[Top][All Lists]
[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
syntax-ppss-flush-cache
From: |
martin rudalics |
Subject: |
syntax-ppss-flush-cache |
Date: |
Thu, 20 Oct 2005 10:50:27 +0200 |
User-agent: |
Mozilla Thunderbird 1.0 (Windows/20041206) |
Consider the following stretch of elisp code:
(defvar foo
"string ..
.............
............."
( "doc ..
.........
.........
..........")
Suppose I change some text on the line below "doc. `syntax-ppss' using
`syntax-begin-function' will record something like (1 X ... (X)) where X
is the position of the left paren preceding "doc. `parse-partial-sexp'
run from point-min would produce a value like (2 X ... (.. X)) instead.
Assume `syntax-ppss' stores the recorded state in `syntax-ppss-last'.
I next delete the paren preceding "doc. `syntax-ppss-flush-cache' will
now set the car of `syntax-ppss-last' to nil since BEG is not less than
(car (nth 10 syntax-ppss-last)) equalling X. After this `syntax-ppss'
will execute
((and (not old-pos) old-ppss
(setq pt-min (or (car (nth 9 old-ppss))
...
setting `pt-min' to X. In the sequel `syntax-ppss' for any position
following X may produce a value as if the paren were still there.
Replacing
(if (< beg (or (car (nth 10 syntax-ppss-last))
by
(if (<= beg (or (car (nth 10 syntax-ppss-last))
in `syntax-ppss-flush-cache' resolves this for me. I'm not sure whether
(while (and syntax-ppss-cache (> (caar syntax-ppss-cache) beg))
should become
(while (and syntax-ppss-cache (>= (caar syntax-ppss-cache) beg))
as well.
A related question: Shouldn't
(nth 2 syntax-ppss-last)
become
(nth 3 syntax-ppss-last)
in `syntax-ppss-flush-cache'? After all (nth 2 syntax-ppss-last) can be
non-nil iff (car (nth 10 ...) is non-nil. With other words, if and when
this gets evaluated it always returns nil. On the other hand you need a
valid third field of `syntax-ppss-last' with
(setq pt-min (or (car (nth 9 old-ppss))
(nth 8 old-ppss)
(nth 2 old-ppss)))
in `syntax-ppss' provided the ninth and tenth fields are nil.
- syntax-ppss-flush-cache,
martin rudalics <=