Let's say you have a collection of animal sightings partitioned on
species. Now you want to add a derivation that only wants to process sightings of
felis catus. The derivation could read every single document in the collection and simply filter out those that it doesn't care about. But doing so means that you use a lot of compute and network capacity reading all the documents that you end up filtering out. Derivations may define a partition selector instead to only read from the partitions they actually care about. For example, you might use the following selector in a derivation if you only wanted to process sightings of house cats.