joto left a comment (osm2pgsql-dev/osm2pgsql#2482)
> If performance is a problem, I know for sure I could make it much faster if
> the generalization could receive the exact ways that were edited, like
> carto's flex lua code, rather than Z/X/Y tiles. That way the slow part (the
> &&tile_box) can be skipped entirely, and the remaining steps would only walk
> the network of the actually edited way(s).
You don't need to base this on tiles, you can base this on the ways that
changed! For instance, you can have a separate table where you store (the "way"
processing function) all the ids of ways that changed, then use that table as
input for the rest of the processing in the "gen" step. But it is not as easy
as one would think to implement this. You have to correctly handle deleted ways
and all the different kinds of change, including cases like this: Say you have
two ways that are not connected, now a third way is changed which makes it
connect to both existing and unchanged ways allowing all of them to be merged.
> so given multithreading this should be basically fine I'd hope
All of this is I/O bound so multithreading is not going to help.
--
Reply to this email directly or view it on GitHub:
https://github.com/osm2pgsql-dev/osm2pgsql/pull/2482#issuecomment-4612569189
You are receiving this because you are subscribed to this thread.
Message ID: <osm2pgsql-dev/osm2pgsql/pull/2482/[email protected]>
_______________________________________________
Tile-serving mailing list
[email protected]
https://lists.openstreetmap.org/listinfo/tile-serving
