[Talk-us-massachusetts] MassGIS address import -- on duplicates and weird housenumbers

[Greg Troxel]

"Alan & Ruth Bragg" <alan.ruth.bragg at gmail.com> writes:

> Yuri,
> Great work, I agree and am fine with rolling up stacked unit numbers into a
> list addr:unit=A;B;C
> I've looked at a couple pages on syntax and think the proper delimiter
> might be a semi-colon. Maybe someone knows better.
> I don't think the quotes are necessary.

I think semi-colon is right, but I'm not sure that's the right approach.

It may be that the horizontal positions of a/b/c are really the same.
But it may be an artifact of how they came up with them.  It might be
that having multiple nearby nodes each with one address mitht be more
useful for OSM data consumers and more easy for human contributors to
build from.  But it might not be.

> There are already 35 buildings in MA with an address of 0. It's certainly
> suspicious. I'm in favor of not importing zeros. Import the addr:street etc
> but not the addr:housenumber=0
> There are two locations with the 0 address in Bedford.
>
> My town gis map shows a building at 32 YD Road that's identified as 0 YD
> Road in the massgis point data file.

So it sounds like this case of 0 addr is wrong.  If the fraction of 0
addresses that we check that are wrong is significant (>5%), then it's
probably best to entirely omit them, and to do the 35 by hand after
checking.  If part of the address is confused, it makes me suspicious
about the rest and the process.

> The other location with an address of 0 is in a stack of 110 nodes
> containing 77 different addresses  next to a road on Hanscom AFB. The
> POINT_TYPE is BMP.  BMP is noted in the massgis definitions as "oddly
> distributed buildings or unusually configured property boundaries ....we
> will consider shifting these points to more intuitive and suitable
> locations"

So probably BMP should just be omitted from this import; it seems hard
and a special case.