From 823077aa2b998781957fd4dde2afe356a0d6ebda Mon Sep 17 00:00:00 2001
From: "Mr. Outis" <mroutis@protonmail.com>
Date: Fri, 18 Jan 2019 23:05:00 -0600
Subject: [PATCH] stage: Checkout cached output when possible

Fix #1348
---
 dvc/output/base.py  |  8 ++++++
 dvc/stage.py        | 18 ++++++++++++--
 tests/test_repro.py | 59 +++++++++++++++++++++++++++++++++++++++++++--
 3 files changed, 81 insertions(+), 4 deletions(-)

diff --git a/dvc/output/base.py b/dvc/output/base.py
index b52727d58..3b8b2b445 100644
--- a/dvc/output/base.py
+++ b/dvc/output/base.py
@@ -113,6 +113,14 @@ class OutputBase(object):
         return getattr(self.project.cache, self.scheme).changed(self.path_info,
                                                                 self.info)
 
+    def changed_cache(self):
+        if not self.use_cache or not self.checksum:
+            return True
+
+        cache = self.project.cache.__getattribute__(self.scheme)
+
+        return cache.changed_cache(self.checksum)
+
     def status(self):
         if self.changed():
             # FIXME better msgs
diff --git a/dvc/stage.py b/dvc/stage.py
index 7bb718f7e..024c24efb 100644
--- a/dvc/stage.py
+++ b/dvc/stage.py
@@ -545,7 +545,10 @@ class Stage(object):
             logger.info("Importing '{dep}' -> '{out}'"
                         .format(dep=self.deps[0].path, out=self.outs[0].path))
             if not dry:
-                self.deps[0].download(self.outs[0].path_info)
+                if self._already_cached():
+                    self.outs[0].checkout()
+                else:
+                    self.deps[0].download(self.outs[0].path_info)
 
         elif self.is_data_source:
             msg = u"Verifying data sources in '{}'".format(self.relpath)
@@ -556,7 +559,10 @@ class Stage(object):
         else:
             logger.info('Running command:\n\t{}'.format(self.cmd))
             if not dry:
-                self._run()
+                if self._already_cached():
+                    self.checkout()
+                else:
+                    self._run()
 
         if not dry:
             self.save()
@@ -598,3 +604,11 @@ class Stage(object):
             return {self.relpath: ret}
 
         return {}
+
+    def _already_cached(self):
+        return (not self.changed_md5()
+                and all(not dep.changed() for dep in self.deps)
+                and all(not out.changed_cache()
+                        if out.use_cache
+                        else not out.changed()
+                        for out in self.outs))
diff --git a/tests/test_repro.py b/tests/test_repro.py
index 33ac54108..871e71b32 100644
--- a/tests/test_repro.py
+++ b/tests/test_repro.py
@@ -617,7 +617,57 @@ class TestReproExternalBase(TestDvc):
     def sep(self):
         return '/'
 
-    def test(self):
+    def check_already_cached(self, stage):
+        stage.outs[0].remove()
+
+        patch_download = patch.object(stage.deps[0],
+                                      'download',
+                                      wraps=stage.deps[0].download)
+
+        patch_checkout = patch.object(stage.outs[0],
+                                      'checkout',
+                                      wraps=stage.outs[0].checkout)
+
+        patch_run = patch.object(stage, '_run', wraps=stage._run)
+
+        with self.dvc.state, \
+                patch_download as mock_download, \
+                patch_checkout as mock_checkout, \
+                patch_run as mock_run:
+
+            stage.run()
+
+            mock_run.assert_not_called()
+            mock_download.assert_not_called()
+            mock_checkout.assert_called_once()
+
+    def corrupted_cache(self):
+        os.unlink('bar.dvc')
+
+        stage = self.dvc.run(deps=[self.FOO],
+                             outs=[self.BAR],
+                             cmd='echo bar > bar')
+
+        with open(self.BAR, 'w') as fd:
+            fd.write('corrupting the cache')
+
+        patch_checkout = patch.object(stage.outs[0],
+                                      'checkout',
+                                      wraps=stage.outs[0].checkout)
+
+        patch_run = patch.object(stage, '_run', wraps=stage._run)
+
+        with self.dvc.state, \
+                patch_checkout as mock_checkout, \
+                patch_run as mock_run:
+
+            stage.run()
+
+            mock_run.assert_called_once()
+            mock_checkout.assert_not_called()
+
+    @patch('dvc.prompt.confirm', return_value=True)
+    def test(self, mock_prompt):
         if not self.should_test():
             return
 
@@ -654,9 +704,11 @@ class TestReproExternalBase(TestDvc):
         self.write(self.bucket, foo_key, self.FOO_CONTENTS)
 
         import_stage = self.dvc.imp(out_foo_path, 'import')
+
         self.assertTrue(os.path.exists('import'))
         self.assertTrue(filecmp.cmp('import', self.FOO, shallow=False))
         self.assertEqual(self.dvc.status(import_stage.path), {})
+        self.check_already_cached(import_stage)
 
         import_remote_stage = self.dvc.imp(out_foo_path,
                                            out_foo_path + '_imported')
@@ -665,9 +717,10 @@ class TestReproExternalBase(TestDvc):
         cmd_stage = self.dvc.run(outs=[out_bar_path],
                                  deps=[out_foo_path],
                                  cmd=self.cmd(foo_path, bar_path))
-        self.assertEqual(self.dvc.status(cmd_stage.path), {})
 
+        self.assertEqual(self.dvc.status(cmd_stage.path), {})
         self.assertEqual(self.dvc.status(), {})
+        self.check_already_cached(cmd_stage)
 
         self.write(self.bucket, foo_key, self.BAR_CONTENTS)
 
@@ -697,6 +750,8 @@ class TestReproExternalBase(TestDvc):
         self.dvc.checkout(cmd_stage.path, force=True)
         self.assertEqual(self.dvc.status(cmd_stage.path), {})
 
+        self.corrupted_cache()
+
 
 class TestReproExternalS3(TestReproExternalBase):
     def should_test(self):