Complicated addresses and buildings

I have some complicated addresses and buildings around Cardigan to deal with. I’ll start with one where the complications can be ignored (that’s what I did when I mapped them) and work my way up. I’m giving them all here in the hope that a cohesive strategy emerges that embraces all of them rather than contradictory ad hoc conflicting solutions.

The first problem is five terraced houses (row houses in Merkin). They were built by the same builder at the same time. These days they are known as 1, 3, 5, 7 and 9, North Road (the numbering sequence continues beyond them and the even numbers on the opposite side of the road shows that they are, indeed, on North Road as far as official postal addresses go), Some have house names: 1 is Clarington, 5 is North Gate house, 7 is Kilhue and 9 is Bryngwyn. And that’s how I’ve mapped them. However, they are also collectively known as Priory Terrace. These days not many people know them as Priory Terrace, and the words “Priory Terrace” are not in their official postal address, However, they are all Grade II buildings listed by Cadw (Welsh equivalent of Historic England) under the addresses 1, 3, 5, 7 and 9 Priory Terrace rather than North Road. So wrap another building around the lot of them with just a common name or house name of Priory Terrace? Or use a relation somehoiw? Or just ignore the name Priory Terrace because it’s mainly historical?

The second problem affects official postal addresses. I haven’t even bothered adding them to the map yet because I can’t figure out a non-bad way of handling the addresses (I don’t add buildings without address details because that makes the task too much like painting by numbers). Quay Street (Stryd y Cei) runs east to west. At the east end of the street, on the south side, are numbers 1 and 2, then a church (no number). At the west end, on the north side, are 26, 25, 24, etc. Fairly normal, apart from the boustrophedon numbering.

The problem comes on the south side, after the church (which is a fairly new building compared to the surroundings and might have replaced 3 and 4). After the church comes Rook Terrace. A terrace of 5 houses, 4 of which were built by the same builder at the same time (and have Cadw listing), the fifth appears to be of a slightly different vintage. All five are collectively known as Rook Terrace and that is part of their official postal address. So there is 1 Rook Terrace, Quay Street and there is 1 Quay Street. Some of them have house names, so there is Roby, 2 Rook Terrace, Quay Street as well as 2 Quay Street. I can’t rename a section of Quay Street as Rook Terrace for two reasons: firstly, directly opposite them on the north side of the street are 16, 17, 18 and 19 Quay Street; secondly, continue further west and there are more buildings on Quay Street with official postal addresses of Quay Street.

The format of addresses doesn’t seem able to cope with stuff like this. I suppose I could pretend they were units of Rook Terrace, which is on Quay Street. But that isn’t the reality, they’re not units or flats but individual dwellings. Unlike Priory Terrace above, wrapping a building around them doesn’t fix up the postal addressing situation. I can’t come up with a way that reflects reality and would also let somebody get the correct postal address from the mapping data.

The third problem is Quay Street again. There are 3 houses, 24, 25 and 26 (all part of a terrace). These are separate houses, with separate doors. However, the Castle Cafe operates on the ground floor of 25 and 26 (each house has flats above the cafe) with an entrance at 26. The Cellar Bar operates in the basement of 24, 25 and 26 and shares a common entrance with the Castle Cafe. Current signage indicates that these were once the same business. However, in a lengthy Facebook chat with the guy running the Cellar Bar I learned that they are now separate businesses which overlap in their usage of three separate buildings and have a common entrance. (Just to complicate matters further, the Cellar Bar uses the ground floor of number 24 as a performer’s relaxation/changing room but as it’s not accessible by the public I’m ignoring that fact).

Anybody have good solutions to any of these problems?

Hi Brian,

Very quick response: I have a long unposted blog article about the peculiarities of terraces in Wales, and what that means for mapping. I really ought to publish it. The problem is now you have contributed yet more examples!



“Terrace names” are pretty common elsewhere too - see for example what the OS has for “Ingleton Road” in Chesterfield: . OSM only has the terrace names because that’s all that’s visible on the ground from the western street: .

A slightly different problem to the one that you’ve got is different names on the left and right of the street like . At least there there’s a somewhat settled solution - name:left and name:right (also see ).

I think it would be good to publish the blog article, even in imperfect form. Incorporate comments on the post itself together with the examples I’ve given to give a revised post later.

One bad idea I just thought of, for the Rook Terrace one, is to abuse the house number. So have the house number be “3 Rook Terrace” and the street is Quay Street. It probably causes all sorts of problems down the line, like maybe with Nominatim or other data parsers, but it’s the least-worst solution I’ve thought of for Rook Terrace. The alternative would be to have the number be “3” and the street name be “Rook Terrace, Quay Street” but that seems (for reasons I can’t put into words) worse.

There’s actually a similar problem further along Quay Street. A former pub, the Royal Oak, was divided up into four separate sections. One is a curry house so the common name distinguishes it. One is a mosque, so again the common name distinguishes it. Another is a vacant shop and another is of unknown function (possibly an unoccupied residence) so I have a problem with them. There are no house names or numbers on them.

BTW, Quay Street used to end at the Mwldan river. Until a car park was built there with the Mwldan culverted under the car park. So then Quay Street extended (as far as postal addresses go, not as far as maps depict) to a stationary boat along a footpath from the carpark. The stationary boat is another curry house, so must be mapped. :slight_smile:

Never mind, at the bottom of Quay Street (or halfway along, if you include the car park and footpath) I can turn north and enter the utter confusion that is Lower, Middle and Upper Mwldan. So there’s worse ahead.

That doesn’t seem to match precisely what I’m describing, but a different example of how bad things can get.

Ah, name:left and name:right might be useful with some of the other weird things I’ll have to deal with at some point. Although in one place it’s weird enough I may need to resort to name:upside-down or name:moebius-strip or something. :slight_smile:

Your abuse solution is actually what Royal Mail do, or ignore the name of the terrace or villas altogether. One of my favourites is on Garth Road in Bangor. The houses are now numbered sequentially along the road. I imagine this was done about the time that Upper Garth Road houses received numbers too (some time post 1963, when we moved from Bangor; IIRC there were no house numbers in the 1970s). However, all the bins have the number in the terrace & the name of the terrace painted on their sides.

@SomeoneElse: yes it is common elsewhere, particularly in industrial areas. However proportionately it is greater in Wales and very significant in more rural locations.

I’ll work on tidying it up.

And if you want odd addresses try: The road did originally get that far, but the address has never been updated to reflect the current road layout.

Not around here. Not every time. They are very inventive around here. Lots of different solutions. Perhaps one day they’ll come up with a solution that is logical.

Unless somebody speaks out against it soon, I’ll employ that bit of abuse to get the terrace on the map. Maybe with a fixme to invite somebody to come up with something better, maybe just a note saying perhaps there’s a better way. I expect various purists will moan if I do that, but unless they come up with something better…

I just found another one today, but worse. There is a terrace where the side of one member is on The Strand, but the front of all three houses are on Carrier’s Lane. Nevertheless, these are all “Brick Row, The Strand.” At least they have a different postcode from the places that are really on The Strand.

Many thanks. I suspect I may be able to provide a few more different examples as I map more of Cardigan.

I know of one like that here. Well, I don’t know if the road layout changed, or original names went out of use (I have dug out a historical map of Cardigan showing that several roads changed name, and part of one road changed name]) but the end result is similar to your example. Partway along High Street there one building with the address Grosvenor Hill. There is an alley that runs along the side of that building to another building, Grosvenor Cottage, which has the address High Street (not Grosvenor Hill as you might expect). There is a bus stop further along High Street which the timetables show as Grosvenor Hill although buildings around it are on High Street. Surprisingly, given the illogic that abounds, that building with the address Grosvenor Hill actually is partway up a hill on High Street.

I can understand what prompted the invention of “what3words,” MapCode, Makaney codes, etc.

I can also understand why Royal Mail decided that house name or number plus postcode uniquely specified an address and that anything else is superfluous. Maybe Royal Mail postcodes were designed by somebody who lived in Wales for a while.

Looking at it looks as the National Land and Property Gazeteer allows for blocks of flats within a terrace as its deepest hierarchy, but does it recursively, rather than the flattened structure that OSM uses. NLPG basically treats a terrace of houses much the same way as a block of flats, directly on the street. The hierarchy is constrained not to go beyond grand child level, even though the data structure might allow more.

I’d therefore suggest that OSM needs an intermediate level between street and house/name number to represent all address known to UK local authorities.

OMFG! How did you ferret that out? How did you then stay awake long enough to read through and get to the bit about terraces? I’d have given up after 6 pages of errata and never hit the paydirt. You did a very good job there.

It does indeed. I can see it applying to three situations in Cardigan, one of which breaks those rules as I understand them (it has its own access road) but is addressed using that convention anyway.

I assume that here you’re talking about the NLPG database schema, which seems somewhat complicated. A sort of denormalized normal form in an unrelational database. Who cares about database internals (except people writing code that makes it all work)?

All we have to do is invent something in the addr: namespace. I note that the document mentions this applies to both named terraces (at least 3 in Cardigan) and parades (at least 1 in Cardigan). And you say that it could be conceptualized as a horizontal block of flats. So how about addr:blockname? Maybe (probably?) somebody will come up with a better tag key, but block can apply horizontally as well as vertically. Only problem with blockname is possible confusion in the US, where cities are divided into blocks, referred to by two orthogonal street names. I see from a thesaurus that “block” in that sense came from “a compact mass of buildings.” Most of the synonyms for block (meaning area) have other usages in Merkin. How about addr:demesne? :slight_smile:

It doesn’t matter that current renderers won’t understand it because the map will reflect what I see on the ground: there are two houses numbered 1 on Quay Street (actually three if I count the block with its own access road). Until iD catches up (if it ever does) those of us that need to do it can enter the tag manually. The only other thing necessary would be to amend the wiki accordingly.

I also suspect other countries may have a similar intermediate level in some circumstances, so maybe addr:block needs a better name, depending upon what they use that intermediate level for. It’s also possible some countries may have more than one intermediate level. Do we worry about that now or deal with an immediate need and slap another kluge on the system later if it becomes necessary? I vote for immediate need, simply because these buildings are nagging at me to get mapped.

It seems to me it’s just a matter of agreeing the appropriate tag (whether here or in a formal discussion) then amending the wiki and waiting for editors (and maybe renderers) to catch up.