Caching Strategy for location requests

Question

I am building REST APIs that return data (lets say events ) in particular area. The REST URL is a simple GET

/api/v1/events?lat=<lat>&lng=<lng>&radius=<radius>.

with parameters lat, lng and radius (10 miles by default), the latitude and longitude are what the device or browser APIs return. Now needless to say that the lat and lng change continuously as the user moves and also two users can be same vicinity with different lat / lng. What is the best way to cache such kind of requests on the server so that I don't have to dip into business logic everytime. The URL is not going to unique since lat/lng change.

Thanks

Answer 1

I'm assuming you have some sort of "grid", and when a user requests a specific coordinate, you return the grid tile(s) around the location. So you have an infinite URL space (coordinates) that is mapped to a finite number of tiles. One solution is to redirect every request to the "canonical", cache friendly URL for that tile, e.g.

GET /api/v1/events?lat=123&lng=456
=>
302 Found
Location: /api/v1/events?tile=abc

Or, if you want to retain the lat/long info in the URL, you could use the location of the center of the tile.

Answer 2

I think the best approach is for you to store the results in a cache with the center coordinates as a key, and later query the points within the circle for the new request.

I'm not aware of any cache engines that would allow you to perform spatial queries, so I think you'll have to use a database that allows easy querying and indexing of spatial data. You may use that database for caching your results, or at least store a key to that result in a cache engine somewhere else, and later you can query them with spatial coordinates, asking for all points with a threshold distance to your new request.

There's PostGis for PostgreSQL, which should be quite straightforward since it has full support for latitude/longitude distance computations. Once you have it setup with proper indexes, it should be as easy as:

SELECT * FROM your_cache_table 
WHERE ST_Distance_Sphere(the_geom, ST_MakePoint(new_lon, new_lat)) <= 160.934

MySQL has some support for the OpenGis extensions, however it doesn't have support for latitude/longitude distance computations. Maybe you'll need to do some calculations by yourself, maybe the simple cartesian distance works for you. Check the documentation here, and this answer should also help.

I also believe even MySQL 5.6 still has support for spatial indexes only in MyISAM tables, but that shouldn't be an issue since you're using them only for cache.

Managing the cache may be a little more complicated than usual. If you need expiration, you should probably store only keys in the database and set an expire parameter on the cache server. When you hit a database point for which there's no longer a valid key, you clean it from the database. You'll probably need a way to invalidate cache when the primary data changes, removing from both the database and the cache server.