Re: [PATCH v3 00/16] Add new command 'restore'

[Duy Nguyen <pclouds@gmail.com>]

On Wed, May 8, 2019 at 1:31 AM Emily Shaffer <emilyshaffer@google.com> wrote:
> > > I found myself in a situation where I had accidentally
> > > staged all my changes to tracked files (I think resulting from a stash
> > > pop which generated a merge conflict?) and didn't see a good way to
> > > unstage everything using restore.
> > >
> > > I tried out `git restore --staged *` and it tried to restore every build
> > > artifact in my working tree, all of which should be ignored, made a lot of
> > > noisy errors, and left me with my changes still staged.
> >
> > For the record, "git restore --staged :/" should do the trick and it
> > is documented as an example (but without --staged).
>
> Yeah, this worked, and today I also noted `git restore --staged .`
> works, as does Junio's suggestion on the other mail (`git restore
> --staged revision.\*`), and quoting the * (`git restore --staged '*'`).
> So maybe I didn't think outside the box enough before mailing :)

No it's good. It actually got me thinking. Actually I take that back.
It's bad, I'm quite stuck at thinking here :D

> > Either way. I think you raise a good point about "*" (or patterns
> > matching more than expected in general). I need to sleep on it and see
> > if the old way of handling pattern matching failure is still a good
> > way to go.
>
> I think it's worth considering, especially as `git reset HEAD *` and
> `git reset HEAD` both work. (`git restore --staged` complains that it
> hasn't got any paths, but reset seems to figure you just mean
> everything.) It was a surprising failure to me coming away from years of
> using reset.

Yeah I agree that doing it the "git reset" way makes sense for
--staged because shell expansion is tied to worktree files, and
--staged is completely disconnected from worktree. Shell expansion
cannot get this case right, no point punishing the user because of it.

My problem is "what about the other two cases, --worktree alone or
--worktree with --staged (iow do we care about consistency)? what
about git-checkout? does the old way even make sense for git-checkout?
any better way to do if the reason behind is still valid?". I
obviously haven't worked it all out yet.
-- 
Duy