Testing optional packages via ci.yml

[soehms]

In #40105, an attempt was made to test the optional packages snappy, regina, and database_knotinfo using the GitHub workflow ci.yml. Unfortunately, this was not yet successful.

As per #40105 (comment), we are separating this issue from #40105 here.

This pull request is related to #42102, as it also enables testing of optional packages via CI. However, the goal here is different: to achieve better sustainability (see #40448 (comment)). On the other hand, the approach presented here may not be suitable for all optional packages.

📝 Checklist

• The title is concise and informative.
• The description explains in detail what this PR is about.
• I have linked a relevant issue or discussion.
• I have created tests covering the changes.
• I have updated the documentation and checked the documentation preview.

⌛ Dependencies

[github-actions]

Documentation preview for this PR (built with commit edcf60d; changes) is ready! 🎉
This preview will update shortly after each push to this PR.

[soehms]

@tobiasdiez, do you know what causes this linter error?

[project.optional-dependencies]
@@ -7,0 +7,1 @@
+R = ['rpy2 >=3.3']
@@ -7,10 +8,10 @@
database_knotinfo = ['database_knotinfo']
extra = [
  'igraph',
  'lrcalc ~=2.1; sys_platform != "win32"',
  #'p_group_cohomology >=3.3', # Only used in tests, not available on PyPI
  'pycosat >=0.6.3',
  'pynormaliz >=2.18; platform_machine != "aarch64" and platform_machine != "arm64"', # Not yet available for Linux aarch64
  'sage_numerical_backends_coin',
  'symengine >= 0.6.1', # Only used in tests
]
@@ -17,1 +18,0 @@
-R = ['rpy2 >=3.3']
@@ -18,7 +18,7 @@
ERROR operation failed error=some files were not properly formatted
regina = ['regina']
snappy = ['snappy']
#giac = ['sagemath_giac'] # Not yet available on PyPI (https://github.com/sagemath/sagemath-giac/issues/3)

There are other similar lines in pyproject.toml that resemble the ones I inserted. uv was also able to read them. I've already made a change as suggested by Copilote:

2. Sorted the entries alphabetically for consistency: database_knotinfo, extra, R, regina, snappy
3. Moved the commented giac entry to the end to keep related entries together

But this didn't help (as expected). Any ideas?

[tobiasdiez]

If I understand the diff correctly, it wants to move the R = ... line to be the first in the block.

[soehms]

@tobiasdiez, if I follow your suggestion at #40105 (comment), I receive the following message:

error: Default group `snappy` (from `tool.uv.default-groups`) is not defined in the project's `dependency-groups` table

Consequently, I followed the suggestions of an AI to use the uv sync options --all-extras or (more specifically) --extra snappy --extra regina --extra database-knotinfo in the workflow.

I ran the corresponding jobs in my fork repository. The first attempt failed during the build at sage_numerical_backends_coin, and the second failed during one of the initial tests (inconsistently across the entire matrix; for example, at src/doc/de/a_tour_of_sage/index.rst for Fedora 41).

Note, however, that even without my change to ci.yml, not many tests were successful—as you can see from the unchanged job for MacOS 15.

[tobiasdiez]

Extra is of course the correct flag.

Those CI runs are sadly not in a very good shape when it comes to the tests. Hopefully the situation looks better after #42276 and #42229.