1 ;;; nxml-rap.el --- low-level support for random access parsing for nXML mode
3 ;; Copyright (C) 2003, 2004, 2007, 2008 Free Software Foundation, Inc.
8 ;; This file is part of GNU Emacs.
10 ;; GNU Emacs is free software: you can redistribute it and/or modify
11 ;; it under the terms of the GNU General Public License as published by
12 ;; the Free Software Foundation, either version 3 of the License, or
13 ;; (at your option) any later version.
15 ;; GNU Emacs is distributed in the hope that it will be useful,
16 ;; but WITHOUT ANY WARRANTY; without even the implied warranty of
17 ;; MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
18 ;; GNU General Public License for more details.
20 ;; You should have received a copy of the GNU General Public License
21 ;; along with GNU Emacs. If not, see <http://www.gnu.org/licenses/>.
25 ;; This uses xmltok.el to do XML parsing. The fundamental problem is
26 ;; how to handle changes. We don't want to maintain a complete parse
27 ;; tree. We also don't want to reparse from the start of the document
28 ;; on every keystroke. However, it is not possible in general to
29 ;; parse an XML document correctly starting at a random point in the
30 ;; middle. The main problems are comments, CDATA sections and
31 ;; processing instructions: these can all contain things that are
32 ;; indistinguishable from elements. Literals in the prolog are also a
33 ;; problem. Attribute value literals are not a problem because
34 ;; attribute value literals cannot contain less-than signs.
36 ;; Our strategy is to keep track of just the problematic things.
37 ;; Specifically, we keep track of all comments, CDATA sections and
38 ;; processing instructions in the instance. We do this by marking all
39 ;; except the first character of these with a non-nil nxml-inside text
40 ;; property. The value of the nxml-inside property is comment,
41 ;; cdata-section or processing-instruction. The first character does
42 ;; not have the nxml-inside property so we can find the beginning of
43 ;; the construct by looking for a change in a text property value
44 ;; (Emacs provides primitives for this). We use text properties
45 ;; rather than overlays, since the implementation of overlays doesn't
46 ;; look like it scales to large numbers of overlays in a buffer.
48 ;; We don't in fact track all these constructs, but only track them in
49 ;; some initial part of the instance. The variable `nxml-scan-end'
50 ;; contains the limit of where we have scanned up to for them.
52 ;; Thus to parse some random point in the file we first ensure that we
53 ;; have scanned up to that point. Then we search backwards for a
54 ;; <. Then we check whether the < has an nxml-inside property. If it
55 ;; does we go backwards to first character that does not have an
56 ;; nxml-inside property (this character must be a <). Then we start
57 ;; parsing forward from the < we have found.
59 ;; The prolog has to be parsed specially, so we also keep track of the
60 ;; end of the prolog in `nxml-prolog-end'. The prolog is reparsed on
61 ;; every change to the prolog. This won't work well if people try to
62 ;; edit huge internal subsets. Hopefully that will be rare.
64 ;; We keep track of the changes by adding to the buffer's
65 ;; after-change-functions hook. Scanning is also done as a
66 ;; prerequisite to fontification by adding to fontification-functions
67 ;; (in the same way as jit-lock). This means that scanning for these
68 ;; constructs had better be quick. Fortunately it is. Firstly, the
69 ;; typical proportion of comments, CDATA sections and processing
70 ;; instructions is small relative to other things. Secondly, to scan
71 ;; we just search for the regexp <[!?].
73 ;; One problem is unclosed comments, processing instructions and CDATA
74 ;; sections. Suppose, for example, we encounter a <!-- but there's no
75 ;; matching -->. This is not an unexpected situation if the user is
76 ;; creating a comment. It is not helpful to treat the whole of the
77 ;; file starting from the <!-- onwards as a single unclosed comment
78 ;; token. Instead we treat just the <!-- as a piece of not well-formed
79 ;; markup and continue. The problem is that if at some later stage a
80 ;; --> gets added to the buffer after the unclosed <!--, we will need
81 ;; to reparse the buffer starting from the <!--. We need to keep
82 ;; track of these reparse dependencies; they are called dependent
83 ;; regions in the code.
90 (defvar nxml-prolog-end nil
91 "Integer giving position following end of the prolog.")
92 (make-variable-buffer-local 'nxml-prolog-end
)
94 (defvar nxml-scan-end nil
95 "Marker giving position up to which we have scanned.
96 nxml-scan-end must be >= nxml-prolog-end. Furthermore, nxml-scan-end
97 must not an inside position in the following sense. A position is
98 inside if the following character is a part of, but not the first
99 character of, a CDATA section, comment or processing instruction.
100 Furthermore all positions >= nxml-prolog-end and < nxml-scan-end that
101 are inside positions must have a non-nil nxml-inside property whose
102 value is a symbol specifying what it is inside. Any characters with a
103 non-nil fontified property must have position < nxml-scan-end and the
104 correct face. Dependent regions must also be established for any
105 unclosed constructs starting before nxml-scan-end.
106 There must be no nxml-inside properties after nxml-scan-end.")
107 (make-variable-buffer-local 'nxml-scan-end
)
109 (defsubst nxml-get-inside
(pos)
110 (get-text-property pos
'nxml-inside
))
112 (defsubst nxml-clear-inside
(start end
)
113 (remove-text-properties start end
'(nxml-inside nil
)))
115 (defsubst nxml-set-inside
(start end type
)
116 (put-text-property start end
'nxml-inside type
))
118 (defun nxml-inside-end (pos)
119 "Return the end of the inside region containing POS.
120 Return nil if the character at POS is not inside."
121 (if (nxml-get-inside pos
)
122 (or (next-single-property-change pos
'nxml-inside
)
126 (defun nxml-inside-start (pos)
127 "Return the start of the inside region containing POS.
128 Return nil if the character at POS is not inside."
129 (if (nxml-get-inside pos
)
130 (or (previous-single-property-change (1+ pos
) 'nxml-inside
)
134 ;;; Change management
136 (defun nxml-scan-after-change (start end
)
137 "Restore `nxml-scan-end' invariants after a change.
138 The change happened between START and END.
139 Return position after which lexical state is unchanged.
140 END must be > nxml-prolog-end."
141 (if (>= start nxml-scan-end
)
144 (nxml-move-outside-backwards)
146 (let ((inside-remove-start start
)
148 xmltok-dependent-regions
)
149 (while (or (when (xmltok-forward-special (min end nxml-scan-end
))
150 (when (memq xmltok-type
153 processing-instruction
))
154 (nxml-clear-inside inside-remove-start
156 (nxml-set-inside (1+ xmltok-start
)
159 (setq inside-remove-start
(point)))
160 (if (< (point) (min end nxml-scan-end
))
164 ;; The end of the change was inside but is now outside.
165 ;; Imagine something really weird like
166 ;; <![CDATA[foo <!-- bar ]]> <![CDATA[ stuff --> <!-- ]]> -->
167 ;; and suppose we deleted "<
HCoop / bpt / emacs.git / blob