How to "cat" a file in JGit?

Question

A while back I was looking for an embeddable distributed version control system in Java, and I think I have found it in JGit, which is a pure Java implementation of git. However, there is not much in the way of sample code or tutorials.

How can I use JGit to retrieve the HEAD version of a certain file (just like svn cat or hg cat whould do)?

I suppose this involves some rev-tree-walking and am looking for a code sample.

The developers are pretty quick to answer on the mailing list : https://dev.eclipse.org/mailman/listinfo/egit-dev . I suggest you give it a try. — Robert Munteanu, Nov 07 '09 at 16:44

score 23 · Accepted Answer · edited Jul 04 '12 at 17:58

23

Unfortunately Thilo's answer does not work with the latest JGit API. Here is the solution I found:

File repoDir = new File("test-git");
// open the repository
Repository repository = new Repository(repoDir);
// find the HEAD
ObjectId lastCommitId = repository.resolve(Constants.HEAD);
// now we have to get the commit
RevWalk revWalk = new RevWalk(repository);
RevCommit commit = revWalk.parseCommit(lastCommitId);
// and using commit's tree find the path
RevTree tree = commit.getTree();
TreeWalk treeWalk = new TreeWalk(repository);
treeWalk.addTree(tree);
treeWalk.setRecursive(true);
treeWalk.setFilter(PathFilter.create(path));
if (!treeWalk.next()) {
  return null;
}
ObjectId objectId = treeWalk.getObjectId(0);
ObjectLoader loader = repository.open(objectId);

// and then one can use either
InputStream in = loader.openStream()
// or
loader.copyTo(out)

I wish it was simpler.

edited Jul 04 '12 at 17:58

imsandli

23
4

answered Sep 15 '11 at 08:05

morisil

1,345
8
19

46

Who in the world designed this API? – Jochen Apr 19 '12 at 22:59
3

Do you know what the nth value from treeWalk.getObjectId(nth) does? (i.e. what are the cases where we pass a value to treeWalk.getObjectId bigger than 0 ?) – Dinis Cruz Jul 15 '13 at 22:50
@DinisCruz `TreeWalk` can walk over more than one tree (by calling `addTree` multiple times). In that case, you can use `getObjectId(N)` to get the object ID from tree N (which might the the same or different, depending on the trees). – robinst Aug 16 '13 at 13:07
See creinig's answer for how to simplify this a bit by using `TreeWalk.forPath`. – robinst Aug 16 '13 at 13:10
How can I save the output of `loader` into a variable? – Evgenij Reznik Aug 26 '14 at 22:31

score 18 · Answer 2 · edited Apr 03 '20 at 03:59

18

Here's a simpler version of @morisil's answer, using some of the concepts from @directed laugh's and tested with JGit 2.2.0:

private String fetchBlob(String revSpec, String path) throws MissingObjectException, IncorrectObjectTypeException,
        IOException {

    // Resolve the revision specification
    final ObjectId id = this.repo.resolve(revSpec);

    // Makes it simpler to release the allocated resources in one go
    ObjectReader reader = this.repo.newObjectReader();

    try {
        // Get the commit object for that revision
        RevWalk walk = new RevWalk(reader);
        RevCommit commit = walk.parseCommit(id);

        // Get the revision's file tree
        RevTree tree = commit.getTree();
        // .. and narrow it down to the single file's path
        TreeWalk treewalk = TreeWalk.forPath(reader, path, tree);

        if (treewalk != null) {
            // use the blob id to read the file's data
            byte[] data = reader.open(treewalk.getObjectId(0)).getBytes();
            return new String(data, "utf-8");
        } else {
            return "";
        }
    } finally {
        reader.close();
    }
}

repo is a Repository object as created in the other answers.

edited Apr 03 '20 at 03:59

Emadpres

3,466
2
29
44

answered Feb 13 '13 at 14:55

creinig

1,560
12
23

1

Looks good, except for returning getBytes() while the return type is String. Note that you also have to call `release` on `walk` and `treeWalk` for releasing resources. To have to do that only once, call `ObjectReader reader = repo.newObjectReader()` and pass that to Revwalk and Treewalk instead of the repository. Then call `reader.release()` in a finally block. – robinst Aug 16 '13 at 13:03
RevWalk also needs to be released, move the call to `newObjectReader` up and use `new RevWalk(reader)` instead. – robinst Sep 04 '13 at 11:02
Another comment (sorry :)), `this.repo.open` should also be replaced by `reader.open`. – robinst Sep 04 '13 at 13:48
@robinst: Fixed. And feel free to edit the answer as needed ;) – creinig Sep 05 '13 at 07:16

score 12 · Answer 3 · answered Feb 12 '12 at 19:37

I followed @Thilo's and @morisil's answer to get this, compatible with JGit 1.2.0:

File repoDir = new File("test-git/.git");
// open the repository
Repository repo = new Repository(repoDir);
// find the HEAD
Commit head = repo.mapCommit(Constants.HEAD);
// retrieve the tree in HEAD
Tree tree = head.getTree();

// 1.2.0 api version here
// find a file (as a TreeEntry, which contains the blob object id)
TreeWalk treewalk = TreeWalk.forPath(repo, "b/test.txt", tree);
// use the blob id to read the file's data
byte[] data = repo.open(treewalk.getObjectId(0)).getBytes();

I didn't test the Java version but it should work. It translates from

(.getBytes (.open repo (.getObjectId (TreeWalk/forPath repo "b/test.txt" tree) 0)))

in clojure (following the same setup as the top section), which does work.

Works great! I went straight to file: `FileOutputStream fileOS = new FileOutputStream(path); if (treewalk != null){repo.open(treewalk.getObjectId(0)).copyTo(fileOS);}` then `fileOS.close;` — Mark Mikofski, Oct 09 '13 at 07:53

score 4 · Answer 4 · answered Nov 08 '09 at 00:42

4

Figured it out by myself. The API is quite low-level, but it's not too bad:

File repoDir = new File("test-git/.git");
// open the repository
Repository repo = new Repository(repoDir);
// find the HEAD
Commit head = repo.mapCommit(Constants.HEAD);
// retrieve the tree in HEAD
Tree tree = head.getTree();
// find a file (as a TreeEntry, which contains the blob object id)
TreeEntry entry = tree.findBlobMember("b/test.txt");
// use the blob id to read the file's data
byte[] data = repo.openBlob(entry.getId()).getBytes();

answered Nov 08 '09 at 00:42

Thilo

257,207
101
511
656

4

This seems to be an outdated example with the current JGit release. The API has changed a bit, beware. – Jonathan Dumaine Mar 24 '11 at 18:33
1

@Jonathan Dumaine: Please update the post if necessary (and you know how) – Thilo Mar 25 '11 at 02:29

score 4 · Answer 5 · answered Nov 03 '11 at 14:04

I have started writing a library called gitective that contains many helpers for working with blobs, commits, and trees using JGit and is MIT-licensed and available on GitHub.

Get content of file in HEAD commit

Repository repo = new FileRepository("/repos/project/.git");
String content = BlobUtils.getHeadContent(repo, "src/Buffer.java");

Get content of a file on a branch

Repository repo = new FileRepository("/repos/project/.git");
String content = BlobUtils.getContent(repo, "master", "src/Buffer.java");

Diff two files

Repository repo = new FileRepository("/repos/project/.git");
ObjectId current = BlobUtils.getId(repo, "master", "Main.java");
ObjectId previous = BlobUtils.getId(repo, "master~1", "Main.java");
Collection<Edit> edit = BlobUtils.diff(repo, previous, current);

More examples of utilities provided are detailed in the README.

It looks nice. If I hadn't already written my jGit project I would have definitely used this. — Jason Wheeler, Jul 25 '13 at 21:58

score 3 · Answer 6 · answered Nov 07 '09 at 02:00

3

There is some info at JGit Tutorial (but that also is neither really helpful nor complete and probably outdated since they switched to eclipse where no documentation is available yet).

answered Nov 07 '09 at 02:00

jitter

53,475
11
111
124

score 3 · Answer 7 · answered Apr 05 '18 at 20:16

You can read the content of a given filepath as follows. Please be aware that the TreeWalk can be null if no path was found in the given tree. So it requires some specific handling.

public String readFile(RevCommit commit, String filepath) throws IOException {
    try (TreeWalk walk = TreeWalk.forPath(repo, filepath, commit.getTree())) {
        if (walk != null) {
            byte[] bytes = repo.open(walk.getObjectId(0)).getBytes();
            return new String(bytes, StandardCharsets.UTF_8);
        } else {
            throw new IllegalArgumentException("No path found.");
        }
    }
}

For example:

ObjectId head = repo.resolve(Constants.HEAD);
RevCommit last = repo.parseCommit(head);
readFile(last, "docs/README.md")

This answer is written with JGit 4.8.0.

How to "cat" a file in JGit?

7 Answers7

Linked