[OSM-dev] osm2pgsql tile expiry freaks me out

[Matt Amos]

On Thu, Nov 5, 2009 at 4:22 PM, Steve Hill <steve at nexusuk.org> wrote:
> On Thu, 5 Nov 2009, Jon Burgess wrote:
>> He found that even though the osm2pgsql code should in theory produce
>> more accurate results, the ruby scripts tended to work better overall.
>
> If the Ruby scripts produce more accurate results, we should certainly
> look at why and try to improve the osm2pgsql code.

they don't. i think when jon said they work better overall, he meant
that they're usually "good enough" and are quicker to run, not that
they're more accurate.

> The ruby scripts
> probably don't expire as many tiles - I think they only expire metatiles
> that have a node associated with a changed object on them, which means
> that it won't catch cases where a way crosses the corner of a metatile
> (with no nodes on the metatile) and where a filled polygon changes
> (resulting in a need to re-render all the tiles within it).

indeed. they're quick and dirty. the algorithm is just: expire all
metatiles which contain a node which changed, or a node which is part
of a way which changed. where "changed" means "was in the osc file".

cheers,

matt