acrosscanadatrails at gmail.com
Mon Apr 6 22:02:12 BST 2009
The data set does have the "city:left" &"city :right" fields. (as part
of the DBF file) Although probably not useful because the city boundry
will eventually get shown.
And ya, use the last sample of canvec (as its the best) -Niagara falls Ontario
"addr:interpolation" i'll add that tag, thanks. ... but however these
arn't 'odd' 'even' ..
What about (from http://wiki.openstreetmap.org/wiki/Key:addr)
If a object has two numbers. Better use "addr:housenumber=<first>;<second>"
So the line can show a long stretch from <second> where it would be
the 'last'? Would this work?
>From the sample, it looks like the off-set way can be extracted from
the data. So then we still get the other details from the road shown.
Something that i dont know how todo. But will certainly keep the
learning curve going to help out.
Also, it appears from that sample that the road DO show the
'first/last' house number. .. as it's showing the start/finish of the
house numbers for the odd side of the road, and the star/finish for
the even side of the road.
FYI, in Canada.. "generally" Odd numbers are on 1 side and Even are on
the other. .. but that's not always true. So your right, by viewing
more data it might be possible to see just how 'generally' this is.
And for the off-set, ill leave that for the expert. ... since the
canvec2osm (using simple java) script is not really 'automatated'.
.the geobase2osm (using python) script is.
So I think best option is still the same, keep the streetnames.osm
files in the canvec2osm script (but as a reference, & not to be
imported), and let the geobase2osm script deal with it.. So hopefully
we can help out SteveS with the scripting as much as we can.
Across Canada Trails
On 4/6/09, Marcus Wolschon <Marcus at wolschon.biz> wrote:
> On Mon, Apr 6, 2009 at 8:44 PM, andrzej zaborowski <balrogg at gmail.com>
>> 2009/4/6 Sam Vekemans <acrosscanadatrails at gmail.com>:
>>> so for whats available in the dataset, along with the way listing the
>>> roadname;is_in and/or addr:city ... Is the listing of the first&last
>>> house number on the left side, as well as first/last on the right.
> it is "addr:city".
> There is no "left" or "right" as we use separate ways just for this.
> (There are reasond for it.) The perfect way are single houses
> drawn as polygons with building=yes and addr:housenumber=xy ,
> the next best thing nodes places where each house is and
> the next best thing is 2 or more nodes where single house-numbers
> are known connected by a way tagged addr:interpolaton=even/odd/...
> to we can calculate where each house would be.
> You can use Traveling Salesman to test actual, working
> house-number search.
>>> For the geobase2osm script we excluded this feature, as we couldnt
>>> figure the best way to tag.
>>> Now I found that the data is also available in the canvec set.
>>> So one option is to have it as a way over top of the roadway.
>> We're importing a small number of streets from a project called UMP
>> pcPL that has similar housenumber data as you describe (for *some*
>> streets). Here's how my script converts it to OSM:
>> We add a addr:interpolation on each side of the way with an arbitrary
>> offset and with some simple heuristics to make it look correct
>> (although obviously it needs a manual check like all imported data -
>> to see that it corresponds with reality). I didn't add any address
>> relations but addr:street value can be used to add relations
>> automatically if needed. Here's a sample:
> Sound cool!
> One suggestion:
> As this is automated anyway, try to add hints about what street the
> houses belong to.
> easy to implement version:
> Simply add a tag add:street=nam to the interpolation-ways
> slightly less easy to implement but easier to evaluate in address-search:
> add an associateStreet- relation
> I will download that area one of these days and play around with my
> address-seach a bit, okay?
> With so much test-data I may be able to improve the algorithm I use
> to associate the street with the house-numbers :) .
More information about the talk