[PATCH 2 of 2] store: reuse direncoded path in _hybridencode

[Bryan O'Sullivan]

On Tue, Sep 18, 2012 at 3:01 PM, Adrian Buehlmann <adrian at cadifra.com>wrote:

>
> I think I could try doing the rebasing and fixing of your patches for
> you tomorrow, if you like.
>

No thanks, I'm in the middle of that now.


> My plan was not to deliberately invalidate your patches. I was trying to
> see how far I can get with speeding up the Python code a bit. I think
> I've probably pretty much reached what's possible with Python now.
>

It looks like good progress so far.

In case you haven't been following along, bug 3366 discusses the importance
of making the encoding fast: http://bz.selenic.com/show_bug.cgi?id=3366

I was amazed that moving that code to C sped up clone by a factor of 20.

Next steps would probably indeed be to start taking form your patch
> queue, at least for the non-hashed code path.
>

Yeah, I intend to push the basic stuff to crew later today, but nothing
else.


> Or I should do a proposal for that fncache2 format we've been talking
> about.
>

I've had that in mind too, as it's on my critical path.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://selenic.com/pipermail/mercurial-devel/attachments/20120918/005a6ce0/attachment.html>