2 ms·

Before I even clocked on this I knew it would be in Tower Hamlets, famous for being the most densely populated council in the UK... I wonder what a quadtree ba

Before I even clocked on this I knew it would be in Tower Hamlets, famous for being the most densely populated council in the UK... I wonder what a quadtree based approach would look like. I mean... I'd almost guarantee you're going to wind up in Tower Hamlets again. So the question of how much it costs to optimize for so little is there. But still, interesting alternative method for someone to gnaw on if they wanted.

fifilura 2y ago
How i would do it: 1. Create a table with quad tiles of appropriate size/level as Parquet. Put it in s3 and map it as a table in AWS Athena (Trino) 2. Put the census data in s3/Athena. 3. Join using Trino geospatial functions. https://trino.io/docs/current/functions/geospatial.html https://trino.io/docs/current/functions/geospatial.html The result for all quad tiles will be there within a couple of minutes.
- jeffbee 2y ago
  Why though? We're talking about much less than a million geothings here. Any laptop can process this much data at interactive speeds.
  - fifilura 2y ago
    True. 0. Look at data size.