On 15/07/06, Bram Moolenaar <[EMAIL PROTECTED]> wrote:

Maciej Kalisiak wrote:

> On 14/07/06, Bram Moolenaar <[EMAIL PROTECTED]> wrote:
> > Please be specific about what kind of scanning needs to finish before it
> > notices the typed key.  All types of scans are implemented differently.
> > Each should check for a typed key at regular intervals, but if there is
> > one that doesn't I need to know which one to be able to look into that.
>
> OK, here's a more detailed rundown:
> - default 'complete' is .,w,b,u,t,i
> - when I hit ^N, I see, on the bottom line in Vim:
>     Scanning tags.   (for about 5 seconds)
>     Scanning included file: C:\...\file1
>     Scanning included file: C:\...\file2
>     Scanning included file: C:\...\file3  (each of these included files: ~ 5s)
>      ... etc ...

So these are all tags files?  5 seconds is a long time to scan a tags
file.  How big are they?

No, all the files (except the ./tags file in "Scanning tags.") are
Python files, and so is the current file in which I hit ^N.  They are
all less than 1000 lines, and most are less than even 200 lines.

But I've more or less resolved the issue while tweaking Vim on a
different tangent.  The culprit seems to have been too broad a 'path'
variable.  In particular, I've had "C:\Python24\Lib\**" in 'path',
which is an understandably large subtree, with non-trivial depth and a
lot of branching.  Basically having a "foo\**" in path, where foo is
any non-trivial directory, brings the whole thing to a crawl.  The
symptoms then are: the insert-completion takes very long to scan the
tags file as well as any included files, and also, :checkpath becomes
terribly slow, printing out a single line per second.

I guess I was just asking too much of Vim.  Now that I've trimmed down
to a more modest 'path', things are going at decent speeds again, even
the scanning of the tags file, incidentally.

The thing I don't quite follow is that, I would expect a broader
'path' setting to merely result in a larger number of files to scan
through in :checkpath or when looking for completions, but not
necessarily affect the scanning time of a single particular file,
which is what I was seeing...

Reply via email to