diff: analyse working tree when dirty (#3534)

dvc/repo/diff.py

@@ -3,6 +3,7 @@ import os
 from dvc.exceptions import DvcException
 from dvc.repo import locked
 from dvc.scm.git import Git
+from dvc.scm.tree import is_working_tree
 
 
 @locked
@@ -36,10 +37,23 @@ def diff(self, a_rev="HEAD", b_rev=None):
                 else os.path.join(str(output), "")
             )
 
+        on_working_tree = is_working_tree(self.tree)
+
+        def _to_checksum(output):
+            if on_working_tree:
+                return self.cache.local.get_checksum(output.path_info)
+            return output.checksum
+
+        def _exists(output):
+            if on_working_tree:
+                return output.exists
+            return True
+
         return {
-            _to_path(output): output.checksum
+            _to_path(output): _to_checksum(output)
             for stage in self.stages
             for output in stage.outs
+            if _exists(output)
         }
 
     working_tree = self.tree

tests/func/test_diff.py

@@ -159,21 +159,51 @@ def test_directories(tmp_dir, scm, dvc):
 
 
 def test_diff_no_cache(tmp_dir, scm, dvc):
-    (stage,) = tmp_dir.dvc_gen(
-        {"dir": {"file": "file content"}}, commit="first"
-    )
+    tmp_dir.dvc_gen({"dir": {"file": "file content"}}, commit="first")
     scm.tag("v1")
+
     tmp_dir.dvc_gen(
         {"dir": {"file": "modified file content"}}, commit="second"
     )
+    scm.tag("v2")
 
-    remove(first(stage.outs).cache_path)
-    remove("dir")
+    remove(dvc.cache.local.cache_dir)
 
     # invalidate_dir_info to force cache loading
     dvc.cache.local._dir_info = {}
 
-    diff = dvc.diff("v1")
+    diff = dvc.diff("v1", "v2")
     assert diff["added"] == []
     assert diff["deleted"] == []
     assert first(diff["modified"])["path"] == os.path.join("dir", "")
+
+
+def test_diff_dirty(tmp_dir, scm, dvc):
+    tmp_dir.dvc_gen(
+        {"file": "file_content", "dir": {"dir_file1": "dir file content"}},
+        commit="initial",
+    )
+
+    (tmp_dir / "file").unlink()
+    tmp_dir.gen({"dir": {"dir_file2": "dir file 2 content"}})
+    tmp_dir.dvc_gen("new_file", "new_file_content")
+
+    result = dvc.diff()
+
+    assert result == {
+        "added": [
+            {"hash": "86d049de17c76ac44cdcac146042ec9b", "path": "new_file"}
+        ],
+        "deleted": [
+            {"hash": "7f0b6bb0b7e951b7fd2b2a4a326297e1", "path": "file"}
+        ],
+        "modified": [
+            {
+                "hash": {
+                    "new": "38175ad60f0e58ac94e0e2b7688afd81.dir",
+                    "old": "92daf39af116ca2fb245acaeb2ae65f7.dir",
+                },
+                "path": os.path.join("dir", ""),
+            }
+        ],
+    }