cd3d1ca9ed
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.
408 lines
16 KiB
Rust
408 lines
16 KiB
Rust
//! `GraphBuilder` — name-based blueprint authoring (spec 0039).
|
|
//!
|
|
//! A fluent, additive authoring surface that wires a blueprint by typed node
|
|
//! handles and port/field *names*, resolving them to the raw-index `Composite`
|
|
//! at a single fallible `build()`. The flat graph stays index-wired (C23): names
|
|
//! are resolved at the authoring boundary and never reach `FlatGraph` — the same
|
|
//! posture param-name resolution already has (`Binder`, blueprint.rs). Resolution
|
|
//! mirrors `PrimitiveBuilder::bind`'s collect-then-reject (node.rs), but returns
|
|
//! `Err(BuildError)` instead of panicking.
|
|
|
|
use aura_core::{NodeSchema, ScalarKind};
|
|
|
|
use crate::blueprint::{BlueprintNode, Composite, OutField, Role};
|
|
use crate::harness::{Edge, Target};
|
|
|
|
/// A typed, `Copy` reference to a node added to a [`GraphBuilder`], carrying the
|
|
/// node's index in the builder's `nodes` Vec.
|
|
#[derive(Clone, Copy, Debug, PartialEq, Eq)]
|
|
pub struct NodeHandle(usize);
|
|
|
|
/// A typed, `Copy` reference to a role reserved on a [`GraphBuilder`].
|
|
#[derive(Clone, Copy, Debug, PartialEq, Eq)]
|
|
pub struct RoleHandle(usize);
|
|
|
|
/// An unresolved consumer endpoint: a node handle's index plus an input-port name.
|
|
#[derive(Clone, Copy, Debug)]
|
|
pub struct InPort {
|
|
node: usize,
|
|
name: &'static str,
|
|
}
|
|
|
|
/// An unresolved producer endpoint: a node handle's index plus an output-field name.
|
|
#[derive(Clone, Copy, Debug)]
|
|
pub struct OutPort {
|
|
node: usize,
|
|
name: &'static str,
|
|
}
|
|
|
|
impl NodeHandle {
|
|
/// Address an input port of this node by name (resolved at [`GraphBuilder::build`]).
|
|
pub fn input(self, name: &'static str) -> InPort {
|
|
InPort { node: self.0, name }
|
|
}
|
|
/// Address this node's output field by name (resolved at [`GraphBuilder::build`]).
|
|
pub fn output(self, name: &'static str) -> OutPort {
|
|
OutPort { node: self.0, name }
|
|
}
|
|
}
|
|
|
|
/// An authoring-layer fault surfaced at [`GraphBuilder::build`]. Resolution by
|
|
/// name is necessary, not sufficient: a name-resolved edge that connects
|
|
/// mismatched scalar kinds still surfaces downstream as a bootstrap kind-check.
|
|
#[derive(Clone, Debug, PartialEq, Eq)]
|
|
pub enum BuildError {
|
|
/// A handle whose index is out of range (i.e. minted by a different builder).
|
|
BadHandle { node: usize },
|
|
/// No input port of `node` is named `name`.
|
|
UnknownInPort { node: usize, name: String },
|
|
/// More than one input port of `node` is named `name`.
|
|
AmbiguousInPort { node: usize, name: String },
|
|
/// No output field of `node` is named `name`.
|
|
UnknownOutPort { node: usize, name: String },
|
|
/// More than one output field of `node` is named `name`.
|
|
AmbiguousOutPort { node: usize, name: String },
|
|
}
|
|
|
|
/// A fluent, additive accumulator that authors a [`Composite`] by typed handles
|
|
/// and port/field names. See the module docs and spec 0039.
|
|
pub struct GraphBuilder {
|
|
name: String,
|
|
nodes: Vec<BlueprintNode>,
|
|
schemas: Vec<NodeSchema>,
|
|
edges: Vec<(OutPort, InPort)>,
|
|
roles: Vec<(String, Option<ScalarKind>, Vec<InPort>)>,
|
|
out: Vec<(String, OutPort)>,
|
|
}
|
|
|
|
impl GraphBuilder {
|
|
/// Start a builder for a composite named `name` (a non-load-bearing render symbol).
|
|
pub fn new(name: impl Into<String>) -> Self {
|
|
Self {
|
|
name: name.into(),
|
|
nodes: Vec::new(),
|
|
schemas: Vec::new(),
|
|
edges: Vec::new(),
|
|
roles: Vec::new(),
|
|
out: Vec::new(),
|
|
}
|
|
}
|
|
|
|
/// Add a node (primitive builder or nested composite); returns its handle.
|
|
pub fn add(&mut self, item: impl Into<BlueprintNode>) -> NodeHandle {
|
|
let node = item.into();
|
|
let schema = node.signature();
|
|
let idx = self.nodes.len();
|
|
self.nodes.push(node);
|
|
self.schemas.push(schema);
|
|
NodeHandle(idx)
|
|
}
|
|
|
|
/// Reserve an open input role (wired by an enclosing graph); returns its handle.
|
|
pub fn input_role(&mut self, name: &str) -> RoleHandle {
|
|
let idx = self.roles.len();
|
|
self.roles.push((name.to_string(), None, Vec::new()));
|
|
RoleHandle(idx)
|
|
}
|
|
|
|
/// Reserve a bound ingestion role of `kind` (a root source); returns its handle.
|
|
pub fn source_role(&mut self, name: &str, kind: ScalarKind) -> RoleHandle {
|
|
let idx = self.roles.len();
|
|
self.roles.push((name.to_string(), Some(kind), Vec::new()));
|
|
RoleHandle(idx)
|
|
}
|
|
|
|
/// Wire a producer's output field into a consumer's input port (resolved at build).
|
|
pub fn connect(&mut self, from: OutPort, to: InPort) {
|
|
self.edges.push((from, to));
|
|
}
|
|
|
|
/// Fan a role's value into one or more consumer input ports.
|
|
pub fn feed(&mut self, role: RoleHandle, into: impl IntoIterator<Item = InPort>) {
|
|
self.roles[role.0].2.extend(into);
|
|
}
|
|
|
|
/// Re-export a producer's output field at the composite boundary under `name`.
|
|
pub fn expose(&mut self, from: OutPort, name: &str) {
|
|
self.out.push((name.to_string(), from));
|
|
}
|
|
|
|
/// Resolve every accumulated name to an index and assemble the [`Composite`].
|
|
pub fn build(self) -> Result<Composite, BuildError> {
|
|
let mut edges = Vec::with_capacity(self.edges.len());
|
|
for (from, to) in &self.edges {
|
|
let from_field = self.resolve_field(from)?;
|
|
let (to_node, slot) = self.resolve_slot(to)?;
|
|
edges.push(Edge { from: from.node, to: to_node, slot, from_field });
|
|
}
|
|
|
|
let mut roles = Vec::with_capacity(self.roles.len());
|
|
for (name, source, ports) in &self.roles {
|
|
let mut targets = Vec::with_capacity(ports.len());
|
|
for p in ports {
|
|
let (node, slot) = self.resolve_slot(p)?;
|
|
targets.push(Target { node, slot });
|
|
}
|
|
roles.push(Role { name: name.clone(), targets, source: *source });
|
|
}
|
|
|
|
let mut output = Vec::with_capacity(self.out.len());
|
|
for (name, from) in &self.out {
|
|
let field = self.resolve_field(from)?;
|
|
output.push(OutField { node: from.node, field, name: name.clone() });
|
|
}
|
|
|
|
Ok(Composite::new(self.name, self.nodes, edges, roles, output))
|
|
}
|
|
|
|
/// Resolve an `InPort` to `(node-index, slot)` by exactly-one-match on the
|
|
/// node's input-port names (the `PrimitiveBuilder::bind` posture, fallible).
|
|
fn resolve_slot(&self, p: &InPort) -> Result<(usize, usize), BuildError> {
|
|
let schema = self
|
|
.schemas
|
|
.get(p.node)
|
|
.ok_or(BuildError::BadHandle { node: p.node })?;
|
|
let m: Vec<usize> = schema
|
|
.inputs
|
|
.iter()
|
|
.enumerate()
|
|
.filter(|(_, port)| port.name == p.name)
|
|
.map(|(i, _)| i)
|
|
.collect();
|
|
match m.as_slice() {
|
|
[i] => Ok((p.node, *i)),
|
|
[] => Err(BuildError::UnknownInPort { node: p.node, name: p.name.to_string() }),
|
|
_ => Err(BuildError::AmbiguousInPort { node: p.node, name: p.name.to_string() }),
|
|
}
|
|
}
|
|
|
|
/// Resolve an `OutPort` to a `from_field` by exactly-one-match on the node's
|
|
/// output-field names.
|
|
fn resolve_field(&self, p: &OutPort) -> Result<usize, BuildError> {
|
|
let schema = self
|
|
.schemas
|
|
.get(p.node)
|
|
.ok_or(BuildError::BadHandle { node: p.node })?;
|
|
let m: Vec<usize> = schema
|
|
.output
|
|
.iter()
|
|
.enumerate()
|
|
.filter(|(_, f)| f.name == p.name)
|
|
.map(|(i, _)| i)
|
|
.collect();
|
|
match m.as_slice() {
|
|
[i] => Ok(*i),
|
|
[] => Err(BuildError::UnknownOutPort { node: p.node, name: p.name.to_string() }),
|
|
_ => Err(BuildError::AmbiguousOutPort { node: p.node, name: p.name.to_string() }),
|
|
}
|
|
}
|
|
}
|
|
|
|
#[cfg(test)]
|
|
mod tests {
|
|
use super::{BuildError, GraphBuilder};
|
|
use crate::test_fixtures::sma_cross as hand_sma_cross;
|
|
use crate::{Composite, OutField, Role, Target};
|
|
use aura_core::{Firing, ScalarKind};
|
|
use aura_std::{Exposure, Recorder, SimBroker, Sma, Sub};
|
|
use std::sync::mpsc;
|
|
|
|
/// The `sma_cross` sub-composite authored through the builder.
|
|
fn built_sma_cross() -> Composite {
|
|
let mut g = GraphBuilder::new("sma_cross");
|
|
let fast = g.add(Sma::builder().named("fast"));
|
|
let slow = g.add(Sma::builder().named("slow"));
|
|
let sub = g.add(Sub::builder());
|
|
let price = g.input_role("price");
|
|
g.feed(price, [fast.input("series"), slow.input("series")]);
|
|
g.connect(fast.output("value"), sub.input("lhs"));
|
|
g.connect(slow.output("value"), sub.input("rhs"));
|
|
g.expose(sub.output("value"), "out");
|
|
g.build().expect("sma_cross resolves")
|
|
}
|
|
|
|
#[test]
|
|
fn builder_sma_cross_structurally_equals_hand_wired() {
|
|
let built = built_sma_cross();
|
|
let hand = hand_sma_cross();
|
|
assert_eq!(built.edges(), hand.edges());
|
|
assert_eq!(built.input_roles(), hand.input_roles());
|
|
assert_eq!(built.output(), hand.output());
|
|
}
|
|
|
|
#[test]
|
|
fn builder_harness_compiles_identically_to_hand_wired() {
|
|
use crate::test_fixtures::composite_sma_cross_harness;
|
|
use aura_core::Scalar;
|
|
|
|
// Shared param vector for the harness param-space (sma_cross.fast.length,
|
|
// sma_cross.slow.length, exposure.scale) — applied to both sides, so it
|
|
// cancels in the comparison. Params size buffers, not topology (C11), so
|
|
// edges/sources are param-independent; a fixed vector just satisfies the
|
|
// arity gate compile() (no-param) trips on this value-empty harness.
|
|
let params = [Scalar::i64(2), Scalar::i64(4), Scalar::f64(0.5)];
|
|
|
|
// hand-wired reference harness -> FlatGraph
|
|
let (hand_bp, _rx_eq, _rx_ex) = composite_sma_cross_harness();
|
|
let hand_flat = hand_bp.compile_with_params(¶ms).expect("hand harness compiles");
|
|
|
|
// builder-authored harness (nests a builder-authored sma_cross) -> FlatGraph
|
|
let (tx_eq, _r1) = mpsc::channel();
|
|
let (tx_ex, _r2) = mpsc::channel();
|
|
let mut g = GraphBuilder::new("root");
|
|
let xross = g.add(built_sma_cross());
|
|
let expo = g.add(Exposure::builder());
|
|
let broker = g.add(SimBroker::builder(0.0001));
|
|
let eq = g.add(Recorder::builder(vec![ScalarKind::F64], Firing::Any, tx_eq));
|
|
let ex = g.add(Recorder::builder(vec![ScalarKind::F64], Firing::Any, tx_ex));
|
|
let src = g.source_role("src", ScalarKind::F64);
|
|
g.feed(src, [xross.input("price"), broker.input("price")]);
|
|
g.connect(xross.output("out"), expo.input("signal"));
|
|
g.connect(expo.output("exposure"), broker.input("exposure"));
|
|
g.connect(broker.output("equity"), eq.input("col[0]"));
|
|
g.connect(expo.output("exposure"), ex.input("col[0]"));
|
|
let built_flat = g
|
|
.build()
|
|
.expect("root resolves")
|
|
.compile_with_params(¶ms)
|
|
.expect("built compiles");
|
|
|
|
assert_eq!(built_flat.edges, hand_flat.edges);
|
|
assert_eq!(built_flat.sources, hand_flat.sources);
|
|
}
|
|
|
|
#[test]
|
|
fn unconnected_port_rejected_on_builder_surface() {
|
|
use crate::CompileError;
|
|
use aura_core::Scalar;
|
|
// The forgotten-exposure-leg harness: every port/field name resolves, so
|
|
// build() succeeds; but broker.input("exposure") is never connected, so the
|
|
// wiring-totality check rejects it at compile (broker is node 1, slot 0).
|
|
let (tx_eq, _r1) = mpsc::channel();
|
|
let mut g = GraphBuilder::new("root");
|
|
let expo = g.add(Exposure::builder());
|
|
let broker = g.add(SimBroker::builder(0.0001));
|
|
let eq = g.add(Recorder::builder(vec![ScalarKind::F64], Firing::Any, tx_eq));
|
|
let price = g.source_role("price", ScalarKind::F64);
|
|
g.feed(price, [expo.input("signal"), broker.input("price")]);
|
|
// BUG: g.connect(expo.output("exposure"), broker.input("exposure")) missing.
|
|
g.connect(broker.output("equity"), eq.input("col[0]"));
|
|
let compiled = g
|
|
.build()
|
|
.expect("all port/field names resolve")
|
|
.compile_with_params(&[Scalar::f64(0.5)]);
|
|
assert_eq!(
|
|
compiled.err(),
|
|
Some(CompileError::UnconnectedPort { node: 1, slot: 0 })
|
|
);
|
|
}
|
|
|
|
#[test]
|
|
fn unknown_in_port_is_reported_at_build() {
|
|
let mut g = GraphBuilder::new("x");
|
|
let a = g.add(Sma::builder());
|
|
let b = g.add(Sub::builder());
|
|
g.connect(a.output("value"), b.input("nope"));
|
|
// `.err()`, not `.unwrap_err()`: the latter would require `Composite: Debug`
|
|
// (the Ok type), which it is not (it holds a `Box<dyn Fn>` builder closure).
|
|
assert_eq!(
|
|
g.build().err(),
|
|
Some(BuildError::UnknownInPort { node: 1, name: "nope".into() })
|
|
);
|
|
}
|
|
|
|
#[test]
|
|
fn unknown_out_port_is_reported_at_build() {
|
|
let mut g = GraphBuilder::new("x");
|
|
let a = g.add(Sma::builder());
|
|
let b = g.add(Sub::builder());
|
|
g.connect(a.output("nope"), b.input("lhs"));
|
|
assert_eq!(
|
|
g.build().err(),
|
|
Some(BuildError::UnknownOutPort { node: 0, name: "nope".into() })
|
|
);
|
|
}
|
|
|
|
#[test]
|
|
fn out_of_range_handle_is_bad_handle() {
|
|
// g1 mints index 2; g2 has only index 0 — using g1's handle in g2 is out of range.
|
|
let mut g1 = GraphBuilder::new("g1");
|
|
g1.add(Sma::builder());
|
|
g1.add(Sma::builder());
|
|
let third = g1.add(Sub::builder()); // NodeHandle(2)
|
|
let mut g2 = GraphBuilder::new("g2");
|
|
let only = g2.add(Sub::builder()); // NodeHandle(0)
|
|
g2.connect(third.output("value"), only.input("lhs"));
|
|
assert_eq!(g2.build().err(), Some(BuildError::BadHandle { node: 2 }));
|
|
}
|
|
|
|
#[test]
|
|
fn ambiguous_in_port_is_reported() {
|
|
// a composite with two roles both named "x" derives two input ports named "x"
|
|
let inner = Composite::new(
|
|
"dup_in",
|
|
vec![Sub::builder().into()],
|
|
vec![],
|
|
vec![
|
|
Role { name: "x".into(), targets: vec![Target { node: 0, slot: 0 }], source: None },
|
|
Role { name: "x".into(), targets: vec![Target { node: 0, slot: 1 }], source: None },
|
|
],
|
|
vec![OutField { node: 0, field: 0, name: "o".into() }],
|
|
);
|
|
let mut g = GraphBuilder::new("outer");
|
|
let c = g.add(inner);
|
|
let src = g.source_role("s", ScalarKind::F64);
|
|
g.feed(src, [c.input("x")]);
|
|
assert_eq!(
|
|
g.build().err(),
|
|
Some(BuildError::AmbiguousInPort { node: 0, name: "x".into() })
|
|
);
|
|
}
|
|
|
|
#[test]
|
|
fn ambiguous_out_port_is_reported() {
|
|
// a composite re-exporting the same interior field twice under one name
|
|
let inner = Composite::new(
|
|
"dup_out",
|
|
vec![Sub::builder().into()],
|
|
vec![],
|
|
vec![],
|
|
vec![
|
|
OutField { node: 0, field: 0, name: "o".into() },
|
|
OutField { node: 0, field: 0, name: "o".into() },
|
|
],
|
|
);
|
|
let mut g = GraphBuilder::new("outer");
|
|
let c = g.add(inner);
|
|
let snk = g.add(Sub::builder());
|
|
g.connect(c.output("o"), snk.input("lhs"));
|
|
assert_eq!(
|
|
g.build().err(),
|
|
Some(BuildError::AmbiguousOutPort { node: 0, name: "o".into() })
|
|
);
|
|
}
|
|
|
|
#[test]
|
|
fn simbroker_legs_resolve_by_name_and_a_typo_is_caught() {
|
|
// exposure/price (both f64, slot 0/1) addressed by NAME — the #21 legibility win
|
|
let mut ok = GraphBuilder::new("ok");
|
|
let expo = ok.add(Exposure::builder());
|
|
let broker = ok.add(SimBroker::builder(0.0001));
|
|
let src = ok.source_role("p", ScalarKind::F64);
|
|
ok.feed(src, [expo.input("signal"), broker.input("price")]);
|
|
ok.connect(expo.output("exposure"), broker.input("exposure"));
|
|
ok.expose(broker.output("equity"), "eq");
|
|
assert!(ok.build().is_ok());
|
|
|
|
// a transposed/typo'd port name is caught, where a bare slot index is not
|
|
let mut typo = GraphBuilder::new("typo");
|
|
let expo2 = typo.add(Exposure::builder());
|
|
let broker2 = typo.add(SimBroker::builder(0.0001));
|
|
typo.connect(expo2.output("exposure"), broker2.input("pirce"));
|
|
assert_eq!(
|
|
typo.build().err(),
|
|
Some(BuildError::UnknownInPort { node: 1, name: "pirce".into() })
|
|
);
|
|
}
|
|
}
|