RFC: index.rangemap

[John Arbash Meinel]

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

Robert Collins wrote:
| This patch is for folk to play with, not for merging [yet].
|
| Its a cleanup of the bisection done in the index layer.
|
| I'd be very interested in the following:
|
|  - when indices are slow, does switching in this branch affect
|    performance
|  - if someone felt like doing a code review, that would be nice
|
| -Rob
|

Just a quick performance note.

With 18 packs locally, I see:

~  time bzr log --long -r -10..-1
~  0m2.930s

~  time bzr.2 log --long -r -10..-1
~  0m2.689s

So it isn't amazingly better, but it *is* better.

For --short the difference is 440ms versus 462ms. Though I should note
that "time bzr root" takes 398ms. So most of that 440ms is loading
bzrlib and finding the branch.

So you could look at it as 40ms versus 62ms, which is a bit more impressive.

For "time bzr status" with your revision merged in, it takes 2.14s
versus 2.02s. So again, a small but measurable improvement.

John
=:->
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.4.9 (Cygwin)
Comment: Using GnuPG with Mozilla - http://enigmail.mozdev.org

iEYEARECAAYFAkhcEMwACgkQJdeBCYSNAANxfwCgnlpiUiH0c+tTDT+7jhjLZFh8
e10AnjTsP6+WN9bGEB+h9/CNaE4Nqq6y
=p4mW
-----END PGP SIGNATURE-----