[darcs-users] Should the hunk editor edit files, or in-memory patches? (was: Re: [patch123] Name the separators in the hunk split...)

Antoine Latter aslatter at gmail.com
Thu Jan 21 03:29:18 UTC 2010


On Wed, Jan 20, 2010 at 8:14 PM, Mark Stosberg <mark at summersault.com> wrote:
>> > It's my understanding that the edit will not affect the working copy
>> > *immediately*, but will affect it if I say "yes" to the resulting
>> > patch. If that's correct, clearer text might be this:
>>
>> It won't affect it at all!  (which is why editing patches leaves the
>> confusing detritus that restores your working copy from the edited
>> state)
>
> OH. I thought all along I was doing the former: editting a small bit
> of a file, with darcs perhaps keeping track in temporary memory of where
> it should split hunks in that file.
>
> It was said that a goal with the current implementation was to be consist with
> the fact that "record" never modifies the working copy. But I didn't see what
> was happening as inconsistent. In my mind, I was leaving record, opening an
> editor to edit a file (or at least, part of a file), and then returning back to
> an interrupted "record" operation, which would resume again just before the
> changed hunk.  And My conception is not just abstract: I literally leave darcs
> and enter my text editor, exit it again and return back to darcs.
>
> I don't see that darcs has modified my files: I have. There is nothing surprising
> to me about making a changes in a text editor having the changes take affect in
> file on the file system. To me, this is the normal, expected behavior.
>
> The concept of "editing a file" is already well understood, and "editting a
> small piece of a file" is a natural extension of that for me, and does not
> create this new and intuitive notion of "filler patches", or this abstract
> experience of editting what appears a file fragment, but which is actually some
> abstract structure in memory.
>
> So I ask the rest of the user list, would you expect that "edit" during
> darcs record would trigger editting a piece of a file, or would you
> expect it to edit one more patches in memory?

There isn't much utility it editing a file within darcs - I already
have a file editor for that. I want to edit patches. I want to have
more control over what I'm recording.

And if we were editing files in place I really don't see how we could
use it for hunk-splitting, which is one of the major drivers for this
(from what I understand).

Antoine


More information about the darcs-users mailing list