[plugin] dirstate experiments

[Bryan O'Sullivan]

On 6/13/06, John Arbash Meinel <john at arbash-meinel.com> wrote:

> take about 139ms for the kernel tree that I'm processing. And there are
> a couple of commits in 'crew' which claim to extract a little bit more
> performance. Though I can say I see a big glaring:
> if '\0' in f:
>   f,c = f.split('\0')

You might think that looks expensive, and so did I. But replacing it
with an unconditional split was no faster for me.

> I think being around 250ms is pretty darn good, considering we are
> handling unicode, and file ids, and full stat information.

Yeah, that's good. You're probably running into the same wall at us at
that point, the apparent limit being dict insert performance.