malariagen / malariagen-data-python

Analyse MalariaGEN data from Python
https://malariagen.github.io/malariagen-data-python/latest/
MIT License
13 stars 23 forks source link

Install failing on Colab (package incompatibility) #586

Closed ahernank closed 2 weeks ago

ahernank commented 2 weeks ago

When installing via pip in Google Colab, there seem to be some issues around solving dependencies using the pyproject.toml; the solver seems to get stuck trying to install versions of numpy (2.0.2) & pandas (1.5.3) that are incompatible with current versions of (external) packages on Colab. These issues seem to relate mostly (but not exclusively) to the packages upgrades on the latest Colab release.

This should be sorted by pinning compatible versions on our requirements but I don't fully understand how these incompatibilities are only showing now (and why the solver is trying to use numpy==2.0.2 to resolve them). Wondered if it could be due to the pip/setuptools upgrades on Colab, but this didn't seem to be the case. It seems to be due to numpy itself, as regular np installs with the latest versions are failing too.

Screenshot 2024-08-28 at 06 53 49
ahernank commented 2 weeks ago

This issue can also be seen on local installations. So far, I have only managed to make it work using Python3.9 and downgrading numpy (i.e., python3.9 -m pip install malariagen_data 'numpy==1.26.4').

Other versions are having issues around cython compiling with this solution (from statsmodels<=0.13.0).

JaneNjeri commented 2 weeks ago

Hi Anastasia, I also got a conflict with the version of pandas on my end with local installation. Error: "cudf-cu12 24.4.1 requires pandas<2.2.2dev0,>=2.0, but you have pandas 1.5.3 which is incompatible"

Would it be an option to have exact versions in the pyproject.toml?

ahernank commented 2 weeks ago

Current working dependencies setup on Colab:

malariagen_data==12.0.0
numpy<2.0.0
statsmodels>0.13.0
xarray>2023.4.0
numba>0.22.0
dask>0.8.2
alimanfoo commented 2 weeks ago

@ahernank has just released v12.0.1 which resolves this issue. I just checked on colab and confirm install is working fine.

Suggest we close here and raise a follow-up issue to add support for numpy version 2.x so we are ready for it when colab decides to upgrade (#588).

Also suggest we drop support for Python 3.9 to align with numpy etc (#589).