[Talk-us-massachusetts] Stacked address points from MassGIS

[Greg Troxel]

Yury Yatsynovich <yury.yatsynovich at gmail.com> writes:

>> So what's the real benefit for importing stacked units at all, in terms
> of helping map users?
> A lot of addresses (in Boston, I guess, majority) are addresses for
> buildings with 2 and more housenumbers -- and all of them are stacked on
> top of each other in MassGIS. So if we ignore stacked points then a
> substantial share of buildings will remain untagged. It seems better to
> have housenumber=93;95 on a building than nothing at all.

[longer reply since i was pretty terse earlier]

I didn't mean to ignore points that hae the same coords, and
addr:street, and

  addr:housenumber=93
  addr:housenumber=95

That implies that those two addresses either really are the same point
or that they are close and the database doesn't really know and is
assuming.  So having them as 93;95 on the same node seems reasonable.

What I meant is that if we have

  addr:housenumber=93
  addr:housenumber=95
  addr:housenumber=95 addr:unit=A

Then the 95A should be skipped, because 1) everybody expects 95 and 95A
to be in the same place, unless there is a different point for 95A, so
it doesn't hurt those seaching for it to be missing, and 2) there is no
agreed-upon way to represent all of the above in one node, and adding a
second node seems like more clutter than it is worth.  Especially since
it's likely that 95A in many cases really is in a different place.

In essence this is a middle ground skipping a form of stacked points
that are too hard to represent and the loss of which does not cause any
trouble.  Again, for now - if this becomes our biggest problem in a
phase N, we can check into it.  But we will need to hand-evaluate a
bunch of such somewhat-irregular points to see if the data is good.

> I agree with the comment on housenumbers like 93-95 -- probably, they
> should be omitted from import, especially that, as a rule, at the same
> location are the points with housenumbers "93" and "95".

Definitely if they are dups, but I think anything with a hyphen is
suspect.  So one approach is just to skip them (for now - I keep
pointing out that declining to import some data does not really
constrain the future, but once imported we have to deal with it if
wrong).  The other would be to hand check a bunch of them and see if
substantially all are correct.  If not, they don't meet quality
standards.  I suspect things with hyphens are going to turn out to have
a non-trivial number of issues, mostly because I have never (maybe super
rarely) seen an actual address of a building/etc. with hyphens, but I do
seem to see them on parcels, often such that there are two actual
addresses on the parcel.
-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 162 bytes
Desc: not available
URL: <http://lists.openstreetmap.org/pipermail/talk-us-massachusetts/attachments/20180911/e49f6e17/attachment-0001.sig>