fix(parity): close P11 — SELECT alias on the LHS of an IN-subquery

[TimelordUK]

Summary

Closes P11: SELECT symbol, price * 2 AS dbl FROM trades WHERE dbl IN (SELECT price * 2 FROM trades WHERE price > 185) errored with Column 'dbl' not found. Parity board moves 76 → 77 AGREE, 6 → 5 GAP; the corpus case select_alias_in_in_subquery now AGREEs and the --check gate locks it.

Fixing it took two layers — the second was a pre-existing bug the alias case merely exposed.

1. R2 migration of WhereAliasExpander

The transformer was a hand-rolled recursion ending in a _ => (clone, false) catch-all (an R3 pattern), so it silently dropped InSubquery / NotInSubquery / the tuple forms — the same-scope LHS operand was skipped along with the subquery body.

Rewrote expand_expression to intercept only the two nodes that carry a real rule (a bare column that names an alias; a method-call string receiver the walker can't reach) and delegate all structural recursion to walk::map_children. That:

• retires ~230 lines of hand-rolled arms and the catch-all;
• picks up the four subquery-LHS variants for free — the walker visits their same-scope operands while treating the nested SelectStatement as an opaque scope boundary, so an alias name reused inside the subquery body is correctly left untouched.

Regression: where_alias_expander::test_expands_alias_on_in_subquery_lhs_not_body.

2. Expression LHS in evaluate_in_list / evaluate_between

Once dbl expands to price * 2, the LHS is a compound expression — and price * 2 IN (SELECT ...) fails even with no alias. The InOperatorLifter normally lifts an expression-LHS IN-list into a computed CTE column, but it only matches InList/NotInList; an IN-subquery's InList is synthesized later by SubqueryExecutor, after the lifter has run, so evaluate_in_list received a raw expression LHS and its extract_column_name errored.

Added RecursiveWhereEvaluator::evaluate_operand_value, which looks up a plain column but evaluates any other expression through the ArithmeticEvaluator — the same delegation evaluate_binary_op already does for its LHS. IN and BETWEEN now both accept a compound LHS. This stands on its own (it also fixes alias-free expr IN (subquery)); the long-term home is the R7 per-row subquery evaluation.

Testing

• Parity harness: 77 AGREE / 1 DIFFER / 5 GAP, contract holds (83 cases).
• cargo test --lib: 687 passed, 0 failed (incl. the new expander test).
• cargo test --test integration: 407 passed, 0 failed.
• cargo fmt clean.

🤖 Generated with Claude Code