feat(rust): #259 slice 2 — own-lowered, typed Layer 2 replay/emitter - #300
Merged
Conversation
The typed Rust half of the frozen Layer 2 contract (LOWERED_VERSION 1): a strict data model of the normalized lowered representation plus the canonical emitter, replaying the Python-authored goldens byte-for-byte. No lowering, no OwnIR validation, no MOS, no analysis wiring — this slice proves Rust can carry and emit the exact surface before anything derives it. * crates/own-lowered — a LEAF data crate (own-diagnostics precedent; the DAG lock in own-diagnostics/tests/dag.rs is widened deliberately with the leaf entry and the rationale: the future own-bridge will CONSTRUCT these types, never the reverse). Model mirrors ownlang/lowered.py's frozen normalization decisions field-for-field: declaration order = canonical JSON order; always-written nullable fields are Option WITHOUT skip; the handle-entry allowlist keys are Option WITH skip, `handle` first; the closed Stmt vocabulary under the `stmt` tag; Rejected carries the fail-loud error text. Every shape is deny_unknown_fields — a Python-side surface change cannot slip past the typed replay. * tests/replay.rs — reads manifest.json (typed, strict), requires lowered_version == LOWERED_VERSION, both fixture halves on disk for every case, then for each rust_replay: true case parses the golden and asserts the canonical re-emit is byte-identical (>= 25 shared cases). The rust_replay: false set is asserted to be exactly the OD-2/#294 snapshot — Python-only by ledger decision, not silently skipped. serde_json's pretty printer reproduces Python json.dumps(indent=2, ensure_ascii=False) exactly on this surface — proven by the byte-equality suite, not assumed. cargo fmt/clippy clean, full workspace tests green (incl. the DAG fitness test), Python suite untouched and green. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01MdYFKUaygHz1T9H1qJ7BqK
📝 WalkthroughWalkthroughAdds the ChangesLowered surface parity
Estimated code review effort: 3 (Moderate) | ~25 minutes Sequence Diagram(s)sequenceDiagram
participant LoweredGolden
participant own_lowered
participant CanonicalEmitter
participant ReplayTest
LoweredGolden->>own_lowered: parse_document(golden JSON)
own_lowered->>CanonicalEmitter: typed Surface
CanonicalEmitter->>ReplayTest: canonical JSON
ReplayTest->>LoweredGolden: compare byte-for-byte
🚥 Pre-merge checks | ✅ 4✅ Passed checks (4 passed)
✨ Finishing Touches📝 Generate docstrings
🧪 Generate unit tests (beta)
Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out. Comment |
…uality The typed model collapses a MISSING handle key and an explicit null into one state, so a valid Python golden carrying the schema-nullable trio (type/source/source_type as null) cannot round-trip byte-exactly. Pinned by a new shared fixture (handles_null_metadata, rust_replay: true) whose first handle carries all three explicit nulls and whose second carries none — the Python emitter distinguishes them by key membership. Also pinned red: * lowered_version parsed but never enforced — version 99 documents (accepted and rejected surfaces alike) currently parse fine; * Param.type is nullable in Rust while Python's AST declares TypeRef, never TypeRef | None — "type": null must be rejected; * a non-nullable optional handle key with explicit null (released: null) is silently decayed to missing instead of rejected; * the replay suite outsourced ledger integrity to Python — it now asserts unique(manifest names) == facts files == golden files itself (green from birth; enforcement, not a bug pin). Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01MdYFKUaygHz1T9H1qJ7BqK
….type * Maybe<T> (Missing / Null / Value) for the schema-nullable handle trio (type, source, source_type): explicit JSON null now survives a parse->emit round trip instead of collapsing into "missing" — the Python emitter distinguishes the two by key membership, so the typed model must too. * Every other optional handle key rejects a present null (deserialize_with = "present") rather than decaying it to absent and silently deleting it on re-emit. * Param.type is non-nullable (TypeShape, not Option): Python's AST declares TypeRef, never TypeRef | None, so "type": null is a shape the emitter cannot produce and the model no longer accepts. * parse_document enforces lowered_version == LOWERED_VERSION on BOTH surfaces (accepted and rejected) — the lockstep promise is now a parse error, not something only the manifest check notices. All 26 rust_replay cases (incl. the new handles_null_metadata fixture) replay byte-identically; ledger equality (unique manifest names == facts files == golden files) is enforced Rust-side. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01MdYFKUaygHz1T9H1qJ7BqK
Owner
Author
|
Ревизия по review (REQUEST CHANGES) — применена, head Пакет ревизии, по пунктам:
Red-состояние зафиксировано: 6 падающих тестов (byte-mismatch replay на null-fixture; v99 принимался ×2; Локальные гейты: python suite (27 lowered cases in sync), ruff — зелёные; cargo fmt / clippy (0 warnings) / полный workspace test — зелёные; повторный Nonblocking ( PR остаётся draft — жду повторного review. Не мержить, не self-accept. Generated by Claude Code |
Owner
Author
|
Independent review verdict (head
Merge authorized by the reviewer on exact head Post-merge, #259 slice 2 is ACCEPTED / MERGED / FROZEN: Generated by Claude Code |
PhysShell
marked this pull request as ready for review
July 18, 2026 22:55
![coderabbitai[bot]](https://avatars.githubusercontent.com/in/347564?s=60&v=4){kind=small}
There was a problem hiding this comment.
Actionable comments posted: 1
🤖 Prompt for all review comments with AI agents
Verify each finding against current code. Fix only still-valid issues, skip the
rest with a brief reason, keep changes minimal, and validate.
Inline comments:
In `@rust/crates/own-lowered/tests/replay.rs`:
- Around line 111-114: Update the replay count assertion in the replay test to
require at least 26 shared cases instead of 25, while preserving the existing
failure message and replayed-count interpolation.
🪄 Autofix (Beta)
Fix all unresolved CodeRabbit comments on this PR:
- Push a commit to this branch (recommended)
- Create a new PR with the fixes
ℹ️ Review info
⚙️ Run configuration
Configuration used: Path: .coderabbit.yaml
Review profile: CHILL
Plan: Pro Plus
Run ID: 615c2702-36bb-408d-a142-47e1f9e09ced
⛔ Files ignored due to path filters (1)
rust/Cargo.lockis excluded by!**/*.lock
📒 Files selected for processing (10)
rust/Cargo.tomlrust/crates/own-diagnostics/tests/dag.rsrust/crates/own-lowered/Cargo.tomlrust/crates/own-lowered/src/lib.rsrust/crates/own-lowered/src/model.rsrust/crates/own-lowered/tests/replay.rsrust/crates/own-lowered/tests/strictness.rstests/fixtures/lowered/handles_null_metadata.facts.jsontests/fixtures/lowered/handles_null_metadata.golden.jsontests/fixtures/lowered/manifest.json
![chatgpt-codex-connector[bot]](https://avatars.githubusercontent.com/in/1144995?s=60&v=4){kind=small}
There was a problem hiding this comment.
💡 Codex Review
Here are some automated review suggestions for this pull request.
Reviewed commit: 159f45cc93
ℹ️ About Codex in GitHub
Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you
- Open a pull request for review
- Mark a draft as ready
- Comment "@codex review".
If Codex has suggestions, it will comment; otherwise it will react with 👍.
Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".
13 tasks
PhysShell
pushed a commit
that referenced
this pull request
Jul 19, 2026
The normative documents described the pre-slice-3 world. Recorded now: * spec/Bridge.md §6 Layer 2: #299 (Python emitter + fixture foundation), #300 (typed own-lowered surface/emitter, presence-aware metadata, per-document version gate), #301 (own-bridge constructs Layer 2 from facts and reproduces all 26 shared goldens byte-exact; golden is expected output only); tolerant_unknown_kind stays Python-only under #294 (the Rust bridge fails loud on an unknown kind instead of adopting the tolerant fallback); Layer 1, Layer 3, analysis wiring and #259 as a whole remain open. * spec/BridgeBehaviorMatrix.md: the Layer 2 summary paragraph updated to the same completed state. * tests/test_lowered_fixtures.py docstring: future tense ("will replay", "until the Rust emitter exists") moved to the present. * BR-L7 wording nit folded in (recorded on #297): condition (4) guards the post-merge reference/discharge, not only a release — the use-only hoist fixture is in the gate too. Docs-only; no Rust code, facts, manifest or golden changes. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01MdYFKUaygHz1T9H1qJ7BqK
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
2 participants
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Что и зачем
Второй slice #259 в границе, зафиксированной при merge #299: typed Rust replay/emitter только для
rust_replay: truefixtures — без lowering, без OwnIR validation, без MOS, без analysis wiring и без решения #294 попутно. Новый leaf data crateown-lowered(прецедент —own-diagnostics): строгая (deny_unknown_fieldsна каждом shape) типизированная модель замороженной Layer 2 поверхности (LOWERED_VERSION = 1), зеркалящая нормализационные решенияownlang/lowered.pyполе-в-поле (declaration order = canonical JSON order; always-written nullable поля —Optionбез skip; handle-allowlist — skip-при-отсутствии,handleпервым: schema-nullable тройкаtype/source/source_type—Maybe<T>(Missing / explicit Null / Value — членство ключа и явныйnullразличимы и переживают round-trip), остальные optional-ключи отвергают presentnullчерез presence-deserializer вместо тихого удаления;Param.typenon-nullable; закрытыйStmt-словарь подstmt-тегом;Rejectedс fail-loud текстом), плюс канонический emitter.parse_documentтребуетlowered_version == LOWERED_VERSIONна каждом документе (accepted и rejected).tests/replay.rsчитает typed manifest, сам перечисляет fixture-директорию и требует ledger-равенства (unique(manifest names) == *.facts.json == *.golden.json), и для каждого из 26 shared кейсов доказывает: golden парсится в типизированную модель и re-emit байт-в-байт равен Python-golden.rust_replay: falseнабор утверждается равным ровно{tolerant_unknown_kind}(OD-2/#294 снапшот) — пропуск по решению ledger, не молча. DAG-lock (own-diagnostics/tests/dag.rs) расширен сознательно:own-lowered— leaf; будущий own-bridge будет конструировать эти типы, не наоборот.Факт, который доказан тестом, а не предположен:
serde_jsonpretty printer воспроизводит Pythonjson.dumps(indent=2, ensure_ascii=False)байт-в-байт на всей поверхности (включая raw-UTF-8 em-dash в error-тексте).Ревизия по review (red
8c41e22→ green159f45c): presence-awareMaybe<T>+ null-отвергающие optional-ключи; новый shared fixturehandles_null_metadata(explicit nulls vs отсутствие ключей, авторизован Python--write); non-nullableParam.type+ негативный тест; per-document version gate + негативные тесты v99; Rust-side ledger equality.Тип изменения
Как проверено
cargo test— весь workspace зелёный, включая новый replay (26 shared кейсов byte-exact), strictness-негативы и DAG fitness-тест с новым leafcargo fmt --check,cargo clippy --all-targets— 0 warningspython tests/run_tests.py,ruff,mypy— зелёные; Python verdict-поведение не менялось (diff вrust/+ новая fixture-пара/manifest + одна запись в dag.rs); повторный--write— no-opdeny_unknown_fieldsвезде — Python-side изменение поверхности ломает replay, а не проскальзываетСвязанные issue
Refs #259 (slice 2 — НЕ закрывает issue), #258, #294 (не решается — снапшот остаётся Python-only), #250.
Чеклист
Независимый review: APPROVED на head
159f45c(после REQUEST CHANGES → ревизии); merge авторизован reviewer'ом.🤖 Generated with Claude Code
https://claude.ai/code/session_01MdYFKUaygHz1T9H1qJ7BqK
Generated by Claude Code
Summary by CodeRabbit
New Features
Bug Fixes
Tests