On 13/07/06, Bram Moolenaar <[EMAIL PROTECTED]> wrote:
Hmm, perhaps you are not talking about omni completion but about normal Insert mode completion.
Thanks to Hari's post earlier, I now believe so, yes.
This now scans other files sooner than in previous versions, so that the menu can be filled. Note that you don't have to wait for this to complete, typing a character will stop it.
Ah, interesting. The issue for me is that it doesn't really stop immediately; it stops at the end of the current scanning "stage" (i.e., it will stop once it *finishes* scanning the current object being scanned, be it tags or an actual file). This wasn't noticeable earlier, or at least it never drew my attention, but now each "object scan" takes somewhere between 2 to 5 seconds, which is odd. This is on a fairly fast machine. The included Python files, as well as the tags file, are fairly simple and short (most <200 lines, a handful ~1000 lines; the tags file in current directory has about 1000 lines). 2 to 5 seconds seems absurdly too much time for said files, no? Hence, although an improvement, hitting a key immediately after a ^P, still takes about 5 seconds to stop... way too much if you're used to pressing ^P every 3rd or 4th identifier... BTW, my previous estimate of the full scan was incorrect; I've never had the stomach to wait through the whole thing... I just timed it more explicitly, but aborted past the 30 seconds mark... it wasn't nearly done 10% of included files...
