PYTHON-3298 Add flag to create_collection to skip listCollections pre-check (#1006)

Author: juliusgeo

doc/changelog.rst

@@ -13,6 +13,9 @@ PyMongo 4.2 brings a number of improvements including:
   changes may be made before the final release.  See :ref:`automatic-queryable-client-side-encryption` for example usage.
 - Provisional (beta) support for :func:`pymongo.timeout` to apply a single timeout
   to an entire block of pymongo operations.
+- Added ``check_exists`` option to :meth:`~pymongo.database.Database.create_collection`
+  that when True (the default)  runs an additional ``listCollections`` command to verify that the
+  collection does not exist already.
 
 Bug fixes
 .........

pymongo/database.py

@@ -305,6 +305,7 @@ def create_collection(
         read_concern: Optional["ReadConcern"] = None,
         session: Optional["ClientSession"] = None,
         timeout: Optional[float] = None,
+        check_exists: Optional[bool] = True,
         **kwargs: Any,
     ) -> Collection[_DocumentType]:
         """Create a new :class:`~pymongo.collection.Collection` in this
@@ -336,6 +337,8 @@ def create_collection(
             :class:`~pymongo.collation.Collation`.
           - `session` (optional): a
             :class:`~pymongo.client_session.ClientSession`.
+          - ``check_exists`` (optional): if True (the default), send a listCollections command to
+            check if the collection already exists before creation.
           - `**kwargs` (optional): additional keyword arguments will
             be passed as options for the `create collection command`_
 
@@ -402,7 +405,7 @@ def create_collection(
             enabling pre- and post-images.
 
         .. versionchanged:: 4.2
-           Added the ``clusteredIndex`` and ``encryptedFields`` parameters.
+           Added the ``check_exists``, ``clusteredIndex``, and  ``encryptedFields`` parameters.
 
         .. versionchanged:: 3.11
            This method is now supported inside multi-document transactions
@@ -441,8 +444,10 @@ def create_collection(
         with self.__client._tmp_session(session) as s:
             # Skip this check in a transaction where listCollections is not
             # supported.
-            if (not s or not s.in_transaction) and name in self.list_collection_names(
-                filter={"name": name}, session=s
+            if (
+                check_exists
+                and (not s or not s.in_transaction)
+                and name in self.list_collection_names(filter={"name": name}, session=s)
             ):
                 raise CollectionInvalid("collection %s already exists" % name)
             return Collection(

test/test_database.py

@@ -220,6 +220,21 @@ def test_list_collection_names_filter(self):
             self.assertIn("nameOnly", command)
             self.assertTrue(command["nameOnly"])
 
+    def test_check_exists(self):
+        listener = OvertCommandListener()
+        results = listener.results
+        client = rs_or_single_client(event_listeners=[listener])
+        self.addCleanup(client.close)
+        db = client[self.db.name]
+        db.drop_collection("unique")
+        db.create_collection("unique", check_exists=True)
+        self.assertIn("listCollections", listener.started_command_names())
+        listener.reset()
+        db.drop_collection("unique")
+        db.create_collection("unique", check_exists=False)
+        self.assertTrue(len(results["started"]) > 0)
+        self.assertNotIn("listCollections", listener.started_command_names())
+
     def test_list_collections(self):
         self.client.drop_database("pymongo_test")
         db = Database(self.client, "pymongo_test")

test/unified_format.py

@@ -283,7 +283,6 @@ def __init__(
             self._observe_sensitive_commands = False
             self._ignore_commands = _SENSITIVE_COMMANDS | set(ignore_commands)
             self._ignore_commands.add("configurefailpoint")
-        self.ignore_list_collections = False
         self._event_mapping = collections.defaultdict(list)
         self.entity_map = entity_map
         if store_events:
@@ -314,10 +313,7 @@ def add_event(self, event):
             )
 
     def _command_event(self, event):
-        if not (
-            event.command_name.lower() in self._ignore_commands
-            or (self.ignore_list_collections and event.command_name == "listCollections")
-        ):
+        if not event.command_name.lower() in self._ignore_commands:
             self.add_event(event)
 
     def started(self, event):
@@ -1032,13 +1028,8 @@ def _databaseOperation_listCollections(self, target, *args, **kwargs):
 
     def _databaseOperation_createCollection(self, target, *args, **kwargs):
         # PYTHON-1936 Ignore the listCollections event from create_collection.
-        for listener in target.client.options.event_listeners:
-            if isinstance(listener, EventListenerUtil):
-                listener.ignore_list_collections = True
+        kwargs["check_exists"] = False
         ret = target.create_collection(*args, **kwargs)
-        for listener in target.client.options.event_listeners:
-            if isinstance(listener, EventListenerUtil):
-                listener.ignore_list_collections = False
         return ret
 
     def __entityOperation_aggregate(self, target, *args, **kwargs):

test/utils.py

@@ -202,23 +202,14 @@ class OvertCommandListener(EventListener):
     ignore_list_collections = False
 
     def started(self, event):
-        if self.ignore_list_collections and event.command_name.lower() == "listcollections":
-            self.ignore_list_collections = False
-            return
         if event.command_name.lower() not in _SENSITIVE_COMMANDS:
             super(OvertCommandListener, self).started(event)
 
     def succeeded(self, event):
-        if self.ignore_list_collections and event.command_name.lower() == "listcollections":
-            self.ignore_list_collections = False
-            return
         if event.command_name.lower() not in _SENSITIVE_COMMANDS:
             super(OvertCommandListener, self).succeeded(event)
 
     def failed(self, event):
-        if self.ignore_list_collections and event.command_name.lower() == "listcollections":
-            self.ignore_list_collections = False
-            return
         if event.command_name.lower() not in _SENSITIVE_COMMANDS:
             super(OvertCommandListener, self).failed(event)
 
@@ -1114,6 +1105,7 @@ def prepare_spec_arguments(spec, arguments, opname, entity_map, with_txn_callbac
         elif opname == "create_collection":
             if arg_name == "collection":
                 arguments["name"] = arguments.pop(arg_name)
+            arguments["check_exists"] = False
             # Any other arguments to create_collection are passed through
             # **kwargs.
         elif opname == "create_index" and arg_name == "keys":

test/utils_spec_runner.py

@@ -307,15 +307,7 @@ def run_operation(self, sessions, collection, operation):
             args.update(arguments)
             arguments = args
 
-        try:
-            if name == "create_collection" and (
-                "encrypted" in operation["arguments"]["name"]
-                or "plaintext" in operation["arguments"]["name"]
-            ):
-                self.listener.ignore_list_collections = True
-            result = cmd(**dict(arguments))
-        finally:
-            self.listener.ignore_list_collections = False
+        result = cmd(**dict(arguments))
         # Cleanup open change stream cursors.
         if name == "watch":
             self.addCleanup(result.close)