GitHub Git Merge 2013, 8 Jun 2013

Previous Meeting

Next Meeting

⏯

youtube image

►

From YouTube: Git Merge • Current status of libgit2 (Russell Belfer)

Description

Subscribe: http://bit.ly/subgithub • Russell Belfer, a member of the libgit2 core team, describes the current status of the library.

About Git Merge
Git Merge is a Git user conference, held to bring together Git users and Git developers. This May, they came together in Berlin to improve the world of Git for everyone.

A

Hello. my name's russel belfer.

A

I work at github on libgit2 and a couple folks have asked to hear.

A

A little bit more just about the state of libgit2, because some of you may be aware that we are actually not at a 1.0 release. Yet.

A

So, the question that I've probably gotten the most during this conference has been, "when. Will core git be rewritten.

A

On top of libgit2?" and the answer to that, in case there's, any question, is never.

A

We serve different masters, right? everything that [vicent] talked about when he was speaking is,.

A

All the things that core git does that make it not so useful in a server setting.

A

Are actually good things for it to be doing,, right? there--, it helps it run quickly.. They solve the problem in as efficient way as possible for executing a particular command and that's what they should be doing.

A

and so,. There's some hope that in the future, we may share code modules.

A

You know if, for example, if there's a fantastic merge engine or fantastic file, similarity detection or whatever it might be that we may share some code,, but they are independent projects and you should not be thinking of them as somehow,. One will replace the other.

A

So,, if you hang on just a moment.

A

So,, what is the state of libgit2? libgit2 is very close to a 1.0 release.

A

Anything that you wanna do,, typically with a git repository locally, you can probably do, almost anything..

A

You can read and write objects,.

A

You can create trees, commits..

A

You can manipulate the index.

A

essentially, anything that you would do with references, objects, configuration settings.

A

Should all be available to you., we support fetch and push over http at git.

A

With ssh, well, there's a ssh pull request.

A

That's probably ready to be merged in the next couple weeks or so., which should give us fairly good network support. on windows,. We support winhttp.

A

So it's nicely integrated with the rest of the system,, a pretty good experience.. We support diff status with ignores.

A

And crlf filtering, rename detection,.

A

A pretty wide arrange of things there.

A

history, walking,, ref, logs,, branching,,.

A

There are no conflicts,.

A

Stash, notes, submodules,, all of the.

A

Access to the object, data base, to references and to configurations actually support, plugable storage, back ends.

A

so, out of the box libgit2 ships with.

A

Your traditional file system backends for dealing with git repositories, as you know, and use all the time., but you can actually create your own back-end objects to store those.

A

However, you want to. and there's some experimentation with some alternatives. There. it also supports nicely configurable caching., so you can say, "oh, I'd like to keep cache, trees and memories. So, as you do have these long-lived processes, you can do, you can sort of make some intelligent decisions about what you'd like to save in memory, so that things can be more efficient in the long run.

A

No,, I actually just have notes, so, sorry.

A

I can put this up in a state of.

A

Libgit2 document in the libgit2 repo, but right now,, I'm sorry.. I just have notes.

A

sorry, it's a little last minute.. What libgit2 can do is,.

A

"What can't it do?. What's missing?" so,, probably the biggest thing that's missing is we do not have a merge engine that is complete.. There is a merge in progress.

A

some beginning parts of it have been merged., [chuckles] and then,.

A

There's other stuff, that's on the way,, but it is not done and as a result,, you can't really do pull, because we don't have a good way of merging in things when you pull..

A

We also don't have blame right. Now., that's also a work in progress.

A

this is, in case it's not incredibly obvious to everyone, libgit2 is--.

A

There we go.--in case it's not obvious to everyone. Libgit2 is fully open source, so you could actually just go onto github, libgit2 organization, libgit2 project and see all the pull requests in progress. Comment on them,, say, "I. Don't like the way this api is designed" and we actually care.

A

There are some things that are-- some other small things that are still missing..

A

Our status doesn't currently support rename detection in status,, that's probably on the way.. There's lots of little things.

A

diff doesn't support, word diff or function contacts.. You can't do recursive clones right now., some rough manipulation doesn't automatically write, ref logs and there's a bunch of things you would have to potentially do manually.

A

If you were looking at logs or things like that, but it's pretty good.. There are a few known bugs, so that they are pretty small.

A

a lot: of--if you're running on case-insensitive file systems or particularly,.

A

If you're running on case-insensitive, but case-preserving file systems, that is…evil. and well, not evil,,.

A

It's very convenient, but it is not so convenient for git. and then on windows, there's still some oddities with some links and things like that., but for the most part, it is pretty good.

A

and I will actually talk later about ways that you can help with those very things..

A

So libgit2 runs on a lot of platforms,.

A

Right? so linux, mac, os,, windows,, solaris,.

A

A bunch of bsd's, amigaos,.

A

For those of you who are amiga fans.,.

A

Most of the operations across those platform-- many of the operation are thread-safe..

A

That is not everything, but we've certainly concentrated on access to the objects, and so you can have multiple threads that are access--that are at least looking up and reading the data in the repository, and that seems to work pretty well across threads.

A

Again, we'll probably get better over time, but mostly through you all using it and writing your own git applications.

A

So bindings, most people don't write.

A

C, apparently., I don't know why.

A

and so,. There are bindings in a lot of different languages.. Libgit2 sharp is probably the most advanced of the bindings right now.. It covers almost 100 percent of the functionality that is in libgit2, and it is a very.

A

Idiomatic binding, so if you are a c sharp fan, it should feel very natural.

A

Rugged is the ruby binding.. It is further behind.

A

But it is an actually, well--.

A

Rugged and objective git,, which is the objective c binding and pygit 2,, which is the python binding,.

A

Those three, none of them are complete bindings, but actually a lot-- they're, all fairly active and a lot of the work that they've been doing. Recently, have driven api changes in the underlying library to make it actually easier for them to provide idiomatic support for.

A

Accessing git., it's natural to users of that language in particular,.

A

I think that there's a couple of pygit2 folks that are here and pygit has been doing a lot of work on trying to become more idiomatic so that it feels more natural to the language users.

A

So,, the last thing, really that I have to say is.

A

If you are interested in libgit2 and you'd like to get involved, it is a fairly easy project to get involved with, because you can access it from so many languages..

A

If you're interested in looking at git, data,, manipulating git data,, you can probably pull down a binding in your language of choice and just start playing around. and to the extent that you do that and are willing to contribute back sample code,. You know, here's a little experiment that I did in this thing,.

A

That's amazing for us, right?, because that is how people use the library and learn to kind of have fun with their git repositories, and so anything that you do that you're willing to share is helpful.. You are also certainly welcome to open issues because, I think,.

A

It is still a young library and you will find them and we, like that.

A

you're, also very welcome to open pr's.

A

Pull requests. libgit2 is very,.

A

Very typical, open source, project, right?, we have, you know, on any given month,.

A

We're likely to have 15- 20 contributors.

A

Lot's of people will just fork: it, open pr's and the team--we really try to make a lot of effort to be very supportive.. So it's a pretty easy project to get involved. In. actually,! If you look at our read me,, there's a section on good starter projects..

A

If you wanna start playing around with things- and I think most of the binding teams are also very receptive to people who wanna come in and start finding small things that they can--ways, that they can enhance things or play around with it.

A

so,. I dunno.. If you have any questions,, I'm happy to answer them. sure. ago, actually cut a release.

A

[Belfer] we did. [man], which was nice.. I tried the python bindings the week before.. That was not really fun.--. [Belfer] yeah.

A

milestones going forward?.

A

A pretty good track.

A

And it's--and these things, happen.

A

So, we were shooting for a release,, basically last october, as the microsoft folks.

A

Started getting more involved and we started getting actually a little bit more ambitious because the team of folks who were actively working in the library, grew., we're like, "oh,, well we'll shoot for a 1.0 release by the end of the year.".

A

These things happen, anyhow.. So we did just cut the 0.18 release.

A

I think we are going to make an effort to do more frequent releases and, you know, at least we had been doing pretty well on every three to four months, doing releases prior to that, and so hopefully we'll be at least back on that track. If not, more frequent., our big thing has been.

A

And, as you may have noticed, if you looked at the differences between the last two releases,, as we've tried to get all of the breaking api changes in now,, so that we can actually get to a 1.0 release.

A

Where we won't break the api., and so that's been actually one of the big challenges for the project over like the last six months, as we've been kind of pushing in many breaking api changes.

A

hopefully,, it is resulting in a more consistent and standard api..

A

So it should get us to a good place.

A

But we welcome--you're free to open issues and say, "you, guys suck., please stop doing that.". Actually the release.

A

That did happen,, I dunno if you followed it,, but it was actually largely driven over an issue that was opened,. I think, by the pygit folks, just saying like,.

A

"Hey, c'mon it's time for a release.", but it's fine, we're totally happy to have that..

A

We had a little back and forth discussion. And, you know,, ultimately did try to understand why having that particular tag was important and pushed it out, so.

A

Anything else? okay,, thanks for your time.