Make queries per-table

[rmarianski]

Connects to https://github.com/mapzen/tile-tasks/issues/267

[pnorman]

404 error, could you provide some details on what it connects to?

[nvkelso]

Matt can provide more details Monday, but it's baby steps away from Postgres and then full separation using raw data file assets on S3.

…

On Jul 28, 2017, at 18:25, Paul Norman ***@***.***> wrote: 404 error, could you provide some details on what it connects to? ― You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHub, or mute the thread.

[zerebubuth]

I've added more information to the README for raw tiles, which is still on a branch for the moment. That describes the overall goals and what we're trying to achieve, but it's also useful to know how we're getting there.

We've split the work up into several "phases", of which one is the tile-tasks issue linked to above (phase 2). The summary is:

1. Phase 1 was about being able to calculate in Python all the min_zoom and feature properties that we're currently calculating in SQL (generated from YAML for the bonus confusement), and has now finished.
2. Phase 2 is about querying the database on a per-table basis rather than a per-layer basis.
3. Phase 3 is about generating flat files (RAW(R) tiles) for each table at zoom 10, and querying those instead of hitting the database.
4. Phase 4 extends that to all zoom levels.

The main benefits (in more detail in that README) are that the PostgreSQL server is no longer on any "hot path", so doesn't ever become a bottleneck, and that we should be able to run multiple different styles/versions from the same infrastructure stack.