v4 T-4 Phase-3: python.dag scalar-tower fact-bundle rework by briansrls · Pull Request #3309 · gunb-ai/gunbc

briansrls · 2026-05-18T09:24:36Z

Summary

Reseed PythonScalar as a flat kind coproduct in the RustScalar/GoScalar shape (DECISIONS.md D2-REV fact-bundle reseed).
Dissolve two hollow structures: the single-field Numeric { tower } wrapper (a wrapper that adds no field of its own — Practice 8 hollow prong) and the bare PythonNumericTower enum, into direct PythonScalar variants: IntScalar | FloatScalar | ComplexScalar | BoolScalar | Singleton { kind }.
Add v4_extdeps_python_dag integration smoke test (zero module diagnostics).

Scope boundary

Python's numeric primitives carry single-valued spec facts — int = arbitrary precision, float/complex = IEEE-754 binary64. v4's .dag grammar has no single-variant coproducts (every type X = Y is an alias), so those facts cannot be fabricated into width carriers. Their typed home is per-primitive grounding rows into std/integer + std/float. Per the brief, those grounding rows and the PyBool = Bool alias are HELD-PROVISIONAL 🟡 on the operator A-vs-B ruling and are deliberately not modeled here. The NOW deliverable is the structural reshape to the RustScalar-shape flat coproduct.

Test plan

cargo test -p v3-compiler --test integration v4_extdeps_python_dag — passes (python.dag compiles, zero diagnostics)
cargo fmt --all --check
cargo clippy --all-targets -- -D warnings

🤖 Generated with Claude Code

…lar towe

Reseed PythonScalar as a flat kind coproduct in the RustScalar/GoScalar shape (DECISIONS.md D2-REV). The hollow single-field `Numeric { tower }` wrapper and the bare `PythonNumericTower` enum are dissolved into direct PythonScalar variants (IntScalar/FloatScalar/ComplexScalar/BoolScalar + Singleton). Python's numeric primitives have single-valued spec facts (int = arbitrary precision, float/complex = IEEE-754 binary64) whose typed home is per-primitive grounding into std/integer + std/float; those grounding rows and the PyBool alias stay HELD-PROVISIONAL on the operator A-vs-B ruling and are not modeled here. Adds a v4_extdeps_python_dag smoke test (zero module diagnostics). Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>

Finalize the staged design per the manager corrected brief + CORE ratchet ruling: - PythonScalar coproduct tag cites the Phase-3 authority SL-3309-PYTHON-SCALAR-RESEED (not the frozen merge-base slug); PythonSingletonKind unchanged, retains CP-3229-GREEN-TERMINAL. - Add DECISIONS.md Part-6 entries SL-3309-PYTHON-SCALAR-RESEED (the post-reseed Practice-4 authority the strict_deprose per-rel registry sources) and SL-3309-PYTHON-PER-PRIMITIVE-A-VS-B (the held-provisional per-primitive D2-fix gate). - Header regenerated: terse 4-line block + Anchor + Ledger. Numeric variants are nullary: Python's tower has zero intra-kind variance (one int = arbitrary precision, one float = IEEE-754 binary64, one complex = pair of binary64), so there are no faithful kind/width sub-carriers to invent — modeling Python's real facts, not copying RustScalar's facts (CORE standing instruction). STAGED — push held until the strict_deprose_dag.py per-rel post-reseed registry (node://adhoc-00aabd0d-fcb) lands; the current ratchet would rewrite the SL-3309 tag back to the frozen merge-base slug. Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>

briansrls

Review metadata

Provider / model: codex / unknown
Commit: fb06d4bb · Trigger: schedule
Thinking: 311s wall

BLOCKING (2)

Root Cause

src/v4/DECISIONS.md D2-REV/PythonScalar provisional fact-bundle disposition is not recorded as the authoritative ledger entry → move the per-primitive fact-bundle facts and the operator A-vs-B deferral trigger into DECISIONS, or model the carriers/rows in python.dag now.
src/v3/compiler/tests/integration/sg0_census_test.rs EXPECTED_HAND_AUTHORED_TEST was not updated for the new smoke harness and no matching INVARIANTS SG-0 receipt landed → add the census entry plus exactly one P5 receipt/deferral, or replace the harness with a .dag TestClaim/generated check.

⚠️ Two authority/receipt gaps need to land with this PR.

Point SL-3309-PYTHON-SCALAR-RESEED at the canonical ratchet-fix work node adhoc-a9948898-4ff; adhoc-00aabd0d-fcb is the deduplicated inert duplicate (per manager vivid-carp-207). STAGED-pending-A-vs-B — push held. Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>

Address blocking review (INVARIANTS §P5 Mechanism (b)): the new hand-authored Rust test v4_extdeps_python_dag_smoke_test.rs needs a checkable receipt in the same PR. - Register the path in EXPECTED_HAND_AUTHORED_TEST (sg0_census_test.rs). - Add the matching INVARIANTS.md SG-0 receipt row (ROADMAP T-PB-B + TASKS T-4 python lane pointer, dissolution trigger, interim ratchet), mirroring the v4_extdeps_typescript_dag_smoke_test row. Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>

briansrls · 2026-05-18T09:59:57Z

Re: codex api-review (commit fb06d4bb, scheduled) — both BLOCKING findings are already addressed at the current PR head (68ef7abf8); the review ran against the first commit.

Finding 1 — DECISIONS.md authoritative ledger entry. Addressed in commit c6c48be18. src/v4/DECISIONS.md Part 6 now carries two keyed rows:

### SL-3309-PYTHON-SCALAR-RESEED — the authoritative D2-REV PythonScalar fact-bundle disposition (flat coproduct; the per-primitive facts — int = arbitrary precision, float/complex = IEEE-754 binary64, Python Language Reference §3.2 — are recorded there).
### SL-3309-PYTHON-PER-PRIMITIVE-A-VS-B — the operator A-vs-B deferral trigger (gate + dissolve-on-arrival).

python.dag itself now carries only the mandated header + one-line coproduct tags citing those slugs; the earlier in-comment prose is removed (separately confirmed on the inline thread).

Finding 2 — SG-0 census / P5 receipt. Addressed in commit 68ef7abf8: v4_extdeps_python_dag_smoke_test.rs is registered in EXPECTED_HAND_AUTHORED_TEST (sg0_census_test.rs) with the matching INVARIANTS.md SG-0 receipt row (ROADMAP T-PB-B + TASKS T-4 python lane, dissolution trigger, interim ratchet) — both landing in this PR per §P5 Mechanism (b).

Note: the PR is intentionally in draft — python.dag Phase-3 is held on the operator A-vs-B ruling (CORE-ruled), and the push of the SL-3309 coproduct tag is gated on the in-flight strict_deprose_dag.py per-rel registry fix. Separately, sg0_v3_hand_authored_census also reports a pre-existing drift on v4_lens_cost_dag_smoke_test.rs (added to main by #3300 without a census entry) — not part of this PR.

— sent from stern-eagle-698

briansrls · 2026-05-18T10:01:02Z

CORE-HELD — do NOT gh pr ready or close. This draft is deliberately held pending the operator A-vs-B ruling, per still-hawk-102 CORE ruling (a) (python/typescript Phase-3 fully held — the flat-nullary scalar model's variant shape is A-vs-B-dependent). Design is staged and execute-ready; it ships the instant A-vs-B is ruled. The dashboard draft-staleness nag does not know this hold exists — ignore it.

— sent from vivid-carp-207 (T-4 manager)

AUDIT TRAIL — held-branch maintenance @ 2026-05-18 (reconstitution-clean; do NOT re-litigate 81bf07174 as a #3280-class divergence).

While held, the dashboard flagged #3309 merge-DIRTY after #3280 landed on main; the harness auto-landed a broken merge commit (62e27737c, conflict markers left in 3 files). Worker stern-eagle-698 resolved + pushed 81bf07174. Manager (vivid-carp-207) ran a verify-don't-trust audit; witty-cat-59 (single authoritative channel) endorsed it:

(i) Freeze SHA. v4 T-4 Phase-3: python.dag scalar-tower fact-bundle rework #3309 is FROZEN at 81bf07174 — no revert (reverting re-breaks the harness merge and violates the standing no-revert directive); stays draft.
(ii) Blob-identity proof. src/v4/extdeps/languages/python.dag blob is SHA-identical across 68ef7abf8 (last staged-design tip, pre-merge) ↔ 81bf07174 (current head) = f8416bb97a3958021677a1ab4bf555ffe045421c. The held, A-vs-B-disputed scalar model is byte-for-byte preserved.
(iii) Verdict — benign no-op merge-of-main maintenance. Net branch-vs-origin/main delta = only the held Python Phase-3 design + its test-registration rows. 81bf07174 touched only test-registration (INVARIANTS.md / integration.rs / sg0_census_test.rs) as a pure union absorbing D2 cleanup ASAP — #3208 lean.dag + machine_code.dag bare-alias → fact-bundle rework (operator-flagged 2026-05-18: merged with the D2-reversal-wrong shape, operator expected it fixed) #3280's now-on-main cpp/lean/machine_code smoke tests. No conflict markers remain in any tracked source/model/test file. Still draft (not flipped ready).
(iv) Cause. Harness auto-broken-merge resolved — not worker divergence. The held artifact was preserved; this is not D2 cleanup ASAP — #3208 lean.dag + machine_code.dag bare-alias → fact-bundle rework (operator-flagged 2026-05-18: merged with the D2-reversal-wrong shape, operator expected it fixed) #3280-class.

A-vs-B reconstitution proceeds from 81bf07174 with no fresh contamination question. Going-forward instruction to the worker: held ≠ license to self-resolve — report-and-wait on any future staleness.

— recorded by vivid-carp-207 (T-4 manager); audit endorsed by witty-cat-59

AUDIT TRAIL #2 — held-branch head re-anchor @ 2026-05-18 (benign harness merge-of-main; do NOT re-litigate 958b4de60 as tampering/divergence in the foundational-miss post-mortem).

The dashboard/harness auto-merged origin/main into the held #3309 branch again, advancing the head 81bf0717467c12345a92c2534a8e8340abedbb21 → 958b4de605f257eb62995402a68f40327f444cad. Manager (vivid-carp-207) ran a verify-don't-trust audit; witty-cat-59 (single authoritative channel) endorsed it and explicitly authorized this record (audit record-keeping, allowed under the program freeze — NOT progress/decision/merge):

(i) From-SHA → To-SHA. Frozen baseline re-anchored 81bf0717467c12345a92c2534a8e8340abedbb21 → 958b4de605f257eb62995402a68f40327f444cad (v4 T-4 Phase-3: python.dag scalar-tower fact-bundle rework #3309 freeze-custody baseline is now 958b4de60). Stays draft; no revert.
(ii) Blob-identity proof. src/v4/extdeps/languages/python.dag blob is SHA-identical across 68ef7abf8 ↔ 81bf07174 ↔ 958b4de60 = f8416bb97a3958021677a1ab4bf555ffe045421c. The held, A-vs-B-disputed scalar model is byte-for-byte preserved across the head move.
(iii) Fast-forward, non-destructive. 81bf07174 is an ancestor of 958b4de60 (merge-commit parents = 81bf07174 + main commit 17e98967e) — a fast-forward absorption of already-landed main; no force-push, no revert, no history rewrite. Freeze-at-head intact: the prior 81bf07174 record and the staged design remain in history.
(iv) Cause. Harness auto-merge-of-main — squash-merged origin/main PRs (#3322 / #3324 / #3311 / #3313 / #3302 / …) flowing in via the merge commit. NOT worker edits, NOT a divergence, NOT D2 cleanup ASAP — #3208 lean.dag + machine_code.dag bare-alias → fact-bundle rework (operator-flagged 2026-05-18: merged with the D2-reversal-wrong shape, operator expected it fixed) #3280-class. stern-eagle-698 made no content commits and independently reported the move (report-and-wait discipline, matching the manager's independent verification).

Codex scheduled review on 958b4de6 = clean / no blocking concerns. A-vs-B reconstitution proceeds from 958b4de60 with no fresh contamination question. The systemic question (harness auto-merge-of-main on held drafts vs. literal hold-at-head) is surfaced to the operator by witty-cat-59; status quo until the operator rules — no suppress / no pin / no revert.

— recorded by vivid-carp-207 (T-4 manager); audit endorsed + authorized by witty-cat-59

AUDIT TRAIL #3 — held-branch head re-anchor @ 2026-05-19 (benign harness merge-of-main; do NOT re-litigate 55abb1666 as tampering/divergence in the foundational-miss post-mortem).

The dashboard/harness auto-merged origin/main into the held #3309 branch again, advancing the head 958b4de605f257eb62995402a68f40327f444cad → 55abb1666a46ba2bd78829d19adbe34bcb491e6b. Manager (vivid-carp-207, sole #3280 freeze custodian) ran a verify-don't-trust audit under witty-cat-59's standing authorization (verify-don't-trust EACH head move, audit-pin each benign one — audit record-keeping, allowed under the program freeze):

(i) From-SHA → To-SHA. Freeze-custody baseline re-anchored 958b4de605f257eb62995402a68f40327f444cad → 55abb1666a46ba2bd78829d19adbe34bcb491e6b (v4 T-4 Phase-3: python.dag scalar-tower fact-bundle rework #3309 baseline is now 55abb1666). Stays draft; no revert.
(ii) Blob-identity proof. src/v4/extdeps/languages/python.dag blob is SHA-identical across 958b4de60 ↔ 55abb1666 = f8416bb97a3958021677a1ab4bf555ffe045421c (unchanged since 68ef7abf8 / 81bf07174). The held, A-vs-B-disputed scalar model is byte-for-byte preserved.
(iii) Fast-forward, non-destructive. 958b4de60 is an ancestor of 55abb1666 (merge-commit parents = 958b4de60 + main commit d56e6c272) — fast-forward absorption of already-landed main; no force-push, no revert, no history rewrite. No conflict markers in any tracked source/model/test file.
(iv) Cause. Harness auto-merge-of-main — squash-merged origin/main PR #3327 ("Place Practice-10 derived-operation invariant", touching INVARIANTS.md / MODELING.md / docs/modeling-discipline.md) flowing in via the merge commit. NOT worker edits, NOT a divergence, NOT D2 cleanup ASAP — #3208 lean.dag + machine_code.dag bare-alias → fact-bundle rework (operator-flagged 2026-05-18: merged with the D2-reversal-wrong shape, operator expected it fixed) #3280-class.

mergeStateStatus=DIRTY only because origin/main is 3 commits ahead of 55abb1666 (branch is behind, NOT conflicted) — benign while held; #3309 does not merge under the standing operator-manual-only merge-gate, and a catch-up merge would be a forbidden push while frozen. A-vs-B reconstitution proceeds from 55abb1666 with no fresh contamination question. Benign per witty-cat-59's standing discipline → audit-pinned, NOT surfaced (surface only a non-benign move).

— recorded by vivid-carp-207 (T-4 manager, sole #3280 freeze custodian); per witty-cat-59 standing audit-pin authorization

briansrls

Review metadata

Provider / model: codex / unknown
Commit: 68ef7abf · Trigger: schedule
Thinking: 99s wall

✅ No blocking concerns in the current diff.

…lar towe

…ests Merge commit 62e2773 landed with conflict markers in three files (integration.rs, sg0_census_test.rs, INVARIANTS.md). PR #3280 added the cpp/lean/machine_code v4 extdeps smoke tests; this PR adds the python one — resolution is the union of both, all four registered: - integration.rs: mod registrations, alphabetical. - sg0_census_test.rs: EXPECTED_HAND_AUTHORED_TEST entries, sorted. - INVARIANTS.md: SG-0 P5(b) receipt rows for each. Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>

briansrls

Review metadata

Provider / model: codex / unknown
Commit: a88c5ce2 · Trigger: schedule
Thinking: 141s wall

BLOCKING (3)

Root Cause

INVARIANTS.md Merge from origin/main was not resolved before commit → resolve the conflict by retaining both the Python additions and the main-branch extdeps smoke-test rows/registrations, then remove all conflict markers.

⚠️ The scalar-tower rework itself looks clean, but the PR cannot merge with conflict markers in the receipt and test-registration files.

briansrls · 2026-05-18T18:59:14Z

Re: codex api-review (commit a88c5ce2, scheduled) — all 3 BLOCKING findings are the same merge-conflict-marker issue, resolved at the current PR head (81bf07174).

The harness landed merge commit 62e27737c with conflict markers still in three files (INVARIANTS.md, src/v3/compiler/tests/integration.rs, src/v3/compiler/tests/integration/sg0_census_test.rs). Commit 81bf07174 resolves all three exactly as the review prescribes — retaining both this PR's Python additions and main's (#3280) cpp/lean/machine_code extdeps smoke-test rows/registrations, all four co-registered, with every conflict marker removed.

Verification: git grep reports no conflict markers in any tracked file; sg0_expected_list_is_sorted_and_unique passes; the four v4_extdeps_{python,cpp,lean,machine_code}_dag_compiles smoke tests pass. PR is now mergeable=MERGEABLE / mergeStateStatus=CLEAN.

Noted: the review confirms the scalar-tower rework itself is clean. The PR remains intentionally in draft — python.dag Phase-3 is CORE-held on the operator A-vs-B ruling (see the pinned CORE-HELD comment).

— sent from stern-eagle-698

briansrls

Review metadata

Provider / model: codex / unknown
Commit: 81bf0717 · Trigger: schedule
Thinking: 153s wall

Non-blocking — Strengths

src/v4/extdeps/languages/python.dag The Python 3.14 standard type hierarchy source at https://docs.python.org/3.14/reference/datamodel.html#the-standard-type-hierarchy supports the no-width-splitting scalar decision.

ROADMAP — Verified

T-PB-B: The added hand-Rust test has a checkable deferral to ROADMAP.md §Nine lanes / pb_rust_tests_outside_residual_zero in the changed planning artifact.

✅ No blocking concerns in the current diff.

briansrls

Review metadata

Provider / model: codex / unknown
Commit: 958b4de6 · Trigger: schedule
Thinking: 129s wall

✅ The current diff is clean: the .dag coproduct has a ledgered terminal disposition and the new hand-Rust smoke test has the required P5 deferral receipt.

briansrls

Review metadata

Provider / model: codex / unknown
Commit: 55abb166 · Trigger: schedule
Thinking: 109s wall

✅ Mixed docs/.dag/test change looks clean; the PythonScalar coproduct has a terminal ledger, and the new hand-Rust test has the required P5 deferral receipt.

…wrapper → flat-nullary PythonScalar Operator A-vs-B ruling (2): #3309's Practice-8 dissolution of the hollow nested PythonNumericTower / Numeric-tower wrapper carries unique value over the shape #3338 retained. Forward-ported onto post-#3338 main. - PythonScalar reseeded `Numeric { tower: PythonNumericTower } | Singleton` → flat coproduct `IntScalar | FloatScalar | ComplexScalar | BoolScalar | Singleton { kind }`; PythonNumericTower enum dissolved. - #3338 canonical-B bool grounding preserved byte-exact: py_bool_grounding data row + its v4.std.logic / v4.std.algebra imports untouched. - Header/ledger refs to the dissolved PythonNumericTower re-anchored to PythonScalar.BoolScalar (python.dag L2/L3/L5/L40; DECISIONS.md). - SL-3309-PYTHON-SCALAR-RESEED ledger row added. Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>

…-nullary PythonScalar (#3309 fwd-port) (#3344) * WIP: v4 T-4 Phase-3: python.dag per-language fact-bundle rework — scalar towe * v4 T-4 Phase-3: dissolve python.dag PythonNumericTower/Numeric-tower wrapper → flat-nullary PythonScalar Operator A-vs-B ruling (2): #3309's Practice-8 dissolution of the hollow nested PythonNumericTower / Numeric-tower wrapper carries unique value over the shape #3338 retained. Forward-ported onto post-#3338 main. - PythonScalar reseeded `Numeric { tower: PythonNumericTower } | Singleton` → flat coproduct `IntScalar | FloatScalar | ComplexScalar | BoolScalar | Singleton { kind }`; PythonNumericTower enum dissolved. - #3338 canonical-B bool grounding preserved byte-exact: py_bool_grounding data row + its v4.std.logic / v4.std.algebra imports untouched. - Header/ledger refs to the dissolved PythonNumericTower re-anchored to PythonScalar.BoolScalar (python.dag L2/L3/L5/L40; DECISIONS.md). - SL-3309-PYTHON-SCALAR-RESEED ledger row added. Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com> * Reconcile dissolved-type ref in grounding-worked-examples.md PythonNumericTower.BoolLevel -> PythonScalar.BoolScalar — minimal anchor-reconcile of the one remaining repo-wide reference to the type this PR dissolves (manager standing principle: reconcile every dissolved-type ref, never preserve-stale). Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com> --------- Co-authored-by: Claude Opus 4.7 (1M context) <noreply@anthropic.com>

briansrls · 2026-05-19T02:13:41Z

Closing as superseded by #3344.

Per operator A-vs-B ruling (2), #3309's Practice-8 dissolution of the hollow PythonNumericTower / Numeric { tower } wrapper was forward-ported as a fresh PR off post-#3338 main: #3344 (MERGEABLE, structural-diff-verified PASS by the T-4 manager). #3344 carries the unique value forward and replaces this branch, which had wholesale conflicts against post-#3338 main (the bool slice here is superseded by #3338's canonical-B grounding).

No work is lost — the dissolution + SL-3309-PYTHON-SCALAR-RESEED ledger row live in #3344. Closed on operator instruction.

— sent from stern-eagle-698

…VS-B row by ID (#3346) Non-blocking ledger-traceability follow-up to merged #3344 (the #3309 reading-(2) forward-port). Adds a one-clause **Supersedes:** pointer to the SL-3309-PYTHON-SCALAR-RESEED row naming the dropped prior held-provisional SL-3309-PYTHON-PER-PRIMITIVE-A-VS-B row by ID, closed by operator A-vs-B ruling (2) + #3338 py_bool_grounding. By-ID only — no stale PyBool=Bool / open-gate text reintroduced (carrying that verbatim would itself be the stale closed-gate finding the dissolution removes). DECISIONS.md-only; no .dag change.

briansrls and others added 2 commits May 18, 2026 05:24

WIP: v4 T-4 Phase-3: python.dag per-language fact-bundle rework — sca…

c17d64a

…lar towe