Map Buildings next to each other with a space between them?

I have heard in mapathons (And I think it is something many people do) that adjacent buildings should be mapped with separated nodes and a small space between them (even if there is no space between them). Like this:

 x-------x
 |       |
 |       | x----x
 |       | |    |
 |       | |    |
 x-------x x----x

However, I have always mapped adjacent buildings like that.

 x-------x
 |       |
 |       x----x
 |       |    |
 |       |    |
 x-------x----x

For example, this kind of house, even when they have different architecture (not terraces - very similar architecture for all adjacent houses).
https://www.mapillary.com/app/?pKey=1476776676184810

I prefer the second way (that I have been using), but I want to confirm because proposing to map with different nodes to newcomers makes the mapping more complex.

5 Likes

Agree, houses that share a wall physically should also share nodes in OSM. Town houses and row houses should be one building=* for each unit, while appartments should be one building=*.

17 Likes

I try to map what is there: If two buildings physically touch then I will use your second example. If there is a space between the buildings then I use your first example.

In my town there are some areas where it is not obvious from the available aerial imagery which the case is, you really need to perform an in-person survey. I assume that is true for other places so I defer to local mappers if they correct what I have entered while “arm chair mapping”.

9 Likes

I do map them separately, I even do fix the case of sharing nodes between buildings made by other mappers because buildings have their own walls, at least in Mexico that is the norm. Even if the walls are adjacent, they are not one and the same. The way you do it is probably adequate for indoor mapping shared walls.

EDIT: I guess a lot of people end up snapping nodes because it is the default behavior in all the editors I know, but most of them have one key that, while pressed, will deactivate snapping.

EDIT 2: Also, editing an area that was mapped sharing nodes is hell and very, very hard to fix. You can’t just move one building without messing with the snapped ways.

sharing nodes or connecting ways is not making a statement whether there is one wall for each building (typical situation) or two buildings share the same wall (would be a rare exception, probably in most of the world, have seen it with historic buildings sometimes).
We don’t map walls when we map a building=*

Even our barrier=wall is somehow abstract and there is not really a satisfying way for example to map explicitly 2 walls, one touching the other. We are not prepared to map at this scale.

Still, we do try to describe topology, so when 2 buildings are attached, ideally it should be possible to see it from the data.

3 Likes

Sharing nodes does not mean that the buildings have a shared wall, just that their walls are directly adjacent without a gap between them. Mapping adjacent buildings with shared nodes is what I would consider best practice – it’s certainly the standard in those parts of the world I’ve been mapping.

I’ve even taken this assumption into account when writing software for 3D rendering OSM data to some extent. Specifically, I assume building polygons represent the outer edge of the building’s walls, not the center of those walls.

10 Likes

no, most people map joined buildings as joined because these buildings are joined

if people currently map in Mexico in other way I would strongly encourage to drop this divergent local rule

to avoid potential confusion, I am thinking about cases like this:


(image from File:Sławkowska Street in Kraków, Poland, September 2019.jpg - Wikimedia Commons )

relevant buildings in OSM are Way: ‪Dom pod Globusem‬ (‪152797352‬) | OpenStreetMap (building with the globe thingy on top) and Way: 152797448 | OpenStreetMap (light building on the right) and they correctly share nodes where walls are joined

building=* outline does not map wall centerline (imagine fortress with 5m wide walls - would you really put building= outline in the middle of the wall?

I understand that changing local diverging practice is irritating, especially when it turns out that some improvements were actually making map data worse - but I would strongly encourage to map building geometries like elsewhere, this would remove one of traps/blockers for using Mexico data or mapping there.

5 Likes

Yes, we don’t map walls, I’m well aware, thank you guys. We map entities and, in the case of buildings, they are generally limited by walls. I am not saying you can’t have a case where you can share nodes, but it is not the general case in buildings. They are usually separate entities, physically, that is probably why this is the consensus.

@Mateusz_Konieczny I’m sorry, but I see two different buildings in your example; your opinion is that sharing nodes between these buildings is correct, but I don’t share that opinion, nor is that the recommended practice in OSM. The separation may be minimal, but it exists, I even think it is visible in the photograph.

EDIT: Please, before replying take a look at this area. I am talking about this practice, not a fringe case where maybe sharing nodes could be okay.

1 Like

In my opinion, as long as their walls touch each other all the way from ground, they can share nodes. For sure though such thing can make future edits on either building’s outline more difficult.
What happens in the case where the walls have few centimeters distance from each other, but at some point upwards they touch each other (one building falling towards the other)? It sounds rare case, but it does occur in cities. Personally, I would make them separate.

As it was said before, shared nodes would be for adjacent walls with no gap between them, which means that from imagery it will be difficult to determine, especially when it’s sideways, so either ground-level imagery or on-site search (go with a magnifying glass at each building like an engineer :P).

4 Likes

The thing is, even if they are adjacent, they are separate entities. I know close to nothing about 3D rendering software, but this is not an issue on that. This is fundamentally a geo-ontological question, but for us, it is about standard practices. What you think it is a best practice is not necessarily standard practice, and we, as a community, shouldn’t map for any renderer in particular. I don’t know where you contribute, but where I and --apparently-- the OP contribute, we map buildings trying not to share nodes.

I believe your assumption is correct, though. The issue is where a building ends, and we agree this is the outer edge of its own exterior walls, in general. Even if two buildings are so close to each other, they generally all have their own walls that are a part of each building, not shared. If these walls are next to each other, there is still a minimal gap between them, it could be 1 mm width, variable, 1 cm, 1 dm, 1 m, several meters… Strictly speaking, these are all separate entities in space, but I still can understand sharing nodes in most of these cases.

At least in Mexico, the buildings that are in different property can’t share walls, by construction norms they should have their own, inside the property of the land. Frequently, two buildings will be next to each other, but in general, they are still physically and legally separated by the outer edge of its own walls.

  1. yes, these are two different buildings

  2. no, there is no separation whatsoever between them, see Node: 1655757628 | OpenStreetMap that is on edge between them on Google Maps StreetView

On assumption that these buildings have walls with no separation between them (looks likely from aerial imagery) then these look like correctly mapped buildings, sharing nodes as expected for joint buildings? See say Node: 5119559809 | OpenStreetMap Though I am unable to verify exact splitting based on aerial imagery.

And if in reality there is gap between buildings, not visible on aerial imagery, then changes would be needed.

Do you have photo of on the ground situation?

It can be also 0. In such cases nodes should be shared. Maybe in Mexico this type of construction is rare.

In “1 cm, 1 dm, 1 m, several meters…” cases there definitely should be separation.

1mm seems unstable and problematic for multiple reasons, but until there is such gap between buildings I guess that there should be separation (but building in this way is likely to go wrong - such gap will quickly fill with debris and is likely going to end with water infiltration issues, so it seems to be avoided)

3 Likes

I agree that if it is 0, they could share nodes, though, it is advantageous from my POV to not do it this way for future edits and circumstances (higher spatial resolution imagery). I will insist in Mexico this is generally not the case for normative reasons. Being a country with intense seismic activity, we are just maybe stricter with construction norms than in other places.

In your Google View image I don’t see evidence there is no gap between the walls, since a pipe is preventing us (me?) to see what’s really happening behind it. I will admit I thought the pipe was the gap between buildings in your previous image. I think I’ve given my 2 cents here, and I hope it helps others with the problem.

that sounds like something likely resulting in drastically different norms, Poland has basically no seismic activity (except cases of ground moving above active and former mines).

I’m not sure I follow this point. In California, United States, another region with strict building codes due to seismic activity, buildings in urban areas frequently touch each other, with truly 0 gap between adjacent buildings. These buildings typically share nodes along their touching walls when mapped in OSM, as is standard practice throughout most of the world as far as I know.

4 Likes

Yes, it is a case of no separation. This is typical in many cities, in such case nodes should be shared by buildings. See closeup at File:Border of buildings.jpg - Wikimedia Commons

See File:Sławkowska Street in Kraków, Poland, September 2019.jpg - Wikimedia Commons for view of entire building.

2 Likes

I can’t say about California. I’m just trying to convey the idea that your local cases do not necessarily make sense for other places. For example, about the area I linked before, @Mateusz_Konieczny said:

On assumption that these buildings have walls with no separation between them (looks likely from aerial imagery) then these look like correctly mapped buildings

But the problem I’m seeing is precisely that assumption. I know that area, I know for a fact that several of those buildings have a variable gap between their walls (some of them you can watch in Street View). It is not a safe assumption to snap all the nodes you can in blocks like these, at least not here.

I may add that, usually, I only map in Mexico. We don’t have a lot of contributors and these edits proliferated during the 2017 HOT Project following the two earthquakes that devastated various cities in the country. Obviously, a lot of foreigner contributors helped to map these areas. I can see now where this issue comes from. It’s been interesting to discuss it.

1 Like

I know that area, I know for a fact that several of those buildings have a variable gap between their walls (some of them you can watch in Street View)

then this is a completely different situation and the nodes should not be connected and a gap should be kept.

5 Likes

This might be the main-reason for the discrepancy. Because the example you brought up

looks like a regular city to me. For example where i map a lot it looks like OpenStreetMap that. But i can assure you, everything is correct mapped here, buildingwise and there are not gaps.

So to answer the question from @AngocA :slight_smile: Thats advice you got on the mapathon is false, if the two buildings are direct adjacent without a gap then your second example is correct. Just separate the building in OSM if there is also a gap in the real world, no matter if its just 5cm or a meter - its not adjacent then.

1 Like

I still see many people do this, and that’s why I created these posts to try to correct this mispractice:

1 Like

I’m not sure if I understood your comments correctly, but… (I’m a non-English speaking editor).
OSM accepts the big principles of ‘drawing as you see it’ and ‘drawing the building along the outermost area of the final building’.
Regardless of whether the walls are shared or not, I think it is correct to draw them together if they are attached.
Also in East Asia (and in some other places too) there are even buildings that look overlapping because of their roofs.

1 Like