[PATCH 2 of 5] revlog: use hashlib.sha1 directly instead of through util

[Matt Mackall]

On Thu, 2016-06-23 at 21:51 -0400, Augie Fackler wrote:
> > 
> > On Jun 10, 2016, at 5:41 PM, Gregory Szorc <gregory.szorc at gmail.com> wrote:
> > 
> > On Fri, Jun 10, 2016 at 2:36 PM, Augie Fackler <raf at durin42.com <mailto:raf@
> > durin42.com>> wrote:
> > 
> > > 
> > > On Jun 10, 2016, at 4:47 PM, Matt Mackall <mpm at selenic.com <mailto:mpm at sel
> > > enic.com>> wrote:
> > > 
> > > On Fri, 2016-06-10 at 00:41 -0400, Augie Fackler wrote:
> > > > 
> > > > # HG changeset patch
> > > > # User Augie Fackler <raf at durin42.com <mailto:raf at durin42.com>>
> > > > # Date 1465531834 14400
> > > > #      Fri Jun 10 00:10:34 2016 -0400
> > > > # Node ID bd2caf6762219aba40edd4dd697fee70de306940
> > > > # Parent  16f5b31b1228212dc175b5a0baf3601cca48330d
> > > > revlog: use hashlib.sha1 directly instead of through util
> > > > 
> > > > Also remove module-local _sha alias, which was barely used.
> > > @@ -40,7 +41,6 @@ from . import (
> > > > 
> > > >  _unpack = struct.unpack
> > > >  _compress = zlib.compress
> > > >  _decompress = zlib.decompress
> > > > -_sha = util.sha1
> > > These are each a micro-optimization on Python's abysmal lookup path.
> > Ugh. Does it make a measurable difference on a perf command?
> > 
> > I would go for a revset that accesses non-indexed changelog revision data.
> > Something like author() or date(). That will perform a sha1 for every
> > changeset without involving too many other operations.
> Curiously, this appears to make things worse? See http://hg.durin42.com/hg-wip
> /rev/997c8cf8f586 for a patch and some perfrevset results.
> 
> I’m a little suspicious of my results, but this result lines up with what I’ve
> seen elsewhere - it’s either a little worse or too close to be a meaningful
> result.

I couldn't measure it last week. Power management effects seemed to dominate. Oh
well.

-- 
Mathematics is the supreme nostalgia of our time.