The case of terraced houses being misaligned and offset by an entire building still worries me. Postcodes are less of a worry, but if we are putting `ref:GB:uprn`on each building, those could all be off-by-one.
I don’t know what the answer is, beyond a visual sanity-check in JOSM that the buildings aren’t misaligned.
That’s precisely what I’ll be doing, using the OSMUK LR Polygons overlay as a visual check before uploading. If there’s a terrace which is misaligned by one building, I will either (1) exclude it from the upload, or (2) abort the upload, realign the terrace, re-run my script and upload. (I’ll add this to the wiki page).
For terraces in much of Greater London, I can also use the “NLS - OS 1:1,250/1:2,500 National Grid Maps, 1947-63” layer to confirm the alignment of terraces.
House buildings are mapped as buildings, but it is not the default to add address data to the building.
Address data in the UK is commonly added to either a node within a building, or directly to the building. Adding the address data to the building is more common, but I’d suggest this is due to the way tagging is led through Editors.
We don’t have a correct or default way of adding address data in the UK, but I’m concerned this import will effectively impose address data to buildings
I originally added address data to buildings, but changed to follow local practice and now always add to a node within the building. Primarily because the data we use is from a node, and I came across examples where it was wrong to add the address data to a building.
The data proposed to be added is very useful, and I need to think of a solution, rather than just dump problems into this thread.
Some time ago, I considered writing a straight import proposal that would add UPRNs without the manual checking, by only adding those where everything matched exactly within an area. The presence of UPRNs for various types of things (which can’t be readily identified from the available open data), and the number of objects that they could match to led me to abandon that idea.
I have been adding UPRNs, aided by some scripting in JOSM, and can confirm that the checking is definitely needed. It is also quite straightforward to check visually - I use a combination of a UPRN overlay and a custom style to see where things have been missed, or are misaligned.
The checking that Robert is proposing as part of this would allay any worries I would have about this.
There certainly are buildings which shouldn’t have addresses, blocks of flats with separate entrances being the example which springs most readily to mind. In the parts of London with which I’m most familiar, addresses for houses may start off as interpolation nodes, but usually end up as tags on the buildings. There are also areas where every address is a node within the building polygon, which this import won’t touch.
At present, at least 30% of the 15 million+ buildings in Great Britain have addr:* tags (taginfo).
I’ve always added address to buildings where there is only one address associated with the building. I suppose it would be possible to check whether this is the prevalent way of doing things in the UK, but my impression is that it is. It’s certainly prevalent round my manor.
Same unless it’s something like a semi-detached house or apartment complex then I will try to add entrances and add addr: to those. I will only use nodes for residential if in a rush or what entrance is what is not clear. That’s my way of working at least.
Looking at Geofabrik’s taginfo data for Great Britain from 2025-10-27, there are currently:
building, of which 27 493 are nodes (probably placeholders)building, 4 854 014 are also tagged with addr:street and 3 764 034 tagged with addr:housenumber (these will obviously overlap)addr:housenumber, of which 3 772 791 are on ways and 665 332 on nodesAs addresses are already over 5 times more common on ways than nodes, I don’t think there is any reason to worry that this import will impose addresses on buildings as a mapping style, as it is already the most common way of mapping addresses. This import will not add tags to a building polygon which contains nodes with address tags, so where local mappers prefer adding address nodes that preference will be respected.
What proportion of nodes with addresses are tagged as entrances?
I’ve run a couple of Overpass Turbo queries for nodes with addr:housenumber or addr:street in Great Britain:
entrance: 59 030 (9%), including entrance=staircase 3356 (0.5%) and entrance=home 14 575 (2%)name: 210 771 (32%) (probably POIs)addr:interpolation way: 50 746 (8%)I feel it’s reasonable to describe the style as imposed. There is more than one way to tag the data in question, but this import will only use one style. A very large and very useful addition, but it will lead to one tagging style completely dominating. Saying that, I can’t think of another way to do it 
.
If local mappers are adding address to nodes, but there are properties without an address on a node or building, I assume this will add postcode to that building?
Although I have issues with the proposal I’ll accept it because overall outcome is positive.
It will do, but if a local mapper subsequently moves the address tags to a node, I’m absolutely fine with that. I could add something to the wiki page (to which a changeset tag will link) making it clear that several ways of mapping addresses are in use and people should feel free to use the tags on whichever object they feel appropriate. If you have a better way of phrasing that, I’m happy to use it.
Also, if I notice when doing manual post-processing in JOSM to add things like addr:street that everything nearby uses addresses on nodes, I’ll move the tags to nodes myself.
There is a patch of addresses on nodes in west London that were created by a mapper who was banned for using questionable sources. Maybe we shouldnʼt follow those ones.
Unless anyone objects, I would like to make a start with the first small imports this Wednesday, 2025-11-05.
No objection here - I look forward to seeing it pop up on my OSMcha filter!
Yeh, weirdly looking forward to it myself. I’ll have to get more houses mapped in my local area so they can catch tags 