[Mariehane]

This does some clever stuff in the background, e.g. query optimization and performing the computations of a 'for'-loop concurrently (potentially over multiple machines in a cluster). So in my experience there was a bit of latency for the query to execute, but then it executes faster and on much larger datasets than regular JS could handle.

Just like how you could easily manipulate tabular data using numpy or pandas (or excel), but SQL allows you to do it declaratively, which has benefits in some cases.

[mr_toad]

CouchDB has been running distributed Map/Reduce queries on JSON data for over a decade. It’ll be interesting to see how this compares.