[Talk-GB] Postcode finder based on OSM data

[Matt Williams]

On 17 March 2011 12:29, Ed Loach <ed at loach.me.uk> wrote:
> Craig wrote:
>
>> A practical approach would seem to be for an implementation to
>> accept
>> multiple instances of the street role, as either ways or
> relations.
>> Where a child-relation of role=street is treated as a collection
> of
>> segments.
>
> It does seem logical. Especially as any street currently mapped with
> an associatedStreet relation which is then split at a later date (by
> someone adding a bus route, change in speed limit, or whatever
> reason they split it) will end up as an associatedStreet relation
> with two street members anyway in all likelihood. I doubt the
> editors are at a stage where they can split such a relation into two
> separate relations and get the right houses with the correct street
> section automatically (and for some streets I can imagine this will
> never be possible automatically reliably anyway). I'd be surprised
> if there aren't already a number of associatedStreet relation with
> more than one street member, but not sure how easy it would be to
> find out.

I'm commonly creating associatedStreet relations with multiple
role=street members since there's no other way to easily map reality.
I don't think it should be a problem. Most likely the people who first
defined the relation didn't consider the possibility of split streets.
I certainly don't consider it a 'bug' in the postcode finder.

As for statistics on this, it should be quite easy for me to make a
measurement this evening.

-- 
Matt Williams
http://milliams.com