Implement the full spec for the old Docker V1 registry search API

This API is still (apparently) being used by the Docker CLI for `docker search` (why?!) and we therefore have customers expecting this to work the same way as the DockerHub.

endpoints/v1/index.py

@@ -289,27 +289,28 @@ def put_repository_auth(namespace_name, repo_name):
 def get_search():
   query = request.args.get('q')
 
+  try:
+    limit = min(100, max(1, int(request.args.get('n', 25))))
+  except ValueError:
+    limit = 25
+
+  try:
+    page = max(0, int(request.args.get('page', 1)))
+  except ValueError:
+    page = 1
+
   username = None
   user = get_authenticated_user()
   if user is not None:
     username = user.username
 
-  results = []
-  if query:
-    _conduct_repo_search(username, query, results)
-
-  data = {
-    "query": query,
-    "num_results": len(results),
-    "results" : results
-  }
-
+  data = _conduct_repo_search(username, query, limit, page)
   resp = make_response(json.dumps(data), 200)
   resp.mimetype = 'application/json'
   return resp
 
 
-def _conduct_repo_search(username, query, results):
+def _conduct_repo_search(username, query, limit=25, page=1):
   """ Finds matching repositories. """
   def can_read(repo):
     if repo.is_public:
@@ -317,13 +318,32 @@ def _conduct_repo_search(username, query, results):
 
     return ReadRepositoryPermission(repo.namespace_user.username, repo.name).can()
 
-  only_public = username is None
-  matching_repos = model.get_sorted_matching_repositories(query, only_public, can_read, limit=5)
+  # Note: We put a max 5 page limit here. The Docker CLI doesn't seem to use the
+  # pagination and most customers hitting the API should be using V2 catalog, so this
+  # is a safety net for our slow search below, since we have to use the slow approach
+  # of finding *all* the results, and then slicing in-memory, because this old API requires
+  # the *full* page count in the returned results.
+  _MAX_PAGE_COUNT = 5
+  page = min(page, _MAX_PAGE_COUNT)
 
-  for repo in matching_repos:
+  only_public = username is None
+  matching_repos = model.get_sorted_matching_repositories(query, only_public, can_read,
+                                                          limit=limit*_MAX_PAGE_COUNT)
+  results = []
+  for repo in matching_repos[(page - 1) * _MAX_PAGE_COUNT:limit]:
     results.append({
       'name': repo.namespace_name + '/' + repo.name,
       'description': repo.description,
       'is_public': repo.is_public,
       'href': '/repository/' + repo.namespace_name + '/' + repo.name
     })
+
+  # Defined: https://docs.docker.com/v1.6/reference/api/registry_api/
+  return {
+    'query': query,
+    'num_results': len(results),
+    'num_pages': (len(matching_repos) / limit) + 1,
+    'page': page,
+    'page_size': limit,
+    'results': results,
+  }

test/registry_tests.py

@@ -1289,6 +1289,31 @@ class V1RegistryTests(V1RegistryPullMixin, V1RegistryPushMixin, RegistryTestsMix
     self.assertEquals(1, len(data['results']))
 
 
+  def test_search_pagination(self):
+    # Check for the first page.
+    resp = self.conduct('GET', '/v1/search', params=dict(q='s', n='1'))
+    data = resp.json()
+    self.assertEquals('s', data['query'])
+
+    self.assertEquals(1, data['num_results'])
+    self.assertEquals(1, len(data['results']))
+
+    self.assertEquals(1, data['page'])
+    self.assertTrue(data['num_pages'] > 1)
+
+    # Check for the followup pages.
+    for page_index in range(1, data['num_pages']):
+      resp = self.conduct('GET', '/v1/search', params=dict(q='s', n='1', page=page_index))
+      data = resp.json()
+      self.assertEquals('s', data['query'])
+
+      self.assertEquals(1, data['num_results'])
+      self.assertEquals(1, len(data['results']))
+
+      self.assertEquals(1, data['page'])
+      self.assertTrue(data['num_pages'] > 1)
+
+
   def test_users(self):
     # Not logged in, should 404.
     self.conduct('GET', '/v1/users', expected_code=404)