Pin max numpy version in `setup.py`

Feature request

As of 6/22/22, numpy v1.23 has released. It is possible to install numpy v1.23 alongside numba, though when importing numba, the following error is thrown:

File "/home/divya/external_dependencies/miniconda3/envs/truera-c492f339dfe9d66881580e3ea4cc559e6dae9d7e/lib/python3.9/site-packages/numba/__init__.py", line 140, in _ensure_critical_deps
    raise ImportError("Numba needs NumPy 1.22 or less")

Can we specify a max numpy version in install_requires of numba’s setup.py file? This way if numba is specified in a package list, it will force an installation of numpy that is compatible with numba.

Obviously long term is it best to make sure numba is compatible with the latest numpy, but it would be great to pin a max numpy version during install rather than throwing an error at runtime.

[esc]

@divya-gopinath thank you for asking about this. In general we tend to pin NumPy on release branches only. Looking at the most recent release branch release_0.55, it seems like this is already setup via install_requires(?):

https://github.com/numba/numba/blob/release0.55/setup.py#L361 https://github.com/numba/numba/blob/release0.55/setup.py#L403

So, I am not sure how you ended up in that situation, but the runtime checks are in there in case something does go wrong during installation.

How did you install NumPy and Numba?

[esc]

Although, perhaps I should add, that if you use the advertised pip install numba it will match an appropriate NumPy version, (I presume this is install_requires working):

 💣 zsh» pip install numba
Collecting numba
  Using cached numba-0.55.2-cp310-cp310-macosx_10_14_x86_64.whl (2.3 MB)
Collecting numpy<1.23,>=1.18
  Using cached numpy-1.22.4-cp310-cp310-macosx_10_15_x86_64.whl (17.7 MB)
Collecting llvmlite<0.39,>=0.38.0rc1
  Using cached llvmlite-0.38.1-cp310-cp310-macosx_10_9_x86_64.whl (25.5 MB)
Requirement already satisfied: setuptools in ./miniconda3/envs/issue_8188/lib/python3.10/site-packages (from numba) (61.2.0)
Installing collected packages: numpy, llvmlite, numba
Successfully installed llvmlite-0.38.1 numba-0.55.2 numpy-1.22.4
pip install numba  5.95s user 1.33s system 91% cpu 7.963 total