Observations of things that did not look right to be about the proposed implementation of our new API (in no specific order):
We currently don't handle dictionary input for sample points for cube.interpolate
More importantly, we don't support single pair support (coord, value) only ((coord, value)) for cube.interpolate
or ((coord)) and ((value)) for LinearInterpolator.__call__
Why was the extrapolation dtype changed to nan from its default ('linear') on some of the existing 'master' tests?
collapse_scalar does not belong to the LinearInterpolator class, should be only in the cube.interpolate method (has nothing to do with the interpolation itself).
I have chosen not to modify the tests which belong outside our new testing framework (i.e. lib/iris/tests/test_interpolation.py etc.)
Observations of things that did not look right to be about the proposed implementation of our new API (in no specific order):
LinearInterpolator.__call__
collapse_scalar
does not belong to theLinearInterpolator
class, should be only in the cube.interpolate method (has nothing to do with the interpolation itself).I have chosen not to modify the tests which belong outside our new testing framework (i.e. lib/iris/tests/test_interpolation.py etc.)