update multipolygon incorrectly removes tags from inner ways

[Luzandro]

You can get "fake" holes by multiple touching inner, see for example the illustration here, where the green area in the center inherits the attributes of the outer area as it is part of no inner: ​https://wiki.openstreetmap.org/w/images/e/e7/Multipolygon_illustration_-_touching_inner_rings_with_hole.png

This is in compliance with the definition of multipolygons in OSM, but OSMI raises an issue for this case (see ​https://github.com/geofabrik/osmi_simple_views/issues/49 ), so people try to make that center part explicit and add an additional inner object with the same tags as the outer. Using the update multipolygon tool will remove those tags, resulting in an inner hole without any tags.

[stoecker]

Why report it here? As far as I can see JOSM's behaviour is correct based on the data. An inner with identical tags makes no sense so it's assumed to be old style and gets removed.

[Luzandro]

I don't think it's always the case that it doesn't make sense, it depends if it's seen as part of the outer object or as a different object (probably with additional tags).

In any way there have been a few problems where this tool was given as reason for the incorrect result. Even though I think it is primarily a user error, I still think the behaviour of the multipolygon tool isn't correct in this case.

[Luzandro]

Replying to stoecker:

An inner with identical tags makes no sense so it's assumed to be old style and gets removed.

This would be the case if it was an OUTER, where it is correct to remove the tags but leave it as an outer member of the MP relation. On the other hand, if you argue that an INNER with the same tags doesn't make sense, the way should also be removed from the MP relation and deleted entirely, instead of just removing the tags.

[stoecker]

Replying to Luzandro:

Replying to stoecker:

An inner with identical tags makes no sense so it's assumed to be old style and gets removed.

This would be the case if it was an OUTER, where it is correct to remove the tags but leave it as an outer member of the MP relation. On the other hand, if you argue that an INNER with the same tags doesn't make sense, the way should also be removed from the MP relation and deleted entirely, instead of just removing the tags.

As I said "An inner with identical tags makes no sense so it's assumed to be old style and gets removed". If you don't understand this sentence go back in the history of OSM areas. There were a lot of iterations until the current state was reached.

[Map47]

It transpires from reading the area discussions that this type of mapping is to be avoided and that this case should be documented on the Wiki. Too bad that after 5 years this has not happened.

In any case, it is clear that between adjacent areas there should be no gaps as a software cannot determine whether this is an error or instead is an intended situation.

It is my opinion that all three areas must be mapped, because they are real, even in the case that the innermost area matches the MP area thus avoiding an arbitrary interpretation, of area type, by the software.

The behavior of the MP update tool, in this case, by deleting the area tags, creates a hole, not a virtual one, with the loss of the area information, specially inserted.

I also highlight that, in each case, the tool signals that the inner rings are adjacent even in the simplest document case in the Wiki.

[Luzandro]

I've updated the wiki pages for ​old style multipolygon and JOSM update multipolygon tool to better reflect what's happening

[Map47]

Perfect, but it should also be pointed out in the Wiki that for these reasons there cannot be holes between adjacent inner areas of the multipolygon as in the case under consideration.