Conversation
scranford1
force-pushed
the
bugfix/interpolation-dimensions
branch
from
b2df68d to
8bd8c32
Compare
|
|
Unit tests are now passing at same percentage as develop. (18 failed, 987 passed, 48 skipped, 9 deselected, 2 xfailed) |
|
@dhinckley-creare Tagging you in this pull request as well, mostly related to stacked coordinate changes and one change for CRS formats. |
| else: | ||
| dtype = object | ||
| return np.dstack([c.coordinates.astype(dtype) for c in self._coords]).squeeze() | ||
| return np.stack([c.coordinates.astype(dtype) for c in self._coords], axis=-1) |
There was a problem hiding this comment.
@scranford1 this is the only line that I'm not too sure about.
What gives me pause here is that squeeze would crush a pointless dimension regardless of order but I'm not familiar enough with stacked_coordinates to know if axis=-1 is always correct.
There was a problem hiding this comment.
From my understanding the squeeze is being used to remove the extra dimension that is introduced during the conversion to a 3D shape. However for single values it can end up removing a dimension which should be kept. With stack and axis=-1, 1D arrays are staked to go from (N,) -> (N,k) and 2-D Arrays are stacked to go from (N,M,) to (N,M,k) where k is number of arrays in stack.
2 participants
Changes