Ian posted some numbers on bzr repository size vs hg and git. The short story: bzr’s 2.0 format is better than hg on every project tested, and within epsilon for git.
Ian posted some numbers on bzr repository size vs hg and git. The short story: bzr’s 2.0 format is better than hg on every project tested, and within epsilon for git.
Theme: Rubric. Blog at WordPress.com.
Why in the world are all of the hg repo’s imported trunk branch larger than the entire imported repo (i.e. all branches)? That doesn’t make any sense to me. Same with Amarock (imported from git). Are the numbers backwards? Or is there some strange repacking going on (or not going on)?
Comment by nicholas a. evans — October 16, 2009 @ 6:52 pm
Hi Nicholas,
That’s a good question and one I don’t have an answer to. Yes, some repacking is going on as part of the branch operation and it’s proving a net negative in some cases. I’ll raise a bug and ask John to take a look.
Comment by Ian Clatworthy — October 27, 2009 @ 3:47 am
See https://bugs.launchpad.net/bzr/+bug/461651. I’ve updated the Benchmarks page with an explanation but basically, it working-tree metadata (“dirstate”) overhead.
Comment by Ian Clatworthy — October 27, 2009 @ 5:29 am