[codex] Consolidate benchmark support changes

[bernalde]

Summary

• Extract benchmark summary generation into gdplib.benchmark_summary and keep generate_benchmark_summary_all.py as a wrapper.
• Standardize GAMS benchmark optcr=1e-6, record it in benchmark metadata, and update the PR58 med-term solver routing rows.
• Add the optional Pixi gurobi environment metadata/docs and focused benchmark coverage from the open model PRs.

Validation

• python -m pytest tests/test_benchmark.py -v --tb=short
• python -m black -S -C --target-version py310 --check --diff benchmark.py gdplib/benchmark.py gdplib/benchmark_summary.py generate_benchmark_summary_all.py tests/test_benchmark.py
• flake8 gdplib/ --count --select=E9,F63,F7,F82 --show-source --statistics
• git diff --check

Notes

• Source PRs inspected: Fix gdp_col hull reformulation #124, Fix methanol GLOA benchmark summary generation #128, Fix hda hull and GLOA benchmark paths #130, Fix batch_processing GLOA bounds #132, Fix biofuel site module constraints #133, Bound batch processing cycle times #134, Fix Kaibel GDP reformulation failures #136, Fix med-term benchmark solver routing #137, Prune fixed mod_hens GDP choices #140, Fix modprodnet distributed pair activation #141. The consolidated changes come from the ancillary parts of Fix gdp_col hull reformulation #124, Fix methanol GLOA benchmark summary generation #128, Bound batch processing cycle times #134, and Fix med-term benchmark solver routing #137.
• I did not run Pixi install or solver-backed benchmarks to avoid heavy WSL/solver activity.

[bernalde]

GitHub would not accept REQUEST_CHANGES from the PR author account, so this is submitted as COMMENT. An eligible reviewer should treat the blocking item below as a request-changes review.

Blocking issues:

1. The packaged benchmark summary path now depends on pandas.to_markdown(), but tabulate is not declared as a runtime dependency. Fresh pip installs can finish benchmark solves and then fail during the default summary step.

Nonblocking issues:

1. The no-results summary guard still treats Path.glob() iterators as truthy, so campaigns with no JSON results can still raise during summary generation.

Questions: None.

Tests run and outcomes:

• /home/bernalde/.pixi/bin/pixi run pytest tests/test_benchmark.py -v --tb=short: passed, 20 tests.
• /home/bernalde/.pixi/bin/pixi run test: passed, 287 passed, 1 skipped.
• /home/bernalde/.pixi/bin/pixi run lint: passed; the second flake8 inventory reported existing style findings under --exit-zero.
• gh pr checks 144 --repo SECQUOIA/gdplib: all reported checks passed.
• Additional focused smoke with tabulate blocked: generate_benchmark_summary raises ImportError from pandas.to_markdown().

The PR should not be merged as-is. I would not merge this until the blocking issues above are addressed.

[bernalde]

Blocking: DataFrame.to_markdown() imports pandas' optional tabulate dependency, but tabulate is only in requirements-dev.txt/Pixi and is not declared in pyproject.toml, setup.py, or requirements.txt. Because gdplib-benchmark run now calls this packaged summary path by default, a normal pip installation can complete the benchmark solves and then fail while generating summaries. Add tabulate>=0.9.0 to the runtime dependency metadata, or make Markdown output conditional while still writing the CSV summaries when tabulate is absent. Please add a fresh-install or metadata regression test for whichever behavior is intended.

[bernalde]

Addressed in f8b4ff9 by adding tabulate>=0.9.0 to runtime dependency metadata (pyproject.toml, setup.py, requirements.txt) and covering it in CI/installation metadata tests. pixi.lock was updated for the new runtime metadata.

[bernalde]

Nonblocking: this guard does not actually check whether any JSON files exist. Path.glob() returns an iterator object, so the outer any() is truthy for any non-empty folder list and _generate_summary() still calls generate_benchmark_summary() with no frames after a campaign that produced no JSON results. Consider if not any(any(Path(folder).glob('*.json')) for folder in folders): return False and add a no-results regression test.

[bernalde]

Addressed in f8b4ff9 by checking for actual JSON files before summary generation and adding test_generate_summary_returns_false_without_json_results.

[bernalde]

Pushed commit:

• f8b4ff9 Address benchmark summary review feedback

Main changes:

• Added tabulate>=0.9.0 as a runtime dependency in pyproject.toml, setup.py, and requirements.txt; removed the duplicate dev-only requirement and updated the Pixi lock metadata.
• Updated CI and installation metadata checks so tabulate stays covered as a runtime dependency.
• Fixed _generate_summary() so it returns False when the selected run has no JSON result files.
• Added a no-results regression test for summary generation.

Tests run:

• /home/bernalde/.pixi/bin/pixi run pytest tests/test_benchmark.py::test_generate_summary_uses_packaged_summary_module tests/test_benchmark.py::test_generate_summary_returns_false_without_json_results tests/test_pip_installation.py::TestPipInstallation::test_runtime_dependencies_specified tests/test_pip_installation.py::TestPipInstallation::test_dependency_list_consistency tests/test_installation.py::TestInstallation::test_pip_installation_dependencies -v --tb=short: passed, 5 tests.
• /home/bernalde/.pixi/bin/pixi run test: passed, 288 passed, 1 skipped.
• /home/bernalde/.pixi/bin/pixi run lint: passed; the second flake8 inventory reported existing findings under --exit-zero.
• git diff --check: passed.

Comments intentionally not addressed:

• None. Both unresolved inline comments were addressed.

Remaining risks or follow-up:

• No solver-backed benchmark campaign was run; the changes are limited to dependency metadata and summary-generation control flow.