[RFC] New core command: graft

[Matt Mackall]

On Tue, 2011-10-11 at 20:28 +0200, Laurens Holst wrote:
> Op 10-10-2011 17:58, Matt Mackall schreef:
> >> Otherwise, I like the idea but don’t grok the name. I can’t see me remerbering
> >> the name graft to copy changesets…
> > Well the choices I'm aware of:
> >
> > cherry-pick - cherry-pick originally implied various merge/DAG magic
> > that no one can make good on (though maybe I'm the only one who
> > remembers this) and isn't an especially good name
> 
> Well in common conversation among developers I think it is the word that 
> is most commonly used for this kind of operation, at least we’ve always 
> used it at the places where I’ve worked, so it seems the most intuitive 
> choice to me. I am not worrying about myself remembering the term given 
> my somewhat above-average interest in Mercurial, but I would worry about 
> my co-workers.
> 
> It’s not even an SVN keyword, and that’s from the systems I’ve worked 
> with the only one that even tries to pretend there is any merge magic 
> (since 1.5, june 2008). Git also uses this term. Personally I never 
> associated such tracking with the term, I saw that more as a feature of 
> the VCS, and only a recent one. Also as you say there is no real sane 
> implementation of it, so why reserve the term for something that can not 
> be done :).

Well, remember, I was around when Git and Mercurial and Bzr and whatnot
were invented and I remember the discussions. In these discussions,
"cherry-picking" was -defined- as "do the right thing with respect to
pulling changesets off other branches". Again, I may be the only one who
remembers this, but I think you'll still hear people talking about Git
having "real cherry-picking" even though in reality it's just a
not-really-much-better-than-patch-when-you-get-right-down-to-it merge.

> So I think it’s a good choice. It’s a familiar term to users, and 
> matches git.

It matches git.. superficially.

-- 
Mathematics is the supreme nostalgia of our time.