SPEC 0: Set minimum supported versions to pandas>=1.5 and xarray>=2022.03

[seisman]

Description of proposed changes

• requirements.txt
• environment.yml
• pyproject.toml
• .github/workflows/ci_tests.yaml
• doc/minversions.rst https://pygmt-dev--3043.org.readthedocs.build/en/3043/minversions.html

Closes #2863.

[seisman]

In #3037 (comment), @weiji14 proposed to add two more columns for pandas and xarray. I think this is not good, because the table is already too wide. I prefer to list the core dependencies in a single column. See https://pygmt-dev--3043.org.readthedocs.build/en/3043/minversions.html for preview.

I also prefer to combine the "Python version" and "Documentation" columns into a single column. The new table will look like:

Thoughts?

[weiji14]

I also prefer to combine the "Python version" and "Documentation" columns into a single column.

Yep, ok with this part. We can add a note at the bottom to say that 'Dev' reflects the main branch and is for the upcoming release.

In #3037 (comment), @weiji14 proposed to add two more columns for pandas and xarray. I think this is not good, because the table is already too wide. I prefer to list the core dependencies in a single column.

Once we combine the PyGMT and Docs columns, maybe there's enough space for separate numpy/pandas/xarray columns? Could you try splitting into individual columns to see how it looks?

[seisman]

Done in 5f22ecb. Not bad. See preview at https://pygmt-dev--3043.org.readthedocs.build/en/3043/minversions.html.

[weiji14]

I also prefer to combine the "Python version" and "Documentation" columns into a single column.

Yep, ok with this part. We can add a note at the bottom to say that 'Dev' reflects the main branch and is for the upcoming release.

In #3037 (comment), @weiji14 proposed to add two more columns for pandas and xarray. I think this is not good, because the table is already too wide. I prefer to list the core dependencies in a single column.

Once we combine the PyGMT and Docs columns, maybe there's enough space for separate numpy/pandas/xarray columns? Could you try splitting into individual columns to see how it looks?

[weiji14]

Great, thanks @seisman!

[seisman]

After pinning pandas and xarray with Python 3.10, now we have many failures on macOS/Windows. See https://github.com/GenericMappingTools/pygmt/actions/runs/7946102470/job/21694725831. Need to find out why.

[weiji14]

After pinning pandas and xarray with Python 3.10, now we have many failures on macOS/Windows. See https://github.com/GenericMappingTools/pygmt/actions/runs/7946102470/job/21694725831. Need to find out why.

pytest-mpl=0.17.0 now adds a metadata key into our savefig call (added in https://github.com/matplotlib/pytest-mpl/pull/196/files I think), and this is not handled by psconvert. We'll need to ignore that metadata key somehow.

[seisman]

After pinning pandas and xarray with Python 3.10, now we have many failures on macOS/Windows. See https://github.com/GenericMappingTools/pygmt/actions/runs/7946102470/job/21694725831. Need to find out why.

pytest-mpl=0.17.0 now adds a metadata key into our savefig call (added in https://github.com/matplotlib/pytest-mpl/pull/196/files I think), and this is not handled by psconvert. We'll need to ignore that metadata key somehow.

Thanks. Will address the issue after this PR is merged.