[OSM-talk-be] Address stats in Belgium

Glenn Plas glenn at byte-consult.be
Tue Apr 16 09:25:40 UTC 2013


On 04/16/2013 11:01 AM, Marc Gemis wrote:
> Maybe I should start using the fixaddresses plugin
> I've been using the plugins terracer, building tool, etc. for over 1 
> 1/2 year. I even wrote a page on it: 
> http://wiki.openstreetmap.org/wiki/User:Escada/JOSM_and_Housenumbers  :-)
>
> But now that I have a python script to convert my GPX trails, they do 
> not fit perfectly in my workflow. I start from nodes that are tagged 
> with building=house and addr:housenumber=XXX (after importing my 
> modified GPX trail). I still use the terrace plugin to split though. I 
> do use the stylesheets too.
If the nodes area already a building node, I understand why that messes 
things up.  you will probably loose information when using terracer on 
them (there is some precedence).   What I really like about it is that 
when selecting the trio (ctrl-select) street/addrnode/building(way) , 
after pressing SHIFT-T, it keeps the street selected, that saves you a 
click when doing the next building.
>
> Did you try looking up businesses represented as nodes in buildings 
> with address information ? Nomatim does not understand this (I think). 
> Should we map for Nomatim, not sure ...
>
I just did this:"century 21 compas" , nominatim finds it right back but 
you're right from the results set it looks like it didn't pick up the 
housenumber of the building that encompasses it.  bummer, as that could 
count for a implicit relation between both.

There are still some duplicates in there on my fix list, like this 
take-away chinese restaurant:

"Xing Fu Lou"

There both building + node have housenumber, which I think I should 
remove on 1, but that was using the logic we just debunked.

The other way around doesn't seem to function either :

"Brusselsesteenweg 86, Zemst", finds back the house but not the 
amenity/shop.   The question now is, how can we 'fix' this? Duplicate 
address information will fix nominatim, but it messes up the map.

We should defenitely not map for nominatim more than we map for the .... 
map.   The difference is that nominatim is a lot more sensitive to the 
way data is presented/available than the map is.  The latter is slightly 
more 'forgiving' as it's a visual thing, more than a content thing.   So 
by nature it's more sensitive to this sort of situations.

Glenn





More information about the Talk-be mailing list