The answer is that until recently the Bosnia and Herzegovina polygon (http://www.openstreetmap.org/browse/relation/214908) was broken. Without a valid polygon all Nominatim can do is to select the nearest country node which is mostly wrong, but better than nothing. Unfortunately, because of resource limits, Nominatim can not currently update an entire countries worth of data when the polygon was fixed and so country level polygons are deliberately excluded from triggering re-indexing. When you recently edited the Lukavac node you caused it to re-index this point and the data now works. It would have started working without the is_in tag - it just needed anything that triggered re-indexing. This is a problem and it is being worked on. answered 20 Jan '11, 13:32 twain |
Perhaps adding is_in:country = "Bosnia and Herzegovina" to the node for Lukavac, the search is only as good as the data. Compare the tags:
The Tuzla node (which shown as Bosnia & Herzegovina
The Stockholm node has a lot of tags for these kind of searches. e.g.:
answered 19 Jan '11, 01:17 emj 1
Well, loading the node with even more of the controversial is_in tags might solve the problem, but I think it would be interesting to understand the root cause why Nominatim thinks Lukavac is in Serbia before hiding the symptoms.
(19 Jan '11, 01:53)
gnurk
Nominatim isn't very consistent one thing works one week is changed the other, so change the data and hope that some other city is put in Serbia and then you have more data.
(19 Jan '11, 07:40)
emj
The is_in haters have no other solution, so practically it is atm better with is_in than any other non existing solution.
(19 Jan '11, 08:07)
emj
Sure there is another solution to is_in. Closed (and hopefully correct) administrative boundaries as either ways or relations.
(19 Jan '11, 08:54)
petschge
I have now added is_in:country = Bosnia and Herzegovina to the Lukavac node. I guess we have to wait some days to see the result. Nominatim currently says the data is from 2011/01/16.
(19 Jan '11, 10:56)
gnurk
emj, please don't be so aggressive ("and you know it", "it's not like I'm telling you"). This is a help site. If you want to start a flame-war please use the mailing lists.
(20 Jan '11, 10:47)
Richard ♦
1
Richard: Not intended to sound aggressive at all, but I know I was annoyed to have stuff recommended that just doesn't work with no rationale about the question that is supposed to be solved. Deleting the comment seems to work ok, and doesn't make the discussion seem strange.
(20 Jan '11, 12:06)
emj
Thanks! :)
(20 Jan '11, 13:34)
Richard ♦
showing 5 of 8
show 3 more comments
|
Someone will probably soon tell us that this is not the right place for issues like this, and that you instead should use http://openstreetbugs.schokokeks.org/ for bugs in the map.
But I think the problem is still interesting: Why does Lukavac show up in Serbia in Nominatim, when for example Tuzla does not?
http://open.mapquestapi.com/nominatim/v1/details.php?place_id=7618777 shows a polygon which I can't find when I open the area in JOSM. Is that (deleted?) polygon the cause - and how can I find it? Or maybe the polygon is virtual and just built up of the surburbs where Lukavac is parent?
Yes it's virtual, if they don't use any other source of data which would be very bad of them imo.