[Imports] New Orleans: importing buildings and addresses

Matt Toups mtoups at cs.uno.edu
Wed Oct 22 23:39:49 UTC 2014


On 10/22/2014 09:03 AM, Eric Ladner wrote:
> Random points in no particular order...
>
> One thing that might be helpful for querying the data and perfecting 
> the name conversion (expanding abbreviated street names can be tricky) 
> is using PostGIS + QGIS in your workflow for validation.
>
> Having all your address data loaded into Postgres allows the use of 
> SQL to query/manipulate the data and QGIS to check the GIS side of things.
>
> Look at the process on http://wiki.openstreetmap.org/wiki/MGC
>
> I'm not suggesting you change your process if it's working for you, 
> though.  But, working in a SQL database has it's advantages (rollback, 
> repeatable transform from source table to destination, able to query 
> the whole data set at once for arbitrary conditions, etc).

Thanks, this is good info. I think the python scripts are working well, 
but I have used QGIS to check a number of things.

> Also, check your address assignment and make sure you're not assigning 
> duplicate address nodes to buildings on the same lot (separate 
> garages, out buildings, sheds, etc.).

I have checked various parts of the city and not seen this happen yet, 
but I will add this to the list of things our reviewers should look out for.

> Looking at your street expansion code, it doesn't take into account 
> that French-like street names have the type at the beginning, not the 
> end (i.e. Rue Jessan, not Jessan Rue). Also, it doesn't look like it 
> handles expanding "St" in the street name to Saint.

Currently there are two expansions. One is the expansion of "DIR" which 
is one of N,S,E,W into "North" etc. There are only four simple cases. 
The other is on the "TYPE" field of the address, your ST BLVD etc. There 
are actually no instances of TYPE=RUE in the data. (You see it in 
expansions.py because I borrowed that list from elsewhere)

There are a small number of streets with "Rue" in the STREET field, with 
no TYPE field. These come out fine with no expansion: see houses on "Rue 
Delphine" in ward 15, precinct 18A. (Aside: you might think the "Rue" 
streets would be in the French Quarter, but you'd be wrong: they're all 
found in recently built subdivisions on the westbank.)


You're right, I'm not expanding ST to Saint. I never see this written 
anywhere, I think the street names are officially "St Philip" etc on all 
tax/legal/etc documents I've seen. Some would argue it is improper to 
expand ST to Saint, for example here: 
https://help.openstreetmap.org/questions/19609/saint-or-st-is-there-an-official-osm-policy

I'm leaning toward keeping the spellings as ex. "St Philip" in keeping 
with what's on the ground here. But we kind of have the worst of both 
right now: some Saints have been expanded by bots, others haven't. 
Here's an extreme example I just found of inconsistencies intersecting 
each other: http://www.openstreetmap.org/#map=17/29.92865/-90.06907

I'd like to correct those inconsistencies, but for now I think it's 
outside the scope of this import.


Other than the ST/Saint expansion, I don't see any other expansions that 
are a problem. The numbered streets ("1st" "2nd") are consistent with 
the non-expanded street names.

If you can point other cases out to me, that would be great. Individual 
.osm files are available here: 
https://github.com/mtoupsUNO/nola-buildings/releases/tag/v1.1 or as one 
.zip file here: 
https://github.com/mtoupsUNO/nola-buildings/releases/tag/v1.0

Eric, given your experience both in similar imports and in mapping this 
area, it would be great to add you to our list of import participants! 
As you see we have quite a few small precinct-size chunks to divide up 
and review.

thanks!
Matt
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.openstreetmap.org/pipermail/imports/attachments/20141022/bd1087fc/attachment.html>


More information about the Imports mailing list