Re: git-stash: RFC: Adopt the default behavior to other commands

From: Johannes Schindelin <johannes.schindelin@gmx.de>
Date: Dec 19, 2007, 12:01 UTC
Message-ID: <Pine.LNX.4.64.0712191150240.23902@racer.site>
In-Reply-To: <46a038f90712182346t5309448egebfd3726f4d493c5@mail.gmail.com>

Hi,

On Wed, 19 Dec 2007, Martin Langhoff wrote:

Show 16 quoted lines

> On Dec 19, 2007 8:33 PM, Wincent Colaiuta <win@wincent.com> wrote:
> > El 19/12/2007, a las 0:41, Martin Langhoff escribi?:
> >
> > > On Dec 19, 2007 4:42 AM, J?rg Sommer <joerg@alea.gnuu.de> wrote:
> > >> I vote for stash print the list, because I dropped in the pitfall.
> > >
> > > I've dropped there myself, and work with a large team where we are 
> > > both fans of stash, and scarred by it. Any newcomer to git that 
> > > "discovers" stash gets hit by it a dozen times, this is completely 
> > > unnecesary.
> >
> > I may be missing something here, but what's the danger here? An
> 
> Surprise. Your working directory has *just* changed under your feet. 
> Maybe you have an editor with further unsaved changes that is about to 
> act confused whether you undo the stash or not.

The funny thing: why do you call stash with unsaved changes anyway? I mean, come on, even if you were planning to see the usage of stash (which you would get with -h), you undoubtedly wanted to either stash or unstash some changes. Otherwise you would not _need_ to see the usage to begin with.

Show 5 quoted lines

> > unexpected stash is incredibly easy to revert, unless I'm missing
> 
> Once you know about it, yes it is. Once you know about the reflog, you
> can sing and dance and never be worried. But for starting users, it's
> a dangerous command.

No.

Reflogs have nothing to do with it.

If you want to stash changes, you _got_ to know how to unstash those changes. Otherwise "git reset --hard" would have been your friend.

And so, I cannot help but believe that all users of stash -- even the new ones -- know about the fundamental concept of stashing/unstashing, and therefore it is more something like this:

	$ git stash
	[no usage?  What the heck... Oh, it stashed the changes.  I did 
	 not want to stash them just yet.  Okay:]
	$ git unstash
	No such command
	[Oh, maybe I should have read the man page.  Oh, well, the usage 
	 should do...]
	$ git stash -h
	...

Yes, if you start "git stash" blindly, believing that it does what "git stash -h" does, you will be bitten by it once.

But as I said, you will be bitten by a plenty commands if you are not familiar with the concept that "-h" shows the synopsis, "--help" the man page, and that there is a default action where possible.

But stash really is about things like

	$ git stash
	$ git pull
	$ git stash apply

I do not _want_ to see the stash list in _most_ cases, because I do not stash away million dirty changes.

And that workflow was what "git stash" was designed for.

Show 10 quoted lines

> > And nobody commented on the idea I posted earlier which seems to 
> > address the concerns about newbies not knowing what "git stash" with 
> > no params does:
> 
> I agree with making stash more verbose -- if the unlucky new user is
> paying close attention, they'll have instructions on to how to get out
> of trouble. But I agree more with making it "just verbose, no action"
> by default. There are two strong hints:
> 
>  - all other state-changing commands take parameters

What about "git commit", "git reset", "git fetch"? These definitely change state of the repository.

>  - quite a few people in this list have gotten burned with it

Okay, let's have a little calculation: IIRC 3 people (let's make it 10, to be conservative) have said that they had this problem. I must assume that all other people did not have the problem, or at least saw why the default action of "git stash" is what it is.

That said, if you _have_ to have the default changed, then I'll probably implement a "git push" and "git pop" alias here, so that I cannot get annoyed by stash's "new and improved" behaviour.

Ciao, Dscho