Add check for ref and deep clone

If the ref does not exist in the shallow clone, create a full clone of
the repository and retry.

repo2docker/app.py

@@ -91,24 +91,62 @@ class Repo2Docker(Application):
     )
 
     def fetch(self, url, ref, checkout_path):
+        def _clone(depth=None):
+            if depth is not None:
+                command = ['git', 'clone', '--depth', str(depth),
+                           url, checkout_path]
+            else:
+                command = ['git', 'clone', url, checkout_path]
+
             try:
-            for line in execute_cmd(['git', 'clone', '--depth', '50',
+                for line in execute_cmd(command, capture=self.json_logs):
-                                     url, checkout_path],
-                                    capture=self.json_logs):
                     self.log.info(line, extra=dict(phase='fetching'))
             except subprocess.CalledProcessError:
-            self.log.error('Failed to clone repository!', extra=dict(phase='failed'))
+                self.log.error('Failed to clone repository!',
+                               extra=dict(phase='failed'))
                 sys.exit(1)
 
-        if ref:
+        def _unshallow():
             try:
-                for line in execute_cmd(['git', 'reset', '--hard', ref], cwd=checkout_path,
+                for line in execute_cmd(['git', 'fetch', '--unshallow'],
+                                        capture=self.json_logs,
+                                        cwd=checkout_path):
+                    self.log.info(line, extra=dict(phase='fetching'))
+            except subprocess.CalledProcessError:
+                self.log.error('Failed to unshallow repository!',
+                               extra=dict(phase='failed'))
+                sys.exit(1)
+
+        def _contains(ref):
+            try:
+                for line in execute_cmd(['git', 'cat-file', '-t', ref],
+                                        capture=self.json_logs,
+                                        cwd=checkout_path):
+                    self.log.debug(line, extra=dict(phase='fetching'))
+            except subprocess.CalledProcessError:
+                return False
+
+            return True
+
+        def _checkout(ref):
+            try:
+                for line in execute_cmd(['git', 'reset', '--hard', ref],
+                                        cwd=checkout_path,
                                         capture=self.json_logs):
                     self.log.info(line, extra=dict(phase='fetching'))
             except subprocess.CalledProcessError:
-                self.log.error('Failed to check out ref %s', ref, extra=dict(phase='failed'))
+                self.log.error('Failed to check out ref %s', ref,
+                               extra=dict(phase='failed'))
                 sys.exit(1)
 
+        # create a shallow clone first
+        _clone(depth=50)
+        if ref:
+            if not _contains(ref):
+                # have to create a full clone
+                _unshallow()
+            _checkout(ref)
+
     def get_argparser(self):
         argparser = argparse.ArgumentParser()
         argparser.add_argument(

tests/external/datasci-shallow-clone.repos.yaml

@@ -0,0 +1,6 @@
+# Check that we correctly detect that this ref is more than 50 commits ago
+# and trigger a full clone of the repository
+- name: Jake's Data Science Book
+  url: https://github.com/jakevdp/PythonDataScienceHandbook
+  ref: 8761de29a853f0c187286b7c7bc1e4767e7c5574
+  verify: python -c 'import matplotlib'