Lane 2 Stage 2a follow-up: collapse DerivedOpEffect into OperationEffect

[briansrls]

Summary

• Delete DerivedOpEffect; derive_op_effect(...) now returns OperationEffect? directly. The method and path_template fields were never read downstream — modifier falsification and obligation generation both project through shape alone, and ReadEffect already encodes "method was GET/HEAD/OPTIONS" structurally.
• check_modifier_vs_derivation and generate_idempotency_obligations now take OperationEffect — bodies unchanged (they only ever read .operation_name and .shape).
• Clears half the §Lane 2 Stage 2a / Track 17a boundary deferral in src/v3/ROADMAP.md; ComposedEffect reshape remains the one open gate before Stage 2b. docs/lane2-compile-time-proofs.md Stage 2b pre-start gate updated accordingly.

Scope is v3-only: dsl/std/effects.dag + src/v2/effect_derivation.dag untouched. v2 is retiring and ModifierCheck has already diverged structurally from v3 (single agreement vs two-axis ModifierAxisCheck), so the "keep in sync" note in the v3 file header is aspirational for this shape.

Test plan

• cargo test -p v3-compiler --test lane2_stage_2a_effects_smoke — 4/4 pass (smoke test asserts function-name presence, not arg types; DerivedOpEffect wasn't in the asserted type list so deletion is safe)
• cargo test -p v3-compiler --test real_stdlib_parse_smoke — 10/10 pass (effects_dag_parses green)
• cargo test -p v3-compiler — full suite green, 0 failures
• cargo test --workspace --exclude v2-compiler-tests — 34 "ok" test-result lines, 0 failures
• cargo fmt --all --check — clean
• cargo clippy --all-targets -- -D warnings — clean

🤖 Generated with Claude Code

[briansrls]

ChatGPT review in progress... (view conversation)

Check back in ~30 minutes for the full review.

[briansrls]

Clean execution of the brief. Three files, no surprise scope, ROADMAP and lane2 docs updated to track the remaining gate. Approve.

What's right

• Option (a) chosen as recommended. DerivedOpEffect deleted entirely; derive_op_effect returns OperationEffect? directly. The transport fields (method, path_template) are gone — the rationale in the doc updates is correct: nothing downstream consumed them ("the modifier check and obligation generator both project through shape alone, and ReadEffect already encodes 'method was GET/HEAD/OPTIONS' structurally"). ✅
• Single-authority restored. OperationEffect is now the durable carrier — same shape compose_effects walks, same shape derive_op_effect produces. No more two-types-for-one-fact at the Stage 2a → Stage 2b boundary. ✅
• Consumer migrations are minimal because the bodies didn't actually depend on the deleted fields. check_modifier_vs_derivation(op: OperationEffect, ...) and generate_idempotency_obligations(ops: List<OperationEffect>) — signature changes only, body shapes unchanged. That confirms the original DerivedOpEffect shape was over-wide; this PR's evidence backs the original "lossy compression at substrate" diagnosis. ✅
• Doc updates are explicit and honest.
  • docs/lane2-compile-time-proofs.md: boundary note rewritten as "(cleared for DerivedOpEffect)" with the rationale inline; pre-start gate at line 90 narrowed from "DerivedOpEffect or ComposedEffect" to "ComposedEffect" only.
  • ROADMAP.md: deferral row title narrowed from DerivedOpEffect / ComposedEffect boundary cleanup to just ComposedEffect; "Cleared (this PR)" subsection appended with the structural rationale.
• ComposedEffect correctly preserved as the remaining gate. The PR doesn't try to do too much. ComposedEffect's reshape is its own work item; this PR doesn't touch it. Scope discipline good. ✅

Forward-looking note (not a blocker for this PR)

The closing sentence in the ROADMAP "Cleared" entry is the right framing for future debt avoidance: "Future diagnostic rendering that wants the originating method/path should attach a separate evidence carrier to the diagnostic, not smuggle transport facts onto the effect record." That's exactly the principle that should govern the eventual ComposedEffect reshape too — separate concerns, don't bundle.

Verdict

APPROVE. Smallest of tonight's PRs, cleanest scope, most directly executes its brief. Ready to merge.

[briansrls]

codex · gpt-5.4 · 01fb105d

✅ Review (blocking: 0, non-blocking: 2+/0-)

Non-blocking — Strengths

• src/v3/std/effects.dag Collapsing DerivedOpEffect into OperationEffect removes a parallel substrate carrier and keeps downstream consumers on the same authority they already read: operation_name plus shape.
• docs/lane2-compile-time-proofs.md The Stage 2a boundary note is honest about the discard: it explains why method and path_template do not need to survive on the effect record instead of letting the convenience shape harden by inertia.

ROADMAP — Verified

• DerivedOpEffect boundary cleanup: src/v3/ROADMAP.md and docs/lane2-compile-time-proofs.md now agree that this deferral is cleared and that ComposedEffect remains the only Stage 2a boundary cleanup still open.

✅ This looks clean; the PR simplifies the effect substrate without dropping any fact that current downstream consumers actually use, and it keeps the remaining debt explicitly tracked.

[briansrls]

codex · gpt-5.4 · 01fb105d

✅ Review (blocking: 0, non-blocking: 2+/0-)

Non-blocking — Strengths

• src/v3/std/effects.dag Collapsing DerivedOpEffect into OperationEffect removes a parallel substrate carrier and keeps downstream consumers on the same authority they already read: operation_name plus shape.
• docs/lane2-compile-time-proofs.md The Stage 2a boundary note is honest about the discard: it explains why method and path_template do not need to survive on the effect record instead of letting the convenience shape harden by inertia.

ROADMAP — Verified

• DerivedOpEffect boundary cleanup: src/v3/ROADMAP.md and docs/lane2-compile-time-proofs.md now agree that this deferral is cleared and that ComposedEffect remains the only Stage 2a boundary cleanup still open.

✅ This looks clean; the PR simplifies the effect substrate without dropping any fact that current downstream consumers actually use, and it keeps the remaining debt explicitly tracked.

[briansrls]

ChatGPT Review

Principle audit.

Fail-closed. Satisfied for this follow-up. The diff does not introduce a new silent-success path; it narrows the boundary so derivation emits the same carrier the downstream checks already consume. The carrier alignment is structural, even though the underlying partial derivation remains inherited Stage 2a behavior rather than something this PR changes.

Illegal states unrepresentable. Improved. DerivedOpEffect let the model carry transport facts (method, path_template) alongside the already-derived effect fact, which admitted redundant states the downstream API never needed. Collapsing to OperationEffect removes that redundancy structurally.

Facts flow forward. Satisfied. The PR does not silently lose method/path_template; it explicitly records that those facts stop at the Stage 2a boundary because the only surviving consumers project through shape. That discard is documented behaviorally in the lane doc / roadmap, while the surviving carrier alignment is structural in the type signature.

Coproduct dissolution. Satisfied. No new substrate enum is introduced, and the modeling move is in the right direction: a short-lived bridge carrier is dissolved before Stage 2b learns it as canonical. The remaining provisional shape, ComposedEffect, stays explicitly tracked instead of being normalized by inertia.

Single-authority metadata. Strongly satisfied. Before this change, derivation produced one carrier and composition/checking walked another; after it, derivation, modifier falsification, obligation generation, and future composition all share the same OperationEffect boundary. This is the clearest win in the PR.

API-level enforcement. Improved. The important invariant—“derivation outputs the same shape composition consumes”—is now enforced by the API instead of depending on discipline across two parallel records. The remaining ComposedEffect issue is still only tracked by documentation, not by a stronger type shape yet.

Design question.

Non-blocking: when Stage 2b or Track 17a eventually wants richer diagnostics for a misderived effect, where will method/path provenance live without reintroducing a second effect carrier? The stake is preserving the win this PR banks: transport provenance should attach as separate diagnostic/test evidence, not grow OperationEffect back into DerivedOpEffect under a new name.

Verdict.

APPROVE. This is the right cleanup: it removes a tracked bridge, improves illegal-state modeling, and makes the Stage 2a/2b seam use one durable carrier. I don’t see a real concern in this commit beyond the already-tracked ComposedEffect follow-up.

LOOP HEALTH: converging — this round dissolves one explicitly tracked scaffold (DerivedOpEffect) and leaves the remaining ComposedEffect debt named, bounded, and gated before new consumers land.

---

View conversation