[OSM-talk] Mapnik maxing CPU with Corine data set

[Dermot McNally]

On 17 January 2011 19:19, Lennard <ldp at xs4all.nl> wrote:

> If you have some polygons that span half the country, that could be slowing
> down mapnik. If that's the case, and you import this as-is, you'd likely
> affect everyone else rendering Ireland in their mapnik renderer stacks.

Assuming it's our data to blame (which seems most likely) then yes, it
would be bad to have this stuff in the main OSM database. So far this
is just test data on a test PostGIS. No polygons of the scale you
suggest ought to be _getting rendered_. I emphasise this because there
are such polygons in the data set and they even have OSM tagging
(invented). Corine has an ocean layer, which we kept in the conversion
in case it might be useful for coastlines or suchlike.

And here I begin to have a doubt - the sea polygons are tagged
natural=water_sea. My assumption was that these would be ignored in
rendering because no such tag is understood. But as it's a "natural"
tag, won't it actually be considered a polygon and stored in the DB
anyway? If so, it must be the culprit. What do you think?

> Also, I presume your indexes are good?

The DB is freshly imported with osm2pgsql and vacuum analysed. This is
the same approach as on the other Database used on this Mapnik
instance and it works fine, so I think that's unlikely to be the
problem.

But I'm about to create a version of the data file without the
natural=water_sea polygons. They seem unlikely to be useful anyway,
and they are of the kind of size you mention...

Thanks for the pointers, I'll report back.
Dermot

-- 
--------------------------------------
Igaühel on siin oma laul
ja ma oma ei leiagi üles