[darcs-users] darcs unapply

John Lato jwlato at gmail.com
Thu Apr 5 15:24:32 UTC 2012


> From: Eric Kow <eric.kow at gmail.com>
>
>
> On 2 Apr 2012, at 19:25, Ben Franksen wrote:
>> Even cooler (if that is possible) would be if there were a standard
>> directory for that under _darcs, say _darcs/obliterated, empty by default,
>> so that one could put
>>
>>  obliterate -O_darcs/obliterated
>>
>> in one's ~/.darcs/defaults.
>
> Hmm, http://bugs.darcs.net/issue75 has some history

That discussion has convinced me of

1.  "darcs obliterate" should actually, (unrecoverably?) delete stuff.
2.  There is some room for darcs to be more helpful, although I don't
think there's agreement yet as to how.

It seems to me that the most common use of "obliterate -O" isn't for
obliterating stuff at all, and it should really go by another name.

What if a new command "darcs stash" is added, which is just an alias
to "darcs obliterate -O_darcs/stash/$timestamp" ?  Or some other
helpful name, which again is simply an alias to the current
"obliterate -O" behavior?

> I wonder if what follows is a helpfully new idea:
>
> 1. without flags darcs should default to creating _darcs/oblterated/YYYY-MM-DD-usual-patchname.dpatch
> 2. (and tell you did so, which is implicit forgivingness, eg, ?Didn't mean to obliterate that? try darcs apply -i _darcs/obliterated/YYYY-MM-DD-usual-patchname.dpatch)
> 3. user/dorcs could do cheap/cheerful gc by getting rid of old patches by name

Personally I would rather that obliterated stuff were erased unless
you specifically tell darcs otherwise. "Obliterate", more than any
other name I've come across in technology, denotes an unrecoverable
loss of data.  But that's the viewpoint of somebody who hasn't
contributed any patches to darcs, so weight it accordingly.

John L.


More information about the darcs-users mailing list