[OSM-dev] [OSM-talk] Circular relation by user mapper_07

[Matt Amos]

On 7/16/09, Chris Browet <cbro at semperpax.com> wrote:
> 2009/7/16 Matt Amos <zerebubuth at gmail.com>
>
>> On Thu, Jul 16, 2009 at 10:41 AM, Chris Browet<cbro at semperpax.com> wrote:
>> > 2009/7/16 Frederik Ramm <frederik at remote.org>
>> >> Chris Browet wrote:
>> >>>
>> >>> I'd be curious to know how the api calculates the bounding box of
>> >>> such
>> a
>> >>> relation
>> >>
>> >>
>> >>
>> http://wiki.openstreetmap.org/wiki/OSM_Protocol_Version_0.6#Bounding_box_computation
>> >>
>> >>> assuming the api definition is "the smallest bbox containg all
>> >>> children bbox'es".
>> >>
>> >> It isn't.
>> >>
>> >
>> > Ok. If I understand well, it adds to a relation bbox all the nodes and
>> ways
>> > of children relations (+ the "pure" ways and nodes, of course).
>> >
>> > That won't remove the problem of infinite recursion: R1 -> R2 -> R1; it
>> will
>> > add to R1 all ways/nodes of R2, which itself will add all ways/nodes of
>> R1,
>> > which will...
>> >
>> > Bottomline question is: How does the api handle the case?
>>
>> short answer: it doesn't.
>>
>> long answer: the position taken by the API for both map calls and
>> changeset bboxes is that relations don't have physical extent - only
>> their node and way members do. bboxes are calculated only from the
>> immediate members of the relation, so recursion is necessary.
>>
>>
> Err... I'm feeling thick, today...
>
> You are saying that the api IS recursing thru the children relations and
> that the api also has an infinite loop problem, right?
> OTOH, you saying "only from the immediate members", would mean it DOESN'T
> recurse thru the relations?? (But it cannot be that or what would happen
> with relations only comprising relations).
>
> Sorry for bothering about this, I'd just like to be sure on how to best
> handle the case...

the second. it doesn't recurse through the relations.

cheers,

matt