remove mamba's pin #105
remove mamba's pin #105
Conversation
ioos_qc/qartod.py
Outdated
| return w.apply(np.ptp, raw=True, engine='numba') | ||
| except (ImportError, TypeError): | ||
| except (ImportError, TypeError, numba.core.errors.NumbaTypeError): |
There was a problem hiding this comment.
In a way numba never worked with masked arrays. See numba/numba@25f17bb
However, if we do want numba's speed-ups, we could try to use it when the Window inputs do not create NaNs and require masked-arrays. I could not figure this out yet so, for now, I'm proposing we just check for this numba error and let it default to non-numba behavior. If/when we implement a masked-array workaround we should keep this check anyway to future proof any mask-array that may pass here.
There was a problem hiding this comment.
OK. I think I got to the bottom of this one. In fact, if the Windows produces NaN, then there is nothing we can do but skip numba and this implementation is OK to do so. However, if the Windows doesn't create NaN, numba will be used without a problem.
TL;DR numba acceleration will happen when the window parameters allow it.
|
@iwensu0313 can you take a look at this one? it will allow us to install ioos_qc in environments modern Python + mamba. Although, the mamba acceleration there won't be available. See my comment above on the reason why. |
|
Self-merging this one b/c this will allow us to install |
Follow up on #102.