As noted in mongodb-user (https://groups.google.com/d/msg/mongodb-user/7bt38w9RFFU/VmUJQZ0pld0J) it would be useful to have a server-side $merge operator for use when updating documents with nested sub-documents.

From the group:

Say I have this object:

> obj = { "subobj" : { "name": "mysub", "mymap" : { "a" : { "id": "a_obj" },
                                                    "b" : { "id" : "b_obj" },
                                                    "c": {"id": "c_obj" } } } }

And I insert it into my collection:

> db.xtest.insert(obj)

And now I want to update the "id" of the "b" field in the embedded "mymap".
I know I can do this with dot notation:

> db.xtest.update({"_id" : ObjectId("55146d3e497e7087b30da8cf") },
                  { "$set" : { "subobj.mymap.b.id" : "new_b_id" } } )

The dot notation becomes tedious pretty quickly, as you need to write code to flatten the sub-objects to generate the update values. When subobj is complex, the query can become large very quickly.

This would be a win for consistency and efficiency (presumably a smaller update query would go out over the wire)

Suggested syntax:

 db.xtest.update({"_id" : ObjectId("55146d3e497e7087b30da8cf") },
                  { "$merge" : { "subobj": {"mymap" : { "a" : { "id": "a_obj" },
                                                    "b" : { "id" : "b_obj" },
                                                    "c": {"id": "c_obj" } } )

Behaviour:

• Existing values not referenced in the $merge document would be unaffected
• Values that do not exist in the DB version of the document would be created
• Values that do exist in the DB version of the document would be updated to the value in the $merge document
• Atomic operation