[Tagging] RFC 2 - addr:interval

Jmapb jmapb at gmx.com
Thu Jan 7 19:41:38 UTC 2021

On 1/7/2021 11:27 AM, Kevin Kenny wrote:
> On Wed, Jan 6, 2021 at 6:38 PM Jmapb <jmapb at gmx.com
> <mailto:jmapb at gmx.com>> wrote:
>     It's entirely reasonable, in context, that the official address of the
>     entrance in question might be "174-190 23rd Street".  Yes, it's
>     obviously derived from a range, but using the whole range as the
>     housenumber in the official address does happen!
> And having housenumbers that are hyphenated and NOT ranges happens
> too.   Look around
> https://www.openstreetmap.org/#map=19/40.73343/-73.80230
> <https://www.openstreetmap.org/#map=19/40.73343/-73.80230> - there are
> a vast number of them.

Yes, offhand there are 3 different meanings of the hyphen in housenumbers:
1) Indicating a range of addresses, including the bounds and possibly
numbers in between, usually ambiguous as to evens/odds/all. Example,
1-31 Helen Drive https://www.openstreetmap.org/way/694324667
2) Indicating a range of addresses where the entire range is associated
with a feature and used as an atomic housenumber. Examples, Hercules
Heat Treating Corporation https://www.openstreetmap.org/way/250351628
... KFC https://www.openstreetmap.org/way/249606377
3) Hyphen used as literal part of the housenumber string, not indicating
a range. Examples, thousands upon thousands of addresses in Queens NY,
some of which look like ranges out of context. (Sometimes also found
before an alphabetical suffix, eg 101-B, though I tend to transcribe
these without the hyphen.)

(Are there negative housenumbers in use? I exhausted my Overpass quota
looking for one that wasn't an obvious error. But if so, they'd fall
under #3 above. Here's hoping we don't have to deal with negative ranges!!)

If there were a way to assert that the value of the housenumber tag is a
range (and specify the interpolation), it would allow geocoding software
to correctly locate addresses in that range. Currently a mapper who
wants to convey this info to a geocoder has to use an interpolation way,
map individual housenumber nodes, or set the housenumber to a
semicoloned list of all possible values in the range. These are all ugly
but especially bad if the mapper doesn't know whether the range is
question is #1 or #2 above -- a Nominatim search for 100-104 does not
find housenumber 100 or housenumber 100;102;104.

Would it also be advantageous to assert that a range-like address is not
a range? I'd hope that, knowing that any hyphens found in a housenumber
are just for formatting, a geocoder could then index the housenumber
without hyphens. This would allow digit-only address searches to work in
Queens without requiring the correct hyphen placement. Which would be
good because it's not uncommon for Queensians to omit the hyphen in both
speech and writing. (This currently works in Apple, Google, Here, and
Bing maps, by the way.)

> (Note that a number like 166-11 can't possibly be a range.)

No it can't, but beware abbreviated ranges like 305-11 Douglas Street:

Hopefully these won't make it onto the map in that form but it may
happen on occasion. Theoretically a QA tool could sniff them out.

Cheers, Jason
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.openstreetmap.org/pipermail/tagging/attachments/20210107/9b4bc596/attachment.htm>

More information about the Tagging mailing list