Last Updated: February 25, 2016
·
448
· st3v3nhunt

Do not include $near in count() queries

Using MongoDB to build an API providing some search and filtering functionality. One of the searches was using the $near query. As part of the call the total number of documents matching the query also needed to be returned.

There was a query builder responsible for building the queries. Rather than distinguish between a query for a count and a query to return the documents the same query was built and used for both operations. During performance testing there was a noticeable delay in returning data for calls including $near queries.

A bit of investigation and a little more clearer thinking and the $near query element being included within the count() query makes no sense whatsoever. Plus, it is no good for performance. No reduction in the number of results occurs through $near so why would it be included within count() queries? Well, it should not be.

Removing $near from count() has improved the performance markedly. In this specific case it reduced the time taken to run the query by at least an order of magnitude.

A silly oversight that should have been picked up during a code review and one that could have proved a little more costly than it did, had it not been found through performance testing. Interesting nevertheless.