[m-messer]

Problem

Some criteria gave learners an error instead of feedback. These bugs came up during the context refactor (bug report). This is the first PR of the bug-fix stack, and it is stacked on #283.

• Bugs 1 + 12: matches criteria whose sides weren't exactly response/answer crashed or applied the tolerance to the wrong values. Sets and None standard values also crashed _within_tolerance.
• Bug 10: comparisons involving complex infinity (e.g. 1/0) sometimes crashed. Every comparison operator was evaluated, not only the one requested, so whether it crashed depended on hash randomisation.
• Bug 14: order criteria such as response > 1.5 m raised TypeError (14a). A quantity written out in the criterion was not parsed with units (14b).
• Bug 16: response matches 2 m was always false. 14b fixes this.
• Bug 17: order criteria crashed for quantities without units ('Rational' * 'NoneType'). 14b fixes this.

Changes

• _within_tolerance (app/context/physical_quantity/graphs/match.py): returns False unless a tolerance is set, the criterion is response/answer, and both standard values are single sympy expressions.
• do_comparison_infinite (app/context/symbolic/comparison.py): looks up only the requested operator in COMPARISON_OPERATORS and evaluates it inside the try.
• comparison_function (app/context/physical_quantity/comparison.py): returns None (unknown) when sympy cannot decide the truth value, instead of raising TypeError.
• CriterionSides (app/context/physical_quantity/comparison.py): parse_criterion_sides now returns a frozen dataclass. A side written out as a quantity (INPUT label) is parsed with the quantity parser and stands in the expression as a criterion_lhs/criterion_rhs symbol. Its ParsedQuantity goes into sides.quantities next to the response and answer.
  • match.py (_quantity_match, _dimension_match, _unit_comparison) and order.py (comparison_base_graph) substitute from sides.quantities.
  • order.py compares a quantity without a unit by its value alone.
  • parsed_quantity is split out of parse_quantity in app/context/physical_quantity/parsing.py.

Checklist

• Tests added/updated
• Docs updated (app/docs.md) if user-facing behaviour changed: n/a, these are bug fixes to behaviour that is already documented

Testing

• New regression tests:
  • test_matches_criteria_other_than_response_and_answer_when_values_differ, test_order_criteria_against_written_out_quantity and test_matches_criteria_against_written_out_quantity in physical_quantity_evaluation_test.py.
  • test_complex_infinity and test_do_comparison_infinite_only_evaluates_requested_comparison in symbolic_evaluation_test.py. The direct test is there because the end-to-end crash only happened on some runs.
• Each regression test failed before its fix.
• pytest app: 1241 passed, 2 skipped, 1 failed. The failure is test_radians_to_frequency, which already failed before this PR.
• mypy app: no issues.

Still open in the bug report, for the next PRs in the stack: 2, 3, 4, 6, 7, 8, 9, 11, 13, 15, and the feedback strings.

This is the bottom of the stack and merges into main. These lines close issues fixed higher up the stack:

• Closes Feature: Support Sig Fig #270 (significant figures, Add significant figures and decimal places parameters #300)
• Closes Feature: Inequalities #271 (inequalities, Support inequalities in the answer and response #301)

🤖 Generated with Claude Code