Merge pull request #787 from efiop/master

dvc: support dependecy files under tracked directory

dvc/dependency/base.py

@@ -39,6 +39,10 @@ class DependencyBase(object):
     def supported(cls, url):
         return cls.match(url) != None
 
+    @property
+    def sep(self):
+        return '/'
+
     def changed(self):
         raise NotImplementedError
 

dvc/dependency/local.py

@@ -39,6 +39,10 @@ class DependencyLOCAL(DependencyBase):
         self.path_info = {'scheme': 'local',
                           'path': self.path}
 
+    @property
+    def sep(self):
+        return os.sep
+
     @property
     def rel_path(self):
         return os.path.relpath(self.path)

dvc/project.py

@@ -399,22 +399,20 @@ class Project(object):
     def graph(self):
         G = nx.DiGraph()
         stages = self.stages()
-
-        outs_map = {}
-        for stage in stages:
-            for o in stage.outs:
-                outs_map[o.path] = stage
+        outs = self.outs()
 
         for stage in stages:
             node = os.path.relpath(stage.path, self.root_dir)
             G.add_node(node, stage=stage)
             for dep in stage.deps:
-                dep_stage = outs_map.get(dep.path, None)
-                if not dep_stage:
-                    continue
-                dep_node = os.path.relpath(dep_stage.path, self.root_dir)
-                G.add_node(dep_node, stage=dep_stage)
-                G.add_edge(node, dep_node)
+                for out in outs:
+                    if out.path != dep.path and not dep.path.startswith(out.path + out.sep):
+                        continue
+
+                    dep_stage = out.stage
+                    dep_node = os.path.relpath(dep_stage.path, self.root_dir)
+                    G.add_node(dep_node, stage=dep_stage)
+                    G.add_edge(node, dep_node)
 
         return G
 

tests/test_repro.py

@@ -36,6 +36,27 @@ class TestRepro(TestDvc):
                      cmd='python {} {} {}'.format(self.CODE, self.FOO, self.file1))
 
 
+class TestReproDepUnderDir(TestDvc):
+    def test(self):
+        self.dir_stage = self.dvc.add(self.DATA_DIR)
+
+        self.file1 = 'file1'
+        self.file1_stage = self.file1 + '.dvc'
+        self.dvc.run(fname=self.file1_stage,
+                     outs=[self.file1],
+                     deps=[self.DATA, self.CODE],
+                     cmd='python {} {} {}'.format(self.CODE, self.DATA, self.file1))
+
+        self.assertTrue(filecmp.cmp(self.file1, self.DATA))
+
+        os.unlink(self.DATA)
+        shutil.copyfile(self.FOO, self.DATA)
+
+        stages = self.dvc.reproduce(self.file1_stage)
+        self.assertEqual(len(stages), 2)
+        self.assertTrue(filecmp.cmp(self.file1, self.FOO))
+
+
 class TestReproNoDeps(TestRepro):
     def test(self):
         out = 'out'