[OSM-talk] Wood & Park mapnik carto anomaly?

Paul Norman penorman at mac.com
Tue Sep 16 18:55:48 UTC 2014


On Sep 16, 2014, at 06:33 AM, "Dave F." <davefox at madasafish.com> wrote:

On 16/09/2014 13:41, Matthijs Melissen wrote:
        > In general, we render smaller landuse on top of larger landuse.

I find it surprising something as arbitrary as size is used as the 
defining factor. Comparing actual tags would surely make more sense.

As a recent bug (https://github.com/gravitystorm/openstreetmap-carto/issues/950) has shown, it's important to have *some* well-defined ordering in cases where the ordering could make a visual distinction, or the rendered result is undefined and potentially not deterministic. This can lead to subtle bugs with clipped labels.

The two criteria are OSM ID and area. The first is truly arbitrary being a computer-assigned number, while the second is well-founded and is the standard way to order within a layer. 

What you're more interested in is why are parks and trees both in the same landuse layer. It would certainly simplify the SQL (https://github.com/gravitystorm/openstreetmap-carto/blob/master/project.yaml#L102) to split it up into different tags, but the problem is there is no universally acceptable ordering of tags. You've pointed at a case where it'd be good to have trees on top of parks, but I can point to cases where parks should be on top of trees.

There's another layer for overlays like military or nature reserves but without trying it out, I'm not sure if that'd add clarity. Someone is welcome to try it out (https://github.com/gravitystorm/openstreetmap-carto/blob/master/CONTRIBUTING.md)

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.openstreetmap.org/pipermail/talk/attachments/20140916/9e859d95/attachment.html>


More information about the talk mailing list