[Talk-us] What's causing rendering artifacts in the Southeast?

[Matthias Julius]

"David ``Smith''" <vidthekid at gmail.com> writes:

> On Fri, Nov 27, 2009 at 2:18 AM, Dale Puch <dale.puch at gmail.com> wrote:
>> It sounds like tile at home needs to trim data at the tile border, and then
>> close open ways along the border of the tile...
>> Just my uneducated opinion.
>
> Let's say you have a tile with two ways crossing it, and they don't
> intersect in the tile.  Both of these ways are "outer" members of the
> same multipolygon relation, and they don't share any nodes.  There's
> no way to determine whether to color in the space between the two
> ways, versus the rest of the tile except for that space, unless the
> renderer knows whether the ways wrap around the polygon clockwise or
> counterclockwise.  One might think that it's safe to assume (as with
> coastlines) the answer is clockwise, but when multipolygons are used
> to map adjacent features with shared ways, that condition cannot be
> true for both/all multipolygon relations; so that assumption must be
> thrown out.  In order to paint the polygon correctly, the entire
> polygon must be known, which means loading the entire multipolygon
> relation*.

Well, it would be fairly easy if the ways in a multipolygon were
required to be ordered.  Then, when there is a gap a renderer could just
connect the ways.

At least this would work in most cases.  It doesn't work when the area
is being inverted or becomes self-intersecting when you do that.

Matthias