threads / discuss / 28648

Symmetric of rebase / more intelligent cherry-pick

Subject: Symmetric of rebase / more intelligent cherry-pick

## tl;dr

4 messages between Oct 11, 2011 and Oct 11, 2011.

replies: 3people: 3as markdown or json

Lionel Elie Mamane· Oct 11, 2011, 15:54 UTC · lore
Hi,

I was thinking about a sort of symmetric to rebase, a kind of "intelligent" mass cherry-pick. For the sake of example let's call it "basere", although a better name would have to be found.

While
 git rebase UPSTREAM
moves the *current* branch (and any commit in it, but not in UPSTREAM)
on top of UPSTREAM,
 git basere UPSTREAM
would "cherry-pick" all commits in UPSTREAM that are not in HEAD to
the current branch.

git cherry-pick ..UPSTREAM *nearly* does what I want, except that it lacks rebase's intelligence of skipping commits that do the same textual changes as a commit already in the current branch. So maybe it should be an option to cherry-pick rather than a new command? E.g. --skip-same? What about making it cherry-picks *default* behaviour?

Why do I think it is useful? Take the case of a developer following two different origins A and B, and B uses rebase (not merge) to integrate changes from A, but not as often as one would like. A could for example be the "official" tree (e.g. linux Linus tree), and B a feature branch of a developer that is meant to eventually go into A "cleanly" at some future time and A's policy is that one has to rebase before sending a merge request. Then another developer wants to keep up with changes from *both* A and B. So, the situation could look like:

      A---A---A remotes/A/master
     /
X---X---B0---B1---L master
     \
      B0---B1---B2--B3 remotes/B/master
So one does "git rebase remotes/A/master master" and one gets
      A---A---A---B0---B1---L master
     /
X---X
     \
      B0---B1---B2--B3 remotes/B/master
I'd like to go to the situation:
      A---A---A---B0---B1---L---B2--B3 master
     /
X---X
     \
      B0---B1---B2--B3 remotes/B/master

Without having to manually find the cutoff point "B2" (or worse, if the history is more messy than that, as in:

      A---A---A---B1---L master
     /
X---X
     \
      B0---B1---B2--B3 remotes/B/master
)
The best way I have found to do that is:

git fetch B git checkout --no-track -b TMP remotes/B/master git rebase master git checkout master git merge --ff-only TMP git branch -d TMP

or less commands but more messy/fragile (and more work for git):

git fetch B git tag TMP git rebase remotes/B/master git rebase TMP git tag -d TMP

but that is rather onerous for something that should be just one command.

So what are your thoughts on that?
 - Good idea? Would you take a patch that does something like that?
 - Should it be cherry-picks's default behaviour?
 - An option to cherry-pick?
 - A new command? What name? Maybe "multi-cherry-pick"?
 - An option to rebase, e.g. "--pull" or "--incoming" or "--here"?
   ("--pull" is probably too confusing wrt to "git pull"; people will
    have trouble remembering which is which between "git pull
    --rebase" and "git rebase --pull"...)
-- 
Lionel
Junio C Hamano· Oct 11, 2011, 19:36 UTC · re: Lionel Elie Mamane · lore

Re: Symmetric of rebase / more intelligent cherry-pick

Lionel Elie Mamane <lionel@mamane.lu> writes:
> git cherry-pick ..UPSTREAM
> *nearly* does what I want, except that it lacks rebase's intelligence
> of skipping commits that do the same textual changes as a commit
> already in the current branch.

I think in the longer term "--ignore-if-in-upstream" that is known only to format-patch, which is the true source the intelligence of rebase you observed comes from, should be factored out into a helper function that can be used to filter output from get_revision() in other commands, or perhaps get_revision() itself might want to learn it.

I say "or perhaps might" above, because I do not think the general revision traversal machinery used by the log family (which cherry-pick's multi-pick option relies on) has enough information to decide what the caller means by "upstream" at the point setup_revisions() is called.

Junio C Hamano· Oct 11, 2011, 21:28 UTC · re: Junio C Hamano · lore

Re: Symmetric of rebase / more intelligent cherry-pick

Junio C Hamano <gitster@pobox.com> writes:
Show 12 quoted lines
> Lionel Elie Mamane <lionel@mamane.lu> writes:
>
>> git cherry-pick ..UPSTREAM
>> *nearly* does what I want, except that it lacks rebase's intelligence
>> of skipping commits that do the same textual changes as a commit
>> already in the current branch.
>
> I think in the longer term "--ignore-if-in-upstream" that is known only to
> format-patch, which is the true source the intelligence of rebase you
> observed comes from, should be factored out into a helper function that
> can be used to filter output from get_revision() in other commands, or
> perhaps get_revision() itself might want to learn it.

We actually have half of that filtering in "--cherry-pick" that was introduced in d7a17ca (git-log --cherry-pick A...B, 2007-04-09).

Perhaps the recent cherry-pick that allows multiple commits to be picked should use that option (i.e. a "log --cherry-pick --right-only ..@{u}" equivalent) when coming up with which commits to apply?

Jonathan Nieder· Oct 11, 2011, 22:23 UTC · re: Junio C Hamano · lore

Re: Symmetric of rebase / more intelligent cherry-pick

Junio C Hamano wrote:
Show 6 quoted lines
> We actually have half of that filtering in "--cherry-pick" that was
> introduced in d7a17ca (git-log --cherry-pick A...B, 2007-04-09).
>
> Perhaps the recent cherry-pick that allows multiple commits to be picked
> should use that option (i.e. a "log --cherry-pick --right-only ..@{u}"
> equivalent) when coming up with which commits to apply?
No UI ideas from me, but I have a script like this sitting in $HOME/bin:
 base=$1 topic=$2
 git rev-list $base..HEAD |
 git diff-tree -s --pretty=raw --stdin |
 sed -ne 's/(cherry picked from commit \([0-9a-f]*\))/\1/p' |
 xargs git cherry-pick -x -s \
   --cherry-pick --right-only --no-merges HEAD...$topic \
   --not $base

← back to recent threads