Re: [PATCH 09/16] prune: factor out loose-object directory traversal

[Jeff King <peff@peff.net>]

On Fri, Oct 03, 2014 at 03:19:29PM -0700, Junio C Hamano wrote:

> Jeff King <peff@peff.net> writes:
> 
> > Prune has to walk $GIT_DIR/objects/?? in order to find the
> > set of loose objects to prune. Other parts of the code
> > (e.g., count-objects) want to do the same. Let's factor it
> > out into a reusable for_each-style function.
> 
> Doesn't fsck also look at these as well?  I recall that we also have
> a "sort by inum" trick employed there.  Would it also be applicable
> to these two callers?

I didn't think to check fsck. It does have a similar function and could
be converted.

Fsck does store the inode and later visits the files in inode order. I
don't know if it matters as much here, because we are not opening most
of the files, just stat()ing them.

I'd also be kind of surprised if this optimization really matters in
practice these days. The inode-sorting comes from 7e8c174 (fsck-cache:
sort entries by inode number, 2005-05-02). That predates packfiles by
several months, I believe, and you are not likely to have a very large
number of loose objects anymore (and if you do, you will get _way_
better performance simply by packing).

-Peff