[OSM-dev] multipolygon relation and non-closed ways

[Pedro Larroy]

Hi

Thanks for you edits.

So for overlapping inner polygons the only way to differentiate them
is by tagging as far as I can see.

>From the link you posted, in case#9 you need to use the role to build
the polygons for example:
http://postgis.refractions.net/docs/images/st_isvalid09.png

Otherwise, looking at the Algorithm page that you mentioned, doesn't
the step RA-4 become ambiguous?  In this particular case there would
be 3 ways that start at one of the nodes where the inner polygon is
touching the outer thus several ring possibilities depending on which
way is choosen.

I was looking also at complete_ring in osmium's builder.hpp, and seems
to me that it just chooses the first suitable.

The algorithm that I'm using stores the endpoint on a hash to make the
rings, so if there's more than two is not possible to have an clear
result.

Am I missing something here?

Pedro.


On Sat, Oct 13, 2012 at 9:49 PM, Frederik Ramm <frederik at remote.org> wrote:
> Hi,
>
>
> On 13.10.2012 21:35, Pedro Larroy wrote:
>>
>> Thanks for the link. I have updated the wiki with requirements for
>> valid multipolygon relations:
>>
>>
>> http://wiki.openstreetmap.org/wiki/Relation:multipolygon#Valid_Multipolygon_conditions
>
>
> I made a few changes to your list in the Wiki. (Frankly I am surprised that
> you thought the list was necessary, as I believed everything to be explained
> properly already.)
>
> First of all, the "role" is not actually required to determine validity, and
> I have removed your references of "... belonging to the same role". The
> algorithm laid out on the multipolygon page takes care of any valid
> multipolygon whatever the role; and algorithms relying on the role are
> broken. The use of the inner/outer roles is recommended because it adds
> redundancy and will often make it clearer what the mapper intended to do,
> but these roles are not a condition for validity.
>
> Also, the ways in a mutlipolygon relation do not have to form "a closed
> chain" (your words) but "one or more closed chains".
>
> You wrote that inner polygons "should" not overlap with outers; I changed
> that to "must" because the polygon is invalid otherwise.
>
> Also, your rule that no more than two unclosed ways may meet in a point is
> not correct in the case of touching inner rings, where any even number of
> unclosed ways might touch in one point.
>
> Bye
> Frederik
>
> --
> Frederik Ramm  ##  eMail frederik at remote.org  ##  N49°00'09" E008°23'33"
>
>
> _______________________________________________
> dev mailing list
> dev at openstreetmap.org
> http://lists.openstreetmap.org/listinfo/dev