Using maxTimeMS parameter with aggregation queries

问题:

• I'm unable to use maxTimeMS parameter with Mongo 2.6 and Pymongo 2.7.1

• As per the documentation on this page Official Mongodb Aggregation Page the aggregation method should return a Cursor object. However, when I run the query locally on a mongod instance (2.6+) with pymongo 2.7.1, I get a dict object!

In [14]: obj = coll.aggregate({'$group': {'_id': '$l', 'n': {'$sum': 1}}})

In [15]: type(obj) Out[15]: dict

Can anyone help me understand what is happening here?

回答1:

Yes, you can use maxTimeMS with pymongo aggregation.

c.foo.bar.aggregate([], maxTimeMS=1000)
{u'ok': 1.0, u'result': []}

If you want a cursor:

for result in c.foo.bar.aggregate([], cursor={}, maxTimeMS=1000):
... print result

The aggregate command didn't support cursors before MongoDB 2.6 so it had to be added as an option to avoid breaking existing applications.

回答2:

This is covered in the driver documentation where it is described that in order to return a cursor, you need to specify the arguments in addition to the pipeline in your .aggregate() method:

cursor = coll.aggregate([{'$group': { '_id': '$l', 'n': {'$sum': 1} }}],cursor={})

Note that the returned object here is a CommandCursor and not a cursor.

This is because various modifiers such as .limit() and .skip() and other options do not apply in the context of an aggregation result. As such $maxTimeMS is not a valid option for this type of cursor.

In addition, it would not do what you think it will even where valid. The reason being that the "cursor" execution is only counted "after" the "aggregation pipeline" execution is complete, so in this case, just fetching the results.

Look at the .currentOp() and .killOp() implementations for other ways to control long running aggregation tasks.