refs #32
21 KiB
Param-sweep grid — Iteration 1 (engine primitive) — Implementation Plan
Parent spec:
docs/specs/0028-param-sweep-grid.mdFor agentic workers: REQUIRED SUB-SKILL: use the
implementskill to run this plan. Steps use- [ ]checkboxes for tracking.
Goal: Ship the engine-side param-sweep primitive — GridSpace (validated
cartesian enumeration), sweep() (disjoint std::thread::scope execution), and
SweepFamily (ordered collection) — in a new crates/aura-engine/src/sweep.rs.
Architecture: A new module with three parts: enumeration (GridSpace::new
validates one value-list per param slot against param_space(); points()
materializes the cartesian product in odometer order), execution (sweep()
derives the worker count from available_parallelism() and delegates to a
private sweep_with_threads() that runs points disjointly over a shared atomic
cursor, tagging results by index and sorting after join for enumeration-order
determinism), and collection (SweepFamily/SweepPoint). The per-point
run-to-metrics closure is the author's. Zero external dependency (C16) — std
only, no rayon.
Tech Stack: crates/aura-engine (sweep.rs new, lib.rs re-export);
reuses Composite::param_space/bootstrap_with_params (#30/#31),
summarize/f64_field/RunMetrics (report), Scalar/ScalarKind/ParamSpec
(aura-core), and a duplicated value-empty SMA-cross two-sink test fixture.
Files this plan creates or modifies:
- Create:
crates/aura-engine/src/sweep.rs— the sweep module (GridSpace,SweepError,SweepPoint,SweepFamily,sweep, privatesweep_with_threads) + its#[cfg(test)]suite. - Modify:
crates/aura-engine/src/lib.rs:34-44— addmod sweep;to the module list and apub use sweep::{…}re-export.
Task 1: GridSpace — cartesian enumeration + the typed fault gate
Files:
-
Create:
crates/aura-engine/src/sweep.rs -
Modify:
crates/aura-engine/src/lib.rs:34-44 -
Step 1: Create
sweep.rswith the test module only (RED)
Create crates/aura-engine/src/sweep.rs with exactly this content (the
production types referenced by the tests do not exist yet, so the crate will not
compile — that is the RED state):
//! Grid param-sweep (C12.1): enumerate a cartesian grid over a blueprint's
//! param-space and run each point disjointly (C1). This module owns enumeration
//! (`GridSpace`), execution (`sweep`), and collection (`SweepFamily`); the
//! per-point run-to-metrics closure is the author's (harness-specific sink glue
//! the engine cannot generically own — C8/C18).
#[cfg(test)]
mod tests {
use super::*;
use aura_core::{ParamSpec, Scalar, ScalarKind};
fn i64_space(n: usize) -> Vec<ParamSpec> {
(0..n)
.map(|i| ParamSpec { name: format!("p{i}"), kind: ScalarKind::I64 })
.collect()
}
#[test]
fn points_enumerate_in_odometer_order() {
let space = i64_space(2);
let grid = GridSpace::new(
&space,
vec![
vec![Scalar::I64(2), Scalar::I64(3)],
vec![Scalar::I64(4), Scalar::I64(5)],
],
)
.expect("valid grid");
assert_eq!(
grid.points(),
vec![
vec![Scalar::I64(2), Scalar::I64(4)],
vec![Scalar::I64(2), Scalar::I64(5)],
vec![Scalar::I64(3), Scalar::I64(4)],
vec![Scalar::I64(3), Scalar::I64(5)],
],
);
}
#[test]
fn len_is_product_of_axis_lengths() {
let space = vec![
ParamSpec { name: "a".into(), kind: ScalarKind::I64 },
ParamSpec { name: "b".into(), kind: ScalarKind::I64 },
ParamSpec { name: "c".into(), kind: ScalarKind::F64 },
];
let grid = GridSpace::new(
&space,
vec![
vec![Scalar::I64(2), Scalar::I64(3)],
vec![Scalar::I64(4), Scalar::I64(5)],
vec![Scalar::F64(0.5)],
],
)
.expect("valid grid");
assert_eq!(grid.len(), 4);
assert!(!grid.is_empty());
}
#[test]
fn arity_mismatch_is_an_error() {
let space = i64_space(2);
let err = GridSpace::new(&space, vec![vec![Scalar::I64(2)]]).unwrap_err();
assert_eq!(err, SweepError::Arity { expected: 2, got: 1 });
}
#[test]
fn wrong_kind_is_a_kind_mismatch() {
let space = i64_space(1);
let err = GridSpace::new(&space, vec![vec![Scalar::F64(1.0)]]).unwrap_err();
assert_eq!(
err,
SweepError::KindMismatch {
slot: 0,
value_index: 0,
expected: ScalarKind::I64,
got: ScalarKind::F64,
},
);
}
#[test]
fn empty_axis_is_an_error() {
let space = i64_space(1);
let err = GridSpace::new(&space, vec![vec![]]).unwrap_err();
assert_eq!(err, SweepError::EmptyAxis { slot: 0 });
}
}
- Step 2: Wire the module + partial re-export into
lib.rs
In crates/aura-engine/src/lib.rs, add mod sweep; after line 37 (mod report;), so the module list reads:
mod blueprint;
mod graph_model;
mod harness;
mod report;
mod sweep;
And add the partial re-export after line 44 (pub use report::{…}):
pub use sweep::{GridSpace, SweepError};
(The full re-export — adding sweep, SweepFamily, SweepPoint — lands in
Task 2 once those symbols exist.)
- Step 3: Run the test to verify it fails (RED)
Run: cargo test -p aura-engine --lib sweep::
Expected: FAIL — compile error, cannot find type GridSpace/SweepError in this
scope (the production types are not written yet).
- Step 4: Write the production
GridSpace+SweepError(GREEN)
Insert this block at the top of crates/aura-engine/src/sweep.rs, directly
after the //! module doc-comment and before the #[cfg(test)] line:
use aura_core::{ParamSpec, Scalar, ScalarKind};
/// A validated cartesian grid over a blueprint's param-space: one discrete
/// value-list per param slot, in `param_space()` order. Enumerates a family of
/// points (C12.1 grid axis).
pub struct GridSpace {
axes: Vec<Vec<Scalar>>,
}
impl GridSpace {
/// Validate `axes` against `space` (the blueprint's `param_space()`): one
/// axis per slot (`Arity`), every value the slot's declared kind
/// (`KindMismatch`), no empty axis (`EmptyAxis`, which would yield zero
/// points). On success the grid enumerates `∏ |axis_i|` points.
pub fn new(space: &[ParamSpec], axes: Vec<Vec<Scalar>>) -> Result<Self, SweepError> {
if axes.len() != space.len() {
return Err(SweepError::Arity { expected: space.len(), got: axes.len() });
}
for (slot, (axis, ps)) in axes.iter().zip(space).enumerate() {
if axis.is_empty() {
return Err(SweepError::EmptyAxis { slot });
}
for (value_index, v) in axis.iter().enumerate() {
if v.kind() != ps.kind {
return Err(SweepError::KindMismatch {
slot,
value_index,
expected: ps.kind,
got: v.kind(),
});
}
}
}
Ok(Self { axes })
}
/// The number of grid points (`∏ |axis_i|`), always `>= 1` (a valid grid has
/// no empty axis; a zero-param grid is the single empty point).
pub fn len(&self) -> usize {
self.axes.iter().map(|a| a.len()).product()
}
/// Always `false` — a valid `GridSpace` rejects empty axes, so it has at
/// least one point. Present to satisfy clippy's `len_without_is_empty`.
pub fn is_empty(&self) -> bool {
false
}
/// The cartesian product, in odometer order: the **last** axis varies
/// fastest. Deterministic — the same grid yields the same point sequence.
pub fn points(&self) -> Vec<Vec<Scalar>> {
let mut out = Vec::with_capacity(self.len());
let mut idx = vec![0usize; self.axes.len()];
loop {
out.push(self.axes.iter().zip(&idx).map(|(a, &i)| a[i]).collect());
// odometer increment from the last axis
let mut k = self.axes.len();
loop {
if k == 0 {
return out;
}
k -= 1;
idx[k] += 1;
if idx[k] < self.axes[k].len() {
break;
}
idx[k] = 0;
}
}
}
}
/// A structural fault constructing a `GridSpace` — the typed gate before any run.
#[derive(Clone, Debug, PartialEq, Eq)]
pub enum SweepError {
/// The number of axes does not equal the param-space slot count.
Arity { expected: usize, got: usize },
/// A grid value's kind does not match its slot's declared kind. `slot` is the
/// flat param-space index; `value_index` is the position within that axis.
KindMismatch { slot: usize, value_index: usize, expected: ScalarKind, got: ScalarKind },
/// A slot was given no values (would collapse the product to zero points).
EmptyAxis { slot: usize },
}
- Step 5: Run the test to verify it passes (GREEN)
Run: cargo test -p aura-engine --lib sweep::
Expected: PASS — 5 tests (points_enumerate_in_odometer_order,
len_is_product_of_axis_lengths, arity_mismatch_is_an_error,
wrong_kind_is_a_kind_mismatch, empty_axis_is_an_error).
- Step 6: Lint gate for this task
Run: cargo clippy -p aura-engine --all-targets -- -D warnings
Expected: clean (no warnings). GridSpace/SweepError are reachable via the
Step-2 re-export, so no dead-code lint fires.
Task 2: sweep() — disjoint parallel execution + the ordered family
Files:
-
Modify:
crates/aura-engine/src/sweep.rs -
Modify:
crates/aura-engine/src/lib.rs:44 -
Step 1: Replace the test-module imports + add the fixtures and run-tests (RED)
In crates/aura-engine/src/sweep.rs, replace the Task-1 test-module import lines
use super::*;
use aura_core::{ParamSpec, Scalar, ScalarKind};
with the full import block:
use super::*;
use crate::{
f64_field, summarize, BlueprintNode, Composite, Edge, OutField, ParamAlias, Role,
RunMetrics, Target,
};
use aura_core::{Firing, ParamSpec, Scalar, ScalarKind, Timestamp};
use aura_std::{Exposure, Recorder, SimBroker, Sma, Sub};
use std::sync::mpsc;
Then, inside the mod tests { … } block, after the existing
empty_axis_is_an_error test and before the module's closing }, append the
fixtures and the three execution tests:
fn synthetic_prices() -> Vec<(Timestamp, Scalar)> {
[
(1_i64, 1.0000_f64),
(2, 1.0010),
(3, 1.0030),
(4, 1.0060),
(5, 1.0040),
(6, 1.0010),
(7, 0.9990),
]
.iter()
.map(|&(t, p)| (Timestamp(t), Scalar::F64(p)))
.collect()
}
/// The SMA-cross signal as a reusable value-empty composite (duplicated from
/// `blueprint.rs`'s test module — test fixtures stay module-local, the same
/// way `report.rs` copies its harness helper).
fn sma_cross() -> Composite {
Composite::new(
"sma_cross",
vec![Sma::builder().into(), Sma::builder().into(), Sub::builder().into()],
vec![
Edge { from: 0, to: 2, slot: 0, from_field: 0 },
Edge { from: 1, to: 2, slot: 1, from_field: 0 },
],
vec![Role {
name: "price".into(),
targets: vec![Target { node: 0, slot: 0 }, Target { node: 1, slot: 0 }],
source: None,
}],
vec![
ParamAlias { name: "fast".into(), node: 0, slot: 0 },
ParamAlias { name: "slow".into(), node: 1, slot: 0 },
],
vec![OutField { node: 2, field: 0, name: "out".into() }],
)
}
/// The signal-quality harness as a value-empty composite blueprint with two
/// recording sinks (duplicated from `blueprint.rs`'s test module).
#[allow(clippy::type_complexity)]
fn composite_sma_cross_harness() -> (
Composite,
mpsc::Receiver<(Timestamp, Vec<Scalar>)>,
mpsc::Receiver<(Timestamp, Vec<Scalar>)>,
) {
let (tx_eq, rx_eq) = mpsc::channel();
let (tx_ex, rx_ex) = mpsc::channel();
let bp = Composite::new(
"root",
vec![
BlueprintNode::Composite(sma_cross()),
Exposure::builder().into(),
SimBroker::builder(0.0001).into(),
Recorder::builder(vec![ScalarKind::F64], Firing::Any, tx_eq).into(),
Recorder::builder(vec![ScalarKind::F64], Firing::Any, tx_ex).into(),
],
vec![
Edge { from: 0, to: 1, slot: 0, from_field: 0 }, // composite out -> Exposure
Edge { from: 1, to: 2, slot: 0, from_field: 0 }, // exposure -> broker slot 0
Edge { from: 2, to: 3, slot: 0, from_field: 0 }, // equity -> sink
Edge { from: 1, to: 4, slot: 0, from_field: 0 }, // exposure -> sink
],
vec![Role {
name: "src".into(),
targets: vec![Target { node: 0, slot: 0 }, Target { node: 2, slot: 1 }],
source: Some(ScalarKind::F64),
}],
vec![], // params: the interior sma_cross + Exposure carry the knobs
vec![], // output: the root ends in sinks
);
(bp, rx_eq, rx_ex)
}
/// Build + bootstrap + run + drain + summarize one grid point. A free `fn`
/// (Copy + Sync) so it serves as the `sweep` closure AND a direct reference
/// for the "sweep == N independent runs" comparison.
fn run_point(point: &[Scalar]) -> RunMetrics {
let (bp, rx_eq, rx_ex) = composite_sma_cross_harness();
let mut h = bp
.bootstrap_with_params(point.to_vec())
.expect("grid points are kind-checked against param_space");
h.run(vec![synthetic_prices()]);
let equity = f64_field(&rx_eq.try_iter().collect::<Vec<_>>(), 0);
let exposure = f64_field(&rx_ex.try_iter().collect::<Vec<_>>(), 0);
summarize(&equity, &exposure)
}
fn sma_cross_grid() -> GridSpace {
let space = composite_sma_cross_harness().0.param_space();
GridSpace::new(
&space,
vec![
vec![Scalar::I64(2), Scalar::I64(3)], // fast ∈ {2, 3}
vec![Scalar::I64(4), Scalar::I64(5)], // slow ∈ {4, 5}
vec![Scalar::F64(0.5)], // scale ∈ {0.5}
],
)
.expect("grid matches the sample param-space")
}
#[test]
fn sweep_equals_n_independent_runs() {
let grid = sma_cross_grid();
let family = sweep(&grid, run_point);
assert_eq!(family.points.len(), 4);
// params carried in enumeration (odometer) order, self-describing
assert_eq!(
family.points[0].params,
vec![Scalar::I64(2), Scalar::I64(4), Scalar::F64(0.5)],
);
assert_eq!(
family.points[3].params,
vec![Scalar::I64(3), Scalar::I64(5), Scalar::F64(0.5)],
);
// each point's metrics equal a direct, independent run of the same point:
// the sweep adds enumeration + execution, never a metrics change (C1).
for pt in &family.points {
assert_eq!(pt.metrics, run_point(&pt.params));
assert!(pt.metrics.total_pips.is_finite());
}
}
#[test]
fn family_is_deterministic_across_thread_counts() {
let grid = sma_cross_grid();
let one = sweep_with_threads(&grid, 1, run_point);
let many = sweep_with_threads(&grid, 8, run_point);
// same family at 1 worker and at N (C1: order = enumeration, not completion)
assert_eq!(one, many);
// and identical to the public `sweep`, which derives the worker count
assert_eq!(one, sweep(&grid, run_point));
}
#[test]
fn distinct_points_produce_distinct_metrics() {
let family = sweep(&sma_cross_grid(), run_point);
// not a constant family: differing SMA lengths produce differing equity
let first = family.points[0].metrics.total_pips;
assert!(
family.points.iter().any(|p| p.metrics.total_pips != first),
"a 4-point SMA-length grid must not collapse to one metric",
);
}
- Step 2: Run the tests to verify they fail (RED)
Run: cargo test -p aura-engine --lib sweep::
Expected: FAIL — compile error, cannot find function sweep /
cannot find function sweep_with_threads / cannot find type SweepPoint /
SweepFamily (the execution layer is not written yet).
- Step 3: Add the execution imports to
sweep.rs
In crates/aura-engine/src/sweep.rs, change the top production import line
use aura_core::{ParamSpec, Scalar, ScalarKind};
to add the two execution-layer imports immediately after it:
use aura_core::{ParamSpec, Scalar, ScalarKind};
use crate::RunMetrics;
use std::sync::atomic::{AtomicUsize, Ordering};
- Step 4: Write
SweepPoint,SweepFamily,sweep,sweep_with_threads(GREEN)
In crates/aura-engine/src/sweep.rs, insert this block after the
SweepError enum and before the #[cfg(test)] line:
/// One enumerated point and the metrics its run produced. Self-describing: the
/// `params` vector is the point's coordinate in `param_space()` order.
#[derive(Clone, Debug, PartialEq)]
pub struct SweepPoint {
pub params: Vec<Scalar>,
pub metrics: RunMetrics,
}
/// The ordered result family of a sweep — one `SweepPoint` per grid point, in
/// enumeration (odometer) order, independent of thread completion order.
#[derive(Clone, Debug, PartialEq)]
pub struct SweepFamily {
pub points: Vec<SweepPoint>,
}
/// Run `run_one` over every grid point, disjointly in parallel (C1), and collect
/// the family in enumeration order. `run_one` builds + bootstraps + runs +
/// summarizes one point; it shares nothing mutable, so it is `Sync` and the runs
/// are lock-free. Parallelism is `available_parallelism()` workers — `std` only,
/// no external dependency (C16).
pub fn sweep<F>(space: &GridSpace, run_one: F) -> SweepFamily
where
F: Fn(&[Scalar]) -> RunMetrics + Sync,
{
let nthreads = std::thread::available_parallelism().map(|n| n.get()).unwrap_or(1);
sweep_with_threads(space, nthreads, run_one)
}
/// The thread-count-explicit core of [`sweep`]. Module-private: the public
/// `sweep` derives the count, while the tests drive it at 1 and at N to pin
/// determinism under concurrency (C1). Workers pull point indices from a shared
/// atomic cursor (work-stealing load-balances uneven per-point cost); each tags
/// its result with the point index, and the family is assembled by sorting on
/// that index after the scope joins — so the order is the enumeration order, not
/// the completion order. Only the cursor is shared; the results side is lock-free.
fn sweep_with_threads<F>(space: &GridSpace, nthreads: usize, run_one: F) -> SweepFamily
where
F: Fn(&[Scalar]) -> RunMetrics + Sync,
{
let points = space.points();
// `points.len() >= 1` always (GridSpace rejects empty axes), so the clamp
// lower bound never exceeds the upper; this also coerces a 0 to 1 worker.
let nthreads = nthreads.clamp(1, points.len());
let cursor = AtomicUsize::new(0);
let mut results: Vec<(usize, RunMetrics)> = std::thread::scope(|scope| {
let handles: Vec<_> = (0..nthreads)
.map(|_| {
scope.spawn(|| {
let mut local: Vec<(usize, RunMetrics)> = Vec::new();
loop {
let i = cursor.fetch_add(1, Ordering::Relaxed);
if i >= points.len() {
break;
}
local.push((i, run_one(&points[i])));
}
local
})
})
.collect();
handles.into_iter().flat_map(|h| h.join().unwrap()).collect()
});
results.sort_by_key(|&(i, _)| i);
SweepFamily {
points: results
.into_iter()
.map(|(i, metrics)| SweepPoint { params: points[i].clone(), metrics })
.collect(),
}
}
- Step 5: Expand the
lib.rsre-export
In crates/aura-engine/src/lib.rs, replace the Task-1 partial re-export line
pub use sweep::{GridSpace, SweepError};
with the full re-export:
pub use sweep::{sweep, GridSpace, SweepError, SweepFamily, SweepPoint};
- Step 6: Run the tests to verify they pass (GREEN)
Run: cargo test -p aura-engine --lib sweep::
Expected: PASS — 8 tests total (the 5 enumeration/fault tests from Task 1 plus
sweep_equals_n_independent_runs, family_is_deterministic_across_thread_counts,
distinct_points_produce_distinct_metrics).
- Step 7: Full workspace build + test + lint gate
Run: cargo build --workspace
Expected: success.
Run: cargo test --workspace
Expected: PASS — the whole suite green, including the new sweep:: tests and all
pre-existing tests (no regression).
Run: cargo clippy --workspace --all-targets -- -D warnings
Expected: clean (no warnings).