dulwich-users team mailing list archive

[Jelmer Vernooij <jelmer@xxxxxxxxx>]

On Tue, 2011-06-07 at 15:38 +0100, arkaitzj@xxxxxxxxx wrote:
> Hi,
> Nope, no reason for fastimport/export, its just the only place I saw
> something that looked like commit(commit).
> 
> So, I have the repo, and I have the list of commits(in other branches)
> that I want already extracted from the repo with dulwich. Now, how do
> I combine my Repo() object, my [Commit()] list and  a merge algo(I
> guess I also need one) to merge all those commits to
> 'refs/heads/master'.
> Where would I plug the merge algo to get those commits merged into my
> base ref?
You can get the commits and tree objects from the repo object; e.g.:

r = Repo("/path/to/repo")
c = r[some_commit_id]
print c.committer
print c.author
print c.message
root_tree = r[c.tree]
(where root Tree is a tree object)

As dulwich doesn't include any merge infrastructure, your merge
algorithm will have to take it from there.

Cheers,

Jelmer

> Arkaitz
> 
> 
> On Tue, Jun 7, 2011 at 3:34 PM, Jelmer Vernooij <jelmer@xxxxxxxxx>
> wrote:
>         Hi,
>         
>         On Tue, 2011-06-07 at 15:19 +0100, arkaitzj@xxxxxxxxx wrote:
>         > I am evaluating dulwich for a project I have, I just need to
>         handle a
>         > local repo where people do commits.
>         >
>         > I want to be able to check if different branches merge
>         cleanly and to
>         > actually do the merge. I know the software works basically
>         at the data
>         > model level and i've been able to extract the commits
>         between the
>         > different refs I need, I just can't find a way of start
>         applying all
>         > those commits into the base one and see if they succeed or
>         not.
>         
>         > I've been looking at the exporter/importer for this but
>         haven't been
>         > able to find a way as things like emit_commit do not
>         actually commit
>         > anything and I am unable to find how to do it with the text
>         stream
>         > that comes out of the exporter.
>         
>         Is there a particular reason you've been looking at the
>         exporter/importer? Is your initial data in
>         fastimport/fastexport?
>         
>         If you have an existing git repository you should be able to
>         open it
>         with dulwich.repo.Repo and then get whatever objects from it
>         that you
>         need; you can then just use them in memory without having to
>         write to
>         the repository.
>         
>         It should be quite easy to create objects that aren't in a
>         database
>         using the classes in dulwich.objects. If you do want to use a
>         fastimport
>         stream to import existing data, you can import into a
>         MemoryObjectStore
>         as well, which only exists in memory and does not require any
>         data on
>         disk.
>         
>         The main problem I see is that dulwich doesn't have a merge
>         implementation at the moment; are you using something else for
>         the
>         merges?
>         
>         Cheers,
>         
>         Jelmer
>

Attachment: signature.asc
Description: This is a digitally signed message part