[RFC/performance] workingtree.commit

[Robert Collins]

On Wed, 2006-09-06 at 07:46 -0500, John Arbash Meinel wrote:
> Robert Collins wrote:


> I think you left in an extra for loop. What you wanted was:
> if deleted_ids:
>     self.work_tree.unversion([file_id for path, file_id in deleted_ids])

yeah my bad.

> Now, I'm okay with having an unversion() function, and you and I already
> agree that the inventory should only be held in memory as long as we
> have a lock. (So it shouldn't be read at startup, and only on request,
> and held inside the Transaction).
> 
> However, we've also discussed that 'commit()' shouldn't be deleting
> entries. They should be marked as missing, and perhaps it should be
> considered a conflict, and should fail to commit, but a missing file
> should not be automatically deleted.

That is a defect true. But I dont want commit to stop deleting entries -
I want it to be optional - just commit -> no deletions. commit -A -
automatically add & delete. This is what hg have and its very slick to
work with. 

> So I'm thinking that unversion() is a misfeature. Because it is enabling
> something we don't want to be doing. I would be okay with adding it in
> the short-term, because it could certainly help our general performance,
> but if it would be a similar amount of effort, I would rather get
> 'missing' files handled properly, and require users to 'bzr rm' files
> that have been deleted.

What do you think of it being an option, in which scenario unversion()
is not a misfeature at all, rather an optional code path for commit.

> So +0, if we decide not to do missing() yet, the go ahead with unversion.
> John
> =:->

-Rob
-- 
GPG key available at: <http://www.robertcollins.net/keys.txt>.
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 191 bytes
Desc: This is a digitally signed message part
Url : https://lists.ubuntu.com/archives/bazaar/attachments/20060907/1cd65dba/attachment.pgp