Can a multipolygon outer share part of a way with an inner of the same multipolygon? asked 01 Mar '11, 02:28 upstream |
No. The outer and inner of a multipolygon relation must not touch or cross. No sharing of points either. answered 01 Mar '11, 02:35 Richard Weait 1
The reason being because if the 'inner' touches the 'outer', then it is no longer 'inner' - the 'outer' way can just divert around what was the 'inner' to exclude the inner area.
(01 Mar '11, 09:30)
EdLoach ♦
1
Interestingly in OSM we allow inner rings to touch. This is not conform to standard OGC Simple Features but allows for simpler modeling in OSM when you have multiple different adjacent features inside the outer way which you want to exclude from the polygon.
(01 Mar '11, 11:59)
dieterdreist
|
I'm seeing a lot of people putting "simple" concave multipolygons (multipolygon relations that have no holes and folds into itself without crossing) with the concave parts as "inner" - am I missing something, or despite being concave, should it not still be outer?
Since I'm seeing this a lot, just need to make sure and there's no clear examples anywhere though I can see why someone might call it "inner" ... incorrectly I assume.
Example: think of a "PacMan" figure, and imagine a way that traces the edges. This way that traces the edges no matter how it's split up into a multipolygon relation, all of the ways should be "outer" including the concave "mouth" portion, and the eye, if drawn, would be the only part that could be "inner"?
Example: if tracing a multipolygon that looks like ⛻ there should be zero "inner" ways.
Another example: think of a spiral shaped hedge. All ways should be "outer" including the center of the spiral.
All three of these examples could be drawn without a multipolygon but ignore that for now... (except if PacMan had an eye...)
Tagging those ways as inner would almost certainly be considered an error. Can you give an example of one of these?
I've been correcting them as my gut feel was that it was indeed an error. Here's one I haven't touched yet though it wasn't as egregious as others I've seen:
https://www.openstreetmap.org/relation/11117214
specifically at this member: https://www.openstreetmap.org/way/805895181
This member should be marked as "outer" despite being "inside" as far as I know about how topology works.
Yes, that way should be an outer. The JOSM validator throws an error that the multipolygon isn't closed, which is true because there isn't a continuous ring of outer ways due to that one mis-roled way.