[Talk-us] Proposed import cleanup: NYSDEClands

[Kevin Kenny]

Oh, another question.  Some of the New York state land parcels have
rather complicated topology, and the previous import didn't get them
entirely right: duplicated nodes, crossing ways, nodes close to other
ways, and so on. Moreover, the upstream data are fairly arbitrarily
divided. An example is Burnt-Rossman Hills State Forest (which has
some recent changes from me that consisted of detaching the boundary
from crossing ways and deduplicating nodes).

>From the upstream system, this arrives as six separate chunks,
corresponding to ways 32035570, 32026630, 32002834, 32047624, 32035988
and 39186229. Some of the fragmentation appears to be simply to avoid
having holes in any of the polygons (Why this is done is unclear: the
shapefile uses multipolygons to represent the parcels, so they can
support inner rings.)

My inclination would be to use PostGIS to coalesce all of these using
ST_Union, and then import the simplified multipolygon, which the tools
surely know how to do. I think that would be more in keeping with our
data model, and would keep us from rendering internal borders on the
parcels. It loses the LANDS_UID of the parcels, but I don't think
that's a particularly useful thing to keep around.

The rule for coalescing would be to group by facility number, so all
the parcels of Burnt-Rossman Hills State Forest would be one relation,
while the ones of adjacent Mallet Pond State Forest would be another.

With all this said, I'm better at PostGIS programming than at OSM
modeling, so I could be off in the weeds here. Does this idea make
sense?