[darcs-users] Git v. Darcs: megapatches and 'darcs shunt'

[Petr Rockai]

Isaac Dupree <ml at isaac.cedarswampstudios.org> writes:
> Hmm interesting.  If that's the semantics we want it could maybe be called
> darcs pull --under.  For the reasons I described before though, I'm not sure
> whether that merging strategy is disciplined enough to produce the results that
> people will actually want.  (Which maybe is a critique of darcs-pull using it
> in the normal way too!  But, well... :)

> From a UI point of view, I might start by thinking about an optimize command.
> I'm thinking it'd be clearly nice to have even if pull has a similar operation,
> and once we have it (or see the challenges in doing it), we'll be able to see
> more clearly whether we also want a pull flag and what it is that we'd want.  I
> hope there aren't a proliferation of reordering strategies that we end up
> wanting.

I would be in favour of a "darcs splay" or "darcs shuffle" (probably
better) or even "darcs bring(up)" kind of operation, where I could
interactively (or with matchers) select what I want to bring near the
top of the stack. It would be useful to both asses dependencies of a
patch, examine it in the current context (diffs) and to do things like
trackdown in the vicinity of a given patch, amend it or temporarily
unpull/unrecord.

(Then, if we had the already-proposed (IIRC) matchers to refer to
presence of patches in external repositories, we could also have
whatever pull --under would do, probably as a pattern using the more
primitive notions.)

Yours,
   Petr

PS: I may be repeating myself, it's hard to recall what I already said
and what I just thought. In that case, sorry for noise.

-- 
id' Ash = Ash; id' Dust = Dust; id' _ = undefined