Streamingfeatures

[bitner]

Test to see if performance can be gained and memory usage minimized by using cursors and streaming response for all outputs.

[bitner]

@vincentsarago This is super ugly right now, but looks possibly very promising.

This separates out the count to a separate query and by default will just use an estimated count for the numberMatched total.

For returning GeoJson and GeoJsonSeq, rather than creating the FeatureCollection on the database, it gets the Features as just straight text and returns them directly as a streaming response using a cursor so that it can grab data a chunk at a time. For GeoJson, it just returns the FeatureCollection wrapper before and after yielding the Features. Right now, it doesn't have any of the link stuff in there, so it is just taking geojson as text from the database and directly yielding it rather than decoding/encoding the json -- adding the links back in will probably add a chunk of the overhead.

Using the countries sample dataset, it returns the geojson of all countries (http://localhost:8000/collections/public.countries/items?f=geojson&limit=1000) in 270ms while on master, the same query takes 6.86 seconds.