Proper skipping of Raster datasource when map.srs != layer.srs

[artemp]

Until we have proper on-the-fly reprojection support for raster datasources when map and layer srs's don't match it seems like we can skip the featureset being called. Need to look into best place to check for (non)intersection in native srs, whether in feature_style_processor or in datasource...

[artemp]

[springmeyer] see r1717, needs more thought in trunk.

[artemp]

[albertov] I'd say that a good place to check is at the beginning of [source:trunk/plugins/input/gdal/gdal_featureset.cpp#L79 gdal_featureset::get_feature](assuming the Query carries the srs information). ie: If the srss don't match then return a null feature.

The reason is for doing it here is that we won't need to modify anything but this method when we implement on-the-fly reprojection, the above conditional would change to: If the srss don't match then reproject query.extent into native extent and return reprojected raster.

Sounds sensible?

Alberto

[artemp]

[albertov] BTW, I think that feature_style_processor already checks for extent intersection with srs-normalized extents, doesn't it? source:/trunk/include/mapnik/feature_style_processor.hpp#L130.

[artemp]

[springmeyer] Replying to [comment:2 albertov]:

I'd say that a good place to check is at the beginning of [source:trunk/plugins/input/gdal/gdal_featureset.cpp#L79 gdal_featureset::get_feature](assuming the Query carries the srs information). ie: If the srss don't match then return a null feature.

The reason is for doing it here is that we won't need to modify anything but this method when we implement on-the-fly reprojection, the above conditional would change to: If the srss don't match then reproject query.extent into native extent and return reprojected raster.

Sounds sensible?

right on, yes sounds sensible. mapnik::query does not currently contain srs or whether map.srs == layer.srs but we could add. I'll look into this after the 0.7.1 release.

• dane

[artemp]

[springmeyer] ps. two things:

1. thanks so much for quick comments on tickets - its awesome as I tend to just stash ideas on tickets for later reference and having quick feedback is invaluable once I get around to clearly thinking about something.

2. I've got a patch to optimize gdal input in small ways (mainly moving a few more things into class members to we don't have to re query the extent, etc) that gives a slight speedup. I'll add this soon to trunk (it was a patch originally against 0.7).

[artemp]

[springmeyer] see also crashes in #599

[artemp]

[springmeyer] artem just addressed this in: r2228