Fix GDPopt LOA maximization objective sense

[bernalde]

Fixes #3937.

Summary/Motivation:

GDPopt LOA currently creates the augmented OA master objective with minimization sense even when the original GDP objective is a maximization objective. For maximization GDPs, this can make the discrete OA master pursue the wrong objective direction.

This is a replacement for #3938, which GitHub would not allow me to reopen after it was closed. The prior review discussion and history remain available there.

This was observed while validating the GDPLib methanol benchmark:

• Track GDPopt LOA objective-sense discrepancy SECQUOIA/gdplib#114 tracks the downstream methanol LOA mismatch.
• Fix methanol GDPopt solve options SECQUOIA/gdplib#113 documents and preserves the downstream sign-convention workaround.

In the methanol reproducer, the algebraically equivalent formulations minimize(-profit) and maximize(profit) selected different disjuncts before this fix. With this branch, both formulations return the same profit and active disjuncts.

Changes proposed in this PR:

• Preserve the original objective sense when GDPopt LOA creates the augmented OA master objective.
• Keep the placeholder OA objective expressionless until the augmented expression is assigned.
• Add regression coverage for a small GDP where minimize(-x) and maximize(x) should select the same disjunct.
• Validation performed locally:
  • python -m pytest -q pyomo/contrib/gdpopt/tests/test_gdpopt.py::TestGDPoptUnit::test_LOA_maximize_matches_minimize_negated_objective
    • Result: 1 passed in 0.54s
  • python -m pytest -q pyomo/contrib/gdpopt/tests/test_gdpopt.py
    • Result: 59 passed, 21 skipped, 3 deselected in 16.93s
  • python -m black --check --diff pyomo/contrib/gdpopt/loa.py pyomo/contrib/gdpopt/tests/test_gdpopt.py
    • Result: passed, 2 files would be left unchanged

AI-Use Disclosure

• AI tools were NOT used during the preparation of this PR

or

• AI tools contributed to the development of this PR

  • AI tools generated documentation (including the PR description/comments, code comments, and/or Sphinx documentation)
  • AI tools generated tests (baselines, examples, and/or code)
  • AI tools generated code (apart from tests)

  Review process (select ONE):

  • Rewritten: All AI-generated content was rewritten by me before being committed.
  • Reviewed/verified: I retained AI-generated content and verified it before committing. Verification included (as applicable):
    • Ran the code and fixed issues
    • Added and ran tests
    • Checked correctness/logic of code and tests
    • Checked for alignment with the contribution guide
    • Considered security implications
  • As-is: AI-generated content was commited directly to the repository

Notes for reviewers (optional): This replacement PR carries over the implementation and tests from #3938, then refreshes the branch against current main. The replacement PR description and branch-refresh workflow were prepared with AI assistance and reviewed before posting. Reviewers may want to focus on the LOA objective-sense change and the regression coverage around equivalent minimize/maximize formulations.

Legal Acknowledgement

By contributing to this software project, I have read the contribution guide and agree to the following terms and conditions for my contribution:

1. I agree my contributions are submitted under the BSD license.
2. I represent I am authorized to make the contributions and grant the license. If my employer has rights to intellectual property that includes these contributions, I represent that I have received permission to make contributions and grant the required license on behalf of that employer.

[emma58]

This looks good to me. I would strongly prefer you don't call private methods in tests, though, so perhaps you could remove test_loa_augmented_penalty_objective_preserves_objective_sense and keep just test_LOA_maximize_matches_minimize_negated_objective (since the latter fully tests the change).

[emma58]

It would be ideal if the test didn't rely on calling a private method. The test below covers the change, so I think you could just remove this one, keep the one below, and dodge the problem.

[bernalde]

Addressed in a18df5d85: I removed the private-method test and its now-unused import, while keeping the end-to-end LOA regression. I also mutation-checked the remaining test: it passes with the fix and fails with the prior behavior (x = 1 instead of 9). The full GDPopt module and Black check pass.

[bernalde]

Addressed the remaining review feedback in a18df5d85 (Remove private-method GDPopt test).

• Removed test_loa_augmented_penalty_objective_preserves_objective_sense, which called the private _setup_augmented_penalty_objective method, and removed its now-unused minimize import.
• Retained the end-to-end test_LOA_maximize_matches_minimize_negated_objective regression. A mutation check confirmed it is non-vacuous: it passes on the fixed code and fails with the prior minimization-only objective behavior (x = 1 instead of 9).
• Updated the PR validation block to remove the deleted test and record the current results.

Verification:

• python -m pytest -q pyomo/contrib/gdpopt/tests/test_gdpopt.py::TestGDPoptUnit::test_LOA_maximize_matches_minimize_negated_objective — 1 passed in 0.54s
• python -m pytest -q pyomo/contrib/gdpopt/tests/test_gdpopt.py — 59 passed, 21 skipped, 3 deselected in 16.93s
• python -m black --check --diff pyomo/contrib/gdpopt/loa.py pyomo/contrib/gdpopt/tests/test_gdpopt.py — passed; 2 files unchanged

No review comments were intentionally left unaddressed. The PR remains approved. GitHub CI and Read the Docs are pending for the new head; the review thread was replied to but not resolved.