Basic support for varying longitude ranges. #350
Conversation
| from iris.cube import Cube | ||
|
|
||
|
|
||
| class Test_linear__circular_wrapping(tests.IrisTest): |
There was a problem hiding this comment.
There must be other interpolation tests somewhere? Is it worth putting this with those? Or moving those to here (in a separate PR?).
There was a problem hiding this comment.
There must be other interpolation tests somewhere?
Indeed there are. And yes, it would be good to move them here in a separate PR (I've deliberately not done that in this PR to keep it obvious what is going on).
There was a problem hiding this comment.
Is there a follow-on issue to do this?
There was a problem hiding this comment.
There is now ... #353 😉
|
@rhattersley - looks like the gribapi build has fallen over on travis-ci. @pelson showed me that you can restart the build on travis-ci (for your own PRs) using the drop-down menu. Could you try this? |
| # | ||
| # You should have received a copy of the GNU Lesser General Public License | ||
| # along with Iris. If not, see <http://www.gnu.org/licenses/>. | ||
| """ |
There was a problem hiding this comment.
I'm not sure what impact having a sub-folder will have our visual testing. I know cartopy can handle it, but not 100% that the standard mpl image testing functionality can. We can cross that bridge when we get to it though.
bump. |
Have done, but I don't hold out much hope - the build of the GRIB API has failed on my other completely unrelated PR (#354). |
|
This time the travis build timed-out while waiting for apt-get update to work its magic. Given we don't even have full test coverage on travis (and hence must not rely on travis anyway), please can we not hold up this PR because of some travis strangeness? |
| # Check that we can interpolate onto [0, 360) from a Cube | ||
| # defined over [-180, 180). | ||
| cube = self._create_cube([-180, -90, 0, 90]) | ||
| samples = [('longitude', [0, 90, 180, 270])] |
There was a problem hiding this comment.
Please add some intermediate points (e.g. 45 deg and 315 deg) to test the linear interpolation. e.g. for 315 deg is it using the -90 and 0 values to estimate the desired sample point?
There was a problem hiding this comment.
Perhaps it's over the top 😉, but I've added quite a few more points (including intermediate 45 deg points) to see what happens with wrapping at each end.
And since we seem to be a gnat's whisker from merge, I've rebased & squashed.
Thanks.
Rebased to include #357 ... now working fine. :-) |
Basic support for varying longitude ranges.
This pull request adds better support to
iris.analysis.interpolate.linearfor circular longitude ranges with different starting points.(*)For example, starting with a Cube defined with longitudes covering the range [-180, 180) it is now possible to re-sample correctly with longitudes in the range [0, 360).
(*) Created in response to a user bug report where
iris.analysis.interpolate.regridwas failing to deal with two Cubes with different longitude ranges.