yet another dvcs shootout (this time python-devel)

[Robert Collins]

On Wed, 2008-11-05 at 23:57 +0000, Paul Moore wrote:

> >>timer on & "\Documents and Settings\Gustav\Desktop\Bazaar"\bzr branch http://co
> > de.python.org/python/3.0 bzrpy3 & timer off
> > Timer 1 on: 23:11:37
> > Branched 37415 revision(s).
> > Timer 1 off: 23:28:31  Elapsed: 0:16:53.52
> >
> > I'm not sure what that proves, but I offer the figures for comparison.
> > For me, it looks like a simple bzr branch is still pretty slow. Maybe
> > it says network performance is slow, but why is my time 4x longer than
> > yours?
> 
> For comparison, here are local copy timings:
> 
> >timer on & copy /q /e /s bzrpy3 bzrpy3b & timer off
> Timer 1 on: 23:50:27
> Timer 1 off: 23:50:48  Elapsed: 0:00:21.47
> 
> 23:50 C:\Data
> >timer on & "\Documents and Settings\Gustav\Desktop\Bazaar"\bzr branch bzrpy3 bz
> rpy3a & timer off
> Timer 1 on: 23:51:56
> Branched 37415 revision(s).
> Timer 1 off: 23:55:10  Elapsed: 0:03:13.73
> 
> So bzr branch is 9x slower than a simple directory copy (is bzr like
> Mercurial, where a directory copy is a working clone, as long as there
> are no concurrent users of the branch when you are copying?).

Yes, its like mercurial in that regard. *if* you are using a shared repo
you can only do that copy within the repo, if you are not then every
branch is entirely standalone.

That 9 times overhead will be composed of a couple of things - one is
that we're processing all the history just like we would for a network
copy (so that tells me, that a network copy for you with more bandwidth
will not be less than 3m13s with current bzr), and at the end of the
history copy we have to extract the current code from the repository
(which a plain copy doesn't because its copying the working files).

As you don't like shared repositories, can I suggest using stacked
branches locally? That may work well for you.

e.g. bzr branch --stacked bzrpy3 featurebranch1

This creates a featurebranch1 which depends on bzrpy3 for deep history,
and accumulates new content locally.

-Rob


-- 
GPG key available at: <http://www.robertcollins.net/keys.txt>.
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 189 bytes
Desc: This is a digitally signed message part
Url : https://lists.ubuntu.com/archives/bazaar/attachments/20081106/2722814a/attachment-0001.pgp