72bfdf79d1c4a24ae0166bb960b2f2c6e38f37da
7 Commits
| Author | SHA1 | Message | Date | |
|---|---|---|---|---|
 Brummel
|
43716be10e |
refactor(aura-engine): Cell into the construction path — param-plane base/frontend split
Cell becomes the carrier of the construction path; Scalar narrows to the author/render boundaries. The validated/enumerated param point carries no redundant kind (it lives once, in the declared param-space); at the author edge the kind is a checksum — two independent sources, the typed value vs. the slot — so the self-describing Scalar stays there. The name->slot binding is dynamic (C23), so the check is necessarily a runtime one and the value must self-describe for it. Base/frontend split (the AnyColumn push/push_cell pattern one level up): compile_with_cells / bootstrap_with_cells are the kind-check-free base; compile_with_params / bootstrap_with_params are the frontend that adds only the per-value kind checksum, strips to cells (new Scalar::cell(), the partner of Scalar::from_cell), and delegates. lower_items loses its per-primitive kind-check; PrimitiveBuilder::build and the std node builders (sma/ema/exposure/lincomb) read cells. Boundary: construction -> Cell (PrimitiveBuilder::build, lower_items, GridSpace.axes/points, SweepPoint.params, the sweep closure). Author edges stay Scalar (GridSpace::new, bind, compile_with_params/bootstrap_with_params). walkforward chosen_params stays a self-describing Scalar report record (Option A — WalkForwardResult carries no space); the cell winner is reconstructed once at the WindowRun site via from_cell. injective-check moved ahead of the arity-check in the frontend to preserve the pre-split error order (DuplicateParamPath before ParamArity). The lossy i64->f64 param projection (scalar_as_f64 / scalar_as_param_f64) is deliberately untouched — a separate follow-up. Behaviour-preserving (C1): build --all-targets / test / clippy -D warnings / doc all clean across the workspace. |
||
|
Brummel
|
82635fa259 |
refactor: Cell becomes the hot-path value carrier (C7/C8 realization)
Motivation ---------- The C7 realization note ( |
||
|
Brummel
|
cd3d1ca9ed |
refactor(aura-core): split Scalar into a tag-free Cell + ScalarKind
Motivation
----------
`Scalar` was a tagged enum (I64/F64/Bool/Ts), so every scalar value
physically carried its own kind tag. But the kind is already known from
the schema/port/column the value flows through (C7: the type is a
property of the column, not of the value — the hot path is already
columnar `Column<T>`, and `AnyColumn::get` *reconstructs* the tag from
the column on the way out). The per-value tag was therefore redundant
with the kind the surrounding context already holds.
That redundancy had three costs:
* It baked an implicit `match` (a branch) into every function that read
a Scalar payload — even where the caller statically knew the type.
The tag could never be exploited away.
* Size: a tagged enum is tag + payload = 16 bytes (f64/i64 alignment),
twice the 8 bytes the value needs. A `Column<Scalar>` would be double
the memory and half the cache utilisation.
* It is the shared root of several downstream papercuts we keep hitting
— the lossy f64 manifest field, the `unreachable!` panic on a
non-numeric param, the serde-tag question — all symptoms of "the type
is baked into the value".
Change
------
Introduce `Cell`: a type-erased 64-bit word (`struct Cell(u64)`) that is
not readable without external type context. It is constructed per base
type (`from_i64/from_f64/from_bool/from_ts`) and read only by naming the
type at the call site (`i64()/f64()/bool()/ts()`) — each a branch-free
bit-cast. The hot path resolves the kind once at the boundary (from the
schema) and then reads natively, with no per-value branch. `Cell` knows
nothing of `Scalar` or `ScalarKind`; the dependency is strictly one-way,
and it lives in its own `cell.rs` (more is planned on top of it).
`Scalar` becomes `struct { kind: ScalarKind, cell: Cell }` — the
self-describing form for the dynamic boundaries (builder binding,
serialization, rendering), built on top of `Cell`. Its `as_*` accessors
now `debug_assert` the kind and return the native value (free in
release); calling the wrong accessor is a caller bug, not a checked
`Option`. The variant constructors `Scalar::I64(..)` become associated
fns `Scalar::i64(..)`.
`PartialEq` is hand-written (not derived) to preserve the former enum's
value semantics: kinds must match, then native payloads compare, so f64
keeps IEEE-754 behaviour (`NaN != NaN`, `+0.0 == -0.0`) and a kind
mismatch is never equal even when the raw words coincide. A fixture
(`scalar_eq_is_value_not_bitwise`) pins exactly the cases where bit- and
value-equality diverge, so it can't silently regress. `Cell`'s own
`Eq`/`Hash` stay bitwise — correct for a raw word.
The change is behaviour-preserving: Scalar's observable behaviour is
identical to the pre-Cell enum (the value-equality fixture proves it);
only the internal representation changed. The ~440 call sites across the
workspace are a mechanical constructor rename plus ~12 destructuring
sites (match-arms / `let`-patterns) rewritten to `kind()` + `as_*`.
Verified: cargo build --workspace --all-targets, cargo clippy --workspace
--all-targets -- -D warnings, cargo test --workspace — all green.
|
||
|
Brummel
|
a8bd52eaff |
refactor(aura-core): own FieldSpec.name to drop the derive_signature heap leak
FieldSpec.name was &'static str, forcing derive_signature to Box::leak every re-exported composite OutField name. Because compile_with_params reruns derive_signature per sweep point, an N-point sweep over an M-output-field composite leaked N×M boundary names unbounded — a scaling wall the moment the World/sweep layer compiles families per point. Pull FieldSpec.name to String, mirroring the already-owned PortSpec.name and ParamSpec.name: a composite re-exports interior fields under runtime boundary names, so ownership is the correct model, not a workaround. FieldSpec loses Copy; the only fallout is one borrow at the render site. leak_name/Box::leak deleted. Behaviour-preserving: signature contents unchanged, all 231 tests green. |
||
|
Brummel
|
e304dbaae1 |
feat(aura-core): name input ports
PortSpec gains a non-load-bearing `name: String`, so an input port is named just as FieldSpec.name (output) and ParamSpec.name (param) already are — input ports were the lone unnamed member of the node signature. Identity stays positional by slot (C23); the name is render/debug only, never read by bootstrap or the run loop. PortSpec drops Copy (String is not Copy), exactly as ParamSpec already does. - Every aura-std node names its input slots: SMA/EMA "series", Sub/Add "lhs"/"rhs", Exposure "signal", SimBroker "exposure"/"price" (the slots become self-documenting), LinComb "term[i]" and Recorder "col[i]" generated in their build loops (mirroring LinComb's existing weights[i] param loop). - derive_signature carries a composite's Role.name into the derived input port (it was dropped before — the output side already carried FieldSpec.name), so the graph model is homogeneously named at both levels. - model_to_json (port_json + the composite-header inputs in scope_json) emits the name as a third tuple element: ["f64","any","exposure"]. The byte golden was re-captured (machine bytes) and its substring twins updated; the model is now fully named across inputs/outputs/params. - All 16 PortSpec construction sites threaded in one compile-gate change; test fixtures carry fixture names. C8 realization note added to the design ledger. Why name-only, no validation: the name is a pure debug symbol. Wire-by-name was rejected (it would be a C23 contract change). Bootstrap slot-wiring validation (which would close #21's same-kind swap footgun) is deferred to its own cycle — a name alone does not catch the swap; it makes the slots self-documenting and gives a future validation something to check against. Verified: cargo test --workspace 168 green; clippy --all-targets -D warnings clean; cargo build --workspace clean. Read-only render path (C9), no serde (C14), scalar kinds unchanged (C4). closes #50 refs #21 refs #51 |
||
|
Brummel
|
1b3909316e |
feat(aura): node signature lives in the blueprint; collapse Blueprint into Composite
Consolidate the node data structure so every node's signature (NodeSchema:
inputs/output/params) is declared once and exists in the blueprint pre-build,
and dissolve the special "root graph" type. Behaviour-preserving (C1).
Signature vs sizing
- NodeSchema is now the static signature only: InputSpec -> PortSpec{kind,firing},
with lookback removed. The signature is fully static per blueprint (input
kinds/firing, output fields, params); LinComb's variable arity is a builder arg,
not an injected param.
- The one param-dependent quantity, an input's buffer lookback (e.g. Sma's window =
its injected length), moves out of the signature to Node::lookbacks() -> Vec<usize>,
read only by bootstrap for sizing. Node::schema() is removed.
- LeafFactory -> PrimitiveBuilder, which carries the full NodeSchema. The built node
no longer re-declares it: closes the params-declared-twice drift (#36, the 8
per-node factory_params_match_built_node_schema lockstep tests are deleted — their
subject is now structurally impossible) and a value-empty recipe exposes its full
I/O interface pre-build (#43).
Root is just a bound composite
- struct Blueprint is deleted; its compile/bootstrap/param_space methods move onto
Composite. Role gains source: Option<ScalarKind> (None = open interior port,
Some = bound ingestion feed). A composite is runnable iff every root role is bound;
the "main graph" is no longer a category, only the fully-source-bound composite.
New error CompileError::UnboundRootRole for an open root role.
- BlueprintNode::signature() answers uniformly for both arms: Primitive returns the
builder's declared schema, Composite derives it from the interior (role kinds in,
OutField kinds out, aggregated params), pre-build, no build.
compile -> FlatGraph -> bootstrap
- compile validates structure pre-build via signature() (validate_wiring: range +
kind, returning the same variants as before, so an edge kind fault is now caught
before any build closure fires) and emits FlatGraph{nodes,signatures,sources,edges}.
- bootstrap consumes the FlatGraph: kinds/firing/output from the carried signatures,
buffer depth from node.lookbacks(). SourceSpec survives as the flat descriptor.
Renames: BlueprintNode::Leaf -> Primitive, LeafFactory -> PrimitiveBuilder.
Render (aura-cli/src/graph.rs) is migrated compile-only: it takes &Composite, maps
bound roles to the same source-entry shape, so both render goldens reproduce
byte-identical output (no re-capture needed). Render-fidelity tuning is the next cycle.
Verification (orchestrator-run, not agent-reported): cargo build --workspace green;
cargo test --workspace 150 passed / 0 failed; cargo clippy --workspace --all-targets
-D warnings clean. All pinned determinism/run-output tests pass with values unchanged;
no behavioural assertion was altered to go green. 5 new tests assert the signature is
pre-build and uniform, that compile rejects a kind mismatch without building (via a
panicking builder), UnboundRootRole, and lookbacks()/signature arity agreement.
Deferred to cycle-close audit (per plan): docs/design/INDEX.md and some aura-std
module docs still name the old Node::schema()/LeafFactory/BlueprintNode::Leaf/
Blueprint::param_space contracts; prose reconciliation is the architect's at audit.
closes #43 #36
|
||
|
Brummel
|
ae1d4564f5 |
feat(aura-std): EMA node (SMA-seeded, O(1) per cycle)
Add Ema, the std-lib's first recursive/stateful producer: exponential moving average with alpha = 2/(length+1). The building block MACD (and DEMA, MACD-histogram, etc.) needs, and a worked example of a node whose param sizes its math, not its input window. Seeding is ratified as SMA-seeded (ta-lib convention): the EMA stays silent until it has seen length samples, seeds with their SMA, then runs the recurrence. Chosen over a first-value seed for three substantive reasons -- warm-up consistency with Sma (both emit None until length samples), parity with the EMA traders expect (ta-lib MACD seeds this way), and statistical honesty (no average claimed from a single observation). A first-value seed was the proof-of-concept convenience; rejected. Performance: recursive state means lookback = 1 (length sizes alpha, not the window); eval is O(1) time and state via ema += alpha*(x - ema), with no hot-path allocation (the output buffer is reused). Tested: warm-up-then-smooth (a step input separates it from a plain SMA), length-1 identity, label, and factory/schema agreement. |