I’ve 3D mapped a building at my university and am generally happy with the results. However, that building has a part which has a distinct name (more specifically, it’s named as an extension to the main building). This part is composed of two of the 3D building_parts.
It also has a different address and ref to the main building (my university assigns each building a code, the main building is C and the extension is CA). It is also somewhat poorly connected to the main building internally, so I’m honestly not sure if I should just map it as a separate building entirely. I should note that this building isn’t the only one on campus to have a similar extension with a different building code, although the other building is better connected internally and the extension doesn’t have its own address.
There’s no solution for this except repeating name= + ref= , if it’s not (modeled as) a different building= , or not addr:unit=
One possibility is at least to allow type=building to relate only building:part= (no building=outline ), as in a reverse of a public_transport=stop_area_group
Unfortunately, type=building relations are also used for 3D mapping and using them for the purpose of grouping building parts sharing a name would therefore break 3D rendering, too.
I think this is another situation where mapping physical reality and mapping the things people give names to cannot always be achieved with the same OSM element. Similar to how, say, a plaza isn’t always represented by exactly one highway=pedestrian area (so we invented place=square).
The problem is that a building:part can only belong to exactly 1 building. If it’s already part of a building relation, it won’t also be part of the building represented by the building outline.
(Specific implementations will be renderer-dependent, but that’s the overall logic I’d expect from a renderer following Simple 3D Buildings.)
type=site is for a single feature/object with different parts. You have to define what parent feature it is first. This is multiple features / same category of objects.
The most generic is type=multiobject , further than =multilinestring mostly used in OHM only. Neither has any possible support now.
Map the combined buildings as one building=university with name=C (or ref=C, whichever is most appropriate), covering the entire area. Add whatever other tags are appropriate.
Map the extension (only) as one building:part=university with name=CA (or ref=CA, whichever is most appropriate), covering the extension area. Add whatever other tags are appropriate.
Add note=* to each of the above to explain the complexity.
Then use building:part=yes and whatever other tags are appropriate to map the 3D details.
building:part=university should show exactly like building:part=yes, but the different tag will cue people to this situation being different. As I understand things, the vast majority of building:part=* act like building:part=yes, with building:part=roof being the most obvious exception.
Alternatively, just map them as two attached buildings.
as you speak about “not well connected to main building”, I wonder if it is not just a proper building that is connected to other buildings. Would you mind posting a link to the object in question? IMHO there is a fair chance that mapping it as its own building is the best alternative for the current state of affairs, as long as it seems justifyable.
Relation: Henry Grattan Building (13005754) | OpenStreetMap This is the building in question (the link leads to the entire building). The part I’m referring to (which has a separate address and a well known name) is the Henry Grattan Extension, which consists of 3D building parts Way: 1321391384 | OpenStreetMap and Way: 1321391385 | OpenStreetMap . It’s only connected to the main Henry Grattan building through a bridge on the first floor (and there are several clearly distinct buildings on the uni that are also connected by similar bridges), and perhaps a staff only entrance or two.
To complicate things further there are also two additional ways that may or may not be considered part of the extension, which are Way: 1321391383 | OpenStreetMap (which could be called the School of Applied Language & Intercultural Studies since offices of professors from that school are pretty much the only thing there) and Way: The Street (1321391386) | OpenStreetMap. I’m pretty sure they were built at around the same time but the room numbers in these parts begin with C instead of CA (although the SALIS wing’s room numbers are in the format Cf1xx instead of Cfxx as with the rest of the Henry Grattan, where f means floor). The Street is connected to the main Henry Grattan building (and in fact serves as the main entrance to that building) and also contains a side entrance to the SALIS wing, but has no public entrance to the parts which are definitely the Henry Grattan extension and contain rooms numbered CA.