refactor(controller): trait impls + main.rs alias removal (PR #130 TODO 3+4)

[zzylol]

Summary

Two of the four PR #130 follow-up TODOs land here. The other two are blocked by architectural reality — surfaced explicitly so the follow-up plan can be reshaped.

TODO 3 — Placeholder traits made functional ✅

• OptimizerRule trait extended (added Cse, Decorrelate categories). All 12 engine rules get per-rule OptimizerRule impls in optimizer/engine.rs. New default_rules_as_optimizer_rules() returns Vec<Box<dyn OptimizerRule>>. Blanket impl for every sketch_algebra::rules::Rule in optimizer/mod.rs.
• PlanEmitter trait gets 4 concrete impls in emit/trait_def.rs: OpampEmitter, OpampGatewayEmitter, StreamingConfigEmitter, InferenceConfigEmitter. Each wraps the existing free-function emitter. InferenceConfigInput<'a> bundles per-tenant inputs.
• DeploymentModelRegistry extended from stub to a real HashMap<DeploymentModelId, DeploymentModel>. Default::default() registers asaplifecycle with the engine's 12-rule library + 4 demo emitters.

13 new unit tests across the three trait files.

TODO 4 — main.rs migrated off back-compat aliases ✅

Rewrote all 52 reference sites:

• algebra::* → optimizer::engine::* / physical::* / intent_algebra::legacy_expr::*
• analyzer::* → pipeline::*
• config::* → emit::* / workload::*
• planner::* → optimizer::* / physical::stage_split / optimizer::cost::online
• stage_split::* → physical::colored_dag::*

Removed all 5 back-compat shims from lib.rs (pub use emit as config, pub use pipeline as analyzer, pub mod algebra, pub use physical::colored_dag as stage_split, pub mod planner). lib.rs shrank 184 → 95 lines.

TODO 1 — legacy_expr migration: BLOCKED 🚫

The canonical intent_algebra::{query_expr,agg_intent,schema} is NOT a structural superset of the legacy intent_algebra::legacy_expr. The canonical QueryExpr has 5 variants (Scan, Window, Aggregate, LetBinding, Ref); legacy has ~26 plus a full ScalarExpr IR with no canonical equivalent.

The canonical IR's own module docs flag this: "the full design.md list is larger; they are deferred to follow-up phases as the planner grows consumers for them".

Migrating consumers requires first growing the canonical IR to absorb the variant set — substantial separate refactor work, not a follow-up cleanup. 50+ consumer sites depend on the legacy variants. Left intact.

Recommendation: retitle the follow-up from "migrate consumers" to "grow canonical L3 to absorb legacy variants" so the scope is honest.

TODO 2 — stage_config.rs split: DEFERRED 🟡

3,020-line emit/stage_config.rs mixes OTel YAML / backend JSON / storage-routing emit with intricately shared helpers and 1,580 lines of interleaved tests.

Critically: TODO 3's wrapper structs (OpampEmitter, StreamingConfigEmitter, InferenceConfigEmitter) already provide the polymorphic PlanEmitter surface. The file-level split is no longer load-bearing for the trait work — it would be purely cosmetic at this point.

Recommendation: address as a separate PR with dedicated time budget if/when the monolith becomes a review bottleneck.

Build + test

• cargo build --release -p controller — clean (4 main + 19 lib warnings, all pre-existing)
• cargo build --release -p query_engine_rust — clean
• cargo test --release -p controller --lib — 646 passed (was 633; +13 trait tests)
• cargo test --release -p controller --bin controller — 27 passed
• cargo test --release -p query_engine_rust --lib -- engines::warm_tier — 13 passed

Diff: 8 files, +785 / -207

File	Net
controller/src/deployment_model.rs	+202
controller/src/emit/mod.rs	+5
controller/src/emit/trait_def.rs	+282
controller/src/lib.rs	-92
controller/src/main.rs	+12 (rewritten paths, similar line count)
controller/src/optimizer/engine.rs	+86
controller/src/optimizer/mod.rs	+21
controller/src/optimizer/trait_def.rs	+118

Caveat — controller/src/store/ (out of scope, flagging)

lib.rs declares pub mod store; but controller/src/store/ is in repo root .gitignore (broad store/ pattern). The directory is untracked; build works because it exists locally on dev boxes. Fresh clones would fail. Needs a separate decision: track the module or remove the declaration.

🤖 Generated with Claude Code