Re: [PATCH] git-add--interactive: manual hunk editing mode

From: Thomas Rast <trast@student.ethz.ch>
Date: May 29, 2008, 19:10 UTC
Message-ID: <200805292110.02151.trast@student.ethz.ch>
In-Reply-To: <alpine.DEB.1.00.0805291708500.13507@racer.site.net>

Johannes Schindelin wrote:

Show 18 quoted lines

> Hi,
> 
> On Thu, 29 May 2008, Thomas Rast wrote:
> 
> > You wrote:
> 
> Who "You"?  You did not say a name.
> 
> > > Adds a new option 'e' to the 'add -p' command loop that lets you 
> > > discard or keep one hunk line at a time.  This is useful if there are 
> > > no unchanged lines in the middle of the hunk, so 's' will not work, 
> > > but you would still like to split it.
> > 
> 
> The splitting (even without common lines at the borders) is something I 
> needed myself, but the concept got rebuked in
> 
> http://thread.gmane.org/gmane.comp.version-control.git/67854/focus=68108

Aha, thanks for the pointer. I'm not sure I'm in a position to restart that discussion, but let's try anyway:

Note that what Junio said in

http://thread.gmane.org/gmane.comp.version-control.git/67854/focus=68108

Show 7 quoted lines

> That is, , if you want to do finer grained hunk splitting than what "git
> add -p" lets you do, you do _not_ let user specify "I want to split the
> hunk into two, before this point and after this point".  Instead, let
> the user pick zero or more '-' line and zero or more '+' line, and
> adjust the context around it.  An unpicked '-' line becomes the common
> context, and an unpicked '+' line disappears.  After that, you recount
> the diff.  That way, you do not have to do any "unidiff zero" cop-out.

is precisely what my patch implements. I did not implement the stashing he recommends in the same message, but I believe the fact that 'add -p' stages the end result should be enough (and works at least for my use case).

Johannes Sixt replied:

http://thread.gmane.org/gmane.comp.version-control.git/67854/focus=68108

> In this case I would expect two adjacent hunks: one that covers the selected
> changes, another one with the remaining changes, but each against the original:

I didn't do this because there is no way to ensure that the "other half" can be applied cleanly to the unpatched state, thus requiring extra tracking of "hunk dependencies", so to speak. It could be done of course, if anyone really needs it.

You also mention

http://thread.gmane.org/gmane.comp.version-control.git/67854/focus=68108

> I thought about [the line-wise picking], but the UI is not trivial.

The y/n loop with count prefix works well for me. It would be nice to have it in an actual GUI of course, but I don't know enough Tcl/Tk to implement that.

- Thomas

-- 
Thomas Rast
trast@student.ethz.ch