Re: fontlock-speed

Hrvoje> Failing that, could the syntax code be modified to better Hrvoje> interact with how XEmacs extents work? Like, is Hrvoje> scan-lists really calling get-text-properties that often Hrvoje> now? Could it be modified to call lower-level functions, I think g-t-p already basically does that. If you need to actually get a property that might be on any of the extents, g-t-p probably can't be much improved on. Hrvoje> and/or to cache the results of those calls? It already does that, however the cache is getting invalidated a lot more often than it needs to. (It gets invalidated on every extent endpoint, whether that extent has a syntax table or not.) In particular, it should be possible to keep a buffer local variable which does not _default_ whether local syntax tables are used or not, but rather is updated when you set a local syntax table property on an extent in the buffer. Since only cperl mode uses them now, that would be enough for 21.4. Then, rather than just a flag, you generalize to keeping a list of the extents that have syntax table properties on them, and consult that rather than doing g-t-p. Hrvoje> The redisplay code already contains similar optimizations, Hrvoje> which include the concept of "extent runs" and "stacks of Hrvoje> extents". This is another good idea, since font-locking closely resembles redisplay in a lot of ways. -- 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.