Add support for inhomogenous error breakdowns

[20DM]

Proposal solution for the feature discussed in #229 :

The HepData validator already allows for cases where different bins have different error breakdowns, but hepdata_lib currently does not support this.

With this change set one can suppress individual sources by setting them to None.

Closes #266.

---

📚 Documentation preview 📚: https://hepdata-lib--265.org.readthedocs.build/en/265/

[codecov]

Codecov Report

All modified and coverable lines are covered by tests ✅

Project coverage is 90.42%. Comparing base (dcc23bd) to head (2b8c075).
Report is 7 commits behind head on main.

Additional details and impacted files

@@            Coverage Diff             @@
##             main     #265      +/-   ##
==========================================
+ Coverage   90.37%   90.42%   +0.04%     
==========================================
  Files           5        5              
  Lines        1112     1117       +5     
  Branches      251      254       +3     
==========================================
+ Hits         1005     1010       +5     
  Misses         78       78              
  Partials       29       29              

Flag	Coverage Δ
unittests-3.10	90.42% <100.00%> (+0.04%)	⬆️
unittests-3.11	90.42% <100.00%> (+0.04%)	⬆️
unittests-3.6	90.14% <100.00%> (+0.04%)	⬆️
unittests-3.7	90.14% <100.00%> (+0.04%)	⬆️
unittests-3.8	90.24% <100.00%> (+0.04%)	⬆️
unittests-3.9	90.24% <100.00%> (+0.04%)	⬆️

Flags with carried forward coverage won't be shown. Click here to find out more.

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

[GraemeWatt]

Thanks for the PR! Can you please add a test in tests/test_uncertainty.py to cover the new feature and add some explanation under Uncertainties in the documentation?

[GraemeWatt]

Thanks for adding the test and expanding the docs. The CI is failing due to issues with pylint. Can you please modify the code so that the pylint checks pass? See Analysing the code in the developer docs.

[GraemeWatt]

Thanks for fixing the linting for the tests. However, the CI still gives a message:

************* Module hepdata_lib
hepdata_lib/__init__.py:237:4: R0912: Too many branches (13/12) (too-many-branches)

-----------------------------------
Your code has been rated at 9.99/10

Could you please fix this too? If you don't see an easy way to refactor the make_dict function to avoid this warning, it can be suppressed by adding a line # pylint: disable=too-many-branches at the top of the function, but this should be avoided if possible. I think it's a current bug in the CI that the "Run pylint" still passes, so I've opened a separate PR #267 to address it.

[GraemeWatt]

PR #267 has been merged now, so if you update your branch with the last commit to main (dcc23bd), the CI should now fail at the "Run pylint on hepdata_lib" step.

[20DM]

Thanks for fixing the linting for the tests. However, the CI still gives a message:

************* Module hepdata_lib
hepdata_lib/__init__.py:237:4: R0912: Too many branches (13/12) (too-many-branches)

-----------------------------------
Your code has been rated at 9.99/10

Could you please fix this too? If you don't see an easy way to refactor the make_dict function to avoid this warning, it can be suppressed by adding a line # pylint: disable=too-many-branches at the top of the function, but this should be avoided if possible. I think it's a current bug in the CI that the "Run pylint" still passes, so I've opened a separate PR #267 to address it.

Urgh - well, this can be circumvented by moving the additional if statement into a separate list comprehensions, e.g. like so

for unc in [ unc for unc in self.uncertainties if not unc.values[i] is None ]:
    ...

but that adds an extra loop for no good reason. So my preference would be to silence the linter here, but I'd leave that call to you. 🤷

[GraemeWatt]

I've added # pylint: disable=too-many-branches at the top of the make_dict function, added NoneType in the docstring of the sanitize_value function, and merged the last commit to the main branch. The workflow run for the separate "Run pylint on hepdata_lib" step now fails before adding the # pylint: disable=too-many-branches line, as expected. I'll let @clelange perform the final review and merge.

[clelange]

Thanks for the contribution @20DM and for the review @GraemeWatt !