Can building:part features have addresses and other functional tags?

Terraced houses are very common in Finland. They’re usually recognizable as one obviously continuous building, sharing the same superstructure, roof, and top-level housenumber. When the geometries of individual homes are known, it is attractive to map the roof overhang and home footprints separately. Normally this would involve tagging the entire building outline with building=*, with the roof being building:part=roof and homes being building:part=house.

However, the individual homes are as important as the building structure itself, and necessitate individual address tagging for accurate mapping if these addresses are known. Hence it would be desirable to tag each building:part=house with the necessary addr:* details, such as addr:housenumber or addr:unit. Tagging only the entrance with these address details seems like a weak solution to me, if more information than just the entrance location is known.

The wiki (Key:building:part and Simple 3D Buildings) does not make clear if building:part=* is intended purely for renderers and should not carry any functional information, or if it is accepted that building:part=* carries functional information about address, use, etc. Taginfo reveals that a bit less than 1% of building:part=* features carry addr:*=* or name=*. Does the community accept an amendment or clarification of the wiki to accept adding address or other functional tags to building:part=* features?

Here’s a poll to probe opinions, not to decide anything.

  • building:part=* features could carry other tags with functional meaning
  • building:part=* features should only be used for information for 3D rendering
0 voters
1 Like

every feature can get address tags and you do not need any feature to map an address. The legalities where addresses are assigned to in your country or area may depend on the jurisdiction, but in OpenStreetMap we use addr:* tags not only for assigned addresses but also as a property for a feature at this assigned address, so even if addresses are not assigned to building parts in your area it could still eventually be correct tagging in OpenStreetMap if they had addr:* tags

every feature can get address tags and you do not need any feature to map an address

That is encouraging to hear.

To further illustrate why I map the roof overhang and the individual home footprints separately; people’s backyards, mapped as leisure=garden with garden:type=residential, obviously stop at the edge of the building and not at the outer edge of the roof overhang. It’s logical for the garden polygon to overlap the roof overhang polygon, as the roof is genuinely above the garden in real life. If the whole building outline was mapped as a singular building without building:part=* distinction, the garden would overlap a feature tagged as a house, which is suspicious.

Feel free to drawn a polygon on the area that should have an address. Not our way to map addresses in France, but IMHO it’s way simpler as to have many identical addresses.
For the record, we prefer to do it the other way round: you do not need any feature to map an address, adding a node with the address at the limit between the street and the property (usually on the main entry point).
Side note: the garden is at this address, not elsewhere, isn’t it?

Is it okay if I clarify on the wiki that building parts may be used for other purposes than strictly rendering?

Out of interest, is there a reason they are mapped as building parts rather than as individual adjoining houses (e.g., building=house + house=terraced)? It seems like they are separate houses (they have their own front doors and gardens). Is it just convention in Finland mapping?

1 Like

this does not make them separate houses

They are all part of the same building. I think the tagging should reflect that. If I were to map a pair of semi-detached houses as two adjoining houses, it would not be clear whether they are part of the same structure or if they’re wholly separate aside from one shared wall.

Here’s some random architectural visualization I grabbed online to demonstrate how they often look.

2 Likes

What does then?

I guess that’s the interesting part. What makes them the same building? Is it that there are no load bearing walls between perhaps? Or that the roof could only be replaced across the whole span rather than being replaced for individual properties? Are they more like units, rather than individual properties?

3 Likes

being a separate structure

admittedly, sometimes a lot of effort is to put to fake it in one direction or another - but in vast majority of cases it is clearly visible

being a single unified physical structure (note: legal status does not matter)

in this case they have shared roof, not indication of being structurally split, unified continuous facade

admittedly it is possible that internal structure is separate with full isolation and each part having separate loadbearing walls

There could be plenty of reasons why someone leans more towards seeing them as one building vs distinct buildings. What actually delineates it is vibes. I look at it and ask “does this look and feel like one building” and it’s usually a yes.

2 Likes

The main issue is that building:part is used to model the outer hull of a building and may not have any relation to actual internal subdivisions of the building in question.

So I’m kind of sceptical of using them to carry addresses (it is trivial to come up with examples where this immediately breaks; in particular using building:part for its intended purpose) and would prefer addresses nodes, best on entrances.

8 Likes

Could you give examples? I don’t see any obvious ones.

1 Like

Any thing where you are using building:part to actually model the outer hull and it has any kind of protrusions on different levels, floor plans that are not just rectangular etc.

2 Likes

God damn this mapping! If only outsiders knew how many problems we run into. Maybe the most ideal solution would be a new primary tag for ways that carries the address detail and delineates what space the address applies to, or something like that. It should still mark the boundary of a domicile. For primary tags there’s place=plot, landuse=residential etc. but none of them just fit right.

I have seen a number of cases where it is a good idea to tag the entrance to a building with the address. Typically, the number is attached to the door or right next to it - so this also implements “map what you see on the ground”. It’s also a good idea to knock on the front door if you want to deliver a parcel or a pizza to the house. Maybe this would be a useful solution for your problem, too?

1 Like

Fully agree on this. building:part is defined for somehow getting the shape of a building into a 3D model. They do not carry any additional representation of a real world thing.

In the above case the best place of the address would be at the entrance node.

This is a viable solution, but I find it very unsatisfying in the case where I know the exact layout of a subdivided building. It’s a loss of data and it feels painful!

I suppose that in this situation the main building gets tagged with addr:flats=*, e.g. addr:flats=A-F for a building divided into six lettered homes.

But is there really no present solution to tagging individual addressed homes with polygons if they’re not building=*?

If the individual addresses are mapped, there’s generally no need to tag addresses on the terrace unless the terrace as whole has a name. It’s quite different from an apartment block where the flats don’t have their own street entrance.

E.g for this terrace, the building only carries information about the structure (number of levels). The addresses are mapped on entrance nodes on the polygon, where the property is accessed from the public street.

Of course there are other ways of mapping addresses and conventions vary widely by region, but it sounds like you are suggesting yet another way which may not be a great idea!

1 Like

please note that there are historical examples where different houses lean on the same loadbearing wall, e.g. from medieval times. They are clearly separate houses but are not completely independent.
Also structurally “completely independent” buildings can also rely on each other, as they put weight on the ground, and removing one the other one might, in some circumstances, move and get damaged although their structure was separate, as there was a mechanical coupling through the soil.

5 Likes