Use column indexes in git-cvsserver where necessary.
commit178e015c0543b581a40adbf4822f44fa592ff68b
authorShawn Pearce <spearce@spearce.org>
Mon, 23 Oct 2006 05:09:35 +0000 (23 01:09 -0400)
committerJunio C Hamano <junkio@cox.net>
Mon, 23 Oct 2006 06:06:00 +0000 (22 23:06 -0700)
tree013e56dd790038307605523514f4492b135928d1
parentb6b7fc7283bd091822541c0286340e78b0c497a2
Use column indexes in git-cvsserver where necessary.

Tonight I found a git-cvsserver instance spending a lot of time in
disk IO while trying to process operations against a Git repository
with >30,000 objects contained in it.

Blowing away my SQLLite database and rebuilding all tables with
indexes on the attributes that git-cvsserver frequently runs queries
against seems to have resolved the issue quite nicely.

Since the indexes shouldn't hurt performance on small repositories
and always helps on larger repositories we should just always create
them when creating the revision storage tables.

Signed-off-by: Shawn O. Pearce <spearce@spearce.org>
Signed-off-by: Junio C Hamano <junkio@cox.net>
git-cvsserver.perl