| blob | a901ee074544b2009990e153ffc379b9f625c45c |
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
5 ;; Free Software Foundation, Inc.
7 ;; Authors: 2001- Alan Mackenzie
8 ;; 1998- Martin Stjernholm
9 ;; 1992-1999 Barry A. Warsaw
10 ;; 1987 Dave Detlefs and Stewart Clamen
11 ;; 1985 Richard M. Stallman
12 ;; Maintainer: bug-cc-mode@gnu.org
13 ;; Created: 22-Apr-1997 (split from cc-mode.el)
14 ;; Version: See cc-mode.el
15 ;; Keywords: c languages oop
17 ;; This file is part of GNU Emacs.
19 ;; GNU Emacs is free software; you can redistribute it and/or modify
20 ;; it under the terms of the GNU General Public License as published by
21 ;; the Free Software Foundation; either version 2, or (at your option)
22 ;; any later version.
24 ;; GNU Emacs is distributed in the hope that it will be useful,
25 ;; but WITHOUT ANY WARRANTY; without even the implied warranty of
26 ;; MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
27 ;; GNU General Public License for more details.
29 ;; You should have received a copy of the GNU General Public License
30 ;; along with this program; see the file COPYING. If not, write to
31 ;; the Free Software Foundation, Inc., 51 Franklin Street, Fifth Floor,
32 ;; Boston, MA 02110-1301, USA.
34 ;;; Commentary:
36 ;; The functions which have docstring documentation can be considered
37 ;; part of an API which other packages can use in CC Mode buffers.
38 ;; Otoh, undocumented functions and functions with the documentation
39 ;; in comments are considered purely internal and can change semantics
40 ;; or even disappear in the future.
41 ;;
42 ;; (This policy applies to CC Mode as a whole, not just this file. It
43 ;; probably also applies to many other Emacs packages, but here it's
44 ;; clearly spelled out.)
46 ;; Hidden buffer changes
47 ;;
48 ;; Various functions in CC Mode use text properties for caching and
49 ;; syntactic markup purposes, and those of them that might modify such
50 ;; properties but still don't modify the buffer in a visible way are
51 ;; said to do "hidden buffer changes". They should be used within
52 ;; `c-save-buffer-state' or a similar function that saves and restores
53 ;; buffer modifiedness, disables buffer change hooks, etc.
54 ;;
55 ;; Interactive functions are assumed to not do hidden buffer changes,
56 ;; except in the specific parts of them that do real changes.
57 ;;
58 ;; Lineup functions are assumed to do hidden buffer changes. They
59 ;; must not do real changes, though.
60 ;;
61 ;; All other functions that do hidden buffer changes have that noted
62 ;; in their doc string or comment.
63 ;;
64 ;; The intention with this system is to avoid wrapping every leaf
65 ;; function that do hidden buffer changes inside
66 ;; `c-save-buffer-state'. It should be used as near the top of the
67 ;; interactive functions as possible.
68 ;;
69 ;; Functions called during font locking are allowed to do hidden
70 ;; buffer changes since the font-lock package run them in a context
71 ;; similar to `c-save-buffer-state' (in fact, that function is heavily
72 ;; inspired by `save-buffer-state' in the font-lock package).
74 ;; Use of text properties
75 ;;
76 ;; CC Mode uses several text properties internally to mark up various
77 ;; positions, e.g. to improve speed and to eliminate glitches in
78 ;; interactive refontification.
79 ;;
80 ;; Note: This doc is for internal use only. Other packages should not
81 ;; assume that these text properties are used as described here.
82 ;;
83 ;; 'syntax-table
84 ;; Used to modify the syntax of some characters. Currently used to
85 ;; mark the "<" and ">" of angle bracket parens with paren syntax.
86 ;;
87 ;; This property is used on single characters and is therefore
88 ;; always treated as front and rear nonsticky (or start and end open
89 ;; in XEmacs vocabulary). It's therefore installed on
90 ;; `text-property-default-nonsticky' if that variable exists (Emacs
91 ;; >= 21).
92 ;;
93 ;; 'c-is-sws and 'c-in-sws
94 ;; Used by `c-forward-syntactic-ws' and `c-backward-syntactic-ws' to
95 ;; speed them up. See the comment blurb before `c-put-is-sws'
96 ;; below for further details.
97 ;;
98 ;; 'c-type
99 ;; This property is used on single characters to mark positions with
100 ;; special syntactic relevance of various sorts. Its primary use is
101 ;; to avoid glitches when multiline constructs are refontified
102 ;; interactively (on font lock decoration level 3). It's cleared in
103 ;; a region before it's fontified and is then put on relevant chars
104 ;; in that region as they are encountered during the fontification.
105 ;; The value specifies the kind of position:
106 ;;
107 ;; 'c-decl-arg-start
108 ;; Put on the last char of the token preceding each declaration
109 ;; inside a declaration style arglist (typically in a function
110 ;; prototype).
111 ;;
112 ;; 'c-decl-end
113 ;; Put on the last char of the token preceding a declaration.
114 ;; This is used in cases where declaration boundaries can't be
115 ;; recognized simply by looking for a token like ";" or "}".
116 ;; `c-type-decl-end-used' must be set if this is used (see also
117 ;; `c-find-decl-spots').
118 ;;
119 ;; 'c-<>-arg-sep
120 ;; Put on the commas that separate arguments in angle bracket
121 ;; arglists like C++ template arglists.
122 ;;
123 ;; 'c-decl-id-start and 'c-decl-type-start
124 ;; Put on the last char of the token preceding each declarator
125 ;; in the declarator list of a declaration. They are also used
126 ;; between the identifiers cases like enum declarations.
127 ;; 'c-decl-type-start is used when the declarators are types,
128 ;; 'c-decl-id-start otherwise.
129 ;;
130 ;; 'c-awk-NL-prop
131 ;; Used in AWK mode to mark the various kinds of newlines. See
132 ;; cc-awk.el.
134 ;;; Code:
141 load-path)))
148 ;; Silence the compiler.
151 \f
152 ;; Make declarations for all the `c-lang-defvar' variables in cc-langs.
165 \f
166 ;;; Internal state variables.
168 ;; Internal state of hungry delete key feature
172 ;; The electric flag (toggled by `c-toggle-electric-state').
173 ;; If t, electric actions (like automatic reindentation, and (if
174 ;; c-auto-newline is also set) auto newlining) will happen when an electric
175 ;; key like `{' is pressed (or an electric keyword like `else').
179 ;; Internal state of auto newline feature.
183 ;; Included in the mode line to indicate the active submodes.
184 ;; (defvar c-submode-indicators nil)
185 ;; (make-variable-buffer-local 'c-submode-indicators)
188 ;; Calculate the new state of PREVSTATE, t or nil, based on arg. If
189 ;; arg is nil or zero, toggle the state. If arg is negative, turn
190 ;; the state off, and if arg is positive, turn the state on
196 ;; Dynamically bound cache for `c-in-literal'.
199 \f
200 ;; Basic handling of preprocessor directives.
202 ;; This is a dynamically bound cache used together with
203 ;; `c-query-macro-start' and `c-query-and-set-macro-start'. It only
204 ;; works as long as point doesn't cross a macro boundary.
213 c-macro-start))
221 c-macro-start))
224 "Go to the beginning of a preprocessor directive.
225 Leave point at the beginning of the directive and return t if in one,
226 otherwise return nil and leave point unchanged.
228 Note that this function might do hidden buffer changes. See the
229 comment at the start of cc-engine.el for more info."
240 t
242 nil)))))
245 "Go to the end of a preprocessor directive.
246 More accurately, move the point to the end of the closest following
247 line that doesn't end with a line continuation backslash - no check is
248 done that the point is inside a cpp directive to begin with.
250 Note that this function might do hidden buffer changes. See the
251 comment at the start of cc-engine.el for more info."
257 t))))
260 ;; Assuming point is at the "#" that introduces a preprocessor
261 ;; directive, it's moved forward to the end of the identifier which is
262 ;; "#define"d (or whatever c-opt-cpp-macro-define specifies). Non-nil
263 ;; is returned in this case, in all other cases nil is returned and
264 ;; point isn't moved.
265 ;;
266 ;; This function might do hidden buffer changes.
272 ;; Assuming point is at the "#" that introduces a preprocessor
273 ;; directive, it's moved forward to the start of the definition body
274 ;; if it's a "#define" (or whatever c-opt-cpp-macro-define
275 ;; specifies). Non-nil is returned in this case, in all other cases
276 ;; nil is returned and point isn't moved.
277 ;;
278 ;; This function might do hidden buffer changes.
284 \f
285 ;;; Basic utility functions.
288 ;; Return the given region as a string where all syntactic
289 ;; whitespace is removed or, where necessary, replaced with a single
290 ;; space. If PAREN-LEVEL is given then all parens in the region are
291 ;; collapsed to "()", "[]" etc.
292 ;;
293 ;; This function might do hidden buffer changes.
337 ;; Shift the indentation of the current line with the specified
338 ;; amount (positive inwards). The buffer is modified only if
339 ;; SHIFT-AMT isn't equal to zero.
342 tmp-char-inserted)
344 nil
345 ;; If we're on an empty line inside a macro, we take the point
346 ;; to be at the current indentation and shift it to the
347 ;; appropriate column. This way we don't treat the extra
348 ;; whitespace out to the line continuation as indentation.
364 ;; If initial point was within line's indentation and we're not on
365 ;; a line with a line continuation in a macro, position after the
366 ;; indentation. Else stay at same point in text.
374 ;; Return non-nil if the string KEYWORD is a known keyword. More
375 ;; precisely, the value is the symbol for the keyword in
376 ;; `c-keywords-obarray'.
380 ;; Return non-nil if the symbol KEYWORD-SYM, as returned by
381 ;; `c-keyword-sym', is a member of LANG-CONSTANT, which is the name
382 ;; of a language constant that ends with "-kwds". If KEYWORD-SYM is
383 ;; nil then the result is nil.
386 ;; String syntax chars, suitable for skip-syntax-(forward|backward).
391 ;; Regexp matching string limit syntax.
396 ;; Regexp matching WS followed by string limit syntax.
400 ;; Holds formatted error strings for the few cases where parse errors
401 ;; are reported.
408 c-parsing-error)
410 ;; Faces given to comments and string literals. This is used in some
411 ;; situations to speed up recognition; it isn't mandatory that font
412 ;; locking is in use. This variable is extended with the face in
413 ;; `c-doc-face-name' when fontification is activated in cc-fonts.el.
417 ;; New in Emacs 22.
421 ;; Put a c-type property with the given value at POS.
425 ;; Remove all occurences of the c-type property that has the given
426 ;; value in the region between FROM and TO. VALUE is assumed to not
427 ;; be nil.
428 ;;
429 ;; Note: This assumes that c-type is put on single chars only; it's
430 ;; very inefficient if matching properties cover large regions.
439 \f
440 ;; Some debug tools to visualize various special positions. This
441 ;; debug code isn't as portable as the rest of CC Mode.
477 \f
478 ;; `c-beginning-of-statement-1' and accompanying stuff.
480 ;; KLUDGE ALERT: c-maybe-labelp is used to pass information between
481 ;; c-crosses-statement-barrier-p and c-beginning-of-statement-1. A
482 ;; better way should be implemented, but this will at least shut up
483 ;; the byte compiler.
486 ;; New awk-compatible version of c-beginning-of-statement-1, ACM 2002/6/22
488 ;; Macros used internally in c-beginning-of-statement-1 for the
489 ;; automaton actions.
492 stack)))
497 t
498 ,do-if-done
503 ;; Throw nil if stack is empty, else throw non-nil.
527 noerror comma-delim)
528 "Move to the start of the current statement or declaration, or to
529 the previous one if already at the beginning of one. Only
530 statements/declarations on the same level are considered, i.e. don't
531 move into or out of sexps (not even normal expression parentheses).
535 is within the continuation. If starting at such a token, move to the
536 corresponding statement start. If at the beginning of a statement,
537 move to the closest containing statement if there is any. This might
538 also stop at a continuation clause.
540 Labels are treated as part of the following statements if
541 IGNORE-LABELS is non-nil. (FIXME: Doesn't work if we stop at a known
542 statement start keyword.) Otherwise, each label is treated as a
543 separate statement.
545 Macros are ignored \(i.e. skipped over) unless point is within one, in
546 which case the content of the macro is treated as normal code. Aside
547 from any normal statement starts found in it, stop at the first token
550 macros before leaving them.
552 Return 'label if stopped at a label, 'same if stopped at the beginning
553 of the current statement, 'up if stepped to a containing statement,
554 'previous if stepped to a preceding statement, 'beginning if stepped
555 from a statement continuation clause to its start clause, or 'macro if
556 stepped to a macro start. Note that 'same and not 'label is returned
557 if stopped at the same label without crossing the colon character.
559 LIM may be given to limit the search. If the search hits the limit,
560 point will be left at the closest following token, or at the start
561 position if that is less ('same is returned in this case).
563 NOERROR turns off error logging to `c-parsing-error'.
565 Normally only ';' and virtual semicolons are considered to delimit
566 statements, but if COMMA-DELIM is non-nil then ',' is treated
567 as a delimiter too.
569 Note that this function might do hidden buffer changes. See the
570 comment at the start of cc-engine.el for more info."
572 ;; The bulk of this function is a pushdown automaton that looks at statement
573 ;; boundaries and the tokens (such as "while") in c-opt-block-stmt-key. Its
574 ;; purpose is to keep track of nested statements, ensuring that such
575 ;; statments are skipped over in their entirety (somewhat akin to what C-M-p
576 ;; does with nested braces/brackets/parentheses).
577 ;;
578 ;; Note: The position of a boundary is the following token.
579 ;;
580 ;; Beginning with the current token (the one following point), move back one
581 ;; sexp at a time (where a sexp is, more or less, either a token or the
582 ;; entire contents of a brace/bracket/paren pair). Each time a statement
583 ;; boundary is crossed or a "while"-like token is found, update the state of
584 ;; the PDA. Stop at the beginning of a statement when the stack (holding
585 ;; nested statement info) is empty and the position has been moved.
586 ;;
587 ;; The following variables constitute the PDA:
588 ;;
589 ;; sym: This is either the "while"-like token (e.g. 'for) we've just
590 ;; scanned back over, 'boundary if we've just gone back over a
591 ;; statement boundary, or nil otherwise.
592 ;; state: takes one of the values (nil else else-boundary while
593 ;; while-boundary catch catch-boundary).
594 ;; nil means "no "while"-like token yet scanned".
595 ;; 'else, for example, means "just gone back over an else".
596 ;; 'else-boundary means "just gone back over a statement boundary
597 ;; immediately after having gone back over an else".
598 ;; saved-pos: A vector of either saved positions (tok ptok pptok, etc.) or
599 ;; of error reporting information.
600 ;; stack: The stack onto which the PDA pushes its state. Each entry
601 ;; consists of a saved value of state and saved-pos. An entry is
602 ;; pushed when we move back over a "continuation" token (e.g. else)
603 ;; and popped when we encounter the corresponding opening token
604 ;; (e.g. if).
605 ;;
606 ;;
607 ;; The following diagram briefly outlines the PDA.
608 ;;
609 ;; Common state:
610 ;; "else": Push state, goto state `else'.
611 ;; "while": Push state, goto state `while'.
612 ;; "catch" or "finally": Push state, goto state `catch'.
613 ;; boundary: Pop state.
614 ;; other: Do nothing special.
615 ;;
616 ;; State `else':
617 ;; boundary: Goto state `else-boundary'.
618 ;; other: Error, pop state, retry token.
619 ;;
620 ;; State `else-boundary':
621 ;; "if": Pop state.
622 ;; boundary: Error, pop state.
623 ;; other: See common state.
624 ;;
625 ;; State `while':
626 ;; boundary: Save position, goto state `while-boundary'.
627 ;; other: Pop state, retry token.
628 ;;
629 ;; State `while-boundary':
630 ;; "do": Pop state.
631 ;; boundary: Restore position if it's not at start, pop state. [*see below]
632 ;; other: See common state.
633 ;;
634 ;; State `catch':
635 ;; boundary: Goto state `catch-boundary'.
636 ;; other: Error, pop state, retry token.
637 ;;
638 ;; State `catch-boundary':
639 ;; "try": Pop state.
640 ;; "catch": Goto state `catch'.
641 ;; boundary: Error, pop state.
642 ;; other: See common state.
643 ;;
644 ;; [*] In the `while-boundary' state, we had pushed a 'while state, and were
645 ;; searching for a "do" which would have opened a do-while. If we didn't
646 ;; find it, we discard the analysis done since the "while", go back to this
647 ;; token in the buffer and restart the scanning there, this time WITHOUT
648 ;; pushing the 'while state onto the stack.
649 ;;
650 ;; In addition to the above there is some special handling of labels
651 ;; and macros.
655 macro-start
658 c-stmt-delim-chars-with-comma
659 c-stmt-delim-chars))
660 c-in-literal-cache c-maybe-labelp saved
661 ;; Current position.
662 pos
663 ;; Position of last stmt boundary character (e.g. ;).
664 boundary-pos
665 ;; The position of the last sexp or bound that follows the
666 ;; first found colon, i.e. the start of the nonlabel part of
667 ;; the statement. It's `start' if a colon is found just after
668 ;; the start.
669 after-labels-pos
670 ;; Like `after-labels-pos', but the first such position inside
671 ;; a label, i.e. the start of the last label before the start
672 ;; of the nonlabel part of the statement.
673 last-label-pos
674 ;; The last position where a label is possible provided the
675 ;; statement started there. It's nil as long as no invalid
676 ;; label content has been found (according to
677 ;; `c-nonlabel-token-key'. It's `start' if no valid label
678 ;; content was found in the label. Note that we might still
679 ;; regard it a label if it starts with `c-label-kwds'.
680 label-good-pos
681 ;; Symbol just scanned back over (e.g. 'while or 'boundary).
682 ;; See above.
683 sym
684 ;; Current state in the automaton. See above.
685 state
686 ;; Current saved positions. See above.
687 saved-pos
688 ;; Stack of conses (state . saved-pos).
689 stack
690 ;; Regexp which matches "for", "if", etc.
693 ;; Return value.
695 ;; Positions of the last three sexps or bounds we've stopped at.
696 tok ptok pptok)
706 ;; Try to skip back over unary operator characters, to register
707 ;; that we've moved.
711 ;; Protect post-++/-- operators just before a virtual semicolon.
715 ;; Skip back over any semicolon here. If it was a bare semicolon, we're
716 ;; done. Later on we ignore the boundaries for statements that don't
717 ;; contain any sexp. The only thing that is affected is that the error
718 ;; checking is a little less strict, and we really don't bother.
728 pos saved)
730 ;; Begin at start and not pos to detect macros if we stand
731 ;; directly after the #.
734 ;; Record this as the first token if not starting inside it.
737 ;; The following while loop goes back one sexp (balanced parens,
738 ;; etc. with contents, or symbol or suchlike) each iteration. This
739 ;; movement is accomplished with a call to scan-sexps approx 130 lines
740 ;; below.
758 ;; Stop at the first token in the content of the macro.
762 ret 'macro
763 ignore-labels t))
766 ;; Do a round through the automaton if we've just passed a
767 ;; statement boundary or passed a "while"-like token.
775 ;; The PDA state handling.
776 ;;
777 ;; Refer to the description of the PDA in the opening
778 ;; comments. In the following OR form, the first leaf
779 ;; attempts to handles one of the specific actions detailed
780 ;; (e.g., finding token "if" whilst in state `else-boundary').
781 ;; We drop through to the second leaf (which handles common
782 ;; state) if no specific handler is found in the first cond.
783 ;; If a parsing error is detected (e.g. an "else" with no
784 ;; preceding "if"), we throw to the enclosing catch.
785 ;;
786 ;; Note that the (eq state 'else) means
787 ;; "we've just passed an else", NOT "we're looking for an
788 ;; else".
805 ;; Since this can cause backtracking we do a
806 ;; little more careful analysis to avoid it:
807 ;; If there's a label in front of the while
808 ;; it can't be part of a do-while.
837 ;; This is state common. We get here when the previous
838 ;; cond statement found no particular state handler.
840 ;; If we have a boundary at the start
841 ;; position we push a frame to go to the
842 ;; previous statement.
851 ;; Is this a real while, or a do-while?
852 ;; The next `when' triggers unless we are SURE that
853 ;; the `while' is not the tailend of a `do-while'.
856 ;; The following kludge is to prevent
857 ;; infinite recursion when called from
858 ;; c-awk-after-if-for-while-condition-p,
859 ;; or the like.
863 ;; Since this can cause backtracking we do a
864 ;; little more careful analysis to avoid it: If
865 ;; the while isn't followed by a (possibly
866 ;; virtual) semicolon it can't be a do-while.
875 ;; We're either past a statement boundary or at the
876 ;; start of a statement, so throw away any label data
877 ;; for the previous one.
879 last-label-pos nil
880 c-maybe-labelp nil))))
882 ;; Step to the previous sexp, but not if we crossed a
883 ;; boundary, since that doesn't consume an sexp.
887 ;; HERE IS THE SINGLE PLACE INSIDE THE PDA LOOP WHERE WE MOVE
888 ;; BACKWARDS THROUGH THE SOURCE.
890 ;; This is typically fast with the caching done by
891 ;; c-(backward|forward)-sws.
895 ;; Set as long as we have to continue jumping by sexps.
896 ;; It's the position to use as end in the next round.
897 sexp-loop-continue-pos
898 ;; The end position of the area to search for statement
899 ;; barriers in this round.
902 ;; The following while goes back one sexp per iteration.
906 ;; Give up if we hit an unbalanced block. Since the
907 ;; stack won't be empty the code below will report a
908 ;; suitable error.
911 ;; Check if the sexp movement crossed a statement or
912 ;; declaration boundary. But first modify the point
913 ;; so that `c-crosses-statement-barrier-p' only looks
914 ;; at the non-sexp chars following the sexp.
917 boundary-pos
920 nil
923 ;; Set continuation position in case
924 ;; `c-crosses-statement-barrier-p'
925 ;; doesn't detect anything below.
927 ;; If the sexp movement took us into a
928 ;; macro then there were only some non-sexp
929 ;; chars after it. Skip out of the macro
930 ;; to analyze them but not the non-sexp
931 ;; chars that might be inside the macro.
939 ;; Passed a block sexp. That's a boundary
940 ;; alright.
944 ;; Passed some other paren. Only analyze
945 ;; the non-sexp chars after it.
947 before-sws-pos)))
948 ;; We're at a valid token start position
949 ;; (outside the `save-excursion') if
950 ;; `c-crosses-statement-barrier-p' failed.
955 ;; Passed a symbol sexp or line
956 ;; continuation. It doesn't matter that
957 ;; it's included in the analyzed region.
960 t
961 ;; If it was a line continuation then we
962 ;; have to continue looping.
965 nil))))
968 ptok tok
969 tok boundary-pos
971 ;; Like a C "continue". Analyze the next sexp.
977 sexp-loop-continue-pos nil))))
979 ;; ObjC method def?
986 ;; Handle labels.
989 ;; `c-crosses-statement-barrier-p' has found a colon, so we
990 ;; might be in a label now. Have we got a real label
991 ;; (including a case label) or something like C++'s "public:"?
996 ; wrong for AWK. 2006/1/14.
1003 label-good-pos nil))
1007 ; been found.
1009 ;; We're in a potential label and it's the first
1010 ;; time we've found something that isn't allowed in
1011 ;; one.
1014 ;; We've moved back by a sexp, so update the token positions.
1016 pptok ptok
1017 ptok tok
1021 ;; If the stack isn't empty there might be errors to report.
1030 ;; Need to investigate closer whether we've crossed
1031 ;; between a substatement and its containing statement.
1033 ptok
1034 pptok))
1041 after-labels-pos
1047 last-label-pos
1048 pos))
1050 ;; We're in a label. Maybe we should step to the statement
1051 ;; after it.
1056 ;; Might have jumped over several labels. Go to the last one.
1059 ;; Skip over the unary operators that can start the statement.
1063 ;; protect AWK post-inc/decrement operators, etc.
1068 ret)))
1071 "Return non-nil if buffer positions FROM to TO cross one or more
1072 statement or declaration boundaries. The returned value is actually
1073 the position of the earliest boundary char. FROM must not be within
1074 a string or comment.
1076 The variable `c-maybe-labelp' is set to the position of the first `:' that
1077 might start a label (i.e. not part of `::' and not preceded by `?'). If a
1078 single `?' is found, then `c-maybe-labelp' is cleared.
1080 For AWK, a statement which is terminated by an EOL (not a \; or a }) is
1082 the line. If this virtual semicolon is _at_ from, the function recognises it.
1084 Note that this function might do hidden buffer changes. See the
1085 comment at the start of cc-engine.el for more info."
1087 lit-range)
1100 ;; Ignore scope operators.
1104 ;; A question mark. Can't be a label, so stop
1105 ;; looking for more : and ?.
1116 ;; In trailing space after an as yet undetected virtual semicolon?
1121 nil)))))
1124 "Return non-nil if the point is at the first token in a statement
1125 or somewhere in the syntactic whitespace before it.
1128 Any kind of declaration-like construct that occur outside function
1131 Note that this function might do hidden buffer changes. See the
1132 comment at the start of cc-engine.el for more info."
1136 c-maybe-labelp)
1147 "Return non-nil if the point is at the first token in an expression or
1148 statement, or somewhere in the syntactic whitespace before it.
1151 grammar sense: It's any sequence of expression tokens except commas,
1152 unless they are enclosed inside parentheses of some kind. Also, an
1153 expression never continues past an enclosing parenthesis, but it might
1154 contain parenthesis pairs of any sort except braces.
1156 Since expressions never cross statement boundaries, this function also
1157 recognizes statement beginnings, just like `c-at-statement-start-p'.
1159 Note that this function might do hidden buffer changes. See the
1160 comment at the start of cc-engine.el for more info."
1165 c-maybe-labelp)
1173 \f
1174 ;; A set of functions that covers various idiosyncrasies in
1175 ;; implementations of `forward-comment'.
1177 ;; Note: Some emacsen considers incorrectly that any line comment
1178 ;; ending with a backslash continues to the next line. I can't think
1179 ;; of any way to work around that in a reliable way without changing
1180 ;; the buffer, though. Suggestions welcome. ;) (No, temporarily
1181 ;; changing the syntax for backslash doesn't work since we must treat
1182 ;; escapes in string literals correctly.)
1185 "Move forward past whitespace and the closest following comment, if any.
1186 Return t if a comment was found, nil otherwise. In either case, the
1187 point is moved past the following whitespace. Line continuations,
1188 i.e. a backslashes followed by line breaks, are treated as whitespace.
1189 The line breaks that end line comments are considered to be the
1190 comment enders, so the point will be put on the beginning of the next
1191 line if it moved past a line comment.
1193 This function does not do any hidden buffer changes."
1201 ;; Some emacsen (e.g. XEmacs 21) return t when moving
1202 ;; forwards at eob.
1203 nil
1205 ;; Emacs includes the ending newline in a b-style (c++)
1206 ;; comment, but XEmacs doesn't. We depend on the Emacs
1207 ;; behavior (which also is symmetric).
1211 t))))
1214 "Move forward past all following whitespace and comments.
1215 Line continuations, i.e. a backslashes followed by line breaks, are
1216 treated as whitespace.
1218 Note that this function might do hidden buffer changes. See the
1219 comment at the start of cc-engine.el for more info."
1222 ;; If forward-comment in at least XEmacs 21 is given a large
1223 ;; positive value, it'll loop all the way through if it hits
1224 ;; eob.
1226 ;; Some emacsen (e.g. XEmacs 21) return t when moving
1227 ;; forwards at eob.
1232 t))))
1235 "Move backward past whitespace and the closest preceding comment, if any.
1236 Return t if a comment was found, nil otherwise. In either case, the
1237 point is moved past the preceding whitespace. Line continuations,
1238 i.e. a backslashes followed by line breaks, are treated as whitespace.
1239 The line breaks that end line comments are considered to be the
1240 comment enders, so the point cannot be at the end of the same line to
1241 move over a line comment.
1243 This function does not do any hidden buffer changes."
1246 ;; When we got newline terminated comments, forward-comment in all
1247 ;; supported emacsen so far will stop at eol of each line not
1248 ;; ending with a comment when moving backwards. This corrects for
1249 ;; that, and at the same time handles line continuations.
1257 ;; Some emacsen (e.g. Emacs 19.34) return t when moving
1258 ;; backwards at bob.
1259 nil
1261 ;; Leave point after the closest following newline if we've
1262 ;; backed up over any above, since forward-comment won't move
1263 ;; backward over a line comment if point is at the end of the
1264 ;; same line.
1269 ;; If forward-comment above succeeded and we're at eol
1270 ;; then the newline we moved over above didn't end a
1271 ;; line comment, so we give it another go.
1273 t))
1275 ;; Emacs <= 20 and XEmacs move back over the closer of a
1276 ;; block comment that lacks an opener.
1279 t)))))
1282 "Move backward past all preceding whitespace and comments.
1283 Line continuations, i.e. a backslashes followed by line breaks, are
1284 treated as whitespace. The line breaks that end line comments are
1285 considered to be the comment enders, so the point cannot be at the end
1286 of the same line to move over a line comment. Unlike
1287 c-backward-syntactic-ws, this function doesn't move back over
1288 preprocessor directives.
1290 Note that this function might do hidden buffer changes. See the
1291 comment at the start of cc-engine.el for more info."
1295 ;; `forward-comment' in some emacsen (e.g. XEmacs 21.4)
1296 ;; return t when moving backwards at bob.
1301 ;; Emacs <= 20 and XEmacs move back over the
1302 ;; closer of a block comment that lacks an opener.
1304 t)
1306 ;; XEmacs treats line continuations as whitespace but
1307 ;; only in the backward direction, which seems a bit
1308 ;; odd. Anyway, this is necessary for Emacs.
1313 t))))))
1315 \f
1316 ;; Tools for skipping over syntactic whitespace.
1318 ;; The following functions use text properties to cache searches over
1319 ;; large regions of syntactic whitespace. It works as follows:
1320 ;;
1321 ;; o If a syntactic whitespace region contains anything but simple
1322 ;; whitespace (i.e. space, tab and line breaks), the text property
1323 ;; `c-in-sws' is put over it. At places where we have stopped
1324 ;; within that region there's also a `c-is-sws' text property.
1325 ;; That since there typically are nested whitespace inside that
1326 ;; must be handled separately, e.g. whitespace inside a comment or
1327 ;; cpp directive. Thus, from one point with `c-is-sws' it's safe
1328 ;; to jump to another point with that property within the same
1329 ;; `c-in-sws' region. It can be likened to a ladder where
1330 ;; `c-in-sws' marks the bars and `c-is-sws' the rungs.
1331 ;;
1332 ;; o The `c-is-sws' property is put on the simple whitespace chars at
1333 ;; a "rung position" and also maybe on the first following char.
1334 ;; As many characters as can be conveniently found in this range
1335 ;; are marked, but no assumption can be made that the whole range
1336 ;; is marked (it could be clobbered by later changes, for
1337 ;; instance).
1338 ;;
1339 ;; Note that some part of the beginning of a sequence of simple
1340 ;; whitespace might be part of the end of a preceding line comment
1341 ;; or cpp directive and must not be considered part of the "rung".
1342 ;; Such whitespace is some amount of horizontal whitespace followed
1343 ;; by a newline. In the case of cpp directives it could also be
1344 ;; two newlines with horizontal whitespace between them.
1345 ;;
1346 ;; The reason to include the first following char is to cope with
1347 ;; "rung positions" that doesn't have any ordinary whitespace. If
1348 ;; `c-is-sws' is put on a token character it does not have
1349 ;; `c-in-sws' set simultaneously. That's the only case when that
1350 ;; can occur, and the reason for not extending the `c-in-sws'
1351 ;; region to cover it is that the `c-in-sws' region could then be
1352 ;; accidentally merged with a following one if the token is only
1353 ;; one character long.
1354 ;;
1355 ;; o On buffer changes the `c-in-sws' and `c-is-sws' properties are
1356 ;; removed in the changed region. If the change was inside
1357 ;; syntactic whitespace that means that the "ladder" is broken, but
1358 ;; a later call to `c-forward-sws' or `c-backward-sws' will use the
1359 ;; parts on either side and use an ordinary search only to "repair"
1360 ;; the gap.
1361 ;;
1362 ;; Special care needs to be taken if a region is removed: If there
1363 ;; are `c-in-sws' on both sides of it which do not connect inside
1364 ;; the region then they can't be joined. If e.g. a marked macro is
1365 ;; broken, syntactic whitespace inside the new text might be
1366 ;; marked. If those marks would become connected with the old
1367 ;; `c-in-sws' range around the macro then we could get a ladder
1368 ;; with one end outside the macro and the other at some whitespace
1369 ;; within it.
1370 ;;
1371 ;; The main motivation for this system is to increase the speed in
1372 ;; skipping over the large whitespace regions that can occur at the
1373 ;; top level in e.g. header files that contain a lot of comments and
1374 ;; cpp directives. For small comments inside code it's probably
1375 ;; slower than using `forward-comment' straightforwardly, but speed is
1376 ;; not a significant factor there anyway.
1378 ; (defface c-debug-is-sws-face
1379 ; '((t (:background "GreenYellow")))
1380 ; "Debug face to mark the `c-is-sws' property.")
1381 ; (defface c-debug-in-sws-face
1382 ; '((t (:underline t)))
1383 ; "Debug face to mark the `c-in-sws' property.")
1385 ; (defun c-debug-put-sws-faces ()
1386 ; ;; Put the sws debug faces on all the `c-is-sws' and `c-in-sws'
1387 ; ;; properties in the buffer.
1388 ; (interactive)
1389 ; (save-excursion
1390 ; (c-save-buffer-state (in-face)
1391 ; (goto-char (point-min))
1392 ; (setq in-face (if (get-text-property (point) 'c-is-sws)
1393 ; (point)))
1394 ; (while (progn
1395 ; (goto-char (next-single-property-change
1396 ; (point) 'c-is-sws nil (point-max)))
1397 ; (if in-face
1398 ; (progn
1399 ; (c-debug-add-face in-face (point) 'c-debug-is-sws-face)
1400 ; (setq in-face nil))
1401 ; (setq in-face (point)))
1402 ; (not (eobp))))
1403 ; (goto-char (point-min))
1404 ; (setq in-face (if (get-text-property (point) 'c-in-sws)
1405 ; (point)))
1406 ; (while (progn
1407 ; (goto-char (next-single-property-change
1408 ; (point) 'c-in-sws nil (point-max)))
1409 ; (if in-face
1410 ; (progn
1411 ; (c-debug-add-face in-face (point) 'c-debug-in-sws-face)
1412 ; (setq in-face nil))
1413 ; (setq in-face (point)))
1414 ; (not (eobp)))))))
1417 ;;`(message ,@args)
1418 )
1421 ;; This macro does a hidden buffer change.
1428 ;; This macro does a hidden buffer change.
1435 ;; This macro does a hidden buffer change.
1442 ;; This macro does a hidden buffer change.
1449 ;; This macro does a hidden buffer change.
1457 ;; Called from `after-change-functions'. Note that if
1458 ;; `c-forward-sws' or `c-backward-sws' are used outside
1459 ;; `c-save-buffer-state' or similar then this will remove the cache
1460 ;; properties right after they're added.
1461 ;;
1462 ;; This function does hidden buffer changes.
1465 ;; Adjust the end to remove the properties in any following simple
1466 ;; ws up to and including the next line break, if there is any
1467 ;; after the changed region. This is necessary e.g. when a rung
1468 ;; marked empty line is converted to a line comment by inserting
1469 ;; "//" before the line break. In that case the line break would
1470 ;; keep the rung mark which could make a later `c-backward-sws'
1471 ;; move into the line comment instead of over it.
1481 ;; Ensure that an `c-in-sws' range gets broken. Note that it isn't
1482 ;; safe to keep a range that was continuous before the change. E.g:
1483 ;;
1484 ;; #define foo
1485 ;; \
1486 ;; bar
1487 ;;
1488 ;; There can be a "ladder" between "#" and "b". Now, if the newline
1489 ;; after "foo" is removed then "bar" will become part of the cpp
1490 ;; directive instead of a syntactically relevant token. In that
1491 ;; case there's no longer syntactic ws from "#" to "b".
1498 ;; Used by `c-forward-syntactic-ws' to implement the unbounded search.
1499 ;;
1500 ;; This function might do hidden buffer changes.
1503 ;; unmarked part of the simple ws region.
1505 rung-is-marked next-rung-is-marked simple-ws-end
1506 ;; `safe-start' is set when it's safe to cache the start position.
1507 ;; It's not set if we've initially skipped over comments and line
1508 ;; continuations since we might have gone out through the end of a
1509 ;; macro then. This provision makes `c-forward-sws' not populate the
1510 ;; cache in the majority of cases, but otoh is `c-backward-sws' by far
1511 ;; more common.
1512 safe-start)
1514 ;; Skip simple ws and do a quick check on the following character to see
1515 ;; if it's anything that can't start syntactic ws, so we can bail out
1516 ;; early in the majority of cases when there just are a few ws chars.
1523 ;; Find the last rung position to avoid setting properties in all
1524 ;; the cases when the marked rung is complete.
1525 ;; (`next-single-property-change' is certain to move at least one
1526 ;; step forward.)
1529 ;; Got no marked rung here. Since the simple ws might have started
1530 ;; inside a line comment or cpp directive we must set `rung-pos' as
1531 ;; high as possible.
1540 ;; The following search is the main reason that `c-in-sws'
1541 ;; and `c-is-sws' aren't combined to one property.
1545 ;; If the `c-in-sws' region extended past the last
1546 ;; `c-is-sws' char we have to go back a bit.
1553 "c-forward-sws cached move %s -> %s (max %s)"
1560 ;; We'll loop here if there is simple ws after the last rung.
1561 ;; That means that there's been some change in it and it's
1562 ;; possible that we've stepped into another ladder, so extend
1563 ;; the previous one to join with it if there is one, and try to
1564 ;; use the cache again.
1566 "c-forward-sws extending rung with [%s..%s] (max %s)"
1569 ;; Remove any `c-in-sws' property from the last char of
1570 ;; the rung before we mark it with `c-is-sws', so that we
1571 ;; won't connect with the remains of a broken "ladder".
1577 last-put-in-sws-pos rung-pos)))
1584 ;; Skipped over comments. Don't cache at eob in case the buffer
1585 ;; is narrowed.
1596 ;; Skip a preprocessor directive.
1603 ;; Don't cache at eob in case the buffer is narrowed.
1606 ;; We've searched over a piece of non-white syntactic ws. See if this
1607 ;; can be cached.
1613 ;; Cache if we haven't skipped comments only, and if we started
1614 ;; either from a marked rung or from a completely uncached
1615 ;; position.
1620 ;; See if there's a marked rung in the encountered simple ws. If
1621 ;; so then we can cache, unless `safe-start' is nil. Even then
1622 ;; we need to do this to check if the cache can be used for the
1623 ;; next step.
1627 safe-start))
1631 "c-forward-sws caching [%s..%s] - [%s..%s] (max %s)"
1635 ;; Remove the properties for any nested ws that might be cached.
1636 ;; Only necessary for `c-is-sws' since `c-in-sws' will be set
1637 ;; anyway.
1645 last-put-in-sws-pos rung-pos))
1647 ;; Remove any `c-in-sws' property from the last char of
1648 ;; the rung before we mark it with `c-is-sws', so that we
1649 ;; won't connect with the remains of a broken "ladder".
1652 rung-end-pos))
1655 "c-forward-sws not caching [%s..%s] - [%s..%s] (max %s)"
1659 ;; Set `rung-pos' for the next rung. It's the same thing here as
1660 ;; initially, except that the rung position is set as early as
1661 ;; possible since we can't be in the ending ws of a line comment or
1662 ;; cpp directive now.
1669 ;; Make sure that the newly marked `c-in-sws' region doesn't connect to
1670 ;; another one after the point (which might occur when editing inside a
1671 ;; comment or macro).
1675 "c-forward-sws clearing at %s for cache separation"
1676 last-put-in-sws-pos)
1680 ;; If at eob we have to clear the last character before the end
1681 ;; instead since the buffer might be narrowed and there might
1682 ;; be a `c-in-sws' after (point-max). In this case it's
1683 ;; necessary to clear both properties.
1685 "c-forward-sws clearing thoroughly at %s for cache separation"
1688 last-put-in-sws-pos))))
1689 )))
1692 ;; Used by `c-backward-syntactic-ws' to implement the unbounded search.
1693 ;;
1694 ;; This function might do hidden buffer changes.
1697 ;; part of the simple ws region.
1699 rung-is-marked simple-ws-beg cmt-skip-pos)
1701 ;; Skip simple horizontal ws and do a quick check on the preceding
1702 ;; character to see if it's anying that can't end syntactic ws, so we can
1703 ;; bail out early in the majority of cases when there just are a few ws
1704 ;; chars. Newlines are complicated in the backward direction, so we can't
1705 ;; skip over them.
1712 ;; Try to find a rung position in the simple ws preceding point, so that
1713 ;; we can get a cache hit even if the last bit of the simple ws has
1714 ;; changed recently.
1720 ;; `rung-pos' will be the earliest marked position, which means that
1721 ;; there might be later unmarked parts in the simple ws region.
1722 ;; It's not worth the effort to fix that; the last part of the
1723 ;; simple ws is also typically edited often, so it could be wasted.
1734 ;; The following search is the main reason that `c-in-sws'
1735 ;; and `c-is-sws' aren't combined to one property.
1739 ;; If the `c-in-sws' region extended past the first
1740 ;; `c-is-sws' char we have to go forward a bit.
1745 "c-backward-sws cached move %s <- %s (min %s)"
1757 t
1759 nil))
1761 ;; We'll loop here if there is simple ws before the first rung.
1762 ;; That means that there's been some change in it and it's
1763 ;; possible that we've stepped into another ladder, so extend
1764 ;; the previous one to join with it if there is one, and try to
1765 ;; use the cache again.
1767 "c-backward-sws extending rung with [%s..%s] (min %s)"
1770 ;; Remove any `c-in-sws' property from the last char of
1771 ;; the rung before we mark it with `c-is-sws', so that we
1772 ;; won't connect with the remains of a broken "ladder".
1775 rung-pos)
1779 last-put-in-sws-pos rung-pos))
1788 ;; Inside a cpp directive. See if it should be skipped over.
1791 ;; Move back over all line continuations in the region skipped
1792 ;; over by `c-backward-comments'. If we go past it then we
1793 ;; started inside the cpp directive.
1802 ;; Don't move past the cpp directive if we began inside
1803 ;; it. Note that the position at the end of the last line
1804 ;; of the macro is also considered to be within it.
1806 nil)
1808 ;; It's worthwhile to spend a little bit of effort on finding
1809 ;; the end of the macro, to get a good `simple-ws-beg'
1810 ;; position for the cache. Note that `c-backward-comments'
1811 ;; could have stepped over some comments before going into
1812 ;; the macro, and then `simple-ws-beg' must be kept on the
1813 ;; same side of those comments.
1820 ;; Might happen if comments after the macro were skipped
1821 ;; over.
1825 t)))
1830 simple-ws-beg)
1831 ;; Skipped over comments. Must put point at the end of
1832 ;; the simple ws at point since we might be after a line
1833 ;; comment or cpp directive that's been partially
1834 ;; narrowed out, and we can't risk marking the simple ws
1835 ;; at the end of it.
1837 t)))
1839 ;; We've searched over a piece of non-white syntactic ws. See if this
1840 ;; can be cached.
1845 ;; Cache if we started either from a marked rung or from a
1846 ;; completely uncached position.
1847 rung-is-marked
1850 ;; Cache if there's a marked rung in the encountered simple ws.
1858 "c-backward-sws caching [%s..%s] - [%s..%s] (min %s)"
1863 ;; Remove the properties for any nested ws that might be cached.
1864 ;; Only necessary for `c-is-sws' since `c-in-sws' will be set
1865 ;; anyway.
1870 ;; Remove any `c-in-sws' property from the last char of
1871 ;; the rung before we mark it with `c-is-sws', so that we
1872 ;; won't connect with the remains of a broken "ladder".
1875 rung-end-pos)
1878 last-put-in-sws-pos simple-ws-beg)
1879 rung-pos)
1884 "c-backward-sws not caching [%s..%s] - [%s..%s] (min %s)"
1890 ))
1892 ;; Make sure that the newly marked `c-in-sws' region doesn't connect to
1893 ;; another one before the point (which might occur when editing inside a
1894 ;; comment or macro).
1898 "c-backward-sws clearing at %s for cache separation"
1901 last-put-in-sws-pos))
1903 ;; If at bob and the buffer is narrowed, we have to clear the
1904 ;; character we're standing on instead since there might be a
1905 ;; `c-in-sws' before (point-min). In this case it's necessary
1906 ;; to clear both properties.
1908 "c-backward-sws clearing thoroughly at %s for cache separation"
1909 last-put-in-sws-pos)
1912 )))
1914 \f
1915 ;; Other whitespace tools
1917 ;; Is the region (beg end) WS, and is there WS (or BOB/EOB) next to the
1918 ;; region? This is a "heuristic" function. .....
1919 ;;
1920 ;; The motivation for the second bit is to check whether the removal of this
1921 ;; space is to check whether removing this region would coalesce two
1922 ;; symbols.
1923 ;;
1924 ;; FIXME!!! This function doesn't check virtual semicolons in any way. Be
1925 ;; careful about using this function for, e.g. AWK. (2007/3/7)
1934 \f
1935 ;; A system for finding noteworthy parens before the point.
1939 ;; The state cache used by `c-parse-state' to cut down the amount of
1940 ;; searching. It's the result from some earlier `c-parse-state' call.
1941 ;;
1942 ;; The use of the cached info is more effective if the next
1943 ;; `c-parse-state' call is on a line close by the one the cached state
1944 ;; was made at; the cache can actually slow down a little if the
1945 ;; cached state was made very far back in the buffer. The cache is
1946 ;; most effective if `c-parse-state' is used on each line while moving
1947 ;; forward.
1951 ;; This is (point-min) when `c-state-cache' was calculated, since a
1952 ;; change of narrowing is likely to affect the parens that are visible
1953 ;; before the point.
1957 ;; This is a position where `c-state-cache' is known to be correct.
1958 ;; It's a position inside one of the recorded unclosed parens or the
1959 ;; top level, but not further nested inside any literal or subparen
1960 ;; that is closed before the last recorded position.
1961 ;;
1962 ;; The exact position is chosen to try to be close to yet earlier than
1963 ;; the position where `c-state-cache' will be called next. Right now
1964 ;; the heuristic is to set it to the position after the last found
1965 ;; closing paren (of any type) before the line on which
1966 ;; `c-parse-state' was called. That is chosen primarily to work well
1967 ;; with refontification of the current line.
1970 ;; Invalidate all info on `c-state-cache' that applies to the buffer
1971 ;; at POS or higher. This is much like `c-whack-state-after', but
1972 ;; it never changes a paren pair element into an open paren element.
1973 ;; Doing that would mean that the new open paren wouldn't have the
1974 ;; required preceding paren pair element.
1978 nil))
1984 nil))
1988 nil)))))
1992 ;; Return the start position for building `c-state-cache' from
1993 ;; scratch.
1995 ;; Go back 2 bods, but ignore any bogus positions returned by
1996 ;; beginning-of-defun (i.e. open paren in column zero).
2006 ;; Find and record all noteworthy parens between some good point
2007 ;; earlier in the file and point. That good point is at least the
2008 ;; beginning of the top-level construct we are in, or the beginning
2009 ;; of the preceding top-level construct if we aren't in one.
2010 ;;
2011 ;; The returned value is a list of the noteworthy parens with the
2012 ;; last one first. If an element in the list is an integer, it's
2013 ;; the position of an open paren which has not been closed before
2014 ;; the point. If an element is a cons, it gives the position of a
2015 ;; closed brace paren pair; the car is the start paren position and
2016 ;; the cdr is the position following the closing paren. Only the
2017 ;; last closed brace paren pair before each open paren and before
2018 ;; the point is recorded, and thus the state never contains two cons
2019 ;; elements in succession.
2020 ;;
2021 ;; Currently no characters which are given paren syntax with the
2022 ;; syntax-table property are recorded, i.e. angle bracket arglist
2023 ;; parens are never present here. Note that this might change.
2024 ;;
2025 ;; BUG: This function doesn't cope entirely well with unbalanced
2026 ;; parens in macros. E.g. in the following case the brace before
2027 ;; the macro isn't balanced with the one after it:
2028 ;;
2029 ;; {
2030 ;; #define X {
2031 ;; }
2032 ;;
2033 ;; This function might do hidden buffer changes.
2040 old-state last-pos brace-pair-open brace-pair-close
2041 pos save-pos)
2044 ;; If the minimum position has changed due to narrowing then we
2045 ;; have to fix the tail of `c-state-cache' accordingly.
2048 ;; If point-min has moved forward then we just need to cut
2049 ;; off a bit of the tail.
2060 ;; If point-min has moved backward then we drop the state
2061 ;; completely. It's possible to do a better job here and
2062 ;; recalculate the top only.
2067 ;; Get the latest position we know are directly inside the
2068 ;; closest containing paren of the cached state.
2076 ;; Take the opportunity to move the cached good position
2077 ;; further down.
2081 ;; Check if `last-pos' is in a macro. If it is, and we're not
2082 ;; in the same macro, we must discard everything on
2083 ;; `c-state-cache' that is inside the macro before using it.
2089 ;; Set `last-pos' again just like above except that there's
2090 ;; no use looking at `c-state-cache-good-pos' here.
2097 ;; If we've moved very far from the last cached position then
2098 ;; it's probably better to redo it from scratch, otherwise we
2099 ;; might spend a lot of time searching from `last-pos' down to
2100 ;; here.
2102 ;; First get the fallback start position. If it turns out
2103 ;; that it's so far back that the cached state is closer then
2104 ;; we'll keep it afterall.
2109 c-state-cache nil
2112 ;; Find the start position for the forward search. (Can't
2113 ;; search in the backward direction since the point might be in
2114 ;; some kind of literal.)
2119 ;; There's a cached state with a containing paren. Pop off
2120 ;; the stale containing sexps from it by going forward out of
2121 ;; parens as far as possible.
2135 ;; The last paren pair we moved out from was a brace
2136 ;; pair. Modify the state to record this as a closed
2137 ;; pair now.
2141 c-state-cache))))
2143 ;; Check if the preceding balanced paren is within a
2144 ;; macro; it should be ignored if we're outside the
2145 ;; macro. There's no need to check any further upwards;
2146 ;; if the macro contains an unbalanced opening paren then
2147 ;; we're smoked anyway.
2157 ;; Have to adjust the cached good position if state has been
2158 ;; popped off.
2165 old-state c-state-cache))
2170 ;; Get the fallback start position.
2173 c-state-cache nil
2180 brace-pair-open nil)
2182 ;; Find the balanced brace pairs. This loop is hot, so it
2183 ;; does ugly tricks to go faster.
2193 ;; Update the cached good position and record the brace
2194 ;; pair, whichever is applicable for the paren we've
2195 ;; just jumped over. But first check that it isn't
2196 ;; inside a macro and the point isn't inside the same
2197 ;; one.
2207 brace-pair-close pos))))))
2209 ;; Record the last brace pair.
2217 brace-pair-close)
2218 c-state-cache)))))
2221 ;; Prepare to loop, but record the open paren only if it's
2222 ;; outside a macro or within the same macro as point, and
2223 ;; if it is a legitimate open paren and not some character
2224 ;; that got an open paren syntax-table property.
2231 ;; Check for known types of parens that we
2232 ;; want to record. The syntax table is not to
2233 ;; be trusted here since the caller might be
2234 ;; using e.g. `c++-template-syntax-table'.
2241 ;; Found a close paren without a corresponding opening
2242 ;; one. Maybe we didn't go back far enough, so try to
2243 ;; scan backward for the start paren and then start over.
2246 c-state-cache nil
2247 c-state-cache-good-pos c-state-cache-start)
2249 ;; Emacs (up to at least 21.2) can get confused by
2250 ;; open parens in column zero inside comments: The
2251 ;; sexp functions can then misbehave and bring us
2252 ;; back to the same point again. Check this so that
2253 ;; we don't get an infinite loop.
2256 c-parsing-error
2262 ;;(message "c-parse-state: %S end: %S" c-state-cache c-state-cache-good-pos)
2263 c-state-cache)))
2265 ;; Debug tool to catch cache inconsistencies.
2277 ;; The cache can actually go further back due to the ad-hoc way
2278 ;; the first paren is found, so try to whack off a bit of its
2279 ;; start before complaining.
2288 res1 res2))))
2289 res1))
2294 'c-debug-parse-state
2301 ;; Whack off any state information from PAREN-STATE which lies
2302 ;; before BUFPOS. Not destructive on PAREN-STATE.
2305 car)
2316 ;; Whack off any state information from PAREN-STATE which lies at or
2317 ;; after BUFPOS. Not destructive on PAREN-STATE.
2322 ;; just check the car, because in a balanced brace
2323 ;; expression, it must be impossible for the corresponding
2324 ;; close brace to be before point, but the open brace to
2325 ;; be after.
2327 nil ; whack it off
2329 ;; its possible that the open brace is before
2330 ;; bufpos, but the close brace is after. In that
2331 ;; case, convert this to a non-cons element. The
2332 ;; rest of the state is before bufpos, so we're
2333 ;; done.
2335 ;; we know that both the open and close braces are
2336 ;; before bufpos, so we also know that everything else
2337 ;; on state is before bufpos.
2340 nil ; whack it off
2341 ;; it's before bufpos, so everything else should too.
2344 nil)))
2347 ;; Return the bufpos of the innermost enclosing open paren before
2348 ;; bufpos, or nil if none was found.
2358 enclosingp))
2361 ;; Return the bufpos of the outermost enclosing open paren, or nil
2362 ;; if none was found.
2369 pos))
2372 ;; Return the closest "safe" position recorded on PAREN-STATE that
2373 ;; is higher up than BUFPOS. Return nil if PAREN-STATE doesn't
2374 ;; contain any. Return nil if BUFPOS is nil, which is useful to
2375 ;; find the closest limit before a given limit that might be nil.
2376 ;;
2377 ;; A "safe" position is a position at or after a recorded open
2378 ;; paren, or after a recorded close paren. The returned position is
2379 ;; thus either the first position after a close brace, or the first
2380 ;; position after an enclosing paren, or at the enclosing paren in
2381 ;; case BUFPOS is immediately after it.
2391 ;; See below.
2394 ;; elem is the position at and not after the opening paren, so
2395 ;; we can go forward one more step unless it's equal to
2396 ;; bufpos. This is useful in some cases avoid an extra paren
2397 ;; level between the safe position and bufpos.
2402 ;; This is used for `font-lock-beginning-of-syntax-function'. It
2403 ;; goes to the closest previous point that is known to be outside
2404 ;; any string literal or comment. `c-state-cache' is used if it has
2405 ;; a position in the vicinity.
2407 elem
2410 ;; Note: Similar code in `c-safe-position'. The
2411 ;; difference is that we accept a safe position at
2412 ;; the point and don't bother to go forward past open
2413 ;; parens.
2428 ;; The position is far back. Try `c-beginning-of-defun-1'
2429 ;; (although we can't be entirely sure it will go to a position
2430 ;; outside a comment or string in current emacsen). FIXME:
2431 ;; Consult `syntax-ppss' here.
2436 \f
2437 ;; Tools for scanning identifiers and other tokens.
2440 "Return non-nil if the point is on or directly after an identifier.
2441 Keywords are recognized and not considered identifiers. If an
2442 identifier is detected, the returned value is its starting position.
2443 If an identifier ends at the point and another begins at it \(can only
2444 happen in Pike) then the point for the preceding one is returned.
2446 Note that this function might do hidden buffer changes. See the
2447 comment at the start of cc-engine.el for more info."
2449 ;; FIXME: Shouldn't this function handle "operator" in C++?
2456 ;; Check for a normal (non-keyword) identifier.
2462 ;; Handle the `<operator> syntax in Pike.
2466 t
2473 ;; Handle the "operator +" syntax in C++.
2490 )))
2493 ;; If the point is at the end of a symbol then skip backward to the
2494 ;; beginning of it. Don't move otherwise. Return non-nil if point
2495 ;; moved.
2496 ;;
2497 ;; This function might do hidden buffer changes.
2500 ;; Handle the `<operator> syntax in Pike.
2506 t
2508 nil)))))
2511 ;; Move to the beginning of the current token. Do not move if not
2512 ;; in the middle of one. BACK-LIMIT may be used to bound the
2513 ;; backward search; if given it's assumed to be at the boundary
2514 ;; between two tokens. Return non-nil if the point is move, nil
2515 ;; otherwise.
2516 ;;
2517 ;; This function might do hidden buffer changes.
2524 ;; `c-nonsymbol-token-regexp' should always match
2525 ;; since we've skipped backward over punctuator
2526 ;; or paren syntax, but consume one char in case
2527 ;; it doesn't so that we don't leave point before
2528 ;; some earlier incorrect token.
2535 ;; Move to the end of the current token. Do not move if not in the
2536 ;; middle of one. BACK-LIMIT may be used to bound the backward
2537 ;; search; if given it's assumed to be at the boundary between two
2538 ;; tokens. Return non-nil if the point is moved, nil otherwise.
2539 ;;
2540 ;; This function might do hidden buffer changes.
2548 ;; `c-nonsymbol-token-regexp' should always match since
2549 ;; we've skipped backward over punctuator or paren
2550 ;; syntax, but move forward in case it doesn't so that
2551 ;; we don't leave point earlier than we started with.
2567 "Move forward by tokens.
2568 A token is defined as all symbols and identifiers which aren't
2570 treated properly). Point is always either left at the beginning of a
2571 token or not moved at all. COUNT specifies the number of tokens to
2572 move; a negative COUNT moves in the opposite direction. A COUNT of 0
2573 moves to the next token beginning only if not already at one. If
2574 BALANCED is true, move over balanced parens, otherwise move into them.
2575 Also, if BALANCED is true, never move out of an enclosing paren.
2577 LIMIT sets the limit for the movement and defaults to the point limit.
2578 The case when LIMIT is set in the middle of a token, comment or macro
2579 is handled correctly, i.e. the point won't be left there.
2581 Return the number of tokens left to move \(positive or negative). If
2582 BALANCED is true, a move over a balanced paren counts as one. Note
2583 that if COUNT is 0 and no appropriate token beginning is found, 1 will
2584 be returned. Thus, a return value of 0 guarantees that point is at
2585 the requested position and a return value less \(without signs) than
2586 COUNT guarantees that point is at the beginning of some token.
2588 Note that this function might do hidden buffer changes. See the
2589 comment at the start of cc-engine.el for more info."
2596 c-jump-syntax-balanced
2597 c-jump-syntax-unbalanced))
2602 ;; If count is zero we should jump if in the middle of a token.
2609 ;; Skip whitespace. Count this as a move if we did in
2610 ;; fact move.
2614 ;; Moved out of bounds. Make sure the returned count isn't zero.
2619 ;; Use `condition-case' to avoid having the limit tests
2620 ;; inside the loop.
2628 t)
2631 t)
2632 ;; `c-nonsymbol-token-regexp' above should always
2633 ;; match if there are correct tokens. Try to
2634 ;; widen to see if the limit was set in the
2635 ;; middle of one, else fall back to treating
2636 ;; the offending thing as a one character token.
2641 nil)
2644 t))))
2654 count)))
2657 "Move backward by tokens.
2658 See `c-forward-token-2' for details."
2666 c-jump-syntax-balanced
2667 c-jump-syntax-unbalanced))
2671 ;; The count is zero so try to skip to the beginning of the
2672 ;; current token.
2676 ;; The limit is inside the same token, so return 1.
2679 ;; We're not in the middle of a token. If there's
2680 ;; whitespace after the point then we must move backward,
2681 ;; so set count to 1 in that case.
2683 ;; If we're looking at a '#' that might start a cpp
2684 ;; directive then we have to do a more elaborate check.
2697 ;; Use `condition-case' to avoid having to check for buffer
2698 ;; limits in `backward-char', `scan-sexps' and `goto-char' below.
2707 ;; This can be very inefficient if there's a long
2708 ;; sequence of operator tokens without any separation.
2709 ;; That doesn't happen in practice, anyway.
2719 count)))
2722 "Like `c-forward-token-2' but doesn't treat multicharacter operator
2724 characters are jumped over character by character. This function is
2725 for compatibility only; it's only a wrapper over `c-forward-token-2'."
2730 "Like `c-backward-token-2' but doesn't treat multicharacter operator
2732 characters are jumped over character by character. This function is
2733 for compatibility only; it's only a wrapper over `c-backward-token-2'."
2737 \f
2738 ;; Tools for doing searches restricted to syntactically relevant text.
2741 paren-level not-inside-token
2742 lookbehind-submatch)
2743 "Like `re-search-forward', but only report matches that are found
2744 in syntactically significant text. I.e. matches in comments, macros
2745 or string literals are ignored. The start point is assumed to be
2746 outside any comment, macro or string literal, or else the content of
2747 that region is taken as syntactically significant text.
2749 If PAREN-LEVEL is non-nil, an additional restriction is added to
2750 ignore matches in nested paren sexps. The search will also not go
2751 outside the current list sexp, which has the effect that if the point
2752 should be moved to BOUND when no match is found \(i.e. NOERROR is
2753 neither nil nor t), then it will be at the closing paren if the end of
2754 the current list sexp is encountered first.
2756 If NOT-INSIDE-TOKEN is non-nil, matches in the middle of tokens are
2757 ignored. Things like multicharacter operators and special symbols
2759 constants.
2761 If LOOKBEHIND-SUBMATCH is non-nil, it's taken as a number of a
2762 subexpression in REGEXP. The end of that submatch is used as the
2763 position to check for syntactic significance. If LOOKBEHIND-SUBMATCH
2764 isn't used or if that subexpression didn't match then the start
2766 subexpression is never tested before the starting position, so it
2770 subexpression can match the end of nonwhite syntactic whitespace,
2771 i.e. the end of comments or cpp directives. This since the function
2772 skips over such things before resuming the search. It's on the other
2774 matches syntactic whitespace.
2776 Bug: Unbalanced parens inside cpp directives are currently not handled
2777 correctly \(i.e. they don't get ignored as they should) when
2778 PAREN-LEVEL is set.
2780 Note that this function might do hidden buffer changes. See the
2781 comment at the start of cc-engine.el for more info."
2786 ;;(message "c-syntactic-re-search-forward %s %s %S" (point) bound regexp)
2789 tmp
2790 ;; Start position for the last search.
2791 search-pos
2792 ;; The `parse-partial-sexp' state between the start position
2793 ;; and the point.
2794 state
2795 ;; The current position after the last state update. The next
2796 ;; `parse-partial-sexp' continues from here.
2798 ;; The position at which to check the state and the state
2799 ;; there. This is separate from `state-pos' since we might
2800 ;; need to back up before doing the next search round.
2801 check-pos check-state
2802 ;; Last position known to end a token.
2804 ;; Set when a valid match is found.
2805 found)
2823 state-pos check-pos paren-level nil state))
2825 check-state state))
2827 ;; NOTE: If we got a look behind subexpression and get
2828 ;; an insignificant match in something that isn't
2829 ;; syntactic whitespace (i.e. strings or in nested
2830 ;; parentheses), then we can never skip more than a
2831 ;; single character from the match start position
2832 ;; (i.e. `state-pos' here) before continuing the
2833 ;; search. That since the look behind subexpression
2834 ;; might match the end of the insignificant region in
2835 ;; the next search.
2839 ;; Match inside a line comment. Skip to eol. Use
2840 ;; `re-search-forward' instead of `skip-chars-forward' to get
2841 ;; the right bound behavior.
2845 ;; Match inside a block comment. Skip to the '*/'.
2852 ;; Match in the middle of the opener of a block or line
2853 ;; comment.
2858 ;; The last `parse-partial-sexp' above might have
2859 ;; stopped short of the real check position if the end
2860 ;; of the current sexp was encountered in paren-level
2861 ;; mode. The checks above are always false in that
2862 ;; case, and since they can do better skipping in
2863 ;; lookbehind-submatch mode, we do them before
2864 ;; checking the paren level.
2868 ;; Check the paren level first since we're short of the
2869 ;; syntactic checking position if the end of the
2870 ;; current sexp was encountered by `parse-partial-sexp'.
2873 ;; Inside a nested paren sexp.
2875 ;; See the NOTE above.
2877 ;; Skip out of the paren quickly.
2881 ;; Have exited the current paren sexp.
2884 ;; The last `parse-partial-sexp' call above
2885 ;; has left us just after the closing paren
2886 ;; in this case, so we can modify the bound
2887 ;; to leave the point at the right position
2888 ;; upon return.
2890 nil)
2894 ;; Match inside a string.
2897 ;; See the NOTE above.
2899 ;; Skip to the end of the string before continuing.
2908 continue)))
2913 ;; Match inside a macro. Skip to the end of it.
2921 (< check-pos
2927 ;; Inside a token.
2929 ;; See the NOTE above.
2934 ;; A real match.
2936 nil)))
2938 ;; Should loop to search again, but take care to avoid
2939 ;; looping on the same spot.
2943 nil
2946 t))))
2952 ;;(message "c-syntactic-re-search-forward done %s" (or (match-end 0) (point)))
2959 ;; Search failed. Set point as appropriate.
2963 nil)))
2966 "Like `skip-chars-backward' but only look at syntactically relevant chars,
2967 i.e. don't stop at positions inside syntactic whitespace or string
2968 literals. Preprocessor directives are also ignored, with the exception
2969 of the one that the point starts within, if any. If LIMIT is given,
2970 it's assumed to be at a syntactically relevant position.
2972 If PAREN-LEVEL is non-nil, the function won't stop in nested paren
2973 sexps, and the search will also not go outside the current paren sexp.
2974 However, if LIMIT or the buffer limit is reached inside a nested paren
2975 then the point will be left at the limit.
2977 Non-nil is returned if the point moved, nil otherwise.
2979 Note that this function might do hidden buffer changes. See the
2980 comment at the start of cc-engine.el for more info."
2983 state
2984 ;; A list of syntactically relevant positions in descending
2985 ;; order. It's used to avoid scanning repeatedly over
2986 ;; potentially large regions with `parse-partial-sexp' to verify
2987 ;; each position.
2988 safe-pos-list
2989 ;; The position at the beginning of `safe-pos-list'.
2990 safe-pos
2991 ;; The result from `c-beginning-of-macro' at the start position or the
2992 ;; start position itself if it isn't within a macro. Evaluated on
2993 ;; demand.
2994 start-macro-beg
2995 ;; The earliest position after the current one with the same paren
2996 ;; level. Used only when `paren-level' is set.
3003 ;; Use `parse-partial-sexp' from a safe position down to
3004 ;; the point to check if it's outside comments and
3005 ;; strings.
3007 ;; Pick a safe position as close to the point as
3008 ;; possible.
3009 ;;
3010 ;; FIXME: Consult `syntax-ppss' here if our
3011 ;; cache doesn't give a good position.
3023 ;; Cache positions along the way to use if we have to
3024 ;; back up more. We cache every closing paren on the
3025 ;; same level. If the paren cache is relevant in this
3026 ;; region then we're typically already on the same
3027 ;; level as the target position. Note that we might
3028 ;; cache positions after opening parens in case
3029 ;; safe-pos is in a nested list. That's both uncommon
3030 ;; and harmless.
3040 ;; Inside string or comment. Continue search at the
3041 ;; beginning of it.
3043 t)
3051 ;; Not at the right level.
3054 ;; We stop above if we go out of a paren.
3055 ;; Now check whether it precedes or is
3056 ;; nested in the starting sexp.
3060 pps-end-pos paren-level-pos
3061 nil nil state-2))
3064 ;; We've stopped short of the starting position
3065 ;; so the hit was inside a nested list. Go up
3066 ;; until we are at the right level.
3075 nil)
3076 t))
3079 nil))
3081 ;; The hit was outside the list at the start
3082 ;; position. Go to the start of the list and exit.
3084 nil))
3087 ;; Inside a macro.
3095 t
3097 ;; It's inside the same macro we started in so it's
3098 ;; a relevant match.
3100 nil)))))
3102 ;; If the state contains the start of the containing sexp we
3103 ;; cache that position too, so that parse-partial-sexp in the
3104 ;; next run has a bigger chance of starting at the same level
3105 ;; as the target position and thus will get more good safe
3106 ;; positions into the list.
3113 ;; Skip syntactic ws afterwards so that we don't stop at the
3114 ;; end of a comment if `skip-chars' is something like "^/".
3118 ;; We might want to extend this with more useful return values in
3119 ;; the future.
3122 ;; The following is an alternative implementation of
3123 ;; `c-syntactic-skip-backward' that uses backward movement to keep
3124 ;; track of the syntactic context. It turned out to be generally
3125 ;; slower than the one above which uses forward checks from earlier
3126 ;; safe positions.
3127 ;;
3128 ;;(defconst c-ssb-stop-re
3129 ;; ;; The regexp matching chars `c-syntactic-skip-backward' needs to
3130 ;; ;; stop at to avoid going into comments and literals.
3131 ;; (concat
3132 ;; ;; Match comment end syntax and string literal syntax. Also match
3133 ;; ;; '/' for block comment endings (not covered by comment end
3134 ;; ;; syntax).
3135 ;; "\\s>\\|/\\|\\s\""
3136 ;; (if (memq 'gen-string-delim c-emacs-features)
3137 ;; "\\|\\s|"
3138 ;; "")
3139 ;; (if (memq 'gen-comment-delim c-emacs-features)
3140 ;; "\\|\\s!"
3141 ;; "")))
3142 ;;
3143 ;;(defconst c-ssb-stop-paren-re
3144 ;; ;; Like `c-ssb-stop-re' but also stops at paren chars.
3145 ;; (concat c-ssb-stop-re "\\|\\s(\\|\\s)"))
3146 ;;
3147 ;;(defconst c-ssb-sexp-end-re
3148 ;; ;; Regexp matching the ending syntax of a complex sexp.
3149 ;; (concat c-string-limit-regexp "\\|\\s)"))
3150 ;;
3151 ;;(defun c-syntactic-skip-backward (skip-chars &optional limit paren-level)
3152 ;; "Like `skip-chars-backward' but only look at syntactically relevant chars,
3153 ;;i.e. don't stop at positions inside syntactic whitespace or string
3154 ;;literals. Preprocessor directives are also ignored. However, if the
3155 ;;point is within a comment, string literal or preprocessor directory to
3156 ;;begin with, its contents is treated as syntactically relevant chars.
3157 ;;If LIMIT is given, it limits the backward search and the point will be
3158 ;;left there if no earlier position is found.
3159 ;;
3160 ;;If PAREN-LEVEL is non-nil, the function won't stop in nested paren
3161 ;;sexps, and the search will also not go outside the current paren sexp.
3162 ;;However, if LIMIT or the buffer limit is reached inside a nested paren
3163 ;;then the point will be left at the limit.
3164 ;;
3165 ;;Non-nil is returned if the point moved, nil otherwise.
3166 ;;
3167 ;;Note that this function might do hidden buffer changes. See the
3168 ;;comment at the start of cc-engine.el for more info."
3169 ;;
3170 ;; (save-restriction
3171 ;; (when limit
3172 ;; (narrow-to-region limit (point-max)))
3173 ;;
3174 ;; (let ((start (point)))
3175 ;; (catch 'done
3176 ;; (while (let ((last-pos (point))
3177 ;; (stop-pos (progn
3178 ;; (skip-chars-backward skip-chars)
3179 ;; (point))))
3180 ;;
3181 ;; ;; Skip back over the same region as
3182 ;; ;; `skip-chars-backward' above, but keep to
3183 ;; ;; syntactically relevant positions.
3184 ;; (goto-char last-pos)
3185 ;; (while (and
3186 ;; ;; `re-search-backward' with a single char regexp
3187 ;; ;; should be fast.
3188 ;; (re-search-backward
3189 ;; (if paren-level c-ssb-stop-paren-re c-ssb-stop-re)
3190 ;; stop-pos 'move)
3191 ;;
3192 ;; (progn
3193 ;; (cond
3194 ;; ((looking-at "\\s(")
3195 ;; ;; `paren-level' is set and we've found the
3196 ;; ;; start of the containing paren.
3197 ;; (forward-char)
3198 ;; (throw 'done t))
3199 ;;
3200 ;; ((looking-at c-ssb-sexp-end-re)
3201 ;; ;; We're at the end of a string literal or paren
3202 ;; ;; sexp (if `paren-level' is set).
3203 ;; (forward-char)
3204 ;; (condition-case nil
3205 ;; (c-backward-sexp)
3206 ;; (error
3207 ;; (goto-char limit)
3208 ;; (throw 'done t))))
3209 ;;
3210 ;; (t
3211 ;; (forward-char)
3212 ;; ;; At the end of some syntactic ws or possibly
3213 ;; ;; after a plain '/' operator.
3214 ;; (let ((pos (point)))
3215 ;; (c-backward-syntactic-ws)
3216 ;; (if (= pos (point))
3217 ;; ;; Was a plain '/' operator. Go past it.
3218 ;; (backward-char)))))
3219 ;;
3220 ;; (> (point) stop-pos))))
3221 ;;
3222 ;; ;; Now the point is either at `stop-pos' or at some
3223 ;; ;; position further back if `stop-pos' was at a
3224 ;; ;; syntactically irrelevant place.
3225 ;;
3226 ;; ;; Skip additional syntactic ws so that we don't stop
3227 ;; ;; at the end of a comment if `skip-chars' is
3228 ;; ;; something like "^/".
3229 ;; (c-backward-syntactic-ws)
3230 ;;
3231 ;; (< (point) stop-pos))))
3232 ;;
3233 ;; ;; We might want to extend this with more useful return values
3234 ;; ;; in the future.
3235 ;; (/= (point) start))))
3237 \f
3238 ;; Tools for handling comments and string literals.
3241 "Return the type of literal point is in, if any.
3242 The return value is `c' if in a C-style comment, `c++' if in a C++
3243 style comment, `string' if in a string literal, `pound' if DETECT-CPP
3244 is non-nil and in a preprocessor line, or nil if somewhere else.
3245 Optional LIM is used as the backward limit of the search. If omitted,
3246 or nil, `c-beginning-of-defun' is used.
3248 The last point calculated is cached if the cache is enabled, i.e. if
3249 `c-in-literal-cache' is bound to a two element vector.
3251 Note that this function might do hidden buffer changes. See the
3252 comment at the start of cc-engine.el for more info."
3268 ;; cache this result if the cache is enabled
3271 rtn)))
3273 ;; XEmacs has a built-in function that should make this much quicker.
3274 ;; I don't think we even need the cache, which makes our lives more
3275 ;; complicated anyway. In this case, lim is only used to detect
3276 ;; cpp directives.
3277 ;;
3278 ;; Note that there is a bug in Xemacs's buffer-syntactic-context when used in
3279 ;; conjunction with syntax-table-properties. The bug is present in, e.g.,
3280 ;; Xemacs 21.4.4. It manifested itself thus:
3281 ;;
3282 ;; Starting with an empty AWK Mode buffer, type
3283 ;; /regexp/ {<C-j>
3284 ;; Point gets wrongly left at column 0, rather than being indented to tab-width.
3285 ;;
3286 ;; AWK Mode is designed such that when the first / is typed, it gets the
3287 ;; syntax-table property "string fence". When the second / is typed, BOTH /s
3288 ;; are given the s-t property "string". However, buffer-syntactic-context
3289 ;; fails to take account of the change of the s-t property on the opening / to
3290 ;; "string", and reports that the { is within a string started by the second /.
3291 ;;
3292 ;; The workaround for this is for the AWK Mode initialisation to switch the
3293 ;; defalias for c-in-literal to c-slow-in-literal. This will slow down other
3294 ;; cc-modes in Xemacs whenever an awk-buffer has been initialised.
3295 ;;
3296 ;; (Alan Mackenzie, 2003/4/30).
3299 ;; This function might do hidden buffer changes.
3312 ;; The defalias above isn't enough to shut up the byte compiler.
3316 "Return a cons of the beginning and end positions of the comment or
3317 string surrounding point (including both delimiters), or nil if point
3319 to start parsing from. If NEAR is non-nil, then the limits of any
3321 spaces and tabs between point and the literal. The search for such a
3322 literal is done first in forward direction. If NOT-IN-DELIMITER is
3323 non-nil, the case when point is inside a starting delimiter won't be
3324 recognized. This only has effect for comments, which have starting
3325 delimiters with more than one character.
3327 Note that this function might do hidden buffer changes. See the
3328 comment at the start of cc-engine.el for more info."
3350 ;; We're standing in a comment starter.
3357 ;; Search forward for a literal.
3369 ;; Search backward.
3380 ;; Block comment. Due to the nature of line
3381 ;; comments, they will always be covered by the
3382 ;; normal case above.
3385 ;; If LIM is bogus, beg will be bogus.
3389 ))))
3391 ;; In case external callers use this; it did have a docstring.
3395 "If the argument is a cons of two buffer positions (such as returned by
3396 `c-literal-limits'), and that range contains a C++ style line comment,
3397 then an extended range is returned that contains all adjacent line
3398 comments (i.e. all comments that starts in the same column with no
3399 empty lines or non-whitespace characters between them). Otherwise the
3400 argument is returned.
3402 Note that this function might do hidden buffer changes. See the
3403 comment at the start of cc-engine.el for more info."
3414 ;; Got to take care in the backward direction to handle
3415 ;; comments which are preceded by code.
3429 range)
3433 "Convenience function that given the result of `c-literal-limits',
3434 returns nil or the type of literal that the range surrounds. It's
3435 much faster than using `c-in-literal' and is intended to be used when
3436 you need both the type of a literal and its limits.
3438 Note that this function might do hidden buffer changes. See the
3439 comment at the start of cc-engine.el for more info."
3450 range))
3452 \f
3453 ;; `c-find-decl-spots' and accompanying stuff.
3455 ;; Variables used in `c-find-decl-spots' to cache the search done for
3456 ;; the first declaration in the last call. When that function starts,
3457 ;; it needs to back up over syntactic whitespace to look at the last
3458 ;; token before the region being searched. That can sometimes cause
3459 ;; moves back and forth over a quite large region of comments and
3460 ;; macros, which would be repeated for each changed character when
3461 ;; we're called during fontification, since font-lock refontifies the
3462 ;; current line for each change. Thus it's worthwhile to cache the
3463 ;; first match.
3464 ;;
3465 ;; `c-find-decl-syntactic-pos' is a syntactically relevant position in
3466 ;; the syntactic whitespace less or equal to some start position.
3467 ;; There's no cached value if it's nil.
3468 ;;
3469 ;; `c-find-decl-match-pos' is the match position if
3470 ;; `c-find-decl-prefix-search' matched before the syntactic whitespace
3471 ;; at `c-find-decl-syntactic-pos', or nil if there's no such match.
3482 ; (defface c-debug-decl-spot-face
3483 ; '((t (:background "Turquoise")))
3484 ; "Debug face to mark the spots where `c-find-decl-spots' stopped.")
3485 ; (defface c-debug-decl-sws-face
3486 ; '((t (:background "Khaki")))
3487 ; "Debug face to mark the syntactic whitespace between the declaration
3488 ; spots and the preceding token end.")
3504 ;; Macro used inside `c-find-decl-spots'. It ought to be a defun,
3505 ;; but it contains lots of free variables that refer to things
3506 ;; inside `c-find-decl-spots'. The point is left at `cfd-match-pos'
3507 ;; if there is a match, otherwise at `cfd-limit'.
3508 ;;
3509 ;; This macro might do hidden buffer changes.
3512 ;; Find the next property match position if we haven't got one already.
3523 ;; Find the next `c-decl-prefix-or-start-re' match if we haven't
3524 ;; got one already.
3534 ;; Match. Check if it's inside a comment or string literal.
3537 ;; Matched the end of a token preceding a decl spot.
3541 ;; Matched a token that start a decl spot.
3544 c-literal-faces)
3546 ;; No match. Finish up and exit the loop.
3548 nil)
3550 ;; Skip out of comments and string literals.
3557 ;; If we matched at the decl start, we have to back up over the
3558 ;; preceding syntactic ws to set `cfd-match-pos' and to catch
3559 ;; any decl spots in the syntactic ws.
3564 ;; Choose whichever match is closer to the start.
3567 cfd-re-match nil)
3569 cfd-prop-match nil))
3574 ;; Skip forward past comments only so we don't skip macros.
3576 ;; Set the position to continue at. We can avoid going over
3577 ;; the comments skipped above a second time, but it's possible
3578 ;; that the comment skipping has taken us past `cfd-prop-match'
3579 ;; since the property might be used inside comments.
3585 ;; Call CFD-FUN for each possible spot for a declaration, cast or
3586 ;; label from the point to CFD-LIMIT. Such a spot is:
3587 ;;
3588 ;; o The first token after bob.
3589 ;; o The first token after the end of submatch 1 in
3590 ;; `c-decl-prefix-or-start-re' when that submatch matches.
3591 ;; o The start of each `c-decl-prefix-or-start-re' match when
3592 ;; submatch 1 doesn't match.
3593 ;; o The first token after the end of each occurence of the
3594 ;; `c-type' text property with the value `c-decl-end', provided
3595 ;; `c-type-decl-end-used' is set.
3596 ;;
3597 ;; Only a spot that match CFD-DECL-RE and whose face is in the
3598 ;; CFD-FACE-CHECKLIST list causes CFD-FUN to be called. The face
3599 ;; check is disabled if CFD-FACE-CHECKLIST is nil.
3600 ;;
3601 ;; If the match is inside a macro then the buffer is narrowed to the
3602 ;; end of it, so that CFD-FUN can investigate the following tokens
3603 ;; without matching something that begins inside a macro and ends
3604 ;; outside it. It's to avoid this work that the CFD-DECL-RE and
3605 ;; CFD-FACE-CHECKLIST checks exist.
3606 ;;
3607 ;; CFD-FUN is called with point at the start of the spot. It's
3608 ;; passed two arguments: The first is the end position of the token
3609 ;; preceding the spot, or 0 for the implicit match at bob. The
3610 ;; second is a flag that is t when the match is inside a macro. If
3611 ;; CFD-FUN adds `c-decl-end' properties somewhere below the current
3612 ;; spot, it should return non-nil to ensure that the next search
3613 ;; will find them.
3614 ;;
3615 ;; The spots are visited approximately in order from top to bottom.
3616 ;; It's however the positions where `c-decl-prefix-or-start-re'
3617 ;; matches and where `c-decl-end' properties are found that are in
3618 ;; order. Since the spots often are at the following token, they
3619 ;; might be visited out of order insofar as more spots are reported
3620 ;; later on within the syntactic whitespace between the match
3621 ;; positions and their spots.
3622 ;;
3623 ;; It's assumed that comments and strings are fontified in the
3624 ;; searched range.
3625 ;;
3626 ;; This is mainly used in fontification, and so has an elaborate
3627 ;; cache to handle repeated calls from the same start position; see
3628 ;; the variables above.
3629 ;;
3630 ;; All variables in this function begin with `cfd-' to avoid name
3631 ;; collision with the (dynamically bound) variables used in CFD-FUN.
3632 ;;
3633 ;; This function might do hidden buffer changes.
3637 ;; The end of the token preceding the decl spot last found
3638 ;; with `c-decl-prefix-or-start-re'. `cfd-limit' if there's
3639 ;; no match.
3640 cfd-re-match
3641 ;; The end position of the last `c-decl-prefix-or-start-re'
3642 ;; match. If this is greater than `cfd-continue-pos', the
3643 ;; next regexp search is started here instead.
3645 ;; The end of the last `c-decl-end' found by
3646 ;; `c-find-decl-prefix-search'. `cfd-limit' if there's no
3647 ;; match. If searching for the property isn't needed then we
3648 ;; disable it by setting it to `cfd-limit' directly.
3650 ;; The end of the token preceding the decl spot last found by
3651 ;; `c-find-decl-prefix-search'. 0 for the implicit match at
3652 ;; bob. `cfd-limit' if there's no match. In other words,
3653 ;; this is the minimum of `cfd-re-match' and `cfd-prop-match'.
3655 ;; The position to continue searching at.
3656 cfd-continue-pos
3657 ;; The position of the last "real" token we've stopped at.
3658 ;; This can be greater than `cfd-continue-pos' when we get
3659 ;; hits inside macros or at `c-decl-end' positions inside
3660 ;; comments.
3662 ;; The end position of the last entered macro.
3665 ;; Initialize by finding a syntactically relevant start position
3666 ;; before the point, and do the first `c-decl-prefix-or-start-re'
3667 ;; search unless we're at bob.
3670 ;; Must back up a bit since we look for the end of the previous
3671 ;; statement or declaration, which is earlier than the first
3672 ;; returned match.
3675 ;; First we need to move to a syntactically relevant position.
3676 ;; Begin by backing out of comment or string literals.
3679 ;; Try to use the faces to back up to the start of the
3680 ;; literal. FIXME: What if the point is on a declaration
3681 ;; inside a comment?
3687 ;; XEmacs doesn't fontify the quotes surrounding string
3688 ;; literals.
3697 ;; Don't trust the literal to contain only literal faces
3698 ;; (the font lock package might not have fontified the
3699 ;; start of it at all, for instance) so check that we have
3700 ;; arrived at something that looks like a start or else
3701 ;; resort to `c-literal-limits'.
3708 ;; The start is in a literal. If the limit is in the same
3709 ;; one we don't have to find a syntactic position etc. We
3710 ;; only check that if the limit is at or before bonl to save
3711 ;; time; it covers the by far most common case when font-lock
3712 ;; refontifies the current line only.
3723 ;; Completely inside a literal. Set up variables to trig the
3724 ;; (< cfd-continue-pos cfd-start-pos) case below and it'll
3725 ;; find a suitable start position.
3728 ;; Check if the region might be completely inside a macro, to
3729 ;; optimize that like the completely-inside-literal above.
3736 ;; (Maybe) completely inside a macro. Only need to trig the
3737 ;; (< cfd-continue-pos cfd-start-pos) case below to make it
3738 ;; set things up.
3740 start-in-macro t))
3743 ;; Back out of any macro so we don't miss any declaration
3744 ;; that could follow after it.
3748 ;; Now we're at a proper syntactically relevant position so we
3749 ;; can use the cache. But first clear it if it applied
3750 ;; further down.
3755 ;; Don't have to do this if the cache is relevant here,
3756 ;; typically if the same line is refontified again. If
3757 ;; we're just some syntactic whitespace further down we can
3758 ;; still use the cache to limit the skipping.
3761 ;; If we hit `c-find-decl-syntactic-pos' and
3762 ;; `c-find-decl-match-pos' is set then we install the cached
3763 ;; values. If we hit `c-find-decl-syntactic-pos' and
3764 ;; `c-find-decl-match-pos' is nil then we know there's no decl
3765 ;; prefix in the whitespace before `c-find-decl-syntactic-pos'
3766 ;; and so we can continue the search from this point. If we
3767 ;; didn't hit `c-find-decl-syntactic-pos' then we're now in
3768 ;; the right spot to begin searching anyway.
3770 c-find-decl-match-pos)
3772 cfd-continue-pos syntactic-pos)
3777 ;; Always consider bob a match to get the first
3778 ;; declaration in the file. Do this separately instead of
3779 ;; letting `c-decl-prefix-or-start-re' match bob, so that
3780 ;; regexp always can consume at least one character to
3781 ;; ensure that we won't get stuck in an infinite loop.
3786 ;; Do an initial search now. In the bob case above it's
3787 ;; only done to search for a `c-decl-end' spot.
3791 cfd-match-pos)))))
3793 ;; Advance `cfd-continue-pos' if it's before the start position.
3794 ;; The closest continue position that might have effect at or
3795 ;; after the start depends on what we started in. This also
3796 ;; finds a suitable start position in the special cases when the
3797 ;; region is completely within a literal or macro.
3802 ;; If we're in a macro then it's the closest preceding token
3803 ;; in the macro. Check this before `start-in-literal',
3804 ;; since if we're inside a literal in a macro, the preceding
3805 ;; token is earlier than any `c-decl-end' spot inside the
3806 ;; literal (comment).
3808 ;; The only syntactic ws in macros are comments.
3814 ;; If we're in a comment it can only be the closest
3815 ;; preceding `c-decl-end' position within that comment, if
3816 ;; any. Go back to the beginning of such a property so that
3817 ;; `c-find-decl-prefix-search' will find the end of it.
3818 ;; (Can't stop at the end and install it directly on
3819 ;; `cfd-prop-match' since that variable might be cleared
3820 ;; after `cfd-fun' below.)
3821 ;;
3822 ;; Note that if the literal is a string then the property
3823 ;; search will simply skip to the beginning of it right
3824 ;; away.
3836 ;; Didn't find any property inside the comment, so we can
3837 ;; skip it entirely. (This won't skip past a string, but
3838 ;; that'll be handled quickly by the next
3839 ;; `c-find-decl-prefix-search' anyway.)
3845 ;; If we started in normal code, the only match that might
3846 ;; apply before the start is what we already got in
3847 ;; `cfd-match-pos' so we can continue at the start position.
3848 ;; (Note that we don't get here if the first match is below
3849 ;; it.)
3852 ;; Delete found matches if they are before our new continue
3853 ;; position, so that `c-find-decl-prefix-search' won't back up
3854 ;; to them later on.
3862 ;; This is the normal case and we got a proper syntactic
3863 ;; position. If there's a match then it's always outside
3864 ;; macros and comments, so advance to the next token and set
3865 ;; `cfd-token-pos'. The loop below will later go back using
3866 ;; `cfd-continue-pos' to fix declarations inside the
3867 ;; syntactic ws.
3875 ;; Have one of the special cases when the region is completely
3876 ;; within a literal or macro. `cfd-continue-pos' is set to a
3877 ;; good start position for the search, so do it.
3880 ;; Now loop. Round what? (ACM, 2006/7/5). We already got the first match.
3887 ;; Kludge to filter out matches on the "<" that
3888 ;; aren't open parens, for the sake of languages
3889 ;; that got `c-recognize-<>-arglists' set.
3894 ;; If `cfd-continue-pos' is less or equal to
3895 ;; `cfd-token-pos', we've got a hit inside a macro
3896 ;; that's in the syntactic whitespace before the last
3897 ;; "real" declaration we've checked. If they're equal
3898 ;; we've arrived at the declaration a second time, so
3899 ;; there's nothing to do.
3903 ;; If `cfd-continue-pos' is less than `cfd-token-pos'
3904 ;; we're still searching for declarations embedded in
3905 ;; the syntactic whitespace. In that case we need
3906 ;; only to skip comments and not macros, since they
3907 ;; can't be nested, and that's already been done in
3908 ;; `c-find-decl-prefix-search'.
3913 ;; Continue if the following token fails the
3914 ;; CFD-DECL-RE and CFD-FACE-CHECKLIST checks.
3921 t)))
3932 ;; Narrow to the end of the macro if we got a hit inside
3933 ;; one, to avoid recognizing things that start inside the
3934 ;; macro and end outside it.
3936 ;; Not in the same macro as in the previous round.
3947 t
3950 t)
3951 ;; The matched token was the last thing in the macro,
3952 ;; so the whole match is bogus.
3954 nil))))
3961 ;; Restore limits if we did macro narrowment above.
3969 \f
3970 ;; A cache for found types.
3972 ;; Buffer local variable that contains an obarray with the types we've
3973 ;; found. If a declaration is recognized somewhere we record the
3974 ;; fully qualified identifier in it to recognize it as a type
3975 ;; elsewhere in the file too. This is not accurate since we do not
3976 ;; bother with the scoping rules of the languages, but in practice the
3977 ;; same name is seldom used as both a type and something else in a
3978 ;; file, and we only use this as a last resort in ambiguous cases (see
3979 ;; `c-forward-decl-or-cast-1').
3980 ;;
3981 ;; Not every type need be in this cache. However, things which have
3982 ;; ceased to be types must be removed from it.
3983 ;;
3984 ;; Template types in C++ are added here too but with the template
3985 ;; arglist replaced with "<>" in references or "<" for the one in the
3986 ;; primary type. E.g. the type "Foo<A,B>::Bar<C>" is stored as
3987 ;; "Foo<>::Bar<". This avoids storing very long strings (since C++
3988 ;; template specs can be fairly sized programs in themselves) and
3989 ;; improves the hit ratio (it's a type regardless of the template
3990 ;; args; it's just not the same type, but we're only interested in
3991 ;; recognizing types, not telling distinct types apart). Note that
3992 ;; template types in references are added here too; from the example
3993 ;; above there will also be an entry "Foo<".
3998 ;; Clears `c-found-types'.
4002 ;; Add the given region as a type in `c-found-types'. If the region
4003 ;; doesn't match an existing type but there is a type which is equal
4004 ;; to the given one except that the last character is missing, then
4005 ;; the shorter type is removed. That's done to avoid adding all
4006 ;; prefixes of a type as it's being entered and font locked. This
4007 ;; doesn't cover cases like when characters are removed from a type
4008 ;; or added in the middle. We'd need the position of point when the
4009 ;; font locking is invoked to solve this well.
4010 ;;
4011 ;; This function might do hidden buffer changes.
4018 ;; Remove the "NAME" from c-found-types, if present.
4022 ;; Return non-nil if the given region contains a type in
4023 ;; `c-found-types'.
4024 ;;
4025 ;; This function might do hidden buffer changes.
4027 c-found-types))
4030 ;; Return all the types in `c-found-types' as a sorted list of
4031 ;; strings.
4035 type-list)))
4036 c-found-types)
4040 ;; An after change function which, in conjunction with the info in
4041 ;; c-maybe-stale-found-type (set in c-before-change), removes a type
4042 ;; from `c-found-types', should this type have become stale. For
4043 ;; example, this happens to "foo" when "foo \n bar();" becomes
4044 ;; "foo(); \n bar();". Such stale types, if not removed, foul up
4045 ;; the fontification.
4046 ;;
4047 ;; Have we, perhaps, added non-ws characters to the front/back of a found
4048 ;; type?
4068 ;; Changing the amount of (already existing) whitespace - don't do anything.
4071 ; (string-match "\\s *\\'" (nth 5 c-maybe-stale-found-type))
4074 ;; The syntactic relationship which defined a "found type" has been
4075 ;; destroyed.
4078 ;; ((eq (car c-maybe-stale-found-type) 'c-decl-type-start) FIXME!!!
4079 )))
4081 \f
4082 ;; Handling of small scale constructs like types and names.
4085 ;; This is called from `after-change-functions' when
4086 ;; c-recognize-<>-arglists' is set. It ensures that no "<" or ">"
4087 ;; chars with paren syntax become part of another operator like "<<"
4088 ;; or ">=".
4089 ;;
4090 ;; This function might do hidden buffer changes.
4122 ;; Dynamically bound variable that instructs `c-forward-type' to also
4123 ;; treat possible types (i.e. those that it normally returns 'maybe or
4124 ;; 'found for) as actual types (and always return 'found for them).
4125 ;; This means that it records them in `c-record-type-identifiers' if
4126 ;; that is set, and that it adds them to `c-found-types'.
4129 ;; Dynamically bound variable that instructs `c-forward-<>-arglist' to
4130 ;; mark up successfully parsed arglists with paren syntax properties on
4131 ;; the surrounding angle brackets and with `c-<>-arg-sep' in the
4132 ;; `c-type' property of each argument separating comma.
4133 ;;
4134 ;; Setting this variable also makes `c-forward-<>-arglist' recurse into
4135 ;; all arglists for side effects (i.e. recording types), otherwise it
4136 ;; exploits any existing paren syntax properties to quickly jump to the
4137 ;; end of already parsed arglists.
4138 ;;
4139 ;; Marking up the arglists is not the default since doing that correctly
4140 ;; depends on a proper value for `c-restricted-<>-arglists'.
4143 ;; Dynamically bound variable that instructs `c-forward-<>-arglist' to
4144 ;; not accept arglists that contain binary operators.
4145 ;;
4146 ;; This is primarily used to handle C++ template arglists. C++
4147 ;; disambiguates them by checking whether the preceding name is a
4148 ;; template or not. We can't do that, so we assume it is a template
4149 ;; if it can be parsed as one. That usually works well since
4150 ;; comparison expressions on the forms "a < b > c" or "a < b, c > d"
4151 ;; in almost all cases would be pointless.
4152 ;;
4153 ;; However, in function arglists, e.g. in "foo (a < b, c > d)", we
4154 ;; should let the comma separate the function arguments instead. And
4155 ;; in a context where the value of the expression is taken, e.g. in
4156 ;; "if (a < b || c > d)", it's probably not a template.
4159 ;; Dynamically bound variables that instructs
4160 ;; `c-forward-keyword-clause', `c-forward-<>-arglist',
4161 ;; `c-forward-name', `c-forward-type', `c-forward-decl-or-cast-1', and
4162 ;; `c-forward-label' to record the ranges of all the type and
4163 ;; reference identifiers they encounter. They will build lists on
4164 ;; these variables where each element is a cons of the buffer
4165 ;; positions surrounding each identifier. This recording is only
4166 ;; activated when `c-record-type-identifiers' is non-nil.
4167 ;;
4168 ;; All known types that can't be identifiers are recorded, and also
4169 ;; other possible types if `c-promote-possible-types' is set.
4170 ;; Recording is however disabled inside angle bracket arglists that
4171 ;; are encountered inside names and other angle bracket arglists.
4172 ;; Such occurrences are taken care of by `c-font-lock-<>-arglists'
4173 ;; instead.
4174 ;;
4175 ;; Only the names in C++ template style references (e.g. "tmpl" in
4176 ;; "tmpl<a,b>::foo") are recorded as references, other references
4177 ;; aren't handled here.
4178 ;;
4179 ;; `c-forward-label' records the label identifier(s) on
4180 ;; `c-record-ref-identifiers'.
4184 ;; This variable will receive a cons cell of the range of the last
4185 ;; single identifier symbol stepped over by `c-forward-name' if it's
4186 ;; successful. This is the range that should be put on one of the
4187 ;; record lists above by the caller. It's assigned nil if there's no
4188 ;; such symbol in the name.
4193 ;; Always true.
4203 ;; Always true.
4211 ;; Dynamically bound variable that instructs `c-forward-type' to
4212 ;; record the ranges of types that only are found. Behaves otherwise
4213 ;; like `c-record-type-identifiers'.
4217 ;; Used internally in `c-forward-keyword-clause' to move forward
4218 ;; over a type (if TYPE is 'type) or a name (otherwise) which
4219 ;; possibly is prefixed by keywords and their associated clauses.
4220 ;; Try with a type/name first to not trip up on those that begin
4221 ;; with a keyword. Return t if a known or found type is moved
4222 ;; over. The point is clobbered if nil is returned. If range
4223 ;; recording is enabled, the identifier is recorded on as a type
4224 ;; if TYPE is 'type or as a reference if TYPE is 'ref.
4225 ;;
4226 ;; This macro might do hidden buffer changes.
4231 nil
4238 t)))
4241 ;; Used internally in `c-forward-keyword-clause' to move forward
4242 ;; over a comma separated list of types or names using
4243 ;; `c-forward-keyword-prefixed-id'.
4244 ;;
4245 ;; This macro might do hidden buffer changes.
4256 ;; Submatch MATCH in the current match data is assumed to surround a
4257 ;; token. If it's a keyword, move over it and any immediately
4258 ;; following clauses associated with it, stopping at the start of
4259 ;; the next token. t is returned in that case, otherwise the point
4260 ;; stays and nil is returned. The kind of clauses that are
4261 ;; recognized are those specified by `c-type-list-kwds',
4262 ;; `c-ref-list-kwds', `c-colon-type-list-kwds',
4263 ;; `c-paren-nontype-kwds', `c-paren-type-kwds', `c-<>-type-kwds',
4264 ;; and `c-<>-arglist-kwds'.
4265 ;;
4266 ;; This function records identifier ranges on
4267 ;; `c-record-type-identifiers' and `c-record-ref-identifiers' if
4268 ;; `c-record-type-identifiers' is non-nil.
4269 ;;
4270 ;; Note that for `c-colon-type-list-kwds', which doesn't necessary
4271 ;; apply directly after the keyword, the type list is moved over
4272 ;; only when there is no unaccounted token before it (i.e. a token
4273 ;; that isn't moved over due to some other keyword list). The
4274 ;; identifier ranges in the list are still recorded if that should
4275 ;; be done, though.
4276 ;;
4277 ;; This function might do hidden buffer changes.
4280 ;; The call to `c-forward-<>-arglist' below is made after
4281 ;; `c-<>-sexp-kwds' keywords, so we're certain they actually
4282 ;; are angle bracket arglists and `c-restricted-<>-arglists'
4283 ;; should therefore be nil.
4295 ;; There's a type directly after a keyword in `c-type-list-kwds'.
4300 ;; There's a name directly after a keyword in `c-ref-list-kwds'.
4305 ;; There's an open paren after a keyword in `c-paren-any-kwds'.
4312 ;; Use `c-forward-type' on every identifier we can find
4313 ;; inside the paren, to record the types.
4338 ;; If we are at the colon already, we move over the type
4339 ;; list after it.
4345 ;; Not at the colon, so stop here. But the identifier
4346 ;; ranges in the type list later on should still be
4347 ;; recorded.
4350 ;; If a keyword matched both one of the types above and
4351 ;; this one, we match `c-colon-type-list-re' after the
4352 ;; clause matched above.
4359 ;; There's a type after the `c-colon-type-list-re' match
4360 ;; after a keyword in `c-colon-type-list-kwds'.
4364 t)))
4367 ;; The point is assumed to be at a "<". Try to treat it as the open
4368 ;; paren of an angle bracket arglist and move forward to the the
4369 ;; corresponding ">". If successful, the point is left after the
4370 ;; ">" and t is returned, otherwise the point isn't moved and nil is
4371 ;; returned. If ALL-TYPES is t then all encountered arguments in
4372 ;; the arglist that might be types are treated as found types.
4373 ;;
4374 ;; The variable `c-parse-and-markup-<>-arglists' controls how this
4375 ;; function handles text properties on the angle brackets and argument
4376 ;; separating commas.
4377 ;;
4378 ;; `c-restricted-<>-arglists' controls how lenient the template
4379 ;; arglist recognition should be.
4380 ;;
4381 ;; This function records identifier ranges on
4382 ;; `c-record-type-identifiers' and `c-record-ref-identifiers' if
4383 ;; `c-record-type-identifiers' is non-nil.
4384 ;;
4385 ;; This function might do hidden buffer changes.
4388 ;; If `c-record-type-identifiers' is set then activate
4389 ;; recording of any found types that constitute an argument in
4390 ;; the arglist.
4398 ;; `nconc' doesn't mind that the tail of
4399 ;; `c-record-found-types' is t.
4401 t)
4404 nil)))
4407 ;; Recursive part of `c-forward-<>-arglist'.
4408 ;;
4409 ;; This function might do hidden buffer changes.
4412 ;; Cover this so that any recorded found type ranges are
4413 ;; automatically lost if it turns out to not be an angle
4414 ;; bracket arglist. It's propagated through the return value
4415 ;; on successful completion.
4417 ;; List that collects the positions after the argument
4418 ;; separating ',' in the arglist.
4419 arg-start-pos)
4421 ;; If the '<' has paren open syntax then we've marked it as an angle
4422 ;; bracket arglist before, so skip to the end.
4430 t
4432 ;; Got unmatched paren angle brackets. We don't clear the paren
4433 ;; syntax properties and retry, on the basis that it's very
4434 ;; unlikely that paren angle brackets become operators by code
4435 ;; manipulation. It's far more likely that it doesn't match due
4436 ;; to narrowing or some temporary change.
4438 nil))
4448 ;; All encountered identifiers are types, so set the
4449 ;; promote flag and parse the type.
4456 ;; Check if this arglist argument is a sole type. If
4457 ;; it's known then it's recorded in
4458 ;; `c-record-type-identifiers'. If it only is found
4459 ;; then it's recorded in `c-record-found-types' which we
4460 ;; might roll back if it turns out that this isn't an
4461 ;; angle bracket arglist afterall.
4467 ;; A found type was recorded but it's not the
4468 ;; only thing in the arglist argument, so reset
4469 ;; `c-record-found-types'.
4471 orig-record-found-types))))))
4475 ;; Must check for '>' at the very start separately,
4476 ;; since the regexp below has to avoid ">>" without
4477 ;; using \\=.
4479 t)
4481 ;; Note: These regexps exploit the match order in \| so
4482 ;; that "<>" is matched by "<" rather than "[^>:-]>".
4485 ;; Stop on ',', '|', '&', '+' and '-' to catch
4486 ;; common binary operators that could be between
4487 ;; two comparison expressions "a<b" and "c>d".
4489 ;; Otherwise we still stop on ',' to find the
4490 ;; argument start positions.
4494 ;; If the arglist starter has lost its open paren
4495 ;; syntax but not the closer, we won't find the
4496 ;; closer above since we only search in the
4497 ;; balanced sexp. In that case we stop just short
4498 ;; of it so check if the following char is the closer.
4501 t)))
4505 ;; Either an operator starting with '>' or the end of
4506 ;; the angle bracket arglist.
4513 ;; The angle bracket arglist is finished.
4525 ;; Either an operator starting with '<' or a nested arglist.
4537 ;; There's always an identifier before an angle
4538 ;; bracket arglist, or a keyword in
4539 ;; `c-<>-type-kwds' or `c-<>-arglist-kwds'.
4556 )))
4558 ;; It was not an angle bracket arglist.
4561 ;; It was an angle bracket arglist.
4564 ;; Record the identifier before the template as a type
4565 ;; or reference depending on whether the arglist is last
4566 ;; in a qualified identifier.
4575 t)
4579 ;; Just another argument. Record the position. The
4580 ;; type check stuff that made us stop at it is at
4581 ;; the top of the loop.
4585 ;; Got a character that can't be in an angle bracket
4586 ;; arglist argument. Abort using `throw', since
4587 ;; it's useless to try to find a surrounding arglist
4588 ;; if we're nested.
4595 ;; The point is assumed to be directly after a ">". Try to treat it
4596 ;; as the close paren of an angle bracket arglist and move back to
4597 ;; the corresponding "<". If successful, the point is left at
4598 ;; the "<" and t is returned, otherwise the point isn't moved and
4599 ;; nil is returned. ALL-TYPES is passed on to
4600 ;; `c-forward-<>-arglist'.
4601 ;;
4602 ;; If the optional LIMIT is given, it bounds the backward search.
4603 ;; It's then assumed to be at a syntactically relevant position.
4604 ;;
4605 ;; This is a wrapper around `c-forward-<>-arglist'. See that
4606 ;; function for more details.
4615 t
4616 ;; See corresponding note in `c-forward-<>-arglist'.
4618 nil)
4625 t
4626 ;; Stopped at bob or a char that isn't allowed in an
4627 ;; arglist, so we've failed.
4629 nil)
4634 ;; If we moved then the "<" was part of some
4635 ;; multicharacter token.
4636 t
4642 ;; Matched the arglist. Break the while.
4644 nil)
4646 ;; We started from a non-paren ">" inside an
4647 ;; arglist.
4649 nil)
4651 ;; Matched a shorter arglist. Can be a nested
4652 ;; one so continue looking.
4654 t))
4655 t))))))
4660 ;; Move forward over a complete name if at the beginning of one,
4661 ;; stopping at the next following token. If the point is not at
4662 ;; something that are recognized as name then it stays put. A name
4663 ;; could be something as simple as "foo" in C or something as
4664 ;; complex as "X<Y<class A<int>::B, BIT_MAX >> b>, ::operator<> ::
4665 ;; Z<(a>b)> :: operator const X<&foo>::T Q::G<unsigned short
4666 ;; int>::*volatile const" in C++ (this function is actually little
4667 ;; more than a `looking-at' call in all modes except those that,
4668 ;; like C++, have `c-recognize-<>-arglists' set). Return nil if no
4669 ;; name is found, 'template if it's an identifier ending with an
4670 ;; angle bracket arglist, 'operator of it's an operator identifier,
4671 ;; or t if it's some other kind of name.
4672 ;;
4673 ;; This function records identifier ranges on
4674 ;; `c-record-type-identifiers' and `c-record-ref-identifiers' if
4675 ;; `c-record-type-identifiers' is non-nil.
4676 ;;
4677 ;; This function might do hidden buffer changes.
4680 ;; Turn off `c-promote-possible-types' here since we might
4681 ;; call `c-forward-<>-arglist' and we don't want it to promote
4682 ;; every suspect thing in the arglist to a type. We're
4683 ;; typically called from `c-forward-type' in this case, and
4684 ;; the caller only wants the top level type that it finds to
4685 ;; be promoted.
4686 c-promote-possible-types)
4692 ;; Check for keyword. We go to the last symbol in
4693 ;; `c-identifier-key' first.
4706 ;; "template" is only valid inside an
4707 ;; identifier if preceded by "::".
4712 t))
4714 ;; Handle a C++ operator or template identifier.
4718 ;; Got "... ::template".
4722 res subres))))
4725 ;; Got a cast operator.
4729 ;; Now we should match a sequence of either
4730 ;; '*', '&' or a name followed by ":: *",
4731 ;; where each can be followed by a sequence
4732 ;; of `c-opt-type-modifier-key'.
4735 t)
4745 t))))
4753 ;; Got some other operator.
4761 nil)
4763 ;; `id-start' is equal to `id-end' if we've jumped over
4764 ;; an identifier that doesn't end with a symbol token.
4765 ;; That can occur e.g. for Java import directives on the
4766 ;; form "foo.bar.*".
4773 res t)))
4783 ;; Got a concatenated identifier. This handles the
4784 ;; cases with tricky syntactic whitespace that aren't
4785 ;; covered in `c-identifier-key'.
4788 t)
4792 ;; Maybe an angle bracket arglist.
4795 c-record-found-types)
4801 c-last-identifier-range nil)
4806 ;; Continue if there's an identifier concatenation
4807 ;; operator after the template argument.
4813 t)
4818 nil)))
4819 )))))
4822 res))
4825 ;; Move forward over a type spec if at the beginning of one,
4826 ;; stopping at the next following token. Return t if it's a known
4827 ;; type that can't be a name or other expression, 'known if it's an
4828 ;; otherwise known type (according to `*-font-lock-extra-types'),
4829 ;; 'prefix if it's a known prefix of a type, 'found if it's a type
4830 ;; that matches one in `c-found-types', 'maybe if it's an identfier
4831 ;; that might be a type, or nil if it can't be a type (the point
4832 ;; isn't moved then). The point is assumed to be at the beginning
4833 ;; of a token.
4834 ;;
4835 ;; Note that this function doesn't skip past the brace definition
4836 ;; that might be considered part of the type, e.g.
4837 ;; "enum {a, b, c} foo".
4838 ;;
4839 ;; This function records identifier ranges on
4840 ;; `c-record-type-identifiers' and `c-record-ref-identifiers' if
4841 ;; `c-record-type-identifiers' is non-nil.
4842 ;;
4843 ;; This function might do hidden buffer changes.
4847 ;; Skip leading type modifiers. If any are found we know it's a
4848 ;; prefix of a type.
4857 ;; Looking at a keyword that prefixes a type identifier,
4858 ;; e.g. "class".
4865 ;; In many languages the name can be used without the
4866 ;; prefix, so we add it to `c-found-types'.
4869 c-last-identifier-range)
4872 ;; Invalid syntax.
4884 id-range c-last-identifier-range))))
4896 id-end)
4898 ;; Looking at a primitive or known type identifier. We've
4899 ;; checked for a name first so that we don't go here if the
4900 ;; known type match only is a prefix of another name.
4911 ;; There might be more keywords for the type.
4939 ;; A normal identifier.
4949 ;; It's an identifier that has been used as
4950 ;; a type somewhere else.
4951 'found
4952 ;; It's an identifier that might be a type.
4955 ;; A template is a type.
4959 ;; Otherwise it's an operator identifier, which is not a type.
4964 ;; Skip trailing type modifiers. If any are found we know it's
4965 ;; a type.
4972 ;; Step over any type suffix operator. Do not let the existence
4973 ;; of these alter the classification of the found type, since
4974 ;; these operators typically are allowed in normal expressions
4975 ;; too.
4982 ;; Look for a trailing operator that concatenates the type
4983 ;; with a following one, and if so step past that one through
4984 ;; a recursive call. Note that we don't record concatenated
4985 ;; types in `c-found-types' - it's the component types that
4986 ;; are recorded when appropriate.
4989 c-promote-possible-types))
4990 ;; If we can't promote then set `c-record-found-types' so that
4991 ;; we can merge in the types from the second part afterwards if
4992 ;; it turns out to be a known type there.
4995 subres)
5004 ;; If either operand certainly is a type then both are, but we
5005 ;; don't let the existence of the operator itself promote two
5006 ;; uncertain types to a certain one.
5025 ;; Merge in the ranges of any types found by the second
5026 ;; `c-forward-type'.
5028 ;; `nconc' doesn't mind that the tail of
5029 ;; `c-record-found-types' is t.
5031 c-record-type-identifiers))))
5039 ;;(message "c-forward-type %s -> %s: %s" start (point) res)
5041 res))
5043 \f
5044 ;; Handling of large scale constructs like statements and declarations.
5046 ;; Macro used inside `c-forward-decl-or-cast-1'. It ought to be a
5047 ;; defsubst or perhaps even a defun, but it contains lots of free
5048 ;; variables that refer to things inside `c-forward-decl-or-cast-1'.
5050 ;; `c-forward-decl-or-cast-1' can consume an arbitrary length list
5051 ;; of types when parsing a declaration, which means that it
5052 ;; sometimes consumes the identifier in the declaration as a type.
5053 ;; This is used to "backtrack" and make the last type be treated as
5054 ;; an identifier instead.
5057 ;; These identifiers are bound only in the inner let.
5059 identifier-start type-start
5060 got-parens nil
5061 got-identifier t
5062 got-suffix t
5063 got-suffix-after-parens id-start
5067 t
5068 backup-at-type))
5070 id-start backup-id-start)
5072 id-start start-pos))
5074 ;; When these flags already are set we've found specifiers that
5075 ;; unconditionally signal these attributes - backtracking doesn't
5076 ;; change that. So keep them set in that case.
5083 ;; This identifier is bound only in the inner let.
5087 ;; Move forward over a declaration or a cast if at the start of one.
5088 ;; The point is assumed to be at the start of some token. Nil is
5089 ;; returned if no declaration or cast is recognized, and the point
5090 ;; is clobbered in that case.
5091 ;;
5092 ;; If a declaration is parsed:
5093 ;;
5094 ;; The point is left at the first token after the first complete
5095 ;; declarator, if there is one. The return value is a cons where
5096 ;; the car is the position of the first token in the declarator.
5097 ;; Some examples:
5098 ;;
5099 ;; void foo (int a, char *b) stuff ...
5100 ;; car ^ ^ point
5101 ;; float (*a)[], b;
5102 ;; car ^ ^ point
5103 ;; unsigned int a = c_style_initializer, b;
5104 ;; car ^ ^ point
5105 ;; unsigned int a (cplusplus_style_initializer), b;
5106 ;; car ^ ^ point (might change)
5107 ;; class Foo : public Bar {}
5108 ;; car ^ ^ point
5109 ;; class PikeClass (int a, string b) stuff ...
5110 ;; car ^ ^ point
5111 ;; enum bool;
5112 ;; car ^ ^ point
5113 ;; enum bool flag;
5114 ;; car ^ ^ point
5115 ;; void cplusplus_function (int x) throw (Bad);
5116 ;; car ^ ^ point
5117 ;; Foo::Foo (int b) : Base (b) {}
5118 ;; car ^ ^ point
5119 ;;
5120 ;; The cdr of the return value is non-nil iff a
5121 ;; `c-typedef-decl-kwds' specifier is found in the declaration,
5122 ;; i.e. the declared identifier(s) are types.
5123 ;;
5124 ;; If a cast is parsed:
5125 ;;
5126 ;; The point is left at the first token after the closing paren of
5127 ;; the cast. The return value is `cast'. Note that the start
5128 ;; position must be at the first token inside the cast parenthesis
5129 ;; to recognize it.
5130 ;;
5131 ;; PRECEDING-TOKEN-END is the first position after the preceding
5132 ;; token, i.e. on the other side of the syntactic ws from the point.
5133 ;; Use a value less than or equal to (point-min) if the point is at
5134 ;; the first token in (the visible part of) the buffer.
5135 ;;
5136 ;; CONTEXT is a symbol that describes the context at the point:
5137 ;; 'decl In a comma-separatded declaration context (typically
5138 ;; inside a function declaration arglist).
5139 ;; '<> In an angle bracket arglist.
5140 ;; 'arglist Some other type of arglist.
5141 ;; nil Some other context or unknown context.
5142 ;;
5143 ;; LAST-CAST-END is the first token after the closing paren of a
5144 ;; preceding cast, or nil if none is known. If
5145 ;; `c-forward-decl-or-cast-1' is used in succession, it should be
5146 ;; the position after the closest preceding call where a cast was
5147 ;; matched. In that case it's used to discover chains of casts like
5148 ;; "(a) (b) c".
5149 ;;
5150 ;; This function records identifier ranges on
5151 ;; `c-record-type-identifiers' and `c-record-ref-identifiers' if
5152 ;; `c-record-type-identifiers' is non-nil.
5153 ;;
5154 ;; This function might do hidden buffer changes.
5157 ;; first type, i.e. after any leading specifiers. It might
5158 ;; also point at the beginning of the preceding syntactic
5159 ;; whitespace.
5161 ;; Set to the result of `c-forward-type'.
5162 at-type
5163 ;; The position of the first token in what we currently
5164 ;; believe is the type in the declaration or cast, after any
5165 ;; specifiers and their associated clauses.
5166 type-start
5167 ;; The position of the first token in what we currently
5168 ;; believe is the declarator for the first identifier. Set
5169 ;; when the type is found, and moved forward over any
5170 ;; `c-decl-hangon-kwds' and their associated clauses that
5171 ;; occurs after the type.
5172 id-start
5173 ;; These store `at-type', `type-start' and `id-start' of the
5174 ;; identifier before the one in those variables. The previous
5175 ;; identifier might turn out to be the real type in a
5176 ;; declaration if the last one has to be the declarator in it.
5177 ;; If `backup-at-type' is nil then the other variables have
5178 ;; undefined values.
5179 backup-at-type backup-type-start backup-id-start
5180 ;; Set if we've found a specifier that makes the defined
5181 ;; identifier(s) types.
5182 at-type-decl
5183 ;; Set if we've found a specifier that can start a declaration
5184 ;; where there's no type.
5185 maybe-typeless
5186 ;; If a specifier is found that also can be a type prefix,
5187 ;; these flags are set instead of those above. If we need to
5188 ;; back up an identifier, they are copied to the real flag
5189 ;; variables. Thus they only take effect if we fail to
5190 ;; interpret it as a type.
5191 backup-at-type-decl backup-maybe-typeless
5192 ;; Whether we've found a declaration or a cast. We might know
5193 ;; this before we've found the type in it. It's 'ids if we've
5194 ;; found two consecutive identifiers (usually a sure sign, but
5195 ;; we should allow that in labels too), and t if we've found a
5196 ;; specifier keyword (a 100% sure sign).
5197 at-decl-or-cast
5198 ;; Set when we need to back up to parse this as a declaration
5199 ;; but not as a cast.
5200 backup-if-not-cast
5201 ;; For casts, the return position.
5202 cast-end
5203 ;; Save `c-record-type-identifiers' and
5204 ;; `c-record-ref-identifiers' since ranges are recorded
5205 ;; speculatively and should be thrown away if it turns out
5206 ;; that it isn't a declaration or cast.
5210 ;; Check for a type. Unknown symbols are treated as possible
5211 ;; types, but they could also be specifiers disguised through
5212 ;; macros like __INLINE__, so we recognize both types and known
5213 ;; specifiers after them too.
5217 ;; Look for a specifier keyword clause.
5225 ;; Found a known or possible type or a prefix of a known type.
5228 ;; Got two identifiers with nothing but whitespace
5229 ;; between them. That can only happen in declarations.
5233 ;; If the previous identifier is a found type we
5234 ;; record it as a real one; it might be some sort of
5235 ;; alias for a prefix like "unsigned".
5242 backup-type-start type-start
5243 backup-id-start id-start
5244 at-type found-type
5245 type-start start
5247 ;; The previous ambiguous specifier/type turned out
5248 ;; to be a type since we've parsed another one after
5249 ;; it, so clear these backup flags.
5250 backup-at-type-decl nil
5251 backup-maybe-typeless nil))
5255 ;; Handle known specifier keywords and
5256 ;; `c-decl-hangon-kwds' which can occur after known
5257 ;; types.
5260 ;; It's a hang-on keyword that can occur anywhere.
5264 ;; Move the identifier start position if
5265 ;; we've passed a type.
5267 ;; Otherwise treat this as a specifier and
5268 ;; move the fallback position.
5272 ;; It's an ordinary specifier so we know that
5273 ;; anything before this can't be the type.
5275 start-pos kwd-clause-end)
5278 ;; It's ambiguous whether this keyword is a
5279 ;; specifier or a type prefix, so set the backup
5280 ;; flags. (It's assumed that `c-forward-type'
5281 ;; moved further than `c-forward-keyword-clause'.)
5293 ;; Haven't matched a type so it's an umambiguous
5294 ;; specifier keyword and we know we're in a
5295 ;; declaration.
5300 ;; If the type isn't known we continue so that we'll jump
5301 ;; over all specifiers and type identifiers. The reason
5302 ;; to do this for a known type prefix is to make things
5303 ;; like "unsigned INT16" work.
5308 ;; If a known type was found, we still need to skip over any
5309 ;; hangon keyword clauses after it. Otherwise it has already
5310 ;; been done in the loop above.
5316 ;; A prefix type is itself a primitive type when it's not
5317 ;; followed by another type.
5321 ;; Got no type but set things up to continue anyway to handle
5322 ;; the various cases when a declaration doesn't start with a
5323 ;; type.
5328 ;; If it's C++ then check if the last "type" ends on the form
5329 ;; "foo::foo" or "foo::~foo", i.e. if it's the name of a
5330 ;; (con|de)structor.
5341 ;; Cheating in the handling of syntactic ws below.
5348 name))
5349 ;; It is a (con|de)structor name. In that case the
5350 ;; declaration is typeless so zap out any preceding
5351 ;; identifier(s) that we might have taken as types.
5354 backup-at-type nil
5355 id-start type-start))))))
5357 ;; Check for and step over a type decl expression after the thing
5358 ;; that is or might be a type. This can't be skipped since we
5359 ;; need the correct end position of the declarator for
5360 ;; `max-type-decl-end-*'.
5362 ;; True if there's a non-open-paren match of
5363 ;; `c-type-decl-prefix-key'.
5364 got-prefix
5365 ;; True if the declarator is surrounded by a parenthesis pair.
5366 got-parens
5367 ;; True if there is an identifier in the declarator.
5368 got-identifier
5369 ;; True if there's a non-close-paren match of
5370 ;; `c-type-decl-suffix-key'.
5371 got-suffix
5372 ;; True if there's a prefix match outside the outermost
5373 ;; paren pair that surrounds the declarator.
5374 got-prefix-before-parens
5375 y ;; True if there's a suffix match outside the outermost
5376 ;; paren pair that surrounds the declarator. The value is
5377 ;; the position of the first suffix match.
5378 got-suffix-after-parens
5379 ;; True if we've parsed the type decl to a token that is
5380 ;; known to end declarations in this context.
5381 at-decl-end
5382 ;; The earlier values of `at-type' and `type-start' if we've
5383 ;; shifted the type backwards.
5384 identifier-type identifier-start
5385 ;; If `c-parse-and-markup-<>-arglists' is set we need to
5386 ;; turn it off during the name skipping below to avoid
5387 ;; getting `c-type' properties that might be bogus. That
5388 ;; can happen since we don't know if
5389 ;; `c-restricted-<>-arglists' will be correct inside the
5390 ;; arglist paren that gets entered.
5395 ;; Skip over type decl prefix operators. (Note similar code in
5396 ;; `c-font-lock-declarators'.)
5400 ;; If the second submatch matches in C++ then
5401 ;; we're looking at an identifier that's a
5402 ;; prefix only if it specifies a member pointer.
5405 ;; We only check for a trailing "::" and
5406 ;; let the "*" that should follow be
5407 ;; matched in the next round.
5409 ;; It turned out to be the real identifier,
5410 ;; so stop.
5411 nil))
5412 t))
5426 ;; Skip over an identifier.
5431 ;; Skip over type decl suffix operators.
5438 t)
5442 t)
5448 ;; No suffix matched. We might have matched the
5449 ;; identifier as a type and the open paren of a
5450 ;; function arglist as a type decl prefix. In that
5451 ;; case we should "backtrack": Reinterpret the last
5452 ;; type as the identifier, move out of the arglist and
5453 ;; continue searching for suffix operators.
5454 ;;
5455 ;; Do this even if there's no preceding type, to cope
5456 ;; with old style function declarations in K&R C,
5457 ;; (con|de)structors in C++ and `c-typeless-decl-kwds'
5458 ;; style declarations. That isn't applicable in an
5459 ;; arglist context, though.
5464 maybe-typeless
5465 backup-maybe-typeless
5472 t))
5479 at-type)
5480 ;; Have found no identifier but `c-typeless-decl-kwds' has
5481 ;; matched so we know we're inside a declaration. The
5482 ;; preceding type must be the identifier instead.
5486 at-decl-or-cast
5490 ;; Encountered something inside parens that isn't matched by
5491 ;; the `c-type-decl-*' regexps, so it's not a type decl
5492 ;; expression. Try to skip out to the same paren depth to
5493 ;; not confuse the cast check below.
5495 ;; If we've found a specifier keyword then it's a
5496 ;; declaration regardless.
5504 ;; Now we've collected info about various characteristics of
5505 ;; the construct we're looking at. Below follows a decision
5506 ;; tree based on that. It's ordered to check more certain
5507 ;; signs before less certain ones.
5514 ;; Got another identifier directly after the type, so it's a
5515 ;; declaration.
5522 maybe-typeless
5523 backup-maybe-typeless))
5524 ;; Got a declaration of the form "foo bar (gnu);" where we've
5525 ;; recognized "bar" as the type and "gnu" as the declarator.
5526 ;; In this case it's however more likely that "bar" is the
5527 ;; declarator and "gnu" a function argument or initializer (if
5528 ;; `c-recognize-paren-inits' is set), since the parens around
5529 ;; "gnu" would be superfluous if it's a declarator. Shift the
5530 ;; type one step backward.
5533 ;; Found no identifier.
5539 ;; Got a plain list of identifiers. If a colon follows it's
5540 ;; a valid label. Otherwise the last one probably is the
5541 ;; declared identifier and we should back up to the previous
5542 ;; type, providing it isn't a cast.
5544 ;; If we've found a specifier keyword then it's a
5545 ;; declaration regardless.
5553 ;; Got a plain list of identifiers followed by some suffix.
5554 ;; If this isn't a cast then the last identifier probably is
5555 ;; the declared one and we should back up to the previous
5556 ;; type.
5561 ;; If the type is known we know that there can't be any
5562 ;; identifier somewhere else, and it's only in declarations in
5563 ;; e.g. function prototypes and in casts that the identifier may
5564 ;; be left out.
5568 ;; Only got a single identifier (parsed as a type so far).
5570 ;; Check that the identifier isn't at the start of an
5571 ;; expression.
5572 at-decl-end
5575 ;; Inside an arglist that contains declarations. If K&R
5576 ;; style declarations and parenthesis style initializers
5577 ;; aren't allowed then the single identifier must be a
5578 ;; type, else we require that it's known or found
5579 ;; (primitive types are handled above).
5584 ;; Inside a template arglist. Accept known and found
5585 ;; types; other identifiers could just as well be
5586 ;; constants in C++.
5589 ;; Can't be a valid declaration or cast, but if we've found a
5590 ;; specifier it can't be anything else either, so treat it as
5591 ;; an invalid/unfinished declaration or cast.
5599 maybe-typeless
5600 backup-maybe-typeless
5604 c-after-suffixed-type-maybe-decl-key))))))
5605 ;; Got an empty paren pair and a preceding type that probably
5606 ;; really is the identifier. Shift the type backwards to make
5607 ;; the last one the identifier. This is analogous to the
5608 ;; "backtracking" done inside the `c-type-decl-suffix-key' loop
5609 ;; above.
5610 ;;
5611 ;; Exception: In addition to the conditions in that
5612 ;; "backtracking" code, do not shift backward if we're not
5613 ;; looking at either `c-after-suffixed-type-decl-key' or "[;,]".
5614 ;; Since there's no preceding type, the shift would mean that
5615 ;; the declaration is typeless. But if the regexp doesn't match
5616 ;; then we will simply fall through in the tests below and not
5617 ;; recognize it at all, so it's better to try it as an abstract
5618 ;; declarator instead.
5621 ;; Still no identifier.
5624 ;; Require `got-prefix' together with either `got-parens' or
5625 ;; `got-suffix' to recognize it as an abstract declarator:
5626 ;; `got-parens' only is probably an empty function call.
5627 ;; `got-suffix' only can build an ordinary expression together
5628 ;; with the preceding identifier which we've taken as a type.
5629 ;; We could actually accept on `got-prefix' only, but that can
5630 ;; easily occur temporarily while writing an expression so we
5631 ;; avoid that case anyway. We could do a better job if we knew
5632 ;; the point when the fontification was invoked.
5638 got-suffix-after-parens
5640 ;; Got a type, no declarator but a paren suffix. I.e. it's a
5641 ;; normal function call afterall (or perhaps a C++ style object
5642 ;; instantiation expression).
5646 ;; By now we've located the type in the declaration that we know
5647 ;; we're in.
5657 ;; Shift the type backward in the case that there's a
5658 ;; single identifier inside parens. That can only
5659 ;; occur in K&R style function declarations so it's
5660 ;; more likely that it really is a function call.
5661 ;; Therefore we only do this after
5662 ;; `c-after-suffixed-type-decl-key' has matched.
5664 got-suffix-after-parens))
5665 ;; A declaration according to `c-after-suffixed-type-decl-key'.
5670 ;; It's a declaration if a known type precedes it and it can't be a
5671 ;; function call.
5674 ;; If we get here we can't tell if this is a type decl or a normal
5675 ;; expression by looking at it alone. (That's under the assumption
5676 ;; that normal expressions always can look like type decl expressions,
5677 ;; which isn't really true but the cases where it doesn't hold are so
5678 ;; uncommon (e.g. some placements of "const" in C++) it's not worth
5679 ;; the effort to look for them.)
5682 ;; If this is a declaration it should end here or its initializer(*)
5683 ;; should start here, so check for allowed separation tokens. Note
5684 ;; that this rule doesn't work e.g. with a K&R arglist after a
5685 ;; function header.
5686 ;;
5687 ;; *) Don't check for C++ style initializers using parens
5688 ;; since those already have been matched as suffixes.
5689 ;;
5690 ;; If `at-decl-or-cast' is then we've found some other sign that
5691 ;; it's a declaration or cast, so then it's probably an
5692 ;; invalid/unfinished one.
5695 ;; Below are tests that only should be applied when we're certain to
5696 ;; not have parsed halfway through an expression.
5699 ;; The expression starts with a known type so treat it as a
5700 ;; declaration.
5704 ;; In C++ we check if the identifier is a known type, since
5705 ;; (con|de)structors use the class name as identifier.
5706 ;; We've always shifted over the identifier as a type and
5707 ;; then backed up again in this case.
5708 identifier-type
5711 ;; `at-type' probably won't be 'found for
5712 ;; destructors since the "~" is then part of the
5713 ;; type name being checked against the list of
5714 ;; known types, so do a check without that
5715 ;; operator.
5720 c-identifier-syntax-table
5724 ;; We have already parsed the type earlier,
5725 ;; so it'd be possible to cache the end
5726 ;; position instead of redoing it here, but
5727 ;; then we'd need to keep track of another
5728 ;; position everywhere.
5737 at-type
5741 ;; Got something like "foo * bar;". Since we're not inside an
5742 ;; arglist it would be a meaningless expression because the
5743 ;; result isn't used. We therefore choose to recognize it as
5744 ;; a declaration. Do not allow a suffix since it could then
5745 ;; be a function call.
5752 ;; Got something like "a (*b) (c);" or "a (b) = c;". It could
5753 ;; be an odd expression or it could be a declaration. Treat
5754 ;; it as a declaration if "a" has been used as a type
5755 ;; somewhere else (if it's a known type we won't get here).
5763 ;; Got a type followed by an abstract declarator. If `got-prefix'
5764 ;; is set it's something like "a *" without anything after it. If
5765 ;; `got-parens' or `got-suffix' is set it's "a()", "a[]", "a()[]",
5766 ;; or similar, which we accept only if the context rules out
5767 ;; expressions.
5770 ;; If we had a complete symbol table here (which rules out
5771 ;; `c-found-types') we should return t due to the disambiguation rule
5772 ;; (in at least C++) that anything that can be parsed as a declaration
5773 ;; is a declaration. Now we're being more defensive and prefer to
5774 ;; highlight things like "foo (bar);" as a declaration only if we're
5775 ;; inside an arglist that contains declarations.
5778 ;; The point is now after the type decl expression.
5781 ;; Check for a cast.
5784 c-cast-parens
5786 ;; Should be the first type/identifier in a cast paren.
5790 ;; The closing paren should follow.
5795 ;; There should be a primary expression after it.
5804 ;; Check if the expression begins with a prefix keyword.
5807 ;; Expression begins with an ambiguous operator. Treat
5808 ;; it as a cast if it's a type decl or if we've
5809 ;; recognized the type somewhere else.
5812 ;; Unless it's a keyword, it's the beginning of a primary
5813 ;; expression.
5815 ;; If `c-primary-expr-regexp' matched a nonsymbol token, check
5816 ;; that it matched a whole one so that we don't e.g. confuse
5817 ;; the operator '-' with '->'. It's ok if it matches further,
5818 ;; though, since it e.g. can match the float '.5' while the
5819 ;; operator regexp only matches '.'.
5823 ;; There should either be a cast before it or something that isn't an
5824 ;; identifier or close paren.
5832 ;; It's a symbol. Accept it only if it's one of the
5833 ;; keywords that can precede an expression (without
5834 ;; surrounding parens).
5837 ;; Check that it isn't a close paren (block close is ok,
5838 ;; though).
5840 ;; Check that it isn't a nonsymbol identifier.
5843 ;; Handle the cast.
5853 ;; We're at a declaration. Highlight the type and the following
5854 ;; declarators.
5860 ;; Make sure to propagate the `c-decl-arg-start' property to
5861 ;; the next argument if it's set in this one, to cope with
5862 ;; interactive refontification.
5874 ;; False alarm. Restore the recorded ranges.
5876 c-record-ref-identifiers save-rec-ref-ids)
5877 nil))))
5880 ;; Assuming that point is at the beginning of a token, check if it starts a
5881 ;; label and if so move over it and return t, otherwise don't move and
5882 ;; return nil. "Label" here means "most things with a colon".
5883 ;;
5884 ;; More precisely, a "label" is regarded as one of:
5885 ;; (i) a goto target like "foo:";
5886 ;; (ii) A case label - either the entire construct "case FOO:" or just the
5887 ;; bare "case", should the colon be missing;
5888 ;; (iii) a keyword which needs a colon, like "default:" or "private:";
5889 ;; (iv) One of QT's "extended" C++ variants of
5890 ;; "private:"/"protected:"/"public:"/"more:" looking like "public slots:".
5891 ;; (v) One of the keywords matched by `c-opt-extra-label-key' (without any
5892 ;; colon). Currently (2006-03), this applies only to Objective C's
5893 ;; keywords "@private", "@protected", and "@public".
5894 ;;
5895 ;; One of the things which will NOT be recognised as a label is a bit-field
5896 ;; element of a struct, something like "int foo:5".
5897 ;;
5898 ;; The end of the label is taken to be just after the colon, or the end of
5899 ;; the first submatch in `c-opt-extra-label-key'. The point is directly
5900 ;; after the end on return. The terminating char gets marked with
5901 ;; `c-decl-end' to improve recognition of the following declaration or
5902 ;; statement.
5903 ;;
5904 ;; If ASSUME-MARKUP is non-nil, it's assumed that the preceding
5905 ;; label, if any, has already been marked up like that.
5906 ;;
5907 ;; If PRECEDING-TOKEN-END is given, it should be the first position
5908 ;; after the preceding token, i.e. on the other side of the
5909 ;; syntactic ws from the point. Use a value less than or equal to
5910 ;; (point-min) if the point is at the first token in (the visible
5911 ;; part of) the buffer.
5912 ;;
5913 ;; The optional LIMIT limits the forward scan for the colon.
5914 ;;
5915 ;; This function records the ranges of the label symbols on
5916 ;; `c-record-ref-identifiers' if `c-record-type-identifiers' (!) is
5917 ;; non-nil.
5918 ;;
5919 ;; This function might do hidden buffer changes.
5922 qt-symbol-idx
5925 ;; "case" or "default" (Doesn't apply to AWK).
5928 ;; Record only the keyword itself for fontification, since in
5929 ;; case labels the following is a constant expression and not
5930 ;; a label.
5934 ;; Find the label end.
5937 ;; Stop on chars that aren't allowed in expressions,
5938 ;; and on operator chars that would be meaningless
5939 ;; there. FIXME: This doesn't cope with ?: operators.
5947 t)
5949 ;; It's an unfinished label. We consider the keyword enough
5950 ;; to recognize it as a label, so that it gets fontified.
5951 ;; Leave the point at the end of it, but don't put any
5952 ;; `c-decl-end' marker.
5954 t)))
5956 ;; @private, @protected, @public, in Objective C, or similar.
5959 ;; For a `c-opt-extra-label-key' match, we record the whole
5960 ;; thing for fontification. That's to get the leading '@' in
5961 ;; Objective-C protection labels fontified.
5966 t)
5968 ;; All other cases of labels.
5971 ;; A colon label must have something before the colon.
5974 ;; Check that we're not after a token that can't precede a label.
5976 ;; Trivially succeeds when there's no preceding token.
5984 ;; Check if we're after a label, if we're after a closing
5985 ;; paren that belong to statement, and with
5986 ;; `c-label-prefix-re'. It's done in different order
5987 ;; depending on `assume-markup' since the checks have
5988 ;; different expensiveness.
6015 ;; Might be after another label, so check it recursively.
6019 ;; Essentially the same as the
6020 ;; `c-syntactic-re-search-forward' regexp below.
6026 ;; Note: the following should work instead of the
6027 ;; narrow-to-region above. Investigate why not,
6028 ;; sometime. ACM, 2006-03-31.
6029 ;; (c-syntactic-skip-backward "^-]:?;}=*/%&|,<>!@+"
6030 ;; macro-start t)
6032 ;; If the caller turned on recording for us,
6033 ;; it shouldn't apply when we check the
6034 ;; preceding label.
6035 c-record-type-identifiers)
6036 ;; A label can't start at a cpp directive. Check for
6037 ;; this, since c-forward-syntactic-ws would foul up on it.
6042 ;; Check that the next nonsymbol token is ":", or that we're in one
6043 ;; of QT's "slots" declarations. Allow '(' for the sake of macro
6044 ;; arguments. FIXME: Should build this regexp from the language
6045 ;; constants.
6061 t)))
6066 ;; Check that `c-nonlabel-token-key' doesn't match anywhere.
6075 t)
6078 ;; Record the identifiers in the label for fontification, unless
6079 ;; it begins with `c-label-kwds' in which case the following
6080 ;; identifiers are part of a (constant) expression that
6081 ;; shouldn't be fontified.
6091 t)))
6094 ;; Not a label.
6096 nil))))
6099 ;; Assuming the point is at the beginning of a token, try to move
6100 ;; forward to the end of the Objective-C directive that starts
6101 ;; there. Return t if a directive was fully recognized, otherwise
6102 ;; the point is moved as far as one could be successfully parsed and
6103 ;; nil is returned.
6104 ;;
6105 ;; This function records identifier ranges on
6106 ;; `c-record-type-identifiers' and `c-record-ref-identifiers' if
6107 ;; `c-record-type-identifiers' is non-nil.
6108 ;;
6109 ;; This function might do hidden buffer changes.
6112 start-char
6114 ;; Turn off recognition of angle bracket arglists while parsing
6115 ;; types here since the protocol reference list might then be
6116 ;; considered part of the preceding name or superclass-name.
6127 t)
6136 ;; Handle the name of the class itself.
6138 ; (c-forward-token-2) ; 2006/1/13 This doesn't move if the token's
6139 ; at EOB.
6145 ;; Look for ": superclass-name" or "( category-name )".
6156 ;; Look for a protocol reference list.
6162 t))))
6168 t)
6171 nil)))
6174 ;; Go to the first non-whitespace after the colon that starts a
6175 ;; multiple inheritance introduction. Optional LIM is the farthest
6176 ;; back we should search.
6177 ;;
6178 ;; This function might do hidden buffer changes.
6186 ;; Return nil if we aren't in a method definition, otherwise the
6187 ;; position of the initial [+-].
6188 ;;
6189 ;; This function might do hidden buffer changes.
6195 ))
6197 ;; Contributed by Kevin Ryde <user42@zip.com.au>.
6199 ;; Return non-nil if point is within a gcc \"asm\" block.
6200 ;;
6201 ;; This should be called with point inside an argument list.
6202 ;;
6203 ;; Only one level of enclosing parentheses is considered, so for
6204 ;; instance `nil' is returned when in a function call within an asm
6205 ;; operand.
6206 ;;
6207 ;; This function might do hidden buffer changes.
6217 "Return a determination as to whether point is at the `top-level'.
6218 Being at the top-level means that point is either outside any
6219 enclosing block (such function definition), or only inside a class,
6220 namespace or other block that contains another declaration level.
6222 If point is not at the top-level (e.g. it is inside a method
6223 definition), then nil is returned. Otherwise, if point is at a
6224 top-level not enclosed within a class definition, t is returned.
6225 Otherwise, a 2-vector is returned where the zeroth element is the
6226 buffer position of the start of the class declaration, and the first
6227 element is the buffer position of the enclosing class's opening
6228 brace.
6230 Note that this function might do hidden buffer changes. See the
6231 comment at the start of cc-engine.el for more info."
6237 ;; Return non-nil if the point is in the region after the argument
6238 ;; list of a function and its opening brace (or semicolon in case it
6239 ;; got no body). If there are K&R style argument declarations in
6240 ;; that region, the point has to be inside the first one for this
6241 ;; function to recognize it.
6242 ;;
6243 ;; If successful, the point is moved to the first token after the
6244 ;; function header (see `c-forward-decl-or-cast-1' for details) and
6245 ;; the position of the opening paren of the function arglist is
6246 ;; returned.
6247 ;;
6248 ;; The point is clobbered if not successful.
6249 ;;
6250 ;; LIM is used as bound for backward buffer searches.
6251 ;;
6252 ;; This function might do hidden buffer changes.
6265 ;; There should not be a '=' or ',' between beg and the
6266 ;; start of the declaration since that means we were in the
6267 ;; "expression part" of the declaration.
6272 ;; Check that there's an arglist paren in the
6273 ;; declaration.
6276 ;; The declarator is a paren expression, so skip past it
6277 ;; so that we don't get stuck on that instead of the
6278 ;; function arglist.
6282 ;; Don't trip up on "operator ()".
6289 ;; Return the position of the first argument declaration if point is
6290 ;; inside a K&R style argument declaration list, nil otherwise.
6291 ;; `c-recognize-knr-p' is not checked. If LIM is non-nil, it's a
6292 ;; position that bounds the backward search for the argument list.
6293 ;;
6294 ;; Point must be within a possible K&R region, e.g. just before a top-level
6295 ;; "{". It must be outside of parens and brackets. The test can return
6296 ;; false positives otherwise.
6297 ;;
6298 ;; This function might do hidden buffer changes.
6302 ;; If we're in a macro, our search range is restricted to it. Narrow to
6303 ;; the searchable range.
6306 before-lparen after-rparen)
6309 ;; Search backwards for the defun's argument list. We give up if we
6310 ;; encounter a "}" (end of a previous defun) or BOB.
6311 ;;
6312 ;; The criterion for a paren structure being the arg list is:
6313 ;; o - there is non-WS stuff after it but before any "{"; AND
6314 ;; o - the token after it isn't a ";" AND
6315 ;; o - it is preceded by either an identifier (the function name) or
6316 ;; a macro expansion like "DEFUN (...)"; AND
6317 ;; o - its content is a non-empty comma-separated list of identifiers
6318 ;; (an empty arg list won't have a knr region).
6319 ;;
6320 ;; The following snippet illustrates these rules:
6321 ;; int foo (bar, baz, yuk)
6322 ;; int bar [] ;
6323 ;; int (*baz) (my_type) ;
6324 ;; int (*) (void) (*yuk) (void) ;
6325 ;; {
6338 ;; We're inside a paren. Could it be our argument list....?
6344 ;; FIXME!!! What about macros between the parens? 2007/01/20
6347 ;; It can't be the arg list if next token is ; or {
6352 ;; Is the thing preceding the list an identifier (the
6353 ;; function name), or a macro expansion?
6363 ;; Have we got a non-empty list of comma-separated
6364 ;; identifiers?
6378 ;; ...Yes. We've identified the function's argument list.
6384 ;; ...No. The current parens aren't the function's arg list.
6391 ;; skip forward over conditional at point, including any predicate
6392 ;; statements in parentheses. No error checking is performed.
6393 ;;
6394 ;; This function might do hidden buffer changes.
6396 ;; else if()
6401 ;; do, else, try, finally
6406 ;; for, if, while, switch, catch, synchronized, foreach
6410 ;; If looking at the token after a conditional then return the
6411 ;; position of its start, otherwise return nil.
6412 ;;
6413 ;; This function might do hidden buffer changes.
6423 ;; If the point is after an operator identifier that isn't handled
6424 ;; like an ordinary symbol (i.e. like "operator =" in C++) then the
6425 ;; position of the start of that identifier is returned. nil is
6426 ;; returned otherwise. The point may be anywhere in the syntactic
6427 ;; whitespace after the last token of the operator identifier.
6428 ;;
6429 ;; This function might do hidden buffer changes.
6441 ;; Assuming point is at a brace that opens a statement block of some
6442 ;; kind, move to the proper anchor point for that block. It might
6443 ;; need to be adjusted further by c-add-stmt-syntax, but the
6444 ;; position at return is suitable as start position for that
6445 ;; function.
6446 ;;
6447 ;; This function might do hidden buffer changes.
6454 ;; Assuming point is at a brace that opens the block of a top level
6455 ;; declaration of some kind, move to the proper anchor point for
6456 ;; that block.
6457 ;;
6458 ;; This function might do hidden buffer changes.
6463 ;; Search forward for the end of the "header" of the current
6464 ;; declaration. That's the position where the definition body
6465 ;; starts, or the first variable initializer, or the ending
6466 ;; semicolon. I.e. search forward for the closest following
6467 ;; (syntactically relevant) '{', '=' or ';' token. Point is left
6468 ;; _after_ the first found token, or at point-max if none is found.
6469 ;;
6470 ;; This function might do hidden buffer changes.
6475 ;; In C++ we need to take special care to handle operator
6476 ;; tokens and those pesky template brackets.
6481 ;; Handle operator identifiers, i.e. ignore any
6482 ;; operator token preceded by "operator".
6489 t
6491 nil)))))
6500 ;; Go to the beginning of the current declaration, or the beginning
6501 ;; of the previous one if already at the start of it. Point won't
6502 ;; be moved out of any surrounding paren. Return a cons cell of the
6503 ;; form (MOVE . KNR-POS). MOVE is like the return value from
6504 ;; `c-beginning-of-statement-1'. If point skipped over some K&R
6505 ;; style argument declarations (and they are to be recognized) then
6506 ;; KNR-POS is set to the start of the first such argument
6507 ;; declaration, otherwise KNR-POS is nil. If LIM is non-nil, it's a
6508 ;; position that bounds the backward search.
6509 ;;
6510 ;; NB: Cases where the declaration continues after the block, as in
6511 ;; "struct foo { ... } bar;", are currently recognized as two
6512 ;; declarations, e.g. "struct foo { ... }" and "bar;" in this case.
6513 ;;
6514 ;; This function might do hidden buffer changes.
6520 ;; `c-beginning-of-statement-1' stops at a block start, but we
6521 ;; want to continue if the block doesn't begin a top level
6522 ;; construct, i.e. if it isn't preceded by ';', '}', ':', bob,
6523 ;; or an open paren.
6525 ;; Go back one "statement" each time round the loop until we're just
6526 ;; after a ;, }, or :, or at BOB or the start of a macro or start of
6527 ;; an ObjC method. This will move over a multiple declaration whose
6528 ;; components are comma separated.
6530 ;; Must check with c-opt-method-key in ObjC mode.
6540 ;; Check that we don't move from the first thing in a
6541 ;; macro to its header.
6547 move tentative-move))
6552 ;; Handle K&R argdecls. Back up after the "statement" jumped
6553 ;; over by `c-beginning-of-statement-1', unless it was the
6554 ;; function body, in which case we're sitting on the opening
6555 ;; brace now. Then test if we're in a K&R argdecl region and
6556 ;; that we started at the other side of the first argdecl in
6557 ;; it.
6567 'previous
6569 knr-argdecl-start))
6572 ;; `c-beginning-of-statement-1' counts each brace block as a separate
6573 ;; statement, so the result will be 'previous if we've moved over any.
6574 ;; So change our result back to 'same if necessary.
6575 ;;
6576 ;; If they were brace list initializers we might not have moved over a
6577 ;; declaration boundary though, so change it to 'same if we've moved
6578 ;; past a '=' before '{', but not ';'. (This ought to be integrated
6579 ;; into `c-beginning-of-statement-1', so we avoid this extra pass which
6580 ;; potentially can search over a large amount of text.). Take special
6581 ;; pains not to get mislead by C++'s "operator=", and the like.
6584 c++-template-syntax-table
6590 ; no more, or get to one which isn't an "operator ="
6593 c-overloadable-operators-regexp
6594 c-opt-op-identifier-prefix
6609 ;; Assuming point is at the start of a declaration (as detected by
6610 ;; e.g. `c-beginning-of-decl-1'), go to the end of it. Unlike
6611 ;; `c-beginning-of-decl-1', this function handles the case when a
6612 ;; block is followed by identifiers in e.g. struct declarations in C
6613 ;; or C++. If a proper end was found then t is returned, otherwise
6614 ;; point is moved as far as possible within the current sexp and nil
6615 ;; is returned. This function doesn't handle macros; use
6616 ;; `c-end-of-macro' instead in those cases.
6617 ;;
6618 ;; This function might do hidden buffer changes.
6621 c++-template-syntax-table
6629 ;; Stopped at the ';' in a K&R argdecl section which is
6630 ;; detected using the same criteria as in
6631 ;; `c-beginning-of-decl-1'. Move to the following block
6632 ;; start.
6636 ;; Encountered a block in the declaration. Jump over it.
6647 ;; Check for `c-opt-block-decls-with-vars-key'
6648 ;; before the first paren.
6651 c-opt-block-decls-with-vars-key
6653 lim t t t)
6656 ;; Check that the first following paren is
6657 ;; the block.
6659 lim t t t)
6661 ;; The declaration doesn't have any of the
6662 ;; `c-opt-block-decls-with-vars' keywords in the
6663 ;; beginning, so it ends here at the end of the block.
6671 nil)))
6674 ;; Assuming the point is at an open brace, check if it starts a
6675 ;; block that contains another declaration level, i.e. that isn't a
6676 ;; statement block or a brace list, and if so return non-nil.
6677 ;;
6678 ;; If the check is successful, the return value is the start of the
6679 ;; keyword that tells what kind of construct it is, i.e. typically
6680 ;; what `c-decl-block-key' matched. Also, if GOTO-START is set then
6681 ;; the point will be at the start of the construct, before any
6682 ;; leading specifiers, otherwise it's at the returned position.
6683 ;;
6684 ;; The point is clobbered if the check is unsuccessful.
6685 ;;
6686 ;; CONTAINING-SEXP is the position of the open of the surrounding
6687 ;; paren, or nil if none.
6688 ;;
6689 ;; The optional LIMIT limits the backward search for the start of
6690 ;; the construct. It's assumed to be at a syntactically relevant
6691 ;; position.
6692 ;;
6693 ;; If any template arglists are found in the searched region before
6694 ;; the open brace, they get marked with paren syntax.
6695 ;;
6696 ;; This function might do hidden buffer changes.
6703 ;; Could be at the end of a template arglist.
6714 ;; Note: Can't get bogus hits inside template arglists below since they
6715 ;; have gotten paren syntax above.
6717 ;; If `goto-start' is set we begin by searching for the
6718 ;; first possible position of a leading specifier list.
6719 ;; The `c-decl-block-key' search continues from there since
6720 ;; we know it can't match earlier.
6723 open-brace t t)
6725 t)
6726 t)
6733 ;; Found a keyword that can't be a type?
6736 ;; Can be a type too, in which case it's the return type of a
6737 ;; function (under the assumption that no declaration level
6738 ;; block construct starts with a type).
6741 ;; Jumped over a type, but it could be a declaration keyword
6742 ;; followed by the declared identifier that we've jumped over
6743 ;; instead (e.g. in "class Foo {"). If it indeed is a type
6744 ;; then we should be at the declarator now, so check for a
6745 ;; valid declarator start.
6746 ;;
6747 ;; Note: This doesn't cope with the case when a declared
6748 ;; identifier is followed by e.g. '(' in a language where '('
6749 ;; also might be part of a declarator expression. Currently
6750 ;; there's no such language.
6754 ;; In Pike a list of modifiers may be followed by a brace
6755 ;; to make them apply to many identifiers. Note that the
6756 ;; match data will be empty on return in this case.
6762 ;; Use this variant to avoid yet another special regexp.
6766 t)))
6768 ;; Got a match.
6771 ;; Back up over any preceding specifiers and their clauses
6772 ;; by going forward from `first-specifier-pos', which is the
6773 ;; earliest possible position where the specifier list can
6774 ;; start.
6780 ;; Accept any plain symbol token on the ground that
6781 ;; it's a specifier masked through a macro (just
6782 ;; like `c-forward-decl-or-cast-1' skip forward over
6783 ;; such tokens).
6784 ;;
6785 ;; Could be more restrictive wrt invalid keywords,
6786 ;; but that'd only occur in invalid code so there's
6787 ;; no use spending effort on it.
6793 ;; Can't parse a declaration preamble and is still
6794 ;; before `kwd-start'. That means `first-specifier-pos'
6795 ;; was in some earlier construct. Search again.
6799 ;; Got no preamble before the block declaration keyword.
6805 kwd-start)))
6808 ;; Check if the closest containing paren sexp is a declaration
6809 ;; block, returning a 2 element vector in that case. Aref 0
6810 ;; contains the bufpos at boi of the class key line, and aref 1
6811 ;; contains the bufpos of the open brace. This function is an
6812 ;; obsolete wrapper for `c-looking-at-decl-block'.
6813 ;;
6814 ;; This function might do hidden buffer changes.
6822 nil))
6827 ;; return the buffer position of the beginning of the brace list
6828 ;; statement if we're inside a brace list, otherwise return nil.
6829 ;; CONTAINING-SEXP is the buffer pos of the innermost containing
6830 ;; paren. PAREN-STATE is the remainder of the state of enclosing
6831 ;; braces
6832 ;;
6833 ;; N.B.: This algorithm can potentially get confused by cpp macros
6834 ;; placed in inconvenient locations. It's a trade-off we make for
6835 ;; speed.
6836 ;;
6837 ;; This function might do hidden buffer changes.
6839 ;; This will pick up brace list declarations.
6852 ;; this will pick up array/aggregate init lists, even if they are nested.
6855 ;; Pike can have class definitions anywhere, so we must
6856 ;; check for the class key here.
6858 c-decl-block-key))
6859 bufpos braceassignp lim next-containing)
6861 containing-sexp)
6872 ;; We're in an in-expression block of some kind. Do not
6873 ;; check nesting. We deliberately set the limit to the
6874 ;; containing sexp, so that c-looking-at-inexpr-block
6875 ;; doesn't check for an identifier before it.
6877 ;; see if the open brace is preceded by = or [...] in
6878 ;; this statement, but watch out for operator=
6881 ;; Checks to do only on the first sexp before the brace.
6884 ;; In Java, an initialization brace list may follow
6885 ;; directly after "new Foo[]", so check for a "new"
6886 ;; earlier.
6892 ;; Carry on looking if this is an
6893 ;; identifier (may contain "." in Java)
6894 ;; or another "[]" sexp.
6897 ;; Checks to do on all sexps before the brace, up to the
6898 ;; beginning of the statement.
6906 ;; We've seen a =, but must check earlier tokens so
6907 ;; that it isn't something that should be ignored.
6913 ;; Check for operator =
6916 nil)
6917 ;; Check for `<opchar>= in Pike.
6920 ;; Special case for Pikes
6921 ;; `[]=, since '[' is not in
6922 ;; the punctuation class.
6925 nil)
6927 ;; make sure we're not in a C++ template
6928 ;; argument assignment
6940 ))))
6941 nil)
6948 ;; Brace lists can't contain a semicolon, so we're done.
6950 ;; Go up one level.
6952 lim nil
6953 next-containing nil))
6954 ;; we've hit the beginning of the aggregate list
6958 )
6959 bufpos))
6960 ))
6963 ;; If we're looking at the start of a pike-style list, ie `({Â })',
6964 ;; `([Â ])', `(<Â >)' etc, a cons of a cons of its starting and ending
6965 ;; positions and its entry in c-special-brace-lists is returned, nil
6966 ;; otherwise. The ending position is nil if the list is still open.
6967 ;; LIM is the limit for forward search. The point may either be at
6968 ;; the `(' or at the following paren character. Tries to check the
6969 ;; matching closer, but assumes it's correct if no balanced paren is
6970 ;; found (i.e. the case `({ ... } ... )' is detected as _not_ being
6971 ;; a special brace list).
6972 ;;
6973 ;; This function might do hidden buffer changes.
6978 inner-beg end type)
6993 nil)))))
7003 ;; Check balancing of the inner paren
7004 ;; below.
7007 t)
7008 ;; If the inner char isn't a paren then
7009 ;; we can't check balancing, so just
7010 ;; check the char before the outer
7011 ;; closing paren.
7026 ;; Return non-nil if between two statements or declarations, assuming
7027 ;; point is not inside a literal or comment.
7028 ;;
7029 ;; Obsolete - `c-at-statement-start-p' or `c-at-expression-start-p'
7030 ;; are recommended instead.
7031 ;;
7032 ;; This function might do hidden buffer changes.
7037 ;; Return non-nil if we're looking at the beginning of a block
7038 ;; inside an expression. The value returned is actually a cons of
7039 ;; either 'inlambda, 'inexpr-statement or 'inexpr-class and the
7040 ;; position of the beginning of the construct.
7041 ;;
7042 ;; LIM limits the backward search. CONTAINING-SEXP is the start
7043 ;; position of the closest containing list. If it's nil, the
7044 ;; containing paren isn't used to decide whether we're inside an
7045 ;; expression or not. If both LIM and CONTAINING-SEXP are used, LIM
7046 ;; needs to be farther back.
7047 ;;
7048 ;; If CHECK-AT-END is non-nil then extra checks at the end of the
7049 ;; brace block might be done. It should only be used when the
7050 ;; construct can be assumed to be complete, i.e. when the original
7051 ;; starting position was further down than that.
7052 ;;
7053 ;; This function might do hidden buffer changes.
7058 ;; Look at the character after point only as a last resort
7059 ;; when we can't disambiguate.
7079 ;; If the class definition is at the start of
7080 ;; a statement, we don't consider it an
7081 ;; in-expression class.
7089 ;; Also, in Pike we treat it as an
7090 ;; in-expression class if it's used in an
7091 ;; object clone expression.
7107 nil)
7115 ;; Accept several square bracket sexps for
7116 ;; Java array initializations.
7124 block-follows
7125 containing-sexp
7134 nil
7137 res))))
7140 ;; Returns non-nil if we're looking at the end of an in-expression
7141 ;; block, otherwise the same as `c-looking-at-inexpr-block'.
7142 ;; PAREN-STATE is the paren state relevant at the current position.
7143 ;;
7144 ;; This function might do hidden buffer changes.
7146 ;; We currently only recognize a block.
7149 containing-sexp)
7158 paren-state)
7159 containing-sexp)))))
7161 \f
7162 ;; `c-guess-basic-syntax' and the functions that precedes it below
7163 ;; implements the main decision tree for determining the syntactic
7164 ;; analysis of the current line of code.
7166 ;; Dynamically bound to t when `c-guess-basic-syntax' is called during
7167 ;; auto newline analysis.
7171 ;; A simple function to prepend a new syntax element to
7172 ;; `c-syntactic-context'. Using `setq' on it is unsafe since it
7173 ;; should always be dynamically bound but since we read it first
7174 ;; we'll fail properly anyway if this function is misused.
7176 c-syntactic-context)))
7179 ;; Like `c-add-syntax' but appends to the end of the syntax list.
7180 ;; (Normally not necessary.)
7185 syntax-extra-args
7186 stop-at-boi-only
7187 containing-sexp
7188 paren-state)
7189 ;; Add the indicated SYNTAX-SYMBOL to `c-syntactic-context', extending it as
7190 ;; needed with further syntax elements of the types `substatement',
7191 ;; `inexpr-statement', `arglist-cont-nonempty', `statement-block-intro', and
7192 ;; `defun-block-intro'.
7193 ;;
7194 ;; Do the generic processing to anchor the given syntax symbol on
7195 ;; the preceding statement: Skip over any labels and containing
7196 ;; statements on the same line, and then search backward until we
7197 ;; find a statement or block start that begins at boi without a
7198 ;; label or comment.
7199 ;;
7200 ;; Point is assumed to be at the prospective anchor point for the
7201 ;; given SYNTAX-SYMBOL. More syntax entries are added if we need to
7202 ;; skip past open parens and containing statements. All the added
7203 ;; syntax elements will get the same anchor point.
7204 ;;
7205 ;; SYNTAX-EXTRA-ARGS are a list of the extra arguments for the
7206 ;; syntax symbol. They are appended after the anchor point.
7207 ;;
7208 ;; If STOP-AT-BOI-ONLY is nil, we can stop in the middle of the line
7209 ;; if the current statement starts there.
7210 ;;
7211 ;; Note: It's not a problem if PAREN-STATE "overshoots"
7212 ;; CONTAINING-SEXP, i.e. contains info about parens further down.
7213 ;;
7214 ;; This function might do hidden buffer changes.
7217 ;; This is by far the most common case, so let's give it special
7218 ;; treatment.
7223 ;; Set when we're on a label, so that we don't stop there.
7224 ;; FIXME: To be complete we should check if we're on a label
7225 ;; now at the start.
7226 on-label)
7230 ;; Loop while we have to back out of containing blocks.
7235 ;; Loop while we have to back up statements.
7237 on-label
7240 ;; Skip past any comments that stands between the
7241 ;; statement start and boi.
7249 ;; Skip to the beginning of this statement or backward
7250 ;; another one.
7258 ;; If we didn't move we're at the start of a block and
7259 ;; have to continue outside it.
7268 ;; Special case to avoid deeper and deeper indentation
7269 ;; of "else if" clauses.
7270 )
7275 ;; If stop-at-boi-only is nil, we shouldn't back up
7276 ;; over previous or containing statements to try to
7277 ;; reach boi, so go back to the last position and
7278 ;; exit.
7285 ;; If we've moved into another statement then we
7286 ;; should no longer try to stop in the middle of a
7287 ;; line.
7290 ;; Record this as a substatement if we skipped up one
7291 ;; level.
7294 )))
7296 containing-sexp)
7298 ;; Now we have to go out of this block.
7301 ;; Don't stop in the middle of a special brace list opener
7302 ;; like "({".
7314 ;; Analyze the construct in front of the block we've stepped out
7315 ;; from and add the right syntactic element for it.
7318 step-type)
7321 ;; Stepped out of a parenthesis block, so we're in an
7322 ;; expression now.
7334 ;; Stepped out of an in-expression statement. This
7335 ;; syntactic element won't get an anchor pos.
7338 ;; A parenthesis normally belongs to an arglist.
7346 on-label nil))
7357 paren-state)
7358 containing-sexp)))
7366 ;; Always done if the open brace was at boi. The
7367 ;; c-beginning-of-statement-1 call above is necessary
7368 ;; anyway, to decide the type of block-intro to add.
7371 ))
7373 ;; Fill in the current point as the anchor for all the symbols
7374 ;; added above.
7380 )))
7383 containing-decl-open
7384 containing-decl-start
7385 containing-decl-kwd
7386 paren-state)
7387 ;; The inclass and class-close syntactic symbols are added in
7388 ;; several places and some work is needed to fix everything.
7389 ;; Therefore it's collected here.
7390 ;;
7391 ;; This function might do hidden buffer changes.
7396 containing-decl-open)
7398 ;; Ought to use `c-add-stmt-syntax' instead of backing up to boi
7399 ;; here, but we have to do like this for compatibility.
7409 char-after-ip
7410 beg-of-same-or-containing-stmt
7411 containing-sexp
7412 paren-state)
7413 ;; This function contains the decision tree reached through both
7414 ;; cases 18 and 10. It's a continued statement or top level
7415 ;; construct of some kind.
7416 ;;
7417 ;; This function might do hidden buffer changes.
7424 ;; (CASE A removed.)
7425 ;; CASE B: open braces for class or brace-lists
7432 ;; CASE B.1: class-open
7439 ;; CASE B.2: brace-list-open
7444 indent-point t t t)))
7445 ;; The most semantically accurate symbol here is
7446 ;; brace-list-open, but we normally report it simply as a
7447 ;; statement-cont. The reason is that one normally adjusts
7448 ;; brace-list-open for brace lists as top-level constructs,
7449 ;; and brace lists inside statements is a completely different
7450 ;; context. C.f. case 5A.3.
7453 ;; Turn off the dwim above when we're
7454 ;; analyzing the nature of the brace
7455 ;; for the auto newline feature.
7456 'brace-list-open
7458 nil nil
7459 containing-sexp paren-state))
7461 ;; CASE B.3: The body of a function declared inside a normal
7462 ;; block. Can occur e.g. in Pike and when using gcc
7463 ;; extensions, but watch out for macros followed by blocks.
7464 ;; C.f. cases E, 16F and 17G.
7470 ;; Turn off recognition of constructs that lacks a
7471 ;; type in this case, since that's more likely to be
7472 ;; a macro followed by a block.
7475 containing-sexp paren-state))
7477 ;; CASE B.4: Continued statement with block open. The most
7478 ;; accurate analysis is perhaps `statement-cont' together with
7479 ;; `block-open' but we play DWIM and use `substatement-open'
7480 ;; instead. The rationaly is that this typically is a macro
7481 ;; followed by a block which makes it very similar to a
7482 ;; statement with a substatement block.
7485 containing-sexp paren-state))
7486 ))
7488 ;; CASE C: iostream insertion or extraction operator
7492 ;; If there is no preceding streamop in the statement
7493 ;; then indent this line as a normal statement-cont.
7497 t))))
7499 ;; CASE E: In the "K&R region" of a function declared inside a
7500 ;; normal block. C.f. case B.3.
7502 ;; Check that the next token is a '{'. This works as
7503 ;; long as no language that allows nested function
7504 ;; definitions allows stuff like member init lists, K&R
7505 ;; declarations or throws clauses there.
7506 ;;
7507 ;; Note that we do a forward search for something ahead
7508 ;; of the indentation line here. That's not good since
7509 ;; the user might not have typed it yet. Unfortunately
7510 ;; it's exceedingly tricky to recognize a function
7511 ;; prototype in a code block without resorting to this.
7519 ;; Turn off recognition of constructs that lacks a
7520 ;; type in this case, since that's more likely to be
7521 ;; a macro followed by a block.
7524 containing-sexp paren-state))
7526 ;; CASE D: continued statement.
7530 containing-sexp paren-state))
7531 )))
7533 ;; The next autoload was added by RMS on 2005/8/9 - don't know why (ACM,
7534 ;; 2005/11/29).
7535 ;;;###autoload
7537 "Return the syntactic context of the current line."
7543 ;; A whole ugly bunch of various temporary variables. Have
7544 ;; to declare them here since it's not possible to declare
7545 ;; a variable with only the scope of a cond test and the
7546 ;; following result clauses, and most of this function is a
7547 ;; single gigantic cond. :P
7548 literal char-before-ip before-ws-ip char-after-ip macro-start
7549 in-macro-expr c-syntactic-context placeholder c-in-literal-cache
7550 step-type tmpsymbol keyword injava-inher special-brace-list tmp-pos
7551 ;; The following record some positions for the containing
7552 ;; declaration block if we're directly within one:
7553 ;; `containing-decl-open' is the position of the open
7554 ;; brace. `containing-decl-start' is the start of the
7555 ;; declaration. `containing-decl-kwd' is the keyword
7556 ;; symbol of the keyword that tells what kind of block it
7557 ;; is.
7558 containing-decl-open
7559 containing-decl-start
7560 containing-decl-kwd
7561 ;; The open paren of the closest surrounding sexp or nil if
7562 ;; there is none.
7563 containing-sexp
7564 ;; The position after the closest preceding brace sexp
7565 ;; (nested sexps are ignored), or the position after
7566 ;; `containing-sexp' if there is none, or (point-min) if
7567 ;; `containing-sexp' is nil.
7568 lim
7569 ;; The paren state outside `containing-sexp', or at
7570 ;; `indent-point' if `containing-sexp' is nil.
7572 ;; There's always at most one syntactic element which got
7573 ;; an anchor pos. It's stored in syntactic-relpos.
7574 syntactic-relpos
7577 ;; Check if we're directly inside an enclosing declaration
7578 ;; level block.
7587 containing-sexp)
7588 t)))
7591 containing-sexp nil)
7596 ;; Init some position variables.
7605 ;; Ignore balanced paren. The next entry
7606 ;; can't be another one.
7609 ;; If there is no surrounding open paren then
7610 ;; put the last balanced pair back on paren-state.
7612 containing-sexp nil)))
7616 ;; If we're in a parenthesis list then ',' delimits the
7617 ;; "statements" rather than being an operator (with the
7618 ;; exception of the "for" clause). This difference is
7619 ;; typically only noticeable when statements are used in macro
7620 ;; arglists.
7625 ;; cache char before and after indent point, and move point to
7626 ;; the most likely position to perform the majority of tests
7635 ;; are we in a literal?
7638 ;; now figure out syntactic qualities of the current line
7641 ;; CASE 1: in a string.
7645 ;; CASE 2: in a C or C++ style comment.
7647 ;; This is a kludge for XEmacs where we use
7648 ;; `buffer-syntactic-context', which doesn't correctly
7649 ;; recognize "\*/" to end a block comment.
7650 ;; `parse-partial-sexp' which is used by
7651 ;; `c-literal-limits' will however do that in most
7652 ;; versions, which results in that we get nil from
7653 ;; `c-literal-limits' even when `c-in-literal' claims
7654 ;; we're inside a comment.
7658 ;; CASE 3: in a cpp preprocessor macro continuation.
7668 ;; If at the beginning of the body of a #define
7669 ;; directive then analyze as cpp-define-intro
7670 ;; only. Go on with the syntactic analysis
7671 ;; otherwise. in-macro-expr is set if we're in a
7672 ;; cpp expression, i.e. before the #define body
7673 ;; or anywhere in a non-#define directive.
7680 nil)))))
7684 ;; CASE 11: an else clause?
7688 containing-sexp paren-state))
7690 ;; CASE 12: while closure of a do/while construct?
7698 containing-sexp paren-state))
7700 ;; CASE 13: A catch or finally clause? This case is simpler
7701 ;; than if-else and do-while, because a block is required
7702 ;; after every try, catch and finally.
7710 t)
7714 t)
7717 t))
7722 containing-sexp paren-state))
7724 ;; CASE 18: A substatement we can recognize by keyword.
7734 ;; Ought to cache the result from the
7735 ;; c-beginning-of-statement-1 calls here.
7745 ;; Step up to the containing statement if we
7746 ;; stayed in the same one.
7754 ;; There was no containing statement afterall.
7756 placeholder))
7758 ;; Require a parenthesis after these keywords.
7759 ;; Necessary to catch e.g. synchronized in Java,
7760 ;; which can be used both as statement and
7761 ;; modifier.
7767 ;; CASE 18A: Simple substatement.
7773 containing-sexp paren-state))
7779 containing-sexp paren-state))
7782 containing-sexp paren-state))))
7784 ;; CASE 18B: Some other substatement. This is shared
7785 ;; with case 10.
7787 char-after-ip
7788 placeholder
7789 lim
7790 paren-state)))
7792 ;; CASE 14: A case or default label
7798 containing-sexp))
7801 ;; Got a bogus label at the top level. In lack of better
7802 ;; alternatives, anchor it on (point-min).
7805 ;; CASE 15: any other label
7812 containing-decl-open
7813 containing-decl-start
7814 containing-decl-kwd
7815 paren-state))
7816 ;; Append access-label with the same anchor point as
7817 ;; inclass gets.
7823 containing-sexp))
7828 ;; If the surrounding statement is a switch then
7829 ;; let's analyze all labels as switch labels, so
7830 ;; that they get lined up consistently.
7831 'case-label
7837 ;; A label on the top level. Treat it as a class
7838 ;; context. (point-min) is the closest we get to the
7839 ;; class open brace.
7842 ;; CASE 4: In-expression statement. C.f. cases 7B, 16A and
7843 ;; 17E.
7846 containing-sexp
7847 ;; Have to turn on the heuristics after
7848 ;; the point even though it doesn't work
7849 ;; very well. C.f. test case class-16.pike.
7850 t))
7855 ;; It's a statement block or an anonymous class.
7857 ;; It's a Pike lambda. Check whether we are between the
7858 ;; lambda keyword and the argument list or at the defun
7859 ;; opener.
7861 'inline-open
7867 paren-state)
7871 ;; CASE 5: Line is inside a declaration level block or at top level.
7875 ;; CASE 5A: we are looking at a defun, brace list, class,
7876 ;; or inline-inclass method opening brace
7883 ;; CASE 5A.1: Non-class declaration block open.
7898 ;; Special case for extern-lang-open.
7899 'extern-lang-open
7901 nil t containing-sexp paren-state))
7903 ;; CASE 5A.2: we are looking at a class opening brace
7912 ;; CASE 5A.3: brace list open
7937 ))))
7941 ;; We're in Java and have found that the open brace
7942 ;; belongs to a "new Foo[]" initialization list,
7943 ;; which means the brace list is part of an
7944 ;; expression and not a top level definition. We
7945 ;; therefore treat it as any topmost continuation
7946 ;; even though the semantically correct symbol still
7947 ;; is brace-list-open, on the same grounds as in
7948 ;; case B.2.
7954 ;; CASE 5A.4: inline defun open
7960 containing-decl-open
7961 containing-decl-start
7962 containing-decl-kwd
7963 paren-state))
7965 ;; CASE 5A.5: ordinary defun open
7970 ;; Bogus to use bol here, but it's the legacy.
7972 )))
7974 ;; CASE 5B: After a function header but before the body (or
7975 ;; the ending semicolon if there's no body).
7981 ;; CASE 5B.1: Member init list.
7986 ;; There is no preceding member init clause.
7987 ;; Indent relative to the beginning of indentation
7988 ;; for the topmost-intro line that contains the
7989 ;; prototype's open paren.
7992 ;; Indent relative to the first member init clause.
7997 ;; CASE 5B.2: K&R arg decl intro
8004 containing-decl-open
8005 containing-decl-start
8006 containing-decl-kwd
8007 paren-state)))
8009 ;; CASE 5B.4: Nether region after a C++ or Java func
8010 ;; decl, which could include a `throws' declaration.
8014 )))
8016 ;; CASE 5C: inheritance line. could be first inheritance
8017 ;; line, or continuation of a multiple inheritance
8025 ;; watch out for scope operator
8030 )))
8039 ;; for Java
8044 cont done)
8049 done t))
8053 )
8055 injava-inher)
8058 ))
8061 ;; CASE 5C.1: non-hanging colon on an inher intro
8065 ;; don't add inclass symbol since relative point already
8066 ;; contains any class offset
8067 )
8069 ;; CASE 5C.2: hanging colon on an inher intro
8075 containing-decl-open
8076 containing-decl-start
8077 containing-decl-kwd
8078 paren-state)))
8080 ;; CASE 5C.3: in a Java implements/extends
8094 )))
8096 ;; CASE 5C.4: a continued inheritance line
8100 ;; don't add inclass symbol since relative point already
8101 ;; contains any class offset
8102 )))
8104 ;; CASE 5D: this could be a top-level initialization, a
8105 ;; member init list continuation, or a template argument
8106 ;; list continuation.
8108 ;; Note: We use the fact that lim always is after any
8109 ;; preceding brace sexp.
8119 t))
8124 t)
8130 nil)
8131 t))
8133 ;; NB: No c-after-special-operator-id stuff in this
8134 ;; clause - we assume only C++ needs it.
8139 ;; CASE 5D.3: perhaps a template list continuation?
8154 ;; In a nested template arglist.
8160 ;; FIXME: Should use c-add-stmt-syntax, but it's not yet
8161 ;; template aware.
8164 ;; CASE 5D.4: perhaps a multiple inheritance line?
8176 t)
8181 ;; CASE 5D.5: Continuation of the "expression part" of a
8182 ;; top level construct.
8191 ;; A preceding comma at the top level means that a
8192 ;; new variable declaration starts here. Use
8193 ;; topmost-intro-cont for it, for consistency with
8194 ;; the first variable declaration. C.f. case 5N.
8195 'topmost-intro-cont
8197 nil nil containing-sexp paren-state))
8198 ))
8200 ;; CASE 5F: Close of a non-class declaration level block.
8204 ;; This is inconsistent: Should use `containing-decl-open'
8205 ;; here if it's at boi, like in case 5J.
8209 ;; Special case for compatibility with the
8210 ;; extern-lang syntactic symbols.
8211 'extern-lang-close
8214 nil t
8216 paren-state))
8218 ;; CASE 5G: we are looking at the brace which closes the
8219 ;; enclosing nested class decl
8224 containing-decl-open
8225 containing-decl-start
8226 containing-decl-kwd
8227 paren-state))
8229 ;; CASE 5H: we could be looking at subsequent knr-argdecls
8235 ;; Do an extra check to avoid tripping up on
8236 ;; statements that occur in invalid contexts
8237 ;; (e.g. in macro bodies where we don't really
8238 ;; know the context of what we're looking at).
8245 ;; CASE 5I: ObjC method definition.
8250 ;; Handle the case when it's the first (non-comment)
8251 ;; thing in the buffer. Can't look for a 'same return
8252 ;; value from cbos1 since ObjC directives currently
8253 ;; aren't recognized fully, so that we get 'same
8254 ;; instead of 'previous if it moved over a preceding
8255 ;; directive.
8259 ;; CASE 5P: AWK pattern or function or continuation
8260 ;; thereof.
8266 'topmost-intro-cont
8268 nil nil
8269 containing-sexp paren-state))
8271 ;; CASE 5N: At a variable declaration that follows a class
8272 ;; definition or some other block declaration that doesn't
8273 ;; end at the closing '}'. C.f. case 5D.5.
8282 ;; Speed up the backward search a bit.
8287 ;; The '}' is unbalanced.
8288 nil
8293 containing-sexp paren-state))
8295 ;; NOTE: The point is at the end of the previous token here.
8297 ;; CASE 5J: we are at the topmost level, make
8298 ;; sure we skip back past any access specifiers
8300 ;; A macro continuation line is never at top level.
8322 nil)))))
8323 ;; For historic reasons we anchor at bol of the last
8324 ;; line of the previous declaration. That's clearly
8325 ;; highly bogus and useless, and it makes our lives hard
8326 ;; to remain compatible. :P
8339 ;; Special case for compatibility with the
8340 ;; extern-lang syntactic symbols.
8341 'inextern-lang
8344 nil t
8346 paren-state))
8348 containing-decl-open
8349 containing-decl-start
8350 containing-decl-kwd
8351 paren-state)))
8353 macro-start
8358 ;; CASE 5K: we are at an ObjC method definition
8359 ;; continuation line.
8368 ;; CASE 5L: we are at the first argument of a template
8369 ;; arglist that begins on the previous line.
8377 ;; CASE 5Q: we are at a statement within a macro.
8382 ;; CASE 5M: we are at a topmost continuation line
8393 ))
8395 ;; (CASE 6 has been removed.)
8397 ;; CASE 7: line is an expression, not a statement. Most
8398 ;; likely we are either in a function prototype or a function
8399 ;; call argument list
8407 ;; CASE 7A: we are looking at the arglist closing paren.
8408 ;; C.f. case 7F.
8420 paren-state))
8422 ;; CASE 7B: Looking at the opening brace of an
8423 ;; in-expression block or brace list. C.f. cases 4, 16A
8424 ;; and 17E.
8428 paren-state))
8432 placeholder
8435 paren-state)
8436 containing-sexp)))
8437 ;; placeholder is nil if it's a block directly in
8438 ;; a function arglist. That makes us skip out of
8439 ;; this case.
8440 )))
8445 paren-state)
8449 ;; CASE 7C: we are looking at the first argument in an empty
8450 ;; argument list. Use arglist-close if we're actually
8451 ;; looking at a close paren or bracket.
8463 paren-state))
8465 ;; CASE 7D: we are inside a conditional test clause. treat
8466 ;; these things as statements
8476 ))
8478 ;; CASE 7E: maybe a continued ObjC method call. This is the
8479 ;; case when we are inside a [] bracketed exp, and what
8480 ;; precede the opening bracket is not an identifier.
8488 )))
8490 ;; CASE 7F: we are looking at an arglist continuation line,
8491 ;; but the preceding argument is on the same line as the
8492 ;; opening paren. This case includes multi-line
8493 ;; mathematical paren groupings, but we could be on a
8494 ;; for-list continuation line. C.f. case 7A.
8511 paren-state))
8513 ;; CASE 7G: we are looking at just a normal arglist
8514 ;; continuation line
8517 ))
8519 ;; CASE 8: func-local multi-inheritance line
8529 ;; CASE 8A: non-hanging colon on an inher intro
8534 ;; CASE 8B: hanging colon on an inher intro
8538 ;; CASE 8C: a continued inheritance line
8542 )))
8544 ;; CASE 9: we are inside a brace-list
8554 ;; CASE 9A: In the middle of a special brace list opener.
8574 ;; CASE 9B: brace-list-close brace
8576 ;; Check special brace list closer.
8583 ;; We were between the special close char and the `)'.
8586 ;; We were before the special close char.
8590 ;; Normal brace list check.
8601 ;; Prepare for the rest of the cases below by going to the
8602 ;; token following the opening brace
8615 ;; CASE 9C: we're looking at the first line in a brace-list
8626 ;; CASE 9D: this is just a later brace-list-entry or
8627 ;; brace-entry-open
8636 ))
8637 ))))
8639 ;; CASE 10: A continued statement or top level construct.
8649 ;; This is shared with case 18.
8651 char-after-ip
8652 placeholder
8653 containing-sexp
8654 paren-state))
8656 ;; CASE 16: block close brace, possibly closing the defun or
8657 ;; the class
8659 ;; From here on we have the next containing sexp in lim.
8664 ;; CASE 16E: Closing a statement block? This catches
8665 ;; cases where it's preceded by a statement keyword,
8666 ;; which works even when used in an "invalid" context,
8667 ;; e.g. a macro argument.
8672 ;; CASE 16A: closing a lambda defun or an in-expression
8673 ;; block? C.f. cases 4, 7B and 17E.
8676 nil))
8678 'inline-close
8688 paren-state)
8692 ;; CASE 16B: does this close an inline or a function in
8693 ;; a non-class declaration level block?
8700 nil))
8710 ;; CASE 16F: Can be a defun-close of a function declared
8711 ;; in a statement block, e.g. in Pike or when using gcc
8712 ;; extensions, but watch out for macros followed by
8713 ;; blocks. Let it through to be handled below.
8714 ;; C.f. cases B.3 and 17G.
8720 ;; Turn off recognition of constructs that
8721 ;; lacks a type in this case, since that's more
8722 ;; likely to be a macro followed by a block.
8729 ;; CASE 16C: If there is an enclosing brace then this is
8730 ;; a block close since defun closes inside declaration
8731 ;; level blocks have been handled above.
8733 ;; If the block is preceded by a case/switch label on
8734 ;; the same line, we anchor at the first preceding label
8735 ;; at boi. The default handling in c-add-stmt-syntax
8736 ;; really fixes it better, but we do like this to keep
8737 ;; the indentation compatible with version 5.28 and
8738 ;; earlier. C.f. case 17H.
8745 ;; c-backward-to-block-anchor not necessary here; those
8746 ;; situations are handled in case 16E above.
8749 ;; CASE 16D: Only top level defun close left.
8755 paren-state))
8756 ))
8758 ;; CASE 17: Statement or defun catchall.
8761 ;; Back up statements until we find one that starts at boi.
8764 containing-sexp)))
8768 nil)
8773 ;; CASE 17B: continued statement
8777 containing-sexp paren-state))
8779 ;; CASE 17A: After a case/default label?
8787 'statement-case-open
8789 nil t containing-sexp paren-state))
8791 ;; CASE 17D: any old statement
8798 containing-sexp paren-state)
8802 ;; CASE 17I: Inside a substatement block.
8804 ;; The following tests are all based on containing-sexp.
8806 ;; From here on we have the next containing sexp in lim.
8811 lim paren-state)
8815 ;; CASE 17E: first statement in an in-expression block.
8816 ;; C.f. cases 4, 7B and 16A.
8819 nil))
8821 'defun-block-intro
8830 paren-state)
8836 ;; CASE 17F: first statement in an inline, or first
8837 ;; statement in a top-level defun. we can tell this is it
8838 ;; if there are no enclosing braces that haven't been
8839 ;; narrowed out by a class (i.e. don't use bod here).
8842 paren-state)))
8848 nil))))
8853 ;; CASE 17G: First statement in a function declared inside
8854 ;; a normal block. This can occur in Pike and with
8855 ;; e.g. the gcc extensions, but watch out for macros
8856 ;; followed by blocks. C.f. cases B.3 and 16F.
8862 ;; Turn off recognition of constructs that lacks
8863 ;; a type in this case, since that's more likely
8864 ;; to be a macro followed by a block.
8870 lim paren-state))
8872 ;; CASE 17H: First statement in a block.
8874 ;; If the block is preceded by a case/switch label on the
8875 ;; same line, we anchor at the first preceding label at
8876 ;; boi. The default handling in c-add-stmt-syntax is
8877 ;; really fixes it better, but we do like this to keep the
8878 ;; indentation compatible with version 5.28 and earlier.
8879 ;; C.f. case 16C.
8886 ;; c-backward-to-block-anchor not necessary here; those
8887 ;; situations are handled in case 17I above.
8889 lim paren-state))
8892 ))
8893 )
8895 ;; now we need to look at any modifiers
8899 ;; are we looking at a comment only line?
8904 ;; we might want to give additional offset to friends (in C++).
8909 ;; Set syntactic-relpos.
8915 nil)
8916 t))
8919 ;; Start of or a continuation of a preprocessor directive?
8934 ;; If inside a cpp expression, i.e. anywhere in a
8935 ;; cpp directive except a #define body, we only let
8936 ;; through the syntactic analysis that is internal
8937 ;; in the expression. That means the arglist
8938 ;; elements, if they are anchored inside the cpp
8939 ;; expression.
8948 ;; Inside a #define body and the syntactic analysis is
8949 ;; anchored on the start of the #define. In this case
8950 ;; we add cpp-define-intro to get the extra
8951 ;; indentation of the #define body.
8954 ;; return the syntax
8955 c-syntactic-context)))
8957 \f
8958 ;; Indentation calculation.
8961 ;; offset can be a number, a function, a variable, a list, or one of
8962 ;; the symbols + or -
8963 ;;
8964 ;; This function might do hidden buffer changes.
8983 langelem symbol))
8989 offset symbol)
8990 nil)
9004 Error evaluating offset %S for %s: \
9005 Cannot combine absolute offset %S with relative %S in `%s' method"
9011 Error evaluating offset %S for %s: \
9012 Cannot combine relative offset %S with absolute %S in `%s' method"
9017 res))
9035 Error evaluating offset %S for %s: \
9036 Cannot combine absolute offsets %S and %S in `add' method"
9040 res))
9049 res))))
9056 nil))))
9060 res
9062 offset symbol res)
9063 nil)))
9066 ;; Get offset from LANGELEM which is a list beginning with the
9067 ;; syntactic symbol and followed by any analysis data it provides.
9068 ;; That data may be zero or more elements, but if at least one is
9069 ;; given then the first is the anchor position (or nil). The symbol
9070 ;; is matched against `c-offsets-alist' and the offset calculated
9071 ;; from that is returned.
9072 ;;
9073 ;; This function might do hidden buffer changes.
9083 offset
9087 ))
9090 ;; This is a compatibility wrapper for `c-calc-offset' in case
9091 ;; someone is calling it directly. It takes an old style syntactic
9092 ;; element on the form (SYMBOL . ANCHOR-POS) and converts it to the
9093 ;; new list form.
9094 ;;
9095 ;; This function might do hidden buffer changes.
9102 ;; Calculate the syntactic indentation from a syntactic description
9103 ;; as returned by `c-guess-syntax'.
9104 ;;
9105 ;; Note that topmost-intro always has an anchor position at bol, for
9106 ;; historical reasons. It's often used together with other symbols
9107 ;; that has more sane positions. Since we always use the first
9108 ;; found anchor position, we rely on that these other symbols always
9109 ;; precede topmost-intro in the LANGELEMS list.
9110 ;;
9111 ;; This function might do hidden buffer changes.
9119 ;; Got an absolute column that overrides any indentation
9120 ;; we've collected so far, but not the relative
9121 ;; indentation we might get for the nested structures
9122 ;; further down the langelems list.
9126 ;; Got a relative change of the current calculated
9127 ;; indentation.
9130 ;; Use the anchor position from the first syntactic
9131 ;; element with one.
9141 indent)))
9143 \f
9146 ;;; arch-tag: 149add18-4673-4da5-ac47-6805e4eae089
9147 ;;; cc-engine.el ends here