Highway polygon drawing order does not follow z_order

[Robot8A]

Expected behavior

An underground railway platform either gets rendered over pedestrian areas or doesn't, but is consistent.

Actual behavior

It is inconsistent. The platform gets rendered over highway=pedestrian type=multipolygon, but not over highway=pedestrian area=yes.

Screenshots with links illustrating the problem

https://www.openstreetmap.org/way/1077719862 (https://www.openstreetmap.org/relation/10860682, https://www.openstreetmap.org/way/295343685)

[imagico]

The sample case you showed no more works because someone could not resist the urge to change tagging so the problem is not visible any more. The really advisable tagging here (as with any case of overlapping road polygons which are vertically separate and non-connected) is to apply appropriate layer=* and this way resolve the topological ambiguity for all data users. level=* and location=underground and not a substitute for that.

The problem you observe is not related to closed way vs. multipolygon mapping, the discerning criterion is way_area. In contrast to road lines, which we sort by layer, z_order and osm_id, for road polygons we sort by layer and way_area. This means that when two road polygons (like highway=pedestrian and railway=platform) with the same layer overlap they get sorted by way_area, just like landcover polygons.

Background is the unresolved question how we want to proceed with the polygon and line features in the road layers. The principal possibilities are outlined in #2046 and #3281.

The shown scenario is rare - most cases of overlapping platforms and pedestrian areas are with different layer values. Still - it can be confusing, because while we render the road polygons within the load layers we sort them like landcover and not according to the sorting rules that otherwise apply in the road layers.

We could change that by sorting according to z_order. But that could be considered to prejudice the above mentioned strategic decision towards #2046. I personally would be fine with that (i chose that direction in the AC-Style - see https://imagico.de/blog/en/navigating-the-maze-part-2/) - but that is just my personal view.

Since introducing ordering according to z_order is the only action i can think of that would actually resolve this issue i have changed the title accordingly. But it is not clear to me if it would really make sense to implement that change without making further progress towards #2046. Because the other consequences of not having the road polygons fully integrated into the road layers - like the lack of bridge/tunnel handing - are IMO much more substantial than this.