[PATCH 1 of 5 STABLE] filecache: pass func name that constructs the full path of the given file

Matt Mackall mpm at selenic.com
Wed Feb 29 19:10:26 UTC 2012


On Wed, 2012-02-29 at 10:54 +0200, Idan Kamara wrote:
> On Wed, Feb 29, 2012 at 2:20 AM, Matt Mackall <mpm at selenic.com> wrote:
> 
> > On Wed, 2012-02-29 at 01:40 +0200, Idan Kamara wrote:
> > > On Wed, Feb 29, 2012 at 12:59 AM, Matt Mackall <mpm at selenic.com> wrote:
> >
> > > > Making this work cleanly is a simple matter of localizing the
> > > > path-joining logic to a single method that the derived class overrides:
> > > >
> > > > class storecache(scmutil.filecache):
> > > >    def join(self, obj, fname):
> > > >        return obj.sjoin(fname)
> > >
> > >
> > > I guess that's better, except it forces all clients of such classes
> > > to name their join function the same.
> >
> > No, because this entire class is now specifically tailored for caching
> > store files for localrepo methods (note the 'sjoin'). The class is
> > tailored to the client, the client just needs to use the appropriate
> > class.
> 
> 
> localrepo and dirstate both need a cache for repo files, so now we have
> two 'repocache' classes, one for localrepo's 'join' and another that uses
> whatever dirstate calls its repo-relative join function?

Yes.

> Unless I misunderstand what you mean here, I don't see how this is
> less yucky than my original suggestion.

If you don't already see that passing around pointers to functions as
strings is hacky in the extreme, I really don't know what to tell you.

The only other place we pass around function names by string is
extensions.wrapfunction, which is inherently introspective[1]. I expect
you'll have a hard time finding anywhere else this is done in the Python
library.

[1] Though we could try to map a symbol back to its string name by
walking the container dictionary, that's unreliable since it may not be
1:1.

-- 
Mathematics is the supreme nostalgia of our time.





More information about the Mercurial-devel mailing list