Helium314

Thanks for the hint with the timestamp! Looks like after uploading the first change (and persisting the changed element), the local timestamp differs from timestamp of the element (re-)downloaded after...

It should be. But I'm not sure whether this is necessarily true, as I don't know enough about how OSM elements may be changed. So if you think it's fine,...

Yes it is, see https://github.com/streetcomplete/StreetComplete/blob/a8d915e022ddf1146b54a3a47ef262eec7baa586/app/src/main/java/de/westnordost/streetcomplete/data/osm/edits/upload/ElementEditsUploader.kt#L59 This is where I found that after uploading the edit, the only difference between downloaded element and local element is the timestamp. Anyway, I think...

> Coin or note only machines are too rare to ask specifically about but might be concentrated in specific regions/systems I don't know any note-only machines, but there are a...

No, this is because of my changes

Anyway I got it now. Needed to learn a bit more about how to use Mockito, and ran into trouble because `any()` apparently doesn't actually mean **any**... that's why there...

Could we just always use `MapDataApi` if `edit.action is SplitWayAction`? > Right now I am asking myself why even the `element: Element?` is passed, after all the edit action could...

> What I am thinking is whether that workaround to throw away all route relations on download (see `IGNORED_RELATION_TYPES`) could be revisited, especially when we can get the persisting of...

See https://github.com/streetcomplete/StreetComplete/issues/2803#issuecomment-827639718 and below. I had identified the the db query for getting members of (huge) relations as the slow part.

I just tested how fast things are when all relations are in the database. As expected `getMapDataWithGeometry` is still very fast when everything is cached, but loading an area with...