Moved files show up in deltas as two entries

[martinheidegger]

If you move a file it is recognized as one operation with clients such as Sourcetree:

but if I check diffDelta's with nodegit like this:

commit.getDiff()
  .then(function (diffs) {
    // note: I am not sure why this I an Array?! It seems to always have only one entry?
    for (var diffNr = 0; diffNr < diffs.length; diffNr++) {
      var diff = diffs[diffNr]
      var deltas = diff.numDeltas()
      for (var i = 0; i < deltas; i++) {
        var delta = diff.getDelta(i)
        console.log(delta.newFile().path(), '←', delta.oldFile.path())
      }
    }
  })

i get the output:

test/data/renamed/test-a.json ← test/data/renamed/test-a.json
test/data/renamed/test.json ← test/data/renamed/test.json

Which is weird because the API obviously says oldFile.path() but it just shows the two delta's separately?!

The only workaround I can see to fix this in my code is to compare every delta in the diff with the file I am looking for an see whether its the same file (to see whether it was a "move")

[martinheidegger]

Further research revealed following workaround:

commit.getDiff()
 .then(function (diffs) {
    for (var diffNr = 0; diffNr < diffs.length; diffNr++) {
      var diff = diffs[diffNr]
      var deltas = diff.numDeltas()
      for (var i = 0; i < deltas; i++) {
        var delta = diff.getDelta(i)
        for (var j = 0; j < deltas; j++) {
          if (j !== i) {
             // Note: the documentation of `.cmp` is rather thin. For some queer reason `.tostrS` seems to work too....
             var otherDelta = diff.getDelta(j)
             if (otherDelta.oldFile().id().cmp(delta.newFile().id()) === 0 /* why 0? */) {
                // oh! this file has moved!
                console.log(otherDelta.newFile().path(), '←', delta.oldFile().path())
                return
             }
          }
        }
        console.log(delta.newFile().path(), '←', delta.oldFile.path())
      }
    }
  })

Which is quirky and weird - to say the least.

[rcjsuen]

@martinheidegger Maybe try "compressing" your diff with the findSimilar function.

var git = require(".");
var repo_dir = "tmp";
var repo;
var diff;

git.Repository.open(repo_dir)
.then(function(_repo) {
  repo = _repo;
  return repo.getHeadCommit();
})
.then(function(c) {
  return c.getDiff();
})
.then(function(d) {
  diff = d[0];
  var opts = {
    flags: git.Diff.FIND.RENAMES
  };
  return diff.findSimilar(opts);
})
.then(function() {
  console.log(diff.numDeltas());
  var delta = diff.getDelta(0);
  console.log(delta.status() === git.Diff.DELTA.RENAMED);
})
.catch(function(err) {
  console.log(err);
});

[rcjsuen]

As to your question in your code's comment, getDiffs is returning an array because the diff is performed against the commit's parent. Most commits usually have one parent so there is usually only going to be one entry. However, if a commit is a merge commit, then it will have two (or more?) parents and then it will return individual diffs against each parent when getDiffs is called.

[wmertens]

@rcjsuen what is the standard way to show diffs from two parents? Is there a way to combine the diff? Or is it better to show diffs from both sides separately?

[wmertens]

self-answer: convention is to show the diff vs the leftmost parent. For stashes you need to add to that the stand-alone diff of the second parent, which contains untracked files