Friendly import error message for dask-expr

[benrutter]

Dask dataframe now uses dask-expr by default which needs to be installed seperately.

I think it could potentially trip new users up a little bit if they install the latest dask since they'll hit a message asking them to python -m pip install "dask[dataframe]", but doing this won't resolve the error (assuming they are using dask expressions which is the default).

This PR expands the error message to the following:

Dask dataframe requirements are not installed.

Please either conda or pip install as follows:

    conda install dask                     # either conda install
    python -m pip install "dask[dataframe]" --upgrade  # or python -m pip install

Dask also now uses dask-expr by default, which requires installing seperately:

    conda install dask-expr
    python -m pip install "dask-expr" --upgrade

I haven't created a github issue for this (but can do)

• Closes #xxxx
• Tests added / passed
• Passes pre-commit run --all-files

[phofl]

dask-expr is a dask[dataframe], is this not getting pulled in automatically?

[benrutter]

No I think this is intentional behavour? (dask.dataframe doesn't pull dask-expr, rather than the other way round)

Looking at this it's possible that it's only if you're updating dask rather than installing fresh? I'll check!

[phofl]

It's pulled for me if I do

pip uninstall dask-expr
pip install "dask[dataframe]" --update

[benrutter]

Yes, @phofl you're right (my bad)! Just pip installing dask completely fresh into an environment installs dask-expr with it.

I still think a mention in the error message is helpful, since depending on workflow tools, commands like pipenv update, pip update dask or rye sync --update-all.

The downside is it'd need updating again in some future version when this isn't a likely risk though (I think a bunch of people are likely to update dask to the latest version, but there's a point where nobodies bumping versions by more than a year or so and not expecting to hit some kind of issue, especially since there's no semvar in dask)

Edit: looking at your comment again, I'm wondering if it's more obscure that I assumed? I hit this issue running "update all" with Rye, but that's probably not the most common workflow? (I'd guess similar things would happen with workflow tools like poetry or hatch but I haven't put that to the test)

[phofl]

I don't want to add a reference to dask-expr to the message for now if we can avoid it. dask will consume dask-expr at some point where this won't be necessary anymore and probably mess with users environments