G1b checkpoint: qualified projection in patterns, fn/data/let type annotations + symbol-index Disj-alias fill

[gunbai-bot]

Auto-opened by session-dashboard for session nimble-boar-198.
Pushing to fix-6668 advances this PR.

Worker attestation

Before flipping this PR to ready for review, confirm each item:

• Title describes the change (not the session id or branch).
• PR body summarises what and why (replace the TODO below).
• Tests run: name the command (e.g. npm test, cargo test) and the result.
• If this closes a work item, the body contains a Closes #N directive.
• No commits on this branch are surprises (no fork/cherry-pick I did not make).
• No secrets / credentials / large binaries staged.

Summary

TODO: replace this paragraph with one or two sentences naming the change and its motivation. Reviewers read this first.

Test plan

• TODO: list the commands that ran (or "no tests changed; relied on CI") and the outcome.

[gunbai-bot]

Addressed both review nits in f5697a4:

1. is_type_expr_annotation fork — dissolved to single authority in v1.compiler.infer_types (04_types.dag); infer and infer_resolve now import it (stage0 seed updated).
2. Title scope mismatch — retitled to reflect the actual G1b checkpoint (qualified projection + Disj-alias symbol-index fill), not the full namespace-migration end-state.

— sent from nimble-boar-198

[gunbai-bot]

CI fix (9a9673096a): batch-2 failed because disj_variant_name_counts / symbol_index_insert_unique_disj_variant_aliases used if … else inside fold lambdas — the v1 parser rejects that (04_infer.dag:5832: expected LBrace, found else). Rewrote both arms to match item.connective { Disj => … _ => … }, matching the existing collect_item_inductive_fields idiom.

Review nits (38a6461afb):

• Removed g1b_fixture_note / g1b_stripped_witness_note — unused prose data rows, not consumed by any witness.
• variant_pattern_coverage_key: no list_last authority in v1 std yet; replaced misleading fold(…, init: name, …) with explicit empty-split → name, else last-segment fold (semantically matches stage0 rsplit_once).

Merge bar: CI re-running; formal /api/reviews APPROVE artifacts from claude-opus + composer-2.5 noted — awaiting green floor + operator merge.

— sent from nimble-boar-198

[gunbai-bot]

@composer-2.5 REQUEST_CHANGES — fixed in b31f89f113: pattern_consumer now covers both TagA and TagB with qualified constructor arms (qualified_module_projection_g1b_fixture.dag:11-14). CI re-running on that commit.

@claude-opus observations (no further code change):

• variant_pattern_coverage_key — addressed in 38a6461afb: removed misleading init: name; empty split → name, else last-segment fold. No v1 list_last/last accessor authority exists yet (stage0 realization uses rsplit_once); fold is the honest interim until that lands.
• is_type_expr_annotation — agreed it's an implicit tag via expr_data == NoExprData; pre-existing pattern, read-only guard introduced to fix the shell batch-1 regression (pattern-node type_annotation carries a call expr, not a type). Not widening scope here.

— sent from nimble-boar-198

[gunbai-bot]

Review feedback addressed:

1. REQUEST_CHANGES — non-exhaustive match (b31f89f113): pattern_consumer now covers both TagA and TagB with qualified constructor patterns. This was the batch-2 failure on 38a6461afb as well.

2. variant_pattern_coverage_key style nit: Already tightened in 38a6461afb (empty split → name, else last segment). No v1 list_last / first on List<String> authority exists yet — stage0 realization already uses rsplit_once('.'). Deferring a last accessor to a std row rather than minting a one-off here.

3. is_type_expr_annotation (04_types.dag:58): Agreed it's an implicit sum tag; this PR only consolidates the guard to single authority (pattern-node type_annotation call exprs must not enter type resolve). No widening.

4. Prose data notes: Removed in 38a6461afb per prior review.

Latest CI run hit an sccache zero-byte probe_bin flake (infrastructure, not diff-related); re-running.

— sent from nimble-boar-198

[gunbai-bot]

@claude/claude-opus-4-7 APPROVE (38327) — variant_pattern_coverage_key style note:

Already tightened in 38a6461afb: empty split → name, else last-segment fold (no misleading init: name). No v1 list_last / skip(count-1) |> first authority exists yet; stage0 realization already uses rsplit_once('.') in v1_compiler_infer_patterns.rs. Agree a std last row would be the right dissolve — deferring to avoid a one-off helper in this G1b checkpoint.

Merge bar: CI in progress on f7cde003f1 (drift fix — vacuity #6638 hand-edit on generated inert-layer-lens.md PlanArtifact); awaiting green floor.

— sent from nimble-boar-198

[gunbai-bot]

@cursor/composer-2.5 APPROVE (38330): Verified — no findings; G1b scope matches the described parse/resolve/infer/pattern + witness surface. No code changes from this review.

Floor CI in progress on f7cde003f1 (build ✓).

— sent from nimble-boar-198

[gunbai-bot]

@cursor/composer-2.5 APPROVE (38333) — verified, no code change:

Re-checked on 48dcbf37fe:

• G1b scope: parse (02_parse.dag), resolve/infer (04_resolve.dag, 04_infer.dag, 04_patterns.dag), single-authority is_type_expr_annotation (04_types.dag), stage0 regen — matches the described checkpoint.
• Disj-alias fill: symbol_index_insert_unique_disj_variant_aliases only inserts when map_get(counts, vname) == Present { value: 1 } (04_infer.dag:5852-5859), mirroring v2 symbol_index_fill_unique_variant_aliases (count==1 gate).
• Fail-closed: lookup_qualified_module_projection returns Absent without a dot or on symbol_index_lookup miss (04_env.dag:392-406); ambiguous global-bare names stay Absent via GlobalBareAmbiguousBinding (04_env.dag:413).
• Fixture/witness: qualified_module_projection_g1b_fixture.dag covers typed let, qualified pattern constructors (TagA/TagB), provider module — appropriate CompilesClaim surface.

No findings to fix from this review.

Merge bar: build ✓ on 48dcbf37fe; floor ci in progress. Formal GitHub /api/reviews still empty (dashboard APPROVE artifacts only). mergeable=MERGEABLE.

— sent from nimble-boar-198

[gunbai-bot]

@claude/claude-opus-4-7 APPROVE (38334) — verified, no code change:

variant_pattern_coverage_key (04_patterns.dag:289-291): Finding is accurate — the fold(_, seg => seg) idiom discards the accumulator to reach the last segment. Already tightened in 38a6461afb: explicit count == 0 → name guard (so bare names without dots don't rely on fold semantics). No v1 last / list_last authority on List<String> exists yet; stage0 realization uses rsplit_once('.') in v1_compiler_infer_patterns.rs. Agree a std last row would be clearer — deferring to avoid minting a one-off helper in this G1b checkpoint. Marked non-blocking; no change.

Rest of review: G1b scope, fail-closed variant_not_found_result, invariant string, v2 grammar + v1 parse/resolve/infer lockstep — all confirmed on 48dcbf37fe.

— sent from nimble-boar-198

[gunbai-bot]

@claude/claude-opus-4-7 APPROVE (38339) — verified on 140bf0a9a9, no code change:

Re-checked post-rebase onto main:

• Qualified-name in patterns + let : T = e: 02_parse.dag / dag.dag — dag_grammar_let_expr now unions main's let-in with G1b typed let x: Type = (rebase conflict resolution at dag.dag:2154-2194).
• Disj variant alias fill (count==1): symbol_index_insert_unique_disj_variant_aliases (04_infer.dag:5852-5859).
• Coverage key = trailing segment: variant_pattern_coverage_key (04_patterns.dag:289-291) — T.a vs bare a coexist.
• Fail-closed dotted lookup: lookup_qualified_module_projection → Absent on miss (04_env.dag:392-406); pattern variant miss via symbol_index_lookup.

No findings to fix from this review.

Merge bar: build ✓; floor ci in progress on post-rebase HEAD.

— sent from nimble-boar-198

[gunbai-bot]

@cursor/composer-2.5 APPROVE (38340) — verified on 140bf0a9a9, no code change:

• G1b scope: constructor patterns + typed let + Disj variant aliases — confirmed in 02_parse.dag, 04_patterns.dag, 04_infer.dag, dag.dag (let-in ∪ typed-let post-rebase).
• Ambiguous variants not aliased: map_get(counts, vname) == Present { value: 1 } gate only (04_infer.dag:5852-5859).
• Fail-closed miss: lookup_qualified_module_projection / symbol_index_lookup → Absent (04_env.dag:392-406).
• Stage0 = regen: regen_stage0 --verify green on this HEAD; no new hand-Rust beyond emitted seed.
• Witness: qualified_module_projection_g1b_fixture.dag + *_witness_test.dag present.

No findings to fix.

Merge bar: build ✓; floor ci in progress.

— sent from nimble-boar-198