openml-python
openml-python copied to clipboard
Added suite.metadata property and to_latex helper and tests (fixes #1126)
What does the PR fix?
This PR implements the OpenMLBenchmarkSuite.metadata property and the
suite.to_latex(...) helper method. This essentially fixes #1126
What this adds
-
suite.metadata
Returns a pandas DataFrame with:- dataset IDs,
- task IDs,
- dataset names,
- #instances (
n), #features (p), #classes (C)
-
suite.to_latex(filename, caption=None, label=None)
A lightweight, non-opinionated LaTeX export helper for the metadata.
It:- prefers pandas Styler formatting if jinja2 is installed,
- otherwise falls back to a minimal internal LaTeX generator,
- never imposes formatting rules/tricks on the user.
Why
Per discussion in #1126, maintainers preferred exposing the DataFrame (suite.metadata) instead of an opinionated Latex formatter. This PR:
- provides the canonical DataFrame for suites,
- includes a small, non-opinionated to_latex convenience helper,
- ships tests (mocked + server-tagged).
Notes
- The to_latex helper will prefer pandas' Styler when jinja2 is available; otherwise it uses a small fallback generator that does not require extra dependencies.
- The server integration test will skip if suite 99 is not present on the configured server.
- Additionally, if the fix is satisfactory, examples and a mention of the same in documentation can be done.
Metadata
- Reference Issue: Fixes #1126
- New Tests Added: Yes
- Documentation Updated: No