feat(api): a result hands back how its solve terminated as one Record

[FBumann]

Prompt: In the records change, does each Result and Runs also have a .record property? … Add it

Note

The following content was generated by AI.

result.record returns the Record that save() writes to record.parquet: one row of the table a sweep keeps as sweep.record, under the same name one dimension wider. Before this, Result exposed only the separate fields, and the Record value existed only inside save(). save() now builds its row from the property, so the row has one home.

Details, mutation table and verification

Behaviour

• result.record computes model_digest() on first use, as save() already did. The digest is cached, so later reads are free.
• An answer read back with load_result or load_archive now keeps the run its record was stamped with. Before, spec_digest, solved_at and model_digest were carried over, but run was dropped. So load_archive(...).answer.record equals the row on disk.
• save() still writes run as null. A plain save is not an archive, and Record documents run as "stamped when the archive is written and null until then".

Tests

• test_a_result_hands_back_the_row_its_save_writes, parametrized over LP, MIP and INFEASIBLE. The property, the saved file and a load_result of that file must all give one row, and objective must be null exactly where nothing was reached.
• test_an_archive_stamps_its_own_name_and_when_the_solve_returned also checks that an archived answer carries its run, and that re-saving it plainly writes run as null.

Mutation table (committed tree, restored with git checkout --, __pycache__ dropped on both sides)

guard	result
load_result carries run (api.py:673, via tools/mutate.py)	caught
save() writes run as null (result.py:791: ._replace(run=None) removed, by hand)	caught

Gates: ruff check, ruff format --check, pyrefly check, the full suite (4033 passed, 319 skipped, 1 xfailed) and bench/test_harness.py (151 passed), in a uv environment. After #1730–#1734 merged, GitHub rebased this branch onto main; the resulting tree is identical to the one those gates ran on.

🤖 Generated with Claude Code

https://claude.ai/code/session_01VgkRT9Kpb2iNvL6dpV8VER

[read-the-docs-community]

Documentation build overview

📚 specsolve | 🛠️ Build #34752666 | 📁 Comparing 01fbfc2 against latest (50c976e)

  🔍 Preview build  

8 files changed · ± 8 modified

± Modified

• about/architecture/index.html
• about/linopy/index.html
• howto/archiving/index.html
• howto/parallel/index.html
• howto/warehouse/index.html
• reference/api/index.html
• reference/glossary/index.html
• reference/sweeps/index.html

[read-the-docs-community]

Documentation build overview

📚 lpspec | 🛠️ Build #34752665 | 📁 Comparing 01fbfc2 against latest (50c976e)

  🔍 Preview build  

8 files changed · ± 8 modified

± Modified

• about/architecture/index.html
• about/linopy/index.html
• howto/archiving/index.html
• howto/parallel/index.html
• howto/warehouse/index.html
• reference/api/index.html
• reference/glossary/index.html
• reference/sweeps/index.html

[codspeed]

Merging this PR will not alter performance

🆕 24 new benchmarks
⏩ 82 skipped benchmarks¹

Performance Changes

	Benchmark	BASE	HEAD	Efficiency
🆕	test_emit[commitment-s-specsolve-highs]	N/A	2 MB	N/A
🆕	test_emit[commitment-s-specsolve-lp]	N/A	2.6 MB	N/A
🆕	test_emit[dispatch-s-specsolve-highs]	N/A	12.4 MB	N/A
🆕	test_emit[dispatch-s-specsolve-lp]	N/A	18.1 MB	N/A
🆕	test_emit[fleet-s-specsolve-highs]	N/A	21.9 MB	N/A
🆕	test_emit[fleet-s-specsolve-lp]	N/A	26.2 MB	N/A
🆕	test_emit[nodal-s-specsolve-highs]	N/A	4.9 MB	N/A
🆕	test_emit[nodal-s-specsolve-lp]	N/A	5.9 MB	N/A
🆕	test_emit[profiled-s-specsolve-highs]	N/A	21.1 MB	N/A
🆕	test_emit[profiled-s-specsolve-lp]	N/A	24.3 MB	N/A
🆕	test_emit[sector-s-specsolve-highs]	N/A	5.3 MB	N/A
🆕	test_emit[sector-s-specsolve-lp]	N/A	5.3 MB	N/A
🆕	test_emit[storage-s-specsolve-highs]	N/A	16.9 MB	N/A
🆕	test_emit[storage-s-specsolve-lp]	N/A	21.7 MB	N/A
🆕	test_emit[transport-s-specsolve-highs]	N/A	15.7 MB	N/A
🆕	test_emit[transport-s-specsolve-lp]	N/A	19.9 MB	N/A
🆕	test_window[commitment-s-specsolve-highs]	N/A	1.6 MB	N/A
🆕	test_window[dispatch-s-specsolve-highs]	N/A	11.3 MB	N/A
🆕	test_window[fleet-s-specsolve-highs]	N/A	17.4 MB	N/A
🆕	test_window[nodal-s-specsolve-highs]	N/A	4.7 MB	N/A
...	...	...	...	...

ℹ️ Only the first 20 benchmarks are displayed. Go to the app to view all benchmarks.

---

_{Comparing claude/result-record (01fbfc2) with main (85ebcd2)²}

Footnotes

1. 82 benchmarks were skipped, so the baseline results were used instead. If they were deleted from the codebase, click here and archive them to remove them from the performance reports. ↩

2. No successful run was found on main (8c783ca) during the generation of this report, so 85ebcd2 was used instead as the comparison base. There might be some changes unrelated to this pull request in this report. ↩