[Imports] Belgium address import
sanderd17 at gmail.com
Mon Oct 6 11:49:25 UTC 2014
2014-10-06 9:59 GMT+02:00 Ben Abelshausen <ben.abelshausen at gmail.com>:
> There was a TODO-list on this last year and I asked the talk-be list for
> help but there was nobody willing to do this at the time:
> - Get a better feel for the quality of the data and formulate this in a
> better way than just saying 'good enough'.
>From what I see, the address nodes are in the exact centre of the parcels.
This is often not on a building, or near the entrance, but the precision of
the parcel boundaries (and thus the precision of those nodes) is way higher
than any OSM data, though for us it's a bit a useless position and needs to
be interpreted by humans. I also haven't found any human errors in their
dataset this far. Though there will certainly be some human errors, I think
the amount of human errors will be smaller than what can be achieved by
- Find a way to keep track of progress.
I thought the French address manager was a good way to track progress. We
just need lists of streets, and the ability to mark them done, which is
exactly what that tool offers (displayed on a map, which is nice though not
neccesary). Next to that, it also offers an easy button to import the data
in JOSM, which is nice.
- Find a way to compare OSM addresses with the CRAB database.
That's a hard one, as there will be differences by design. Sometimes a lot
bigger than you'd expect when you have big parcels. At most, since we don't
have parcel boundaries, the list of housenumbers per street can be
compared. Making an overpass query that allows to extract lists of
housenumbers per street won't be that hard. Anything more complicated will
be impossible to do automatically I guess.
> I stopped after that because coordinating this in a good way will take
> time and resources and cannot be done by one person (me at that time).
> @sander: suggestions have already been made on the import proposal page:
> Feel free to change these. Your suggestions don't seem different from what
> is already there but maybe there is room for improvement on the wiki ;-).
> I'm willing to spend some more time improving the python script to extract
> only the relevant addresses from the DBF-files.
> Met vriendelijke groeten,
> Best regards,
> Ben Abelshausen
The main thing I'm worried about is maintenance. This will be a continuous
effort, and a street should be marked "unreviewed" when there are new
buildings in the CRAB data, or the houses are re-numbered.
The AGIV file is updated continuously, so we should poll for changes. But
they do offer diff files that could be used. This is also why I don't want
to start using the tool yet. As the data will probably have to be imported
again, erasing all previous "complete" markings.
-------------- next part --------------
An HTML attachment was scrubbed...
More information about the Imports