3699: back out attempts to change tempfile.tempdir
[arvados.git] / sdk / python / arvados / commands / keepdocker.py
index fa4e44136b974cc2dd1831248e48e0bd2218a7c8..c96e2cbd60f4a5265c5506ee059c7a86141f47b9 100644 (file)
@@ -22,19 +22,12 @@ STAT_CACHE_ERRORS = (IOError, OSError, ValueError)
 DockerImage = namedtuple('DockerImage',
                          ['repo', 'tag', 'hash', 'created', 'vsize'])
 
-opt_parser = argparse.ArgumentParser(add_help=False)
-opt_parser.add_argument(
+keepdocker_parser = argparse.ArgumentParser(add_help=False)
+keepdocker_parser.add_argument(
     '-f', '--force', action='store_true', default=False,
     help="Re-upload the image even if it already exists on the server")
 
-opt_parser.add_argument(
-    '--project-uuid',
-    help="Add the Docker image and metadata to the specified project.  Goes into user 'home' project by default.")
-opt_parser.add_argument(
-    '--name',
-    help="Name to use for the collection that will contain the docker image.")
-
-_group = opt_parser.add_mutually_exclusive_group()
+_group = keepdocker_parser.add_mutually_exclusive_group()
 _group.add_argument(
     '--pull', action='store_true', default=False,
     help="Try to pull the latest image from Docker registry")
@@ -42,16 +35,19 @@ _group.add_argument(
     '--no-pull', action='store_false', dest='pull',
     help="Use locally installed image only, don't pull image from Docker registry (default)")
 
-opt_parser.add_argument(
+keepdocker_parser.add_argument(
     'image', nargs='?',
     help="Docker image to upload, as a repository name or hash")
-opt_parser.add_argument(
+keepdocker_parser.add_argument(
     'tag', nargs='?', default='latest',
     help="Tag of the Docker image to upload (default 'latest')")
 
+# Combine keepdocker options listed above with run_opts options of arv-put.
+# The options inherited from arv-put include --name, --project-uuid,
+# --progress/--no-progress/--batch-progress and --resume/--no-resume.
 arg_parser = argparse.ArgumentParser(
         description="Upload or list Docker images in Arvados",
-        parents=[opt_parser, arv_put.run_opts])
+        parents=[keepdocker_parser, arv_put.run_opts])
 
 class DockerError(Exception):
     pass
@@ -193,10 +189,10 @@ def list_images_in_arv():
 
     st = sorted(images.items(), lambda a, b: cmp(b[1]["timestamp"], a[1]["timestamp"]))
 
-    fmt = "{:30}  {:10}  {:12}  {:38}  {:20}"
-    print fmt.format("REPOSITORY", "TAG", "IMAGE ID", "KEEP LOCATOR", "CREATED")
+    fmt = "{:30}  {:10}  {:12}  {:29}  {:20}"
+    print fmt.format("REPOSITORY", "TAG", "IMAGE ID", "COLLECTION", "CREATED")
     for i, j in st:
-        print(fmt.format(j["repo"], j["tag"], j["dockerhash"][0:11], i, j["timestamp"].strftime("%c")))
+        print(fmt.format(j["repo"], j["tag"], j["dockerhash"][0:12], i, j["timestamp"].strftime("%c")))
 
 def main(arguments=None):
     args = arg_parser.parse_args(arguments)
@@ -215,18 +211,71 @@ def main(arguments=None):
     except DockerError as error:
         print >>sys.stderr, "arv-keepdocker:", error.message
         sys.exit(1)
+
+    image_repo_tag = '{}:{}'.format(args.image, args.tag) if not image_hash.startswith(args.image.lower()) else None
+
+    if args.name is None:
+        if image_repo_tag:
+            collection_name = 'Docker image {} {}'.format(image_repo_tag, image_hash[0:12])
+        else:
+            collection_name = 'Docker image {}'.format(image_hash[0:12])
+    else:
+        collection_name = args.name
+
+    api = arvados.api('v1')
+
     if not args.force:
-        # Abort if this image is already in Arvados.
-        existing_links = arvados.api('v1').links().list(
+        # Check if this image is already in Arvados.
+
+        # Project where everything should be owned
+        parent_project_uuid = args.project_uuid if args.project_uuid else api.users().current().execute()['uuid']
+
+        # Find image hash tags
+        existing_links = api.links().list(
             filters=[['link_class', '=', 'docker_image_hash'],
                      ['name', '=', image_hash]]).execute()['items']
         if existing_links:
-            message = [
-                "arv-keepdocker: Image {} already stored in collection(s):".
-                format(image_hash)]
-            message.extend(link['head_uuid'] for link in existing_links)
-            print >>sys.stderr, "\n".join(message)
-            sys.exit(0)
+            # get readable collections
+            collections = api.collections().list(
+                filters=[['uuid', 'in', [link['head_uuid'] for link in existing_links]]],
+                select=["uuid", "owner_uuid", "name", "manifest_text"]).execute()['items']
+
+            if collections:
+                # check for repo+tag links on these collections
+                existing_repo_tag = (api.links().list(
+                    filters=[['link_class', '=', 'docker_image_repo+tag'],
+                             ['name', '=', image_repo_tag],
+                             ['head_uuid', 'in', collections]]).execute()['items']) if image_repo_tag else []
+
+                # Filter on elements owned by the parent project
+                owned_col = [c for c in collections if c['owner_uuid'] == parent_project_uuid]
+                owned_img = [c for c in existing_links if c['owner_uuid'] == parent_project_uuid]
+                owned_rep = [c for c in existing_repo_tag if c['owner_uuid'] == parent_project_uuid]
+
+                if owned_col:
+                    # already have a collection owned by this project
+                    coll_uuid = owned_col[0]['uuid']
+                else:
+                    # create new collection owned by the project
+                    coll_uuid = api.collections().create(body={"manifest_text": collections[0]['manifest_text'],
+                                                               "name": collection_name,
+                                                               "owner_uuid": parent_project_uuid},
+                                                         ensure_unique_name=True).execute()['uuid']
+
+                link_base = {'owner_uuid': parent_project_uuid,
+                             'head_uuid':  coll_uuid }
+
+                if not owned_img:
+                    # create image link owned by the project
+                    make_link('docker_image_hash', image_hash, **link_base)
+
+                if not owned_rep and image_repo_tag:
+                    # create repo+tag link owned by the project
+                    make_link('docker_image_repo+tag', image_repo_tag, **link_base)
+
+                print(coll_uuid)
+
+                sys.exit(0)
 
     # Open a file for the saved image, and write it if needed.
     outfile_name = '{}.tar'.format(image_hash)
@@ -236,15 +285,10 @@ def main(arguments=None):
 
     # Call arv-put with switches we inherited from it
     # (a.k.a., switches that aren't our own).
-    put_args = opt_parser.parse_known_args(arguments)[1]
+    put_args = keepdocker_parser.parse_known_args(arguments)[1]
 
     if args.name is None:
-        put_args += ['--name', 'Docker image {}:{} {}'.format(args.image, args.tag, image_hash[0:11])]
-    else:
-        put_args += ['--name', args.name]
-
-    if args.project_uuid is not None:
-        put_args += ['--project-uuid', args.project_uuid]
+        put_args += ['--name', collection_name]
 
     coll_uuid = arv_put.main(
         put_args + ['--filename', outfile_name, image_file.name]).strip()
@@ -263,8 +307,8 @@ def main(arguments=None):
         link_base['owner_uuid'] = args.project_uuid
 
     make_link('docker_image_hash', image_hash, **link_base)
-    if not image_hash.startswith(args.image.lower()):
-        make_link('docker_image_repo+tag', '{}:{}'.format(args.image, args.tag),
+    if image_repo_tag:
+        make_link('docker_image_repo+tag', image_repo_tag,
                   **link_base)
 
     # Clean up.