[OSM-talk] Osmarender, bridges, layers and seams

[Dermot McNally]

The matter of renderer quirks and whether you should work around them
arose on another thread but didn't really go places. One principle
that was strongly backed was "fix the renderer, don't tamper with the
data". But before filing any bug reports I want to check the mood
among those of you hacking on the renderers, particularly
osmarender...

As of one of the major version bumps in osmarender, the way that
"normal" (definition below) bridges were rendered changed. Consider
the following ascii art to be a rendered bridge:

 0
 0
 0
 0
 0
|1|
|1|
|1|
|1|
|1|
 0
 0
 0
 0
 0

IOW, we have the bridge portion itself, which is a way having layer=1
and bridge=yes. Abutting it are ways not having either of these,
therefore implicitly at layer=0. Previously, this would be rendered as
a continuous road of whatever type with "bridge-marks" alongside the
part so-tagged. Brilliant. However, for a while now, an unsightly seam
has been getting shown at the transition.

This has led at least one mapper to get creative (IMHO unreasonably
so) with the layer attribute. For example, to work around the seams in
the above case, you might do the following:

 0
 0
 1
 1
 1
|1|
|1|
|1|
|1|
|1|
 1
 1
 1
 0
 0

This troubles me, though. Firstly, it's less "correct" in the layering
it implies. Secondly, you still get your seam, it just moves away from
the bridge boundary.

So is there any reason we can't get rid of these seams? Their presence
is a temptation to adopt workarounds. Those workarounds have already
cost me a lot of effort reinstating good data destroyed as a result of
rework that would never have been performed if the seams hadn't been
showing.

Do they serve any purpose? Is it hard to avoid them?

Thanks,
Dermot