[MERGE] Make bzr log FILE faster by looking up text_keys one at a time.

John Arbash Meinel john at arbash-meinel.com
Thu Jul 10 03:02:40 BST 2008


-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

Gary van der Merwe wrote:
> Hi
> 
> I stumbled on the fact that by looking up text indexes one at a time
> is much faster than looking up a bunch of them. By doing this, we can
> make bzr log FILE faster.
> 
> Tests with --ls-prof showed a 44% improvement for "bzr log
> bzrlib\tree.py >dev/nul"
> 
> Gary
> 

I wouldn't really trust --lsprof testing of this, as it overly penalizes the
index code for our current indexes. Can you test this without --lsprof and see
if it still shows improvement?

I'm also concerned that our next index format will do as you expect, and
actually *be faster* when doing a batch request.

Otherwise, this certainly seems worth merging.

Also, you might try both before and after running 'bzr pack' on your repository.

John
=:->

-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.4.6 (GNU/Linux)
Comment: Using GnuPG with Mozilla - http://enigmail.mozdev.org

iD8DBQFIdW3AJdeBCYSNAAMRAr/vAKDAH9q6+l7rp9F0jxJ19gVSx8CoXACdE7k4
ie88oQtg7nWFHZbs8mHTNn0=
=YYuf
-----END PGP SIGNATURE-----



More information about the bazaar mailing list