pydata-sphinx-theme 0.8.1 requires sphinx<5,>=3.5.4

[jarrodmillman]

I would like to test Sphinx==5.0.0.b1:

• https://pypi.org/project/Sphinx/5.0.0b1/

But I noticed that the most recent version of pydata-sphinx-theme pins on sphinx<5. Of course, I can just ignore that and test anyway.

According to the 5.0.0 release plan, Sphinx 5 should be released in a few weeks:

• 5.0.0 release plan sphinx-doc/sphinx#10312

According to your CI, this project works just fine with Sphinx 5:

• https://github.com/pydata/pydata-sphinx-theme/runs/6388923592

Would it be possible to to make a pydata-sphinx-theme 0.8.2 with the upperbound unpinned (or at least incremented so folks don't see the warning)? Personally, I would remove the upper bound entirely.

[jarrodmillman]

If you need to keep the upper bound for some reason that I am not aware of, maybe you could atleast make a prerelease that doesn't have the upper bound. That would make the CI configuration more consistent for numpydoc.

Or maybe there is a pip flag that I can use to ignore the upper bound...

[damianavila]

Related issues and PRs:

• Support / test older versions of sphinx?  #446
• Pin sphinx to >=3.5.4,<5 #595

[damianavila]

Or maybe there is a pip flag that I can use to ignore the upper bound...

I think you may use --no-deps and manage the dependencies by yourself... ugly, but it may help, maybe?

[jarrodmillman]

I think you may use --no-deps and manage the dependencies by yourself... ugly, but it may help, maybe?

I recommend against using the --no-deps flag in this case. It just creates extra work.

Here is what I would do assuming you decide not to "officially" support sphinx 5:

pip install pydata-sphinx-theme
pip install sphinx==5.0.0.b1

Here is what you will need to do (if you use the --no-deps flag) and want to achieve the same thing as above:

pip install --no-deps pydata-sphinx-theme
pip install beautifulsoup4
pip install sphinx==5.0.0.b1

Is there a reason you can't support sphinx 5? In my testing everything seems to work fine. It is also passing your CI tests.

[jarrodmillman]

Can we at least make the pypa package consistent with the requirements on conda-forge?

https://github.com/conda-forge/pydata-sphinx-theme-feedstock/blob/main/recipe/meta.yaml

<...>
    - sphinx
<...>
    - sphinx >=4
<...>

[damianavila]

For future readers, the upper bound was removed here: #661