Re: Unbalanced parentheses

Zajcev> Ok, in modes where syntax for `"' is "(string quote) - Zajcev> C-M-u gives error, but why? sjt> Because scanlists is a horrific hack. The problem is that sjt> basically the only way to be sure whether you're inside or sjt> outside of a string is to parse forward from the beginning of sjt> the buffer. And scan_lists (the internal function that handles all the parsing) doesn't even try. It simply assumes you've just moved on to a string delimiter from outside, and skips to the next one. So C-M-u has the following effects, where < and > denote beginning and end of buffer respectively, and ! is point: <xxxxxxx ("abcd!efgh")> ==> unbalanced parentheses # original report <"xxxxx" ("abcd!efgh")> ==> unbalanced parentheses # extra " doesn't help <"x ( x" ("abcd!efgh")> ==> <"x !( x" ("abcdefgh")> # dives into string! For now the answer has to be "don't use C-M-u inside a string." :-( Heuristics are not going to help much. <(" word ")> is obviously a list containing a single space-padded word, but there's no general heuristic to distinguish that from <(concat "func (" arg ");")>. I suspect that at some point in the distant past (cvs blame seems to say this hasn't changed in any significant respect since 21.2 at least, but of course cvs blame doesn't tell you about _deletions_), scan_lists did check that precondition. Then the code got eliminated because it's expensive and looked like a no-op. I think that the way to handle this is to always parse forward from a point known to be outside of a comment or string, and thus determine whether you are now inside a string (so you zip to the matching delimited) or outside (and resume list-oriented parsing). This can be made pretty efficient by caching known places. Note that because this kind of parsing always goes forward, you can always postpone parsing forward past where you are until you need it at no loss. Then any time a relevant text change takes place, you invalidate the cache past the point of the text change. This strategy can be refined in stages: 1. No cache---parse comments and strings from the beginning every time. 2. Cache and invalidate after point on every insertion or deletion. 3. Cache and invalidate after point on insertion or deletion of a delimiter. I need to go to sleep, but anybody who wants to look at how GNU Emacs does this would be a hero. -- School of Systems and Information Engineering http://turnbull.sk.tsukuba.ac.jp University of Tsukuba Tennodai 1-1-1 Tsukuba 305-8573 JAPAN Ask not how you can "do" free software business; ask what your business can "do for" free software.