[Tile-serving] [osm2pgsql] Clustering strategies (#208)

[Paul Norman]

@cquest, do you have any data on how [`maxchars`](http://postgis.net/docs/manual-2.1/ST_GeoHash.html) impacts either clustering time or resulting rendering speed?

-----

For those who are not familiar, a GeoHash is the latitude and longitude each encoded as a binary fraction of the entire range, interleaved starting with longitude, and expressed in a base 32 encoding. This is the same in principle used in quadtiles, and I think identical to the quad_tile encoding used by the main API, save that that is expressed as a single number rather than multiple base-32 digits.

We can't expect latlong quad tiles to line up with web mercator xyz tiles exactly (projections), but we can get an order of magnitude comparison. 

A standard (8x8) z19 metatiling is the earth divided into 2^16 segments each way. A 8 character geohash is 2^20 each way. So there are theoretical reasons to not expect major gains past 8 characters, as any additional data locality is "wasted" as all the data would need to be retrieved anyways, with only some minor improvements at edges. 

---
Reply to this email directly or view it on GitHub:
https://github.com/openstreetmap/osm2pgsql/issues/208#issuecomment-62293364
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.openstreetmap.org/pipermail/tile-serving/attachments/20141108/94f25bf1/attachment.html>