[PATCH 2 of 6] scmutil.addremove: use the dirstate's map directly

Kevin Bullock kbullock+mercurial at ringworld.org
Tue Apr 2 16:15:33 CDT 2013


On 2 Apr 2013, at 3:35 PM, Siddharth Agarwal wrote:

> On 04/02/2013 12:49 PM, Kevin Bullock wrote:
>> Boo, hiss. If dirstate's map needs to be a public object (and it looks like there's several places where it does), let's refactor it into a proper external collaborator of the dirstate object.
> 
> I was expecting this :)
> 
> So I went back and decided to try just pulling the repo.dirstate fetch out of the loop. Turns out that gets us most of the win:
> Before: 2.78s
> repo.dirstate out of the loop: 2.40s
> directly accessing the dirstate map: 2.29s
> 
> Would just pulling the dirstate fetch out of the loop be fine for now?

Totally fine. I think we really _do_ want to eventually factor the map out of dirstate, but that doesn't have to be in scope for this series.

pacem in terris / мир / शान्ति / ‎‫سَلاَم‬ / 平和
Kevin R. Bullock



More information about the Mercurial-devel mailing list