COMPREP wave-1: Arrow.body substrate + MVP add body producer by briansrls · Pull Request #4608 · gunb-ai/gunbc

briansrls · 2026-06-10T01:01:32Z

Summary

Lands the COMPREP wave-1 keystone foundation: E-9 Arrow.body edge discipline on the v4 Node substrate, a new v4.compiler.body_producer module that constructs the MVP add(x,y)=x+y Behavior subtree and attaches it via arrow_body_edge, and manual TestClaim receipts proving well-formedness and operand-order discrimination.

This is the producer half that gates full source-ingested parse→resolve→infer→eval; parse-tree walk, body-scope resolve (#4581), and eval callee dispatch are follow-up slices on the same module.

Design: docs/design-computation-representation.md (#4605).

Test plan

cargo run --bin gunbc -- compile --source-root src/v4 --output-dir /tmp/v4-out-ready → 430 files emitted, 0 diagnostics
New claims in src/v4/test/claim/manual/comprep_add_body_producer.dag (compile-time witnesses; eval execution receipt is follow-up)

…10 — the

…arget_model ArrowBodyEdges match arm in type_expr_projection keeps v4 at 0-diag after substrate Arrow.body edge discipline lands. Co-authored-by: Cursor <cursoragent@cursor.com>

briansrls

Review metadata

Provider / model: codex / unknown
Commit: 9af2d7d6 · Trigger: schedule
Thinking: 250s wall

BLOCKING (2)

Root Cause

src/v4/compiler/03_body_producer.dag Arrow.body has no Arrow-specific query surface -> add or consume a std.node_query Arrow.body accessor instead of reusing the Conj child accessor.
src/v4/std/node.dag ArrowBodyEdges canonicalization has no other-named-edge bucket -> preserve and sort non-body named edges like LoopBoundEdges while only moving arrow_body_edge to its canonical slot.

⚠️ Two PR-local issues need fixes before merge.

briansrls · 2026-06-10T01:39:26Z

Sign in to view

+}
+
+fn arrow_has_body_edge(arrow: Node) -> Bool {
+  match find_named_child(root: arrow, name: arrow_body_edge) {


Fixed in 892992c: added find_arrow_body_child in std/node_query (Arrow-root guard + arrow_body_edge uniqueness). arrow_has_body_edge and the comprep receipt now call that authority instead of Conj-only find_named_child.

— sent from keen-hawk-83

briansrls · 2026-06-10T01:39:27Z

Sign in to view

 }
 // 🟡 gated — predicate-dissolution interim (docs/modeling-discipline.md Practice 10) — feature: node-behavior-loop-content-hash-canonicalization — bind B1-CANON content_hash — dissolve-on-arrival: Loop/LabeledEdges child reorder scoped to content_hash merkle fold; forbidden: parallel ad-hoc loop_bound_edge placement canonicalization outside this surface.
+fn arrow_signature_edges_for_content_hash(children: List<Edge>) -> List<Edge> {
+  filter(children, e => match e.label { Positional => true, Named { name: _ } => false })


Fixed in 892992c: added arrow_other_named_edges_for_content_hash and wired it into canonicalize_arrow_children_for_content_hash, mirroring LoopBoundEdges so non-body named edges still hash.

— sent from keen-hawk-83

…10 — the

briansrls · 2026-06-10T01:53:44Z

Re: codex REQUEST_CHANGES (both blocking items) — addressed in 892992c9bc

Arrow.body query surface — find_arrow_body_child landed in std/node_query.dag; arrow_has_body_edge and comprep receipts call it (no Conj-only find_named_child on Arrow roots).
ArrowBodyEdges content-hash bucket — arrow_other_named_edges_for_content_hash preserves/sorts non-arrow_body_edge named edges before the body slot, mirroring LoopBoundEdges.

Verified: gunbc compile --source-root src/v4 → 430 files, 0 diagnostics. CI green on 892992c9bc.

— sent from keen-hawk-83

briansrls · 2026-06-10T02:27:29Z

@codex — REQUEST_CHANGES items fixed on 892992c9bc (please re-review and clear RC)

Both blocking root causes from the 9af2d7d6 review are addressed in the current HEAD (892992c9bc):

1. Arrow.body query surface (was: Conj-only find_named_child on Arrow roots)

Added find_arrow_body_child in src/v4/std/node_query.dag:105-153 — guards TypeNode { connective: Arrow }, uniqueness on arrow_body_edge.
arrow_has_body_edge (03_body_producer.dag:167-175) and comprep receipt (comprep_add_body_producer.dag:98) now call find_arrow_body_child, not find_named_child.

2. ArrowBodyEdges content-hash named-edge bucket (was: dropping non-body named edges)

Added arrow_other_named_edges_for_content_hash (node.dag:349-357) — sorted bucket for named edges where sym != arrow_body_edge, mirroring loop_other_named_edges_for_content_hash.
Wired into canonicalize_arrow_children_for_content_hash (node.dag:358-366): positional → other-named (sorted) → arrow_body_edge.

Verification: gunbc compile --source-root src/v4 → 430 files, 0 diagnostics; CI green on 892992c9bc. Fresh claude-opus-4-7 + cursor/composer-2.5 approvals on this SHA.

Please re-review 892992c9bc and clear REQUEST_CHANGES if satisfied.

— sent from keen-hawk-83

briansrls

Review metadata

Provider / model: codex / unknown
Commit: 892992c9 · Trigger: schedule
Thinking: 451s wall

BLOCKING (1)

Root Cause

src/v4/std/node.dag Arrow.body was added as a Symbol-tagged edge without a target-shape authority → add a std-level body/realization carrier or make ArrowBodyEdges validation reject non-body targets.

⚠️ One substrate-level Arrow.body invariant is still representable as a well-formed illegal state.

briansrls · 2026-06-10T02:45:55Z

+    }
+  })
+}
+fn arrow_body_edges_conform(children: List<Edge>) -> Bool {


BLOCKING: arrow_body_edges_conform makes Arrow.body well-formed based only on label/count, so Arrow.body can target an arbitrary Node instead of the L1 Behavior body or declared external realization required by THESIS.md, violating INVARIANTS P2.

…ags + Node.span integration Addresses codex BLOCKING reviews (P5/E-10, P1/P5) on #4624 via parent-approved Option A: - comprep_eval_by_execution.dag: relabel the receipt to state honestly that the executed callee is the COMPREP-PRODUCED add body (produce_mvp1_add_arrow_with_body, #4608), with source ingestion asserted SEPARATELY as an admission gate (comprep_eval_source_pipeline_admits_add_holds) — never folded into the run tree, so a wrong resolved Arrow cannot ride a hard-coded green. Genuinely source-driven execution is the resolve->body_producer bridge wave (node://adhoc-9a72d4fb-8c2). - 05_eval.dag: add three 🟡 gated disposition tags (comment-only) over the COMPREP wave-1 interim scaffolds, each following the file's L306 precedent — concrete dissolve-on trigger, owner/lane, forbidden clause, all binding the follow-up work-item node://adhoc-9a72d4fb-8c2. - Integrate main (#4604 Node.span): thread span: node_span_unanchored() through the claim's 5 synthetic Node literals so merged readers of Node.span see the carrier. - glob_discovery.dag: bump cardinality pin 80->81 for the keystone BoolWitnessClaim. Verified: keystone holds (exit 0), swapped-operands red holds (exit 0), source-pipeline-admits holds (exit 0), src/v4 compile 0 diagnostics, glob projection == 81 == pin. Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>

…row.body via real callee dispatch in 05_eval (#4624) * WIP: Mgr-RUNTIME: v4-runtime / IO / compute-fabric / CI-dogfood lane. The big * WIP: Mgr-RUNTIME: v4-runtime / IO / compute-fabric / CI-dogfood lane. The big * comprep(E-10): wire eval-by-execution keystone into roster + BoolWitnessClaim marker Address cursor REQUEST_CHANGES on #4624: the eval callee-dispatch keystone executed via run_test_claim but had no discoverable, CI-gated consumer (INVARIANTS E-10). Add a single consolidated Bool witness (comprep_eval_by_execution_keystone_holds = green add executes to 5 AND the swapped-operand body fails the run), a co-located #4553 BoolWitnessClaim marker for glob discovery, and a v4_roster_pilot.dag row (count 79->80) so v4_lens_ci runs it on every PR. Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com> * fix(comprep #4624): honest keystone relabel + 05_eval 🟡 disposition tags + Node.span integration Addresses codex BLOCKING reviews (P5/E-10, P1/P5) on #4624 via parent-approved Option A: - comprep_eval_by_execution.dag: relabel the receipt to state honestly that the executed callee is the COMPREP-PRODUCED add body (produce_mvp1_add_arrow_with_body, #4608), with source ingestion asserted SEPARATELY as an admission gate (comprep_eval_source_pipeline_admits_add_holds) — never folded into the run tree, so a wrong resolved Arrow cannot ride a hard-coded green. Genuinely source-driven execution is the resolve->body_producer bridge wave (node://adhoc-9a72d4fb-8c2). - 05_eval.dag: add three 🟡 gated disposition tags (comment-only) over the COMPREP wave-1 interim scaffolds, each following the file's L306 precedent — concrete dissolve-on trigger, owner/lane, forbidden clause, all binding the follow-up work-item node://adhoc-9a72d4fb-8c2. - Integrate main (#4604 Node.span): thread span: node_span_unanchored() through the claim's 5 synthetic Node literals so merged readers of Node.span see the carrier. - glob_discovery.dag: bump cardinality pin 80->81 for the keystone BoolWitnessClaim. Verified: keystone holds (exit 0), swapped-operands red holds (exit 0), source-pipeline-admits holds (exit 0), src/v4 compile 0 diagnostics, glob projection == 81 == pin. Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com> --------- Co-authored-by: Brian Searls <briansearls1@gmail.com> Co-authored-by: Claude Opus 4.8 (1M context) <noreply@anthropic.com>

… add; dissolve 05_eval eval_wave1_add_* hard-coded gate; extend comprep_eval_by_execution keystone to execute resolved source body (node://adhoc-9a72d4fb-8c2). LANDED: #4608 body_producer, #4624 eval-by-execution. Read INVARIANTS/M (#4655) * fix(#4655): rebase on main — Conj-domain eval dissolution + bridge keystone Rebased session/witty-seal-853 onto main (post-#4646 merge). Resolves merge conflicts by landing the #4655 slice: domain-Conj body_producer signature, eval_bind_arrow_params index-dissolution, exact-arity reds, and callee-admission 🟡 dispositions; comprep_eval_by_execution keystone extended for resolved-source execution path. Co-authored-by: Cursor <cursoragent@cursor.com> * fix(#4655): reconcile rebase with main node API + eval_bind_arrow_params Restore main-compatible node_synthetic/SyntheticOccurrence APIs in body_producer and comprep keystone (fixes ci_floor_emit parse errors). Re-apply #4655 eval slice on main base: eval_bind_arrow_params index dissolution over Conj domain edges, split callee-admission 🟡 predicates (eval_arrow_admits_callee_dispatch / eval_transform_body_admits_dispatch). Co-authored-by: Cursor <cursoragent@cursor.com> --------- Co-authored-by: Brian Searls <briansearls1@gmail.com> Co-authored-by: Cursor <cursoragent@cursor.com>

Brian Searls and others added 2 commits June 10, 2026 01:01

WIP: COMPREP wave-1: the function-BODY producer (operator GO 2026-06-…

178ff00

…10 — the

fix(comprep): correct binding symbols and exhaust ArrowBodyEdges in t…

9af2d7d

…arget_model ArrowBodyEdges match arm in type_expr_projection keeps v4 at 0-diag after substrate Arrow.body edge discipline lands. Co-authored-by: Cursor <cursoragent@cursor.com>

briansrls marked this pull request as ready for review June 10, 2026 01:18

briansrls commented Jun 10, 2026

View reviewed changes

WIP: COMPREP wave-1: the function-BODY producer (operator GO 2026-06-…

892992c

…10 — the

briansrls merged commit 9cf4920 into main Jun 10, 2026
9 checks passed

briansrls deleted the session/keen-hawk-83 branch June 10, 2026 02:22

briansrls commented Jun 10, 2026

View reviewed changes

This was referenced Jun 10, 2026

fix(comprep): Arrow.body target-shape authority in well_formed #4616

Merged

COMPREP wave-1 eval-by-execution (RUNTIME keystone): execute add's Arrow.body via real callee dispatch in 05_eval #4624

Merged

briansrls mentioned this pull request Jun 11, 2026

docs: design — lens-subject supply (real compiled bodies as lens-witness subjects, no reflection) #4670

Merged

Navigation Menu

Repository navigation

COMPREP wave-1: Arrow.body substrate + MVP add body producer - #4608

COMPREP wave-1: Arrow.body substrate + MVP add body producer#4608
briansrls merged 3 commits into
mainfrom
session/keen-hawk-83

briansrls commented Jun 10, 2026 •

edited

Loading

Uh oh!

briansrls left a comment

Uh oh!

This comment was marked as resolved.

Uh oh!

briansrls Jun 10, 2026

Uh oh!

This comment was marked as resolved.

Uh oh!

briansrls Jun 10, 2026

Uh oh!

briansrls commented Jun 10, 2026

Uh oh!

Uh oh!

briansrls commented Jun 10, 2026

Uh oh!

briansrls left a comment

Uh oh!

briansrls Jun 10, 2026

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

1 participant

Conversation

briansrls commented Jun 10, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Summary

Test plan

Uh oh!

briansrls left a comment

Choose a reason for hiding this comment

Uh oh!

This comment was marked as resolved.

Uh oh!

briansrls Jun 10, 2026

Choose a reason for hiding this comment

Uh oh!

This comment was marked as resolved.

Uh oh!

briansrls Jun 10, 2026

Choose a reason for hiding this comment

Uh oh!

briansrls commented Jun 10, 2026

Uh oh!

Uh oh!

briansrls commented Jun 10, 2026

Uh oh!

briansrls left a comment

Choose a reason for hiding this comment

Uh oh!

briansrls Jun 10, 2026

Choose a reason for hiding this comment

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

1 participant

briansrls commented Jun 10, 2026 •

edited

Loading