Reject scalar acceptance input rows

[goktugozkanmd]

This keeps AcceptanceInput from accepting a bare string row such as gate_results=("ok",) or { "gate_results": ["ok"] } as ("o", "k").

Why it matters:

• gate_results and contradictions are record-shaped fields, so scalar strings should fail at the boundary instead of being split into characters.
• from_dict() now leaves row-shape validation to the constructor, keeping the direct and serialized paths consistent.

Boundary and compatibility:

• Stable public API names are unchanged.
• Serialized AcceptanceInput output shape is unchanged.
• Well-formed list/tuple rows still round-trip as before.
• The only behavior change is rejecting malformed scalar row inputs that were previously accepted by accidental character splitting.
• Trust-boundary impact: malformed host observations fail at construction instead of reaching policy evaluation as synthetic two-character rows.

Checklist:

• Problem and intended behavior stated.
• Kernel boundary change only; no host/runtime adapter expansion.
• Public API and serialized-record compatibility checked.
• Failure mode covered by regression tests.
• Local validation evidence listed below.

Evidence:

• uv run --frozen --extra dev python -m pytest tests/test_kernel.py -q
• uv run --frozen --extra dev python -m pytest -q
• uv run --frozen --extra dev python -m ruff check src tests scripts
• uv run --frozen --extra dev python -m ruff format --check src tests scripts
• python3 -m compileall -q src tests scripts

Summary by CodeRabbit

• Bug Fixes

  • Improved validation for acceptance data entries.
  • Invalid scalar, string, byte, and non-iterable row values now produce clearer validation errors.
  • Consistent validation is applied whether data is constructed directly or loaded from a dictionary.

• Tests

  • Added regression coverage for invalid gate result and contradiction entries.

[coderabbitai]

No actionable comments were generated in the recent review. 🎉

ℹ️ Recent review info

⚙️ Run configuration

Configuration used: defaults

Review profile: CHILL

Plan: Pro Plus

Run ID: 8b898ea1-1f07-49c0-846c-e5e30854542e

📥 Commits

Reviewing files that changed from the base of the PR and between 7f3eae4 and e575701.

📒 Files selected for processing (2)

• src/evalopt_graph/kernel.py
• tests/test_kernel.py

🚧 Files skipped from review as they are similar to previous changes (1)

• src/evalopt_graph/kernel.py

---

📝 Walkthrough

Walkthrough

AcceptanceInput now normalizes and validates row-like fields, rejects string and byte rows, and reports conversion or shape errors as ValueError. Dictionary construction delegates normalization to this logic, with regression tests covering direct and dictionary-based inputs.

Changes

Acceptance input validation

Layer / File(s)	Summary
Row normalization and validation src/evalopt_graph/kernel.py	The row normalizer rejects str and bytes, converts rows to tuples, and validates row widths and string elements.
Construction wiring and regression coverage src/evalopt_graph/kernel.py, tests/test_kernel.py	from_dict passes raw row structures to the constructor, and tests verify scalar and non-iterable entries raise the expected ValueError through the supported construction paths.

Estimated code review effort: 2 (Simple) | ~10 minutes

Suggested reviewers: ajaysurya1221

🚥 Pre-merge checks | ✅ 4 | ❌ 1

❌ Failed checks (1 warning)

Check name	Status	Explanation	Resolution
Docstring Coverage	⚠️ Warning	Docstring coverage is 0.00% which is insufficient. The required threshold is 80.00%.	Write docstrings for the functions missing them to satisfy the coverage threshold.

✅ Passed checks (4 passed)

Check name	Status	Explanation
Title check	✅ Passed	The title clearly summarizes the main change: rejecting scalar acceptance input rows.
Description check	✅ Passed	The description covers the problem, boundary/compatibility impact, testing evidence, and a checklist, with only minor template formatting differences.
Linked Issues check	✅ Passed	Check skipped because no linked issues were found for this pull request.
Out of Scope Changes check	✅ Passed	Check skipped because no linked issues were found for this pull request.

✨ Finishing Touches

🧪 Generate unit tests (beta)

• Create PR with unit tests

---

Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out.

❤️ Share

• X
• Mastodon
• Reddit
• LinkedIn

_{Comment @coderabbitai help to get the list of available commands.}

[coderabbitai]

Actionable comments posted: 1

🧹 Nitpick comments (1)

tests/test_kernel.py (1)

248-266: 📐 Maintainability & Code Quality | 🔵 Trivial | ⚡ Quick win

Cover contradictions and byte-string rows too.

This regression test exercises only gate_results with str values. Since the implementation changes the shared normalizer for both gate_results and contradictions, add direct and from_dict cases for contradictions, plus bytes cases, preferably via parametrization.

🤖 Prompt for AI Agents

Verify each finding against current code. Fix only still-valid issues, skip the
rest with a brief reason, keep changes minimal, and validate.

In `@tests/test_kernel.py` around lines 248 - 266, Add parametrized regression
coverage in
test_gate_result_rows_reject_bare_strings_instead_of_splitting_characters for
both gate_results and contradictions, covering direct AcceptanceInput
construction and AcceptanceInput.from_dict inputs with str and bytes rows.
Preserve the existing ValueError assertion and message match while reusing the
parametrized field/value cases.

🤖 Prompt for all review comments with AI agents

Verify each finding against current code. Fix only still-valid issues, skip the
rest with a brief reason, keep changes minimal, and validate.

Inline comments:
In `@src/evalopt_graph/kernel.py`:
- Around line 160-172: Update the validation block that normalizes values so
top-level iteration failures from inputs such as None or integers are caught and
re-raised as the existing ValueError message. Wrap the loop over values,
including iterator acquisition, while preserving the current row-level
normalization and validation behavior.

---

Nitpick comments:
In `@tests/test_kernel.py`:
- Around line 248-266: Add parametrized regression coverage in
test_gate_result_rows_reject_bare_strings_instead_of_splitting_characters for
both gate_results and contradictions, covering direct AcceptanceInput
construction and AcceptanceInput.from_dict inputs with str and bytes rows.
Preserve the existing ValueError assertion and message match while reusing the
parametrized field/value cases.

🪄 Autofix (Beta)

Fix all unresolved CodeRabbit comments on this PR:

• Push a commit to this branch (recommended)
• Create a new PR with the fixes

---

ℹ️ Review info

⚙️ Run configuration

Configuration used: defaults

Review profile: CHILL

Plan: Pro Plus

Run ID: cbf25d7a-8fe1-4744-b793-cbdad78fd113

📥 Commits

Reviewing files that changed from the base of the PR and between cdf51d3 and 7f3eae4.

📒 Files selected for processing (2)

• src/evalopt_graph/kernel.py
• tests/test_kernel.py

[goktugozkanmd]

Addressed the CodeRabbit notes in e5757016ebf18423451fb12ac7868a91842fa66d.

What changed:

• top-level non-iterable row containers such as None and integers now raise the existing ValueError shape message
• regression coverage now includes gate_results and contradictions, with both str and bytes rows, through direct construction and from_dict

Local validation on the pushed branch:

• uv run --frozen --extra dev python -m pytest -q
• uv run --frozen --extra dev python -m ruff check src tests scripts
• uv run --frozen --extra dev python -m ruff format --check src tests scripts
• python3 -m compileall -q src tests scripts