[OSM-dev] tiles tiles tiles

Jon Burgess jburgess at uklinux.net
Wed Nov 29 23:05:54 GMT 2006


On Wed, 2006-11-29 at 23:55 +0100, Raphaël Jacquot wrote:
> Jon Burgess wrote:
> > On Wed, 2006-11-29 at 21:23 +0000, SteveC wrote:
> >> Using a bugfixed version of almiens tilecount.pl (which figures out what
> >> tiles actually have map data on them), I've written a script to render
> >> them (from zoom 0-18).
> > 
> > I believe there are a couple of edge cases which we may need to
> > consider. 
> > 
> > 1) Way segements which cross a tile even though the nodes are no nodes
> > inside, e.g.
> > 
> >    ------
> >   |      |
> > +-----------+
> >   |      |
> >    ------
> 
> this is handled by the postgis operators
> 
> > 2) Text labels and icons which extend beyond the edge of the tile containing the node 
> > 
> >  ------- -------
> > |       |       |
> > |   *City-Name  |
> > |       |       |
> >  ------- -------
> > 
> > Perhaps we need to ensure the 8 surrounding tiles of and on the list is also rendered.
> > In many cases these tiles will already have nodes so the overall increase in the number
> > of tiles to be rendered may not be as big as first appears.
> 
> if you look at the python program, it renders on a 512x512 tile and cuts 
> out the sides to keep a 256x256 center
> 

I completely agree with the above 2 statements, however neither will
address this issue. The statements below are based on reading the
scripts in utils/mapnik/all_tiles/...

In (1), the scripts will decide that the tile does not contain any nodes
and therefore not even add it into the lists which mapnik will try
rendering, hence PostGIS will not even get asked if the tile contains
any relevant data.

Similarly in (2), we will use the data +- 128 when rendering the tile
containing the node. However, we will not select the tile on the side
for rendering since it does not contain the node.



	Jon







More information about the dev mailing list