Re: fontlock-speed

nix> On 08 Feb 2002, Stephen J. Turnbull spake: nix> Um, since C is a (somewhat) context-sensitive language, and I doubt the minimal degree of context sensitivity in C/C++/Java is what's relevant here. I'm sure it's dominated by the fact that the indentation itself is context-sensitive. But I thought context was relative, that is invariably local to the current list and its prefix (eg, there are all kinds of switch () {} conventions, and they generally are quite different from things like function definitions that look sort of like that). I gather you're trying to tell me that's false? nix> In some (admittedly pathological) cases involving K&R nix> declarations and/or hanging-brace K&R brace style, it may nix> have to chase all the way up to the top of the buffer and nix> back down to work out what your indentation level is. Yeah, but it doesn't seem to be an issue of pathology. It simply seems to be a matter of buffer size. It seems that something is _always_ chasing indentation or syntax over the whole buffer. Did you look at the profile? One indentation causes 998 scan-lists in a 7000 line file! What with comment lines and the like, it sounds to me like it's doing a scan-list for a very large proportion of the lists in the file. This could very easily mean that get-text-property happens on the order of N^2 due to nesting of lists. Even microseconds add up quickly in units of 100^2. nix> I don't think this will be fixable without truly massive nix> changes to cc-mode (which would probably cripple its nix> indentation engine again). That would be a bummer. But it would also be a bummer to find ourselves forced to give up local syntax tables just because one important application is wedded to an N^2 algorithm or something like that. -- Institute of Policy and Planning Sciences http://turnbull.sk.tsukuba.ac.jp University of Tsukuba Tennodai 1-1-1 Tsukuba 305-8573 JAPAN Don't ask how you can "do" free software business; ask what your business can "do for" free software.