1

My goal is to clone/pull/whatever a copy of the latest version of a branch of my repository to my test web server, to test before production.

Note: I realize there are a ton of similar questions on here, but the ones I've looked into were different enough from my situation that I still came away a bit confused.

I am not very well-versed in git on the command line (I tend to use GUI tools), so I'm not sure if I have this right. What I've got so far works, but the initial pull seemed pretty big, so I'm wondering if I am getting the whole history, or only the latest version.

I do not ever want to make changes live on the server or push back to origin. This only needs to be a one-way street.

Let's say I have a branch called test, which is what I will load onto the test web server. However, we don't want the root directory of the repository, because there's some cruft in there. We only want the /sub directory and everything below. The initial setup on my server:

$ cd ~/webapps/test_dir
$ git init
$ git remote add origin https://github.com/user/repository.git
$ git config core.sparseCheckout true
$ echo "sub/*" >> .git/info/sparse-checkout

Then each time I want to update the code on the server, this is what I will run:

$ git pull --depth=1 origin test

Is this command correct, to replace what's on the server with only the latest code, with no history? Is there a more efficient command to accomplish what I want?

Update

In light of Github's lack of support for git archive (which seems like the way to go otherwise), I am trying to work up a solution getting an archive using their APIv3, which they suggested in a support email response.

I've gotten as far as getting a downloaded archive of the correct branch of my repo:

$ curl -L --user "username" https://api.github.com/repos/<username>/<reponame>/tarball/<branchname> > example.tar.gz

But that gets me everything from the root of the branch. What I really want is only one subfolder, but I have yet to figure out how to filter for that.

Update update

Got confirmation from GitHub that there is no way to retrieve a subfolder through this API method. So… back to the way I was doing it in the first place, I think.

swizzlevixen
  • 325
  • 1
  • 12

2 Answers2

1

You probably want to consider using git archive rather than a pull to render out data with no history attached to it. The formal docs for that command are at https://git-scm.com/docs/git-archive.

Something like git archive branchname | tar xf - should be about right to dump the commit at the head of the relevant branch to the local directory.

If you want to put your exported bits into a subdirectory, you can use something like git archive foo --prefix branch/ | tar xf - and that will put all the commits from the HEAD of branch foo into the directory branch. You could also redirect the tar command to move it elsewhere - but since you're in the directory with your git enlistment, the local directory from the first part of this will still have a .git in it.

heckj
  • 7,136
  • 3
  • 39
  • 50
  • it seems that Github may not support `git archive` for some reason? Perplexing, but I've sent in a support request to clarify. – swizzlevixen Apr 24 '17 at 21:29
  • It is true. From their support email: “git-archive isn't something we support I'm afraid. You would need to first do a clone and then git-archive the local copy. If you're looking to download a specific version of a repository and specific files, perhaps our API could be of use to you: https://developer.github.com/v3/git/” So, in light of that, I'm thinking a `curl` or `wget` solution via the API might be the ticket. Working on it. – swizzlevixen Apr 25 '17 at 20:25
  • if you're working with one or two files, curl or wget will get you there, but at more than a handful you're probably better off with with a sparse clone and manually killing the `.git` directory, as crappy as that is. If you're starting with a remote-only github hosted repo, then something like: `git clone --depth=1 https://github.com/bobtiki/whatever;` `cd whatever;` `rm -rf .git` Seriously ghetto, but will probably do what you want. – heckj Apr 25 '17 at 21:40
  • Ok, so I guess my original way was on the right track, but I'll just go in and clean up manually. Thanks for the confirmation! – swizzlevixen Apr 26 '17 at 05:19
0

Referencing this answer by Anthony Hatzopoulos, it seems the best way to do this may be to take advantage of GitHub’s Subversion compatibility, and do this:

$ svn export --force https://github.com/<user>/<repo>/branches/<branch>/<subfolder>

Does exactly what I wanted it to do, without any cruft.

The --force is to replace any current files/folders with the new ones. Of course that means I'm replacing everything instead of incremental updates, but… tradeoffs? ¯\_(ツ)_/¯

Community
  • 1
  • 1
swizzlevixen
  • 325
  • 1
  • 12