Incompatibility between Dask 2023.1.1 and Pandas 2.03

[SWeav02]

I'm working to merge the warrenapp into Simmate and I've been running into an error when running simmate workflows explore. The error I get seems to be caused by the following import:

import dask.dataframe as dd

and throws the following error

AttributeError: module 'pandas.core.strings' has no attribute 'StringMethods'

There is a stack overflow discussion on this error (here) which seems to indicate that there's an incompatibility between some versions of dask and pandas. Checking my conda environment, I noticed that my dev install of simmate defaulted to dask 2023.1.1 and pandas 2.0.3. I'm not to familiar with how conda handles dependencies so I'm not sure what's making dask stay on this older version since Simmate seems to have the dependency set up to dask 2023.6.1. I used pip to force an update to dask 2023.6.1 and calling the command works fine now, but there should be some way to make sure a conda install of Simmate will function without having to do this. Maybe increase the dependency range in the pyproject.toml to something higher?

Any suggestions?

Full error is in the attached txt file.
dask pandas error.txt

[jacksund]

import dask.dataframe as dd

^^ where is this line? I'm actually not a big fan of dask.dataframe + import abbreviations so that doesn't look like a line I would write. Maybe it's a bug from another package like django_pandas...?

But pandas went through major api changes (v2.0) this year, so that probably caused this everywhere. Sounds like we need to just bump the minimum dask version in our CI. I can do that in #524

[SWeav02]

Sorry, I should have mentioned where this actually came from. This is from my BadELF code since I use dask.dataframe to parallelize part of the algorithm. Located here in the warrenapp repo. It makes sense that you haven't run into it yet if you're not a fan of dask.dataframe since that seems to be the only thing causing the error. Importing other dask modules doesn't seem to cause problems.

Increasing the minimum dask version sounds good to me. Thanks!