Re: Optimize sub query with large group by

[Alex K]

I don't use object_id anywhere else because I don't care which object it was. I just want to make sure that only the latest person to access each object gets credit.

The innermost query sorts all the of the events by date (most recent first). The query outside that uses the group by trick to select only the first event for each object. The outermost query then counts up how many per person.

I am not sure how this would benefit from compound indexes as each query is only grouped/ordered by one column and there are no where clauses.

Should I just expect that this sort of query is bound to take several minutes when there are around 10 million events.