[PATCH 3 of 3] use per-directory clustered stat calls even in cases where known tree is walked

[Benoit Boissinot]

On Tue, Sep 30, 2008 at 11:32 PM, Petr Kodl <petrkodl at gmail.com> wrote:
> # HG changeset patch
> # User Petr Kodl <petrkodl at gmail.com>
> # Date 1222809788 14400
> # Node ID 88d9e471e94e2914234f49c0d4e0cc49795aabe1
> # Parent  89c54a451554b0f9475fe08acbf167dbd065212d
> use per-directory clustered stat calls even in cases where known tree is walked
>
>
> diff -r 89c54a451554 -r 88d9e471e94e mercurial/dirstate.py
> --- a/mercurial/dirstate.py     Tue Sep 30 17:23:08 2008 -0400
> +++ b/mercurial/dirstate.py     Tue Sep 30 17:23:08 2008 -0400
> @@ -507,6 +507,15 @@
>                         if (nf in dmap or not ignore(nf)) and matchfn(nf):
>                             results[nf] = None
>
> +        # if stat on whole directory is fast compared to stat on files
> +        # avoid per-file lstat call in favor of per-directory call
> +        # proceed normally if specific files are listed
> +        recursive = 1
> +        if util.fastdirstat() and not len(match.files()) and not len(work):
> +            work.append('.')
> +            work.extend(self._dirs.keys())
> +            recursive = 0
> +

I'm not sure but I think it would be cleaner not to do this
and to do the clustering in step3.

using a fonction util.statfiles() that would yield (nf, stat or None
if not found)

then step 3 would look like:

for nf, st in util.statfiles(visit):
    if st is not None:
        kind = getkind(st.st_mode)
        if ...:
            st = None
    results[nf] = st

(it could be a one-liner with dict.update() + some lambda, but since
we can't use
generators I don't think it's worth it :)

Then util.statfiles could cluster by directories on win32
(or just sort the files and lstat on unix).

Would it work ?

regards,

Benoit