test(ITL-10): add dataclass → DataFrame conversion test coverage

[kurodo3]

Summary

• Adds 6 tests (2 new files) confirming that dataclass-output FunctionNodes produce correct Polars DataFrames
• Adds converter-level unit tests confirming plain Arrow structs round-trip safely through the converter as dynamic TypedDicts
• No production code changes — the core bug (FN1) was already fixed; this PR closes ITL-10 by adding the required test coverage

Staleness finding

The exact reproducer from ITL-10 already runs without error. as_table() now derives its schema from output_schema() directly (FN1 fix), so the original ValueError: Unsupported Python type: typing.Any crash no longer occurs. arrow_schema_to_python_schema correctly returns the original dataclass type (not Any) when the extension type wrapper is present. When a plain Arrow struct (no extension metadata) is passed to arrow_schema_to_python_schema, the converter creates a dynamic TypedDict that round-trips back to the original struct type.

Files added

• tests/test_core/nodes/test_function_node_dataclass.py — 3 end-to-end pipeline tests
• tests/test_extension_types/test_universal_converter_struct.py — 3 converter-level unit tests

Fixes ITL-10

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.

📢 Thoughts on this report? Let us know!

[Copilot]

Pull request overview

This PR adds regression test coverage for ITL-10, ensuring dataclass-returning FunctionNodes produce stable, correctly-typed Polars DataFrames via .as_df(), and that plain Arrow struct schemas are inferred as dynamic TypedDicts that round-trip back to the identical Arrow struct type through the universal converter.

Changes:

• Added end-to-end pipeline tests validating dataclass output → as_df() behavior for populated and unrun nodes (including schema stability).
• Added converter-level unit tests verifying plain Arrow struct → dynamic TypedDict inference, round-trip type mapping, and write-back table construction.
• Added internal design/spec and implementation plan documentation for the ITL-10 test work.

Reviewed changes

Copilot reviewed 4 out of 4 changed files in this pull request and generated 1 comment.

File	Description
tests/test_extension_types/test_universal_converter_struct.py	New unit tests for struct → dynamic TypedDict inference + round-trip/write-back behavior.
tests/test_core/nodes/test_function_node_dataclass.py	New end-to-end tests for dataclass-returning FunctionNode .as_df() output and schema stability.
superpowers/specs/2026-07-08-itl-10-dataclass-dataframe-conversion-tests-design.md	New design/spec documenting ITL-10 staleness assessment and success criteria for tests.
docs/metamorphic/plans/2026-07-08-itl-10-dataclass-dataframe-tests.md	New implementation plan documenting the steps to add/verify the new tests.

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

[kurodo3]

Fixed. The docstring now accurately states that get_default_context() returns a cached singleton and that the converter may be shared across tests. Added a note that the tests are written to be order-independent under that constraint.

[kurodo3]

Review response (round 1)

One comment addressed:

• Fixture docstring (): Removed the misleading word "Fresh". The docstring now accurately states that get_default_context() returns a cached singleton and that the converter instance may be shared across tests. Added a note that the tests are written to be order-independent under that constraint. (commit f58e35d)