OpenStreetMap
shops as closed-Way building outlines, but also as Nodes in the center?
Posted by Peter Dobratz on 7 June 2015 in English.I seem to be going around in circles in discussion of changeset 31269952 with a professional GIS user. He’s added on purpose, a node for a hardware store for which there is already a closed-Way polygon for the same hardware store.
Maybe someone can weigh in on the changeset discussion and add clarity to the discussion.
Discussion
Comment from Stalfur on 7 June 2015 at 16:15
I’m in the POI camp now myself. It makes it easier to fix if the shop moves or if a second one opens up in the same building. Address on building, node inside building with shopname - inherits the address of building.
If you have 4 shops in the same building then they don’t need address on the POI.
Comment from SomeoneElse on 7 June 2015 at 18:15
As you’ve said, except in very special cases, there shouldn’t be two things in OSM for one thing in the real world (you’ve already linked to http://wiki.openstreetmap.org/wiki/One_feature,_one_OSM_element on the changeset discussion).
It sounds like they’re trying to use some database representation of OSM data (“ the problem with that approach is if you need to use a derivative of the polygon dataset for other purposes”) but it’s not clear which one, or for what. If they’re assuming that all objects in OSM have a duplicate node representing them then they are (thankfully) sadly mistaken.
That said, there’s nothing wrong with having “shop” node POIs within a non-shop “building” way. As Stalfur said, it does have some advantages - although a downside can be when you want to show the relationship of different shops to each other.
Comment from MarkusHD on 7 June 2015 at 20:52
The main problem does not seem to be the POI tagging style (node vs. way), but the ignorance of existing POI data and strange explanations (at least I did not understand it at all) to justify this duplication.
I don’t want to misdirect the discussion, but want to add that if tagging the POI as separate node I put the address on both the POI node and the building. Yes, duplication, but not comparable to POI duplication. Nominatim didn’t inherit the building address, but guessed from the nearest street, which failed in some cases - this may have changed meanwhile, I don’t know. Practically it’s the same reason as for adding more address data than just housenumber and street: reliability without hoping the data consumer does it right.
Comment from SK53 on 8 June 2015 at 09:21
As someone who has been working to create a global POI dataset of retail outlets I can confidently state that this is fairly straightforward if one uses osmconvert (or for more limited data overpass with the centroid feature).
Processing the data to remove duplicate node/polygon elements turns the simple code into something far more complex. Interestingly following MarkusHD’s approach means that things like address data can be incorporated trivially in the data set. So I’m happy with duplication of addresses (straightforward as we, like many others, treat addresses as attributes of elements, rather than elements themselves), but vehemently object to duplicating elements themselves.