Clone a git repository into an InMemoryRepository with JGit
There is a proper solution without any files on disk, even temporary:
// sample usage
private void loadFromGit() throws Exception {
ObjectLoader loader = loadRemote("https://github.com/msangel/promisified-resource-loader", "master", "README.md");
loader.copyTo(System.out);
}
private ObjectLoader loadRemote(String uri, String branch, String filename) throws Exception {
DfsRepositoryDescription repoDesc = new DfsRepositoryDescription();
InMemoryRepository repo = new InMemoryRepository(repoDesc);
Git git = new Git(repo);
git.fetch()
.setRemote(uri)
.setRefSpecs(new RefSpec("+refs/heads/*:refs/heads/*"))
.call();
repo.getObjectDatabase();
ObjectId lastCommitId = repo.resolve("refs/heads/"+ branch);
RevWalk revWalk = new RevWalk(repo);
RevCommit commit = revWalk.parseCommit(lastCommitId);
RevTree tree = commit.getTree();
TreeWalk treeWalk = new TreeWalk(repo);
treeWalk.addTree(tree);
treeWalk.setRecursive(true);
treeWalk.setFilter(PathFilter.create(filename));
if (!treeWalk.next()) {
return null;
}
ObjectId objectId = treeWalk.getObjectId(0);
ObjectLoader loader = repo.open(objectId);
return loader;
}
So, actually you have TreeWalk here (which is default to recursive), but you can change this and manually iterate all from repo.
The CloneCommand
will always create a file-based repository. The lines from your post that create an InMemoryRepository
have no effect on the clone command.
I suggest to clone into a temporary location if you only need the repository to make a change and push the result.
For example:
Git.cloneRepository().setURI( ... ).setDirectory( new File("/path/to/empty/dir" ) ).call();
An InMemoryRepository
in JGit still requires a work directory for operations like checkout, etc. Only the object database which is usually found in the .git
directory is stored in memory