1 /* Routines to compute the current syntactic context, for font-lock mode.
2 Copyright (C) 1992, 1993, 1994 Free Software Foundation, Inc.
3 Copyright (C) 1995 Sun Microsystems, Inc.
5 This file is part of XEmacs.
7 XEmacs is free software; you can redistribute it and/or modify it
8 under the terms of the GNU General Public License as published by the
9 Free Software Foundation; either version 2, or (at your option) any
12 XEmacs is distributed in the hope that it will be useful, but WITHOUT
13 ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
14 FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License
17 You should have received a copy of the GNU General Public License
18 along with XEmacs; see the file COPYING. If not, write to
19 the Free Software Foundation, Inc., 59 Temple Place - Suite 330,
20 Boston, MA 02111-1307, USA. */
22 /* Synched up with: Not in FSF. */
24 /* This code computes the syntactic context of the current point, that is,
25 whether point is within a comment, a string, what have you. It does
26 this by picking a point "known" to be outside of any syntactic constructs
27 and moving forward, examining the syntax of each character.
29 Two caches are used: one caches the last point computed, and the other
30 caches the last point at the beginning of a line. This makes there
31 be little penalty for moving left-to-right on a line a character at a
32 time; makes starting over on a line be cheap; and makes random-accessing
33 within a line relatively cheap.
35 When we move to a different line farther down in the file (but within the
36 current top-level form) we simply continue computing forward. If we move
37 backward more than a line, or move beyond the end of the current tlf, or
38 switch buffers, then we call `beginning-of-defun' and start over from
41 #### We should really rewrite this to keep extents over the buffer
42 that hold the current syntactic information. This would be a big win.
43 This way there would be no guessing or incorrect results.
54 Lisp_Object Qblock_comment;
55 Lisp_Object Qbeginning_of_defun;
57 enum syntactic_context
62 context_block_comment,
63 context_generic_comment,
64 context_generic_string
67 enum block_comment_context
84 Bufpos start_point; /* beginning of defun */
85 Bufpos cur_point; /* cache location */
86 Bufpos end_point; /* end of defun */
87 struct buffer *buffer; /* does this need to be staticpro'd? */
88 enum syntactic_context context; /* single-char-syntax state */
89 enum block_comment_context ccontext; /* block-comment state */
90 enum comment_style style; /* which comment group */
91 Emchar scontext; /* active string delimiter */
92 int depth; /* depth in parens */
93 int backslash_p; /* just read a backslash */
94 int needs_its_head_reexamined; /* we're apparently outside of
95 a top level form, and far away
96 from it. This is a bad situation
97 because it will lead to constant
98 slowness as we keep going way
99 back to that form and moving
100 forward again. In this case,
101 we try to compute a "pseudo-
102 top-level-form" where the
103 depth is 0 and the context
104 is none at both ends. */
107 /* We have two caches; one for the current point and one for
108 the beginning of line. We used to rely on the caller to
109 tell us when to invalidate them, but now we do it ourselves;
110 it lets us be smarter. */
112 static struct context_cache context_cache;
114 static struct context_cache bol_context_cache;
118 #define reset_context_cache(cc) memset (cc, 0, sizeof (struct context_cache))
120 /* This function is called from signal_after_change() to tell us when
121 textual changes are made so we can flush our caches when necessary.
123 We make the following somewhat heuristic assumptions:
125 (remember that current_point is always >= start_point, but may be
126 less than or greater than end_point (we might not be inside any
129 1) Textual changes before the beginning of the current top-level form
130 don't affect anything; all we need to do is offset the caches
132 2) Textual changes right at the beginning of the current
133 top-level form messes things up and requires that we flush
135 3) Textual changes after the beginning of the current top-level form
136 and before one or both or the caches invalidates the corresponding
138 4) Textual changes after the caches and before the end of the
139 current top-level form don't affect anything; all we need to do is
140 offset the caches appropriately.
141 5) Textual changes right at the end of the current top-level form
142 necessitate recomputing that end value.
143 6) Textual changes after the end of the current top-level form
148 font_lock_maybe_update_syntactic_caches (struct buffer *buf, Bufpos start,
149 Bufpos orig_end, Bufpos new_end)
151 /* Note: either both context_cache and bol_context_cache are valid and
152 point to the same buffer, or both are invalid. If we have to
153 invalidate just context_cache, we recopy it from bol_context_cache.
155 if (context_cache.buffer != buf)
156 /* caches don't apply */
158 /* NOTE: The order of the if statements below is important. If you
159 change them around unthinkingly, you will probably break something. */
160 if (orig_end <= context_cache.start_point - 1)
162 /* case 1: before the beginning of the current top-level form */
163 Charcount diff = new_end - orig_end;
165 stderr_out ("font-lock; Case 1\n");
166 context_cache.start_point += diff;
167 context_cache.cur_point += diff;
168 context_cache.end_point += diff;
169 bol_context_cache.start_point += diff;
170 bol_context_cache.cur_point += diff;
171 bol_context_cache.end_point += diff;
173 else if (start <= context_cache.start_point)
176 stderr_out ("font-lock; Case 2\n");
177 /* case 2: right at the current top-level form (paren that starts
178 top level form got deleted or moved away from the newline it
180 reset_context_cache (&context_cache);
181 reset_context_cache (&bol_context_cache);
183 /* OK, now we know that the start is after the beginning of the
184 current top-level form. */
185 else if (start < bol_context_cache.cur_point)
188 stderr_out ("font-lock; Case 3 (1)\n");
189 /* case 3: after the beginning of the current top-level form
190 and before both of the caches */
191 reset_context_cache (&context_cache);
192 reset_context_cache (&bol_context_cache);
194 else if (start < context_cache.cur_point)
197 stderr_out ("font-lock; Case 3 (2)\n");
198 /* case 3: but only need to invalidate one cache */
199 context_cache = bol_context_cache;
201 /* OK, now we know that the start is after the caches. */
202 else if (start >= context_cache.end_point)
205 stderr_out ("font-lock; Case 6\n");
206 /* case 6: after the end of the current top-level form
207 and after the caches. */
209 else if (orig_end <= context_cache.end_point - 2)
211 /* case 4: after the caches and before the end of the
212 current top-level form */
213 Charcount diff = new_end - orig_end;
215 stderr_out ("font-lock; Case 4\n");
216 context_cache.end_point += diff;
217 bol_context_cache.end_point += diff;
222 stderr_out ("font-lock; Case 5\n");
223 /* case 5: right at the end of the current top-level form */
224 context_cache.end_point = context_cache.start_point - 1;
225 bol_context_cache.end_point = context_cache.start_point - 1;
229 /* This function is called from Fkill_buffer(). */
232 font_lock_buffer_was_killed (struct buffer *buf)
234 if (context_cache.buffer == buf)
236 reset_context_cache (&context_cache);
237 reset_context_cache (&bol_context_cache);
242 beginning_of_defun (struct buffer *buf, Bufpos pt)
244 /* This function can GC */
245 Bufpos opt = BUF_PT (buf);
246 if (pt == BUF_BEGV (buf))
248 BUF_SET_PT (buf, pt);
249 /* There used to be some kludginess to call c++-beginning-of-defun
250 if we're in C++ mode. There's no point in this any more;
251 we're using cc-mode. If you really want to get the old c++
252 mode working, fix it rather than the C code. */
253 call0_in_buffer (buf, Qbeginning_of_defun);
255 BUF_SET_PT (buf, opt);
260 end_of_defun (struct buffer *buf, Bufpos pt)
262 Lisp_Object retval = scan_lists (buf, pt, 1, 0, 0, 1);
266 return XINT (retval);
269 /* Set up context_cache for attempting to determine the syntactic context
270 in buffer BUF at point PT. */
273 setup_context_cache (struct buffer *buf, Bufpos pt)
275 int recomputed_start_point = 0;
276 /* This function can GC */
277 if (context_cache.buffer != buf || pt < context_cache.start_point)
281 stderr_out ("reset context cache\n");
282 /* OK, completely invalid. */
283 reset_context_cache (&context_cache);
284 reset_context_cache (&bol_context_cache);
286 if (!context_cache.buffer)
288 /* Need to recompute the start point. */
290 stderr_out ("recompute start\n");
291 context_cache.start_point = beginning_of_defun (buf, pt);
292 recomputed_start_point = 1;
293 bol_context_cache.start_point = context_cache.start_point;
294 bol_context_cache.buffer = context_cache.buffer = buf;
296 if (context_cache.end_point < context_cache.start_point)
298 /* Need to recompute the end point. */
300 stderr_out ("recompute end\n");
301 context_cache.end_point = end_of_defun (buf, context_cache.start_point);
302 bol_context_cache.end_point = context_cache.end_point;
304 if (bol_context_cache.cur_point == 0 ||
305 pt < bol_context_cache.cur_point)
308 stderr_out ("reset to start\n");
309 if (pt > context_cache.end_point
310 /* 3000 is some arbitrary delta but seems reasonable;
311 about the size of a reasonable function */
312 && pt - context_cache.end_point > 3000)
313 /* If we're far past the end of the top level form,
314 don't trust it; recompute it. */
316 /* But don't get in an infinite loop doing this.
317 If we're really far past the end of the top level
318 form, try to compute a pseudo-top-level form. */
319 if (recomputed_start_point)
320 context_cache.needs_its_head_reexamined = 1;
322 /* force recomputation */
325 /* Go to the nearest end of the top-level form that's before
327 if (pt > context_cache.end_point)
328 pt = context_cache.end_point;
330 pt = context_cache.start_point;
331 /* Reset current point to start of buffer. */
332 context_cache.cur_point = pt;
333 context_cache.context = context_none;
334 context_cache.ccontext = ccontext_none;
335 context_cache.style = comment_style_none;
336 context_cache.scontext = '\000';
337 context_cache.depth = 0;
338 /* #### shouldn't this be checking the character's syntax instead of
339 explicitly testing for backslash characters? */
340 context_cache.backslash_p = ((pt > 1) &&
341 (BUF_FETCH_CHAR (buf, pt - 1) == '\\'));
342 /* Note that the BOL context cache may not be at the beginning
343 of the line, but that should be OK, nobody's checking. */
344 bol_context_cache = context_cache;
347 else if (pt < context_cache.cur_point)
350 stderr_out ("reset to bol\n");
351 /* bol cache is OK but current_cache is not. */
352 context_cache = bol_context_cache;
355 else if (pt <= context_cache.end_point)
358 stderr_out ("everything is OK\n");
359 /* in same top-level form. */
363 /* OK, we're past the end of the top-level form. */
364 Bufpos maxpt = max (context_cache.end_point, context_cache.cur_point);
370 stderr_out ("past end\n");
375 /* This appears to cause huge slowdowns in files which have no
378 In any case, it's not really necessary that we know for
379 sure the top-level form we're in; if we're in a form
380 but the form we have recorded is the previous one,
383 scan_buffer (buf, '\n', maxpt, pt, 1, &shortage, 1);
385 /* If there was a newline in the region past the known universe,
386 we might be inside another top-level form, so start over.
387 Otherwise, we're outside of any top-level forms and we know
388 the one directly before us, so it's OK. */
394 /* GCC 2.95.4 seems to need the casts */
395 #define SYNTAX_START_STYLE(c1, c2) \
396 ((enum comment_style) \
397 (SYNTAX_CODES_MATCH_START_P (c1, c2, SYNTAX_COMMENT_STYLE_A) ? \
399 SYNTAX_CODES_MATCH_START_P (c1, c2, SYNTAX_COMMENT_STYLE_B) ? \
403 #define SYNTAX_END_STYLE(c1, c2) \
404 ((enum comment_style) \
405 (SYNTAX_CODES_MATCH_END_P (c1, c2, SYNTAX_COMMENT_STYLE_A) ? \
407 SYNTAX_CODES_MATCH_END_P (c1, c2, SYNTAX_COMMENT_STYLE_B) ? \
411 #define SINGLE_SYNTAX_STYLE(c) \
412 ((enum comment_style) \
413 (SYNTAX_CODE_MATCHES_1CHAR_P (c, SYNTAX_COMMENT_STYLE_A) ? \
415 SYNTAX_CODE_MATCHES_1CHAR_P (c, SYNTAX_COMMENT_STYLE_B) ? \
419 /* Set up context_cache for position PT in BUF. */
422 find_context (struct buffer *buf, Bufpos pt)
424 /* This function can GC */
426 Lisp_Char_Table *mirrortab = XCHAR_TABLE (buf->mirror_syntax_table);
427 Lisp_Object syntaxtab = buf->syntax_table;
430 int prev_syncode, syncode;
432 setup_context_cache (buf, pt);
433 pt = context_cache.cur_point;
435 SCS_STATISTICS_SET_FUNCTION (scs_find_context);
436 SETUP_SYNTAX_CACHE (pt - 1, 1);
437 if (pt > BUF_BEGV (buf))
439 c = BUF_FETCH_CHAR (buf, pt - 1);
440 syncode = SYNTAX_CODE_FROM_CACHE (mirrortab, c);
444 c = '\n'; /* to get bol_context_cache at point-min */
445 syncode = Swhitespace;
448 for (; pt < target; pt++, context_cache.cur_point = pt)
450 if (context_cache.needs_its_head_reexamined)
452 if (context_cache.depth == 0
453 && context_cache.context == context_none)
455 /* We've found an anchor spot.
456 Try to put the start of defun within 6000 chars of
457 the target, and the end of defun as close as possible.
458 6000 is also arbitrary but tries to strike a balance
459 between two conflicting pulls when dealing with a
460 file that has lots of stuff sitting outside of a top-
463 a) If you move past the start of defun, you will
464 have to recompute defun, which in this case
465 means that start of defun goes all the way back
466 to the beginning of the file; so you want
467 to set start of defun a ways back from the
469 b) If you move a line backwards but within start of
470 defun, you have to move back to start of defun;
471 so you don't want start of defun too far from
474 if (target - context_cache.start_point > 6000)
475 context_cache.start_point = pt;
476 context_cache.end_point = pt;
477 bol_context_cache = context_cache;
481 UPDATE_SYNTAX_CACHE_FORWARD (pt);
483 prev_syncode = syncode;
484 c = BUF_FETCH_CHAR (buf, pt);
485 syncode = SYNTAX_CODE_FROM_CACHE (mirrortab, c);
488 bol_context_cache = context_cache;
490 if (context_cache.backslash_p)
492 context_cache.backslash_p = 0;
496 switch (SYNTAX_FROM_CACHE (mirrortab, c))
499 context_cache.backslash_p = 1;
503 if (context_cache.context == context_none)
504 context_cache.depth++;
508 if (context_cache.context == context_none)
509 context_cache.depth--;
513 if (context_cache.context == context_none)
515 context_cache.context = context_comment;
516 context_cache.ccontext = ccontext_none;
517 context_cache.style = SINGLE_SYNTAX_STYLE (syncode);
518 if (context_cache.style == comment_style_none) ABORT ();
523 if (context_cache.style != SINGLE_SYNTAX_STYLE (syncode))
525 else if (context_cache.context == context_comment)
527 context_cache.context = context_none;
528 context_cache.style = comment_style_none;
530 else if (context_cache.context == context_block_comment &&
531 (context_cache.ccontext == ccontext_start2 ||
532 context_cache.ccontext == ccontext_end1))
534 context_cache.context = context_none;
535 context_cache.ccontext = ccontext_none;
536 context_cache.style = comment_style_none;
542 if (context_cache.context == context_string &&
543 context_cache.scontext == c)
545 context_cache.context = context_none;
546 context_cache.scontext = '\000';
548 else if (context_cache.context == context_none)
550 Lisp_Object stringtermobj =
551 syntax_match (syntax_cache.current_syntax_table, c);
554 if (CHARP (stringtermobj))
555 stringterm = XCHAR (stringtermobj);
558 context_cache.context = context_string;
559 context_cache.scontext = stringterm;
560 context_cache.ccontext = ccontext_none;
567 if (context_cache.context == context_generic_comment)
569 context_cache.context = context_none;
571 else if (context_cache.context == context_none)
573 context_cache.context = context_generic_comment;
574 context_cache.ccontext = ccontext_none;
581 if (context_cache.context == context_generic_string)
583 context_cache.context = context_none;
585 else if (context_cache.context == context_none)
587 context_cache.context = context_generic_string;
588 context_cache.ccontext = ccontext_none;
597 /* That takes care of the characters with manifest syntax.
598 Now we've got to hack multi-char sequences that start
599 and end block comments.
601 if ((SYNTAX_CODE_COMMENT_BITS (syncode) &
602 SYNTAX_SECOND_CHAR_START) &&
603 context_cache.context == context_none &&
604 context_cache.ccontext == ccontext_start1 &&
605 SYNTAX_CODES_START_P (prev_syncode, syncode) /* the two chars match */
608 context_cache.ccontext = ccontext_start2;
609 context_cache.style = SYNTAX_START_STYLE (prev_syncode, syncode);
610 if (context_cache.style == comment_style_none) ABORT ();
612 else if ((SYNTAX_CODE_COMMENT_BITS (syncode) &
613 SYNTAX_FIRST_CHAR_START) &&
614 context_cache.context == context_none &&
615 (context_cache.ccontext == ccontext_none ||
616 context_cache.ccontext == ccontext_start1))
618 context_cache.ccontext = ccontext_start1;
619 context_cache.style = comment_style_none; /* should be this already*/
621 else if ((SYNTAX_CODE_COMMENT_BITS (syncode) &
622 SYNTAX_SECOND_CHAR_END) &&
623 context_cache.context == context_block_comment &&
624 context_cache.ccontext == ccontext_end1 &&
625 SYNTAX_CODES_END_P (prev_syncode, syncode) &&
626 /* the two chars match */
627 context_cache.style ==
628 SYNTAX_END_STYLE (prev_syncode, syncode)
631 context_cache.context = context_none;
632 context_cache.ccontext = ccontext_none;
633 context_cache.style = comment_style_none;
635 else if ((SYNTAX_CODE_COMMENT_BITS (syncode) &
636 SYNTAX_FIRST_CHAR_END) &&
637 context_cache.context == context_block_comment &&
639 /* #### pre-Matt code had: */
640 (context_cache.style ==
641 SYNTAX_END_STYLE (c, BUF_FETCH_CHAR (buf, pt+1))) &&
642 /* why do these differ here?! */
644 context_cache.style == SINGLE_SYNTAX_STYLE (syncode) &&
645 (context_cache.ccontext == ccontext_start2 ||
646 context_cache.ccontext == ccontext_end1))
647 /* check end1, to detect a repetition of the first char of a
648 comment-end sequence. ie, '/xxx foo xxx/' or '/xxx foo x/',
649 where 'x' = '*' -- mct */
651 if (context_cache.style == comment_style_none) ABORT ();
652 context_cache.ccontext = ccontext_end1;
655 else if (context_cache.ccontext == ccontext_start1)
657 if (context_cache.context != context_none) ABORT ();
658 context_cache.ccontext = ccontext_none;
660 else if (context_cache.ccontext == ccontext_end1)
662 if (context_cache.context != context_block_comment) ABORT ();
663 context_cache.context = context_none;
664 context_cache.ccontext = ccontext_start2;
667 if (context_cache.ccontext == ccontext_start2 &&
668 context_cache.context == context_none)
670 context_cache.context = context_block_comment;
671 if (context_cache.style == comment_style_none) ABORT ();
673 else if (context_cache.ccontext == ccontext_none &&
674 context_cache.context == context_block_comment)
676 context_cache.context = context_none;
680 context_cache.needs_its_head_reexamined = 0;
684 context_to_symbol (enum syntactic_context context)
688 case context_none: return Qnil;
689 case context_string: return Qstring;
690 case context_comment: return Qcomment;
691 case context_block_comment: return Qblock_comment;
692 case context_generic_comment: return Qblock_comment;
693 case context_generic_string: return Qstring;
694 default: ABORT (); return Qnil; /* suppress compiler warning */
698 DEFUN ("buffer-syntactic-context", Fbuffer_syntactic_context, 0, 1, 0, /*
699 Return the syntactic context of BUFFER at point.
700 If BUFFER is nil or omitted, the current buffer is assumed.
701 The returned value is one of the following symbols:
703 nil ; meaning no special interpretation
704 string ; meaning point is within a string
705 comment ; meaning point is within a line comment
706 block-comment ; meaning point is within a block comment
708 See also the function `buffer-syntactic-context-depth', which returns
709 the current nesting-depth within all parenthesis-syntax delimiters
710 and the function `syntactically-sectionize', which will map a function
711 over each syntactic context in a region.
713 WARNING: this may alter match-data.
717 /* This function can GC */
718 struct buffer *buf = decode_buffer (buffer, 0);
719 find_context (buf, BUF_PT (buf));
720 return context_to_symbol (context_cache.context);
723 DEFUN ("buffer-syntactic-context-depth", Fbuffer_syntactic_context_depth,
725 Return the depth within all parenthesis-syntax delimiters at point.
726 If BUFFER is nil or omitted, the current buffer is assumed.
727 WARNING: this may alter match-data.
731 /* This function can GC */
732 struct buffer *buf = decode_buffer (buffer, 0);
733 find_context (buf, BUF_PT (buf));
734 return make_int (context_cache.depth);
738 DEFUN ("syntactically-sectionize", Fsyntactically_sectionize, 3, 4, 0, /*
739 Call FUNCTION for each contiguous syntactic context in the region.
740 Call the given function with four arguments: the start and end of the
741 region, a symbol representing the syntactic context, and the current
742 depth (as returned by the functions `buffer-syntactic-context' and
743 `buffer-syntactic-context-depth'). When this function is called, the
744 current buffer will be set to BUFFER.
746 WARNING: this may alter match-data.
748 (function, start, end, buffer))
750 /* This function can GC */
753 enum syntactic_context this_context;
754 Lisp_Object extent = Qnil;
756 struct buffer *buf = decode_buffer (buffer, 0);
758 get_buffer_range_char (buf, start, end, &s, &e, 0);
761 find_context (buf, pt);
767 /* skip over "blank" areas, and bug out at end-of-buffer. */
768 while (context_cache.context == context_none)
771 if (pt >= e) goto DONE_LABEL;
772 find_context (buf, pt);
774 /* We've found a non-blank area; keep going until we reach its end */
775 this_context = context_cache.context;
778 /* Minor kludge: consider the comment-start character(s) a part of
781 if (this_context == context_block_comment &&
782 context_cache.ccontext == ccontext_start2)
784 else if (this_context == context_comment
785 || this_context == context_generic_comment
789 edepth = context_cache.depth;
790 while (context_cache.context == this_context && pt < e)
793 find_context (buf, pt);
798 /* Minor kludge: consider the character which terminated the comment
799 a part of the comment.
801 if ((this_context == context_block_comment ||
802 this_context == context_comment
803 || this_context == context_generic_comment
810 /* Make sure not to pass in values that are outside the
811 actual bounds of this function. */
812 call4_in_buffer (buf, function, make_int (max (s, estart)),
813 make_int (eend == e ? e : eend - 1),
814 context_to_symbol (this_context),
823 syms_of_font_lock (void)
825 defsymbol (&Qcomment, "comment");
826 defsymbol (&Qblock_comment, "block-comment");
827 defsymbol (&Qbeginning_of_defun, "beginning-of-defun");
829 DEFSUBR (Fbuffer_syntactic_context);
830 DEFSUBR (Fbuffer_syntactic_context_depth);
831 DEFSUBR (Fsyntactically_sectionize);
835 reinit_vars_of_font_lock (void)
837 xzero (context_cache);
838 xzero (bol_context_cache);
842 vars_of_font_lock (void)
844 reinit_vars_of_font_lock ();