[Imports-us] Portland Metro Area Building Import

[Darrell Fuhriman]

> So it's about a dataset of 1:1 building correlations and then another
> dataset of 1:multiple and then handling it that way? So there will be
> some "holes" in the intermediate?

Yes, that is correct.

> You have dataset A released today. In a year or two you'll have
> dataset B. There will be a delta between the two that will hopefully
> capture new buildings, removed buildings, new addresses, etc.
> 
> I'm not even asking how to handle this delta from a community
> standpoint (though that'd sure be nice) but rather just "Have you
> tried to perform this delta and looked at the results?"


It’s an issue I’ve been thinking about a lot. Given that there is an allergic reaction to storing keys (plus I have issues with the current method of defining keys in the source data anyway, and there’s talk of changing them, so I’m not sure I’d want to use them in the first place), there needs to be another method.

Given that the number of building changes should be pretty small (address changes are a different issue, in particular point addresses, as opposed to way addresses — luckily addresses are their own natural key).

My current thinking is that it will have to be done with a spatial query and maybe some data caching.

Adding buildings and addresses is easy, and is essentially no different from what we’re proposing here.

Deletions and changes are more problematic, but my inclination is to do this:

If the way/point has not been changed since it was imported, then trust the source data, and update OSM. 

If it *has* changed, then I think flagging for manual review/conflation is appropriate. I had considered other things like checking for whether or not the geometry had changed, but I think that’s more difficult to track and would probably require as much manual checking as to not make it worth the effort.

 There are obviously a few potential edge cases here, but this should cover the vast majority of issues.

Given that the expectation is that this will be done (hopefully) quarterly, my hope is that the strange cases will get fleshed out pretty quickly.

Darrell