You can clone with
No one assigned
I've got layers in PostGIS in latlong, some of them are the entire world, e.g. -180,-90,180,90
When making a map in Lambert conformal conic (and probably all other conical projections), the logic for intersecting envelopes and calculating layer bbox isn't quite right.
Currently, feature_style_processor.hpp converts the layer extents into the map projection before comparing. If you convert -180,-90 to lambert it's usually above the north pole, the resulting bbox is non-sensical. (see http://www.mgaqua.net/AquaDoc/Projections/img/Lambert%20Conformal%20Conic.jpg)
This patch reverses things a bit, and converts the map extents into the layer's projection, then compares and clips. It's actually faster because you end up doing 2 fewer projection transforms.
It's not perfect, but works well for well-centered conical maps.
Forgive me if my patch doesn't work, this is my first time doing something like this.
[springmeyer] roka, thanks for the patch. I'm traveling atm, but I'll look forward to taking a look soon. See BoundsClipping for more info on this issue, particularly #549, which came up last month, and I forgot to address so I just stashed link to mailing list entry in that ticket.
[springmeyer] fixed in r2784.
first forward map ext to layer srs, falling back to backward projecti…
…ng layer to map (previous behavior). Also leverage new clip function of box2d<double> class - closes #402, #548, and #549 (and indirectly fixes #308 combined with #506) - see also http://trac.mapnik.org/wiki/BoundsClipping