numba is broken in 2022.02.04 image

[rabernat]

The following code does not work in our latest docker image 2022.02.04 with

• numpy 1.22.2
• numba 0.53.1

from numba import vectorize, float64
@vectorize([float64(float64)])
def add_one(a):
    return a + 1
add_one(1.)

# UFuncTypeError: ufunc 'add_one' did not contain a loop with signature matching types <class 'numpy.dtype[float64]'> -> None

I can fix it by installing the a newer version of numba, which requires an older version of numpy

• numpy 1.21.5
• numba 0.55.1

This seems like numpy/numpy#20735

xref https://discourse.pangeo.io/t/xarray-aply-ufunc-is-giving-new-error/2309

[rabernat]

Actually, our latest master is on numpy 1.22.3, where this should be fixed (see numpy/numpy#21139)

pangeo-docker-images/pangeo-notebook/conda-linux-64.lock

Line 318 in 1affb9f

https://conda.anaconda.org/conda-forge/linux-64/numpy-1.22.3-py39h18676bf_0.tar.bz2#58c2800b252e3519d855161057223df4

However, numba is still pinning < 1.22. So that's why we are stuck on numba 0.53.1:

pangeo-docker-images/pangeo-notebook/conda-linux-64.lock

Line 398 in 1affb9f

https://conda.anaconda.org/conda-forge/linux-64/numba-0.53.1-py39h56b8d98_1.tar.bz2#2b8674a6cb39ccd00833311aae90640e

So this works in our latest image!