Consolidation 02_parse Level-1b — RE-SPAWN (prior fierce-wren died): migrate the remaining GrammarExpr hand-match queries (expr_direct_left_corner_list, expr_has_undefined_nonterminal, parse_expr) onto fold_grammar_expr (the Level-1 catamorphism, now on main). Additive + equivalence-preserving; pars

[briansrls]

Summary

Migrates the remaining three GrammarExpr hand-match queries in src/v4/compiler/02_parse.dag onto fold_grammar_expr (Level-1 catamorphism from #4637):

• expr_direct_left_corner_list — GrammarExprCornerFold product threads nullable alongside corners through a single fold (Sequence gates right corners on left nullability).
• expr_has_undefined_nonterminal — GrammarExprFold<Bool> algebra; does not follow nonterminal references into productions.
• parse_expr — fold_grammar_expr builds a deferred ParseExprWalk; parse_expr_run_walk interprets with token threading (equivalent to the prior direct-recursion shape, including T-8 span stamping).

Additive + equivalence-preserving: no match expr arms remain in 02_parse.dag.

Test plan

• parse_table_token_position_indices_no_extra_append_holds — green by execution
• parse_table_nullable_morphism_holds — green by execution
• parse_table_grammar_expr_projection_holds — green by execution
• parse_table_left_corner_morphism_holds — green by execution (new)
• parse_table_undefined_nonterminal_morphism_holds — green by execution (new)
• grammar_validation.dag claims — still deferred (T-22 runner / contains-Variant gap); parse_expr exercised there when T-22 lands

[briansrls]

Addressing the review findings in 6be0645c:

Finding 1 (ParseExprWalk parallel representation): Valid. The fold stage here is identity-shaped — structural recursion over GrammarExpr is dissolved into fold_grammar_expr, but interpretation still requires a per-variant match on the interim ParseExprWalk coproduct. That is not full walker dissolution.

Fixed by downgrading both ParseExprWalk and parse_expr from 🟢 to 🟡 with bounded dissolution receipts:

• ParseExprWalk: parallel 1:1 mirror; dissolve-on-arrival via a higher-order GrammarExprFold algebra whose handlers thread tokens and yield ParseExprResult directly (trigger: Level-2 parse-walk algebra in std/grammar.dag or T-8 parse-result projection unification).
• parse_expr: marked as partial migration — fold dissolves structural recursion; parse_expr_run_walk remains interim interpretation. Forbidden to claim full dissolution until the higher-order algebra lands.

The two query consumers (expr_direct_left_corner_list, expr_has_undefined_nonterminal) remain genuine fold-algebra consumers with non-identity combine semantics — unchanged.

Finding 2 (Sequence nullable threading): Confirmed correct. left_fold.nullable is fold_grammar_expr applied with the same Bool nullable algebra as expr_is_nullable_given_set over the left subtree — equivalent to recomputing expr_is_nullable_given_set(expr: left, nullable_set). Added an inline comment at expr_corner_fold_sequence documenting this.

All five parse_table_claims witnesses still green by execution on this HEAD.

— sent from calm-gull-609