Should landcover be ordered by osm_id in addition to way_area?

[imagico]

Recently i saw a landcover area where the color varied arbitrarily depending on zoom level because there are two different areas with different tags but identical geometry and therefore identical way_area:

https://www.openstreetmap.org/way/346109314
https://www.openstreetmap.org/way/49188959

I think cases like this should get a consistent drawing order, probably best by ordering by osm_id in addition.

[matthijsmelissen]

Funny bug. Your proposal makes sense.

[pnorman]

👍

I personally think we should be making decisions ourself on ordering order more often, but we should always have an osm_id in the sorts where rendering order makes a visual difference

[andreynovikov]

Should not it be one way with additional landuse tag?

[polarbearing]

While I understand the issue in general, in the specific example the landuse way makes no sense to me. The use of the land on which the airport sits is not 'grass'. Some of its part may be covered with grass, but probably not the runway.

[dieterdreist]

sent from a phone

On 9. May 2018, at 00:24, polarbearing ***@***.***> wrote: While I understand the issue in general, in the specific example the landuse way makes no sense to me. The use of the land on which the airport sits is not 'grass'. Some of its part may be covered with grass, but probably not the runway.

completely agree, in this example the grass area should be reduced to the actual grass area. As a sidenote I would question in general the semantic sense of „grass“ as a landuse value, and would propose to drop rendering it.

[kocio-pl]

This would be quite a big move to drop landuse=grass. I don't exclude possibility of such action, however we should implement landcover=grass from #2548 then to help make transition, because I can't imagine generic style not showing grass areas.

[pnorman]

Let's keep this confined to the issue at hand, which is fully defined ordering of the landuse layer. Random rendering results based on a lack of ordering is a bug leading to errors at metatile edges, regardless of if the tagging in this case makes sense or not.