Prevent segfault in LambertAzimuthalEqualArea #1100
Conversation
Avoid shifting the antipode out-of-range of the PlateCarree projection before transforming to LambertAzimuthalEqualArea to compute the projection boundary.
|
The approach to me looks like a reasonable tweak to the original. Can you add the test case from #1099 as a test? |
There was a problem hiding this comment.
Works for me. Thanks @ajdawson! @dopplershift's suggestion of a simple test is worth the effort IMO.
| crs = ccrs.LambertAzimuthalEqualArea(central_latitude=latitude) | ||
| expected = ('+ellps=WGS84 +proj=laea +lon_0=0.0 +lat_0={} ' | ||
| '+x_0=0.0 +y_0=0.0 +no_defs'.format(latitude)) | ||
| assert crs.proj4_init == expected |
This comment was marked as resolved.
This comment was marked as resolved.
Sorry, something went wrong.
ajdawson
force-pushed
the
lambert-segfault
branch
from
93066ca to
7500d40
Compare
These tests produce the LamberAzimuthalEqualArea CRS with central latitudes at the poles to ensure the extreme cases work as required.
ajdawson
force-pushed
the
lambert-segfault
branch
from
7500d40 to
2453bfd
Compare
|
Thanks @ajdawson! |
|
Thanks for the fast action on this. This is one of my first issues opened here. Can you please tell me how to figure out in which cartopy version I should expect to see the fix released? |
|
You should expect to see it in 0.17.0, I don't know when that will be though. If your use case can cope with using -89.9 as the central latitude then that is a possible workaround for the current version (e.g. might be ok for plotting, but not for more precise transform requirements). |
|
We finally released v0.17 @mjbrodzik, so an update of your environment should see you right. 👍 |
|
Thank you! I've updated to 0.17.0 and confirmed that I can now initialize LamberAzimuthalEqualArea with central_latitude=-90.0. |
Computing the boundary of the Lambert azimuthal equal area projection requires finding the anitpode of the projection centre in lat and lon, shifting it a little bit in latitude then transforming to the native coordinate system. This shifting is required because the exact antipode is not within the target projection. The implementation of the shift allowed for out-of-range latitudes when the central latitude is less than -89.9 degrees. This change avoids shifting the antipode out-of-range of the
PlateCarreeprojection before transforming toLambertAzimuthalEqualAreato compute the projection boundary.Fixes #1099.