[Imports] NYC building + address import - to merge or not to merge?

Paul Norman penorman at mac.com
Fri Oct 25 02:33:57 UTC 2013

I used a different strategy than Eric, but I also looked at this question


For the queries I ran to see how McDonald's were merged to ways I used this
with an osm2pgsql database to find McDonalds mapped as a node within a
building with no other POIs in the building:



  FROM planet_osm_point AS pt

    JOIN planet_osm_polygon AS bp ON ST_Intersects(pt.way, bp.way)

    LEFT JOIN planet_osm_point AS po ON (

      ST_Intersects(bp.way, po.way)

      AND po.osm_id != pt.osm_id)

  WHERE pt.name='McDonald''s'

    AND bp.building IS NOT NULL

    AND po IS NULL;


Getting the number mapped on buildings is simple, as is the total number
mapped as nodes, and with subtraction you can get the same measures that
Eric did.


There are


* 2208 McDonald's points inside building polygons with no other points

* 3315 McDonald's polygons

* 1484 McDonald's points inside building polygons with other points inside

* 3721 McDonald's points not inside a building polygon at all


The first two are the relevant numbers for this.


To change it to addresses is simple, but expect a much slower query as there
are a lot more addresses.


I believe this is actually a more accurate method than Eric used as it
explicitly considers if a node has other POIs in the same building, but
Eric's result was conservative in finding that it is at least 10:3 in favour
of addresses on the way.



From: Clifford Snow [mailto:clifford at snowandsnow.us] 
Sent: Thursday, October 24, 2013 7:17 PM
To: Alex Barth
Cc: imports at openstreetmap.org
Subject: Re: [Imports] NYC building + address import - to merge or not to



On Thu, Oct 24, 2013 at 5:57 PM, Alex Barth <alex at mapbox.com> wrote:

Eric Fischer and I got curious, he just ran numbers on the entire planet, I
posted a summary here: http://www.openstreetmap.org/user/lxbarth/diary/20261


The short is that for better or worse, tagging addresses on buildings where
possible is most common in OSM. Whether that's good or bad is a different
discussion, for OSM NYC we'll stick to this convention. Especially as the
location of the address point is not always meaningful in this dataset.

Can you post a howto on how you pulled the data? I'm struggling trying to
get address info out. 





OpenStreetMap: Maps with a human touch

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.openstreetmap.org/pipermail/imports/attachments/20131024/ab22f664/attachment.html>

More information about the Imports mailing list