Michael Olson <[EMAIL PROTECTED]> writes:

> Hmm.  At first I didn't like the idea, but being able to diff against
> just the index would indeed be useful.  When it comes to checking in the
> commit log, however, I would want it to only affect the case where a
> partial commit is being done, and *not* the case there a normal commit
> is being done.

That's precisely why I'm suggesting a variable, and not just a
different shortcut for commit.

Note that "partial commit" is slightly incorrect here. I often use the
index, and rarely actually do a partial commit. I just validate what I
want to commit with git-add, and then run commit.

BTW, in the command-line interface, if you've set color.status=auto,
the changes staged for commit appear in green, and the others in red.
So, you just review and git-add untill git-status is all green.

One thing I don't know how to deal with is that in the flow where you
intend to run commit without -a, both "git diff" and "git diff
--cached" make sense. I'd say M-x xgit-diff RET should run "git diff",
which is the least surprising for git users, but there should probably
be a M-x xgit-diff-cached RET too. In an ideal world, there would be
also a view with Emacs splitted in two parts, with "diff" on the one
side, and "diff --cached" on the other, with shortcuts to move patch
hunks from a view to the other.

> The value itself should be a customizable option, defaulting to nil
> (perhaps) in order to keep the surprise factor to a minimum for
> other VCS users.

The default value is indeed hard to choose. nil makes DVC-git less
surprising for users of other tools, but t makes it less surprising
for git users.

There used to be cogito, which was a kind of extra layer on top of
git, which hided the index, but it's deprecated. Tools I know that use
git do make the index visible to the user (but I don't know most of
them indeed). So, it's not stupid to follow the trend in the git
community.

-- 
Matthieu

_______________________________________________
Dvc-dev mailing list
[email protected]
https://mail.gna.org/listinfo/dvc-dev

Reply via email to