[Talk-ca] Vancouver address normalization

[Adam Williamson]

Hi, folks. Just wanted to pass along a heads-up: I went through every
object in Vancouver (City of Vancouver bounding box) with an
addr:housenumber tag today, and did some cleanups. What I was primarily
interested in was correcting addr:street tags: a lot of them used
variant formats. For instance, some things on West Hastings Street would
be listed as "Hastings", some as "Hastings St", some as "Hastings St.",
some as "W Hastings", some as "West Hastings", and some as (the correct)
"West Hastings Street". It's amazing how many different ways there are
to render street names. =)

Anyhow, I made them all match the names used for the actual street
objects themselves, which are all in the 'most correct' form as I
understand it (no contractions, regular casing - "West Hastings Street",
"East 49th Avenue" and so on). I ran this by pnorman before sending it
out and I don't *think* there'd be any reason to object to it, but I
thought I'd let the list know.

Changesets:

http://www.openstreetmap.org/changeset/20153277 (a very large one, I'm
sorry - I'm used to the software convention of committing all related
changes as one large block, but Paul has let me know OSM prefers small
staged commits, and I'll do that in future)
http://www.openstreetmap.org/changeset/20153308
http://www.openstreetmap.org/changeset/20153377

I also did a few other sanitizations along the way, trying to stick only
to things which are completely unambiguous - adding street names for
objects with only a house number, correcting completely incorrect
formatting (like "the entire address stuffed into the addr:street tag"),
fixing spelling errors (yes, I checked to make sure it's not an
intentionally misspelled business name), stuff like that.

One thing I did which I thought was a no-brainer but realized later may
not be quite so clear-cut is to use the addr:unit tag for the common (in
Vancouver) case of a business address which is for a particular
commercial unit in a building. In Vancouver these addresses are commonly
given as "Unit number-Street number Street name", e.g. "101-346 Some
Street". Previously different editors had rendered these differently.
E.g.:

http://www.openstreetmap.org/node/2560866260/history

previously had addr:housename=#642 , I made it addr:unit=642 . I can't
find an example of the other formats right away, but some simply put
both numbers in the 'addr:housenumber' tag - e.g.
addr:housenumber=101-342 - and some did stuff like
addr:housenumber=(unit number), addr:street=(house number) (street
name).

Some of those are pretty inarguably wrong, but now I look at the
guidelines, using addr:housename or putting both numbers in
addr:housenumber seems to be at least acceptable, so if I was starting
over it might have been more conservative to leave those alone. Still, I
don't think using addr:unit could reasonably be considered worse than
any of the above, and at least now they're all consistent.

Most of the changes didn't require any source beyond the existing data
and the BC Mosaic aerial imagery; where I needed to double-check a
business address I went direct to the business's website.

Please do yell at me if I'm messing stuff up :) Thanks!
-- 
adamw