Summary

Fixes Python SDK BSON round-trip handling for Binary and Code values.

Bugs

Code is serialized as a plain string

PyMongo's bson.Code is a subclass of str. Collection.__serialize() currently checks primitive strings before looking up BSON_SERIALIZERS, so Code values never reach the existing Code: lambda v: {"$code": str(v)} serializer. They are sent as plain strings instead of BSON Code Extended JSON.

The deserializer also ignores $scope, so scoped Code values cannot round-trip.

Binary loses subtype

Binary values are currently serialized as only a hex string:

{"$binary": v.hex()}

Deserialization reconstructs them with the default subtype:

Binary(bytes.fromhex(value["$binary"]))

That silently changes UUID/vector/user-defined/etc. BSON binary values into generic subtype 0.

Fix

• Handle Code before the primitive str branch.
• Serialize Code as $code plus optional recursively serialized $scope.
• Deserialize $scope back into Code(code, scope).
• Serialize Binary as { data, subType }.
• Deserialize the new object form while preserving subtype, and keep backward compatibility for the old $binary: "hex" shape.

Validation

• Verified against mongodb/mongo-python-driver 4.11.3 source: Binary stores subtype, and Code subclasses str with an optional scope property.
• Ran python3 -m py_compile onenode-py/onenode/_collection.py.
• Ran git diff --check.
• Ran a local stubbed Python validation covering:
  • current Code string-subclass ordering issue
  • $scope serialization/deserialization
  • Binary subtype 4 serialization
  • hex-string subtype 128 deserialization
  • legacy $binary: "hex" fallback to subtype 0

Full package tests were not run because this environment does not have pymongo/bson installed and shell network cannot reach package indexes.