| blob | 3b7f6e4ea0a3672bccfd00c13bf4808133aa9711 |
1 ;;; cc-engine.el --- core syntax guessing engine for CC mode
3 ;; Copyright (C) 1985, 1987, 1992, 1993, 1994, 1995, 1996, 1997, 1998,
4 ;; 1999, 2000, 2001, 2002, 2003, 2004, 2005, 2006, 2007, 2008, 2009, 2010
5 ;; Free Software Foundation, Inc.
7 ;; Authors: 2001- Alan Mackenzie
8 ;; 1998- Martin Stjernholm
9 ;; 1992-1999 Barry A. Warsaw
10 ;; 1987 Dave Detlefs
11 ;; 1987 Stewart Clamen
12 ;; 1985 Richard M. Stallman
13 ;; Maintainer: bug-cc-mode@gnu.org
14 ;; Created: 22-Apr-1997 (split from cc-mode.el)
15 ;; Version: See cc-mode.el
16 ;; Keywords: c languages oop
18 ;; This file is part of GNU Emacs.
20 ;; GNU Emacs is free software: you can redistribute it and/or modify
21 ;; it under the terms of the GNU General Public License as published by
22 ;; the Free Software Foundation, either version 3 of the License, or
23 ;; (at your option) any later version.
25 ;; GNU Emacs is distributed in the hope that it will be useful,
26 ;; but WITHOUT ANY WARRANTY; without even the implied warranty of
27 ;; MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
28 ;; GNU General Public License for more details.
30 ;; You should have received a copy of the GNU General Public License
31 ;; along with GNU Emacs. If not, see <http://www.gnu.org/licenses/>.
33 ;;; Commentary:
35 ;; The functions which have docstring documentation can be considered
36 ;; part of an API which other packages can use in CC Mode buffers.
37 ;; Otoh, undocumented functions and functions with the documentation
38 ;; in comments are considered purely internal and can change semantics
39 ;; or even disappear in the future.
40 ;;
41 ;; (This policy applies to CC Mode as a whole, not just this file. It
42 ;; probably also applies to many other Emacs packages, but here it's
43 ;; clearly spelled out.)
45 ;; Hidden buffer changes
46 ;;
47 ;; Various functions in CC Mode use text properties for caching and
48 ;; syntactic markup purposes, and those of them that might modify such
49 ;; properties but still don't modify the buffer in a visible way are
50 ;; said to do "hidden buffer changes". They should be used within
51 ;; `c-save-buffer-state' or a similar function that saves and restores
52 ;; buffer modifiedness, disables buffer change hooks, etc.
53 ;;
54 ;; Interactive functions are assumed to not do hidden buffer changes,
55 ;; except in the specific parts of them that do real changes.
56 ;;
57 ;; Lineup functions are assumed to do hidden buffer changes. They
58 ;; must not do real changes, though.
59 ;;
60 ;; All other functions that do hidden buffer changes have that noted
61 ;; in their doc string or comment.
62 ;;
63 ;; The intention with this system is to avoid wrapping every leaf
64 ;; function that do hidden buffer changes inside
65 ;; `c-save-buffer-state'. It should be used as near the top of the
66 ;; interactive functions as possible.
67 ;;
68 ;; Functions called during font locking are allowed to do hidden
69 ;; buffer changes since the font-lock package run them in a context
70 ;; similar to `c-save-buffer-state' (in fact, that function is heavily
71 ;; inspired by `save-buffer-state' in the font-lock package).
73 ;; Use of text properties
74 ;;
75 ;; CC Mode uses several text properties internally to mark up various
76 ;; positions, e.g. to improve speed and to eliminate glitches in
77 ;; interactive refontification.
78 ;;
79 ;; Note: This doc is for internal use only. Other packages should not
80 ;; assume that these text properties are used as described here.
81 ;;
82 ;; 'category
83 ;; Used for "indirection". With its help, some other property can
84 ;; be cheaply and easily switched on or off everywhere it occurs.
85 ;;
86 ;; 'syntax-table
87 ;; Used to modify the syntax of some characters. It is used to
88 ;; mark the "<" and ">" of angle bracket parens with paren syntax, and
89 ;; to "hide" obtrusive characters in preprocessor lines.
90 ;;
91 ;; This property is used on single characters and is therefore
92 ;; always treated as front and rear nonsticky (or start and end open
93 ;; in XEmacs vocabulary). It's therefore installed on
94 ;; `text-property-default-nonsticky' if that variable exists (Emacs
95 ;; >= 21).
96 ;;
97 ;; 'c-is-sws and 'c-in-sws
98 ;; Used by `c-forward-syntactic-ws' and `c-backward-syntactic-ws' to
99 ;; speed them up. See the comment blurb before `c-put-is-sws'
100 ;; below for further details.
101 ;;
102 ;; 'c-type
103 ;; This property is used on single characters to mark positions with
104 ;; special syntactic relevance of various sorts. Its primary use is
105 ;; to avoid glitches when multiline constructs are refontified
106 ;; interactively (on font lock decoration level 3). It's cleared in
107 ;; a region before it's fontified and is then put on relevant chars
108 ;; in that region as they are encountered during the fontification.
109 ;; The value specifies the kind of position:
110 ;;
111 ;; 'c-decl-arg-start
112 ;; Put on the last char of the token preceding each declaration
113 ;; inside a declaration style arglist (typically in a function
114 ;; prototype).
115 ;;
116 ;; 'c-decl-end
117 ;; Put on the last char of the token preceding a declaration.
118 ;; This is used in cases where declaration boundaries can't be
119 ;; recognized simply by looking for a token like ";" or "}".
120 ;; `c-type-decl-end-used' must be set if this is used (see also
121 ;; `c-find-decl-spots').
122 ;;
123 ;; 'c-<>-arg-sep
124 ;; Put on the commas that separate arguments in angle bracket
125 ;; arglists like C++ template arglists.
126 ;;
127 ;; 'c-decl-id-start and 'c-decl-type-start
128 ;; Put on the last char of the token preceding each declarator
129 ;; in the declarator list of a declaration. They are also used
130 ;; between the identifiers cases like enum declarations.
131 ;; 'c-decl-type-start is used when the declarators are types,
132 ;; 'c-decl-id-start otherwise.
133 ;;
134 ;; 'c-awk-NL-prop
135 ;; Used in AWK mode to mark the various kinds of newlines. See
136 ;; cc-awk.el.
138 ;;; Code:
145 load-path)))
152 ;; Silence the compiler.
155 \f
156 ;; Make declarations for all the `c-lang-defvar' variables in cc-langs.
169 \f
170 ;;; Internal state variables.
172 ;; Internal state of hungry delete key feature
176 ;; The electric flag (toggled by `c-toggle-electric-state').
177 ;; If t, electric actions (like automatic reindentation, and (if
178 ;; c-auto-newline is also set) auto newlining) will happen when an electric
179 ;; key like `{' is pressed (or an electric keyword like `else').
183 ;; Internal state of auto newline feature.
187 ;; Included in the mode line to indicate the active submodes.
188 ;; (defvar c-submode-indicators nil)
189 ;; (make-variable-buffer-local 'c-submode-indicators)
192 ;; Calculate the new state of PREVSTATE, t or nil, based on arg. If
193 ;; arg is nil or zero, toggle the state. If arg is negative, turn
194 ;; the state off, and if arg is positive, turn the state on
200 ;; Dynamically bound cache for `c-in-literal'.
203 \f
204 ;; Basic handling of preprocessor directives.
206 ;; This is a dynamically bound cache used together with
207 ;; `c-query-macro-start' and `c-query-and-set-macro-start'. It only
208 ;; works as long as point doesn't cross a macro boundary.
217 c-macro-start))
225 c-macro-start))
228 "Go to the beginning of a preprocessor directive.
229 Leave point at the beginning of the directive and return t if in one,
230 otherwise return nil and leave point unchanged.
232 Note that this function might do hidden buffer changes. See the
233 comment at the start of cc-engine.el for more info."
244 t
246 nil)))))
249 "Go to the end of a preprocessor directive.
250 More accurately, move the point to the end of the closest following
251 line that doesn't end with a line continuation backslash - no check is
252 done that the point is inside a cpp directive to begin with.
254 Note that this function might do hidden buffer changes. See the
255 comment at the start of cc-engine.el for more info."
261 t))))
264 ;; Go to the end of a CPP directive, or a "safe" pos just before.
265 ;;
266 ;; This is normally the end of the next non-escaped line. A "safe"
267 ;; position is one not within a string or comment. (The EOL on a line
268 ;; comment is NOT "safe").
269 ;;
270 ;; This function must only be called from the beginning of a CPP construct.
271 ;;
272 ;; Note that this function might do hidden buffer changes. See the comment
273 ;; at the start of cc-engine.el for more info.
285 ;; Assuming point is at the "#" that introduces a preprocessor
286 ;; directive, it's moved forward to the end of the identifier which is
287 ;; "#define"d (or whatever c-opt-cpp-macro-define specifies). Non-nil
288 ;; is returned in this case, in all other cases nil is returned and
289 ;; point isn't moved.
290 ;;
291 ;; This function might do hidden buffer changes.
297 ;; Assuming point is at the "#" that introduces a preprocessor
298 ;; directive, it's moved forward to the start of the definition body
299 ;; if it's a "#define" (or whatever c-opt-cpp-macro-define
300 ;; specifies). Non-nil is returned in this case, in all other cases
301 ;; nil is returned and point isn't moved.
302 ;;
303 ;; This function might do hidden buffer changes.
309 \f
310 ;;; Basic utility functions.
313 ;; Return the given region as a string where all syntactic
314 ;; whitespace is removed or, where necessary, replaced with a single
315 ;; space. If PAREN-LEVEL is given then all parens in the region are
316 ;; collapsed to "()", "[]" etc.
317 ;;
318 ;; This function might do hidden buffer changes.
362 ;; Shift the indentation of the current line with the specified
363 ;; amount (positive inwards). The buffer is modified only if
364 ;; SHIFT-AMT isn't equal to zero.
367 tmp-char-inserted)
369 nil
370 ;; If we're on an empty line inside a macro, we take the point
371 ;; to be at the current indentation and shift it to the
372 ;; appropriate column. This way we don't treat the extra
373 ;; whitespace out to the line continuation as indentation.
389 ;; If initial point was within line's indentation and we're not on
390 ;; a line with a line continuation in a macro, position after the
391 ;; indentation. Else stay at same point in text.
399 ;; Return non-nil if the string KEYWORD is a known keyword. More
400 ;; precisely, the value is the symbol for the keyword in
401 ;; `c-keywords-obarray'.
405 ;; Return non-nil if the symbol KEYWORD-SYM, as returned by
406 ;; `c-keyword-sym', is a member of LANG-CONSTANT, which is the name
407 ;; of a language constant that ends with "-kwds". If KEYWORD-SYM is
408 ;; nil then the result is nil.
411 ;; String syntax chars, suitable for skip-syntax-(forward|backward).
416 ;; Regexp matching string limit syntax.
421 ;; Regexp matching WS followed by string limit syntax.
425 ;; Holds formatted error strings for the few cases where parse errors
426 ;; are reported.
433 c-parsing-error)
435 ;; Faces given to comments and string literals. This is used in some
436 ;; situations to speed up recognition; it isn't mandatory that font
437 ;; locking is in use. This variable is extended with the face in
438 ;; `c-doc-face-name' when fontification is activated in cc-fonts.el.
442 ;; New in Emacs 22.
446 ;; Put a c-type property with the given value at POS.
450 ;; Remove all occurrences of the c-type property that has the given
451 ;; value in the region between FROM and TO. VALUE is assumed to not
452 ;; be nil.
453 ;;
454 ;; Note: This assumes that c-type is put on single chars only; it's
455 ;; very inefficient if matching properties cover large regions.
464 \f
465 ;; Some debug tools to visualize various special positions. This
466 ;; debug code isn't as portable as the rest of CC Mode.
502 \f
503 ;; `c-beginning-of-statement-1' and accompanying stuff.
505 ;; KLUDGE ALERT: c-maybe-labelp is used to pass information between
506 ;; c-crosses-statement-barrier-p and c-beginning-of-statement-1. A
507 ;; better way should be implemented, but this will at least shut up
508 ;; the byte compiler.
511 ;; New awk-compatible version of c-beginning-of-statement-1, ACM 2002/6/22
513 ;; Macros used internally in c-beginning-of-statement-1 for the
514 ;; automaton actions.
517 stack)))
522 t
523 ,do-if-done
528 ;; Throw nil if stack is empty, else throw non-nil.
552 noerror comma-delim)
553 "Move to the start of the current statement or declaration, or to
554 the previous one if already at the beginning of one. Only
555 statements/declarations on the same level are considered, i.e. don't
556 move into or out of sexps (not even normal expression parentheses).
558 If point is already at the earliest statement within braces or parens,
559 this function doesn't move back into any whitespace preceding it; it
560 returns 'same in this case.
564 is within the continuation. If starting at such a token, move to the
565 corresponding statement start. If at the beginning of a statement,
566 move to the closest containing statement if there is any. This might
567 also stop at a continuation clause.
569 Labels are treated as part of the following statements if
570 IGNORE-LABELS is non-nil. (FIXME: Doesn't work if we stop at a known
571 statement start keyword.) Otherwise, each label is treated as a
572 separate statement.
574 Macros are ignored \(i.e. skipped over) unless point is within one, in
575 which case the content of the macro is treated as normal code. Aside
576 from any normal statement starts found in it, stop at the first token
579 macros before leaving them.
581 Return:
583 'same if stopped at the beginning of the current statement;
584 'up if stepped to a containing statement;
585 'previous if stepped to a preceding statement;
586 'beginning if stepped from a statement continuation clause to
587 its start clause; or
588 'macro if stepped to a macro start.
589 Note that 'same and not 'label is returned if stopped at the same
590 label without crossing the colon character.
592 LIM may be given to limit the search. If the search hits the limit,
593 point will be left at the closest following token, or at the start
594 position if that is less ('same is returned in this case).
596 NOERROR turns off error logging to `c-parsing-error'.
598 Normally only ';' and virtual semicolons are considered to delimit
599 statements, but if COMMA-DELIM is non-nil then ',' is treated
600 as a delimiter too.
602 Note that this function might do hidden buffer changes. See the
603 comment at the start of cc-engine.el for more info."
605 ;; The bulk of this function is a pushdown automaton that looks at statement
606 ;; boundaries and the tokens (such as "while") in c-opt-block-stmt-key. Its
607 ;; purpose is to keep track of nested statements, ensuring that such
608 ;; statements are skipped over in their entirety (somewhat akin to what C-M-p
609 ;; does with nested braces/brackets/parentheses).
610 ;;
611 ;; Note: The position of a boundary is the following token.
612 ;;
613 ;; Beginning with the current token (the one following point), move back one
614 ;; sexp at a time (where a sexp is, more or less, either a token or the
615 ;; entire contents of a brace/bracket/paren pair). Each time a statement
616 ;; boundary is crossed or a "while"-like token is found, update the state of
617 ;; the PDA. Stop at the beginning of a statement when the stack (holding
618 ;; nested statement info) is empty and the position has been moved.
619 ;;
620 ;; The following variables constitute the PDA:
621 ;;
622 ;; sym: This is either the "while"-like token (e.g. 'for) we've just
623 ;; scanned back over, 'boundary if we've just gone back over a
624 ;; statement boundary, or nil otherwise.
625 ;; state: takes one of the values (nil else else-boundary while
626 ;; while-boundary catch catch-boundary).
627 ;; nil means "no "while"-like token yet scanned".
628 ;; 'else, for example, means "just gone back over an else".
629 ;; 'else-boundary means "just gone back over a statement boundary
630 ;; immediately after having gone back over an else".
631 ;; saved-pos: A vector of either saved positions (tok ptok pptok, etc.) or
632 ;; of error reporting information.
633 ;; stack: The stack onto which the PDA pushes its state. Each entry
634 ;; consists of a saved value of state and saved-pos. An entry is
635 ;; pushed when we move back over a "continuation" token (e.g. else)
636 ;; and popped when we encounter the corresponding opening token
637 ;; (e.g. if).
638 ;;
639 ;;
640 ;; The following diagram briefly outlines the PDA.
641 ;;
642 ;; Common state:
643 ;; "else": Push state, goto state `else'.
644 ;; "while": Push state, goto state `while'.
645 ;; "catch" or "finally": Push state, goto state `catch'.
646 ;; boundary: Pop state.
647 ;; other: Do nothing special.
648 ;;
649 ;; State `else':
650 ;; boundary: Goto state `else-boundary'.
651 ;; other: Error, pop state, retry token.
652 ;;
653 ;; State `else-boundary':
654 ;; "if": Pop state.
655 ;; boundary: Error, pop state.
656 ;; other: See common state.
657 ;;
658 ;; State `while':
659 ;; boundary: Save position, goto state `while-boundary'.
660 ;; other: Pop state, retry token.
661 ;;
662 ;; State `while-boundary':
663 ;; "do": Pop state.
664 ;; boundary: Restore position if it's not at start, pop state. [*see below]
665 ;; other: See common state.
666 ;;
667 ;; State `catch':
668 ;; boundary: Goto state `catch-boundary'.
669 ;; other: Error, pop state, retry token.
670 ;;
671 ;; State `catch-boundary':
672 ;; "try": Pop state.
673 ;; "catch": Goto state `catch'.
674 ;; boundary: Error, pop state.
675 ;; other: See common state.
676 ;;
677 ;; [*] In the `while-boundary' state, we had pushed a 'while state, and were
678 ;; searching for a "do" which would have opened a do-while. If we didn't
679 ;; find it, we discard the analysis done since the "while", go back to this
680 ;; token in the buffer and restart the scanning there, this time WITHOUT
681 ;; pushing the 'while state onto the stack.
682 ;;
683 ;; In addition to the above there is some special handling of labels
684 ;; and macros.
688 macro-start
691 c-stmt-delim-chars-with-comma
692 c-stmt-delim-chars))
693 c-in-literal-cache c-maybe-labelp after-case:-pos saved
694 ;; Current position.
695 pos
696 ;; Position of last stmt boundary character (e.g. ;).
697 boundary-pos
698 ;; The position of the last sexp or bound that follows the
699 ;; first found colon, i.e. the start of the nonlabel part of
700 ;; the statement. It's `start' if a colon is found just after
701 ;; the start.
702 after-labels-pos
703 ;; Like `after-labels-pos', but the first such position inside
704 ;; a label, i.e. the start of the last label before the start
705 ;; of the nonlabel part of the statement.
706 last-label-pos
707 ;; The last position where a label is possible provided the
708 ;; statement started there. It's nil as long as no invalid
709 ;; label content has been found (according to
710 ;; `c-nonlabel-token-key'. It's `start' if no valid label
711 ;; content was found in the label. Note that we might still
712 ;; regard it a label if it starts with `c-label-kwds'.
713 label-good-pos
714 ;; Symbol just scanned back over (e.g. 'while or 'boundary).
715 ;; See above.
716 sym
717 ;; Current state in the automaton. See above.
718 state
719 ;; Current saved positions. See above.
720 saved-pos
721 ;; Stack of conses (state . saved-pos).
722 stack
723 ;; Regexp which matches "for", "if", etc.
726 ;; Return value.
728 ;; Positions of the last three sexps or bounds we've stopped at.
729 tok ptok pptok)
739 ;; Try to skip back over unary operator characters, to register
740 ;; that we've moved.
744 ;; Protect post-++/-- operators just before a virtual semicolon.
748 ;; Skip back over any semicolon here. If it was a bare semicolon, we're
749 ;; done. Later on we ignore the boundaries for statements that don't
750 ;; contain any sexp. The only thing that is affected is that the error
751 ;; checking is a little less strict, and we really don't bother.
761 pos saved)
763 ;; Begin at start and not pos to detect macros if we stand
764 ;; directly after the #.
767 ;; Record this as the first token if not starting inside it.
770 ;; The following while loop goes back one sexp (balanced parens,
771 ;; etc. with contents, or symbol or suchlike) each iteration. This
772 ;; movement is accomplished with a call to scan-sexps approx 130 lines
773 ;; below.
791 ;; Stop at the first token in the content of the macro.
795 ret 'macro
796 ignore-labels t))
799 ;; Do a round through the automaton if we've just passed a
800 ;; statement boundary or passed a "while"-like token.
808 ;; The PDA state handling.
809 ;;
810 ;; Refer to the description of the PDA in the opening
811 ;; comments. In the following OR form, the first leaf
812 ;; attempts to handles one of the specific actions detailed
813 ;; (e.g., finding token "if" whilst in state `else-boundary').
814 ;; We drop through to the second leaf (which handles common
815 ;; state) if no specific handler is found in the first cond.
816 ;; If a parsing error is detected (e.g. an "else" with no
817 ;; preceding "if"), we throw to the enclosing catch.
818 ;;
819 ;; Note that the (eq state 'else) means
820 ;; "we've just passed an else", NOT "we're looking for an
821 ;; else".
838 ;; Since this can cause backtracking we do a
839 ;; little more careful analysis to avoid it:
840 ;; If there's a label in front of the while
841 ;; it can't be part of a do-while.
870 ;; This is state common. We get here when the previous
871 ;; cond statement found no particular state handler.
873 ;; If we have a boundary at the start
874 ;; position we push a frame to go to the
875 ;; previous statement.
884 ;; Is this a real while, or a do-while?
885 ;; The next `when' triggers unless we are SURE that
886 ;; the `while' is not the tailend of a `do-while'.
889 ;; The following kludge is to prevent
890 ;; infinite recursion when called from
891 ;; c-awk-after-if-for-while-condition-p,
892 ;; or the like.
896 ;; Since this can cause backtracking we do a
897 ;; little more careful analysis to avoid it: If
898 ;; the while isn't followed by a (possibly
899 ;; virtual) semicolon it can't be a do-while.
908 ;; We're either past a statement boundary or at the
909 ;; start of a statement, so throw away any label data
910 ;; for the previous one.
912 last-label-pos nil
913 c-maybe-labelp nil))))
915 ;; Step to the previous sexp, but not if we crossed a
916 ;; boundary, since that doesn't consume an sexp.
920 ;; HERE IS THE SINGLE PLACE INSIDE THE PDA LOOP WHERE WE MOVE
921 ;; BACKWARDS THROUGH THE SOURCE.
923 ;; This is typically fast with the caching done by
924 ;; c-(backward|forward)-sws.
928 ;; Set as long as we have to continue jumping by sexps.
929 ;; It's the position to use as end in the next round.
930 sexp-loop-continue-pos
931 ;; The end position of the area to search for statement
932 ;; barriers in this round.
935 ;; The following while goes back one sexp per iteration.
939 ;; Give up if we hit an unbalanced block. Since the
940 ;; stack won't be empty the code below will report a
941 ;; suitable error.
944 ;; Check if the sexp movement crossed a statement or
945 ;; declaration boundary. But first modify the point
946 ;; so that `c-crosses-statement-barrier-p' only looks
947 ;; at the non-sexp chars following the sexp.
950 boundary-pos
953 nil
956 ;; Set continuation position in case
957 ;; `c-crosses-statement-barrier-p'
958 ;; doesn't detect anything below.
960 ;; If the sexp movement took us into a
961 ;; macro then there were only some non-sexp
962 ;; chars after it. Skip out of the macro
963 ;; to analyze them but not the non-sexp
964 ;; chars that might be inside the macro.
972 ;; Passed a block sexp. That's a boundary
973 ;; alright.
977 ;; Passed some other paren. Only analyze
978 ;; the non-sexp chars after it.
980 before-sws-pos)))
981 ;; We're at a valid token start position
982 ;; (outside the `save-excursion') if
983 ;; `c-crosses-statement-barrier-p' failed.
988 ;; Passed a symbol sexp or line
989 ;; continuation. It doesn't matter that
990 ;; it's included in the analyzed region.
993 t
994 ;; If it was a line continuation then we
995 ;; have to continue looping.
998 nil))))
1001 ptok tok
1002 tok boundary-pos
1004 ;; Like a C "continue". Analyze the next sexp.
1010 sexp-loop-continue-pos nil))))
1012 ;; ObjC method def?
1019 ;; Handle labels.
1022 ;; `c-crosses-statement-barrier-p' has found a colon, so we
1023 ;; might be in a label now. Have we got a real label
1024 ;; (including a case label) or something like C++'s "public:"?
1025 ;; A case label might use an expression rather than a token.
1037 ; been found.
1039 ;; We're in a potential label and it's the first
1040 ;; time we've found something that isn't allowed in
1041 ;; one.
1044 ;; We've moved back by a sexp, so update the token positions.
1046 pptok ptok
1047 ptok tok
1051 ;; If the stack isn't empty there might be errors to report.
1060 ;; Need to investigate closer whether we've crossed
1061 ;; between a substatement and its containing statement.
1063 ptok
1064 pptok))
1071 after-labels-pos
1077 last-label-pos
1078 pos))
1080 ;; We're in a label. Maybe we should step to the statement
1081 ;; after it.
1086 ;; Might have jumped over several labels. Go to the last one.
1089 ;; Have we got "case <expression>:"?
1099 ;; Skip over the unary operators that can start the statement.
1102 ;; protect AWK post-inc/decrement operators, etc.
1107 ret)))
1110 "Return non-nil if buffer positions FROM to TO cross one or more
1111 statement or declaration boundaries. The returned value is actually
1112 the position of the earliest boundary char. FROM must not be within
1113 a string or comment.
1115 The variable `c-maybe-labelp' is set to the position of the first `:' that
1116 might start a label (i.e. not part of `::' and not preceded by `?'). If a
1117 single `?' is found, then `c-maybe-labelp' is cleared.
1119 For AWK, a statement which is terminated by an EOL (not a \; or a }) is
1121 the line. If this virtual semicolon is _at_ from, the function recognizes it.
1123 Note that this function might do hidden buffer changes. See the
1124 comment at the start of cc-engine.el for more info."
1126 lit-range)
1139 ;; Ignore scope operators.
1143 ;; A question mark. Can't be a label, so stop
1144 ;; looking for more : and ?.
1155 ;; In trailing space after an as yet undetected virtual semicolon?
1160 nil)))))
1163 "Return non-nil if the point is at the first token in a statement
1164 or somewhere in the syntactic whitespace before it.
1167 Any kind of declaration-like construct that occur outside function
1170 Note that this function might do hidden buffer changes. See the
1171 comment at the start of cc-engine.el for more info."
1175 c-maybe-labelp)
1186 "Return non-nil if the point is at the first token in an expression or
1187 statement, or somewhere in the syntactic whitespace before it.
1190 grammar sense: It's any sequence of expression tokens except commas,
1191 unless they are enclosed inside parentheses of some kind. Also, an
1192 expression never continues past an enclosing parenthesis, but it might
1193 contain parenthesis pairs of any sort except braces.
1195 Since expressions never cross statement boundaries, this function also
1196 recognizes statement beginnings, just like `c-at-statement-start-p'.
1198 Note that this function might do hidden buffer changes. See the
1199 comment at the start of cc-engine.el for more info."
1204 c-maybe-labelp)
1212 \f
1213 ;; A set of functions that covers various idiosyncrasies in
1214 ;; implementations of `forward-comment'.
1216 ;; Note: Some emacsen considers incorrectly that any line comment
1217 ;; ending with a backslash continues to the next line. I can't think
1218 ;; of any way to work around that in a reliable way without changing
1219 ;; the buffer, though. Suggestions welcome. ;) (No, temporarily
1220 ;; changing the syntax for backslash doesn't work since we must treat
1221 ;; escapes in string literals correctly.)
1224 "Move forward past whitespace and the closest following comment, if any.
1225 Return t if a comment was found, nil otherwise. In either case, the
1226 point is moved past the following whitespace. Line continuations,
1227 i.e. a backslashes followed by line breaks, are treated as whitespace.
1228 The line breaks that end line comments are considered to be the
1229 comment enders, so the point will be put on the beginning of the next
1230 line if it moved past a line comment.
1232 This function does not do any hidden buffer changes."
1240 ;; Some emacsen (e.g. XEmacs 21) return t when moving
1241 ;; forwards at eob.
1242 nil
1244 ;; Emacs includes the ending newline in a b-style (c++)
1245 ;; comment, but XEmacs doesn't. We depend on the Emacs
1246 ;; behavior (which also is symmetric).
1250 t))))
1253 "Move forward past all following whitespace and comments.
1254 Line continuations, i.e. a backslashes followed by line breaks, are
1255 treated as whitespace.
1257 Note that this function might do hidden buffer changes. See the
1258 comment at the start of cc-engine.el for more info."
1261 ;; If forward-comment in at least XEmacs 21 is given a large
1262 ;; positive value, it'll loop all the way through if it hits
1263 ;; eob.
1265 ;; Some emacsen (e.g. XEmacs 21) return t when moving
1266 ;; forwards at eob.
1271 t))))
1274 "Move backward past whitespace and the closest preceding comment, if any.
1275 Return t if a comment was found, nil otherwise. In either case, the
1276 point is moved past the preceding whitespace. Line continuations,
1277 i.e. a backslashes followed by line breaks, are treated as whitespace.
1278 The line breaks that end line comments are considered to be the
1279 comment enders, so the point cannot be at the end of the same line to
1280 move over a line comment.
1282 This function does not do any hidden buffer changes."
1285 ;; When we got newline terminated comments, forward-comment in all
1286 ;; supported emacsen so far will stop at eol of each line not
1287 ;; ending with a comment when moving backwards. This corrects for
1288 ;; that, and at the same time handles line continuations.
1296 ;; Some emacsen (e.g. Emacs 19.34) return t when moving
1297 ;; backwards at bob.
1298 nil
1300 ;; Leave point after the closest following newline if we've
1301 ;; backed up over any above, since forward-comment won't move
1302 ;; backward over a line comment if point is at the end of the
1303 ;; same line.
1308 ;; If forward-comment above succeeded and we're at eol
1309 ;; then the newline we moved over above didn't end a
1310 ;; line comment, so we give it another go.
1312 t))
1314 ;; Emacs <= 20 and XEmacs move back over the closer of a
1315 ;; block comment that lacks an opener.
1318 t)))))
1321 "Move backward past all preceding whitespace and comments.
1322 Line continuations, i.e. a backslashes followed by line breaks, are
1323 treated as whitespace. The line breaks that end line comments are
1324 considered to be the comment enders, so the point cannot be at the end
1325 of the same line to move over a line comment. Unlike
1326 c-backward-syntactic-ws, this function doesn't move back over
1327 preprocessor directives.
1329 Note that this function might do hidden buffer changes. See the
1330 comment at the start of cc-engine.el for more info."
1334 ;; `forward-comment' in some emacsen (e.g. XEmacs 21.4)
1335 ;; return t when moving backwards at bob.
1340 ;; Emacs <= 20 and XEmacs move back over the
1341 ;; closer of a block comment that lacks an opener.
1343 t)
1345 ;; XEmacs treats line continuations as whitespace but
1346 ;; only in the backward direction, which seems a bit
1347 ;; odd. Anyway, this is necessary for Emacs.
1352 t))))))
1354 \f
1355 ;; Tools for skipping over syntactic whitespace.
1357 ;; The following functions use text properties to cache searches over
1358 ;; large regions of syntactic whitespace. It works as follows:
1359 ;;
1360 ;; o If a syntactic whitespace region contains anything but simple
1361 ;; whitespace (i.e. space, tab and line breaks), the text property
1362 ;; `c-in-sws' is put over it. At places where we have stopped
1363 ;; within that region there's also a `c-is-sws' text property.
1364 ;; That since there typically are nested whitespace inside that
1365 ;; must be handled separately, e.g. whitespace inside a comment or
1366 ;; cpp directive. Thus, from one point with `c-is-sws' it's safe
1367 ;; to jump to another point with that property within the same
1368 ;; `c-in-sws' region. It can be likened to a ladder where
1369 ;; `c-in-sws' marks the bars and `c-is-sws' the rungs.
1370 ;;
1371 ;; o The `c-is-sws' property is put on the simple whitespace chars at
1372 ;; a "rung position" and also maybe on the first following char.
1373 ;; As many characters as can be conveniently found in this range
1374 ;; are marked, but no assumption can be made that the whole range
1375 ;; is marked (it could be clobbered by later changes, for
1376 ;; instance).
1377 ;;
1378 ;; Note that some part of the beginning of a sequence of simple
1379 ;; whitespace might be part of the end of a preceding line comment
1380 ;; or cpp directive and must not be considered part of the "rung".
1381 ;; Such whitespace is some amount of horizontal whitespace followed
1382 ;; by a newline. In the case of cpp directives it could also be
1383 ;; two newlines with horizontal whitespace between them.
1384 ;;
1385 ;; The reason to include the first following char is to cope with
1386 ;; "rung positions" that doesn't have any ordinary whitespace. If
1387 ;; `c-is-sws' is put on a token character it does not have
1388 ;; `c-in-sws' set simultaneously. That's the only case when that
1389 ;; can occur, and the reason for not extending the `c-in-sws'
1390 ;; region to cover it is that the `c-in-sws' region could then be
1391 ;; accidentally merged with a following one if the token is only
1392 ;; one character long.
1393 ;;
1394 ;; o On buffer changes the `c-in-sws' and `c-is-sws' properties are
1395 ;; removed in the changed region. If the change was inside
1396 ;; syntactic whitespace that means that the "ladder" is broken, but
1397 ;; a later call to `c-forward-sws' or `c-backward-sws' will use the
1398 ;; parts on either side and use an ordinary search only to "repair"
1399 ;; the gap.
1400 ;;
1401 ;; Special care needs to be taken if a region is removed: If there
1402 ;; are `c-in-sws' on both sides of it which do not connect inside
1403 ;; the region then they can't be joined. If e.g. a marked macro is
1404 ;; broken, syntactic whitespace inside the new text might be
1405 ;; marked. If those marks would become connected with the old
1406 ;; `c-in-sws' range around the macro then we could get a ladder
1407 ;; with one end outside the macro and the other at some whitespace
1408 ;; within it.
1409 ;;
1410 ;; The main motivation for this system is to increase the speed in
1411 ;; skipping over the large whitespace regions that can occur at the
1412 ;; top level in e.g. header files that contain a lot of comments and
1413 ;; cpp directives. For small comments inside code it's probably
1414 ;; slower than using `forward-comment' straightforwardly, but speed is
1415 ;; not a significant factor there anyway.
1417 ; (defface c-debug-is-sws-face
1418 ; '((t (:background "GreenYellow")))
1419 ; "Debug face to mark the `c-is-sws' property.")
1420 ; (defface c-debug-in-sws-face
1421 ; '((t (:underline t)))
1422 ; "Debug face to mark the `c-in-sws' property.")
1424 ; (defun c-debug-put-sws-faces ()
1425 ; ;; Put the sws debug faces on all the `c-is-sws' and `c-in-sws'
1426 ; ;; properties in the buffer.
1427 ; (interactive)
1428 ; (save-excursion
1429 ; (c-save-buffer-state (in-face)
1430 ; (goto-char (point-min))
1431 ; (setq in-face (if (get-text-property (point) 'c-is-sws)
1432 ; (point)))
1433 ; (while (progn
1434 ; (goto-char (next-single-property-change
1435 ; (point) 'c-is-sws nil (point-max)))
1436 ; (if in-face
1437 ; (progn
1438 ; (c-debug-add-face in-face (point) 'c-debug-is-sws-face)
1439 ; (setq in-face nil))
1440 ; (setq in-face (point)))
1441 ; (not (eobp))))
1442 ; (goto-char (point-min))
1443 ; (setq in-face (if (get-text-property (point) 'c-in-sws)
1444 ; (point)))
1445 ; (while (progn
1446 ; (goto-char (next-single-property-change
1447 ; (point) 'c-in-sws nil (point-max)))
1448 ; (if in-face
1449 ; (progn
1450 ; (c-debug-add-face in-face (point) 'c-debug-in-sws-face)
1451 ; (setq in-face nil))
1452 ; (setq in-face (point)))
1453 ; (not (eobp)))))))
1456 ;;`(message ,@args)
1457 )
1460 ;; This macro does a hidden buffer change.
1467 ;; This macro does a hidden buffer change.
1474 ;; This macro does a hidden buffer change.
1481 ;; This macro does a hidden buffer change.
1488 ;; This macro does a hidden buffer change.
1496 ;; Called from `after-change-functions'. Note that if
1497 ;; `c-forward-sws' or `c-backward-sws' are used outside
1498 ;; `c-save-buffer-state' or similar then this will remove the cache
1499 ;; properties right after they're added.
1500 ;;
1501 ;; This function does hidden buffer changes.
1504 ;; Adjust the end to remove the properties in any following simple
1505 ;; ws up to and including the next line break, if there is any
1506 ;; after the changed region. This is necessary e.g. when a rung
1507 ;; marked empty line is converted to a line comment by inserting
1508 ;; "//" before the line break. In that case the line break would
1509 ;; keep the rung mark which could make a later `c-backward-sws'
1510 ;; move into the line comment instead of over it.
1520 ;; Ensure that an `c-in-sws' range gets broken. Note that it isn't
1521 ;; safe to keep a range that was continuous before the change. E.g:
1522 ;;
1523 ;; #define foo
1524 ;; \
1525 ;; bar
1526 ;;
1527 ;; There can be a "ladder" between "#" and "b". Now, if the newline
1528 ;; after "foo" is removed then "bar" will become part of the cpp
1529 ;; directive instead of a syntactically relevant token. In that
1530 ;; case there's no longer syntactic ws from "#" to "b".
1537 ;; Used by `c-forward-syntactic-ws' to implement the unbounded search.
1538 ;;
1539 ;; This function might do hidden buffer changes.
1542 ;; unmarked part of the simple ws region.
1544 rung-is-marked next-rung-is-marked simple-ws-end
1545 ;; `safe-start' is set when it's safe to cache the start position.
1546 ;; It's not set if we've initially skipped over comments and line
1547 ;; continuations since we might have gone out through the end of a
1548 ;; macro then. This provision makes `c-forward-sws' not populate the
1549 ;; cache in the majority of cases, but otoh is `c-backward-sws' by far
1550 ;; more common.
1551 safe-start)
1553 ;; Skip simple ws and do a quick check on the following character to see
1554 ;; if it's anything that can't start syntactic ws, so we can bail out
1555 ;; early in the majority of cases when there just are a few ws chars.
1562 ;; Find the last rung position to avoid setting properties in all
1563 ;; the cases when the marked rung is complete.
1564 ;; (`next-single-property-change' is certain to move at least one
1565 ;; step forward.)
1568 ;; Got no marked rung here. Since the simple ws might have started
1569 ;; inside a line comment or cpp directive we must set `rung-pos' as
1570 ;; high as possible.
1579 ;; The following search is the main reason that `c-in-sws'
1580 ;; and `c-is-sws' aren't combined to one property.
1584 ;; If the `c-in-sws' region extended past the last
1585 ;; `c-is-sws' char we have to go back a bit.
1592 "c-forward-sws cached move %s -> %s (max %s)"
1599 ;; We'll loop here if there is simple ws after the last rung.
1600 ;; That means that there's been some change in it and it's
1601 ;; possible that we've stepped into another ladder, so extend
1602 ;; the previous one to join with it if there is one, and try to
1603 ;; use the cache again.
1605 "c-forward-sws extending rung with [%s..%s] (max %s)"
1608 ;; Remove any `c-in-sws' property from the last char of
1609 ;; the rung before we mark it with `c-is-sws', so that we
1610 ;; won't connect with the remains of a broken "ladder".
1616 last-put-in-sws-pos rung-pos)))
1623 ;; Skipped over comments. Don't cache at eob in case the buffer
1624 ;; is narrowed.
1635 ;; Skip a preprocessor directive.
1642 ;; Don't cache at eob in case the buffer is narrowed.
1645 ;; We've searched over a piece of non-white syntactic ws. See if this
1646 ;; can be cached.
1652 ;; Cache if we haven't skipped comments only, and if we started
1653 ;; either from a marked rung or from a completely uncached
1654 ;; position.
1659 ;; See if there's a marked rung in the encountered simple ws. If
1660 ;; so then we can cache, unless `safe-start' is nil. Even then
1661 ;; we need to do this to check if the cache can be used for the
1662 ;; next step.
1666 safe-start))
1670 "c-forward-sws caching [%s..%s] - [%s..%s] (max %s)"
1674 ;; Remove the properties for any nested ws that might be cached.
1675 ;; Only necessary for `c-is-sws' since `c-in-sws' will be set
1676 ;; anyway.
1684 last-put-in-sws-pos rung-pos))
1686 ;; Remove any `c-in-sws' property from the last char of
1687 ;; the rung before we mark it with `c-is-sws', so that we
1688 ;; won't connect with the remains of a broken "ladder".
1691 rung-end-pos))
1694 "c-forward-sws not caching [%s..%s] - [%s..%s] (max %s)"
1698 ;; Set `rung-pos' for the next rung. It's the same thing here as
1699 ;; initially, except that the rung position is set as early as
1700 ;; possible since we can't be in the ending ws of a line comment or
1701 ;; cpp directive now.
1708 ;; Make sure that the newly marked `c-in-sws' region doesn't connect to
1709 ;; another one after the point (which might occur when editing inside a
1710 ;; comment or macro).
1714 "c-forward-sws clearing at %s for cache separation"
1715 last-put-in-sws-pos)
1719 ;; If at eob we have to clear the last character before the end
1720 ;; instead since the buffer might be narrowed and there might
1721 ;; be a `c-in-sws' after (point-max). In this case it's
1722 ;; necessary to clear both properties.
1724 "c-forward-sws clearing thoroughly at %s for cache separation"
1727 last-put-in-sws-pos))))
1728 )))
1731 ;; Used by `c-backward-syntactic-ws' to implement the unbounded search.
1732 ;;
1733 ;; This function might do hidden buffer changes.
1736 ;; part of the simple ws region.
1738 rung-is-marked simple-ws-beg cmt-skip-pos)
1740 ;; Skip simple horizontal ws and do a quick check on the preceding
1741 ;; character to see if it's anying that can't end syntactic ws, so we can
1742 ;; bail out early in the majority of cases when there just are a few ws
1743 ;; chars. Newlines are complicated in the backward direction, so we can't
1744 ;; skip over them.
1751 ;; Try to find a rung position in the simple ws preceding point, so that
1752 ;; we can get a cache hit even if the last bit of the simple ws has
1753 ;; changed recently.
1759 ;; `rung-pos' will be the earliest marked position, which means that
1760 ;; there might be later unmarked parts in the simple ws region.
1761 ;; It's not worth the effort to fix that; the last part of the
1762 ;; simple ws is also typically edited often, so it could be wasted.
1773 ;; The following search is the main reason that `c-in-sws'
1774 ;; and `c-is-sws' aren't combined to one property.
1778 ;; If the `c-in-sws' region extended past the first
1779 ;; `c-is-sws' char we have to go forward a bit.
1784 "c-backward-sws cached move %s <- %s (min %s)"
1796 t
1798 nil))
1800 ;; We'll loop here if there is simple ws before the first rung.
1801 ;; That means that there's been some change in it and it's
1802 ;; possible that we've stepped into another ladder, so extend
1803 ;; the previous one to join with it if there is one, and try to
1804 ;; use the cache again.
1806 "c-backward-sws extending rung with [%s..%s] (min %s)"
1809 ;; Remove any `c-in-sws' property from the last char of
1810 ;; the rung before we mark it with `c-is-sws', so that we
1811 ;; won't connect with the remains of a broken "ladder".
1814 rung-pos)
1818 last-put-in-sws-pos rung-pos))
1827 ;; Inside a cpp directive. See if it should be skipped over.
1830 ;; Move back over all line continuations in the region skipped
1831 ;; over by `c-backward-comments'. If we go past it then we
1832 ;; started inside the cpp directive.
1841 ;; Don't move past the cpp directive if we began inside
1842 ;; it. Note that the position at the end of the last line
1843 ;; of the macro is also considered to be within it.
1845 nil)
1847 ;; It's worthwhile to spend a little bit of effort on finding
1848 ;; the end of the macro, to get a good `simple-ws-beg'
1849 ;; position for the cache. Note that `c-backward-comments'
1850 ;; could have stepped over some comments before going into
1851 ;; the macro, and then `simple-ws-beg' must be kept on the
1852 ;; same side of those comments.
1859 ;; Might happen if comments after the macro were skipped
1860 ;; over.
1864 t)))
1869 simple-ws-beg)
1870 ;; Skipped over comments. Must put point at the end of
1871 ;; the simple ws at point since we might be after a line
1872 ;; comment or cpp directive that's been partially
1873 ;; narrowed out, and we can't risk marking the simple ws
1874 ;; at the end of it.
1876 t)))
1878 ;; We've searched over a piece of non-white syntactic ws. See if this
1879 ;; can be cached.
1884 ;; Cache if we started either from a marked rung or from a
1885 ;; completely uncached position.
1886 rung-is-marked
1889 ;; Cache if there's a marked rung in the encountered simple ws.
1897 "c-backward-sws caching [%s..%s] - [%s..%s] (min %s)"
1902 ;; Remove the properties for any nested ws that might be cached.
1903 ;; Only necessary for `c-is-sws' since `c-in-sws' will be set
1904 ;; anyway.
1909 ;; Remove any `c-in-sws' property from the last char of
1910 ;; the rung before we mark it with `c-is-sws', so that we
1911 ;; won't connect with the remains of a broken "ladder".
1914 rung-end-pos)
1917 last-put-in-sws-pos simple-ws-beg)
1918 rung-pos)
1923 "c-backward-sws not caching [%s..%s] - [%s..%s] (min %s)"
1929 ))
1931 ;; Make sure that the newly marked `c-in-sws' region doesn't connect to
1932 ;; another one before the point (which might occur when editing inside a
1933 ;; comment or macro).
1937 "c-backward-sws clearing at %s for cache separation"
1940 last-put-in-sws-pos))
1942 ;; If at bob and the buffer is narrowed, we have to clear the
1943 ;; character we're standing on instead since there might be a
1944 ;; `c-in-sws' before (point-min). In this case it's necessary
1945 ;; to clear both properties.
1947 "c-backward-sws clearing thoroughly at %s for cache separation"
1948 last-put-in-sws-pos)
1951 )))
1953 \f
1954 ;; Other whitespace tools
1956 ;; Is the region (beg end) WS, and is there WS (or BOB/EOB) next to the
1957 ;; region? This is a "heuristic" function. .....
1958 ;;
1959 ;; The motivation for the second bit is to check whether removing this
1960 ;; region would coalesce two symbols.
1961 ;;
1962 ;; FIXME!!! This function doesn't check virtual semicolons in any way. Be
1963 ;; careful about using this function for, e.g. AWK. (2007/3/7)
1972 \f
1973 ;; A system for finding noteworthy parens before the point.
1976 ;; A maximum comfortable scanning distance, e.g. between
1977 ;; `c-state-cache-good-pos' and "HERE" (where we call c-parse-state). When
1978 ;; this distance is exceeded, we take "emergency meausures", e.g. by clearing
1979 ;; the cache and starting again from point-min or a beginning of defun. This
1980 ;; value can be tuned for efficiency or set to a lower value for testing.
1984 ;; The state cache used by `c-parse-state' to cut down the amount of
1985 ;; searching. It's the result from some earlier `c-parse-state' call. See
1986 ;; `c-parse-state''s doc string for details of its structure.
1987 ;;
1988 ;; The use of the cached info is more effective if the next
1989 ;; `c-parse-state' call is on a line close by the one the cached state
1990 ;; was made at; the cache can actually slow down a little if the
1991 ;; cached state was made very far back in the buffer. The cache is
1992 ;; most effective if `c-parse-state' is used on each line while moving
1993 ;; forward.
1997 ;; This is a position where `c-state-cache' is known to be correct, or
1998 ;; nil (see below). It's a position inside one of the recorded unclosed
1999 ;; parens or the top level, but not further nested inside any literal or
2000 ;; subparen that is closed before the last recorded position.
2001 ;;
2002 ;; The exact position is chosen to try to be close to yet earlier than
2003 ;; the position where `c-state-cache' will be called next. Right now
2004 ;; the heuristic is to set it to the position after the last found
2005 ;; closing paren (of any type) before the line on which
2006 ;; `c-parse-state' was called. That is chosen primarily to work well
2007 ;; with refontification of the current line.
2008 ;;
2009 ;; 2009-07-28: When `c-state-point-min' and the last position where
2010 ;; `c-parse-state' or for which `c-invalidate-state-cache' was called, are
2011 ;; both in the same literal, there is no such "good position", and
2012 ;; c-state-cache-good-pos is then nil. This is the ONLY circumstance in which
2013 ;; it can be nil. In this case, `c-state-point-min-literal' will be non-nil.
2014 ;;
2015 ;; 2009-06-12: In a brace desert, c-state-cache-good-pos may also be in
2016 ;; the middle of the desert, as long as it is not within a brace pair
2017 ;; recorded in `c-state-cache' or a paren/bracket pair.
2020 ;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;
2021 ;; We maintain a simple cache of positions which aren't in a literal, so as to
2022 ;; speed up testing for non-literality.
2024 ;; The approximate interval between entries in `c-state-nonlit-pos-cache'.
2028 ;; A list of buffer positions which are known not to be in a literal. This is
2029 ;; ordered with higher positions at the front of the list. Only those which
2030 ;; are less than `c-state-nonlit-pos-cache-limit' are valid.
2034 ;; An upper limit on valid entries in `c-state-nonlit-pos-cache'. This is
2035 ;; reduced by buffer changes, and increased by invocations of
2036 ;; `c-state-literal-at'.
2039 ;; Do a parse-partial-sexp from FROM to TO, returning the bounds of any
2040 ;; literal at TO as a cons, otherwise NIL.
2041 ;; FROM must not be in a literal, and the buffer should already be wide
2042 ;; enough.
2047 nil ; TARGETDEPTH
2048 nil ; STOPBEFORE
2049 s ; OLDSTATE
2054 ;; If position HERE is inside a literal, return (START . END), the
2055 ;; boundaries of the literal (which may be outside the accessible bit of the
2056 ;; buffer). Otherwise, return nil.
2057 ;;
2058 ;; This function is almost the same as `c-literal-limits'. It differs in
2059 ;; that it is a lower level function, and that it rigourously follows the
2060 ;; syntax from BOB, whereas `c-literal-limits' uses a "local" safe position.
2065 pos npos lit)
2066 ;; Trim the cache to take account of buffer changes.
2076 here)
2085 lit))))
2088 ;; Return the start of the literal containing POS, or POS itself.
2090 pos))
2093 ;; Return a position outside of a string/comment at or before POS.
2094 ;; STATE is the parse-partial-sexp state at POS.
2098 pos))
2101 ;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;
2102 ;; Stuff to do with point-min, and coping with any literal there.
2105 ;; This is (point-min) when `c-state-cache' was last calculated. A change of
2106 ;; narrowing is likely to affect the parens that are visible before the point.
2112 ;; These two variables define the literal, if any, containing point-min.
2113 ;; Their values are, respectively, 'string, c, or c++, and the start of the
2114 ;; literal. If there's no literal there, they're both nil.
2118 ;; This is the earliest buffer-pos from which scanning can be done. It is
2119 ;; either the end of the literal containing point-min, or point-min itself.
2120 ;; It becomes nil if the buffer is changed earlier than this point.
2122 ;; Return the lowest valid scanning pos. This will be the end of the
2123 ;; literal enclosing point-min, or point-min itself.
2135 ;; Determine the properties of any literal containing POINT-MIN, setting the
2136 ;; variables `c-state-point-min-lit-type', `c-state-point-min-lit-start',
2137 ;; and `c-state-min-scan-pos' accordingly. The return value is meaningless.
2139 lit)
2154 c-state-point-min-lit-start nil
2155 c-state-min-scan-pos p-min)))))
2158 ;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;
2159 ;; A variable which signals a brace dessert - helpful for reducing the number
2160 ;; of fruitless backward scans.
2163 ;; Used only in `c-append-lower-brace-pair-to-state-cache'. It is set when an
2164 ;; that defun has searched backwards for a brace pair and not found one. Its
2165 ;; value is either nil or a cons (PA . FROM), where PA is the position of the
2166 ;; enclosing opening paren/brace/bracket which bounds the backwards search (or
2167 ;; nil when at top level) and FROM is where the backward search started. It
2168 ;; is reset to nil in `c-invalidate-state-cache'.
2171 ;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;
2172 ;; Lowish level functions/macros which work directly on `c-state-cache', or a
2173 ;; list of like structure.
2175 ;; Return the address of the top left brace/bracket/paren recorded in CACHE
2176 ;; (default `c-state-cache') (or nil).
2183 ;; Return the address of the latest brace/bracket/paren (whether left or
2184 ;; right) recorded in CACHE (default `c-state-cache') or nil.
2191 ;; Return the position just after the latest brace/bracket/paren (whether
2192 ;; left or right) recorded in CACHE (default `c-state-cache') or nil.
2200 ;; From the state-cache, determine the buffer position from which we might
2201 ;; scan forward to HERE to update this cache. This position will be just
2202 ;; after a paren/brace/bracket recorded in the cache, if possible, otherwise
2203 ;; return the earliest position in the accessible region which isn't within
2204 ;; a literal. If the visible portion of the buffer is entirely within a
2205 ;; literal, return NIL.
2207 ;(while (>= (or (c-state-cache-top-lparen c) 1) here)
2223 ;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;
2224 ;; Variables which keep track of preprocessor constructs.
2229 ;; These are the limits of the macro containing point at the previous call of
2230 ;; `c-parse-state', or nil.
2232 ;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;
2233 ;; Defuns which analyse the buffer, yet don't change `c-state-cache'.
2235 ;; Return a start position for building `c-state-cache' from
2236 ;; scratch. This will be at the top level, 2 defuns back.
2238 ;; Go back 2 bods, but ignore any bogus positions returned by
2239 ;; beginning-of-defun (i.e. open paren in column zero).
2249 ;; Return the position of the opening paren/brace/bracket before HERE which
2250 ;; matches the outermost close p/b/b between HERE and TOP, like this:
2251 ;;
2252 ;; ......................................
2253 ;; | |
2254 ;; ( [ ( ........... ) ( ) ] )
2255 ;; ^ ^ ^
2256 ;; | | |
2257 ;; return HERE TOP
2258 ;;
2259 ;; If there aren't enough opening paren/brace/brackets, return the position
2260 ;; of the outermost one found, or HERE it there are none. If there are no
2261 ;; closeing p/b/bs between HERE and TOP, return HERE. HERE and TOP must not
2262 ;; be inside literals. Only the accessible portion of the buffer will be
2263 ;; scanned.
2265 ;; PART 1: scan from `here' up to `top', accumulating ")"s which enclose
2266 ;; `here'. Go round the next loop each time we pass over such a ")". These
2267 ;; probably match "("s before `here'.
2279 ;; PART 2: Scan back before `here' searching for the "("s
2280 ;; matching/mismatching the ")"s found above. We only need to direct the
2281 ;; caller to scan when we've encountered unmatched right parens.
2290 )
2291 pos))
2294 ;; Determine the scanning strategy for adjusting `c-parse-state', attempting
2295 ;; to minimise the amount of scanning. HERE is the pertinent position in
2296 ;; the buffer, GOOD-POS is a position where `c-state-cache' (possibly with
2297 ;; its head trimmed) is known to be good, or nil if there is no such
2298 ;; position.
2299 ;;
2300 ;; The return value is a list, one of the following:
2301 ;;
2302 ;; o - ('forward CACHE-POS START-POINT) - scan forward from START-POINT,
2303 ;; which is not less than CACHE-POS.
2304 ;; o - ('backward CACHE-POS nil) - scan backwards (from HERE).
2305 ;; o - ('BOD nil START-POINT) - scan forwards from START-POINT, which is at the
2306 ;; top level.
2307 ;; o - ('IN-LIT nil nil) - point is inside the literal containing point-min.
2308 ;; , where CACHE-POS is the highest position recorded in `c-state-cache' at
2309 ;; or below HERE.
2311 BOD-pos ; position of 2nd BOD before HERE.
2312 strategy ; 'forward, 'backward, 'BOD, or 'IN-LIT.
2313 start-point
2319 start-point nil
2320 cache-pos nil
2332 start-point cache-pos)))
2334 ;; Might we be better off starting from the top level, two defuns back,
2335 ;; instead?
2340 start-point BOD-pos)))
2343 strategy
2348 ;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;
2349 ;; Routines which change `c-state-cache' and associated values.
2351 ;; The region (more precisely, point-min) has changed since we
2352 ;; calculated `c-state-cache'. Amend `c-state-cache' accordingly.
2354 ;; If point-min has MOVED BACKWARDS then we drop the state completely.
2355 ;; It would be possible to do a better job here and recalculate the top
2356 ;; only.
2360 c-state-cache-good-pos c-state-min-scan-pos
2361 c-state-brace-pair-desert nil))
2363 ;; point-min has MOVED FORWARD.
2365 ;; Is the new point-min inside a (different) literal?
2370 ;; Cut off a bit of the tail from `c-state-cache'.
2372 pa)
2380 c-state-cache-good-pos c-state-min-scan-pos)
2383 )))
2388 ;; If there is a brace pair preceding FROM in the buffer (not necessarily
2389 ;; immediately preceding), push a cons onto `c-state-cache' to represent it.
2390 ;; FROM must not be inside a literal. If UPPER-LIM is non-nil, we append
2391 ;; the highest brace pair whose "}" is below UPPER-LIM.
2392 ;;
2393 ;; Return non-nil when this has been done.
2394 ;;
2395 ;; This routine should be fast. Since it can get called a LOT, we maintain
2396 ;; `c-state-brace-pair-desert', a small cache of "failures", such that we
2397 ;; reduce the time wasted in repeated fruitless searches in brace deserts.
2401 new-cons
2409 ;; If we're essentially repeating a fruitless search, just give up.
2413 ;; Only search what we absolutely need to:
2418 ;; In the next pair of nested loops, the inner one moves back past a
2419 ;; pair of (mis-)matching parens or brackets; the outer one moves
2420 ;; back over a sequence of unmatched close brace/paren/bracket each
2421 ;; time round.
2435 ; brace, ignore it.
2438 ;; We've found the desired brace-pair.
2450 ;; We haven't found a brace pair. Record this.
2454 ;; If BRA+1 is nil, do nothing. Otherwise, BRA+1 is the buffer position
2455 ;; following a {, and that brace has a (mis-)matching } (or ]), and we
2456 ;; "push" "a" brace pair onto `c-state-cache'.
2457 ;;
2458 ;; Here "push" means overwrite the top element if it's itself a brace-pair,
2459 ;; otherwise push it normally.
2460 ;;
2461 ;; The brace pair we push is normally the one surrounding BRA+1, but if the
2462 ;; latter is inside a macro, not being a macro containing
2463 ;; MACRO-START-OR-HERE, we scan backwards through the buffer for a non-macro
2464 ;; base pair. This latter case is assumed to be rare.
2465 ;;
2466 ;; Note: POINT is not preserved in this routine.
2476 c-state-cache)))
2477 ;; N.B. This defsubst codes one method for the simple, normal case,
2478 ;; and a more sophisticated, slower way for the general case. Don't
2479 ;; eliminate this defsubst - it's a speed optimisation.
2483 ;; Scan the buffer from FROM to (point-max), adding elements into
2484 ;; `c-state-cache' for braces etc. Return a candidate for
2485 ;; `c-state-cache-good-pos'.
2486 ;;
2487 ;; FROM must be after the latest brace/paren/bracket in `c-state-cache', if
2488 ;; any. Typically, it is immediately after it. It must not be inside a
2489 ;; literal.
2498 ; Is actually the pos. to scan for a (/{/[ from,
2499 ; which sometimes is after a silly )/}/].
2502 ; good-pos.
2508 ;; Each time round the following loop, we enter a succesively deeper
2509 ;; level of brace/paren nesting. (Except sometimes we "continue at
2510 ;; the existing level".) `pa+1' is a pos inside an opening
2511 ;; brace/paren/bracket, usually just after it.
2514 ;; Each time round the next loop moves forward over an opening then
2515 ;; a closing brace/bracket/paren. This loop is white hot, so it
2516 ;; plays ugly tricks to go fast. DON'T PUT ANYTHING INTO THIS
2517 ;; LOOP WHICH ISN'T ABSOLUTELY NECESSARY!!! It terminates when a
2518 ;; call of `scan-lists' signals an error, which happens when there
2519 ;; are no more b/b/p's to scan.
2530 ;; We've just entered a deeper nesting level.
2532 ;; Insert the brace pair (if present) and the single open
2533 ;; paren/brace/bracket into `c-state-cache' It cannot be
2534 ;; inside a macro, except one around point, because of what
2535 ;; `c-neutralize-syntax-in-CPP' has done.
2537 ;; Insert the opening brace/bracket/paren position.
2539 ;; Clear admin stuff for the next more nested part of the scan.
2543 ;; All open p/b/b's at this nesting level, if any, have probably
2544 ;; been closed by matching/mismatching ones. We're probably
2545 ;; finished - we just need to check for having found an
2546 ;; unmatched )/}/], which we ignore. Such a )/}/] can't be in a
2547 ;; macro, due the action of `c-neutralize-syntax-in-CPP'.
2550 ;; Record the final, innermost, brace-pair if there is one.
2553 ;; Determine a good pos
2561 here-bol))
2570 ;; Remove stale entries from the `c-cache-state', i.e. those which will
2571 ;; not be in it when it is amended for position (point-max).
2572 ;; Additionally, the "outermost" open-brace entry before (point-max)
2573 ;; will be converted to a cons if the matching close-brace is scanned.
2574 ;;
2575 ;; GOOD-POS is a "maximal" "safe position" - there must be no open
2576 ;; parens/braces/brackets between GOOD-POS and (point-max).
2577 ;;
2578 ;; As a second thing, calculate the result of parse-partial-sexp at
2579 ;; PPS-POINT, w.r.t. GOOD-POS. The motivation here is that
2580 ;; `c-state-cache-good-pos' may become PPS-POINT, but the caller may need to
2581 ;; adjust it to get outside a string/comment. (Sorry about this! The code
2582 ;; needs to be FAST).
2583 ;;
2584 ;; Return a list (GOOD-POS SCAN-BACK-POS PPS-STATE), where
2585 ;; o - GOOD-POS is a position where the new value `c-state-cache' is known
2586 ;; to be good (we aim for this to be as high as possible);
2587 ;; o - SCAN-BACK-POS, if not nil, indicates there may be a brace pair
2588 ;; preceding POS which needs to be recorded in `c-state-cache'. It is a
2589 ;; position to scan backwards from.
2590 ;; o - PPS-STATE is the parse-partial-sexp state at PPS-POINT.
2600 ; or nil
2607 ; (point-max)), or nil.
2613 pps-state ; Will be 9 or 10 elements long.
2614 pos
2615 upper-lim ; ,beyond which `c-state-cache' entries are removed
2616 scan-back-pos
2617 pair-beg pps-point-state target-depth)
2619 ;; Remove entries beyond (point-max). Also remove any entries inside
2620 ;; a macro, unless (point-max) is in the same macro.
2629 ;; If `upper-lim' is inside the last recorded brace pair, remove its
2630 ;; RBrace and indicate we'll need to search backwards for a previous
2631 ;; brace pair.
2638 ;; The next loop jumps forward out of a nested level of parens each
2639 ;; time round; the corresponding elements in `c-state-cache' are
2640 ;; removed. `pos' is just after the brace-pair or the open paren at
2641 ;; (car c-state-cache). There can be no open parens/braces/brackets
2642 ;; between `good-pos'/`good-pos-actual-macro-start' and (point-max),
2643 ;; due to the interface spec to this function.
2646 ; marked by a `category' text property.
2647 good-pos))
2656 ;; Do nothing when we've merely reached pps-point.
2657 )
2659 ;; Scan!
2663 target-depth
2664 nil pps-state))
2674 ;; We've just moved out of the paren pair containing the brace-pair
2675 ;; at (car c-state-cache). `pair-beg' is where the open paren is,
2676 ;; and is potentially where the open brace of a cons in
2677 ;; c-state-cache will be.
2683 ; containing Lparen
2687 ; together with enclosed brace pair.
2688 ;; (t nil) ; Ignore an unmated Rparen.
2689 )))
2693 nil nil ; TARGETDEPTH, STOPBEFORE
2694 pps-state)))
2696 ;; If the last paren pair we moved out of was actually a brace pair,
2697 ;; insert it into `c-state-cache'.
2702 c-state-cache)))
2707 ;; Strip stale elements of `c-state-cache' by moving backwards through the
2708 ;; buffer, and inform the caller of the scenario detected.
2709 ;;
2710 ;; HERE is the position we're setting `c-state-cache' for.
2711 ;; CACHE-POS is just after the latest recorded position in `c-state-cache'
2712 ;; before HERE, or a position at or near point-min which isn't in a
2713 ;; literal.
2714 ;;
2715 ;; This function must only be called only when (> `c-state-cache-good-pos'
2716 ;; HERE). Usually the gap between CACHE-POS and HERE is large. It is thus
2717 ;; optimised to eliminate (or minimise) scanning between these two
2718 ;; positions.
2719 ;;
2720 ;; Return a three element list (GOOD-POS SCAN-BACK-POS FWD-FLAG), where:
2721 ;; o - GOOD-POS is a "good position", where `c-state-cache' is valid, or
2722 ;; could become so after missing elements are inserted into
2723 ;; `c-state-cache'. This is JUST AFTER an opening or closing
2724 ;; brace/paren/bracket which is already in `c-state-cache' or just before
2725 ;; one otherwise. exceptionally (when there's no such b/p/b handy) the BOL
2726 ;; before `here''s line, or the start of the literal containing it.
2727 ;; o - SCAN-BACK-POS, if non-nil, indicates there may be a brace pair
2728 ;; preceding POS which isn't recorded in `c-state-cache'. It is a position
2729 ;; to scan backwards from.
2730 ;; o - FWD-FLAG, if non-nil, indicates there may be parens/braces between
2731 ;; POS and HERE which aren't recorded in `c-state-cache'.
2732 ;;
2733 ;; The comments in this defun use "paren" to mean parenthesis or square
2734 ;; bracket (as contrasted with a brace), and "(" and ")" likewise.
2735 ;;
2736 ;; . {..} (..) (..) ( .. { } ) (...) ( .... . ..)
2737 ;; | | | | | |
2738 ;; CP E here D C good
2740 pa ren ; positions of "(" and ")"
2741 dropped-cons ; whether the last element dropped from `c-state-cache'
2742 ; was a cons (representing a brace-pair)
2743 good-pos ; see above.
2744 lit ; (START . END) of a literal containing some point.
2745 here-lit-start here-lit-end ; bounds of literal containing `here'
2746 ; or `here' itself.
2750 ;; Remove completely irrelevant entries from `c-state-cache'.
2756 ;; At this stage, (> pos here);
2757 ;; (< (c-state-cache-top-lparen) here) (or is nil).
2759 ;; CASE 1: The top of the cache is a brace pair which now encloses `here'.
2760 ;; As good-pos, return the address. of the "{".
2763 ;; Since we've no knowledge of what's inside these braces, we have no
2764 ;; alternative but to direct the caller to scan the buffer from the
2765 ;; opening brace.
2770 ; pair entry into a { entry, so the caller
2771 ; needs to search for a brace pair before the
2772 ; {.
2774 ;; ;; `here' might be inside a literal. Check for this.
2779 ;; `here' might be nested inside any depth of parens (or brackets but
2780 ;; not braces). Scan backwards to find the outermost such opening
2781 ;; paren, if there is one. This will be the scan position to return.
2787 ;; CASE 2: Address of outermost ( or [ which now encloses `here',
2788 ;; but didn't enclose the (previous) `c-state-cache-good-pos'. If
2789 ;; there is a brace pair preceding this, it will already be in
2790 ;; `c-state-cache', unless there was a brace pair after it,
2791 ;; i.e. there'll only be one to scan for if we've just deleted one.
2794 ;; `here' isn't enclosed in a (previously unrecorded) bracket/paren.
2795 ;; Further forward scanning isn't needed, but we still need to find a
2796 ;; GOOD-POS. Step out of all enclosing "("s on HERE's line.
2803 ;; CASE 3: After a }/)/] before `here''s BOL.
2806 ;; CASE 4; Best of a bad job: BOL before `here-bol', or beginning of
2807 ;; literal containing it.
2812 ;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;
2813 ;; Externally visible routines.
2817 c-state-cache-good-pos 1
2818 c-state-nonlit-pos-cache nil
2819 c-state-nonlit-pos-cache-limit 1
2820 c-state-brace-pair-desert nil
2821 c-state-point-min 1
2822 c-state-point-min-lit-type nil
2823 c-state-point-min-lit-start nil
2824 c-state-min-scan-pos 1
2825 c-state-old-cpp-beg nil
2826 c-state-old-cpp-end nil)
2830 ;; Invalidate all info on `c-state-cache' that applies to the buffer at HERE
2831 ;; or higher and set `c-state-cache-good-pos' accordingly. The cache is
2832 ;; left in a consistent state.
2833 ;;
2834 ;; This is much like `c-whack-state-after', but it never changes a paren
2835 ;; pair element into an open paren element. Doing that would mean that the
2836 ;; new open paren wouldn't have the required preceding paren pair element.
2837 ;;
2838 ;; This function is called from c-after-change.
2840 ;; The cache of non-literals:
2844 ;; `c-state-cache':
2845 ;; Case 1: if `here' is in a literal containing point-min, everything
2846 ;; becomes (or is already) nil.
2850 c-state-cache-good-pos nil
2851 c-state-min-scan-pos nil)
2853 ;;; Truncate `c-state-cache' and set `c-state-cache-good-pos' to a value below
2854 ;;; `here'. To maintain its consistency, we may need to insert a new brace
2855 ;;; pair.
2857 too-high-pa ; recorded {/(/[ next above here, or nil.
2858 dropped-cons ; was the last removed element a brace pair?
2859 pa)
2860 ;; The easy bit - knock over-the-top bits off `c-state-cache'.
2867 ;; Do we need to add in an earlier brace pair, having lopped one off?
2874 ;; The brace-pair desert marker:
2882 ;; Find and record all noteworthy parens between some good point earlier in
2883 ;; the file and point. That good point is at least the beginning of the
2884 ;; top-level construct we are in, or the beginning of the preceding
2885 ;; top-level construct if we aren't in one.
2886 ;;
2887 ;; The returned value is a list of the noteworthy parens with the last one
2888 ;; first. If an element in the list is an integer, it's the position of an
2889 ;; open paren (of any type) which has not been closed before the point. If
2890 ;; an element is a cons, it gives the position of a closed BRACE paren
2891 ;; pair[*]; the car is the start brace position and the cdr is the position
2892 ;; following the closing brace. Only the last closed brace paren pair
2893 ;; before each open paren and before the point is recorded, and thus the
2894 ;; state never contains two cons elements in succession. When a close brace
2895 ;; has no matching open brace (e.g., the matching brace is outside the
2896 ;; visible region), it is not represented in the returned value.
2897 ;;
2898 ;; [*] N.B. The close "brace" might be a mismatching close bracket or paren.
2899 ;; This defun explicitly treats mismatching parens/braces/brackets as
2900 ;; matching. It is the open brace which makes it a "brace" pair.
2901 ;;
2902 ;; If POINT is within a macro, open parens and brace pairs within
2903 ;; THIS macro MIGHT be recorded. This depends on whether their
2904 ;; syntactic properties have been suppressed by
2905 ;; `c-neutralize-syntax-in-CPP'. This might need fixing (2008-12-11).
2906 ;;
2907 ;; Currently no characters which are given paren syntax with the
2908 ;; syntax-table property are recorded, i.e. angle bracket arglist
2909 ;; parens are never present here. Note that this might change.
2910 ;;
2911 ;; BUG: This function doesn't cope entirely well with unbalanced
2912 ;; parens in macros. (2008-12-11: this has probably been resolved
2913 ;; by the function `c-neutralize-syntax-in-CPP'.) E.g. in the
2914 ;; following case the brace before the macro isn't balanced with the
2915 ;; one after it:
2916 ;;
2917 ;; {
2918 ;; #define X {
2919 ;; }
2920 ;;
2921 ;; Note to maintainers: this function DOES get called with point
2922 ;; within comments and strings, so don't assume it doesn't!
2923 ;;
2924 ;; This function might do hidden buffer changes.
2927 strategy ; 'forward, 'backward etc..
2928 ;; Candidate positions to start scanning from:
2929 cache-pos ; highest position below HERE already existing in
2930 ; cache (or 1).
2931 good-pos
2932 start-point
2933 bopl-state
2934 res
2936 ;; If POINT-MIN has changed, adjust the cache
2940 ;; Strategy?
2948 c-state-cache-good-pos start-point))
2950 ;; SCAN!
2959 ; start-point)
2968 good-pos)))
2977 scan-backward-pos))
2987 c-state-cache-good-pos nil)))))
2989 c-state-cache)
2992 ;; This is a wrapper over `c-invalidate-state-cache-1'.
2993 ;;
2994 ;; It suppresses the syntactic effect of the < and > (template) brackets and
2995 ;; of all parens in preprocessor constructs, except for any such construct
2996 ;; containing point. We can then call `c-invalidate-state-cache-1' without
2997 ;; worrying further about macros and template delimiters.
3002 c-state-old-cpp-beg
3009 ;; This is a wrapper over `c-parse-state-1'. See that function for a
3010 ;; description of the functionality and return value.
3011 ;;
3012 ;; It suppresses the syntactic effect of the < and > (template) brackets and
3013 ;; of all parens in preprocessor constructs, except for any such construct
3014 ;; containing point. We can then call `c-parse-state-1' without worrying
3015 ;; further about macros and template delimiters.
3022 here-cpp-beg)
3024 ;; FIXME!!! Put in a `condition-case' here to protect the integrity of the
3025 ;; subsystem.
3030 here-cpp-beg here-cpp-end
3036 )))
3038 ;; Debug tool to catch cache inconsistencies. This is called from
3039 ;; 000tests.el.
3059 ;; The cache can actually go further back due to the ad-hoc way
3060 ;; the first paren is found, so try to whack off a bit of its
3061 ;; start before complaining.
3070 here res1 res2))))
3071 res1))
3077 'c-debug-parse-state
3083 \f
3085 ;; Whack off any state information from PAREN-STATE which lies
3086 ;; before BUFPOS. Not destructive on PAREN-STATE.
3089 car)
3100 ;; Whack off any state information from PAREN-STATE which lies at or
3101 ;; after BUFPOS. Not destructive on PAREN-STATE.
3106 ;; just check the car, because in a balanced brace
3107 ;; expression, it must be impossible for the corresponding
3108 ;; close brace to be before point, but the open brace to
3109 ;; be after.
3111 nil ; whack it off
3113 ;; its possible that the open brace is before
3114 ;; bufpos, but the close brace is after. In that
3115 ;; case, convert this to a non-cons element. The
3116 ;; rest of the state is before bufpos, so we're
3117 ;; done.
3119 ;; we know that both the open and close braces are
3120 ;; before bufpos, so we also know that everything else
3121 ;; on state is before bufpos.
3124 nil ; whack it off
3125 ;; it's before bufpos, so everything else should too.
3128 nil)))
3131 ;; Return the bufpos of the innermost enclosing open paren before
3132 ;; bufpos, or nil if none was found.
3142 enclosingp))
3145 ;; Return the bufpos of the outermost enclosing open paren, or nil
3146 ;; if none was found.
3153 pos))
3156 ;; Return the closest "safe" position recorded on PAREN-STATE that
3157 ;; is higher up than BUFPOS. Return nil if PAREN-STATE doesn't
3158 ;; contain any. Return nil if BUFPOS is nil, which is useful to
3159 ;; find the closest limit before a given limit that might be nil.
3160 ;;
3161 ;; A "safe" position is a position at or after a recorded open
3162 ;; paren, or after a recorded close paren. The returned position is
3163 ;; thus either the first position after a close brace, or the first
3164 ;; position after an enclosing paren, or at the enclosing paren in
3165 ;; case BUFPOS is immediately after it.
3175 ;; See below.
3178 ;; elem is the position at and not after the opening paren, so
3179 ;; we can go forward one more step unless it's equal to
3180 ;; bufpos. This is useful in some cases avoid an extra paren
3181 ;; level between the safe position and bufpos.
3186 ;; This is used for `font-lock-beginning-of-syntax-function'. It
3187 ;; goes to the closest previous point that is known to be outside
3188 ;; any string literal or comment. `c-state-cache' is used if it has
3189 ;; a position in the vicinity.
3191 elem
3194 ;; Note: Similar code in `c-safe-position'. The
3195 ;; difference is that we accept a safe position at
3196 ;; the point and don't bother to go forward past open
3197 ;; parens.
3212 ;; The position is far back. Try `c-beginning-of-defun-1'
3213 ;; (although we can't be entirely sure it will go to a position
3214 ;; outside a comment or string in current emacsen). FIXME:
3215 ;; Consult `syntax-ppss' here.
3220 \f
3221 ;; Tools for scanning identifiers and other tokens.
3224 "Return non-nil if the point is on or directly after an identifier.
3225 Keywords are recognized and not considered identifiers. If an
3226 identifier is detected, the returned value is its starting position.
3227 If an identifier ends at the point and another begins at it \(can only
3228 happen in Pike) then the point for the preceding one is returned.
3230 Note that this function might do hidden buffer changes. See the
3231 comment at the start of cc-engine.el for more info."
3233 ;; FIXME: Shouldn't this function handle "operator" in C++?
3240 ;; Check for a normal (non-keyword) identifier.
3246 ;; Handle the `<operator> syntax in Pike.
3250 t
3257 ;; Handle the "operator +" syntax in C++.
3274 )))
3277 ;; If the point is at the end of a symbol then skip backward to the
3278 ;; beginning of it. Don't move otherwise. Return non-nil if point
3279 ;; moved.
3280 ;;
3281 ;; This function might do hidden buffer changes.
3284 ;; Handle the `<operator> syntax in Pike.
3290 t
3292 nil)))))
3295 ;; Move to the beginning of the current token. Do not move if not
3296 ;; in the middle of one. BACK-LIMIT may be used to bound the
3297 ;; backward search; if given it's assumed to be at the boundary
3298 ;; between two tokens. Return non-nil if the point is moved, nil
3299 ;; otherwise.
3300 ;;
3301 ;; This function might do hidden buffer changes.
3308 ;; `c-nonsymbol-token-regexp' should always match
3309 ;; since we've skipped backward over punctuator
3310 ;; or paren syntax, but consume one char in case
3311 ;; it doesn't so that we don't leave point before
3312 ;; some earlier incorrect token.
3319 ;; Move to the end of the current token. Do not move if not in the
3320 ;; middle of one. BACK-LIMIT may be used to bound the backward
3321 ;; search; if given it's assumed to be at the boundary between two
3322 ;; tokens. Return non-nil if the point is moved, nil otherwise.
3323 ;;
3324 ;; This function might do hidden buffer changes.
3332 ;; `c-nonsymbol-token-regexp' should always match since
3333 ;; we've skipped backward over punctuator or paren
3334 ;; syntax, but move forward in case it doesn't so that
3335 ;; we don't leave point earlier than we started with.
3351 "Move forward by tokens.
3352 A token is defined as all symbols and identifiers which aren't
3354 treated properly). Point is always either left at the beginning of a
3355 token or not moved at all. COUNT specifies the number of tokens to
3356 move; a negative COUNT moves in the opposite direction. A COUNT of 0
3357 moves to the next token beginning only if not already at one. If
3358 BALANCED is true, move over balanced parens, otherwise move into them.
3359 Also, if BALANCED is true, never move out of an enclosing paren.
3361 LIMIT sets the limit for the movement and defaults to the point limit.
3362 The case when LIMIT is set in the middle of a token, comment or macro
3363 is handled correctly, i.e. the point won't be left there.
3365 Return the number of tokens left to move \(positive or negative). If
3366 BALANCED is true, a move over a balanced paren counts as one. Note
3367 that if COUNT is 0 and no appropriate token beginning is found, 1 will
3368 be returned. Thus, a return value of 0 guarantees that point is at
3369 the requested position and a return value less \(without signs) than
3370 COUNT guarantees that point is at the beginning of some token.
3372 Note that this function might do hidden buffer changes. See the
3373 comment at the start of cc-engine.el for more info."
3380 c-jump-syntax-balanced
3381 c-jump-syntax-unbalanced))
3386 ;; If count is zero we should jump if in the middle of a token.
3393 ;; Skip whitespace. Count this as a move if we did in
3394 ;; fact move.
3398 ;; Moved out of bounds. Make sure the returned count isn't zero.
3403 ;; Use `condition-case' to avoid having the limit tests
3404 ;; inside the loop.
3412 t)
3415 t)
3416 ;; `c-nonsymbol-token-regexp' above should always
3417 ;; match if there are correct tokens. Try to
3418 ;; widen to see if the limit was set in the
3419 ;; middle of one, else fall back to treating
3420 ;; the offending thing as a one character token.
3425 nil)
3428 t))))
3438 count)))
3441 "Move backward by tokens.
3442 See `c-forward-token-2' for details."
3450 c-jump-syntax-balanced
3451 c-jump-syntax-unbalanced))
3455 ;; The count is zero so try to skip to the beginning of the
3456 ;; current token.
3460 ;; The limit is inside the same token, so return 1.
3463 ;; We're not in the middle of a token. If there's
3464 ;; whitespace after the point then we must move backward,
3465 ;; so set count to 1 in that case.
3467 ;; If we're looking at a '#' that might start a cpp
3468 ;; directive then we have to do a more elaborate check.
3481 ;; Use `condition-case' to avoid having to check for buffer
3482 ;; limits in `backward-char', `scan-sexps' and `goto-char' below.
3491 ;; This can be very inefficient if there's a long
3492 ;; sequence of operator tokens without any separation.
3493 ;; That doesn't happen in practice, anyway.
3503 count)))
3506 "Like `c-forward-token-2' but doesn't treat multicharacter operator
3508 characters are jumped over character by character. This function is
3509 for compatibility only; it's only a wrapper over `c-forward-token-2'."
3514 "Like `c-backward-token-2' but doesn't treat multicharacter operator
3516 characters are jumped over character by character. This function is
3517 for compatibility only; it's only a wrapper over `c-backward-token-2'."
3521 \f
3522 ;; Tools for doing searches restricted to syntactically relevant text.
3525 paren-level not-inside-token
3526 lookbehind-submatch)
3527 "Like `re-search-forward', but only report matches that are found
3528 in syntactically significant text. I.e. matches in comments, macros
3529 or string literals are ignored. The start point is assumed to be
3530 outside any comment, macro or string literal, or else the content of
3531 that region is taken as syntactically significant text.
3533 If PAREN-LEVEL is non-nil, an additional restriction is added to
3534 ignore matches in nested paren sexps. The search will also not go
3535 outside the current list sexp, which has the effect that if the point
3536 should be moved to BOUND when no match is found \(i.e. NOERROR is
3537 neither nil nor t), then it will be at the closing paren if the end of
3538 the current list sexp is encountered first.
3540 If NOT-INSIDE-TOKEN is non-nil, matches in the middle of tokens are
3541 ignored. Things like multicharacter operators and special symbols
3543 constants.
3545 If LOOKBEHIND-SUBMATCH is non-nil, it's taken as a number of a
3546 subexpression in REGEXP. The end of that submatch is used as the
3547 position to check for syntactic significance. If LOOKBEHIND-SUBMATCH
3548 isn't used or if that subexpression didn't match then the start
3550 subexpression is never tested before the starting position, so it
3554 subexpression can match the end of nonwhite syntactic whitespace,
3555 i.e. the end of comments or cpp directives. This since the function
3556 skips over such things before resuming the search. It's on the other
3558 matches syntactic whitespace.
3560 Bug: Unbalanced parens inside cpp directives are currently not handled
3561 correctly \(i.e. they don't get ignored as they should) when
3562 PAREN-LEVEL is set.
3564 Note that this function might do hidden buffer changes. See the
3565 comment at the start of cc-engine.el for more info."
3570 ;;(message "c-syntactic-re-search-forward %s %s %S" (point) bound regexp)
3573 tmp
3574 ;; Start position for the last search.
3575 search-pos
3576 ;; The `parse-partial-sexp' state between the start position
3577 ;; and the point.
3578 state
3579 ;; The current position after the last state update. The next
3580 ;; `parse-partial-sexp' continues from here.
3582 ;; The position at which to check the state and the state
3583 ;; there. This is separate from `state-pos' since we might
3584 ;; need to back up before doing the next search round.
3585 check-pos check-state
3586 ;; Last position known to end a token.
3588 ;; Set when a valid match is found.
3589 found)
3607 state-pos check-pos paren-level nil state))
3609 check-state state))
3611 ;; NOTE: If we got a look behind subexpression and get
3612 ;; an insignificant match in something that isn't
3613 ;; syntactic whitespace (i.e. strings or in nested
3614 ;; parentheses), then we can never skip more than a
3615 ;; single character from the match start position
3616 ;; (i.e. `state-pos' here) before continuing the
3617 ;; search. That since the look behind subexpression
3618 ;; might match the end of the insignificant region in
3619 ;; the next search.
3623 ;; Match inside a line comment. Skip to eol. Use
3624 ;; `re-search-forward' instead of `skip-chars-forward' to get
3625 ;; the right bound behavior.
3629 ;; Match inside a block comment. Skip to the '*/'.
3636 ;; Match in the middle of the opener of a block or line
3637 ;; comment.
3642 ;; The last `parse-partial-sexp' above might have
3643 ;; stopped short of the real check position if the end
3644 ;; of the current sexp was encountered in paren-level
3645 ;; mode. The checks above are always false in that
3646 ;; case, and since they can do better skipping in
3647 ;; lookbehind-submatch mode, we do them before
3648 ;; checking the paren level.
3652 ;; Check the paren level first since we're short of the
3653 ;; syntactic checking position if the end of the
3654 ;; current sexp was encountered by `parse-partial-sexp'.
3657 ;; Inside a nested paren sexp.
3659 ;; See the NOTE above.
3661 ;; Skip out of the paren quickly.
3665 ;; Have exited the current paren sexp.
3668 ;; The last `parse-partial-sexp' call above
3669 ;; has left us just after the closing paren
3670 ;; in this case, so we can modify the bound
3671 ;; to leave the point at the right position
3672 ;; upon return.
3674 nil)
3678 ;; Match inside a string.
3681 ;; See the NOTE above.
3683 ;; Skip to the end of the string before continuing.
3692 continue)))
3697 ;; Match inside a macro. Skip to the end of it.
3705 (< check-pos
3711 ;; Inside a token.
3713 ;; See the NOTE above.
3718 ;; A real match.
3720 nil)))
3722 ;; Should loop to search again, but take care to avoid
3723 ;; looping on the same spot.
3727 nil
3730 t))))
3736 ;;(message "c-syntactic-re-search-forward done %s" (or (match-end 0) (point)))
3743 ;; Search failed. Set point as appropriate.
3747 nil)))
3752 ;; Return the start of the literal point is in, or nil.
3753 ;; We read and write the variables `safe-pos', `safe-pos-list', `state'
3754 ;; bound in the caller.
3756 ;; Use `parse-partial-sexp' from a safe position down to the point to check
3757 ;; if it's outside comments and strings.
3760 ;; Pick a safe position as close to the point as possible.
3761 ;;
3762 ;; FIXME: Consult `syntax-ppss' here if our cache doesn't give a good
3763 ;; position.
3776 ;; Cache positions along the way to use if we have to back up more. We
3777 ;; cache every closing paren on the same level. If the paren cache is
3778 ;; relevant in this region then we're typically already on the same
3779 ;; level as the target position. Note that we might cache positions
3780 ;; after opening parens in case safe-pos is in a nested list. That's
3781 ;; both uncommon and harmless.
3789 ;; If the state contains the start of the containing sexp we cache that
3790 ;; position too, so that parse-partial-sexp in the next run has a bigger
3791 ;; chance of starting at the same level as the target position and thus
3792 ;; will get more good safe positions into the list.
3798 ;; Inside string or comment. Continue search at the
3799 ;; beginning of it.
3803 "Like `skip-chars-backward' but only look at syntactically relevant chars,
3804 i.e. don't stop at positions inside syntactic whitespace or string
3805 literals. Preprocessor directives are also ignored, with the exception
3806 of the one that the point starts within, if any. If LIMIT is given,
3807 it's assumed to be at a syntactically relevant position.
3809 If PAREN-LEVEL is non-nil, the function won't stop in nested paren
3810 sexps, and the search will also not go outside the current paren sexp.
3811 However, if LIMIT or the buffer limit is reached inside a nested paren
3812 then the point will be left at the limit.
3814 Non-nil is returned if the point moved, nil otherwise.
3816 Note that this function might do hidden buffer changes. See the
3817 comment at the start of cc-engine.el for more info."
3820 state-2
3821 ;; A list of syntactically relevant positions in descending
3822 ;; order. It's used to avoid scanning repeatedly over
3823 ;; potentially large regions with `parse-partial-sexp' to verify
3824 ;; each position. Used in `c-ssb-lit-begin'
3825 safe-pos-list
3826 ;; The result from `c-beginning-of-macro' at the start position or the
3827 ;; start position itself if it isn't within a macro. Evaluated on
3828 ;; demand.
3829 start-macro-beg
3830 ;; The earliest position after the current one with the same paren
3831 ;; level. Used only when `paren-level' is set.
3832 lit-beg
3837 ;; The next loop "tries" to find the end point each time round,
3838 ;; loops when it hasn't succeeded.
3846 ;; Don't stop inside a literal
3849 t)
3857 ;; Not at the right level.
3860 ;; We stop above if we go out of a paren.
3861 ;; Now check whether it precedes or is
3862 ;; nested in the starting sexp.
3866 pps-end-pos paren-level-pos
3867 nil nil state-2))
3870 ;; We've stopped short of the starting position
3871 ;; so the hit was inside a nested list. Go up
3872 ;; until we are at the right level.
3881 nil)
3882 t))
3885 nil))
3887 ;; The hit was outside the list at the start
3888 ;; position. Go to the start of the list and exit.
3890 nil))
3893 ;; Inside a macro.
3901 t
3903 ;; It's inside the same macro we started in so it's
3904 ;; a relevant match.
3906 nil))))))
3910 ;; Skip syntactic ws afterwards so that we don't stop at the
3911 ;; end of a comment if `skip-chars' is something like "^/".
3915 ;; We might want to extend this with more useful return values in
3916 ;; the future.
3919 ;; The following is an alternative implementation of
3920 ;; `c-syntactic-skip-backward' that uses backward movement to keep
3921 ;; track of the syntactic context. It turned out to be generally
3922 ;; slower than the one above which uses forward checks from earlier
3923 ;; safe positions.
3924 ;;
3925 ;;(defconst c-ssb-stop-re
3926 ;; ;; The regexp matching chars `c-syntactic-skip-backward' needs to
3927 ;; ;; stop at to avoid going into comments and literals.
3928 ;; (concat
3929 ;; ;; Match comment end syntax and string literal syntax. Also match
3930 ;; ;; '/' for block comment endings (not covered by comment end
3931 ;; ;; syntax).
3932 ;; "\\s>\\|/\\|\\s\""
3933 ;; (if (memq 'gen-string-delim c-emacs-features)
3934 ;; "\\|\\s|"
3935 ;; "")
3936 ;; (if (memq 'gen-comment-delim c-emacs-features)
3937 ;; "\\|\\s!"
3938 ;; "")))
3939 ;;
3940 ;;(defconst c-ssb-stop-paren-re
3941 ;; ;; Like `c-ssb-stop-re' but also stops at paren chars.
3942 ;; (concat c-ssb-stop-re "\\|\\s(\\|\\s)"))
3943 ;;
3944 ;;(defconst c-ssb-sexp-end-re
3945 ;; ;; Regexp matching the ending syntax of a complex sexp.
3946 ;; (concat c-string-limit-regexp "\\|\\s)"))
3947 ;;
3948 ;;(defun c-syntactic-skip-backward (skip-chars &optional limit paren-level)
3949 ;; "Like `skip-chars-backward' but only look at syntactically relevant chars,
3950 ;;i.e. don't stop at positions inside syntactic whitespace or string
3951 ;;literals. Preprocessor directives are also ignored. However, if the
3952 ;;point is within a comment, string literal or preprocessor directory to
3953 ;;begin with, its contents is treated as syntactically relevant chars.
3954 ;;If LIMIT is given, it limits the backward search and the point will be
3955 ;;left there if no earlier position is found.
3956 ;;
3957 ;;If PAREN-LEVEL is non-nil, the function won't stop in nested paren
3958 ;;sexps, and the search will also not go outside the current paren sexp.
3959 ;;However, if LIMIT or the buffer limit is reached inside a nested paren
3960 ;;then the point will be left at the limit.
3961 ;;
3962 ;;Non-nil is returned if the point moved, nil otherwise.
3963 ;;
3964 ;;Note that this function might do hidden buffer changes. See the
3965 ;;comment at the start of cc-engine.el for more info."
3966 ;;
3967 ;; (save-restriction
3968 ;; (when limit
3969 ;; (narrow-to-region limit (point-max)))
3970 ;;
3971 ;; (let ((start (point)))
3972 ;; (catch 'done
3973 ;; (while (let ((last-pos (point))
3974 ;; (stop-pos (progn
3975 ;; (skip-chars-backward skip-chars)
3976 ;; (point))))
3977 ;;
3978 ;; ;; Skip back over the same region as
3979 ;; ;; `skip-chars-backward' above, but keep to
3980 ;; ;; syntactically relevant positions.
3981 ;; (goto-char last-pos)
3982 ;; (while (and
3983 ;; ;; `re-search-backward' with a single char regexp
3984 ;; ;; should be fast.
3985 ;; (re-search-backward
3986 ;; (if paren-level c-ssb-stop-paren-re c-ssb-stop-re)
3987 ;; stop-pos 'move)
3988 ;;
3989 ;; (progn
3990 ;; (cond
3991 ;; ((looking-at "\\s(")
3992 ;; ;; `paren-level' is set and we've found the
3993 ;; ;; start of the containing paren.
3994 ;; (forward-char)
3995 ;; (throw 'done t))
3996 ;;
3997 ;; ((looking-at c-ssb-sexp-end-re)
3998 ;; ;; We're at the end of a string literal or paren
3999 ;; ;; sexp (if `paren-level' is set).
4000 ;; (forward-char)
4001 ;; (condition-case nil
4002 ;; (c-backward-sexp)
4003 ;; (error
4004 ;; (goto-char limit)
4005 ;; (throw 'done t))))
4006 ;;
4007 ;; (t
4008 ;; (forward-char)
4009 ;; ;; At the end of some syntactic ws or possibly
4010 ;; ;; after a plain '/' operator.
4011 ;; (let ((pos (point)))
4012 ;; (c-backward-syntactic-ws)
4013 ;; (if (= pos (point))
4014 ;; ;; Was a plain '/' operator. Go past it.
4015 ;; (backward-char)))))
4016 ;;
4017 ;; (> (point) stop-pos))))
4018 ;;
4019 ;; ;; Now the point is either at `stop-pos' or at some
4020 ;; ;; position further back if `stop-pos' was at a
4021 ;; ;; syntactically irrelevant place.
4022 ;;
4023 ;; ;; Skip additional syntactic ws so that we don't stop
4024 ;; ;; at the end of a comment if `skip-chars' is
4025 ;; ;; something like "^/".
4026 ;; (c-backward-syntactic-ws)
4027 ;;
4028 ;; (< (point) stop-pos))))
4029 ;;
4030 ;; ;; We might want to extend this with more useful return values
4031 ;; ;; in the future.
4032 ;; (/= (point) start))))
4034 \f
4035 ;; Tools for handling comments and string literals.
4038 "Return the type of literal point is in, if any.
4039 The return value is `c' if in a C-style comment, `c++' if in a C++
4040 style comment, `string' if in a string literal, `pound' if DETECT-CPP
4041 is non-nil and in a preprocessor line, or nil if somewhere else.
4042 Optional LIM is used as the backward limit of the search. If omitted,
4043 or nil, `c-beginning-of-defun' is used.
4045 The last point calculated is cached if the cache is enabled, i.e. if
4046 `c-in-literal-cache' is bound to a two element vector.
4048 Note that this function might do hidden buffer changes. See the
4049 comment at the start of cc-engine.el for more info."
4065 ;; cache this result if the cache is enabled
4068 rtn)))
4070 ;; XEmacs has a built-in function that should make this much quicker.
4071 ;; I don't think we even need the cache, which makes our lives more
4072 ;; complicated anyway. In this case, lim is only used to detect
4073 ;; cpp directives.
4074 ;;
4075 ;; Note that there is a bug in Xemacs's buffer-syntactic-context when used in
4076 ;; conjunction with syntax-table-properties. The bug is present in, e.g.,
4077 ;; Xemacs 21.4.4. It manifested itself thus:
4078 ;;
4079 ;; Starting with an empty AWK Mode buffer, type
4080 ;; /regexp/ {<C-j>
4081 ;; Point gets wrongly left at column 0, rather than being indented to tab-width.
4082 ;;
4083 ;; AWK Mode is designed such that when the first / is typed, it gets the
4084 ;; syntax-table property "string fence". When the second / is typed, BOTH /s
4085 ;; are given the s-t property "string". However, buffer-syntactic-context
4086 ;; fails to take account of the change of the s-t property on the opening / to
4087 ;; "string", and reports that the { is within a string started by the second /.
4088 ;;
4089 ;; The workaround for this is for the AWK Mode initialisation to switch the
4090 ;; defalias for c-in-literal to c-slow-in-literal. This will slow down other
4091 ;; cc-modes in Xemacs whenever an awk-buffer has been initialised.
4092 ;;
4093 ;; (Alan Mackenzie, 2003/4/30).
4096 ;; This function might do hidden buffer changes.
4109 ;; The defalias above isn't enough to shut up the byte compiler.
4113 "Return a cons of the beginning and end positions of the comment or
4114 string surrounding point (including both delimiters), or nil if point
4116 to start parsing from. If NEAR is non-nil, then the limits of any
4118 spaces and tabs between point and the literal. The search for such a
4119 literal is done first in forward direction. If NOT-IN-DELIMITER is
4120 non-nil, the case when point is inside a starting delimiter won't be
4121 recognized. This only has effect for comments which have starting
4122 delimiters with more than one character.
4124 Note that this function might do hidden buffer changes. See the
4125 comment at the start of cc-engine.el for more info."
4147 ;; We're standing in a comment starter.
4154 ;; Search forward for a literal.
4166 ;; Search backward.
4177 ;; Block comment. Due to the nature of line
4178 ;; comments, they will always be covered by the
4179 ;; normal case above.
4182 ;; If LIM is bogus, beg will be bogus.
4186 ))))
4188 ;; In case external callers use this; it did have a docstring.
4192 "If the argument is a cons of two buffer positions (such as returned by
4193 `c-literal-limits'), and that range contains a C++ style line comment,
4194 then an extended range is returned that contains all adjacent line
4195 comments (i.e. all comments that starts in the same column with no
4196 empty lines or non-whitespace characters between them). Otherwise the
4197 argument is returned.
4199 Note that this function might do hidden buffer changes. See the
4200 comment at the start of cc-engine.el for more info."
4211 ;; Got to take care in the backward direction to handle
4212 ;; comments which are preceded by code.
4226 range)
4230 "Convenience function that given the result of `c-literal-limits',
4231 returns nil or the type of literal that the range surrounds, one
4232 of the symbols 'c, 'c++ or 'string. It's much faster than using
4233 `c-in-literal' and is intended to be used when you need both the
4234 type of a literal and its limits.
4236 Note that this function might do hidden buffer changes. See the
4237 comment at the start of cc-engine.el for more info."
4248 range))
4250 \f
4251 ;; `c-find-decl-spots' and accompanying stuff.
4253 ;; Variables used in `c-find-decl-spots' to cache the search done for
4254 ;; the first declaration in the last call. When that function starts,
4255 ;; it needs to back up over syntactic whitespace to look at the last
4256 ;; token before the region being searched. That can sometimes cause
4257 ;; moves back and forth over a quite large region of comments and
4258 ;; macros, which would be repeated for each changed character when
4259 ;; we're called during fontification, since font-lock refontifies the
4260 ;; current line for each change. Thus it's worthwhile to cache the
4261 ;; first match.
4262 ;;
4263 ;; `c-find-decl-syntactic-pos' is a syntactically relevant position in
4264 ;; the syntactic whitespace less or equal to some start position.
4265 ;; There's no cached value if it's nil.
4266 ;;
4267 ;; `c-find-decl-match-pos' is the match position if
4268 ;; `c-find-decl-prefix-search' matched before the syntactic whitespace
4269 ;; at `c-find-decl-syntactic-pos', or nil if there's no such match.
4280 ; (defface c-debug-decl-spot-face
4281 ; '((t (:background "Turquoise")))
4282 ; "Debug face to mark the spots where `c-find-decl-spots' stopped.")
4283 ; (defface c-debug-decl-sws-face
4284 ; '((t (:background "Khaki")))
4285 ; "Debug face to mark the syntactic whitespace between the declaration
4286 ; spots and the preceding token end.")
4302 ;; Macro used inside `c-find-decl-spots'. It ought to be a defun,
4303 ;; but it contains lots of free variables that refer to things
4304 ;; inside `c-find-decl-spots'. The point is left at `cfd-match-pos'
4305 ;; if there is a match, otherwise at `cfd-limit'.
4306 ;;
4307 ;; This macro might do hidden buffer changes.
4310 ;; Find the next property match position if we haven't got one already.
4321 ;; Find the next `c-decl-prefix-or-start-re' match if we haven't
4322 ;; got one already.
4332 ;; Match. Check if it's inside a comment or string literal.
4335 ;; Matched the end of a token preceding a decl spot.
4339 ;; Matched a token that start a decl spot.
4342 c-literal-faces)
4344 ;; No match. Finish up and exit the loop.
4346 nil)
4348 ;; Skip out of comments and string literals.
4355 ;; If we matched at the decl start, we have to back up over the
4356 ;; preceding syntactic ws to set `cfd-match-pos' and to catch
4357 ;; any decl spots in the syntactic ws.
4362 ;; Choose whichever match is closer to the start.
4365 cfd-re-match nil)
4367 cfd-prop-match nil))
4372 ;; Skip forward past comments only so we don't skip macros.
4374 ;; Set the position to continue at. We can avoid going over
4375 ;; the comments skipped above a second time, but it's possible
4376 ;; that the comment skipping has taken us past `cfd-prop-match'
4377 ;; since the property might be used inside comments.
4383 ;; Call CFD-FUN for each possible spot for a declaration, cast or
4384 ;; label from the point to CFD-LIMIT.
4385 ;;
4386 ;; CFD-FUN is called with point at the start of the spot. It's
4387 ;; passed two arguments: The first is the end position of the token
4388 ;; preceding the spot, or 0 for the implicit match at bob. The
4389 ;; second is a flag that is t when the match is inside a macro. If
4390 ;; CFD-FUN adds `c-decl-end' properties somewhere below the current
4391 ;; spot, it should return non-nil to ensure that the next search
4392 ;; will find them.
4393 ;;
4394 ;; Such a spot is:
4395 ;; o The first token after bob.
4396 ;; o The first token after the end of submatch 1 in
4397 ;; `c-decl-prefix-or-start-re' when that submatch matches.
4398 ;; o The start of each `c-decl-prefix-or-start-re' match when
4399 ;; submatch 1 doesn't match.
4400 ;; o The first token after the end of each occurrence of the
4401 ;; `c-type' text property with the value `c-decl-end', provided
4402 ;; `c-type-decl-end-used' is set.
4403 ;;
4404 ;; Only a spot that match CFD-DECL-RE and whose face is in the
4405 ;; CFD-FACE-CHECKLIST list causes CFD-FUN to be called. The face
4406 ;; check is disabled if CFD-FACE-CHECKLIST is nil.
4407 ;;
4408 ;; If the match is inside a macro then the buffer is narrowed to the
4409 ;; end of it, so that CFD-FUN can investigate the following tokens
4410 ;; without matching something that begins inside a macro and ends
4411 ;; outside it. It's to avoid this work that the CFD-DECL-RE and
4412 ;; CFD-FACE-CHECKLIST checks exist.
4413 ;;
4414 ;; The spots are visited approximately in order from top to bottom.
4415 ;; It's however the positions where `c-decl-prefix-or-start-re'
4416 ;; matches and where `c-decl-end' properties are found that are in
4417 ;; order. Since the spots often are at the following token, they
4418 ;; might be visited out of order insofar as more spots are reported
4419 ;; later on within the syntactic whitespace between the match
4420 ;; positions and their spots.
4421 ;;
4422 ;; It's assumed that comments and strings are fontified in the
4423 ;; searched range.
4424 ;;
4425 ;; This is mainly used in fontification, and so has an elaborate
4426 ;; cache to handle repeated calls from the same start position; see
4427 ;; the variables above.
4428 ;;
4429 ;; All variables in this function begin with `cfd-' to avoid name
4430 ;; collision with the (dynamically bound) variables used in CFD-FUN.
4431 ;;
4432 ;; This function might do hidden buffer changes.
4436 ;; The end of the token preceding the decl spot last found
4437 ;; with `c-decl-prefix-or-start-re'. `cfd-limit' if there's
4438 ;; no match.
4439 cfd-re-match
4440 ;; The end position of the last `c-decl-prefix-or-start-re'
4441 ;; match. If this is greater than `cfd-continue-pos', the
4442 ;; next regexp search is started here instead.
4444 ;; The end of the last `c-decl-end' found by
4445 ;; `c-find-decl-prefix-search'. `cfd-limit' if there's no
4446 ;; match. If searching for the property isn't needed then we
4447 ;; disable it by setting it to `cfd-limit' directly.
4449 ;; The end of the token preceding the decl spot last found by
4450 ;; `c-find-decl-prefix-search'. 0 for the implicit match at
4451 ;; bob. `cfd-limit' if there's no match. In other words,
4452 ;; this is the minimum of `cfd-re-match' and `cfd-prop-match'.
4454 ;; The position to continue searching at.
4455 cfd-continue-pos
4456 ;; The position of the last "real" token we've stopped at.
4457 ;; This can be greater than `cfd-continue-pos' when we get
4458 ;; hits inside macros or at `c-decl-end' positions inside
4459 ;; comments.
4461 ;; The end position of the last entered macro.
4464 ;; Initialize by finding a syntactically relevant start position
4465 ;; before the point, and do the first `c-decl-prefix-or-start-re'
4466 ;; search unless we're at bob.
4469 ;; Must back up a bit since we look for the end of the previous
4470 ;; statement or declaration, which is earlier than the first
4471 ;; returned match.
4474 ;; First we need to move to a syntactically relevant position.
4475 ;; Begin by backing out of comment or string literals.
4478 ;; Try to use the faces to back up to the start of the
4479 ;; literal. FIXME: What if the point is on a declaration
4480 ;; inside a comment?
4486 ;; XEmacs doesn't fontify the quotes surrounding string
4487 ;; literals.
4496 ;; Don't trust the literal to contain only literal faces
4497 ;; (the font lock package might not have fontified the
4498 ;; start of it at all, for instance) so check that we have
4499 ;; arrived at something that looks like a start or else
4500 ;; resort to `c-literal-limits'.
4507 ;; The start is in a literal. If the limit is in the same
4508 ;; one we don't have to find a syntactic position etc. We
4509 ;; only check that if the limit is at or before bonl to save
4510 ;; time; it covers the by far most common case when font-lock
4511 ;; refontifies the current line only.
4522 ;; Completely inside a literal. Set up variables to trig the
4523 ;; (< cfd-continue-pos cfd-start-pos) case below and it'll
4524 ;; find a suitable start position.
4527 ;; Check if the region might be completely inside a macro, to
4528 ;; optimize that like the completely-inside-literal above.
4535 ;; (Maybe) completely inside a macro. Only need to trig the
4536 ;; (< cfd-continue-pos cfd-start-pos) case below to make it
4537 ;; set things up.
4539 start-in-macro t))
4542 ;; Back out of any macro so we don't miss any declaration
4543 ;; that could follow after it.
4547 ;; Now we're at a proper syntactically relevant position so we
4548 ;; can use the cache. But first clear it if it applied
4549 ;; further down.
4554 ;; Don't have to do this if the cache is relevant here,
4555 ;; typically if the same line is refontified again. If
4556 ;; we're just some syntactic whitespace further down we can
4557 ;; still use the cache to limit the skipping.
4560 ;; If we hit `c-find-decl-syntactic-pos' and
4561 ;; `c-find-decl-match-pos' is set then we install the cached
4562 ;; values. If we hit `c-find-decl-syntactic-pos' and
4563 ;; `c-find-decl-match-pos' is nil then we know there's no decl
4564 ;; prefix in the whitespace before `c-find-decl-syntactic-pos'
4565 ;; and so we can continue the search from this point. If we
4566 ;; didn't hit `c-find-decl-syntactic-pos' then we're now in
4567 ;; the right spot to begin searching anyway.
4569 c-find-decl-match-pos)
4571 cfd-continue-pos syntactic-pos)
4576 ;; Always consider bob a match to get the first
4577 ;; declaration in the file. Do this separately instead of
4578 ;; letting `c-decl-prefix-or-start-re' match bob, so that
4579 ;; regexp always can consume at least one character to
4580 ;; ensure that we won't get stuck in an infinite loop.
4585 ;; Do an initial search now. In the bob case above it's
4586 ;; only done to search for a `c-decl-end' spot.
4590 cfd-match-pos)))))
4592 ;; Advance `cfd-continue-pos' if it's before the start position.
4593 ;; The closest continue position that might have effect at or
4594 ;; after the start depends on what we started in. This also
4595 ;; finds a suitable start position in the special cases when the
4596 ;; region is completely within a literal or macro.
4601 ;; If we're in a macro then it's the closest preceding token
4602 ;; in the macro. Check this before `start-in-literal',
4603 ;; since if we're inside a literal in a macro, the preceding
4604 ;; token is earlier than any `c-decl-end' spot inside the
4605 ;; literal (comment).
4607 ;; The only syntactic ws in macros are comments.
4613 ;; If we're in a comment it can only be the closest
4614 ;; preceding `c-decl-end' position within that comment, if
4615 ;; any. Go back to the beginning of such a property so that
4616 ;; `c-find-decl-prefix-search' will find the end of it.
4617 ;; (Can't stop at the end and install it directly on
4618 ;; `cfd-prop-match' since that variable might be cleared
4619 ;; after `cfd-fun' below.)
4620 ;;
4621 ;; Note that if the literal is a string then the property
4622 ;; search will simply skip to the beginning of it right
4623 ;; away.
4635 ;; Didn't find any property inside the comment, so we can
4636 ;; skip it entirely. (This won't skip past a string, but
4637 ;; that'll be handled quickly by the next
4638 ;; `c-find-decl-prefix-search' anyway.)
4644 ;; If we started in normal code, the only match that might
4645 ;; apply before the start is what we already got in
4646 ;; `cfd-match-pos' so we can continue at the start position.
4647 ;; (Note that we don't get here if the first match is below
4648 ;; it.)
4651 ;; Delete found matches if they are before our new continue
4652 ;; position, so that `c-find-decl-prefix-search' won't back up
4653 ;; to them later on.
4661 ;; This is the normal case and we got a proper syntactic
4662 ;; position. If there's a match then it's always outside
4663 ;; macros and comments, so advance to the next token and set
4664 ;; `cfd-token-pos'. The loop below will later go back using
4665 ;; `cfd-continue-pos' to fix declarations inside the
4666 ;; syntactic ws.
4674 ;; Have one of the special cases when the region is completely
4675 ;; within a literal or macro. `cfd-continue-pos' is set to a
4676 ;; good start position for the search, so do it.
4679 ;; Now loop. Round what? (ACM, 2006/7/5). We already got the first match.
4686 ;; Kludge to filter out matches on the "<" that
4687 ;; aren't open parens, for the sake of languages
4688 ;; that got `c-recognize-<>-arglists' set.
4693 ;; If `cfd-continue-pos' is less or equal to
4694 ;; `cfd-token-pos', we've got a hit inside a macro
4695 ;; that's in the syntactic whitespace before the last
4696 ;; "real" declaration we've checked. If they're equal
4697 ;; we've arrived at the declaration a second time, so
4698 ;; there's nothing to do.
4702 ;; If `cfd-continue-pos' is less than `cfd-token-pos'
4703 ;; we're still searching for declarations embedded in
4704 ;; the syntactic whitespace. In that case we need
4705 ;; only to skip comments and not macros, since they
4706 ;; can't be nested, and that's already been done in
4707 ;; `c-find-decl-prefix-search'.
4712 ;; Continue if the following token fails the
4713 ;; CFD-DECL-RE and CFD-FACE-CHECKLIST checks.
4720 t)))
4731 ;; Narrow to the end of the macro if we got a hit inside
4732 ;; one, to avoid recognizing things that start inside the
4733 ;; macro and end outside it.
4735 ;; Not in the same macro as in the previous round.
4746 t
4749 t)
4750 ;; The matched token was the last thing in the macro,
4751 ;; so the whole match is bogus.
4753 nil))))
4760 ;; Restore limits if we did macro narrowment above.
4768 \f
4769 ;; A cache for found types.
4771 ;; Buffer local variable that contains an obarray with the types we've
4772 ;; found. If a declaration is recognized somewhere we record the
4773 ;; fully qualified identifier in it to recognize it as a type
4774 ;; elsewhere in the file too. This is not accurate since we do not
4775 ;; bother with the scoping rules of the languages, but in practice the
4776 ;; same name is seldom used as both a type and something else in a
4777 ;; file, and we only use this as a last resort in ambiguous cases (see
4778 ;; `c-forward-decl-or-cast-1').
4779 ;;
4780 ;; Not every type need be in this cache. However, things which have
4781 ;; ceased to be types must be removed from it.
4782 ;;
4783 ;; Template types in C++ are added here too but with the template
4784 ;; arglist replaced with "<>" in references or "<" for the one in the
4785 ;; primary type. E.g. the type "Foo<A,B>::Bar<C>" is stored as
4786 ;; "Foo<>::Bar<". This avoids storing very long strings (since C++
4787 ;; template specs can be fairly sized programs in themselves) and
4788 ;; improves the hit ratio (it's a type regardless of the template
4789 ;; args; it's just not the same type, but we're only interested in
4790 ;; recognizing types, not telling distinct types apart). Note that
4791 ;; template types in references are added here too; from the example
4792 ;; above there will also be an entry "Foo<".
4797 ;; Clears `c-found-types'.
4801 ;; Add the given region as a type in `c-found-types'. If the region
4802 ;; doesn't match an existing type but there is a type which is equal
4803 ;; to the given one except that the last character is missing, then
4804 ;; the shorter type is removed. That's done to avoid adding all
4805 ;; prefixes of a type as it's being entered and font locked. This
4806 ;; doesn't cover cases like when characters are removed from a type
4807 ;; or added in the middle. We'd need the position of point when the
4808 ;; font locking is invoked to solve this well.
4809 ;;
4810 ;; This function might do hidden buffer changes.
4817 ;; Remove the "NAME" from c-found-types, if present.
4821 ;; Return non-nil if the given region contains a type in
4822 ;; `c-found-types'.
4823 ;;
4824 ;; This function might do hidden buffer changes.
4826 c-found-types))
4829 ;; Return all the types in `c-found-types' as a sorted list of
4830 ;; strings.
4834 type-list)))
4835 c-found-types)
4838 ;; Shut up the byte compiler.
4842 ;; An after change function which, in conjunction with the info in
4843 ;; c-maybe-stale-found-type (set in c-before-change), removes a type
4844 ;; from `c-found-types', should this type have become stale. For
4845 ;; example, this happens to "foo" when "foo \n bar();" becomes
4846 ;; "foo(); \n bar();". Such stale types, if not removed, foul up
4847 ;; the fontification.
4848 ;;
4849 ;; Have we, perhaps, added non-ws characters to the front/back of a found
4850 ;; type?
4870 ;; Changing the amount of (already existing) whitespace - don't do anything.
4875 ;; The syntactic relationship which defined a "found type" has been
4876 ;; destroyed.
4879 ;; ((eq (car c-maybe-stale-found-type) 'c-decl-type-start) FIXME!!!
4880 )))
4882 \f
4883 ;; Setting and removing syntax properties on < and > in languages (C++
4884 ;; and Java) where they can be template/generic delimiters as well as
4885 ;; their normal meaning of "less/greater than".
4887 ;; Normally, < and > have syntax 'punctuation'. When they are found to
4888 ;; be delimiters, they are marked as such with the category properties
4889 ;; c-<-as-paren-syntax, c->-as-paren-syntax respectively.
4891 ;; STRATEGY:
4892 ;;
4893 ;; It is impossible to determine with certainty whether a <..> pair in
4894 ;; C++ is two comparison operators or is template delimiters, unless
4895 ;; one duplicates a lot of a C++ compiler. For example, the following
4896 ;; code fragment:
4897 ;;
4898 ;; foo (a < b, c > d) ;
4899 ;;
4900 ;; could be a function call with two integer parameters (each a
4901 ;; relational expression), or it could be a constructor for class foo
4902 ;; taking one parameter d of templated type "a < b, c >". They are
4903 ;; somewhat easier to distinguish in Java.
4904 ;;
4905 ;; The strategy now (2010-01) adopted is to mark and unmark < and
4906 ;; > IN MATCHING PAIRS ONLY. [Previously, they were marked
4907 ;; individually when their context so indicated. This gave rise to
4908 ;; intractible problems when one of a matching pair was deleted, or
4909 ;; pulled into a literal.]
4910 ;;
4911 ;; At each buffer change, the syntax-table properties are removed in a
4912 ;; before-change function and reapplied, when needed, in an
4913 ;; after-change function. It is far more important that the
4914 ;; properties get removed when they they are spurious than that they
4915 ;; be present when wanted.
4916 ;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;
4918 ;; POS (default point) is at a < character. If it is marked with
4919 ;; open paren syntax-table text property, remove the property,
4920 ;; together with the close paren property on the matching > (if
4921 ;; any).
4936 ;; POS (default point) is at a > character. If it is marked with
4937 ;; close paren syntax-table property, remove the property, together
4938 ;; with the open paren property on the matching < (if any).
4953 ;; POS (default point) is at a < or > character. If it has an
4954 ;; open/close paren syntax-table property, remove this property both
4955 ;; from the current character and its partner (which will also be
4956 ;; thusly marked).
4966 ;; POS (default point) is at a < character. If it is both marked
4967 ;; with open/close paren syntax-table property, and has a matching >
4968 ;; (also marked) which is after LIM, remove the property both from
4969 ;; the current > and its partner.
4985 ;; POS (default point) is at a > character. If it is both marked
4986 ;; with open/close paren syntax-table property, and has a matching <
4987 ;; (also marked) which is before LIM, remove the property both from
4988 ;; the current < and its partner.
5004 ;; Unmark certain pairs of "< .... >" which are currently marked as
5005 ;; template/generic delimiters. (This marking is via syntax-table
5006 ;; text properties).
5007 ;;
5008 ;; These pairs are those which are in the current "statement" (i.e.,
5009 ;; the region between the {, }, or ; before BEG and the one after
5010 ;; END), and which enclose any part of the interval (BEG END).
5011 ;;
5012 ;; Note that in C++ (?and Java), template/generic parens cannot
5013 ;; enclose a brace or semicolon, so we use these as bounds on the
5014 ;; region we must work on.
5015 ;;
5016 ;; This function is called from before-change-functions (via
5017 ;; c-get-state-before-change-functions). Thus the buffer is widened,
5018 ;; and point is undefined, both at entry and exit.
5019 ;;
5020 ;; FIXME!!! This routine ignores the possibility of macros entirely.
5021 ;; 2010-01-29.
5025 ;; Locate the barrier before the changed region
5029 ;; Remove the syntax-table properties from each pertinent <...> pair.
5030 ;; Firsly, the ones with the < before beg and > after beg.
5034 ;; Locate the barrier after END.
5039 ;; Remove syntax-table properties from the remaining pertinent <...>
5040 ;; pairs, those with a > after end and < before end.
5047 ;; This is called from `after-change-functions' when
5048 ;; c-recognize-<>-arglists' is set. It ensures that no "<" or ">"
5049 ;; chars with paren syntax become part of another operator like "<<"
5050 ;; or ">=".
5051 ;;
5052 ;; This function might do hidden buffer changes.
5085 \f
5086 ;; Handling of small scale constructs like types and names.
5088 ;; Dynamically bound variable that instructs `c-forward-type' to also
5089 ;; treat possible types (i.e. those that it normally returns 'maybe or
5090 ;; 'found for) as actual types (and always return 'found for them).
5091 ;; This means that it records them in `c-record-type-identifiers' if
5092 ;; that is set, and that it adds them to `c-found-types'.
5095 ;; Dynamically bound variable that instructs `c-forward-<>-arglist' to
5096 ;; mark up successfully parsed arglists with paren syntax properties on
5097 ;; the surrounding angle brackets and with `c-<>-arg-sep' in the
5098 ;; `c-type' property of each argument separating comma.
5099 ;;
5100 ;; Setting this variable also makes `c-forward-<>-arglist' recurse into
5101 ;; all arglists for side effects (i.e. recording types), otherwise it
5102 ;; exploits any existing paren syntax properties to quickly jump to the
5103 ;; end of already parsed arglists.
5104 ;;
5105 ;; Marking up the arglists is not the default since doing that correctly
5106 ;; depends on a proper value for `c-restricted-<>-arglists'.
5109 ;; Dynamically bound variable that instructs `c-forward-<>-arglist' to
5110 ;; not accept arglists that contain binary operators.
5111 ;;
5112 ;; This is primarily used to handle C++ template arglists. C++
5113 ;; disambiguates them by checking whether the preceding name is a
5114 ;; template or not. We can't do that, so we assume it is a template
5115 ;; if it can be parsed as one. That usually works well since
5116 ;; comparison expressions on the forms "a < b > c" or "a < b, c > d"
5117 ;; in almost all cases would be pointless.
5118 ;;
5119 ;; However, in function arglists, e.g. in "foo (a < b, c > d)", we
5120 ;; should let the comma separate the function arguments instead. And
5121 ;; in a context where the value of the expression is taken, e.g. in
5122 ;; "if (a < b || c > d)", it's probably not a template.
5125 ;; Dynamically bound variables that instructs
5126 ;; `c-forward-keyword-clause', `c-forward-<>-arglist',
5127 ;; `c-forward-name', `c-forward-type', `c-forward-decl-or-cast-1', and
5128 ;; `c-forward-label' to record the ranges of all the type and
5129 ;; reference identifiers they encounter. They will build lists on
5130 ;; these variables where each element is a cons of the buffer
5131 ;; positions surrounding each identifier. This recording is only
5132 ;; activated when `c-record-type-identifiers' is non-nil.
5133 ;;
5134 ;; All known types that can't be identifiers are recorded, and also
5135 ;; other possible types if `c-promote-possible-types' is set.
5136 ;; Recording is however disabled inside angle bracket arglists that
5137 ;; are encountered inside names and other angle bracket arglists.
5138 ;; Such occurrences are taken care of by `c-font-lock-<>-arglists'
5139 ;; instead.
5140 ;;
5141 ;; Only the names in C++ template style references (e.g. "tmpl" in
5142 ;; "tmpl<a,b>::foo") are recorded as references, other references
5143 ;; aren't handled here.
5144 ;;
5145 ;; `c-forward-label' records the label identifier(s) on
5146 ;; `c-record-ref-identifiers'.
5150 ;; This variable will receive a cons cell of the range of the last
5151 ;; single identifier symbol stepped over by `c-forward-name' if it's
5152 ;; successful. This is the range that should be put on one of the
5153 ;; record lists above by the caller. It's assigned nil if there's no
5154 ;; such symbol in the name.
5159 ;; Always true.
5169 ;; Always true.
5177 ;; Dynamically bound variable that instructs `c-forward-type' to
5178 ;; record the ranges of types that only are found. Behaves otherwise
5179 ;; like `c-record-type-identifiers'.
5183 ;; Used internally in `c-forward-keyword-clause' to move forward
5184 ;; over a type (if TYPE is 'type) or a name (otherwise) which
5185 ;; possibly is prefixed by keywords and their associated clauses.
5186 ;; Try with a type/name first to not trip up on those that begin
5187 ;; with a keyword. Return t if a known or found type is moved
5188 ;; over. The point is clobbered if nil is returned. If range
5189 ;; recording is enabled, the identifier is recorded on as a type
5190 ;; if TYPE is 'type or as a reference if TYPE is 'ref.
5191 ;;
5192 ;; This macro might do hidden buffer changes.
5197 nil
5204 t)))
5207 ;; Used internally in `c-forward-keyword-clause' to move forward
5208 ;; over a comma separated list of types or names using
5209 ;; `c-forward-keyword-prefixed-id'.
5210 ;;
5211 ;; This macro might do hidden buffer changes.
5222 ;; Submatch MATCH in the current match data is assumed to surround a
5223 ;; token. If it's a keyword, move over it and any immediately
5224 ;; following clauses associated with it, stopping at the start of
5225 ;; the next token. t is returned in that case, otherwise the point
5226 ;; stays and nil is returned. The kind of clauses that are
5227 ;; recognized are those specified by `c-type-list-kwds',
5228 ;; `c-ref-list-kwds', `c-colon-type-list-kwds',
5229 ;; `c-paren-nontype-kwds', `c-paren-type-kwds', `c-<>-type-kwds',
5230 ;; and `c-<>-arglist-kwds'.
5231 ;;
5232 ;; This function records identifier ranges on
5233 ;; `c-record-type-identifiers' and `c-record-ref-identifiers' if
5234 ;; `c-record-type-identifiers' is non-nil.
5235 ;;
5236 ;; Note that for `c-colon-type-list-kwds', which doesn't necessary
5237 ;; apply directly after the keyword, the type list is moved over
5238 ;; only when there is no unaccounted token before it (i.e. a token
5239 ;; that isn't moved over due to some other keyword list). The
5240 ;; identifier ranges in the list are still recorded if that should
5241 ;; be done, though.
5242 ;;
5243 ;; This function might do hidden buffer changes.
5246 ;; The call to `c-forward-<>-arglist' below is made after
5247 ;; `c-<>-sexp-kwds' keywords, so we're certain they actually
5248 ;; are angle bracket arglists and `c-restricted-<>-arglists'
5249 ;; should therefore be nil.
5261 ;; There's a type directly after a keyword in `c-type-list-kwds'.
5266 ;; There's a name directly after a keyword in `c-ref-list-kwds'.
5271 ;; There's an open paren after a keyword in `c-paren-any-kwds'.
5278 ;; Use `c-forward-type' on every identifier we can find
5279 ;; inside the paren, to record the types.
5304 ;; If we are at the colon already, we move over the type
5305 ;; list after it.
5311 ;; Not at the colon, so stop here. But the identifier
5312 ;; ranges in the type list later on should still be
5313 ;; recorded.
5316 ;; If a keyword matched both one of the types above and
5317 ;; this one, we match `c-colon-type-list-re' after the
5318 ;; clause matched above.
5325 ;; There's a type after the `c-colon-type-list-re' match
5326 ;; after a keyword in `c-colon-type-list-kwds'.
5330 t)))
5333 ;; The point is assumed to be at a "<". Try to treat it as the open
5334 ;; paren of an angle bracket arglist and move forward to the
5335 ;; corresponding ">". If successful, the point is left after the
5336 ;; ">" and t is returned, otherwise the point isn't moved and nil is
5337 ;; returned. If ALL-TYPES is t then all encountered arguments in
5338 ;; the arglist that might be types are treated as found types.
5339 ;;
5340 ;; The variable `c-parse-and-markup-<>-arglists' controls how this
5341 ;; function handles text properties on the angle brackets and argument
5342 ;; separating commas.
5343 ;;
5344 ;; `c-restricted-<>-arglists' controls how lenient the template
5345 ;; arglist recognition should be.
5346 ;;
5347 ;; This function records identifier ranges on
5348 ;; `c-record-type-identifiers' and `c-record-ref-identifiers' if
5349 ;; `c-record-type-identifiers' is non-nil.
5350 ;;
5351 ;; This function might do hidden buffer changes.
5354 ;; If `c-record-type-identifiers' is set then activate
5355 ;; recording of any found types that constitute an argument in
5356 ;; the arglist.
5364 ;; `nconc' doesn't mind that the tail of
5365 ;; `c-record-found-types' is t.
5367 t)
5370 nil)))
5373 ;; Recursive part of `c-forward-<>-arglist'.
5374 ;;
5375 ;; This function might do hidden buffer changes.
5378 ;; Cover this so that any recorded found type ranges are
5379 ;; automatically lost if it turns out to not be an angle
5380 ;; bracket arglist. It's propagated through the return value
5381 ;; on successful completion.
5383 ;; List that collects the positions after the argument
5384 ;; separating ',' in the arglist.
5385 arg-start-pos)
5387 ;; If the '<' has paren open syntax then we've marked it as an angle
5388 ;; bracket arglist before, so skip to the end.
5396 t
5398 ;; Got unmatched paren angle brackets. We don't clear the paren
5399 ;; syntax properties and retry, on the basis that it's very
5400 ;; unlikely that paren angle brackets become operators by code
5401 ;; manipulation. It's far more likely that it doesn't match due
5402 ;; to narrowing or some temporary change.
5404 nil))
5414 ;; All encountered identifiers are types, so set the
5415 ;; promote flag and parse the type.
5422 ;; Check if this arglist argument is a sole type. If
5423 ;; it's known then it's recorded in
5424 ;; `c-record-type-identifiers'. If it only is found
5425 ;; then it's recorded in `c-record-found-types' which we
5426 ;; might roll back if it turns out that this isn't an
5427 ;; angle bracket arglist afterall.
5433 ;; A found type was recorded but it's not the
5434 ;; only thing in the arglist argument, so reset
5435 ;; `c-record-found-types'.
5437 orig-record-found-types))))))
5441 ;; Must check for '>' at the very start separately,
5442 ;; since the regexp below has to avoid ">>" without
5443 ;; using \\=.
5445 t)
5447 ;; Note: These regexps exploit the match order in \| so
5448 ;; that "<>" is matched by "<" rather than "[^>:-]>".
5451 ;; Stop on ',', '|', '&', '+' and '-' to catch
5452 ;; common binary operators that could be between
5453 ;; two comparison expressions "a<b" and "c>d".
5455 ;; Otherwise we still stop on ',' to find the
5456 ;; argument start positions.
5460 ;; If the arglist starter has lost its open paren
5461 ;; syntax but not the closer, we won't find the
5462 ;; closer above since we only search in the
5463 ;; balanced sexp. In that case we stop just short
5464 ;; of it so check if the following char is the closer.
5467 t)))
5471 ;; Either an operator starting with '>' or the end of
5472 ;; the angle bracket arglist.
5479 ;; The angle bracket arglist is finished.
5491 ;; Either an operator starting with '<' or a nested arglist.
5503 ;; There's always an identifier before an angle
5504 ;; bracket arglist, or a keyword in
5505 ;; `c-<>-type-kwds' or `c-<>-arglist-kwds'.
5522 )))
5524 ;; It was not an angle bracket arglist.
5527 ;; It was an angle bracket arglist.
5530 ;; Record the identifier before the template as a type
5531 ;; or reference depending on whether the arglist is last
5532 ;; in a qualified identifier.
5541 t)
5545 ;; Just another argument. Record the position. The
5546 ;; type check stuff that made us stop at it is at
5547 ;; the top of the loop.
5551 ;; Got a character that can't be in an angle bracket
5552 ;; arglist argument. Abort using `throw', since
5553 ;; it's useless to try to find a surrounding arglist
5554 ;; if we're nested.
5561 ;; The point is assumed to be directly after a ">". Try to treat it
5562 ;; as the close paren of an angle bracket arglist and move back to
5563 ;; the corresponding "<". If successful, the point is left at
5564 ;; the "<" and t is returned, otherwise the point isn't moved and
5565 ;; nil is returned. ALL-TYPES is passed on to
5566 ;; `c-forward-<>-arglist'.
5567 ;;
5568 ;; If the optional LIMIT is given, it bounds the backward search.
5569 ;; It's then assumed to be at a syntactically relevant position.
5570 ;;
5571 ;; This is a wrapper around `c-forward-<>-arglist'. See that
5572 ;; function for more details.
5581 t
5582 ;; See corresponding note in `c-forward-<>-arglist'.
5584 nil)
5591 t
5592 ;; Stopped at bob or a char that isn't allowed in an
5593 ;; arglist, so we've failed.
5595 nil)
5600 ;; If we moved then the "<" was part of some
5601 ;; multicharacter token.
5602 t
5608 ;; Matched the arglist. Break the while.
5610 nil)
5612 ;; We started from a non-paren ">" inside an
5613 ;; arglist.
5615 nil)
5617 ;; Matched a shorter arglist. Can be a nested
5618 ;; one so continue looking.
5620 t))
5621 t))))))
5626 ;; Move forward over a complete name if at the beginning of one,
5627 ;; stopping at the next following token. If the point is not at
5628 ;; something that are recognized as name then it stays put. A name
5629 ;; could be something as simple as "foo" in C or something as
5630 ;; complex as "X<Y<class A<int>::B, BIT_MAX >> b>, ::operator<> ::
5631 ;; Z<(a>b)> :: operator const X<&foo>::T Q::G<unsigned short
5632 ;; int>::*volatile const" in C++ (this function is actually little
5633 ;; more than a `looking-at' call in all modes except those that,
5634 ;; like C++, have `c-recognize-<>-arglists' set). Return nil if no
5635 ;; name is found, 'template if it's an identifier ending with an
5636 ;; angle bracket arglist, 'operator of it's an operator identifier,
5637 ;; or t if it's some other kind of name.
5638 ;;
5639 ;; This function records identifier ranges on
5640 ;; `c-record-type-identifiers' and `c-record-ref-identifiers' if
5641 ;; `c-record-type-identifiers' is non-nil.
5642 ;;
5643 ;; This function might do hidden buffer changes.
5646 ;; Turn off `c-promote-possible-types' here since we might
5647 ;; call `c-forward-<>-arglist' and we don't want it to promote
5648 ;; every suspect thing in the arglist to a type. We're
5649 ;; typically called from `c-forward-type' in this case, and
5650 ;; the caller only wants the top level type that it finds to
5651 ;; be promoted.
5652 c-promote-possible-types)
5658 ;; Check for keyword. We go to the last symbol in
5659 ;; `c-identifier-key' first.
5672 ;; "template" is only valid inside an
5673 ;; identifier if preceded by "::".
5678 t))
5680 ;; Handle a C++ operator or template identifier.
5684 ;; Got "... ::template".
5688 res subres))))
5691 ;; Got a cast operator.
5695 ;; Now we should match a sequence of either
5696 ;; '*', '&' or a name followed by ":: *",
5697 ;; where each can be followed by a sequence
5698 ;; of `c-opt-type-modifier-key'.
5701 t)
5711 t))))
5719 ;; Got some other operator.
5727 nil)
5729 ;; `id-start' is equal to `id-end' if we've jumped over
5730 ;; an identifier that doesn't end with a symbol token.
5731 ;; That can occur e.g. for Java import directives on the
5732 ;; form "foo.bar.*".
5739 res t)))
5749 ;; Got a concatenated identifier. This handles the
5750 ;; cases with tricky syntactic whitespace that aren't
5751 ;; covered in `c-identifier-key'.
5754 t)
5758 ;; Maybe an angle bracket arglist.
5761 c-record-found-types)
5767 c-last-identifier-range nil)
5772 ;; Continue if there's an identifier concatenation
5773 ;; operator after the template argument.
5779 t)
5784 nil)))
5785 )))))
5788 res))
5791 ;; Move forward over a type spec if at the beginning of one,
5792 ;; stopping at the next following token. Return t if it's a known
5793 ;; type that can't be a name or other expression, 'known if it's an
5794 ;; otherwise known type (according to `*-font-lock-extra-types'),
5795 ;; 'prefix if it's a known prefix of a type, 'found if it's a type
5796 ;; that matches one in `c-found-types', 'maybe if it's an identfier
5797 ;; that might be a type, or nil if it can't be a type (the point
5798 ;; isn't moved then). The point is assumed to be at the beginning
5799 ;; of a token.
5800 ;;
5801 ;; Note that this function doesn't skip past the brace definition
5802 ;; that might be considered part of the type, e.g.
5803 ;; "enum {a, b, c} foo".
5804 ;;
5805 ;; This function records identifier ranges on
5806 ;; `c-record-type-identifiers' and `c-record-ref-identifiers' if
5807 ;; `c-record-type-identifiers' is non-nil.
5808 ;;
5809 ;; This function might do hidden buffer changes.
5813 ;; Skip leading type modifiers. If any are found we know it's a
5814 ;; prefix of a type.
5823 ;; Looking at a keyword that prefixes a type identifier,
5824 ;; e.g. "class".
5831 ;; In many languages the name can be used without the
5832 ;; prefix, so we add it to `c-found-types'.
5835 c-last-identifier-range)
5838 ;; Invalid syntax.
5850 id-range c-last-identifier-range))))
5862 id-end)
5864 ;; Looking at a primitive or known type identifier. We've
5865 ;; checked for a name first so that we don't go here if the
5866 ;; known type match only is a prefix of another name.
5877 ;; There might be more keywords for the type.
5905 ;; A normal identifier.
5915 ;; It's an identifier that has been used as
5916 ;; a type somewhere else.
5917 'found
5918 ;; It's an identifier that might be a type.
5921 ;; A template is a type.
5925 ;; Otherwise it's an operator identifier, which is not a type.
5930 ;; Skip trailing type modifiers. If any are found we know it's
5931 ;; a type.
5938 ;; Step over any type suffix operator. Do not let the existence
5939 ;; of these alter the classification of the found type, since
5940 ;; these operators typically are allowed in normal expressions
5941 ;; too.
5948 ;; Look for a trailing operator that concatenates the type
5949 ;; with a following one, and if so step past that one through
5950 ;; a recursive call. Note that we don't record concatenated
5951 ;; types in `c-found-types' - it's the component types that
5952 ;; are recorded when appropriate.
5955 c-promote-possible-types))
5956 ;; If we can't promote then set `c-record-found-types' so that
5957 ;; we can merge in the types from the second part afterwards if
5958 ;; it turns out to be a known type there.
5961 subres)
5970 ;; If either operand certainly is a type then both are, but we
5971 ;; don't let the existence of the operator itself promote two
5972 ;; uncertain types to a certain one.
5991 ;; Merge in the ranges of any types found by the second
5992 ;; `c-forward-type'.
5994 ;; `nconc' doesn't mind that the tail of
5995 ;; `c-record-found-types' is t.
5997 c-record-type-identifiers))))
6005 ;;(message "c-forward-type %s -> %s: %s" start (point) res)
6007 res))
6009 \f
6010 ;; Handling of large scale constructs like statements and declarations.
6012 ;; Macro used inside `c-forward-decl-or-cast-1'. It ought to be a
6013 ;; defsubst or perhaps even a defun, but it contains lots of free
6014 ;; variables that refer to things inside `c-forward-decl-or-cast-1'.
6016 ;; `c-forward-decl-or-cast-1' can consume an arbitrary length list
6017 ;; of types when parsing a declaration, which means that it
6018 ;; sometimes consumes the identifier in the declaration as a type.
6019 ;; This is used to "backtrack" and make the last type be treated as
6020 ;; an identifier instead.
6023 ;; These identifiers are bound only in the inner let.
6025 identifier-start type-start
6026 got-parens nil
6027 got-identifier t
6028 got-suffix t
6029 got-suffix-after-parens id-start
6033 t
6034 backup-at-type))
6036 id-start backup-id-start)
6038 id-start start-pos))
6040 ;; When these flags already are set we've found specifiers that
6041 ;; unconditionally signal these attributes - backtracking doesn't
6042 ;; change that. So keep them set in that case.
6049 ;; This identifier is bound only in the inner let.
6053 ;; Move forward over a declaration or a cast if at the start of one.
6054 ;; The point is assumed to be at the start of some token. Nil is
6055 ;; returned if no declaration or cast is recognized, and the point
6056 ;; is clobbered in that case.
6057 ;;
6058 ;; If a declaration is parsed:
6059 ;;
6060 ;; The point is left at the first token after the first complete
6061 ;; declarator, if there is one. The return value is a cons where
6062 ;; the car is the position of the first token in the declarator. (See
6063 ;; below for the cdr.)
6064 ;; Some examples:
6065 ;;
6066 ;; void foo (int a, char *b) stuff ...
6067 ;; car ^ ^ point
6068 ;; float (*a)[], b;
6069 ;; car ^ ^ point
6070 ;; unsigned int a = c_style_initializer, b;
6071 ;; car ^ ^ point
6072 ;; unsigned int a (cplusplus_style_initializer), b;
6073 ;; car ^ ^ point (might change)
6074 ;; class Foo : public Bar {}
6075 ;; car ^ ^ point
6076 ;; class PikeClass (int a, string b) stuff ...
6077 ;; car ^ ^ point
6078 ;; enum bool;
6079 ;; car ^ ^ point
6080 ;; enum bool flag;
6081 ;; car ^ ^ point
6082 ;; void cplusplus_function (int x) throw (Bad);
6083 ;; car ^ ^ point
6084 ;; Foo::Foo (int b) : Base (b) {}
6085 ;; car ^ ^ point
6086 ;;
6087 ;; The cdr of the return value is non-nil iff a `c-typedef-decl-kwds'
6088 ;; specifier (e.g. class, struct, enum, typedef) is found in the
6089 ;; declaration, i.e. the declared identifier(s) are types.
6090 ;;
6091 ;; If a cast is parsed:
6092 ;;
6093 ;; The point is left at the first token after the closing paren of
6094 ;; the cast. The return value is `cast'. Note that the start
6095 ;; position must be at the first token inside the cast parenthesis
6096 ;; to recognize it.
6097 ;;
6098 ;; PRECEDING-TOKEN-END is the first position after the preceding
6099 ;; token, i.e. on the other side of the syntactic ws from the point.
6100 ;; Use a value less than or equal to (point-min) if the point is at
6101 ;; the first token in (the visible part of) the buffer.
6102 ;;
6103 ;; CONTEXT is a symbol that describes the context at the point:
6104 ;; 'decl In a comma-separated declaration context (typically
6105 ;; inside a function declaration arglist).
6106 ;; '<> In an angle bracket arglist.
6107 ;; 'arglist Some other type of arglist.
6108 ;; nil Some other context or unknown context. Includes
6109 ;; within the parens of an if, for, ... construct.
6110 ;;
6111 ;; LAST-CAST-END is the first token after the closing paren of a
6112 ;; preceding cast, or nil if none is known. If
6113 ;; `c-forward-decl-or-cast-1' is used in succession, it should be
6114 ;; the position after the closest preceding call where a cast was
6115 ;; matched. In that case it's used to discover chains of casts like
6116 ;; "(a) (b) c".
6117 ;;
6118 ;; This function records identifier ranges on
6119 ;; `c-record-type-identifiers' and `c-record-ref-identifiers' if
6120 ;; `c-record-type-identifiers' is non-nil.
6121 ;;
6122 ;; This function might do hidden buffer changes.
6125 ;; first type, i.e. after any leading specifiers. It might
6126 ;; also point at the beginning of the preceding syntactic
6127 ;; whitespace.
6129 ;; Set to the result of `c-forward-type'.
6130 at-type
6131 ;; The position of the first token in what we currently
6132 ;; believe is the type in the declaration or cast, after any
6133 ;; specifiers and their associated clauses.
6134 type-start
6135 ;; The position of the first token in what we currently
6136 ;; believe is the declarator for the first identifier. Set
6137 ;; when the type is found, and moved forward over any
6138 ;; `c-decl-hangon-kwds' and their associated clauses that
6139 ;; occurs after the type.
6140 id-start
6141 ;; These store `at-type', `type-start' and `id-start' of the
6142 ;; identifier before the one in those variables. The previous
6143 ;; identifier might turn out to be the real type in a
6144 ;; declaration if the last one has to be the declarator in it.
6145 ;; If `backup-at-type' is nil then the other variables have
6146 ;; undefined values.
6147 backup-at-type backup-type-start backup-id-start
6148 ;; Set if we've found a specifier that makes the defined
6149 ;; identifier(s) types.
6150 at-type-decl
6151 ;; Set if we've found a specifier that can start a declaration
6152 ;; where there's no type.
6153 maybe-typeless
6154 ;; If a specifier is found that also can be a type prefix,
6155 ;; these flags are set instead of those above. If we need to
6156 ;; back up an identifier, they are copied to the real flag
6157 ;; variables. Thus they only take effect if we fail to
6158 ;; interpret it as a type.
6159 backup-at-type-decl backup-maybe-typeless
6160 ;; Whether we've found a declaration or a cast. We might know
6161 ;; this before we've found the type in it. It's 'ids if we've
6162 ;; found two consecutive identifiers (usually a sure sign, but
6163 ;; we should allow that in labels too), and t if we've found a
6164 ;; specifier keyword (a 100% sure sign).
6165 at-decl-or-cast
6166 ;; Set when we need to back up to parse this as a declaration
6167 ;; but not as a cast.
6168 backup-if-not-cast
6169 ;; For casts, the return position.
6170 cast-end
6171 ;; Save `c-record-type-identifiers' and
6172 ;; `c-record-ref-identifiers' since ranges are recorded
6173 ;; speculatively and should be thrown away if it turns out
6174 ;; that it isn't a declaration or cast.
6178 ;; Check for a type. Unknown symbols are treated as possible
6179 ;; types, but they could also be specifiers disguised through
6180 ;; macros like __INLINE__, so we recognize both types and known
6181 ;; specifiers after them too.
6185 ;; Look for a specifier keyword clause.
6193 ;; Found a known or possible type or a prefix of a known type.
6196 ;; Got two identifiers with nothing but whitespace
6197 ;; between them. That can only happen in declarations.
6201 ;; If the previous identifier is a found type we
6202 ;; record it as a real one; it might be some sort of
6203 ;; alias for a prefix like "unsigned".
6210 backup-type-start type-start
6211 backup-id-start id-start
6212 at-type found-type
6213 type-start start
6215 ;; The previous ambiguous specifier/type turned out
6216 ;; to be a type since we've parsed another one after
6217 ;; it, so clear these backup flags.
6218 backup-at-type-decl nil
6219 backup-maybe-typeless nil))
6223 ;; Handle known specifier keywords and
6224 ;; `c-decl-hangon-kwds' which can occur after known
6225 ;; types.
6228 ;; It's a hang-on keyword that can occur anywhere.
6232 ;; Move the identifier start position if
6233 ;; we've passed a type.
6235 ;; Otherwise treat this as a specifier and
6236 ;; move the fallback position.
6240 ;; It's an ordinary specifier so we know that
6241 ;; anything before this can't be the type.
6243 start-pos kwd-clause-end)
6246 ;; It's ambiguous whether this keyword is a
6247 ;; specifier or a type prefix, so set the backup
6248 ;; flags. (It's assumed that `c-forward-type'
6249 ;; moved further than `c-forward-keyword-clause'.)
6261 ;; Haven't matched a type so it's an umambiguous
6262 ;; specifier keyword and we know we're in a
6263 ;; declaration.
6268 ;; If the type isn't known we continue so that we'll jump
6269 ;; over all specifiers and type identifiers. The reason
6270 ;; to do this for a known type prefix is to make things
6271 ;; like "unsigned INT16" work.
6276 ;; If a known type was found, we still need to skip over any
6277 ;; hangon keyword clauses after it. Otherwise it has already
6278 ;; been done in the loop above.
6284 ;; A prefix type is itself a primitive type when it's not
6285 ;; followed by another type.
6289 ;; Got no type but set things up to continue anyway to handle
6290 ;; the various cases when a declaration doesn't start with a
6291 ;; type.
6296 ;; If it's C++ then check if the last "type" ends on the form
6297 ;; "foo::foo" or "foo::~foo", i.e. if it's the name of a
6298 ;; (con|de)structor.
6309 ;; Cheating in the handling of syntactic ws below.
6316 name))
6317 ;; It is a (con|de)structor name. In that case the
6318 ;; declaration is typeless so zap out any preceding
6319 ;; identifier(s) that we might have taken as types.
6322 backup-at-type nil
6323 id-start type-start))))))
6325 ;; Check for and step over a type decl expression after the thing
6326 ;; that is or might be a type. This can't be skipped since we
6327 ;; need the correct end position of the declarator for
6328 ;; `max-type-decl-end-*'.
6330 ;; True if there's a non-open-paren match of
6331 ;; `c-type-decl-prefix-key'.
6332 got-prefix
6333 ;; True if the declarator is surrounded by a parenthesis pair.
6334 got-parens
6335 ;; True if there is an identifier in the declarator.
6336 got-identifier
6337 ;; True if there's a non-close-paren match of
6338 ;; `c-type-decl-suffix-key'.
6339 got-suffix
6340 ;; True if there's a prefix match outside the outermost
6341 ;; paren pair that surrounds the declarator.
6342 got-prefix-before-parens
6343 ;; True if there's a suffix match outside the outermost
6344 ;; paren pair that surrounds the declarator. The value is
6345 ;; the position of the first suffix match.
6346 got-suffix-after-parens
6347 ;; True if we've parsed the type decl to a token that is
6348 ;; known to end declarations in this context.
6349 at-decl-end
6350 ;; The earlier values of `at-type' and `type-start' if we've
6351 ;; shifted the type backwards.
6352 identifier-type identifier-start
6353 ;; If `c-parse-and-markup-<>-arglists' is set we need to
6354 ;; turn it off during the name skipping below to avoid
6355 ;; getting `c-type' properties that might be bogus. That
6356 ;; can happen since we don't know if
6357 ;; `c-restricted-<>-arglists' will be correct inside the
6358 ;; arglist paren that gets entered.
6363 ;; Skip over type decl prefix operators. (Note similar code in
6364 ;; `c-font-lock-declarators'.)
6368 ;; If the second submatch matches in C++ then
6369 ;; we're looking at an identifier that's a
6370 ;; prefix only if it specifies a member pointer.
6373 ;; We only check for a trailing "::" and
6374 ;; let the "*" that should follow be
6375 ;; matched in the next round.
6377 ;; It turned out to be the real identifier,
6378 ;; so stop.
6379 nil))
6380 t))
6394 ;; Skip over an identifier.
6399 ;; Skip over type decl suffix operators.
6406 t)
6410 t)
6416 ;; No suffix matched. We might have matched the
6417 ;; identifier as a type and the open paren of a
6418 ;; function arglist as a type decl prefix. In that
6419 ;; case we should "backtrack": Reinterpret the last
6420 ;; type as the identifier, move out of the arglist and
6421 ;; continue searching for suffix operators.
6422 ;;
6423 ;; Do this even if there's no preceding type, to cope
6424 ;; with old style function declarations in K&R C,
6425 ;; (con|de)structors in C++ and `c-typeless-decl-kwds'
6426 ;; style declarations. That isn't applicable in an
6427 ;; arglist context, though.
6432 maybe-typeless
6433 backup-maybe-typeless
6440 t))
6447 at-type)
6448 ;; Have found no identifier but `c-typeless-decl-kwds' has
6449 ;; matched so we know we're inside a declaration. The
6450 ;; preceding type must be the identifier instead.
6454 at-decl-or-cast
6457 ;; CASE 1
6459 ;; Encountered something inside parens that isn't matched by
6460 ;; the `c-type-decl-*' regexps, so it's not a type decl
6461 ;; expression. Try to skip out to the same paren depth to
6462 ;; not confuse the cast check below.
6464 ;; If we've found a specifier keyword then it's a
6465 ;; declaration regardless.
6473 ;; Now we've collected info about various characteristics of
6474 ;; the construct we're looking at. Below follows a decision
6475 ;; tree based on that. It's ordered to check more certain
6476 ;; signs before less certain ones.
6481 ;; CASE 2
6484 ;; Got another identifier directly after the type, so it's a
6485 ;; declaration.
6492 maybe-typeless
6493 backup-maybe-typeless))
6494 ;; Got a declaration of the form "foo bar (gnu);" where we've
6495 ;; recognized "bar" as the type and "gnu" as the declarator.
6496 ;; In this case it's however more likely that "bar" is the
6497 ;; declarator and "gnu" a function argument or initializer (if
6498 ;; `c-recognize-paren-inits' is set), since the parens around
6499 ;; "gnu" would be superfluous if it's a declarator. Shift the
6500 ;; type one step backward.
6503 ;; Found no identifier.
6508 ;; CASE 3
6510 ;; Got a plain list of identifiers. If a colon follows it's
6511 ;; a valid label. Otherwise the last one probably is the
6512 ;; declared identifier and we should back up to the previous
6513 ;; type, providing it isn't a cast.
6515 ;; If we've found a specifier keyword then it's a
6516 ;; declaration regardless.
6521 ;; CASE 4
6525 ;; Got a plain list of identifiers followed by some suffix.
6526 ;; If this isn't a cast then the last identifier probably is
6527 ;; the declared one and we should back up to the previous
6528 ;; type.
6532 ;; CASE 5
6534 ;; If the type is known we know that there can't be any
6535 ;; identifier somewhere else, and it's only in declarations in
6536 ;; e.g. function prototypes and in casts that the identifier may
6537 ;; be left out.
6541 ;; Only got a single identifier (parsed as a type so far).
6542 ;; CASE 6
6544 ;; Check that the identifier isn't at the start of an
6545 ;; expression.
6546 at-decl-end
6549 ;; Inside an arglist that contains declarations. If K&R
6550 ;; style declarations and parenthesis style initializers
6551 ;; aren't allowed then the single identifier must be a
6552 ;; type, else we require that it's known or found
6553 ;; (primitive types are handled above).
6558 ;; Inside a template arglist. Accept known and found
6559 ;; types; other identifiers could just as well be
6560 ;; constants in C++.
6563 ;; CASE 7
6564 ;; Can't be a valid declaration or cast, but if we've found a
6565 ;; specifier it can't be anything else either, so treat it as
6566 ;; an invalid/unfinished declaration or cast.
6574 maybe-typeless
6575 backup-maybe-typeless
6579 c-after-suffixed-type-maybe-decl-key))))))
6580 ;; Got an empty paren pair and a preceding type that probably
6581 ;; really is the identifier. Shift the type backwards to make
6582 ;; the last one the identifier. This is analogous to the
6583 ;; "backtracking" done inside the `c-type-decl-suffix-key' loop
6584 ;; above.
6585 ;;
6586 ;; Exception: In addition to the conditions in that
6587 ;; "backtracking" code, do not shift backward if we're not
6588 ;; looking at either `c-after-suffixed-type-decl-key' or "[;,]".
6589 ;; Since there's no preceding type, the shift would mean that
6590 ;; the declaration is typeless. But if the regexp doesn't match
6591 ;; then we will simply fall through in the tests below and not
6592 ;; recognize it at all, so it's better to try it as an abstract
6593 ;; declarator instead.
6596 ;; Still no identifier.
6597 ;; CASE 8
6599 ;; Require `got-prefix' together with either `got-parens' or
6600 ;; `got-suffix' to recognize it as an abstract declarator:
6601 ;; `got-parens' only is probably an empty function call.
6602 ;; `got-suffix' only can build an ordinary expression together
6603 ;; with the preceding identifier which we've taken as a type.
6604 ;; We could actually accept on `got-prefix' only, but that can
6605 ;; easily occur temporarily while writing an expression so we
6606 ;; avoid that case anyway. We could do a better job if we knew
6607 ;; the point when the fontification was invoked.
6610 ;; CASE 9
6614 got-suffix-after-parens
6616 ;; Got a type, no declarator but a paren suffix. I.e. it's a
6617 ;; normal function call afterall (or perhaps a C++ style object
6618 ;; instantiation expression).
6621 ;; CASE 10
6623 ;; By now we've located the type in the declaration that we know
6624 ;; we're in.
6627 ;; CASE 11
6635 ;; Shift the type backward in the case that there's a
6636 ;; single identifier inside parens. That can only
6637 ;; occur in K&R style function declarations so it's
6638 ;; more likely that it really is a function call.
6639 ;; Therefore we only do this after
6640 ;; `c-after-suffixed-type-decl-key' has matched.
6642 got-suffix-after-parens))
6643 ;; A declaration according to `c-after-suffixed-type-decl-key'.
6646 ;; CASE 12
6649 ;; It's a declaration if a known type precedes it and it can't be a
6650 ;; function call.
6653 ;; If we get here we can't tell if this is a type decl or a normal
6654 ;; expression by looking at it alone. (That's under the assumption
6655 ;; that normal expressions always can look like type decl expressions,
6656 ;; which isn't really true but the cases where it doesn't hold are so
6657 ;; uncommon (e.g. some placements of "const" in C++) it's not worth
6658 ;; the effort to look for them.)
6661 ;; If this is a declaration it should end here or its initializer(*)
6662 ;; should start here, so check for allowed separation tokens. Note
6663 ;; that this rule doesn't work e.g. with a K&R arglist after a
6664 ;; function header.
6665 ;;
6666 ;; *) Don't check for C++ style initializers using parens
6667 ;; since those already have been matched as suffixes.
6668 ;;
6669 ;; If `at-decl-or-cast' is then we've found some other sign that
6670 ;; it's a declaration or cast, so then it's probably an
6671 ;; invalid/unfinished one.
6674 ;; Below are tests that only should be applied when we're certain to
6675 ;; not have parsed halfway through an expression.
6677 ;; CASE 14
6679 ;; The expression starts with a known type so treat it as a
6680 ;; declaration.
6683 ;; CASE 15
6685 ;; In C++ we check if the identifier is a known type, since
6686 ;; (con|de)structors use the class name as identifier.
6687 ;; We've always shifted over the identifier as a type and
6688 ;; then backed up again in this case.
6689 identifier-type
6692 ;; `at-type' probably won't be 'found for
6693 ;; destructors since the "~" is then part of the
6694 ;; type name being checked against the list of
6695 ;; known types, so do a check without that
6696 ;; operator.
6701 c-identifier-syntax-table
6705 ;; We have already parsed the type earlier,
6706 ;; so it'd be possible to cache the end
6707 ;; position instead of redoing it here, but
6708 ;; then we'd need to keep track of another
6709 ;; position everywhere.
6717 ;; CASE 16
6719 at-type
6723 ;; Got something like "foo * bar;". Since we're not inside an
6724 ;; arglist it would be a meaningless expression because the
6725 ;; result isn't used. We therefore choose to recognize it as
6726 ;; a declaration. Do not allow a suffix since it could then
6727 ;; be a function call.
6730 ;; CASE 17
6735 ;; Got something like "a (*b) (c);" or "a (b) = c;". It could
6736 ;; be an odd expression or it could be a declaration. Treat
6737 ;; it as a declaration if "a" has been used as a type
6738 ;; somewhere else (if it's a known type we won't get here).
6741 ;; CASE 18
6747 ;; Got a type followed by an abstract declarator. If `got-prefix'
6748 ;; is set it's something like "a *" without anything after it. If
6749 ;; `got-parens' or `got-suffix' is set it's "a()", "a[]", "a()[]",
6750 ;; or similar, which we accept only if the context rules out
6751 ;; expressions.
6754 ;; If we had a complete symbol table here (which rules out
6755 ;; `c-found-types') we should return t due to the disambiguation rule
6756 ;; (in at least C++) that anything that can be parsed as a declaration
6757 ;; is a declaration. Now we're being more defensive and prefer to
6758 ;; highlight things like "foo (bar);" as a declaration only if we're
6759 ;; inside an arglist that contains declarations.
6762 ;; The point is now after the type decl expression.
6765 ;; Check for a cast.
6768 c-cast-parens
6770 ;; Should be the first type/identifier in a cast paren.
6774 ;; The closing paren should follow.
6779 ;; There should be a primary expression after it.
6788 ;; Check if the expression begins with a prefix keyword.
6791 ;; Expression begins with an ambiguous operator. Treat
6792 ;; it as a cast if it's a type decl or if we've
6793 ;; recognized the type somewhere else.
6796 ;; Unless it's a keyword, it's the beginning of a primary
6797 ;; expression.
6799 ;; If `c-primary-expr-regexp' matched a nonsymbol token, check
6800 ;; that it matched a whole one so that we don't e.g. confuse
6801 ;; the operator '-' with '->'. It's ok if it matches further,
6802 ;; though, since it e.g. can match the float '.5' while the
6803 ;; operator regexp only matches '.'.
6807 ;; There should either be a cast before it or something that isn't an
6808 ;; identifier or close paren.
6816 ;; It's a symbol. Accept it only if it's one of the
6817 ;; keywords that can precede an expression (without
6818 ;; surrounding parens).
6821 ;; Check that it isn't a close paren (block close is ok,
6822 ;; though).
6824 ;; Check that it isn't a nonsymbol identifier.
6827 ;; Handle the cast.
6837 ;; We're at a declaration. Highlight the type and the following
6838 ;; declarators.
6844 ;; Make sure to propagate the `c-decl-arg-start' property to
6845 ;; the next argument if it's set in this one, to cope with
6846 ;; interactive refontification.
6858 ;; False alarm. Restore the recorded ranges.
6860 c-record-ref-identifiers save-rec-ref-ids)
6861 nil))))
6864 ;; Assuming that point is at the beginning of a token, check if it starts a
6865 ;; label and if so move over it and return non-nil (t in default situations,
6866 ;; specific symbols (see below) for interesting situations), otherwise don't
6867 ;; move and return nil. "Label" here means "most things with a colon".
6868 ;;
6869 ;; More precisely, a "label" is regarded as one of:
6870 ;; (i) a goto target like "foo:" - returns the symbol `goto-target';
6871 ;; (ii) A case label - either the entire construct "case FOO:", or just the
6872 ;; bare "case", should the colon be missing. We return t;
6873 ;; (iii) a keyword which needs a colon, like "default:" or "private:"; We
6874 ;; return t;
6875 ;; (iv) One of QT's "extended" C++ variants of
6876 ;; "private:"/"protected:"/"public:"/"more:" looking like "public slots:".
6877 ;; Returns the symbol `qt-2kwds-colon'.
6878 ;; (v) QT's construct "signals:". Returns the symbol `qt-1kwd-colon'.
6879 ;; (vi) One of the keywords matched by `c-opt-extra-label-key' (without any
6880 ;; colon). Currently (2006-03), this applies only to Objective C's
6881 ;; keywords "@private", "@protected", and "@public". Returns t.
6882 ;;
6883 ;; One of the things which will NOT be recognised as a label is a bit-field
6884 ;; element of a struct, something like "int foo:5".
6885 ;;
6886 ;; The end of the label is taken to be just after the colon, or the end of
6887 ;; the first submatch in `c-opt-extra-label-key'. The point is directly
6888 ;; after the end on return. The terminating char gets marked with
6889 ;; `c-decl-end' to improve recognition of the following declaration or
6890 ;; statement.
6891 ;;
6892 ;; If ASSUME-MARKUP is non-nil, it's assumed that the preceding
6893 ;; label, if any, has already been marked up like that.
6894 ;;
6895 ;; If PRECEDING-TOKEN-END is given, it should be the first position
6896 ;; after the preceding token, i.e. on the other side of the
6897 ;; syntactic ws from the point. Use a value less than or equal to
6898 ;; (point-min) if the point is at the first token in (the visible
6899 ;; part of) the buffer.
6900 ;;
6901 ;; The optional LIMIT limits the forward scan for the colon.
6902 ;;
6903 ;; This function records the ranges of the label symbols on
6904 ;; `c-record-ref-identifiers' if `c-record-type-identifiers' (!) is
6905 ;; non-nil.
6906 ;;
6907 ;; This function might do hidden buffer changes.
6910 label-end
6911 qt-symbol-idx
6912 macro-start ; if we're in one.
6913 label-type
6914 kwd)
6916 ;; "case" or "default" (Doesn't apply to AWK).
6919 ;; Record only the keyword itself for fontification, since in
6920 ;; case labels the following is a constant expression and not
6921 ;; a label.
6925 ;; Find the label end.
6929 ;; Stop on chars that aren't allowed in expressions,
6930 ;; and on operator chars that would be meaningless
6931 ;; there. FIXME: This doesn't cope with ?: operators.
6939 t)
6941 ;; It's an unfinished label. We consider the keyword enough
6942 ;; to recognize it as a label, so that it gets fontified.
6943 ;; Leave the point at the end of it, but don't put any
6944 ;; `c-decl-end' marker.
6946 t))))
6948 ;; @private, @protected, @public, in Objective C, or similar.
6951 ;; For a `c-opt-extra-label-key' match, we record the whole
6952 ;; thing for fontification. That's to get the leading '@' in
6953 ;; Objective-C protection labels fontified.
6960 ;; All other cases of labels.
6963 ;; A colon label must have something before the colon.
6966 ;; Check that we're not after a token that can't precede a label.
6968 ;; Trivially succeeds when there's no preceding token.
6976 ;; Check if we're after a label, if we're after a closing
6977 ;; paren that belong to statement, and with
6978 ;; `c-label-prefix-re'. It's done in different order
6979 ;; depending on `assume-markup' since the checks have
6980 ;; different expensiveness.
7007 ;; Might be after another label, so check it recursively.
7011 ;; Essentially the same as the
7012 ;; `c-syntactic-re-search-forward' regexp below.
7018 ;; Note: the following should work instead of the
7019 ;; narrow-to-region above. Investigate why not,
7020 ;; sometime. ACM, 2006-03-31.
7021 ;; (c-syntactic-skip-backward "^-]:?;}=*/%&|,<>!@+"
7022 ;; macro-start t)
7024 ;; If the caller turned on recording for us,
7025 ;; it shouldn't apply when we check the
7026 ;; preceding label.
7027 c-record-type-identifiers)
7028 ;; A label can't start at a cpp directive. Check for
7029 ;; this, since c-forward-syntactic-ws would foul up on it.
7034 ;; Point is still at the beginning of the possible label construct.
7035 ;;
7036 ;; Check that the next nonsymbol token is ":", or that we're in one
7037 ;; of QT's "slots" declarations. Allow '(' for the sake of macro
7038 ;; arguments. FIXME: Should build this regexp from the language
7039 ;; constants.
7041 ;; public: protected: private:
7051 ;; QT double keyword like "protected slots:" or goto target.
7070 'qt-1kwd-colon
7082 ;; Check that `c-nonlabel-token-key' doesn't match anywhere.
7092 t)
7095 ;; Record the identifiers in the label for fontification, unless
7096 ;; it begins with `c-label-kwds' in which case the following
7097 ;; identifiers are part of a (constant) expression that
7098 ;; shouldn't be fontified.
7110 ;; Not a label.
7112 label-type))
7115 ;; Assuming the point is at the beginning of a token, try to move
7116 ;; forward to the end of the Objective-C directive that starts
7117 ;; there. Return t if a directive was fully recognized, otherwise
7118 ;; the point is moved as far as one could be successfully parsed and
7119 ;; nil is returned.
7120 ;;
7121 ;; This function records identifier ranges on
7122 ;; `c-record-type-identifiers' and `c-record-ref-identifiers' if
7123 ;; `c-record-type-identifiers' is non-nil.
7124 ;;
7125 ;; This function might do hidden buffer changes.
7128 start-char
7130 ;; Turn off recognition of angle bracket arglists while parsing
7131 ;; types here since the protocol reference list might then be
7132 ;; considered part of the preceding name or superclass-name.
7143 t)
7152 ;; Handle the name of the class itself.
7154 ; (c-forward-token-2) ; 2006/1/13 This doesn't move if the token's
7155 ; at EOB.
7161 ;; Look for ": superclass-name" or "( category-name )".
7172 ;; Look for a protocol reference list.
7178 t))))
7184 t)
7187 nil)))
7190 ;; Go to the first non-whitespace after the colon that starts a
7191 ;; multiple inheritance introduction. Optional LIM is the farthest
7192 ;; back we should search.
7193 ;;
7194 ;; This function might do hidden buffer changes.
7202 ;; Return nil if we aren't in a method definition, otherwise the
7203 ;; position of the initial [+-].
7204 ;;
7205 ;; This function might do hidden buffer changes.
7211 ))
7213 ;; Contributed by Kevin Ryde <user42@zip.com.au>.
7215 ;; Return non-nil if point is within a gcc \"asm\" block.
7216 ;;
7217 ;; This should be called with point inside an argument list.
7218 ;;
7219 ;; Only one level of enclosing parentheses is considered, so for
7220 ;; instance `nil' is returned when in a function call within an asm
7221 ;; operand.
7222 ;;
7223 ;; This function might do hidden buffer changes.
7235 a function.
7237 More precisely, being at the top-level means that point is either
7238 outside any enclosing block (such as a function definition), or
7239 directly inside a class, namespace or other block that contains
7240 another declaration level.
7242 If point is not at the top-level (e.g. it is inside a method
7243 definition), then nil is returned. Otherwise, if point is at a
7244 top-level not enclosed within a class definition, t is returned.
7245 Otherwise, a 2-vector is returned where the zeroth element is the
7246 buffer position of the start of the class declaration, and the first
7247 element is the buffer position of the enclosing class's opening
7248 brace.
7250 Note that this function might do hidden buffer changes. See the
7251 comment at the start of cc-engine.el for more info."
7257 ;; Return non-nil if the point is in the region after the argument
7258 ;; list of a function and its opening brace (or semicolon in case it
7259 ;; got no body). If there are K&R style argument declarations in
7260 ;; that region, the point has to be inside the first one for this
7261 ;; function to recognize it.
7262 ;;
7263 ;; If successful, the point is moved to the first token after the
7264 ;; function header (see `c-forward-decl-or-cast-1' for details) and
7265 ;; the position of the opening paren of the function arglist is
7266 ;; returned.
7267 ;;
7268 ;; The point is clobbered if not successful.
7269 ;;
7270 ;; LIM is used as bound for backward buffer searches.
7271 ;;
7272 ;; This function might do hidden buffer changes.
7285 ;; There should not be a '=' or ',' between beg and the
7286 ;; start of the declaration since that means we were in the
7287 ;; "expression part" of the declaration.
7292 ;; Check that there's an arglist paren in the
7293 ;; declaration.
7296 ;; The declarator is a paren expression, so skip past it
7297 ;; so that we don't get stuck on that instead of the
7298 ;; function arglist.
7302 ;; Don't trip up on "operator ()".
7309 ;; Return the position of the first argument declaration if point is
7310 ;; inside a K&R style argument declaration list, nil otherwise.
7311 ;; `c-recognize-knr-p' is not checked. If LIM is non-nil, it's a
7312 ;; position that bounds the backward search for the argument list.
7313 ;;
7314 ;; Point must be within a possible K&R region, e.g. just before a top-level
7315 ;; "{". It must be outside of parens and brackets. The test can return
7316 ;; false positives otherwise.
7317 ;;
7318 ;; This function might do hidden buffer changes.
7322 ;; If we're in a macro, our search range is restricted to it. Narrow to
7323 ;; the searchable range.
7326 before-lparen after-rparen
7330 ;; Search backwards for the defun's argument list. We give up if we
7331 ;; encounter a "}" (end of a previous defun) or BOB.
7332 ;;
7333 ;; The criterion for a paren structure being the arg list is:
7334 ;; o - there is non-WS stuff after it but before any "{"; AND
7335 ;; o - the token after it isn't a ";" AND
7336 ;; o - it is preceded by either an identifier (the function name) or
7337 ;; a macro expansion like "DEFUN (...)"; AND
7338 ;; o - its content is a non-empty comma-separated list of identifiers
7339 ;; (an empty arg list won't have a knr region).
7340 ;;
7341 ;; The following snippet illustrates these rules:
7342 ;; int foo (bar, baz, yuk)
7343 ;; int bar [] ;
7344 ;; int (*baz) (my_type) ;
7345 ;; int (*) (void) (*yuk) (void) ;
7346 ;; {
7360 ;; We're inside a paren. Could it be our argument list....?
7366 ;; FIXME!!! What about macros between the parens? 2007/01/20
7369 ;; It can't be the arg list if next token is ; or {
7374 ;; Is the thing preceding the list an identifier (the
7375 ;; function name), or a macro expansion?
7385 ;; Have we got a non-empty list of comma-separated
7386 ;; identifiers?
7400 ;; ...Yes. We've identified the function's argument list.
7406 ;; ...No. The current parens aren't the function's arg list.
7413 ;; skip forward over conditional at point, including any predicate
7414 ;; statements in parentheses. No error checking is performed.
7415 ;;
7416 ;; This function might do hidden buffer changes.
7418 ;; else if()
7423 ;; do, else, try, finally
7428 ;; for, if, while, switch, catch, synchronized, foreach
7432 ;; If looking at the token after a conditional then return the
7433 ;; position of its start, otherwise return nil.
7434 ;;
7435 ;; This function might do hidden buffer changes.
7445 ;; If the point is after an operator identifier that isn't handled
7446 ;; like an ordinary symbol (i.e. like "operator =" in C++) then the
7447 ;; position of the start of that identifier is returned. nil is
7448 ;; returned otherwise. The point may be anywhere in the syntactic
7449 ;; whitespace after the last token of the operator identifier.
7450 ;;
7451 ;; This function might do hidden buffer changes.
7463 ;; Assuming point is at a brace that opens a statement block of some
7464 ;; kind, move to the proper anchor point for that block. It might
7465 ;; need to be adjusted further by c-add-stmt-syntax, but the
7466 ;; position at return is suitable as start position for that
7467 ;; function.
7468 ;;
7469 ;; This function might do hidden buffer changes.
7476 ;; Assuming point is at a brace that opens the block of a top level
7477 ;; declaration of some kind, move to the proper anchor point for
7478 ;; that block.
7479 ;;
7480 ;; This function might do hidden buffer changes.
7485 ;; Search forward for the end of the "header" of the current
7486 ;; declaration. That's the position where the definition body
7487 ;; starts, or the first variable initializer, or the ending
7488 ;; semicolon. I.e. search forward for the closest following
7489 ;; (syntactically relevant) '{', '=' or ';' token. Point is left
7490 ;; _after_ the first found token, or at point-max if none is found.
7491 ;;
7492 ;; This function might do hidden buffer changes.
7497 ;; In C++ we need to take special care to handle operator
7498 ;; tokens and those pesky template brackets.
7503 ;; Handle operator identifiers, i.e. ignore any
7504 ;; operator token preceded by "operator".
7511 t
7513 nil)))))
7522 ;; Go to the beginning of the current declaration, or the beginning
7523 ;; of the previous one if already at the start of it. Point won't
7524 ;; be moved out of any surrounding paren. Return a cons cell of the
7525 ;; form (MOVE . KNR-POS). MOVE is like the return value from
7526 ;; `c-beginning-of-statement-1'. If point skipped over some K&R
7527 ;; style argument declarations (and they are to be recognized) then
7528 ;; KNR-POS is set to the start of the first such argument
7529 ;; declaration, otherwise KNR-POS is nil. If LIM is non-nil, it's a
7530 ;; position that bounds the backward search.
7531 ;;
7532 ;; NB: Cases where the declaration continues after the block, as in
7533 ;; "struct foo { ... } bar;", are currently recognized as two
7534 ;; declarations, e.g. "struct foo { ... }" and "bar;" in this case.
7535 ;;
7536 ;; This function might do hidden buffer changes.
7542 ;; `c-beginning-of-statement-1' stops at a block start, but we
7543 ;; want to continue if the block doesn't begin a top level
7544 ;; construct, i.e. if it isn't preceded by ';', '}', ':', bob,
7545 ;; or an open paren.
7547 ;; Go back one "statement" each time round the loop until we're just
7548 ;; after a ;, }, or :, or at BOB or the start of a macro or start of
7549 ;; an ObjC method. This will move over a multiple declaration whose
7550 ;; components are comma separated.
7552 ;; Must check with c-opt-method-key in ObjC mode.
7562 ;; Check that we don't move from the first thing in a
7563 ;; macro to its header.
7569 move tentative-move))
7574 ;; Handle K&R argdecls. Back up after the "statement" jumped
7575 ;; over by `c-beginning-of-statement-1', unless it was the
7576 ;; function body, in which case we're sitting on the opening
7577 ;; brace now. Then test if we're in a K&R argdecl region and
7578 ;; that we started at the other side of the first argdecl in
7579 ;; it.
7589 'previous
7591 knr-argdecl-start))
7594 ;; `c-beginning-of-statement-1' counts each brace block as a separate
7595 ;; statement, so the result will be 'previous if we've moved over any.
7596 ;; So change our result back to 'same if necessary.
7597 ;;
7598 ;; If they were brace list initializers we might not have moved over a
7599 ;; declaration boundary though, so change it to 'same if we've moved
7600 ;; past a '=' before '{', but not ';'. (This ought to be integrated
7601 ;; into `c-beginning-of-statement-1', so we avoid this extra pass which
7602 ;; potentially can search over a large amount of text.). Take special
7603 ;; pains not to get mislead by C++'s "operator=", and the like.
7606 c++-template-syntax-table
7612 ; no more, or get to one which isn't an "operator ="
7615 c-overloadable-operators-regexp
7616 c-opt-op-identifier-prefix
7631 ;; Assuming point is at the start of a declaration (as detected by
7632 ;; e.g. `c-beginning-of-decl-1'), go to the end of it. Unlike
7633 ;; `c-beginning-of-decl-1', this function handles the case when a
7634 ;; block is followed by identifiers in e.g. struct declarations in C
7635 ;; or C++. If a proper end was found then t is returned, otherwise
7636 ;; point is moved as far as possible within the current sexp and nil
7637 ;; is returned. This function doesn't handle macros; use
7638 ;; `c-end-of-macro' instead in those cases.
7639 ;;
7640 ;; This function might do hidden buffer changes.
7643 c++-template-syntax-table
7651 ;; Stopped at the ';' in a K&R argdecl section which is
7652 ;; detected using the same criteria as in
7653 ;; `c-beginning-of-decl-1'. Move to the following block
7654 ;; start.
7658 ;; Encountered a block in the declaration. Jump over it.
7669 ;; Check for `c-opt-block-decls-with-vars-key'
7670 ;; before the first paren.
7673 c-opt-block-decls-with-vars-key
7675 lim t t t)
7678 ;; Check that the first following paren is
7679 ;; the block.
7681 lim t t t)
7683 ;; The declaration doesn't have any of the
7684 ;; `c-opt-block-decls-with-vars' keywords in the
7685 ;; beginning, so it ends here at the end of the block.
7693 nil)))
7696 ;; Assuming the point is at an open brace, check if it starts a
7697 ;; block that contains another declaration level, i.e. that isn't a
7698 ;; statement block or a brace list, and if so return non-nil.
7699 ;;
7700 ;; If the check is successful, the return value is the start of the
7701 ;; keyword that tells what kind of construct it is, i.e. typically
7702 ;; what `c-decl-block-key' matched. Also, if GOTO-START is set then
7703 ;; the point will be at the start of the construct, before any
7704 ;; leading specifiers, otherwise it's at the returned position.
7705 ;;
7706 ;; The point is clobbered if the check is unsuccessful.
7707 ;;
7708 ;; CONTAINING-SEXP is the position of the open of the surrounding
7709 ;; paren, or nil if none.
7710 ;;
7711 ;; The optional LIMIT limits the backward search for the start of
7712 ;; the construct. It's assumed to be at a syntactically relevant
7713 ;; position.
7714 ;;
7715 ;; If any template arglists are found in the searched region before
7716 ;; the open brace, they get marked with paren syntax.
7717 ;;
7718 ;; This function might do hidden buffer changes.
7725 ;; Could be at the end of a template arglist.
7736 ;; Note: Can't get bogus hits inside template arglists below since they
7737 ;; have gotten paren syntax above.
7739 ;; If `goto-start' is set we begin by searching for the
7740 ;; first possible position of a leading specifier list.
7741 ;; The `c-decl-block-key' search continues from there since
7742 ;; we know it can't match earlier.
7745 open-brace t t)
7747 t)
7748 t)
7755 ;; Found a keyword that can't be a type?
7758 ;; Can be a type too, in which case it's the return type of a
7759 ;; function (under the assumption that no declaration level
7760 ;; block construct starts with a type).
7763 ;; Jumped over a type, but it could be a declaration keyword
7764 ;; followed by the declared identifier that we've jumped over
7765 ;; instead (e.g. in "class Foo {"). If it indeed is a type
7766 ;; then we should be at the declarator now, so check for a
7767 ;; valid declarator start.
7768 ;;
7769 ;; Note: This doesn't cope with the case when a declared
7770 ;; identifier is followed by e.g. '(' in a language where '('
7771 ;; also might be part of a declarator expression. Currently
7772 ;; there's no such language.
7776 ;; In Pike a list of modifiers may be followed by a brace
7777 ;; to make them apply to many identifiers. Note that the
7778 ;; match data will be empty on return in this case.
7784 ;; Use this variant to avoid yet another special regexp.
7788 t)))
7790 ;; Got a match.
7793 ;; Back up over any preceding specifiers and their clauses
7794 ;; by going forward from `first-specifier-pos', which is the
7795 ;; earliest possible position where the specifier list can
7796 ;; start.
7802 ;; Accept any plain symbol token on the ground that
7803 ;; it's a specifier masked through a macro (just
7804 ;; like `c-forward-decl-or-cast-1' skip forward over
7805 ;; such tokens).
7806 ;;
7807 ;; Could be more restrictive wrt invalid keywords,
7808 ;; but that'd only occur in invalid code so there's
7809 ;; no use spending effort on it.
7815 ;; Can't parse a declaration preamble and is still
7816 ;; before `kwd-start'. That means `first-specifier-pos'
7817 ;; was in some earlier construct. Search again.
7821 ;; Got no preamble before the block declaration keyword.
7827 kwd-start)))
7830 ;; Check if the closest containing paren sexp is a declaration
7831 ;; block, returning a 2 element vector in that case. Aref 0
7832 ;; contains the bufpos at boi of the class key line, and aref 1
7833 ;; contains the bufpos of the open brace. This function is an
7834 ;; obsolete wrapper for `c-looking-at-decl-block'.
7835 ;;
7836 ;; This function might do hidden buffer changes.
7844 nil))
7849 ;; return the buffer position of the beginning of the brace list
7850 ;; statement if we're inside a brace list, otherwise return nil.
7851 ;; CONTAINING-SEXP is the buffer pos of the innermost containing
7852 ;; paren. PAREN-STATE is the remainder of the state of enclosing
7853 ;; braces
7854 ;;
7855 ;; N.B.: This algorithm can potentially get confused by cpp macros
7856 ;; placed in inconvenient locations. It's a trade-off we make for
7857 ;; speed.
7858 ;;
7859 ;; This function might do hidden buffer changes.
7861 ;; This will pick up brace list declarations.
7874 ;; this will pick up array/aggregate init lists, even if they are nested.
7877 ;; Pike can have class definitions anywhere, so we must
7878 ;; check for the class key here.
7880 c-decl-block-key))
7881 bufpos braceassignp lim next-containing)
7883 containing-sexp)
7894 ;; We're in an in-expression block of some kind. Do not
7895 ;; check nesting. We deliberately set the limit to the
7896 ;; containing sexp, so that c-looking-at-inexpr-block
7897 ;; doesn't check for an identifier before it.
7899 ;; see if the open brace is preceded by = or [...] in
7900 ;; this statement, but watch out for operator=
7903 ;; Checks to do only on the first sexp before the brace.
7906 ;; In Java, an initialization brace list may follow
7907 ;; directly after "new Foo[]", so check for a "new"
7908 ;; earlier.
7914 ;; Carry on looking if this is an
7915 ;; identifier (may contain "." in Java)
7916 ;; or another "[]" sexp.
7919 ;; Checks to do on all sexps before the brace, up to the
7920 ;; beginning of the statement.
7928 ;; We've seen a =, but must check earlier tokens so
7929 ;; that it isn't something that should be ignored.
7935 ;; Check for operator =
7938 nil)
7939 ;; Check for `<opchar>= in Pike.
7942 ;; Special case for Pikes
7943 ;; `[]=, since '[' is not in
7944 ;; the punctuation class.
7947 nil)
7949 ;; make sure we're not in a C++ template
7950 ;; argument assignment
7962 ))))
7963 nil)
7970 ;; Brace lists can't contain a semicolon, so we're done.
7972 ;; Go up one level.
7974 lim nil
7975 next-containing nil))
7976 ;; we've hit the beginning of the aggregate list
7980 )
7981 bufpos))
7982 ))
7985 ;; If we're looking at the start of a pike-style list, ie `({Â })',
7986 ;; `([Â ])', `(<Â >)' etc, a cons of a cons of its starting and ending
7987 ;; positions and its entry in c-special-brace-lists is returned, nil
7988 ;; otherwise. The ending position is nil if the list is still open.
7989 ;; LIM is the limit for forward search. The point may either be at
7990 ;; the `(' or at the following paren character. Tries to check the
7991 ;; matching closer, but assumes it's correct if no balanced paren is
7992 ;; found (i.e. the case `({ ... } ... )' is detected as _not_ being
7993 ;; a special brace list).
7994 ;;
7995 ;; This function might do hidden buffer changes.
8000 inner-beg end type)
8015 nil)))))
8025 ;; Check balancing of the inner paren
8026 ;; below.
8029 t)
8030 ;; If the inner char isn't a paren then
8031 ;; we can't check balancing, so just
8032 ;; check the char before the outer
8033 ;; closing paren.
8048 ;; Return non-nil if between two statements or declarations, assuming
8049 ;; point is not inside a literal or comment.
8050 ;;
8051 ;; Obsolete - `c-at-statement-start-p' or `c-at-expression-start-p'
8052 ;; are recommended instead.
8053 ;;
8054 ;; This function might do hidden buffer changes.
8059 ;; Return non-nil if we're looking at the beginning of a block
8060 ;; inside an expression. The value returned is actually a cons of
8061 ;; either 'inlambda, 'inexpr-statement or 'inexpr-class and the
8062 ;; position of the beginning of the construct.
8063 ;;
8064 ;; LIM limits the backward search. CONTAINING-SEXP is the start
8065 ;; position of the closest containing list. If it's nil, the
8066 ;; containing paren isn't used to decide whether we're inside an
8067 ;; expression or not. If both LIM and CONTAINING-SEXP are used, LIM
8068 ;; needs to be farther back.
8069 ;;
8070 ;; If CHECK-AT-END is non-nil then extra checks at the end of the
8071 ;; brace block might be done. It should only be used when the
8072 ;; construct can be assumed to be complete, i.e. when the original
8073 ;; starting position was further down than that.
8074 ;;
8075 ;; This function might do hidden buffer changes.
8080 ;; Look at the character after point only as a last resort
8081 ;; when we can't disambiguate.
8101 ;; If the class definition is at the start of
8102 ;; a statement, we don't consider it an
8103 ;; in-expression class.
8111 ;; Also, in Pike we treat it as an
8112 ;; in-expression class if it's used in an
8113 ;; object clone expression.
8129 nil)
8137 ;; Accept several square bracket sexps for
8138 ;; Java array initializations.
8146 block-follows
8147 containing-sexp
8156 nil
8159 res))))
8162 ;; Returns non-nil if we're looking at the end of an in-expression
8163 ;; block, otherwise the same as `c-looking-at-inexpr-block'.
8164 ;; PAREN-STATE is the paren state relevant at the current position.
8165 ;;
8166 ;; This function might do hidden buffer changes.
8168 ;; We currently only recognize a block.
8171 containing-sexp)
8180 paren-state)
8181 containing-sexp)))))
8183 \f
8184 ;; `c-guess-basic-syntax' and the functions that precedes it below
8185 ;; implements the main decision tree for determining the syntactic
8186 ;; analysis of the current line of code.
8188 ;; Dynamically bound to t when `c-guess-basic-syntax' is called during
8189 ;; auto newline analysis.
8193 ;; BRACEPOS is the position of a brace in a construct like "namespace
8194 ;; Bar {". Return the anchor point in this construct; this is the
8195 ;; earliest symbol on the brace's line which isn't earlier than
8196 ;; "namespace".
8197 ;;
8198 ;; Currently (2007-08-17), "like namespace" means "matches
8199 ;; c-other-block-decl-kwds". It doesn't work with "class" or "struct"
8200 ;; or anything like that.
8210 ;; A simple function to prepend a new syntax element to
8211 ;; `c-syntactic-context'. Using `setq' on it is unsafe since it
8212 ;; should always be dynamically bound but since we read it first
8213 ;; we'll fail properly anyway if this function is misused.
8215 c-syntactic-context)))
8218 ;; Like `c-add-syntax' but appends to the end of the syntax list.
8219 ;; (Normally not necessary.)
8224 syntax-extra-args
8225 stop-at-boi-only
8226 containing-sexp
8227 paren-state)
8228 ;; Add the indicated SYNTAX-SYMBOL to `c-syntactic-context', extending it as
8229 ;; needed with further syntax elements of the types `substatement',
8230 ;; `inexpr-statement', `arglist-cont-nonempty', `statement-block-intro', and
8231 ;; `defun-block-intro'.
8232 ;;
8233 ;; Do the generic processing to anchor the given syntax symbol on
8234 ;; the preceding statement: Skip over any labels and containing
8235 ;; statements on the same line, and then search backward until we
8236 ;; find a statement or block start that begins at boi without a
8237 ;; label or comment.
8238 ;;
8239 ;; Point is assumed to be at the prospective anchor point for the
8240 ;; given SYNTAX-SYMBOL. More syntax entries are added if we need to
8241 ;; skip past open parens and containing statements. Most of the added
8242 ;; syntax elements will get the same anchor point - the exception is
8243 ;; for an anchor in a construct like "namespace"[*] - this is as early
8244 ;; as possible in the construct but on the same line as the {.
8245 ;;
8246 ;; [*] i.e. with a keyword matching c-other-block-decl-kwds.
8247 ;;
8248 ;; SYNTAX-EXTRA-ARGS are a list of the extra arguments for the
8249 ;; syntax symbol. They are appended after the anchor point.
8250 ;;
8251 ;; If STOP-AT-BOI-ONLY is nil, we can stop in the middle of the line
8252 ;; if the current statement starts there.
8253 ;;
8254 ;; Note: It's not a problem if PAREN-STATE "overshoots"
8255 ;; CONTAINING-SEXP, i.e. contains info about parens further down.
8256 ;;
8257 ;; This function might do hidden buffer changes.
8260 ;; This is by far the most common case, so let's give it special
8261 ;; treatment.
8266 ;; Set when we're on a label, so that we don't stop there.
8267 ;; FIXME: To be complete we should check if we're on a label
8268 ;; now at the start.
8269 on-label)
8271 ;; Use point as the anchor point for "namespace", "extern", etc.
8275 syntax-extra-args)
8277 ;; Loop while we have to back out of containing blocks.
8282 ;; Loop while we have to back up statements.
8284 on-label
8287 ;; Skip past any comments that stands between the
8288 ;; statement start and boi.
8296 ;; Skip to the beginning of this statement or backward
8297 ;; another one.
8305 ;; If we didn't move we're at the start of a block and
8306 ;; have to continue outside it.
8315 ;; Special case to avoid deeper and deeper indentation
8316 ;; of "else if" clauses.
8317 )
8322 ;; If stop-at-boi-only is nil, we shouldn't back up
8323 ;; over previous or containing statements to try to
8324 ;; reach boi, so go back to the last position and
8325 ;; exit.
8332 ;; If we've moved into another statement then we
8333 ;; should no longer try to stop in the middle of a
8334 ;; line.
8337 ;; Record this as a substatement if we skipped up one
8338 ;; level.
8341 )))
8343 containing-sexp)
8345 ;; Now we have to go out of this block.
8348 ;; Don't stop in the middle of a special brace list opener
8349 ;; like "({".
8361 ;; Analyze the construct in front of the block we've stepped out
8362 ;; from and add the right syntactic element for it.
8365 step-type)
8368 ;; Stepped out of a parenthesis block, so we're in an
8369 ;; expression now.
8381 ;; Stepped out of an in-expression statement. This
8382 ;; syntactic element won't get an anchor pos.
8385 ;; A parenthesis normally belongs to an arglist.
8393 on-label nil))
8395 ;; Stepped out of a brace block.
8407 containing-sexp)))
8409 'defun-block-intro
8411 nil))
8415 c-other-decl-block-key-in-symbols-alist))
8422 ;; Always done if the open brace was at boi. The
8423 ;; c-beginning-of-statement-1 call above is necessary
8424 ;; anyway, to decide the type of block-intro to add.
8427 ))
8429 ;; Fill in the current point as the anchor for all the symbols
8430 ;; added above.
8439 )))
8442 containing-decl-open
8443 containing-decl-start
8444 containing-decl-kwd
8445 paren-state)
8446 ;; The inclass and class-close syntactic symbols are added in
8447 ;; several places and some work is needed to fix everything.
8448 ;; Therefore it's collected here.
8449 ;;
8450 ;; This function might do hidden buffer changes.
8455 containing-decl-open)
8457 ;; Ought to use `c-add-stmt-syntax' instead of backing up to boi
8458 ;; here, but we have to do like this for compatibility.
8468 char-after-ip
8469 beg-of-same-or-containing-stmt
8470 containing-sexp
8471 paren-state)
8472 ;; This function contains the decision tree reached through both
8473 ;; cases 18 and 10. It's a continued statement or top level
8474 ;; construct of some kind.
8475 ;;
8476 ;; This function might do hidden buffer changes.
8483 ;; (CASE A removed.)
8484 ;; CASE B: open braces for class or brace-lists
8491 ;; CASE B.1: class-open
8498 ;; CASE B.2: brace-list-open
8503 indent-point t t t)))
8504 ;; The most semantically accurate symbol here is
8505 ;; brace-list-open, but we normally report it simply as a
8506 ;; statement-cont. The reason is that one normally adjusts
8507 ;; brace-list-open for brace lists as top-level constructs,
8508 ;; and brace lists inside statements is a completely different
8509 ;; context. C.f. case 5A.3.
8512 ;; Turn off the dwim above when we're
8513 ;; analyzing the nature of the brace
8514 ;; for the auto newline feature.
8515 'brace-list-open
8517 nil nil
8518 containing-sexp paren-state))
8520 ;; CASE B.3: The body of a function declared inside a normal
8521 ;; block. Can occur e.g. in Pike and when using gcc
8522 ;; extensions, but watch out for macros followed by blocks.
8523 ;; C.f. cases E, 16F and 17G.
8529 ;; Turn off recognition of constructs that lacks a
8530 ;; type in this case, since that's more likely to be
8531 ;; a macro followed by a block.
8534 containing-sexp paren-state))
8536 ;; CASE B.4: Continued statement with block open. The most
8537 ;; accurate analysis is perhaps `statement-cont' together with
8538 ;; `block-open' but we play DWIM and use `substatement-open'
8539 ;; instead. The rationaly is that this typically is a macro
8540 ;; followed by a block which makes it very similar to a
8541 ;; statement with a substatement block.
8544 containing-sexp paren-state))
8545 ))
8547 ;; CASE C: iostream insertion or extraction operator
8551 ;; If there is no preceding streamop in the statement
8552 ;; then indent this line as a normal statement-cont.
8556 t))))
8558 ;; CASE E: In the "K&R region" of a function declared inside a
8559 ;; normal block. C.f. case B.3.
8561 ;; Check that the next token is a '{'. This works as
8562 ;; long as no language that allows nested function
8563 ;; definitions allows stuff like member init lists, K&R
8564 ;; declarations or throws clauses there.
8565 ;;
8566 ;; Note that we do a forward search for something ahead
8567 ;; of the indentation line here. That's not good since
8568 ;; the user might not have typed it yet. Unfortunately
8569 ;; it's exceedingly tricky to recognize a function
8570 ;; prototype in a code block without resorting to this.
8578 ;; Turn off recognition of constructs that lacks a
8579 ;; type in this case, since that's more likely to be
8580 ;; a macro followed by a block.
8583 containing-sexp paren-state))
8585 ;; CASE D: continued statement.
8589 containing-sexp paren-state))
8590 )))
8592 ;; The next autoload was added by RMS on 2005/8/9 - don't know why (ACM,
8593 ;; 2005/11/29).
8594 ;;;###autoload
8596 "Return the syntactic context of the current line."
8602 ;; A whole ugly bunch of various temporary variables. Have
8603 ;; to declare them here since it's not possible to declare
8604 ;; a variable with only the scope of a cond test and the
8605 ;; following result clauses, and most of this function is a
8606 ;; single gigantic cond. :P
8607 literal char-before-ip before-ws-ip char-after-ip macro-start
8608 in-macro-expr c-syntactic-context placeholder c-in-literal-cache
8609 step-type tmpsymbol keyword injava-inher special-brace-list tmp-pos
8610 containing-<
8611 ;; The following record some positions for the containing
8612 ;; declaration block if we're directly within one:
8613 ;; `containing-decl-open' is the position of the open
8614 ;; brace. `containing-decl-start' is the start of the
8615 ;; declaration. `containing-decl-kwd' is the keyword
8616 ;; symbol of the keyword that tells what kind of block it
8617 ;; is.
8618 containing-decl-open
8619 containing-decl-start
8620 containing-decl-kwd
8621 ;; The open paren of the closest surrounding sexp or nil if
8622 ;; there is none.
8623 containing-sexp
8624 ;; The position after the closest preceding brace sexp
8625 ;; (nested sexps are ignored), or the position after
8626 ;; `containing-sexp' if there is none, or (point-min) if
8627 ;; `containing-sexp' is nil.
8628 lim
8629 ;; The paren state outside `containing-sexp', or at
8630 ;; `indent-point' if `containing-sexp' is nil.
8632 ;; There's always at most one syntactic element which got
8633 ;; an anchor pos. It's stored in syntactic-relpos.
8634 syntactic-relpos
8637 ;; Check if we're directly inside an enclosing declaration
8638 ;; level block.
8647 containing-sexp)
8648 t)))
8651 containing-sexp nil)
8656 ;; Init some position variables.
8665 ;; Ignore balanced paren. The next entry
8666 ;; can't be another one.
8669 ;; If there is no surrounding open paren then
8670 ;; put the last balanced pair back on paren-state.
8672 containing-sexp nil)))
8676 ;; If we're in a parenthesis list then ',' delimits the
8677 ;; "statements" rather than being an operator (with the
8678 ;; exception of the "for" clause). This difference is
8679 ;; typically only noticeable when statements are used in macro
8680 ;; arglists.
8685 ;; cache char before and after indent point, and move point to
8686 ;; the most likely position to perform the majority of tests
8695 ;; are we in a literal?
8698 ;; now figure out syntactic qualities of the current line
8701 ;; CASE 1: in a string.
8705 ;; CASE 2: in a C or C++ style comment.
8707 ;; This is a kludge for XEmacs where we use
8708 ;; `buffer-syntactic-context', which doesn't correctly
8709 ;; recognize "\*/" to end a block comment.
8710 ;; `parse-partial-sexp' which is used by
8711 ;; `c-literal-limits' will however do that in most
8712 ;; versions, which results in that we get nil from
8713 ;; `c-literal-limits' even when `c-in-literal' claims
8714 ;; we're inside a comment.
8718 ;; CASE 3: in a cpp preprocessor macro continuation.
8728 ;; If at the beginning of the body of a #define
8729 ;; directive then analyze as cpp-define-intro
8730 ;; only. Go on with the syntactic analysis
8731 ;; otherwise. in-macro-expr is set if we're in a
8732 ;; cpp expression, i.e. before the #define body
8733 ;; or anywhere in a non-#define directive.
8740 nil)))))
8744 ;; CASE 11: an else clause?
8748 containing-sexp paren-state))
8750 ;; CASE 12: while closure of a do/while construct?
8758 containing-sexp paren-state))
8760 ;; CASE 13: A catch or finally clause? This case is simpler
8761 ;; than if-else and do-while, because a block is required
8762 ;; after every try, catch and finally.
8770 t)
8774 t)
8777 t))
8782 containing-sexp paren-state))
8784 ;; CASE 18: A substatement we can recognize by keyword.
8794 ;; Ought to cache the result from the
8795 ;; c-beginning-of-statement-1 calls here.
8805 ;; Step up to the containing statement if we
8806 ;; stayed in the same one.
8814 ;; There was no containing statement afterall.
8816 placeholder))
8818 ;; Require a parenthesis after these keywords.
8819 ;; Necessary to catch e.g. synchronized in Java,
8820 ;; which can be used both as statement and
8821 ;; modifier.
8827 ;; CASE 18A: Simple substatement.
8833 containing-sexp paren-state))
8839 containing-sexp paren-state))
8842 containing-sexp paren-state))))
8844 ;; CASE 18B: Some other substatement. This is shared
8845 ;; with case 10.
8847 char-after-ip
8848 placeholder
8849 lim
8850 paren-state)))
8852 ;; CASE 14: A case or default label
8858 containing-sexp))
8861 ;; Got a bogus label at the top level. In lack of better
8862 ;; alternatives, anchor it on (point-min).
8865 ;; CASE 15: any other label
8872 containing-decl-open
8873 containing-decl-start
8874 containing-decl-kwd
8875 paren-state))
8876 ;; Append access-label with the same anchor point as
8877 ;; inclass gets.
8883 containing-sexp))
8888 ;; If the surrounding statement is a switch then
8889 ;; let's analyze all labels as switch labels, so
8890 ;; that they get lined up consistently.
8891 'case-label
8897 ;; A label on the top level. Treat it as a class
8898 ;; context. (point-min) is the closest we get to the
8899 ;; class open brace.
8902 ;; CASE 4: In-expression statement. C.f. cases 7B, 16A and
8903 ;; 17E.
8906 containing-sexp
8907 ;; Have to turn on the heuristics after
8908 ;; the point even though it doesn't work
8909 ;; very well. C.f. test case class-16.pike.
8910 t))
8915 ;; It's a statement block or an anonymous class.
8917 ;; It's a Pike lambda. Check whether we are between the
8918 ;; lambda keyword and the argument list or at the defun
8919 ;; opener.
8921 'inline-open
8927 paren-state)
8931 ;; CASE 5: Line is inside a declaration level block or at top level.
8935 ;; CASE 5A: we are looking at a defun, brace list, class,
8936 ;; or inline-inclass method opening brace
8943 ;; CASE 5A.1: Non-class declaration block open.
8958 ;; Special case for extern-lang-open.
8959 'extern-lang-open
8961 nil t containing-sexp paren-state))
8963 ;; CASE 5A.2: we are looking at a class opening brace
8972 ;; CASE 5A.3: brace list open
8997 ))))
9001 ;; We're in Java and have found that the open brace
9002 ;; belongs to a "new Foo[]" initialization list,
9003 ;; which means the brace list is part of an
9004 ;; expression and not a top level definition. We
9005 ;; therefore treat it as any topmost continuation
9006 ;; even though the semantically correct symbol still
9007 ;; is brace-list-open, on the same grounds as in
9008 ;; case B.2.
9014 ;; CASE 5A.4: inline defun open
9020 containing-decl-open
9021 containing-decl-start
9022 containing-decl-kwd
9023 paren-state))
9025 ;; CASE 5A.5: ordinary defun open
9033 ;; Bogus to use bol here, but it's the legacy. (Resolved,
9034 ;; 2007-11-09)
9035 ))))
9037 ;; CASE 5B: After a function header but before the body (or
9038 ;; the ending semicolon if there's no body).
9044 ;; CASE 5B.1: Member init list.
9049 ;; There is no preceding member init clause.
9050 ;; Indent relative to the beginning of indentation
9051 ;; for the topmost-intro line that contains the
9052 ;; prototype's open paren.
9055 ;; Indent relative to the first member init clause.
9060 ;; CASE 5B.2: K&R arg decl intro
9067 containing-decl-open
9068 containing-decl-start
9069 containing-decl-kwd
9070 paren-state)))
9072 ;; CASE 5B.4: Nether region after a C++ or Java func
9073 ;; decl, which could include a `throws' declaration.
9077 )))
9079 ;; CASE 5C: inheritance line. could be first inheritance
9080 ;; line, or continuation of a multiple inheritance
9088 ;; watch out for scope operator
9093 )))
9102 ;; for Java
9107 cont done)
9112 done t))
9116 )
9118 injava-inher)
9121 ))
9124 ;; CASE 5C.1: non-hanging colon on an inher intro
9128 ;; don't add inclass symbol since relative point already
9129 ;; contains any class offset
9130 )
9132 ;; CASE 5C.2: hanging colon on an inher intro
9138 containing-decl-open
9139 containing-decl-start
9140 containing-decl-kwd
9141 paren-state)))
9143 ;; CASE 5C.3: in a Java implements/extends
9157 )))
9159 ;; CASE 5C.4: a continued inheritance line
9163 ;; don't add inclass symbol since relative point already
9164 ;; contains any class offset
9165 )))
9167 ;; CASE 5D: this could be a top-level initialization, a
9168 ;; member init list continuation, or a template argument
9169 ;; list continuation.
9171 ;; Note: We use the fact that lim is always after any
9172 ;; preceding brace sexp.
9182 t))
9187 t)
9193 nil)
9194 t))
9196 ;; NB: No c-after-special-operator-id stuff in this
9197 ;; clause - we assume only C++ needs it.
9202 ;; CASE 5D.3: perhaps a template list continuation?
9217 ;; In a nested template arglist.
9223 ;; FIXME: Should use c-add-stmt-syntax, but it's not yet
9224 ;; template aware.
9227 ;; CASE 5D.4: perhaps a multiple inheritance line?
9239 t)
9244 ;; CASE 5D.5: Continuation of the "expression part" of a
9245 ;; top level construct. Or, perhaps, an unrecognised construct.
9257 ;; A preceding comma at the top level means that a
9258 ;; new variable declaration starts here. Use
9259 ;; topmost-intro-cont for it, for consistency with
9260 ;; the first variable declaration. C.f. case 5N.
9263 nil nil containing-sexp paren-state))
9264 ))
9266 ;; CASE 5F: Close of a non-class declaration level block.
9270 ;; This is inconsistent: Should use `containing-decl-open'
9271 ;; here if it's at boi, like in case 5J.
9275 ;; Special case for compatibility with the
9276 ;; extern-lang syntactic symbols.
9277 'extern-lang-close
9280 nil t
9282 paren-state))
9284 ;; CASE 5G: we are looking at the brace which closes the
9285 ;; enclosing nested class decl
9290 containing-decl-open
9291 containing-decl-start
9292 containing-decl-kwd
9293 paren-state))
9295 ;; CASE 5H: we could be looking at subsequent knr-argdecls
9302 ;; Do an extra check to avoid tripping up on
9303 ;; statements that occur in invalid contexts
9304 ;; (e.g. in macro bodies where we don't really
9305 ;; know the context of what we're looking at).
9312 ;; CASE 5I: ObjC method definition.
9317 ;; Handle the case when it's the first (non-comment)
9318 ;; thing in the buffer. Can't look for a 'same return
9319 ;; value from cbos1 since ObjC directives currently
9320 ;; aren't recognized fully, so that we get 'same
9321 ;; instead of 'previous if it moved over a preceding
9322 ;; directive.
9326 ;; CASE 5P: AWK pattern or function or continuation
9327 ;; thereof.
9333 'topmost-intro-cont
9335 nil nil
9336 containing-sexp paren-state))
9338 ;; CASE 5N: At a variable declaration that follows a class
9339 ;; definition or some other block declaration that doesn't
9340 ;; end at the closing '}'. C.f. case 5D.5.
9349 ;; Speed up the backward search a bit.
9354 ;; The '}' is unbalanced.
9355 nil
9360 containing-sexp paren-state))
9362 ;; NOTE: The point is at the end of the previous token here.
9364 ;; CASE 5J: we are at the topmost level, make
9365 ;; sure we skip back past any access specifiers
9367 ;; A macro continuation line is never at top level.
9389 nil)))))
9390 ;; For historic reasons we anchor at bol of the last
9391 ;; line of the previous declaration. That's clearly
9392 ;; highly bogus and useless, and it makes our lives hard
9393 ;; to remain compatible. :P
9404 ;; Special case for compatibility with the
9405 ;; extern-lang syntactic symbols.
9406 'inextern-lang
9409 nil t
9411 paren-state))
9413 containing-decl-open
9414 containing-decl-start
9415 containing-decl-kwd
9416 paren-state)))
9418 macro-start
9423 ;; CASE 5K: we are at an ObjC method definition
9424 ;; continuation line.
9433 ;; CASE 5L: we are at the first argument of a template
9434 ;; arglist that begins on the previous line.
9443 ;; CASE 5Q: we are at a statement within a macro.
9448 ;; CASE 5M: we are at a topmost continuation line
9459 ))
9461 ;; (CASE 6 has been removed.)
9463 ;; CASE 19: line is an expression, not a statement, and is directly
9464 ;; contained by a template delimiter. Most likely, we are in a
9465 ;; template arglist within a statement. This case is based on CASE
9466 ;; 7. At some point in the future, we may wish to create more
9467 ;; syntactic symbols such as `template-intro',
9468 ;; `template-cont-nonempty', etc., and distinguish between them as we
9469 ;; do for `arglist-intro' etc. (2009-12-07).
9475 ; '<') before indent-point.
9483 paren-state))
9486 ;; CASE 7: line is an expression, not a statement. Most
9487 ;; likely we are either in a function prototype or a function
9488 ;; call argument list, or a template argument list.
9497 ;; CASE 7A: we are looking at the arglist closing paren.
9498 ;; C.f. case 7F.
9510 paren-state))
9512 ;; CASE 7B: Looking at the opening brace of an
9513 ;; in-expression block or brace list. C.f. cases 4, 16A
9514 ;; and 17E.
9518 paren-state))
9522 placeholder
9525 paren-state)
9526 containing-sexp)))
9527 ;; placeholder is nil if it's a block directly in
9528 ;; a function arglist. That makes us skip out of
9529 ;; this case.
9530 )))
9535 paren-state)
9539 ;; CASE 7C: we are looking at the first argument in an empty
9540 ;; argument list. Use arglist-close if we're actually
9541 ;; looking at a close paren or bracket.
9553 paren-state))
9555 ;; CASE 7D: we are inside a conditional test clause. treat
9556 ;; these things as statements
9566 ))
9568 ;; CASE 7E: maybe a continued ObjC method call. This is the
9569 ;; case when we are inside a [] bracketed exp, and what
9570 ;; precede the opening bracket is not an identifier.
9578 )))
9580 ;; CASE 7F: we are looking at an arglist continuation line,
9581 ;; but the preceding argument is on the same line as the
9582 ;; opening paren. This case includes multi-line
9583 ;; mathematical paren groupings, but we could be on a
9584 ;; for-list continuation line. C.f. case 7A.
9601 paren-state))
9603 ;; CASE 7G: we are looking at just a normal arglist
9604 ;; continuation line
9607 ))
9609 ;; CASE 8: func-local multi-inheritance line
9619 ;; CASE 8A: non-hanging colon on an inher intro
9624 ;; CASE 8B: hanging colon on an inher intro
9628 ;; CASE 8C: a continued inheritance line
9632 )))
9634 ;; CASE 9: we are inside a brace-list
9644 ;; CASE 9A: In the middle of a special brace list opener.
9664 ;; CASE 9B: brace-list-close brace
9666 ;; Check special brace list closer.
9673 ;; We were between the special close char and the `)'.
9676 ;; We were before the special close char.
9680 ;; Normal brace list check.
9691 ;; Prepare for the rest of the cases below by going to the
9692 ;; token following the opening brace
9705 ;; CASE 9C: we're looking at the first line in a brace-list
9716 ;; CASE 9D: this is just a later brace-list-entry or
9717 ;; brace-entry-open
9726 ))
9727 ))))
9729 ;; CASE 10: A continued statement or top level construct.
9739 ;; This is shared with case 18.
9741 char-after-ip
9742 placeholder
9743 containing-sexp
9744 paren-state))
9746 ;; CASE 16: block close brace, possibly closing the defun or
9747 ;; the class
9749 ;; From here on we have the next containing sexp in lim.
9754 ;; CASE 16E: Closing a statement block? This catches
9755 ;; cases where it's preceded by a statement keyword,
9756 ;; which works even when used in an "invalid" context,
9757 ;; e.g. a macro argument.
9762 ;; CASE 16A: closing a lambda defun or an in-expression
9763 ;; block? C.f. cases 4, 7B and 17E.
9766 nil))
9768 'inline-close
9778 paren-state)
9782 ;; CASE 16B: does this close an inline or a function in
9783 ;; a non-class declaration level block?
9790 nil))
9800 ;; CASE 16F: Can be a defun-close of a function declared
9801 ;; in a statement block, e.g. in Pike or when using gcc
9802 ;; extensions, but watch out for macros followed by
9803 ;; blocks. Let it through to be handled below.
9804 ;; C.f. cases B.3 and 17G.
9810 ;; Turn off recognition of constructs that
9811 ;; lacks a type in this case, since that's more
9812 ;; likely to be a macro followed by a block.
9819 ;; CASE 16C: If there is an enclosing brace then this is
9820 ;; a block close since defun closes inside declaration
9821 ;; level blocks have been handled above.
9823 ;; If the block is preceded by a case/switch label on
9824 ;; the same line, we anchor at the first preceding label
9825 ;; at boi. The default handling in c-add-stmt-syntax
9826 ;; really fixes it better, but we do like this to keep
9827 ;; the indentation compatible with version 5.28 and
9828 ;; earlier. C.f. case 17H.
9835 ;; c-backward-to-block-anchor not necessary here; those
9836 ;; situations are handled in case 16E above.
9839 ;; CASE 16D: Only top level defun close left.
9845 paren-state))
9846 ))
9848 ;; CASE 17: Statement or defun catchall.
9851 ;; Back up statements until we find one that starts at boi.
9854 containing-sexp)))
9858 nil)
9863 ;; CASE 17B: continued statement
9867 containing-sexp paren-state))
9869 ;; CASE 17A: After a case/default label?
9877 'statement-case-open
9879 nil t containing-sexp paren-state))
9881 ;; CASE 17D: any old statement
9888 containing-sexp paren-state)
9892 ;; CASE 17I: Inside a substatement block.
9894 ;; The following tests are all based on containing-sexp.
9896 ;; From here on we have the next containing sexp in lim.
9901 lim paren-state)
9905 ;; CASE 17E: first statement in an in-expression block.
9906 ;; C.f. cases 4, 7B and 16A.
9909 nil))
9911 'defun-block-intro
9920 paren-state)
9926 ;; CASE 17F: first statement in an inline, or first
9927 ;; statement in a top-level defun. we can tell this is it
9928 ;; if there are no enclosing braces that haven't been
9929 ;; narrowed out by a class (i.e. don't use bod here).
9932 paren-state)))
9938 nil))))
9943 ;; CASE 17G: First statement in a function declared inside
9944 ;; a normal block. This can occur in Pike and with
9945 ;; e.g. the gcc extensions, but watch out for macros
9946 ;; followed by blocks. C.f. cases B.3 and 16F.
9952 ;; Turn off recognition of constructs that lacks
9953 ;; a type in this case, since that's more likely
9954 ;; to be a macro followed by a block.
9960 lim paren-state))
9962 ;; CASE 17H: First statement in a block.
9964 ;; If the block is preceded by a case/switch label on the
9965 ;; same line, we anchor at the first preceding label at
9966 ;; boi. The default handling in c-add-stmt-syntax is
9967 ;; really fixes it better, but we do like this to keep the
9968 ;; indentation compatible with version 5.28 and earlier.
9969 ;; C.f. case 16C.
9976 ;; c-backward-to-block-anchor not necessary here; those
9977 ;; situations are handled in case 17I above.
9979 lim paren-state))
9982 ))
9983 )
9985 ;; now we need to look at any modifiers
9989 ;; are we looking at a comment only line?
9994 ;; we might want to give additional offset to friends (in C++).
9999 ;; Set syntactic-relpos.
10005 nil)
10006 t))
10009 ;; Start of or a continuation of a preprocessor directive?
10024 ;; If inside a cpp expression, i.e. anywhere in a
10025 ;; cpp directive except a #define body, we only let
10026 ;; through the syntactic analysis that is internal
10027 ;; in the expression. That means the arglist
10028 ;; elements, if they are anchored inside the cpp
10029 ;; expression.
10038 ;; Inside a #define body and the syntactic analysis is
10039 ;; anchored on the start of the #define. In this case
10040 ;; we add cpp-define-intro to get the extra
10041 ;; indentation of the #define body.
10044 ;; return the syntax
10045 c-syntactic-context)))
10047 \f
10048 ;; Indentation calculation.
10051 ;; offset can be a number, a function, a variable, a list, or one of
10052 ;; the symbols + or -
10053 ;;
10054 ;; This function might do hidden buffer changes.
10073 langelem symbol))
10079 offset symbol)
10080 nil)
10094 Error evaluating offset %S for %s: \
10095 Cannot combine absolute offset %S with relative %S in `%s' method"
10101 Error evaluating offset %S for %s: \
10102 Cannot combine relative offset %S with absolute %S in `%s' method"
10107 res))
10125 Error evaluating offset %S for %s: \
10126 Cannot combine absolute offsets %S and %S in `add' method"
10130 res))
10139 res))))
10146 nil))))
10150 res
10152 offset symbol res)
10153 nil)))
10156 ;; Get offset from LANGELEM which is a list beginning with the
10157 ;; syntactic symbol and followed by any analysis data it provides.
10158 ;; That data may be zero or more elements, but if at least one is
10159 ;; given then the first is the anchor position (or nil). The symbol
10160 ;; is matched against `c-offsets-alist' and the offset calculated
10161 ;; from that is returned.
10162 ;;
10163 ;; This function might do hidden buffer changes.
10173 offset
10177 ))
10180 ;; This is a compatibility wrapper for `c-calc-offset' in case
10181 ;; someone is calling it directly. It takes an old style syntactic
10182 ;; element on the form (SYMBOL . ANCHOR-POS) and converts it to the
10183 ;; new list form.
10184 ;;
10185 ;; This function might do hidden buffer changes.
10192 ;; Calculate the syntactic indentation from a syntactic description
10193 ;; as returned by `c-guess-syntax'.
10194 ;;
10195 ;; Note that topmost-intro always has an anchor position at bol, for
10196 ;; historical reasons. It's often used together with other symbols
10197 ;; that has more sane positions. Since we always use the first
10198 ;; found anchor position, we rely on that these other symbols always
10199 ;; precede topmost-intro in the LANGELEMS list.
10200 ;;
10201 ;; This function might do hidden buffer changes.
10209 ;; Got an absolute column that overrides any indentation
10210 ;; we've collected so far, but not the relative
10211 ;; indentation we might get for the nested structures
10212 ;; further down the langelems list.
10216 ;; Got a relative change of the current calculated
10217 ;; indentation.
10220 ;; Use the anchor position from the first syntactic
10221 ;; element with one.
10231 indent)))
10233 \f
10236 ;; arch-tag: 149add18-4673-4da5-ac47-6805e4eae089
10237 ;;; cc-engine.el ends here