|
Hello, When I search "Taiwan Strait" ( https://www.openstreetmap.org/search?query=Taiwan%20Strait ), the correct multipolyon result ( https://www.openstreetmap.org/relation/9348353, it's a big one so might take a while to load ) does not shown in the result. I checked the multipolyon in JOSM and it's quite complex one. Is this a bug in Nominatim or issue in the strait multipolygon? Thanks in advance! |
|
Mapping these bays and straits as multipolygons is error prone (they are easy to break), pollutes the database with countless versions (every time someone splits a pice of coastline, a new version of the relation must be uploaded), creates giant objects that complicate editing for everyone (there is a reason we don't map oceans as polygons...), and serves no purpose that could not equally be served by a node. When I encounter polygons like that, I usually replace them with a node. I have done that here now. Hi @Frederik, I also don't like this kind of super large multipolygon for all sorts of reasons. Thanks for confirming this is bad mapping. However I don't quite agree to replace that with just a node. A strait node is rendered usually insignificantly. How about a simple polygon like English Channel: https://www.openstreetmap.org/way/920269202 It's a rough approximation but it's good enough for different purpose
(12 Apr '21, 09:31)
strongwillow
2
That English Channel polygon is terrible. It doesn't even have holes for the Channel Islands. It will break anything which renders water polygons above land.
(12 Apr '21, 11:03)
Richard ♦
|
|
The main polygon looks broken when I inspect with the JOSM editor. All elements (ways) exist but there's two gaps and some ways are not in the correct order. Sadly http://ra.osmsurround.org/analyzeRelation times out for me for this relation, it's huge. Once fixed Nominatim will import it and link the node to the relation (based on position and wikidata id). 1
for example a gap between https://www.openstreetmap.org/way/207552518 and https://www.openstreetmap.org/way/225761779
(11 Apr '21, 23:09)
mtmail
Thanks for checking. Indeed large multipolygon tends to break
(12 Apr '21, 09:24)
strongwillow
|
