
1232 prevpbyte
= bidi_find_paragraph_start (p
, pbyte
);
1234 pstartbyte
= prevpbyte
;
1238 && no_default_p
&& bidi_it
->paragraph_dir
== NEUTRAL_DIR
);
1243 /* Contrary to UAX#9 clause P3, we only default the paragraph
1244 direction to L2R if we have no previous usable paragraph
1245 direction. This is allowed by the HL1 clause. */
1246 if (bidi_it
->paragraph_dir
!= L2R
&& bidi_it
->paragraph_dir
!= R2L
)
1247 bidi_it
->paragraph_dir
= L2R
; /* P3 and HL1 ``higher-level protocols'' */
1248 if (bidi_it
->paragraph_dir
== R2L
)
1249 bidi_it
->level_stack
[0].level
= 1;
1251 bidi_it
->level_stack
[0].level
= 0;
1253 bidi_line_init (bidi_it
);
1257 /***********************************************************************
1258 Resolving explicit and implicit levels.
1259 The rest of this file constitutes the core of the UBA implementation.
1260 ***********************************************************************/
1263 bidi_explicit_dir_char (int ch
)
1265 bidi_type_t ch_type
;
1267 if (!bidi_initialized
)
1269 ch_type
= (bidi_type_t
) XINT (CHAR_TABLE_REF (bidi_type_table
, ch
));
1270 return (ch_type
== LRE
|| ch_type
== LRO
1271 || ch_type
== RLE
|| ch_type
== RLO
1275 /* A helper function for bidi_resolve_explicit. It advances to the
1276 next character in logical order and determines the new embedding
1277 level and directional override, but does not take into account
1278 empty embeddings. */
1280 bidi_resolve_explicit_1 (struct bidi_it
*bidi_it
)
1286 bidi_dir_t override
;
1287 int string_p
= bidi_it
->string
.s
!= NULL
|| STRINGP (bidi_it
->string
.lstring
);
1289 /* If reseat()'ed, don't advance, so as to start iteration from the
1290 position where we were reseated. bidi_it->bytepos can be less
1291 than BEGV_BYTE after reseat to BEGV. */
1292 if (bidi_it
->bytepos
< (string_p
? 0 : BEGV_BYTE
)
1293 || bidi_it
->first_elt
)
1295 bidi_it
->first_elt
= 0;
1298 const unsigned char *p
=
1299 STRINGP (bidi_it
->string
.lstring
)
1300 ? SDATA (bidi_it
->string
.lstring
) : bidi_it
->string
.s
;
1302 if (bidi_it
->charpos
< 0)
1303 bidi_it
->charpos
= 0;
1304 bidi_it
->bytepos
= bidi_count_bytes (p
, 0, 0, bidi_it
->charpos
,
1305 bidi_it
->string
.unibyte
);
1309 if (bidi_it
->charpos
< BEGV
)
1310 bidi_it
->charpos
= BEGV
;
1311 bidi_it
->bytepos
= CHAR_TO_BYTE (bidi_it
->charpos
);
1314 /* Don't move at end of buffer/string. */
1315 else if (bidi_it
->charpos
< (string_p
? bidi_it
->string
.schars
: ZV
))
1317 /* Advance to the next character, skipping characters covered by
1318 display strings (nchars > 1). */
1319 if (bidi_it
->nchars
<= 0)
1321 bidi_it
->charpos
+= bidi_it
->nchars
;
1322 if (bidi_it
->ch_len
== 0)
1324 bidi_it
->bytepos
+= bidi_it
->ch_len
;
1327 current_level
= bidi_it
->level_stack
[bidi_it
->stack_idx
].level
; /* X1 */
1328 override
= bidi_it
->level_stack
[bidi_it
->stack_idx
].override
;
1329 new_level
= current_level
;
1331 if (bidi_it
->charpos
>= (string_p
? bidi_it
->string
.schars
: ZV
))
1334 bidi_it
->ch_len
= 1;
1335 bidi_it
->nchars
= 1;
1336 bidi_it
->disp_pos
= (string_p
? bidi_it
->string
.schars
: ZV
);
1337 bidi_it
->disp_prop_p
= 0;
1341 /* Fetch the character at BYTEPOS. If it is covered by a
1342 display string, treat the entire run of covered characters as
1343 a single character u+FFFC. */
1344 curchar
= bidi_fetch_char (bidi_it
->bytepos
, bidi_it
->charpos
,
1345 &bidi_it
->disp_pos
, &bidi_it
->disp_prop_p
,
1346 &bidi_it
->string
, bidi_it
->frame_window_p
,
1347 &bidi_it
->ch_len
, &bidi_it
->nchars
);
1349 bidi_it
->ch
= curchar
;
1351 /* Don't apply directional override here, as all the types we handle
1352 below will not be affected by the override anyway, and we need
1353 the original type unaltered. The override will be applied in
1354 bidi_resolve_weak. */
1355 type
= bidi_get_type (curchar
, NEUTRAL_DIR
);
1356 bidi_it
->orig_type
= type
;
1357 bidi_check_type (bidi_it
->orig_type
);
1360 bidi_it
->prev_was_pdf
= 0;
1362 bidi_it
->type_after_w1
= UNKNOWN_BT
;
1368 bidi_it
->type_after_w1
= type
;
1369 bidi_check_type (bidi_it
->type_after_w1
);
1370 type
= WEAK_BN
; /* X9/Retaining */
1371 if (bidi_it
->ignore_bn_limit
<= -1)
1373 if (current_level
<= BIDI_MAXLEVEL
- 4)
1375 /* Compute the least odd embedding level greater than
1376 the current level. */
1377 new_level
= ((current_level
+ 1) & ~1) + 1;
1378 if (bidi_it
->type_after_w1
== RLE
)
1379 override
= NEUTRAL_DIR
;
1382 if (current_level
== BIDI_MAXLEVEL
- 4)
1383 bidi_it
->invalid_rl_levels
= 0;
1384 bidi_push_embedding_level (bidi_it
, new_level
, override
);
1388 bidi_it
->invalid_levels
++;
1389 /* See the commentary about invalid_rl_levels below. */
1390 if (bidi_it
->invalid_rl_levels
< 0)
1391 bidi_it
->invalid_rl_levels
= 0;
1392 bidi_it
->invalid_rl_levels
++;
1395 else if (bidi_it
->prev
.type_after_w1
== WEAK_EN
/* W5/Retaining */
1396 || bidi_it
->next_en_pos
> bidi_it
->charpos
)
1401 bidi_it
->type_after_w1
= type
;
1402 bidi_check_type (bidi_it
->type_after_w1
);
1403 type
= WEAK_BN
; /* X9/Retaining */
1404 if (bidi_it
->ignore_bn_limit
<= -1)
1406 if (current_level
<= BIDI_MAXLEVEL
- 5)
1408 /* Compute the least even embedding level greater than
1409 the current level. */
1410 new_level
= ((current_level
+ 2) & ~1);
1411 if (bidi_it
->type_after_w1
== LRE
)
1412 override
= NEUTRAL_DIR
;
1415 bidi_push_embedding_level (bidi_it
, new_level
, override
);
1419 bidi_it
->invalid_levels
++;
1420 /* invalid_rl_levels counts invalid levels encountered
1421 while the embedding level was already too high for
1422 LRE/LRO, but not for RLE/RLO. That is because
1423 there may be exactly one PDF which we should not
1424 ignore even though invalid_levels is non-zero.
1425 invalid_rl_levels helps to know what PDF is
1427 if (bidi_it
->invalid_rl_levels
>= 0)
1428 bidi_it
->invalid_rl_levels
++;
1431 else if (bidi_it
->prev
.type_after_w1
== WEAK_EN
/* W5/Retaining */
1432 || bidi_it
->next_en_pos
> bidi_it
->charpos
)
1436 bidi_it
->type_after_w1
= type
;
1437 bidi_check_type (bidi_it
->type_after_w1
);
1438 type
= WEAK_BN
; /* X9/Retaining */
1439 if (bidi_it
->ignore_bn_limit
<= -1)
1441 if (!bidi_it
->invalid_rl_levels
)
1443 new_level
= bidi_pop_embedding_level (bidi_it
);
1444 bidi_it
->invalid_rl_levels
= -1;
1445 if (bidi_it
->invalid_levels
)
1446 bidi_it
->invalid_levels
--;
1447 /* else nothing: UAX#9 says to ignore invalid PDFs */
1449 if (!bidi_it
->invalid_levels
)
1450 new_level
= bidi_pop_embedding_level (bidi_it
);
1453 bidi_it
->invalid_levels
--;
1454 bidi_it
->invalid_rl_levels
--;
1457 else if (bidi_it
->prev
.type_after_w1
== WEAK_EN
/* W5/Retaining */
1458 || bidi_it
->next_en_pos
> bidi_it
->charpos
)
1466 bidi_it
->type
= type
;
1467 bidi_check_type (bidi_it
->type
);
1472 /* Given an iterator state in BIDI_IT, advance one character position
1473 in the buffer/string to the next character (in the logical order),
1474 resolve any explicit embeddings and directional overrides, and
1475 return the embedding level of the character after resolving
1476 explicit directives and ignoring empty embeddings. */
1478 bidi_resolve_explicit (struct bidi_it
*bidi_it
)
1480 int prev_level
= bidi_it
->level_stack
[bidi_it
->stack_idx
].level
;
1481 int new_level
= bidi_resolve_explicit_1 (bidi_it
);
1482 EMACS_INT eob
= bidi_it
->string
.s
? bidi_it
->string
.schars
: ZV
;
1483 const unsigned char *s
= STRINGP (bidi_it
->string
.lstring
)
1484 ? SDATA (bidi_it
->string
.lstring
) : bidi_it
->string
.s
;
1486 if (prev_level
< new_level
1487 && bidi_it
->type
== WEAK_BN
1488 && bidi_it
->ignore_bn_limit
== -1 /* only if not already known */
1489 && bidi_it
->charpos
< eob
/* not already at EOB */
1490 && bidi_explicit_dir_char (bidi_char_at_pos (bidi_it
->bytepos
1491 + bidi_it
->ch_len
, s
,
1492 bidi_it
->string
.unibyte
)))
1494 /* Avoid pushing and popping embedding levels if the level run
1495 is empty, as this breaks level runs where it shouldn't.
1496 UAX#9 removes all the explicit embedding and override codes,
1497 so empty embeddings disappear without a trace. We need to
1498 behave as if we did the same. */
1499 struct bidi_it saved_it
;
1500 int level
= prev_level
;
1502 bidi_copy_it (&saved_it
, bidi_it
);
1504 while (bidi_explicit_dir_char (bidi_char_at_pos (bidi_it
->bytepos
1505 + bidi_it
->ch_len
, s
,
1506 bidi_it
->string
.unibyte
)))
1508 /* This advances to the next character, skipping any
1509 characters covered by display strings. */
1510 level
= bidi_resolve_explicit_1 (bidi_it
);
1511 /* If string.lstring was relocated inside bidi_resolve_explicit_1,
1512 a pointer to its data is no longer valid. */
1513 if (STRINGP (bidi_it
->string
.lstring
))
1514 s
= SDATA (bidi_it
->string
.lstring
);
1517 if (bidi_it
->nchars
<= 0)
1519 if (level
== prev_level
) /* empty embedding */
1520 saved_it
.ignore_bn_limit
= bidi_it
->charpos
+ bidi_it
->nchars
;
1521 else /* this embedding is non-empty */
1522 saved_it
.ignore_bn_limit
= -2;
1524 bidi_copy_it (bidi_it
, &saved_it
);
1525 if (bidi_it
->ignore_bn_limit
> -1)
1527 /* We pushed a level, but we shouldn't have. Undo that. */
1528 if (!bidi_it
->invalid_rl_levels
)
1530 new_level
= bidi_pop_embedding_level (bidi_it
);
1531 bidi_it
->invalid_rl_levels
= -1;
1532 if (bidi_it
->invalid_levels
)
1533 bidi_it
->invalid_levels
--;
1535 if (!bidi_it
->invalid_levels
)
1536 new_level
= bidi_pop_embedding_level (bidi_it
);
1539 bidi_it
->invalid_levels
--;
1540 bidi_it
->invalid_rl_levels
--;
1545 if (bidi_it
->type
== NEUTRAL_B
) /* X8 */
1547 bidi_set_paragraph_end (bidi_it
);
1548 /* This is needed by bidi_resolve_weak below, and in L1. */
1549 bidi_it
->type_after_w1
= bidi_it
->type
;
1550 bidi_check_type (bidi_it
->type_after_w1
);
1556 /* Advance in the buffer/string, resolve weak types and return the
1557 type of the next character after weak type resolution. */
1559 bidi_resolve_weak (struct bidi_it
*bidi_it
)
1562 bidi_dir_t override
;
1563 int prev_level
= bidi_it
->level_stack
[bidi_it
->stack_idx
].level
;
1564 int new_level
= bidi_resolve_explicit (bidi_it
);
1566 bidi_type_t type_of_next
;
1567 struct bidi_it saved_it
;
1569 (STRINGP (bidi_it
->string
.lstring
) || bidi_it
->string
.s
)
1570 ? bidi_it
->string
.schars
: ZV
;
1572 type
= bidi_it
->type
;
1573 override
= bidi_it
->level_stack
[bidi_it
->stack_idx
].override
;
1575 if (type
== UNKNOWN_BT
1583 if (new_level
!= prev_level
1584 || bidi_it
->type
== NEUTRAL_B
)
1586 /* We've got a new embedding level run, compute the directional
1587 type of sor and initialize per-run variables (UAX#9, clause
1589 bidi_set_sor_type (bidi_it
, prev_level
, new_level
);
1591 else if (type
== NEUTRAL_S
|| type
== NEUTRAL_WS
1592 || type
== WEAK_BN
|| type
== STRONG_AL
)
1593 bidi_it
->type_after_w1
= type
; /* needed in L1 */
1594 bidi_check_type (bidi_it
->type_after_w1
);
1596 /* Level and directional override status are already recorded in
1597 bidi_it, and do not need any change; see X6. */
1598 if (override
== R2L
) /* X6 */
1600 else if (override
== L2R
)
1604 if (type
== WEAK_NSM
) /* W1 */
1606 /* Note that we don't need to consider the case where the
1607 prev character has its type overridden by an RLO or LRO,
1608 because then either the type of this NSM would have been
1609 also overridden, or the previous character is outside the
1610 current level run, and thus not relevant to this NSM.
1611 This is why NSM gets the type_after_w1 of the previous
1613 if (bidi_it
->prev
.type_after_w1
!= UNKNOWN_BT
1614 /* if type_after_w1 is NEUTRAL_B, this NSM is at sor */
1615 && bidi_it
->prev
.type_after_w1
!= NEUTRAL_B
)
1616 type
= bidi_it
->prev
.type_after_w1
;
1617 else if (bidi_it
->sor
== R2L
)
1619 else if (bidi_it
->sor
== L2R
)
1621 else /* shouldn't happen! */
1624 if (type
== WEAK_EN
/* W2 */
1625 && bidi_it
->last_strong
.type_after_w1
== STRONG_AL
)
1627 else if (type
== STRONG_AL
) /* W3 */
1629 else if ((type
== WEAK_ES
/* W4 */
1630 && bidi_it
->prev
.type_after_w1
== WEAK_EN
1631 && bidi_it
->prev
.orig_type
== WEAK_EN
)
1633 && ((bidi_it
->prev
.type_after_w1
== WEAK_EN
1634 && bidi_it
->prev
.orig_type
== WEAK_EN
)
1635 || bidi_it
->prev
.type_after_w1
== WEAK_AN
)))
1637 const unsigned char *s
=
1638 STRINGP (bidi_it
->string
.lstring
)
1639 ? SDATA (bidi_it
->string
.lstring
) : bidi_it
->string
.s
;
1642 bidi_it
->charpos
+ bidi_it
->nchars
>= eob
1644 : bidi_char_at_pos (bidi_it
->bytepos
+ bidi_it
->ch_len
, s
,
1645 bidi_it
->string
.unibyte
);
1646 type_of_next
= bidi_get_type (next_char
, override
);
1648 if (type_of_next
== WEAK_BN
1649 || bidi_explicit_dir_char (next_char
))
1651 bidi_copy_it (&saved_it
, bidi_it
);
1652 while (bidi_resolve_explicit (bidi_it
) == new_level
1653 && bidi_it
->type
== WEAK_BN
)
1655 type_of_next
= bidi_it
->type
;
1656 bidi_copy_it (bidi_it
, &saved_it
);
1659 /* If the next character is EN, but the last strong-type
1660 character is AL, that next EN will be changed to AN when
1661 we process it in W2 above. So in that case, this ES
1662 should not be changed into EN. */
1664 && type_of_next
== WEAK_EN
1665 && bidi_it
->last_strong
.type_after_w1
!= STRONG_AL
)
1667 else if (type
== WEAK_CS
)
1669 if (bidi_it
->prev
.type_after_w1
== WEAK_AN
1670 && (type_of_next
== WEAK_AN
1671 /* If the next character is EN, but the last
1672 strong-type character is AL, EN will be later
1673 changed to AN when we process it in W2 above.
1674 So in that case, this ES should not be
1676 || (type_of_next
== WEAK_EN
1677 && bidi_it
->last_strong
.type_after_w1
== STRONG_AL
)))
1679 else if (bidi_it
->prev
.type_after_w1
== WEAK_EN
1680 && type_of_next
== WEAK_EN
1681 && bidi_it
->last_strong
.type_after_w1
!= STRONG_AL
)
1685 else if (type
== WEAK_ET
/* W5: ET with EN before or after it */
1686 || type
== WEAK_BN
) /* W5/Retaining */
1688 if (bidi_it
->prev
.type_after_w1
== WEAK_EN
/* ET/BN w/EN before it */
1689 || bidi_it
->next_en_pos
> bidi_it
->charpos
)
1691 else /* W5: ET/BN with EN after it. */
1693 EMACS_INT en_pos
= bidi_it
->charpos
+ bidi_it
->nchars
;
1694 const unsigned char *s
=
1695 STRINGP (bidi_it
->string
.lstring
)
1696 ? SDATA (bidi_it
->string
.lstring
) : bidi_it
->string
.s
;
1698 if (bidi_it
->nchars
<= 0)
1701 bidi_it
->charpos
+ bidi_it
->nchars
>= eob
1703 : bidi_char_at_pos (bidi_it
->bytepos
+ bidi_it
->ch_len
, s
,
1704 bidi_it
->string
.unibyte
);
1705 type_of_next
= bidi_get_type (next_char
, override
);
1707 if (type_of_next
== WEAK_ET
1708 || type_of_next
== WEAK_BN
1709 || bidi_explicit_dir_char (next_char
))
1711 bidi_copy_it (&saved_it
, bidi_it
);
1712 while (bidi_resolve_explicit (bidi_it
) == new_level
1713 && (bidi_it
->type
== WEAK_BN
1714 || bidi_it
->type
== WEAK_ET
))
1716 type_of_next
= bidi_it
->type
;
1717 en_pos
= bidi_it
->charpos
;
1718 bidi_copy_it (bidi_it
, &saved_it
);
1720 if (type_of_next
== WEAK_EN
)
1722 /* If the last strong character is AL, the EN we've
1723 found will become AN when we get to it (W2). */
1724 if (bidi_it
->last_strong
.type_after_w1
!= STRONG_AL
)
1727 /* Remember this EN position, to speed up processing
1729 bidi_it
->next_en_pos
= en_pos
;
1731 else if (type
== WEAK_BN
)
1732 type
= NEUTRAL_ON
; /* W6/Retaining */
1738 if (type
== WEAK_ES
|| type
== WEAK_ET
|| type
== WEAK_CS
/* W6 */
1740 && (bidi_it
->prev
.type_after_w1
== WEAK_CS
/* W6/Retaining */
1741 || bidi_it
->prev
.type_after_w1
== WEAK_ES
1742 || bidi_it
->prev
.type_after_w1
== WEAK_ET
)))
1745 /* Store the type we've got so far, before we clobber it with strong
1746 types in W7 and while resolving neutral types. But leave alone
1747 the original types that were recorded above, because we will need
1748 them for the L1 clause. */
1749 if (bidi_it
->type_after_w1
== UNKNOWN_BT
)
1750 bidi_it
->type_after_w1
= type
;
1751 bidi_check_type (bidi_it
->type_after_w1
);
1753 if (type
== WEAK_EN
) /* W7 */
1755 if ((bidi_it
->last_strong
.type_after_w1
== STRONG_L
)
1756 || (bidi_it
->last_strong
.type
== UNKNOWN_BT
&& bidi_it
->sor
== L2R
))
1760 bidi_it
->type
= type
;
1761 bidi_check_type (bidi_it
->type
);
1765 /* Resolve the type of a neutral character according to the type of
1766 surrounding strong text and the current embedding level. */
1767 static inline bidi_type_t
1768 bidi_resolve_neutral_1 (bidi_type_t prev_type
, bidi_type_t next_type
, int lev
)
1770 /* N1: European and Arabic numbers are treated as though they were R. */
1771 if (next_type
== WEAK_EN
|| next_type
== WEAK_AN
)
1772 next_type
= STRONG_R
;
1773 if (prev_type
== WEAK_EN
|| prev_type
== WEAK_AN
)
1774 prev_type
= STRONG_R
;
1776 if (next_type
== prev_type
) /* N1 */
1778 else if ((lev
& 1) == 0) /* N2 */
1785 bidi_resolve_neutral (struct bidi_it
*bidi_it
)
1787 int prev_level
= bidi_it
->level_stack
[bidi_it
->stack_idx
].level
;
1788 bidi_type_t type
= bidi_resolve_weak (bidi_it
);
1789 int current_level
= bidi_it
->level_stack
[bidi_it
->stack_idx
].level
;
1791 if (!(type
== STRONG_R
1796 || type
== NEUTRAL_B
1797 || type
== NEUTRAL_S
1798 || type
== NEUTRAL_WS
1799 || type
== NEUTRAL_ON
))
1802 if (bidi_get_category (type
) == NEUTRAL
1803 || (type
== WEAK_BN
&& prev_level
== current_level
))
1805 if (bidi_it
->next_for_neutral
.type
!= UNKNOWN_BT
)
1806 type
= bidi_resolve_neutral_1 (bidi_it
->prev_for_neutral
.type
,
1807 bidi_it
->next_for_neutral
.type
,
1811 /* Arrrgh!! The UAX#9 algorithm is too deeply entrenched in
1812 the assumption of batch-style processing; see clauses W4,
1813 W5, and especially N1, which require to look far forward
1814 (as well as back) in the buffer/string. May the fleas of
1815 a thousand camels infest the armpits of those who design
1816 supposedly general-purpose algorithms by looking at their
1817 own implementations, and fail to consider other possible
1819 struct bidi_it saved_it
;
1820 bidi_type_t next_type
;
1822 if (bidi_it
->scan_dir
== -1)
1825 bidi_copy_it (&saved_it
, bidi_it
);
1826 /* Scan the text forward until we find the first non-neutral
1827 character, and then use that to resolve the neutral we
1828 are dealing with now. We also cache the scanned iterator
1829 states, to salvage some of the effort later. */
1830 bidi_cache_iterator_state (bidi_it
, 0);
1832 /* Record the info about the previous character, so that
1833 it will be cached below with this state. */
1834 if (bidi_it
->type_after_w1
!= WEAK_BN
/* W1/Retaining */
1835 && bidi_it
->type
!= WEAK_BN
)
1836 bidi_remember_char (&bidi_it
->prev
, bidi_it
);
1837 type
= bidi_resolve_weak (bidi_it
);
1838 /* Paragraph separators have their levels fully resolved
1839 at this point, so cache them as resolved. */
1840 bidi_cache_iterator_state (bidi_it
, type
== NEUTRAL_B
);
1841 /* FIXME: implement L1 here, by testing for a newline and
1842 resetting the level for any sequence of whitespace
1843 characters adjacent to it. */
1844 } while (!(type
== NEUTRAL_B
1846 && bidi_get_category (type
) != NEUTRAL
)
1847 /* This is all per level run, so stop when we
1848 reach the end of this level run. */
1849 || bidi_it
->level_stack
[bidi_it
->stack_idx
].level
!=
1852 bidi_remember_char (&saved_it
.next_for_neutral
, bidi_it
);
1863 /* N1: ``European and Arabic numbers are treated as
1864 though they were R.'' */
1865 next_type
= STRONG_R
;
1866 saved_it
.next_for_neutral
.type
= STRONG_R
;
1869 if (!bidi_explicit_dir_char (bidi_it
->ch
))
1870 abort (); /* can't happen: BNs are skipped */
1873 /* Marched all the way to the end of this level run.
1874 We need to use the eor type, whose information is
1875 stored by bidi_set_sor_type in the prev_for_neutral
1877 if (saved_it
.type
!= WEAK_BN
1878 || bidi_get_category (bidi_it
->prev
.type_after_w1
) == NEUTRAL
)
1880 next_type
= bidi_it
->prev_for_neutral
.type
;
1881 saved_it
.next_for_neutral
.type
= next_type
;
1882 bidi_check_type (next_type
);
1886 /* This is a BN which does not adjoin neutrals.
1887 Leave its type alone. */
1888 bidi_copy_it (bidi_it
, &saved_it
);
1889 return bidi_it
->type
;
1895 type
= bidi_resolve_neutral_1 (saved_it
.prev_for_neutral
.type
,
1896 next_type
, current_level
);
1897 saved_it
.type
= type
;
1898 bidi_check_type (type
);
1899 bidi_copy_it (bidi_it
, &saved_it
);
1905 /* Given an iterator state in BIDI_IT, advance one character position
1906 in the buffer/string to the next character (in the logical order),
1907 resolve the bidi type of that next character, and return that
1910 bidi_type_of_next_char (struct bidi_it
*bidi_it
)
1914 /* This should always be called during a forward scan. */
1915 if (bidi_it
->scan_dir
!= 1)
1918 /* Reset the limit until which to ignore BNs if we step out of the
1919 area where we found only empty levels. */
1920 if ((bidi_it
->ignore_bn_limit
> -1
1921 && bidi_it
->ignore_bn_limit
<= bidi_it
->charpos
)
1922 || (bidi_it
->ignore_bn_limit
== -2
1923 && !bidi_explicit_dir_char (bidi_it
->ch
)))
1924 bidi_it
->ignore_bn_limit
= -1;
1926 type
= bidi_resolve_neutral (bidi_it
);
1931 /* Given an iterator state BIDI_IT, advance one character position in
1932 the buffer/string to the next character (in the current scan
1933 direction), resolve the embedding and implicit levels of that next
1934 character, and return the resulting level. */
1936 bidi_level_of_next_char (struct bidi_it
*bidi_it
)
1939 int level
, prev_level
= -1;
1940 struct bidi_saved_info next_for_neutral
;
1941 EMACS_INT next_char_pos
= -2;
1943 if (bidi_it
->scan_dir
== 1)
1946 (bidi_it
->string
.s
|| STRINGP (bidi_it
->string
.lstring
))
1947 ? bidi_it
->string
.schars
: ZV
;
1949 /* There's no sense in trying to advance if we hit end of text. */
1950 if (bidi_it
->charpos
>= eob
)
1951 return bidi_it
->resolved_level
;
1953 /* Record the info about the previous character. */
1954 if (bidi_it
->type_after_w1
!= WEAK_BN
/* W1/Retaining */
1955 && bidi_it
->type
!= WEAK_BN
)
1956 bidi_remember_char (&bidi_it
->prev
, bidi_it
);
1957 if (bidi_it
->type_after_w1
== STRONG_R
1958 || bidi_it
->type_after_w1
== STRONG_L
1959 || bidi_it
->type_after_w1
== STRONG_AL
)
1960 bidi_remember_char (&bidi_it
->last_strong
, bidi_it
);
1961 /* FIXME: it sounds like we don't need both prev and
1962 prev_for_neutral members, but I'm leaving them both for now. */
1963 if (bidi_it
->type
== STRONG_R
|| bidi_it
->type
== STRONG_L
1964 || bidi_it
->type
== WEAK_EN
|| bidi_it
->type
== WEAK_AN
)
1965 bidi_remember_char (&bidi_it
->prev_for_neutral
, bidi_it
);
1967 /* If we overstepped the characters used for resolving neutrals
1968 and whitespace, invalidate their info in the iterator. */
1969 if (bidi_it
->charpos
>= bidi_it
->next_for_neutral
.charpos
)
1970 bidi_it
->next_for_neutral
.type
= UNKNOWN_BT
;
1971 if (bidi_it
->next_en_pos
>= 0
1972 && bidi_it
->charpos
>= bidi_it
->next_en_pos
)
1973 bidi_it
->next_en_pos
= -1;
1974 if (bidi_it
->next_for_ws
.type
!= UNKNOWN_BT
1975 && bidi_it
->charpos
>= bidi_it
->next_for_ws
.charpos
)
1976 bidi_it
->next_for_ws
.type
= UNKNOWN_BT
;
1978 /* This must be taken before we fill the iterator with the info
1979 about the next char. If we scan backwards, the iterator
1980 state must be already cached, so there's no need to know the
1981 embedding level of the previous character, since we will be
1982 returning to our caller shortly. */
1983 prev_level
= bidi_it
->level_stack
[bidi_it
->stack_idx
].level
;
1985 next_for_neutral
= bidi_it
->next_for_neutral
;
1987 /* Perhaps the character we want is already cached. If it is, the
1988 call to bidi_cache_find below will return a type other than
1990 if (bidi_cache_idx
> bidi_cache_start
&& !bidi_it
->first_elt
)
1993 (bidi_it
->string
.s
|| STRINGP (bidi_it
->string
.lstring
)) ? 0 : 1;
1995 if (bidi_it
->scan_dir
> 0)
1997 if (bidi_it
->nchars
<= 0)
1999 next_char_pos
= bidi_it
->charpos
+ bidi_it
->nchars
;
2001 else if (bidi_it
->charpos
>= bob
)
2002 /* Implementation note: we allow next_char_pos to be as low as
2003 0 for buffers or -1 for strings, and that is okay because
2004 that's the "position" of the sentinel iterator state we
2005 cached at the beginning of the iteration. */
2006 next_char_pos
= bidi_it
->charpos
- 1;
2007 if (next_char_pos
>= bob
- 1)
2008 type
= bidi_cache_find (next_char_pos
, -1, bidi_it
);
2014 if (type
!= UNKNOWN_BT
)
2016 /* Don't lose the information for resolving neutrals! The
2017 cached states could have been cached before their
2018 next_for_neutral member was computed. If we are on our way
2019 forward, we can simply take the info from the previous
2021 if (bidi_it
->scan_dir
== 1
2022 && bidi_it
->next_for_neutral
.type
== UNKNOWN_BT
)
2023 bidi_it
->next_for_neutral
= next_for_neutral
;
2025 /* If resolved_level is -1, it means this state was cached
2026 before it was completely resolved, so we cannot return
2028 if (bidi_it
->resolved_level
!= -1)
2029 return bidi_it
->resolved_level
;
2031 if (bidi_it
->scan_dir
== -1)
2032 /* If we are going backwards, the iterator state is already cached
2033 from previous scans, and should be fully resolved. */
2036 if (type
== UNKNOWN_BT
)
2037 type
= bidi_type_of_next_char (bidi_it
);
2039 if (type
== NEUTRAL_B
)
2040 return bidi_it
->resolved_level
;
2042 level
= bidi_it
->level_stack
[bidi_it
->stack_idx
].level
;
2043 if ((bidi_get_category (type
) == NEUTRAL
/* && type != NEUTRAL_B */)
2044 || (type
== WEAK_BN
&& prev_level
== level
))
2046 if (bidi_it
->next_for_neutral
.type
== UNKNOWN_BT
)
2049 /* If the cached state shows a neutral character, it was not
2050 resolved by bidi_resolve_neutral, so do it now. */
2051 type
= bidi_resolve_neutral_1 (bidi_it
->prev_for_neutral
.type
,
2052 bidi_it
->next_for_neutral
.type
,
2056 if (!(type
== STRONG_R
2060 || type
== WEAK_AN
))
2062 bidi_it
->type
= type
;
2063 bidi_check_type (bidi_it
->type
);
2065 /* For L1 below, we need to know, for each WS character, whether
2066 it belongs to a sequence of WS characters preceding a newline
2067 or a TAB or a paragraph separator. */
2068 if (bidi_it
->orig_type
== NEUTRAL_WS
2069 && bidi_it
->next_for_ws
.type
== UNKNOWN_BT
)
2072 EMACS_INT clen
= bidi_it
->ch_len
;
2073 EMACS_INT bpos
= bidi_it
->bytepos
;
2074 EMACS_INT cpos
= bidi_it
->charpos
;
2075 EMACS_INT disp_pos
= bidi_it
->disp_pos
;
2076 EMACS_INT nc
= bidi_it
->nchars
;
2077 struct bidi_string_data bs
= bidi_it
->string
;
2079 int fwp
= bidi_it
->frame_window_p
;
2080 int dpp
= bidi_it
->disp_prop_p
;
2082 if (bidi_it
->nchars
<= 0)
2085 ch
= bidi_fetch_char (bpos
+= clen
, cpos
+= nc
, &disp_pos
, &dpp
, &bs
,
2087 if (ch
== '\n' || ch
== BIDI_EOB
/* || ch == LINESEP_CHAR */)
2090 chtype
= bidi_get_type (ch
, NEUTRAL_DIR
);
2091 } while (chtype
== NEUTRAL_WS
|| chtype
== WEAK_BN
2092 || bidi_explicit_dir_char (ch
)); /* L1/Retaining */
2093 bidi_it
->next_for_ws
.type
= chtype
;
2094 bidi_check_type (bidi_it
->next_for_ws
.type
);
2095 bidi_it
->next_for_ws
.charpos
= cpos
;
2096 bidi_it
->next_for_ws
.bytepos
= bpos
;
2099 /* Resolve implicit levels, with a twist: PDFs get the embedding
2100 level of the enbedding they terminate. See below for the
2102 if (bidi_it
->orig_type
== PDF
2103 /* Don't do this if this formatting code didn't change the
2104 embedding level due to invalid or empty embeddings. */
2105 && prev_level
!= level
)
2107 /* Don't look in UAX#9 for the reason for this: it's our own
2108 private quirk. The reason is that we want the formatting
2109 codes to be delivered so that they bracket the text of their
2110 embedding. For example, given the text
2114 we want it to be displayed as
2122 which will result because we bump up the embedding level as
2123 soon as we see the RLO and pop it as soon as we see the PDF,
2124 so RLO itself has the same embedding level as "teST", and
2125 thus would be normally delivered last, just before the PDF.
2126 The switch below fiddles with the level of PDF so that this
2127 ugly side effect does not happen.
2129 (This is, of course, only important if the formatting codes
2130 are actually displayed, but Emacs does need to display them
2131 if the user wants to.) */
2134 else if (bidi_it
->orig_type
== NEUTRAL_B
/* L1 */
2135 || bidi_it
->orig_type
== NEUTRAL_S
2136 || bidi_it
->ch
== '\n' || bidi_it
->ch
== BIDI_EOB
2137 /* || bidi_it->ch == LINESEP_CHAR */
2138 || (bidi_it
->orig_type
== NEUTRAL_WS
2139 && (bidi_it
->next_for_ws
.type
== NEUTRAL_B
2140 || bidi_it
->next_for_ws
.type
== NEUTRAL_S
)))
2141 level
= bidi_it
->level_stack
[0].level
;
2142 else if ((level
& 1) == 0) /* I1 */
2144 if (type
== STRONG_R
)
2146 else if (type
== WEAK_EN
|| type
== WEAK_AN
)
2151 if (type
== STRONG_L
|| type
== WEAK_EN
|| type
== WEAK_AN
)
2155 bidi_it
->resolved_level
= level
;
2159 /* Move to the other edge of a level given by LEVEL. If END_FLAG is
2160 non-zero, we are at the end of a level, and we need to prepare to
2161 resume the scan of the lower level.
2163 If this level's other edge is cached, we simply jump to it, filling
2164 the iterator structure with the iterator state on the other edge.
2165 Otherwise, we walk the buffer or string until we come back to the
2166 same level as LEVEL.
2168 Note: we are not talking here about a ``level run'' in the UAX#9
2169 sense of the term, but rather about a ``level'' which includes
2170 all the levels higher than it. In other words, given the levels
2173 11111112222222333333334443343222222111111112223322111
2176 and assuming we are at point A scanning left to right, this
2177 function moves to point C, whereas the UAX#9 ``level 2 run'' ends
2180 bidi_find_other_level_edge (struct bidi_it
*bidi_it
, int level
, int end_flag
)
2182 int dir
= end_flag
? -bidi_it
->scan_dir
: bidi_it
->scan_dir
;
2185 /* Try the cache first. */
2186 if ((idx
= bidi_cache_find_level_change (level
, dir
, end_flag
))
2187 >= bidi_cache_start
)
2188 bidi_cache_fetch_state (idx
, bidi_it
);
2194 abort (); /* if we are at end of level, its edges must be cached */
2196 bidi_cache_iterator_state (bidi_it
, 1);
2198 new_level
= bidi_level_of_next_char (bidi_it
);
2199 bidi_cache_iterator_state (bidi_it
, 1);
2200 } while (new_level
>= level
);
2205 bidi_move_to_visually_next (struct bidi_it
*bidi_it
)
2207 int old_level
, new_level
, next_level
;
2208 struct bidi_it sentinel
;
2209 struct gcpro gcpro1
;
2211 if (bidi_it
->charpos
< 0 || bidi_it
->bytepos
< 0)
2214 if (bidi_it
->scan_dir
== 0)
2216 bidi_it
->scan_dir
= 1; /* default to logical order */
2219 /* The code below can call eval, and thus cause GC. If we are
2220 iterating a Lisp string, make sure it won't be GCed. */
2221 if (STRINGP (bidi_it
->string
.lstring
))
2222 GCPRO1 (bidi_it
->string
.lstring
);
2224 /* If we just passed a newline, initialize for the next line. */
2225 if (!bidi_it
->first_elt
&& bidi_it
->orig_type
== NEUTRAL_B
)
2226 bidi_line_init (bidi_it
);
2228 /* Prepare the sentinel iterator state, and cache it. When we bump
2229 into it, scanning backwards, we'll know that the last non-base
2230 level is exhausted. */
2231 if (bidi_cache_idx
== bidi_cache_start
)
2233 bidi_copy_it (&sentinel
, bidi_it
);
2234 if (bidi_it
->first_elt
)
2236 sentinel
.charpos
--; /* cached charpos needs to be monotonic */
2238 sentinel
.ch
= '\n'; /* doesn't matter, but why not? */
2239 sentinel
.ch_len
= 1;
2240 sentinel
.nchars
= 1;
2242 bidi_cache_iterator_state (&sentinel
, 1);
2245 old_level
= bidi_it
->resolved_level
;
2246 new_level
= bidi_level_of_next_char (bidi_it
);
2248 /* Reordering of resolved levels (clause L2) is implemented by
2249 jumping to the other edge of the level and flipping direction of
2250 scanning the text whenever we find a level change. */
2251 if (new_level
!= old_level
)
2253 int ascending
= new_level
> old_level
;
2254 int level_to_search
= ascending
? old_level
+ 1 : old_level
;
2255 int incr
= ascending
? 1 : -1;
2256 int expected_next_level
= old_level
+ incr
;
2258 /* Jump (or walk) to the other edge of this level. */
2259 bidi_find_other_level_edge (bidi_it
, level_to_search
, !ascending
);
2260 /* Switch scan direction and peek at the next character in the
2262 bidi_it
->scan_dir
= -bidi_it
->scan_dir
;
2264 /* The following loop handles the case where the resolved level
2265 jumps by more than one. This is typical for numbers inside a
2266 run of text with left-to-right embedding direction, but can
2267 also happen in other situations. In those cases the decision
2268 where to continue after a level change, and in what direction,
2269 is tricky. For example, given a text like below:
2274 (where the numbers below the text show the resolved levels),
2275 the result of reordering according to UAX#9 should be this:
2279 This is implemented by the loop below which flips direction
2280 and jumps to the other edge of the level each time it finds
2281 the new level not to be the expected one. The expected level
2282 is always one more or one less than the previous one. */
2283 next_level
= bidi_peek_at_next_level (bidi_it
);
2284 while (next_level
!= expected_next_level
)
2286 expected_next_level
+= incr
;
2287 level_to_search
+= incr
;
2288 bidi_find_other_level_edge (bidi_it
, level_to_search
, !ascending
);
2289 bidi_it
->scan_dir
= -bidi_it
->scan_dir
;
2290 next_level
= bidi_peek_at_next_level (bidi_it
);
2293 /* Finally, deliver the next character in the new direction. */
2294 next_level
= bidi_level_of_next_char (bidi_it
);
2297 /* Take note when we have just processed the newline that precedes
2298 the end of the paragraph. The next time we are about to be
2299 called, set_iterator_to_next will automatically reinit the
2300 paragraph direction, if needed. We do this at the newline before
2301 the paragraph separator, because the next character might not be
2302 the first character of the next paragraph, due to the bidi
2303 reordering, whereas we _must_ know the paragraph base direction
2304 _before_ we process the paragraph's text, since the base
2305 direction affects the reordering. */
2306 if (bidi_it
->scan_dir
== 1 && bidi_it
->orig_type
== NEUTRAL_B
)
2308 /* The paragraph direction of the entire string, once
2309 determined, is in effect for the entire string. Setting the
2310 separator limit to the end of the string prevents
2311 bidi_paragraph_init from being called automatically on this
2313 if (bidi_it
->string
.s
|| STRINGP (bidi_it
->string
.lstring
))
2314 bidi_it
->separator_limit
= bidi_it
->string
.schars
;
2315 else if (bidi_it
->bytepos
< ZV_BYTE
)
2318 bidi_at_paragraph_end (bidi_it
->charpos
+ bidi_it
->nchars
,
2319 bidi_it
->bytepos
+ bidi_it
->ch_len
);
2320 if (bidi_it
->nchars
<= 0)
2324 bidi_it
->new_paragraph
= 1;
2325 /* Record the buffer position of the last character of the
2326 paragraph separator. */
2327 bidi_it
->separator_limit
=
2328 bidi_it
->charpos
+ bidi_it
->nchars
+ sep_len
;
2333 if (bidi_it
->scan_dir
== 1 && bidi_cache_idx
> bidi_cache_start
)
2335 /* If we are at paragraph's base embedding level and beyond the
2336 last cached position, the cache's job is done and we can
2338 if (bidi_it
->resolved_level
== bidi_it
->level_stack
[0].level
2339 && bidi_it
->charpos
> (bidi_cache
[bidi_cache_idx
- 1].charpos
2340 + bidi_cache
[bidi_cache_idx
- 1].nchars
- 1))
2341 bidi_cache_reset ();
2342 /* But as long as we are caching during forward scan, we must
2343 cache each state, or else the cache integrity will be
2344 compromised: it assumes cached states correspond to buffer
2347 bidi_cache_iterator_state (bidi_it
, 1);
2350 if (STRINGP (bidi_it
->string
.lstring
))
2354 /* This is meant to be called from within the debugger, whenever you
2355 wish to examine the cache contents. */
2356 void bidi_dump_cached_states (void) EXTERNALLY_VISIBLE
;
2358 bidi_dump_cached_states (void)
2363 if (bidi_cache_idx
== 0)
2365 fprintf (stderr
, "The cache is empty.\n");
2368 fprintf (stderr
, "Total of %"pD
"d state%s in cache:\n",
2369 bidi_cache_idx
, bidi_cache_idx
== 1 ? "" : "s");
2371 for (i
= bidi_cache
[bidi_cache_idx
- 1].charpos
; i
> 0; i
/= 10)
2373 fputs ("ch ", stderr
);
2374 for (i
= 0; i
< bidi_cache_idx
; i
++)
2375 fprintf (stderr
, "%*c", ndigits
, bidi_cache
[i
].ch
);
2376 fputs ("\n", stderr
);
2377 fputs ("lvl ", stderr
);
2378 for (i
= 0; i
< bidi_cache_idx
; i
++)
2379 fprintf (stderr
, "%*d", ndigits
, bidi_cache
[i
].resolved_level
);
2380 fputs ("\n", stderr
);
2381 fputs ("pos ", stderr
);
2382 for (i
= 0; i
< bidi_cache_idx
; i
++)
2383 fprintf (stderr
, "%*"pI
"d", ndigits
, bidi_cache
[i
].charpos
);
2384 fputs ("\n", stderr
);