02_parse Level-1d: migrate grammar validity scans, production lookup, and token head-peek onto std/ catamorphisms

[briansrls]

02_parse Level-1d: migrate grammar validity scans, production lookup, and token head-peek onto std/ catamorphisms

Continues the stage-adoption lane (#4639 Level-1b GrammarExpr folds, #4651 Level-1c capture spans → fold_node_tree_span). Per MODELING M9 the std/ DFS came first: every replacement below consumes an operation that already exists in v4.std.algebra / v4.std.collection — no new substrate was added.

Hand-match queries dissolved

hand-written (deleted)	std/ catamorphism consumed
grammar_dup_check_step + GrammarDupCheckResult accumulator coproduct (🟡)	v4.std.algebra.unique over production names
grammar_has_left_recursion_step or-fold	v4.std.algebra.any
grammar_has_undefined_nonterminals_step or-fold	v4.std.algebra.any
grammar_lookup_production_step first-match Witness fold	v4.std.algebra.filter + v4.std.collection.list_at_optional head projection
TokenPeek per-use carrier coproduct (🟡) + hand fold_list_right peek	v4.std.collection.list_at_optional (Optional<Token> head projection — the mark's exact dissolve-on-arrival condition)

Net: two 🟡 carrier types and four hand step-folds deleted; grammar_has_duplicate_productions's 🟡 upgraded to 🟢 substrate-consumer (the marked scan morphism arrived as unique). The left-recursion / undefined-nonterminal 🟡 marks stay (their full reachability/scan morphisms have not arrived; only the or-fold plumbing migrated). parse_token_first survives as a one-line typed bridge pinning T = Token for the bootstrap compiler's monomorphic field-access inference (direct inline match failed to infer tok.class).

Equivalence notes:

• filter is order-preserving, so head-of-filtered == old first-match fold; first match == unique match post-validation since duplicate names are rejected pre-parse.
• unique (pairwise name eq) == old set-membership scan.

Claims (src/v4/test/claim/parse/grammar_validation.dag)

• claim_distinct_productions_not_duplicate — negative dup-scan case; guards the polarity of the unique migration (existing claim only covered the duplicate→true direction).
• claim_lookup_production_present_holds — Holds payload is the production with the requested name (second production in the fixture, so a head-only/wrong-filter lookup goes red).
• claim_lookup_production_absent_violates — fail-closed Violates for a missing name.

Verification

• gunbc run --claim-run green on both new lookup witnesses (true, exit 0).
• Scratch combined witness (not committed) executed all six migrated behaviors against the existing fixtures — dup true/false, undefined-start, dangling-ref, left-recursive true, right-recursive false — plus the full right-recursive parse_production path (exercises the Terminal-arm head projection and table lookup): all green.
• M1 full-tree emit probe (.github/ci-floor/v4-m1-rust-emit-probe.sh): compiled: 442 files emitted, 0 diagnostics.

🤖 Generated with Claude Code

[briansrls]

Re the non-blocking note on match unique(...) { true => false, false => true }: no not primitive exists in v4.std.logic (or elsewhere in std/) today, and the match-negation is the established idiom — v4.std.algebra.non_empty negates is_empty the same way (src/v4/std/algebra.dag:681). Adding a not to std/logic would be new substrate (model-first lane), out of scope for this migration PR and not worth staling the two approvals on the resolved head; happy to fold it into a future std/logic PR if the lane wants it. — sent from sleek-wren-682