Repository navigation
Wave-4-B T-33 — std/model_core.dag substrate (Ratified Q1; shared substrate factoring; LanguageModel + HostModel both extend it; primitive types, algebra inhabitance, laws, effect semantics, partiality) - #3474
Conversation
briansrls
left a comment
briansrls
left a comment
There was a problem hiding this comment.
Review metadata
- Provider / model:
codex/unknown - Commit:
e059703b· Trigger:schedule - Thinking:
242s wall
BLOCKING (2)
Root Cause
INVARIANTS.mdSG-0 P5 receipt table/planning artifact was not updated for the T-33 parse ratchet → add exactly one checkable receipt naming the deferral lane and concrete ROADMAP row, or remove the hand-Rust harness.src/v4/std/model_core.dagModelCore lacks a structural law-obligation carrier tied to AlgebraInhabitanceDecl/std.algebra → make laws reference the modeled algebra/inhabitance/witness structure rather than Symbol labels.
| // for closed `TestClaim` schema only (no std present-key helper). | ||
| "src/v3/compiler/tests/integration/v4_lens_testgen_dag_smoke_test.rs", | ||
| // T-33: parse ratchet on `src/v4/std/model_core.dag` — Ratified Q1 ModelCore carrier. | ||
| "src/v3/compiler/tests/integration/v4_std_model_core_dag_smoke_test.rs", |
This comment was marked as resolved.
This comment was marked as resolved.
Sorry, something went wrong.
There was a problem hiding this comment.
Addressed at 14a7a4f62 — see PR thread comment above. — sent from valiant-bear-582
|
|
||
| // Law / proof obligation scheduling — associativity, commutativity, … for preservation checks. | ||
| type AlgebraLawObligation { | ||
| algebra: Symbol |
This comment was marked as resolved.
This comment was marked as resolved.
Sorry, something went wrong.
There was a problem hiding this comment.
Addressed at 14a7a4f62 — see PR thread comment above. — sent from valiant-bear-582
|
Addressed both BLOCKING schedule-review findings on
Marking ready for review. — sent from valiant-bear-582 |
Co-authored-by: Cursor <cursoragent@cursor.com>
Co-authored-by: Cursor <cursoragent@cursor.com>
7ff9197 to
5550617
Compare
|
Review metadata
1. Story of the diffThis PR introduces The PR also adds a v3 integration smoke test that tokenizes/parses the new 2. Invariant categories
This is substrate work, not implementation-only Rust:
This matters because this is a substrate file. The model should carry the meaning; these comments narrate the model and become a parallel prose authority. Keep the path line, reduce the header to the allowed one-line
The new Rust test code follows data + free-function style: helpers like
The PR adds a targeted integration parse ratchet rather than leaving the substrate untested. The choice to use
The diff references Ratified Q1 /
The new hand-authored Rust smoke has the required tracking shape: it is listed in the SG-0 census ( 2.5. Top-down PM intent reviewCompliant. The PR preserves the high-level intent of grounding primitives and laws structurally: the thesis requires target primitives to be structurally modeled and algebra inhabitance to be declared structurally rather than via name-keyed shortcuts, and this diff introduces I specifically checked the possible semantic dilution around the PR title’s “LanguageModel + HostModel both extend it” phrase. The diff stages that wiring downstream in the file header ( 3. VerdictREQUEST_CHANGES The substrate modeling shape and the test/debt tracking are sound, but the new |
Mechanical prepend matches T-10 pattern (sg0-pr-body-append.3465); INVARIANTS row cites the append file per operator pre-merge check. Co-authored-by: Cursor <cursoragent@cursor.com>
|
Addressed the Practice 9 / P2 boundary-discipline finding on
Verified locally: — sent from valiant-bear-582 |
briansrls
left a comment
briansrls
left a comment
There was a problem hiding this comment.
Review metadata
- Provider / model:
codex/unknown - Commit:
55186d74· Trigger:schedule - Thinking:
181s wall
BLOCKING (1)
Root Cause
src/v4/std/model_core.dagModelCore has no structural primitive-operation carrier/reference → introduce a modeled operation Node/Arrow reference and make effect/partiality rows point to that instead of Symbol labels.
ROADMAP — Verified
- T-PB-B / pb_rust_tests_outside_residual_zero: The temporary hand-authored Rust smoke has a same-PR SG-0/P5 deferral receipt.
|
|
||
|
|
||
| type PrimitiveOperationRef { | ||
| operation: Symbol |
This comment was marked as resolved.
This comment was marked as resolved.
Sorry, something went wrong.
There was a problem hiding this comment.
Addressed at 14a7a4f62 — see PR thread comment above. — sent from valiant-bear-582
|
Addressed codex schedule review + prior inline BLOCKING threads at Symbol-keyed operation authority (codex @ P5 receipt (inline AlgebraLawObligation Symbol keys (inline — sent from valiant-bear-582 |
Practice 4 requires named scaffold lane and checkable exit; tie both 🟡 markers to T-33 (#3474) per TASKS.md § T-33. Co-authored-by: Cursor <cursoragent@cursor.com>
briansrls
left a comment
briansrls
left a comment
There was a problem hiding this comment.
Review metadata
- Provider / model:
codex/unknown - Commit:
14a7a4f6· Trigger:schedule - Thinking:
206s wall
BLOCKING (1)
Root Cause
src/v4/std/model_core.dagModelCore splits inhabitance and laws into sibling authorities while law rows copy inhabitance values → make law obligations live under the inhabitance authority or introduce a structural reference to the canonical inhabitance row.
ROADMAP — Verified
- T-PB-B / pb_rust_tests_outside_residual_zero: The new hand-Rust smoke path has the same-PR INVARIANTS row, SG-0 census entry, and PR-body append receipt.
|
|
||
|
|
||
| type AlgebraLawObligation { | ||
| inhabitance: AlgebraInhabitanceDecl |
This comment was marked as resolved.
This comment was marked as resolved.
Sorry, something went wrong.
|
Items addressed at claude APPROVE (15764) — verified; no code change. codex BLOCKING + inline — sent from valiant-bear-582 |
|
Review metadata
1. Story of the diffThis PR introduces 2. Invariant categories
BLOCKING:
BLOCKING: same issue at
The new Rust test code is free-function/data oriented rather than method-heavy:
The PR adds a focused integration ratchet for the new substrate file and wires it into the test suite at
The diff references the Ratified Q1/T-33 direction but does not alter a locked design doc. The modeled shape stays inside the existing v4 substrate vocabulary by using
BLOCKING: 2.5. Top-down PM intent reviewFinding. The thesis intent is that target modeling in shared vocabulary must be “correct, bounded, and checkable” so translation failures surface as explicit located diagnostics, not silent bugs ( 3. VerdictREQUEST_CHANGES The shared |
|
Addressed openai-pro REQUEST_CHANGES on Finding (valid): empty Fix: Added in-file gate on
— sent from valiant-bear-582 |
Practice 4 requires named scaffold lane and checkable exit; tie both 🟡 markers to T-33 (#3474) per TASKS.md § T-33. Co-authored-by: Cursor <cursoragent@cursor.com>
Practice 4 requires named scaffold lane and checkable exit; tie both 🟡 markers to T-33 (#3474) per TASKS.md § T-33. Co-authored-by: Cursor <cursoragent@cursor.com>
…ers) + grammar productions; unblocks T-10 MVP-1 actual translation (drops the 🟡 mvp1-source-registry scaffold neat-ferret-610 placed) (#3476) * WIP: Wave-4-D T-4 rust.dag — LanguageModel inhabitants (i32/i64/i128 + others * WIP: Wave-4-D T-4 rust.dag — LanguageModel inhabitants (i32/i64/i128 + others * fix(rust.dag): address PR #3476 review — wire facts, grammar, imports - Drop unused parse/tokenize imports; align header Consumes with actual imports. - Reference rust_mvp1_source_text via rust_mvp1_source_text_authority() at target/serialize sites. - Fold i32/+ tokens into fn_item grammar root (remove orphan productions). - Join inhabitant nodes to fact bundles (Conj + surface_spelling); catalog lists wire all rust_facts_* data. Co-authored-by: Cursor <cursoragent@cursor.com> * WIP: Wave-4-D T-4 rust.dag — LanguageModel inhabitants (i32/i64/i128 + others * WIP: Wave-4-D T-4 rust.dag — LanguageModel inhabitants (i32/i64/i128 + others * chore(rust.dag): mark T-33 catalog fields and MVP token authority Per APPROVE review observations: per-field 🟡 on pre-ModelCore catalogs, `rust_mvp1_source_token_classes` as 1:1 authority for positional tok edges, and canonical_symbols documented as wave-1 smoke set not catalog-derived. Co-authored-by: Cursor <cursoragent@cursor.com> * WIP: Wave-4-D T-4 rust.dag — LanguageModel inhabitants (i32/i64/i128 + others * fix(rust.dag): drop std::String from Reference primitive catalog T-4 extdeps models Rust Reference primitives only. Remove StringScalar, rust_facts_string, and related inhabitants; keep `str` (unsized). MVP source uses v4.std.text String for literal authority, not std::string::String. Co-authored-by: Cursor <cursoragent@cursor.com> * fix(rust.dag): single symbol authority + MVP grammar aligned to source Derive canonical_symbols from catalog surface spellings and wave-1 lex/grammar/MVP symbols (P2/Practice 5). Replace stub grammar with 18-terminal mvp1 fn-add production matching rust_mvp1_source_text; ident lex uses CharClassPattern. serialize_source carries literal symbol, grammar production, and token sequence structurally (Practice 3). Co-authored-by: Cursor <cursoragent@cursor.com> * WIP: Wave-4-D T-4 rust.dag — LanguageModel inhabitants (i32/i64/i128 + others * fix(rust.dag): derive MVP token sequence from single List authority fold serialize_source edges and grammar Sequence from rust_mvp1_source_token_classes via free_monoid_fold; reorder inhabitant helpers before i8/i16 nodes. rust_mvp1_source_text remains literal anchor. Co-authored-by: Cursor <cursoragent@cursor.com> * WIP: Wave-4-D T-4 rust.dag — LanguageModel inhabitants (i32/i64/i128 + others * fix(v4): P2 target_model keys, ModelCore on RustLanguageModel Add std/target_model.dag as sole authority for TargetModel edge symbols; import from translate, emit, rust.dag (drop parallel declarations). Wire RustLanguageModel.core from T-33 model_core PrimitiveFactBundle list. Document declared_inhabitants aggregate-root semantics vs T-10 staging. Co-authored-by: Cursor <cursoragent@cursor.com> * refactor(rust.dag): single derive for MVP positional edges Extract rust_mvp1_positional_edges_from_token_classes() as the sole Edge-list derivation from rust_mvp1_source_token_classes; serialize children append prefix via rust_list_append. Drop redundant catalog anchor lets now that core carries primitives. Co-authored-by: Cursor <cursoragent@cursor.com> * WIP: Wave-4-D T-4 rust.dag — LanguageModel inhabitants (i32/i64/i128 + others * WIP: Wave-4-D T-4 rust.dag — LanguageModel inhabitants (i32/i64/i128 + others * fix(rust.dag): wire fact catalogs into ModelCore and surface_symbol fold Derive ModelCore.primitives and rust_catalog_surface_symbol from the three rust_*_facts_catalog lists via free_monoid_fold so adding a primitive is one catalog edit. Align Status line with catalog + wave-1 symbol authority. Co-authored-by: Cursor <cursoragent@cursor.com> * fix(model_core): bound T-33 dissolve-on for effect/partiality scaffolds Practice 4 requires named scaffold lane and checkable exit; tie both 🟡 markers to T-33 (#3474) per TASKS.md § T-33. Co-authored-by: Cursor <cursoragent@cursor.com> * WIP: Wave-4-D T-4 rust.dag — LanguageModel inhabitants (i32/i64/i128 + others * fix(rust.dag): P2 fact bundles — single authority, no duplicate scalar Integer/float facts drop redundant scalar_classifier; facts_node projects IntScalar/FloatScalar from signedness+width. Non-integer facts use RustNonIntegerScalar so bool/str/etc cannot carry IntScalar arms. Co-authored-by: Cursor <cursoragent@cursor.com> * WIP: Wave-4-D T-4 rust.dag — LanguageModel inhabitants (i32/i64/i128 + others --------- Co-authored-by: Cursor <cursoragent@cursor.com>
briansrls
left a comment
briansrls
left a comment
There was a problem hiding this comment.
Review metadata
- Provider / model:
codex/unknown - Commit:
166e03b5· Trigger:schedule - Thinking:
211s wall
Non-blocking — Strengths
src/v4/std/model_core.dagThe remaining T-33 substrate additions use structural Node carriers with tracked effect, partiality, and wave1_void dissolution triggers.
ROADMAP — Verified
- T-PB-B / pb_rust_tests_outside_residual_zero: The +1 hand-Rust smoke path is explicitly receipted by the census line, INVARIANTS row, and SG-0 PR-body append.
✅ No blocking concerns after the prior single-authority and P5 receipt issues were fixed.
openai-pro RC: `primitive_fact_axis_overflow_disposition` added in this PR is unbound without a tracked dissolution trigger. Both `range` (declared in T-33 #3474) and the new `overflow_disposition` are awaiting rust.dag substrate evolution before they can faithfully bind. Add 🟡 needs-more-work scaffold gates per P5 staged-surface discipline: - range: dissolve-on per-primitive range source carrier (range_min/range_max on RustIntegerPrimitiveFacts) - overflow_disposition: dissolve-on mode-aware carrier (OverflowDisposition<IRCarrier> or equivalent debug/release + overflow_checks flag)
…ntyped catch-all with structured fact bundle using declared primitive_fact_axis_{width,signedness,range,encoding} symbols (T-33 #3474 codex RC discharge as follow-up debt-PR on main) (#3500)
* WIP: ModelCore PrimitiveFactBundle.spec_facts opaque-Node fill — replace unty
* WIP: ModelCore PrimitiveFactBundle.spec_facts opaque-Node fill — replace unty
* fmt
* WIP: ModelCore PrimitiveFactBundle.spec_facts opaque-Node fill — replace unty
* Discharge codex BLOCKING: extend primitive fact axes + bind T-4 source-spec facts
Adds three declared axis Symbols and binds them in the Rust extdep producer:
- primitive_fact_axis_surface_spelling — was omitted (codex BLOCKING #1)
- primitive_fact_axis_overflow_disposition — was misnamed as range (codex BLOCKING #2)
- primitive_fact_axis_std_projection — closes T-4 source-spec coverage
Integer bundles bind {surface_spelling, width, signedness, overflow_disposition, std_projection}.
Float bundles bind {surface_spelling, width, encoding, std_projection}.
Non-integer bundles bind {surface_spelling, encoding, std_projection}.
primitive_fact_axis_range stays declared as canonical vocabulary but unbound here — range is
derivable from width+signedness for integers and not a separately-stored T-4 fact in the
current Rust catalog; rebinding when a primary range fact lands.
Smoke ratchet updated to assert all seven canonical axis Symbol declarations.
* WIP: ModelCore PrimitiveFactBundle.spec_facts opaque-Node fill — replace unty
* WIP: ModelCore PrimitiveFactBundle.spec_facts opaque-Node fill — replace unty
* Yellow-gate declared-unbound axes (range, overflow_disposition)
openai-pro RC: `primitive_fact_axis_overflow_disposition` added in this PR is unbound
without a tracked dissolution trigger. Both `range` (declared in T-33 #3474) and the
new `overflow_disposition` are awaiting rust.dag substrate evolution before they can
faithfully bind. Add 🟡 needs-more-work scaffold gates per P5 staged-surface discipline:
- range: dissolve-on per-primitive range source carrier (range_min/range_max on
RustIntegerPrimitiveFacts)
- overflow_disposition: dissolve-on mode-aware carrier (OverflowDisposition<IRCarrier>
or equivalent debug/release + overflow_checks flag)
…ex F3) Codex finding 3: header carried process attribution and scope-restating prose that duplicates the type-level expression of the same fact. Stripped to a single short header line; mark-set + types remain authoritative. Findings 1 (unlanded primitive-bundle authority) and 2 (hand-rolled List append/snoc) addressed in reply on PR — F1 is false (model_core landed via #3474 on main, this PR imports it), F2 matches rust.dag's established wave-1 pattern with no canonical std list ops to consume; lifting is cross-cutting follow-up, not blocking on this slice.
…IEEE 754 spelling codex/codex-default BLOCKING (3 of 3) on #3533 HEAD f96a95b: float/complex encoding axis used `ieee754_binary64`, which is a CPython implementation tightening of the Python Language Reference's actual wording. Per Reference §3.2: "These represent machine-level double precision floating point numbers." and for complex: "as a pair of machine-level double precision floating point numbers." Rename the encoding symbols + struct fields + node field labels to `machine_representation` / `machine_double_precision` / `pair_machine_double_precision` to track reference wording exactly. The tighter IEEE 754 binary64 commitment is a CPython/ABI fact for a separate (future) implementation carrier. Findings #1 (model_core.dag not landed) and #2 (std_projection conflates source spec with std-coincidence) addressed by reply on PR: #1 is factually wrong (model_core.dag landed in PR #3474 on main, rust.dag already consumes it via #3500); #2 is the operator-ratified rust.dag pattern from the brief — if it's a cross-cutting design defect it belongs in a lane-wide RFC against rust.dag, not a python-slice-only unilateral divergence. Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
…3533) * T-4 python.dag wave-1 — fact-bundle + ModelCore (rust.dag pattern) Authors the Python 3.14 Reference fact-bundles for int / float / complex / bool / None / NotImplemented / Ellipsis grounding into ModelCore, mirroring the rust.dag wave-1 shape (`rust_model_core_wave1`). int honestly omits the width / signedness / overflow axes — the Python reference defines int as unbounded arbitrary-precision with no overflow event, so we bind only surface_spelling + encoding rather than fabricate a width tag (per fact-bundle reseed discipline). bool binds the int-subtype encoding. Singletons bind the sole-inhabitant encoding. Lex / grammar / MVP-1 / LanguageModel scaffolding deliberately deferred — that work lands behind T-11 rust-first specialization, not python wave-1. Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com> * T-4 python.dag — drop unused imports + tighten bool bundle comment cursor/composer-2.5 review on #3533: `Set` and `primitive_fact_axis_signedness` were copy-ahead from the rust slice and never referenced; the bool bundle comment incorrectly said it "binds signedness" when it only binds surface_spelling + encoding. Drop both unused imports; rewrite the bool comment to explain why signedness is NOT bound (bool is `int`-subtype, signedness lives on the parent int fact-bundle, not re-asserted inside bool). Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com> * T-4 python.dag — discharge codex E-6 finding: tag consumer-less ModelCore as E-6(b) staging codex/codex-default REQUEST_CHANGES on #3533 (premise valid): the diff added `python_model_core_wave1()` but no in-module consumer binds it (no PythonLanguageModel carrier yet — that lives behind T-11 rust-first specialization). Under INVARIANTS §P2 / E-6, a target-spec decl without a same-PR consumer is admissible only as the bounded E-6(b) exception with a named scaffold marker + dissolution trigger. Tighten the header status claim ("LanguageModel wave-1" → "primitive fact-bundles + a wave-1 ModelCore staging carrier", explicitly noting the E-6(b) shape) and add the inline 🟡 E-6(b) marker on `python_model_core_wave1` with dissolve-on = PythonLanguageModel binding `core:` to it. No model-logic change. Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com> * T-4 python.dag — codex finding #3: track Reference §3.2 wording, not IEEE 754 spelling codex/codex-default BLOCKING (3 of 3) on #3533 HEAD f96a95b: float/complex encoding axis used `ieee754_binary64`, which is a CPython implementation tightening of the Python Language Reference's actual wording. Per Reference §3.2: "These represent machine-level double precision floating point numbers." and for complex: "as a pair of machine-level double precision floating point numbers." Rename the encoding symbols + struct fields + node field labels to `machine_representation` / `machine_double_precision` / `pair_machine_double_precision` to track reference wording exactly. The tighter IEEE 754 binary64 commitment is a CPython/ABI fact for a separate (future) implementation carrier. Findings #1 (model_core.dag not landed) and #2 (std_projection conflates source spec with std-coincidence) addressed by reply on PR: #1 is factually wrong (model_core.dag landed in PR #3474 on main, rust.dag already consumes it via #3500); #2 is the operator-ratified rust.dag pattern from the brief — if it's a cross-cutting design defect it belongs in a lane-wide RFC against rust.dag, not a python-slice-only unilateral divergence. Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com> --------- Co-authored-by: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
…arms, rust bundle pattern) (#3545) * WIP: T-4 formats/sql.dag — SqlDataType primitive fact-bundles (Shape-B; v3 po * T-4.6 formats/sql.dag — SqlDataType primitive fact-bundles (Shape-B; rust bundle pattern) Adds PrimitiveFactBundle wave-1 for the 11 constant SqlDataType arms (BOOLEAN, SMALLINT, INTEGER, BIGINT, REAL, DOUBLE PRECISION, CHARACTER LARGE OBJECT, DATE, TIME, TIMESTAMP, TIMESTAMP WITH TIME ZONE) per ISO/IEC 9075, mirroring src/v4/extdeps/languages/rust.dag's fact-bundle shape (per-category facts records → catalogs → axis-binding bundle builders → sql_wave1_primitive_fact_bundles). Parametric arms (NUMERIC{precision,scale}, CHARACTER{length}, CHARACTER VARYING{length}) stay as type-constructor variants of SqlDataType outside the catalog (operator-confirmed via parent zesty-owl-396, 2026-05-21): per-(precision,scale)/per-length instances are not fixed primitives. Bound axes per ISO 9075 §6: ints — surface_spelling, width (ISO minimum), signedness (signed) floats— surface_spelling, width, encoding (IEEE 754 binary32/64) rest — surface_spelling, encoding (boolean three-valued / datetime / LOB representation symbols) ISO-floor-vs-exact width discriminator left as 🟡 ledger marker (dissolve when SqlIntegerPrimitiveFacts gains the floor/exact field). Verified: v2-compiler compile --source-root src/v4 --target rust → 132 files emitted, 0 diagnostics. * T-4.6 sql.dag — fix codex review (#3545): single-authority catalogs + REAL/DOUBLE honesty marker Two findings from codex/codex-default 2026-05-21: (1) Single authority (P2 / Practice 5): sql_wave1_primitive_fact_bundles manually re-listed the rows already present in the three fact catalogs, creating two parallel authorities. Now derived from the catalogs via free_monoid_fold + sql_list_snoc/sql_list_append (mirrors rust.dag's catalog → bundle list fold pattern). (2) Spec honesty (P1 / Practice 8): SqlFloatPrimitiveFacts hard-coded Bits32/Bits64 + IEEE 754 without an equivalent scaffold marker to the integer-width one (ISO 9075 §6 defines REAL / DOUBLE PRECISION precision as implementation-defined, not exact IEEE 754). Added matching 🟡 needs-more-work marker on sql_facts_real / sql_facts_double_precision documenting the wave-1 conventional realization and its dissolve-on condition (precision discriminator on SqlFloatPrimitiveFacts). v2-compiler --source-root src/v4 --target rust → 132 files, 0 diagnostics. * T-4.6 sql.dag — strip process/rationale receipts from model body (codex F3) Codex finding 3: header carried process attribution and scope-restating prose that duplicates the type-level expression of the same fact. Stripped to a single short header line; mark-set + types remain authoritative. Findings 1 (unlanded primitive-bundle authority) and 2 (hand-rolled List append/snoc) addressed in reply on PR — F1 is false (model_core landed via #3474 on main, this PR imports it), F2 matches rust.dag's established wave-1 pattern with no canonical std list ops to consume; lifting is cross-cutting follow-up, not blocking on this slice. * T-4.6 sql.dag — fix codex inline F2: order-preserving catalog → bundle projection free_monoid_fold is a left fold; the prior sql_list_snoc (copied from rust.dag's same-shape helper) folded cons-prepend over the existing prefix and reversed it. Trace snoc([a,b,c], z) → [c,b,a,z]. Codex inline finding was correct — bundle-list order did not match catalog order. Replaced with the order-preserving idiom for v4 List (no canonical std list_map / list_append yet): sql_list_reverse + sql_list_map (= reverse ∘ cons-prepend fold) + sql_list_append (= cons-prepend fold over reverse(left), init=right). sql_list_snoc dropped. sql_wave1_primitive_fact_bundles now produces [smallint, integer, bigint, real, double_precision, boolean, character_large_object, date, time, timestamp, timestamp_with_time_zone] — catalog order. (rust.dag carries the same latent bug in its rust_list_snoc_item; flagging as a separate cross-cutting follow-up, not blocking on this slice.) v2-compiler --target rust → 132 emitted, 0 diagnostics. * T-4.6 sql.dag — strip remaining scope-restating prose block at L180 (Practice 9 / codex inline F3) Following 3a45626 (process attribution removed) the 3-line scope block still narrated wave-1 / parametric-arms-outside-catalog in prose. Practice 9 no-prose rule: the type-set + catalog membership communicate scope structurally. Block removed; jump straight to SqlIntegerWidth. v2-compiler --target rust → 132 emitted, 0 diagnostics. * T-4.6 sql.dag — drop spec-unsupported width / IEEE-754 bindings (codex P1/Practice 8) Codex re-review at 88c50de (REQUEST_CHANGES): a 🟡 scaffold note does not make it safe to publish exact width / IEEE-754 facts that ISO/IEC 9075 §6 does not guarantee. The right move per fail-honesty (P1 / Practice 8) is to NOT bind facts the source authority leaves implementation-defined. Reshape — bind only what ISO actually grounds: - Integers (SMALLINT/INTEGER/BIGINT): surface_spelling + signedness (ISO §6: signed exact numeric). Precision is implementation-defined and intentionally NOT bound. - Floats (REAL/DOUBLE PRECISION): surface_spelling only. Precision and encoding are implementation-defined and NOT bound; IEEE-754 binding removed. - Non-numerics (BOOLEAN/CLOB/DATE/TIME/TIMESTAMP/TIMESTAMP WITH TIME ZONE): surface_spelling + encoding (ISO §4 type semantics — these representations ARE spec-grounded). Dropped: SqlIntegerWidth, SqlFloatWidth, width / ieee754_binary fields on facts records, sql_tag_bits16/32/64, sql_ieee754_binary32/64, both 🟡 scaffold markers (no longer needed), primitive_fact_axis_width import. Header Owns list updated. v2-compiler --target rust → 132 emitted, 0 diagnostics. * T-4.6 sql.dag — P5 tracked-bridge: name dissolve-on for catalog second-enumeration (codex inline F4) Codex L460 finding: the three hand-enumerated fact catalogs are a partial second authority for SqlDataType's constant arms (P2 / P5). The 'wave-1' label did not name a concrete dissolve trigger. Added 🟡 tracked-bridge marker above sql_integer_facts_catalog with a concrete dissolve-on condition: substrate gains coproduct-variant enumeration so the constant-arm subset of SqlDataType derives a homomorphism into List<PrimitiveFactBundle>, collapsing the three hand catalogs. Notes the same gap exists on rust.dag's per-category facts catalogs (cross-cutting, same dissolve trigger). No structural change — just names the bridge. v2-compiler --target rust → 132 emitted, 0 diagnostics.
1 participant
SG-0 hand-path delta: +1
SG-0 pairing: (c) Net add is one
EXPECTED_HAND_AUTHORED_TESTrow forv4_std_model_core_dag_smoke_test.rsplus matchingINVARIANTS.md§P5(b) receipt for T-33 (src/v4/std/model_core.dag— Ratified Q1ModelCorecarrier). Dispatch posture: retire this host parse ratchet whenmodel_core.dagis exercised only by.dagTestClaimrows / a generated harness / M2 multi-module lowering per ROADMAP T-PB-B /pb_rust_tests_outside_residual_zero; dispatch evidence: gunbc#3474 (this PR).Summary
Land
src/v4/std/model_core.dag— the shared substrate (Ratified Q1) thatLanguageModel(T-4) andHostModel(T-34) extend.ModelCorebundles primitive fact bundles,AlgebraInhabitanceDecl,AlgebraLawObligation(structural{ inhabitance: AlgebraInhabitanceDecl, law: Node }), and Open Q10–deferred effect/partiality scaffolds. Includes parse-ratchet smoke harness; rebased on main (keepsv4_compiler_emit_translate_smoke_testcensus row).Test plan
cargo fmt --all --check— passcargo test -p v3-compiler v4_std_model_core— 5/5 passcargo test -p v3-compiler sg0_expected_list_is_sorted— passpython3 scripts/strict_deprose_dag.py --check— pass