//! `aura` — the programmatic / CLI face of the engine (the surface the LLM and //! automation drive: author a node, run a sim/sweep, emit structured metrics). //! //! The walking skeleton's closing seam: `aura run` bootstraps a built-in sample //! signal-quality harness (synthetic source → SMA-cross → Bias → SimBroker → //! recording sinks), runs it deterministically (C1), and prints the run's //! metrics + manifest (#6) as canonical JSON to stdout (the headline C14 move). //! //! `aura run --real [--from ] [--to ]` feeds that same harness //! real M1 close bars, streamed lazily from the local data-server archive through //! the #71 Source seam (`M1FieldSource`) instead of the synthetic stream — the //! first real-data backtest from the CLI. mod render; mod graph_construct; mod project; mod campaign_run; mod research_docs; mod scaffold; use render::{ChartData, ChartMeta, ChartMode, ReduceKind, Series}; use aura_core::{zip_params, Cell, Firing, ParamSpec, Scalar, ScalarKind, Timestamp}; use aura_composites::{cost_graph, risk_executor, risk_executor_vol_open, StopRule}; use aura_engine::{ blueprint_from_json, blueprint_to_json, f64_field, join_on_ts, monte_carlo, param_stability, r_bootstrap, r_metrics_from_rs, summarize, summarize_r, walk_forward, window_of, BindError, BlueprintNode, ColumnarTrace, Composite, Edge, FamilySelection, FlatGraph, GraphBuilder, Harness, JoinedRow, McAggregate, McFamily, RBootstrap, RollMode, RunManifest, RunMetrics, RunReport, SelectionMode, SourceSpec, SweepFamily, SweepPoint, SyntheticSpec, Target, VecSource, WalkForwardResult, WindowBounds, WindowRoller, WindowRun, }; use aura_registry::{ check_r_metric, generalization, group_families, mc_member_reports, optimize_deflated, optimize_plateau, rank_by, sweep_member_reports, walkforward_member_reports, FamilyKind, FamilyMember, Generalization, NameKind, PlateauMode, Registry, RunTraces, WriteKind, }; use aura_std::{ Add, Bias, CarryCost, ConstantCost, Delay, Ema, GatedRecorder, Gt, Latch, LinComb, LongOnly, Mul, Recorder, RollingMax, RollingMin, SeriesReducer, SimBroker, Sma, Sqrt, Sub, VolSlippageCost, PM_FIELD_NAMES, PM_RECORD_KINDS, }; // `std_vocabulary` is now only reached through `project::Env::resolve` in production // code; the test module still builds reference blueprints against it directly, so // the import is test-only. #[cfg(test)] use aura_std::std_vocabulary; use std::sync::mpsc::{self, Receiver}; use std::sync::LazyLock; use std::collections::HashSet; use clap::{Args, Parser, Subcommand}; /// The pip size the built-in *synthetic* harnesses run at: a 5-decimal FX major /// (`EURUSD`-shaped). The synthetic streams carry no instrument, so there is no /// recorded geometry sidecar to thread; a single named source keeps the broker's /// divisor (`sample_harness` / `SimBroker::builder`) and the recorded broker label /// (`sim_optimal_manifest`) in lockstep, so they cannot silently drift apart. /// The real path threads the sidecar's looked-up `pip_size` instead. const SYNTHETIC_PIP_SIZE: f64 = 0.0001; /// Real walk-forward roller sizes (Fork D/F). `WindowRoller` takes sizes in the /// stream's epoch-unit; for real M1 that is nanoseconds. A classic 3-month /// in-sample / 1-month out-of-sample / 1-month step (contiguous OOS tiling). const WF_DAY_NS: i64 = 86_400_000_000_000; const WF_REAL_IS_NS: i64 = 90 * WF_DAY_NS; const WF_REAL_OOS_NS: i64 = 30 * WF_DAY_NS; const WF_REAL_STEP_NS: i64 = 30 * WF_DAY_NS; /// Trials-deflation resampling budget for walk-forward winner selection. Recorded /// on each winner's manifest (so `overfit_probability` is reproducible by re-run); /// CLI flags for these are a deferred refinement. const DEFLATION_N_RESAMPLES: usize = 1000; const DEFLATION_BLOCK_LEN: usize = 5; const DEFLATION_SEED: u64 = 0xDEF1_A7ED; /// The built-in synthetic price stream: rises through t=4 then reverses, so the /// demo trace carries one exposure sign flip and a real drawdown (C22 populated /// trace). Deterministic and fixed (C1). 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() } /// A warm-up-adequate synthetic stream for the enriched sample/sweep: ~18 ticks /// rising, falling, then rising again so the trend SMA spread and the MACD /// EMA-of-EMA histogram both warm up and flip sign. It shares the proven warm-up /// profile of `macd_prices` (the enriched sample embeds the same `macd` composite, /// so it needs the same warm-up length); the flat `run_sample` keeps the shorter /// `synthetic_prices`. Deterministic and fixed (C1). fn showcase_prices() -> Vec<(Timestamp, Scalar)> { [ 1.0000_f64, 1.0008, 1.0021, 1.0039, 1.0062, 1.0090, 1.0083, 1.0061, 1.0034, 1.0012, 0.9998, 1.0006, 1.0024, 1.0047, 1.0069, 1.0086, 1.0097, 1.0092, ] .iter() .enumerate() .map(|(i, &p)| (Timestamp(i as i64 + 1), Scalar::f64(p))) .collect() } /// Bootstrap the sample signal-quality harness with two recording sinks (equity /// tapped on the SimBroker, exposure tapped on the Bias node). Rust-authored /// wiring (C17/C20) over the raw bootstrap API — no builder DSL this cycle. The /// price taps both SMAs and the broker's price slot (slot 1); exposure feeds the /// broker's slot 0 (slot order is load-bearing — both are f64). // The harness-plus-two-drained-sink-receivers tuple has exactly one call site // (`run_sample`); a named type would be speculative abstraction this cycle. #[allow(clippy::type_complexity)] fn sample_harness(pip_size: f64) -> ( Harness, Receiver<(Timestamp, Vec)>, Receiver<(Timestamp, Vec)>, ) { let (tx_eq, rx_eq) = mpsc::channel(); let (tx_ex, rx_ex) = mpsc::channel(); let f64_recorder_sig = || aura_engine::NodeSchema { inputs: vec![aura_engine::PortSpec { kind: ScalarKind::F64, firing: Firing::Any, name: "in".into() }], output: vec![], params: vec![], }; let h = Harness::bootstrap(FlatGraph { nodes: vec![ Box::new(Sma::new(2)), // 0 fast SMA Box::new(Sma::new(4)), // 1 slow SMA Box::new(Sub::new()), // 2 spread Box::new(Bias::new(0.5)), // 3 bias Box::new(SimBroker::new(pip_size)), // 4 sim-optimal broker Box::new(Recorder::new(&[ScalarKind::F64], Firing::Any, tx_eq)), // 5 equity sink Box::new(Recorder::new(&[ScalarKind::F64], Firing::Any, tx_ex)), // 6 exposure sink ], signatures: vec![ Sma::builder().schema().clone(), Sma::builder().schema().clone(), Sub::builder().schema().clone(), Bias::builder().schema().clone(), SimBroker::builder(pip_size).schema().clone(), f64_recorder_sig(), f64_recorder_sig(), ], sources: vec![SourceSpec { kind: ScalarKind::F64, targets: vec![ Target { node: 0, slot: 0 }, Target { node: 1, slot: 0 }, Target { node: 4, slot: 1 }, // price into the broker's price slot ], }], edges: vec![ Edge { from: 0, to: 2, slot: 0, from_field: 0 }, Edge { from: 1, to: 2, slot: 1, from_field: 0 }, Edge { from: 2, to: 3, slot: 0, from_field: 0 }, Edge { from: 3, to: 4, slot: 0, from_field: 0 }, // exposure into broker slot 0 Edge { from: 4, to: 5, slot: 0, from_field: 0 }, // equity -> sink 5 Edge { from: 3, to: 6, slot: 0, from_field: 0 }, // exposure -> sink 6 ], }) .expect("valid sample signal-quality DAG"); (h, rx_eq, rx_ex) } /// Build the sim-optimal `RunManifest`: the engine-external descriptor fields /// (commit, seed-free synthetic run, broker label) are constant across the CLI's /// built-in harnesses — only `params` and `window` vary. Centralizing the broker /// label keeps it a single source (and a single edit point for per-asset pip /// work): the `pip_size` it renders is the one its caller ran the broker at. fn sim_optimal_manifest( params: Vec<(String, Scalar)>, window: (Timestamp, Timestamp), seed: u64, pip_size: f64, ) -> RunManifest { // Typed params pass straight through: the manifest carries self-describing // Scalars, so a length stays `i64` and a scale `f64` in the record. RunManifest { commit: option_env!("AURA_COMMIT").unwrap_or("unknown").to_string(), params, window, seed, broker: format!("sim-optimal(pip_size={pip_size})"), selection: None, instrument: None, topology_hash: None, project: None, } } /// Persist a run's drained taps to the on-disk trace store under `runs/traces//` /// (beside the run registry's `runs/`). Shared by every run form — all drain the same /// two f64 taps (equity off the broker, exposure off the strategy) and carry a /// `RunManifest`. Pure wiring over `ColumnarTrace` + `TraceStore`; the engine is /// untouched. An I/O failure is a runtime error (stderr + exit 1), per the spec. fn persist_traces( name: &str, manifest: &RunManifest, eq_rows: &[(Timestamp, Vec)], ex_rows: &[(Timestamp, Vec)], env: &project::Env, ) { let taps = vec![ ColumnarTrace::from_rows("equity", &[ScalarKind::F64], eq_rows), ColumnarTrace::from_rows("exposure", &[ScalarKind::F64], ex_rows), ]; if let Err(e) = env.trace_store().write(name, manifest, &taps) { eprintln!("aura: trace persist failed: {e}"); std::process::exit(1); } } /// The maximum length (bytes) of an on-disk member-key path component. Comfortably /// under the 255-byte POSIX `NAME_MAX` and inside Windows' 260-char `MAX_PATH` /// once the `runs/traces//` prefix and `/.json` suffix are added. const MAX_KEY: usize = 200; /// Map any byte outside the portable directory-name charset `[A-Za-z0-9._-]` to /// `_`. The single source of filesystem-portability for an on-disk path component /// (valid on Linux / Windows / macOS, also URL-path- and cloud-sync-safe). fn sanitize_component(s: &str) -> String { s.chars() .map(|c| if c.is_ascii_alphanumeric() || matches!(c, '.' | '_' | '-') { c } else { '_' }) .collect() } /// Render a scalar value case-lessly: integers/timestamps as decimal digits, bool /// as `true`/`false`, f64 via Rust's `Display` (decimal, shortest round-trip, NO /// scientific notation). Case-less rendering is what keeps two members of one /// family from ever differing only by letter case (case-insensitive-FS safety). fn render_value(v: &Scalar) -> String { match v { Scalar::I64(n) => n.to_string(), Scalar::F64(x) => x.to_string(), Scalar::Bool(b) => b.to_string(), Scalar::Timestamp(t) => t.0.to_string(), } } /// FNV-1a-64 over `bytes` — a fixed, version-stable, non-cryptographic hash used /// only as the over-cap member-key disambiguator (`std`'s `DefaultHasher` is /// explicitly unstable across releases, so it cannot name an on-disk artefact). fn fnv1a64(bytes: &[u8]) -> u64 { let mut h: u64 = 0xcbf2_9ce4_8422_2325; for &b in bytes { h ^= b as u64; h = h.wrapping_mul(0x0000_0100_0000_01b3); } h } /// The portable, collision-free member key for a swept grid point: one `name-value` /// token per *varying* axis (in `named`'s param-space slot order), joined by `_`, /// every token sanitised to the portable charset. Pinned (singleton) axes carry no /// information and are omitted. A 1-point grid (no varying axis) keys as `"m"`. A /// key over `MAX_KEY` degrades to a conformant `h-<16-hex>` FNV fallback so the /// key is one valid path component for ANY grid (the #105 generalisation). fn member_key(named: &[(String, Scalar)], varying: &HashSet) -> String { let key: String = named .iter() .filter(|(n, _)| varying.contains(n)) .map(|(n, v)| format!("{}-{}", sanitize_component(n), sanitize_component(&render_value(v)))) .collect::>() .join("_"); let key = if key.is_empty() { "m".to_string() } else { key }; if key.len() <= MAX_KEY { key } else { format!("h-{:016x}", fnv1a64(key.as_bytes())) } } /// Default decimation budget: target horizontal buckets. ~2000 buckets ⇒ ≤ ~4000 /// spine slots (min+max per bucket) — a few-thousand-point page regardless of the /// underlying multi-year M1 point count. const CHART_DECIMATE_BUCKETS: usize = 2000; /// Per-tap decimation kind (#111): the bounded exposure stream (C10, f64 ∈ [-1,+1]) /// reduces by per-bucket mean, so its net/duty-cycle level survives decimation /// instead of collapsing to a -1..+1 band (every bucket of a multi-year exposure /// straddles many sign flips, so min/max would be ±1 everywhere). An unbounded /// cumulative curve (equity) keeps the min/max envelope so drawdowns survive. Keyed /// on the tap name — `exposure` is the only bounded level tap today. fn reduce_for_tap(tap: &str) -> ReduceKind { if tap == "exposure" { ReduceKind::Mean } else { ReduceKind::MinMax } } /// Serve-time decimation on the aligned `ChartData` (#108). Partition the shared /// `xs` into at most `buckets` contiguous index ranges; per non-empty bucket emit the /// bucket's first (and, if it spans >1 index, last) timestamp as shared spine slots, /// and reduce each series per its [`ReduceKind`] (#111): a `MinMax` series emits min /// then max (the envelope — equity drawdowns survive), a `Mean` series emits the /// per-bucket mean in both slots (the net level — a bounded exposure shows its /// duty-cycle instead of a -1..+1 band). An all-null bucket emits null. `meta` passes /// through unchanged. Deterministic (C1). Full data stays on disk; only the served /// page is thinned. No-op when `xs.len() <= 2 * buckets`. fn decimate(data: ChartData, buckets: usize) -> ChartData { let buckets = buckets.max(1); let n = data.xs.len(); if n <= 2 * buckets { return data; } let ChartData { xs, series, meta } = data; // Bucket index bounds (lo, hi_exclusive, two_slots) + the decimated shared spine. // xs is sorted+deduped (strictly increasing) -> boundary timestamps are strictly // increasing across and within buckets, so the spine stays monotonic for uPlot. let mut bounds: Vec<(usize, usize, bool)> = Vec::with_capacity(buckets); let mut out_xs: Vec = Vec::with_capacity(2 * buckets); for b in 0..buckets { let lo = b * n / buckets; let hi = (b + 1) * n / buckets; if lo >= hi { continue; } let two = hi - lo > 1; out_xs.push(xs[lo]); if two { out_xs.push(xs[hi - 1]); } bounds.push((lo, hi, two)); } let out_series: Vec = series .into_iter() .map(|s| { let mut points: Vec> = Vec::with_capacity(out_xs.len()); for &(lo, hi, two) in &bounds { let (first, second) = match s.reduce { ReduceKind::MinMax => { // envelope: min at the first slot, max at the second. let mut mn = f64::INFINITY; let mut mx = f64::NEG_INFINITY; let mut any = false; for v in s.points[lo..hi].iter().flatten() { any = true; if *v < mn { mn = *v; } if *v > mx { mx = *v; } } if any { (Some(mn), Some(mx)) } else { (None, None) } } ReduceKind::Mean => { // net level: the per-bucket mean written to both slots (a flat // step), so a bounded high-flip series shows its duty-cycle // instead of a -1..+1 band (#111). let mut sum = 0.0; let mut cnt = 0u32; for v in s.points[lo..hi].iter().flatten() { sum += *v; cnt += 1; } let m = if cnt > 0 { Some(sum / cnt as f64) } else { None }; (m, m) } }; points.push(first); if two { points.push(second); } } Series { name: s.name, y_scale_id: s.y_scale_id, points, reduce: s.reduce } }) .collect(); ChartData { xs: out_xs, series: out_series, meta } } /// Build the serve-ready `ChartData` from a run's read-back traces by the spec-§6 /// 3-step union-spine alignment — no tap privileged, no point dropped: /// (1) xs = the sorted, deduped union of every tap's timestamps; /// (2) synthesize an empty-payload spine over xs and pass ALL taps (via /// `ColumnarTrace::to_rows`, which yields uniformly-f64 rows) as symmetric sides /// of `join_on_ts`, so no side row is dropped and none occupies the privileged /// `JoinedRow.spine`; /// (3) flatten each (tap, column) to a `Series` of `Option` over xs. fn build_chart_data(name: &str, traces: RunTraces) -> ChartData { let mut xs: Vec = traces.taps.iter().flat_map(|t| t.ts.iter().copied()).collect(); xs.sort_unstable(); xs.dedup(); let spine: Vec<(Timestamp, Vec)> = xs.iter().map(|&t| (Timestamp(t), Vec::new())).collect(); let tap_rows: Vec)>> = traces.taps.iter().map(|t| t.to_rows()).collect(); let sides: Vec<&[(Timestamp, Vec)]> = tap_rows.iter().map(|r| r.as_slice()).collect(); let joined: Vec = join_on_ts(&spine, &sides); let mut series: Vec = Vec::new(); for (i, tap) in traces.taps.iter().enumerate() { for c in 0..tap.columns.len() { let name = if tap.columns.len() == 1 { tap.tap.clone() } else { format!("{}[{c}]", tap.tap) }; let y_scale_id = format!("y_{}", series.len()); let points: Vec> = joined.iter().map(|r| r.sides[i].as_ref().map(|row| row[c].as_f64())).collect(); series.push(Series { name, y_scale_id, points, reduce: reduce_for_tap(&tap.tap) }); } } let m = &traces.manifest; let meta = ChartMeta { kind: "run".to_string(), name: name.to_string(), commit: m.commit.clone(), window: (m.window.0.0, m.window.1.0), broker: m.broker.clone(), seed: m.seed, taps: traces.taps.iter().map(|t| t.tap.clone()).collect(), members: None, params: m.params.iter().map(|(k, v)| (k.clone(), render_value(v))).collect(), }; ChartData { xs, series, meta } } /// One member's contribution to the comparison build: its key (the future series /// name) paired with the chosen tap's drained `(ts, row)` pairs. type MemberRows = (String, Vec<(Timestamp, Vec)>); /// Build the comparison `ChartData` for a family: one `Series` per member (the /// chosen `tap`'s column), labelled by `member.key`, ALL sharing ONE `y_scale_id` /// (the members measure one identical quantity, so a shared scale is what makes /// them comparable — unlike the single-run overlay, whose series are different /// taps). Aligned on the union-ts spine via the same `join_on_ts` build_chart_data /// uses. `Err` if NO member carries `tap` (refuse-don't-guess). fn build_comparison_chart_data( name: &str, members: &[FamilyMember], tap: &str, ) -> Result { let mut member_rows: Vec = Vec::new(); for m in members { if let Some(t) = m.traces.taps.iter().find(|t| t.tap == tap) { member_rows.push((m.key.clone(), t.to_rows())); } } if member_rows.is_empty() { return Err(format!("no family member has a tap named '{tap}'")); } let mut xs: Vec = member_rows.iter().flat_map(|(_, r)| r.iter().map(|(t, _)| t.0)).collect(); xs.sort_unstable(); xs.dedup(); let spine: Vec<(Timestamp, Vec)> = xs.iter().map(|&t| (Timestamp(t), Vec::new())).collect(); let sides: Vec<&[(Timestamp, Vec)]> = member_rows.iter().map(|(_, r)| r.as_slice()).collect(); let joined: Vec = join_on_ts(&spine, &sides); // One shared y-scale across all member series (same quantity). let y_scale_id = format!("y_cmp_{tap}"); let mut series: Vec = Vec::new(); for (i, (key, _)) in member_rows.iter().enumerate() { // Project column 0 — the doc's "chosen tap's column" (singular). The // comparison taps in scope (equity / exposure) are single-column `f64`. // A future multi-column tap selection would need a column index here. let points: Vec> = joined.iter().map(|r| r.sides[i].as_ref().map(|row| row[0].as_f64())).collect(); series.push(Series { name: key.clone(), y_scale_id: y_scale_id.clone(), points, reduce: reduce_for_tap(tap) }); } // member_rows is non-empty here (checked above) => members is non-empty, so // members[0] is safe. commit/broker ARE shared across a family (one frozen // artifact, one broker profile), but the window is NOT: a walk-forward family's // members are disjoint OOS windows (commit 4c64feb), so the family window is the // SPAN across all members — (min from, max to). For sweep/MC, whose members // share one window, the span collapses to that shared window, so this is the one // correct reading for all three kinds. let m = &members[0].traces.manifest; let window = ( members.iter().map(|fm| fm.traces.manifest.window.0.0).min().unwrap(), members.iter().map(|fm| fm.traces.manifest.window.1.0).max().unwrap(), ); let meta = ChartMeta { kind: "family".to_string(), name: name.to_string(), commit: m.commit.clone(), window, broker: m.broker.clone(), seed: m.seed, taps: vec![tap.to_string()], members: Some(members.len()), params: Vec::new(), }; Ok(ChartData { xs, series, meta }) } /// Restrict a single-run `ChartData` to the one series named `tap`. `Err` if the /// run has no such tap (refuse-don't-guess). Used by the `--tap` flag on the /// single-run chart path; without `--tap` the single-run page is unchanged. fn filter_to_tap(data: ChartData, tap: &str) -> Result { if !data.series.iter().any(|s| s.name == tap) { // List the valid taps so the user can correct a typo (#131) — refuse, but help. let available: Vec<&str> = data.series.iter().map(|s| s.name.as_str()).collect(); return Err(format!("run has no tap named '{tap}' (available: {})", available.join(", "))); } let series: Vec = data.series.into_iter().filter(|s| s.name == tap).collect(); let mut meta = data.meta; meta.taps = vec![tap.to_string()]; Ok(ChartData { xs: data.xs, series, meta }) } /// `aura chart [--tap ] [--panels]`: classify the name and render. A /// single run charts all its taps (or the one `--tap` selects); a family overlays /// one tap (default `equity`) across its members; an unknown name is a runtime error /// (stderr + exit 1), never a panic. fn emit_chart(name: &str, tap: Option<&str>, mode: ChartMode, env: &project::Env) { let store = env.trace_store(); match store.name_kind(name) { NameKind::Run => { let traces = match store.read(name) { Ok(t) => t, Err(e) => { eprintln!("aura: {e}"); std::process::exit(1); } }; let mut data = build_chart_data(name, traces); if let Some(t) = tap { data = match filter_to_tap(data, t) { Ok(d) => d, Err(e) => { eprintln!("aura: {e}"); std::process::exit(1); } }; } let data = decimate(data, CHART_DECIMATE_BUCKETS); print!("{}", render::render_chart_html(&data, mode)); } NameKind::Family => { let members = match store.read_family(name) { Ok(m) => m, Err(e) => { eprintln!("aura: {e}"); std::process::exit(1); } }; let data = match build_comparison_chart_data(name, &members, tap.unwrap_or("equity")) { Ok(d) => d, Err(e) => { eprintln!("aura: {e}"); std::process::exit(1); } }; let data = decimate(data, CHART_DECIMATE_BUCKETS); print!("{}", render::render_chart_html(&data, mode)); } NameKind::NotFound => { eprintln!( "aura: no recorded run or family '{name}' under runs/traces \ (run `aura run --trace {name}` or `aura sweep --trace {name}` first)" ); std::process::exit(1); } } } /// Run the sample harness and fold it into a `RunReport` (drain both sinks → /// `f64_field` → `summarize` → pair with a `RunManifest`). Pure and deterministic /// (C1): the same build yields the same report. fn run_sample(trace: Option<&str>, env: &project::Env) -> RunReport { if let Some(n) = trace && let Err(e) = env.trace_store().ensure_name_free(n, WriteKind::Run) { eprintln!("aura: {e}"); std::process::exit(1); } let (mut h, rx_eq, rx_ex) = sample_harness(SYNTHETIC_PIP_SIZE); let sources: Vec> = vec![Box::new(VecSource::new(synthetic_prices()))]; let window = window_of(&sources).expect("non-empty synthetic stream"); h.run(sources); let eq_rows: Vec<(Timestamp, Vec)> = rx_eq.try_iter().collect(); let ex_rows: Vec<(Timestamp, Vec)> = rx_ex.try_iter().collect(); let manifest = sim_optimal_manifest( vec![ ("sma_fast".to_string(), Scalar::i64(2)), ("sma_slow".to_string(), Scalar::i64(4)), ("bias_scale".to_string(), Scalar::f64(0.5)), ], window, 0, SYNTHETIC_PIP_SIZE, ); if let Some(name) = trace { persist_traces(name, &manifest, &eq_rows, &ex_rows, env); } let metrics = summarize(&f64_field(&eq_rows, 0), &f64_field(&ex_rows, 0)); RunReport { manifest, metrics } } /// `aura run --real `: run the built-in sample harness over real M1 close /// bars streamed lazily from the local data-server archive through the #71 Source /// seam (`M1FieldSource`, a `Box`), not the synthetic `VecSource`. Same /// fold as `run_sample`. The manifest window is read from a *separate* probe source /// (a Source is single-pass), so the run source streams the window untouched. /// A no-local-data condition (unknown symbol, or a window overlapping no file / no /// bars) is a runtime error: stderr + exit(1), not a panic. fn run_sample_real( symbol: &str, from_ms: Option, to_ms: Option, trace: Option<&str>, env: &project::Env, ) -> RunReport { if let Some(n) = trace && let Err(e) = env.trace_store().ensure_name_free(n, WriteKind::Run) { eprintln!("aura: {e}"); std::process::exit(1); } let (source, window, pip_size) = open_real_source(symbol, from_ms, to_ms, env); let (mut h, rx_eq, rx_ex) = sample_harness(pip_size); h.run(vec![source]); let eq_rows: Vec<(Timestamp, Vec)> = rx_eq.try_iter().collect(); let ex_rows: Vec<(Timestamp, Vec)> = rx_ex.try_iter().collect(); let manifest = sim_optimal_manifest( vec![ ("sma_fast".to_string(), Scalar::i64(2)), ("sma_slow".to_string(), Scalar::i64(4)), ("bias_scale".to_string(), Scalar::f64(0.5)), ], window, 0, pip_size, ); if let Some(name) = trace { persist_traces(name, &manifest, &eq_rows, &ex_rows, env); } let metrics = summarize(&f64_field(&eq_rows, 0), &f64_field(&ex_rows, 0)); RunReport { manifest, metrics } } /// What `--real` parsing yields: the synthetic default, or a real symbol + an /// optional window (parsed, not yet opened). Pure, so the grammar is unit-testable. #[derive(Debug, Clone, PartialEq)] enum DataChoice { Synthetic, Real { symbol: String, from_ms: Option, to_ms: Option }, } /// The source provider threaded into the family builders: synthetic built-in /// streams, or real M1 close bars from the data-server archive. Replaces the /// hardcoded `VecSource` so a member's source, pip, window, and roller sizes come /// from one place (Fork B/D/F). /// /// `Synthetic` denotes a *consumer-dependent* built-in stream, not one fixed /// series: the full-window consumers (`full_window` / `run_sources`, used by /// sweep / MC) draw the 18-bar `showcase_prices()`, while the windowed consumers /// (`windowed_sources` / `wf_window_sizes`, used by walk-forward) draw the 60-bar /// `walkforward_prices()` so the `(24,12,12)`-bar roller fits its span. The two /// faces never reach one consumer (a family is either full-window or windowed), so /// the split is invisible per call site but real across the type — read both /// family builders to see it whole. enum DataSource { Synthetic, Real { server: std::sync::Arc, symbol: String, from_ms: Option, to_ms: Option, pip: f64, }, } /// No-local-data refusal — stderr + exit(1), mirroring `run_sample_real`. fn no_real_data(symbol: &str, env: &project::Env) -> ! { eprintln!("aura: no local data for symbol '{symbol}' at {}", env.data_path()); std::process::exit(1) } /// Resolve the per-instrument pip from the recorded geometry sidecar, or refuse /// (stderr + exit 1) when the symbol has no recorded geometry — the single home of /// the guessed-pip refusal, shared by `open_real_source` and `from_choice`. Reads /// the symbol's geometry metadata (not bar data) before the run source opens, so an /// instrument with no recorded geometry refuses without a guessed pip; the pip is /// the provider's recorded value, honest by construction. fn pip_or_refuse( server: &std::sync::Arc, symbol: &str, env: &project::Env, ) -> f64 { match aura_ingest::instrument_geometry(server, symbol) { Some(geo) => geo.pip_size, None => { eprintln!( "aura: no recorded geometry for symbol '{symbol}' at {} — \ refusing to run a real instrument with a guessed pip", env.data_path() ); std::process::exit(1); } } } /// Probe the full data window: open a single-pass probe `M1FieldSource`, drain it /// for the first/last timestamp, and return `(first, last)`. Refuses (via /// `no_real_data`) when the symbol/window yields no source or no bars. Shared by /// `run_sample_real` (which needs the manifest window from a probe separate from /// the run source) and `DataSource::full_window`. fn probe_window( server: &std::sync::Arc, symbol: &str, from_ms: Option, to_ms: Option, env: &project::Env, ) -> (Timestamp, Timestamp) { let mut probe = aura_ingest::M1FieldSource::open(server, symbol, from_ms, to_ms, aura_ingest::M1Field::Close) .unwrap_or_else(|| no_real_data(symbol, env)); let first = aura_engine::Source::peek(&probe).unwrap_or_else(|| no_real_data(symbol, env)); let mut last = first; while let Some((t, _)) = aura_engine::Source::next(&mut probe) { last = t; } (first, last) } /// Open a real M1-close source for a recorded symbol over an optional window, returning /// the run source paired with its manifest `window` and per-instrument `pip_size`. /// Single home of the real-source construction the single-run handlers share — the /// sidecar-pip lookup, the `DataServer` `has_symbol` refusal, the probe-window pass, and /// the run-source `open` (each refusal an stderr + exit 1). Pre-data refusals keep the /// pip honest by construction. Shared by `run_sample_real` and `run_r_sma`. fn open_real_source( symbol: &str, from_ms: Option, to_ms: Option, env: &project::Env, ) -> (Box, (Timestamp, Timestamp), f64) { // Per-instrument pip from the recorded sidecar; resolved BEFORE bar-data access // so an instrument with no geometry refuses without touching the archive. let server = std::sync::Arc::new(data_server::DataServer::new(env.data_path())); let pip = pip_or_refuse(&server, symbol, env); if !server.has_symbol(symbol) { no_real_data(symbol, env); } // Manifest window: drain a separate probe (single-pass Source) for first/last ts. let window = probe_window(&server, symbol, from_ms, to_ms, env); let source: Box = match aura_ingest::M1FieldSource::open(&server, symbol, from_ms, to_ms, aura_ingest::M1Field::Close) { Some(s) => Box::new(s), None => no_real_data(symbol, env), }; (source, window, pip) } impl DataSource { /// Build a provider from a parsed choice, or refuse (stderr + exit 1) on a symbol /// with no recorded geometry / absent data — both BEFORE any member runs (Fork C/G), /// via the same `pip_or_refuse` / `no_real_data` helpers as /// `run_sample_real`. fn from_choice(choice: DataChoice, env: &project::Env) -> DataSource { match choice { DataChoice::Synthetic => DataSource::Synthetic, DataChoice::Real { symbol, from_ms, to_ms } => { let server = std::sync::Arc::new(data_server::DataServer::new(env.data_path())); let pip = pip_or_refuse(&server, &symbol, env); if !server.has_symbol(&symbol) { no_real_data(&symbol, env); } DataSource::Real { server, symbol, from_ms, to_ms, pip } } } } fn pip_size(&self) -> f64 { match self { DataSource::Synthetic => SYNTHETIC_PIP_SIZE, DataSource::Real { pip, .. } => *pip, } } /// The full run window, probed once. Synthetic: the showcase span. Real: /// `probe_window` drains a separate single-pass probe source for first/last ts /// (the same helper `run_sample_real` uses for its manifest window). fn full_window(&self, env: &project::Env) -> (Timestamp, Timestamp) { match self { DataSource::Synthetic => { let s: Vec> = vec![Box::new(VecSource::new(showcase_prices()))]; window_of(&s).expect("non-empty showcase stream") } DataSource::Real { server, symbol, from_ms, to_ms, .. } => { probe_window(server, symbol, *from_ms, *to_ms, env) } } } /// The full walk-forward span. Synthetic draws the 60-bar `walkforward_prices` /// span — NOT `showcase_prices` (which `full_window` uses): walk-forward is a /// *windowed* consumer whose roller `(24,12,12)` needs 36 bars, so it uses the /// longer built-in stream (byte-unchanged from the pre-`DataSource` /// `walkforward_family`, which derived its span the same way). Real: the same /// probed `--from..--to` window as `full_window`. fn wf_full_span(&self, env: &project::Env) -> (Timestamp, Timestamp) { match self { DataSource::Synthetic => { let s: Vec> = vec![Box::new(VecSource::new(walkforward_prices()))]; window_of(&s).expect("non-empty walkforward stream") } DataSource::Real { server, symbol, from_ms, to_ms, .. } => { probe_window(server, symbol, *from_ms, *to_ms, env) } } } /// A fresh full-window source per member (single-pass). Synthetic: showcase. fn run_sources(&self, env: &project::Env) -> Vec> { match self { DataSource::Synthetic => vec![Box::new(VecSource::new(showcase_prices()))], DataSource::Real { server, symbol, from_ms, to_ms, .. } => vec![Box::new( aura_ingest::M1FieldSource::open(server, symbol, *from_ms, *to_ms, aura_ingest::M1Field::Close) .unwrap_or_else(|| no_real_data(symbol, env)), )], } } /// A fresh windowed source for an IS/OOS sub-window (walk-forward). Synthetic: /// `walkforward_window_source`. Real: `open_window` (ns-native `Timestamp`). fn windowed_sources( &self, from: Timestamp, to: Timestamp, env: &project::Env, ) -> Vec> { match self { DataSource::Synthetic => vec![Box::new(walkforward_window_source(from, to))], DataSource::Real { server, symbol, .. } => vec![Box::new( aura_ingest::M1FieldSource::open_window(server, symbol, Some(from), Some(to), aura_ingest::M1Field::Close) .unwrap_or_else(|| no_real_data(symbol, env)), )], } } /// WindowRoller sizes per data kind (Fork F): bar-index for synthetic (24/12/12 /// over the 60-bar span), calendar-ns for real. fn wf_window_sizes(&self) -> (i64, i64, i64) { match self { DataSource::Synthetic => (24, 12, 12), DataSource::Real { .. } => (WF_REAL_IS_NS, WF_REAL_OOS_NS, WF_REAL_STEP_NS), } } /// The built-in strategy's length grid, **per data kind**. Synthetic keeps the /// short lengths that fit the 18/60-bar demo streams (a 200-bar MA would never /// warm on a 60-bar stream); real uses realistic M1 lengths so the SMA-cross + /// MACD signal is a real trend cross over tens of thousands of bars, not noise. /// Returns `(trend_fast grid, trend_slow grid, (macd_fast, macd_slow, macd_signal))`; /// only the two trend axes vary (a 2×2 sweep), the MACD lengths are pinned. fn strategy_lengths(&self) -> ([i64; 2], [i64; 2], (i64, i64, i64)) { match self { DataSource::Synthetic => ([2, 3], [4, 5], (2, 4, 3)), // 50/200-style intraday crosses on M1 (minutes), standard 12/26/9 MACD. DataSource::Real { .. } => ([50, 100], [200, 400], (12, 26, 9)), } } } /// The SMA-cross signal as a named composite (price -> fast/slow SMA -> spread). /// CLI-local sample builder; the engine ships no sample (the duplication with /// `blueprint.rs`'s test helper is the dedup tracked in #14). Value-empty: the SMA /// lengths are injected at compile, not baked here. fn sma_cross(name: &str) -> Composite { let mut g = GraphBuilder::new(name); let fast = g.add(Sma::builder().named("fast")); // fast SMA leg let slow = g.add(Sma::builder().named("slow")); // slow SMA leg 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"), "cross"); g.build().expect("sample sma_cross wiring resolves") } /// The blended signal: a trend leg (SMA-cross) and a momentum leg (MACD), combined /// by a weighted sum. A multiply-nested composite (root → signals → {trend, /// momentum}); the blend is a multi-param node living inside it, with one weight /// bound as a structural constant (so it drops out of the sweepable surface). fn signals(name: &str) -> Composite { let mut g = GraphBuilder::new(name); let trend = g.add(sma_cross("trend")); // trend leg (one f64 "cross") let momentum = g.add(macd("momentum")); // momentum leg (3 outputs) // blend: Σ wᵢ·termᵢ over [trend.cross, momentum.histogram, momentum.signal]. // weights[2] is bound (a fixed signal-line weight) → removed from param_space; // weights[0]/[1] stay tunable. `.named("blend")` makes the path signals.blend.* // (and renders the `blend:` prefix). let blend = g.add( LinComb::builder(3) .named("blend") .bind("weights[2]", Scalar::f64(0.5)), ); let price = g.input_role("price"); g.feed(price, [trend.input("price"), momentum.input("price")]); g.connect(trend.output("cross"), blend.input("term[0]")); // trend.cross → blend.term[0] g.connect(momentum.output("histogram"), blend.input("term[1]")); // momentum.histogram → blend.term[1] g.connect(momentum.output("signal"), blend.input("term[2]")); // momentum.signal → blend.term[2] g.expose(blend.output("value"), "signal"); g.build().expect("sample signals wiring resolves") } /// The sample signal-quality blueprint (value-empty) **with its two recording /// sinks reachable**: returns the equity + exposure receivers a per-point sweep /// run drains (the eight free params — the trend SMA lengths, the momentum EMA /// lengths, the two open blend weights, and the exposure scale — are injected at /// compile via the point vector). The root harness of the sample topology, whose /// signal is built by the nested `signals` composite; `build_sample` (the `aura /// graph` entry) is expressed on top of it. #[allow(clippy::type_complexity)] fn sample_blueprint_with_sinks(pip_size: f64) -> ( Composite, Receiver<(Timestamp, Vec)>, Receiver<(Timestamp, Vec)>, ) { let (tx_eq, rx_eq) = mpsc::channel(); let (tx_ex, rx_ex) = mpsc::channel(); let mut g = GraphBuilder::new("sample"); let sig = g.add(signals("signals")); let exposure = g.add(Bias::builder().named("bias")); let broker = g.add(SimBroker::builder(pip_size)); 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 price = g.source_role("price", ScalarKind::F64); g.feed(price, [sig.input("price"), broker.input("price")]); g.connect(sig.output("signal"), exposure.input("signal")); // blended signal -> Bias g.connect(exposure.output("bias"), broker.input("exposure")); // bias -> broker slot 0 g.connect(broker.output("equity"), eq.input("col[0]")); // equity -> sink g.connect(exposure.output("bias"), ex.input("col[0]")); // bias -> sink let bp = g.build().expect("sample blueprint wiring resolves"); (bp, rx_eq, rx_ex) } /// The sample blueprint without its sink receivers — the `aura graph` render /// entry, which never runs the graph (so the receivers are dropped). fn build_sample() -> Composite { sample_blueprint_with_sinks(SYNTHETIC_PIP_SIZE).0 } /// The built-in sample rendered by `aura graph`. fn sample_blueprint() -> Composite { build_sample() } /// Run the built-in sample over a small built-in grid (fast ∈ {2,3}, /// slow ∈ {4,5}, scale ∈ {0.5} — 4 points) and render one JSON line per point in /// enumeration (odometer) order. Pure + deterministic (C1): the same build yields /// the same report. Each point builds a fresh blueprint (fresh sink channels), /// bootstraps it under the point vector, runs it, and folds the drained sinks to /// metrics — the per-point closure the engine `sweep` drives disjointly. fn sweep_family(trace: Option<&str>, data: &DataSource, env: &project::Env) -> SweepFamily { let pip = data.pip_size(); let window = data.full_window(env); let bp = sample_blueprint_with_sinks(pip).0; let space = bp.param_space(); let (tf, ts, (mf, ms, msig)) = data.strategy_lengths(); let binder = bp .axis("signals.trend.fast.length", tf) .axis("signals.trend.slow.length", ts) .axis("signals.momentum.fast.length", [mf]) .axis("signals.momentum.slow.length", [ms]) .axis("signals.momentum.signal.length", [msig]) .axis("signals.blend.weights[0]", [1.0]) .axis("signals.blend.weights[1]", [1.0]) .axis("bias.scale", [0.5]); let varying: HashSet = binder.varying_axes().into_iter().collect(); binder .sweep(|point| { let (bp, rx_eq, rx_ex) = sample_blueprint_with_sinks(pip); let mut h = bp .bootstrap_with_cells(point) .expect("grid points are kind-checked against param_space"); let sources = data.run_sources(env); h.run(sources); let eq_rows = rx_eq.try_iter().collect::>(); let ex_rows = rx_ex.try_iter().collect::>(); let named = zip_params(&space, point); let key = member_key(&named, &varying); let manifest = sim_optimal_manifest(named, window, 0, pip); if let Some(name) = trace { persist_traces(&format!("{name}/{key}"), &manifest, &eq_rows, &ex_rows, env); } let equity = f64_field(&eq_rows, 0); let exposure = f64_field(&ex_rows, 0); RunReport { manifest, metrics: summarize(&equity, &exposure) } }) .expect("the built-in named grid matches the sample param-space") } /// The EMA-distance momentum demo strategy with its two recording sinks reachable. /// The signal is how far price sits above/below its own EMA (`price - ema`), bounded /// into a directional (unsized) bias by `Bias`, then passed through the long-only /// `LongOnly` gate. Three swept /// knobs of three kinds — `ema.length` (i64), `bias.scale` (f64), /// `longonly.enabled` (bool) — with nothing in common with the SMA-cross demo: the /// generic-sweep proof. The exposure sink taps the FINAL (gated) exposure so the /// bool's effect is visible in the trace. #[allow(clippy::type_complexity)] fn momentum_blueprint_with_sinks(pip_size: f64) -> ( Composite, Receiver<(Timestamp, Vec)>, Receiver<(Timestamp, Vec)>, ) { let (tx_eq, rx_eq) = mpsc::channel(); let (tx_ex, rx_ex) = mpsc::channel(); let mut g = GraphBuilder::new("momentum"); // Name the param-bearing nodes explicitly so the swept paths are guaranteed // ema.length / bias.scale / longonly.enabled regardless of default-name // derivation (the member-key examples + the param-space test depend on these). let ema = g.add(Ema::builder().named("ema")); // ema.length let dist = g.add(Sub::builder()); // momentum = price - ema let expo = g.add(Bias::builder().named("bias")); // bias.scale let gate = g.add(LongOnly::builder().named("longonly")); // longonly.enabled (the bool param) let broker = g.add(SimBroker::builder(pip_size)); 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 price = g.source_role("price", ScalarKind::F64); g.feed(price, [ema.input("series"), dist.input("lhs"), broker.input("price")]); g.connect(ema.output("value"), dist.input("rhs")); // momentum = price - ema g.connect(dist.output("value"), expo.input("signal")); g.connect(expo.output("bias"), gate.input("exposure")); g.connect(gate.output("exposure"), broker.input("exposure")); // gated exposure -> broker g.connect(broker.output("equity"), eq.input("col[0]")); // equity -> sink g.connect(gate.output("exposure"), ex.input("col[0]")); // gated exposure -> sink let bp = g.build().expect("momentum blueprint wiring resolves"); (bp, rx_eq, rx_ex) } /// Run the momentum strategy over its built-in grid — `ema.length ∈ {5,10}` × /// `bias.scale ∈ {0.5,1.0}` × `longonly.enabled ∈ {true,false}` = 8 points, /// all three axes varying. Mirrors `sweep_family`: capture the binder's varying /// axes, key each member via the generic portable `member_key`. With `--trace`, /// persist each member under `runs/traces///`. fn momentum_sweep_family(trace: Option<&str>, data: &DataSource, env: &project::Env) -> SweepFamily { let pip = data.pip_size(); let window = data.full_window(env); let bp = momentum_blueprint_with_sinks(pip).0; let space = bp.param_space(); let binder = bp .axis("ema.length", [5, 10]) .axis("bias.scale", [0.5, 1.0]) .axis("longonly.enabled", [true, false]); let varying: HashSet = binder.varying_axes().into_iter().collect(); binder .sweep(|point| { let (bp, rx_eq, rx_ex) = momentum_blueprint_with_sinks(pip); let mut h = bp .bootstrap_with_cells(point) .expect("grid points are kind-checked against param_space"); let sources = data.run_sources(env); h.run(sources); let eq_rows = rx_eq.try_iter().collect::>(); let ex_rows = rx_ex.try_iter().collect::>(); let named = zip_params(&space, point); let key = member_key(&named, &varying); let manifest = sim_optimal_manifest(named, window, 0, pip); if let Some(name) = trace { persist_traces(&format!("{name}/{key}"), &manifest, &eq_rows, &ex_rows, env); } let equity = f64_field(&eq_rows, 0); let exposure = f64_field(&ex_rows, 0); RunReport { manifest, metrics: summarize(&equity, &exposure) } }) .expect("the momentum named grid matches the momentum param-space") } /// The honest broker label for the dual-tap r-sma harness: it runs a RiskExecutor /// branch alongside the SimBroker, so the plain "sim-optimal" label would under-report /// it (#132). Shared by the single run and the sweep so the two cannot drift. fn r_sma_broker_label(pip_size: f64) -> String { format!("sim-optimal+risk-executor(pip_size={pip_size})") } /// `aura sweep --strategy r-sma`: sweep the r-sma harness over a fast×slow /// SIGNAL grid (the stop + sizing stay fixed — see the cycle-0066 / #133 decision /// log: risk_budget is R-invariant and bias.scale is sign-only under flat-1R, both /// degenerate axes; the stop defines the R unit, so varying it would break /// cross-member SQN comparability). Each member folds the dense R-record via /// summarize_r, so its RunReport carries `r: Some(..)` and the family is rankable /// by sqn / sqn_normalized / expectancy_r / net_expectancy_r. With `--trace`, each /// member's equity / exposure / r_equity streams are persisted under /// `runs/traces///` via `persist_traces_r` (mirroring /// `momentum_sweep_family`), so a swept member is chartable. /// The four griddable knobs of the r-sma sweep as value lists (#137): the SMA fast /// / slow lengths and the vol-stop length / `k`-multiplier. The family is the cartesian /// product of the four lists; absent flags fall back to the historical defaults (fast /// `{2,3}`, slow `{6,12}`, stop_length `{3}`, stop_k `{2.0}`) so a no-flags sweep is /// byte-identical to the pre-#137 family. #[derive(Clone, Debug, PartialEq)] struct RGrid { fast: Vec, slow: Vec, stop_length: Vec, stop_k: Vec, channel: Vec, window: Vec, band_k: Vec, } impl Default for RGrid { fn default() -> Self { Self { fast: vec![2, 3], slow: vec![6, 12], stop_length: vec![R_SMA_STOP_LENGTH], stop_k: vec![R_SMA_STOP_K], channel: vec![1920], window: vec![1920], band_k: vec![2.0], } } } /// The path-qualified `param_space()` suffix of the open vol-stop's EWMA length knob /// (the `Ema` named `stop_length` in `risk_executor_vol_open`). The full slot name is /// resolved at runtime from the live param-space by this suffix, so the composite path /// prefix is never hand-synced. const STOP_LENGTH_SUFFIX: &str = ".vol_stop.stop_length.length"; /// The path-qualified `param_space()` suffix of the open vol-stop's `k`-multiplier knob /// (the `LinComb` named `stop_k`, whose single weight is `weights[0]`). const STOP_K_SUFFIX: &str = ".vol_stop.stop_k.weights[0]"; /// The path-qualified `param_space()` suffixes of the two open SMA-cross signal knobs. /// Since cycle 0092 the signal leg is a nested `sma_signal` composite, so its knobs /// land in `param_space` under the composite prefix (`sma_signal.fast.length` / /// `sma_signal.slow.length`); like the stop knobs they are resolved by suffix so the /// prefix is never hand-synced, and rendered back to the bare `fast.length` / `slow.length` /// manifest names by `r_sma_friendly_name` (the pre-0092 family-record names — C18). const FAST_LENGTH_SUFFIX: &str = ".fast.length"; const SLOW_LENGTH_SUFFIX: &str = ".slow.length"; /// The friendly manifest name for an open vol-stop slot, given its path-qualified /// `param_space()` name: the two stop knobs render as `stop_length` / `stop_k` (their /// pre-#137 manual manifest names), every other slot unchanged. Decoupling the manifest /// name from the deep param-space path keeps the family record auditable (C18) while the /// values flow through the real `.axis(..)` grid. fn r_sma_friendly_name(space_name: &str) -> String { if space_name.ends_with(STOP_LENGTH_SUFFIX) { "stop_length".to_string() } else if space_name.ends_with(STOP_K_SUFFIX) { "stop_k".to_string() } else if space_name.ends_with(FAST_LENGTH_SUFFIX) { "fast.length".to_string() } else if space_name.ends_with(SLOW_LENGTH_SUFFIX) { "slow.length".to_string() } else { space_name.to_string() } } fn r_sma_sweep_family(trace: Option<&str>, data: &DataSource, grid: &RGrid, env: &project::Env) -> SweepFamily { let pip = data.pip_size(); let window = data.full_window(env); // a single throwaway floated build, only to resolve param_space (borrow) then // seed the named axes (move) — its taps are never drained, the per-point run_one // rebuilds with live ones. Mirrors momentum_sweep_family: param_space takes &self, // .axis() consumes self, so one build suffices. The vol-stop knobs are left OPEN // (`stop_open = true`) so all four knobs — signal AND stop — grid through the one // `.axis(..)` mechanism (#137). let (tx_eq, _) = mpsc::channel(); let (tx_ex, _) = mpsc::channel(); let (tx_r, _) = mpsc::channel(); let (tx_req, _) = mpsc::channel(); let bp = r_sma_graph(tx_eq, tx_ex, tx_r, tx_req, None, None, true, true, None); let space = bp.param_space(); // resolve the open stop slots' exact path-qualified names from the live param-space // (the composite prefix is never hand-synced — match by the stable suffix). let stop_length_axis = space .iter() .map(|p| p.name.clone()) .find(|n| n.ends_with(STOP_LENGTH_SUFFIX)) .expect("open r-sma vol-stop exposes a stop_length axis"); let stop_k_axis = space .iter() .map(|p| p.name.clone()) .find(|n| n.ends_with(STOP_K_SUFFIX)) .expect("open r-sma vol-stop exposes a stop_k axis"); let fast_axis = space .iter() .map(|p| p.name.clone()) .find(|n| n.ends_with(FAST_LENGTH_SUFFIX)) .expect("open r-sma signal exposes a fast.length axis"); let slow_axis = space .iter() .map(|p| p.name.clone()) .find(|n| n.ends_with(SLOW_LENGTH_SUFFIX)) .expect("open r-sma signal exposes a slow.length axis"); let binder = bp .axis(&fast_axis, grid.fast.clone()) .axis(&slow_axis, grid.slow.clone()) .axis(&stop_length_axis, grid.stop_length.clone()) .axis(&stop_k_axis, grid.stop_k.clone()); // the varying axes drive the member key; translate the deep stop-axis names to the // friendly `stop_length` / `stop_k` the manifest uses, so the key reads the same // names whether a signal or a stop axis is what varies. let varying: HashSet = binder .varying_axes() .into_iter() .map(|n| r_sma_friendly_name(&n)) .collect(); binder .sweep(|point| { let (tx_eq, rx_eq) = mpsc::channel(); let (tx_ex, rx_ex) = mpsc::channel(); let (tx_r, rx_r) = mpsc::channel(); let (tx_req, rx_req) = mpsc::channel(); let reduce = trace.is_none(); let mut h = r_sma_graph(tx_eq, tx_ex, tx_r, tx_req, None, None, true, reduce, None) .bootstrap_with_cells(point) .expect("r-sma grid points are kind-checked against param_space"); h.run(data.run_sources(env)); // record the swept knobs PLUS the fixed R-defining bias scale, matching the // single run: a family member must be reproducible from its own manifest // (C18). The stop knobs are now real swept axes (they appear in `point`), so // they reach the manifest via `zip_params` — under their friendly // `stop_length` / `stop_k` names so cross-member SQN comparability (C10) is // auditable from the family record by the same names as before. let mut named: Vec<(String, Scalar)> = zip_params(&space, point) .into_iter() .map(|(n, v)| (r_sma_friendly_name(&n), v)) .collect(); named.push(("bias_scale".to_string(), Scalar::f64(0.5))); let key = member_key(&named, &varying); let mut manifest = sim_optimal_manifest(named, window, 0, pip); manifest.broker = r_sma_broker_label(pip); let metrics = if reduce { // folded: GatedRecorder emits O(trades) R rows; each SeriesReducer // emits one [last, max_drawdown, sign_flips] summary row. let r_rows: Vec<(Timestamp, Vec)> = rx_r.try_iter().collect(); let (total_pips, max_drawdown) = rx_eq .try_iter() .next() .map(|(_, row)| (row[0].as_f64(), row[1].as_f64())) .unwrap_or((0.0, 0.0)); let bias_sign_flips = rx_ex.try_iter().next().map(|(_, row)| row[2].as_i64() as u64).unwrap_or(0); let mut m = RunMetrics { total_pips, max_drawdown, bias_sign_flips, r: None }; m.r = Some(summarize_r(&r_rows, &[])); m } else { // trace path (--trace set): raw recorders, persist, summarize — today's code. let eq_rows: Vec<(Timestamp, Vec)> = rx_eq.try_iter().collect(); let ex_rows: Vec<(Timestamp, Vec)> = rx_ex.try_iter().collect(); let r_rows: Vec<(Timestamp, Vec)> = rx_r.try_iter().collect(); let req_rows: Vec<(Timestamp, Vec)> = rx_req.try_iter().collect(); if let Some(name) = trace { persist_traces_r(&format!("{name}/{key}"), &manifest, &eq_rows, &ex_rows, &req_rows, &[], env); } let mut m = summarize(&f64_field(&eq_rows, 0), &f64_field(&ex_rows, 0)); m.r = Some(summarize_r(&r_rows, &[])); m }; RunReport { manifest, metrics } }) .expect("the r-sma named grid matches the r-sma param-space") } /// The param-space of the OPEN r-sma blueprint (all four knobs free) — the kinds /// the per-window `WindowRun::chosen_params` are read against (C7). Mirrors the /// throwaway-build param_space resolution inside `r_sma_sweep_family`. fn r_sma_space() -> Vec { let (tx_eq, _) = mpsc::channel(); let (tx_ex, _) = mpsc::channel(); let (tx_r, _) = mpsc::channel(); let (tx_req, _) = mpsc::channel(); r_sma_graph(tx_eq, tx_ex, tx_r, tx_req, None, None, true, true, None).param_space() } /// Windowed reduce-mode r-sma sweep over `[from, to]` — the in-sample leg of the /// r-sma walk-forward. Identical grid/axis/fold logic to `r_sma_sweep_family`, /// but windowed (`windowed_sources`) and always folded (O(trades)/member): each /// member's RunReport carries `metrics.r = Some(summarize_r(..))`, so the family is /// rankable by an R metric. fn r_sma_sweep_over( from: Timestamp, to: Timestamp, data: &DataSource, grid: &RGrid, env: &project::Env, ) -> (SweepFamily, Option>) { let pip = data.pip_size(); let (tx_eq, _) = mpsc::channel(); let (tx_ex, _) = mpsc::channel(); let (tx_r, _) = mpsc::channel(); let (tx_req, _) = mpsc::channel(); let bp = r_sma_graph(tx_eq, tx_ex, tx_r, tx_req, None, None, true, true, None); let space = bp.param_space(); let stop_length_axis = space .iter() .map(|p| p.name.clone()) .find(|n| n.ends_with(STOP_LENGTH_SUFFIX)) .expect("open r-sma vol-stop exposes a stop_length axis"); let stop_k_axis = space .iter() .map(|p| p.name.clone()) .find(|n| n.ends_with(STOP_K_SUFFIX)) .expect("open r-sma vol-stop exposes a stop_k axis"); let fast_axis = space .iter() .map(|p| p.name.clone()) .find(|n| n.ends_with(FAST_LENGTH_SUFFIX)) .expect("open r-sma signal exposes a fast.length axis"); let slow_axis = space .iter() .map(|p| p.name.clone()) .find(|n| n.ends_with(SLOW_LENGTH_SUFFIX)) .expect("open r-sma signal exposes a slow.length axis"); bp.axis(&fast_axis, grid.fast.clone()) .axis(&slow_axis, grid.slow.clone()) .axis(&stop_length_axis, grid.stop_length.clone()) .axis(&stop_k_axis, grid.stop_k.clone()) .sweep_with_lattice(|point| { let (tx_eq, rx_eq) = mpsc::channel(); let (tx_ex, rx_ex) = mpsc::channel(); let (tx_r, rx_r) = mpsc::channel(); let (tx_req, _rx_req) = mpsc::channel(); let mut h = r_sma_graph(tx_eq, tx_ex, tx_r, tx_req, None, None, true, true, None) .bootstrap_with_cells(point) .expect("r-sma grid points are kind-checked against param_space"); let sources = data.windowed_sources(from, to, env); let window = window_of(&sources).expect("non-empty in-sample window"); h.run(sources); let mut named: Vec<(String, Scalar)> = zip_params(&space, point).into_iter() .map(|(n, v)| (r_sma_friendly_name(&n), v)).collect(); named.push(("bias_scale".to_string(), Scalar::f64(0.5))); let mut manifest = sim_optimal_manifest(named, window, 0, pip); manifest.broker = r_sma_broker_label(pip); let r_rows: Vec<(Timestamp, Vec)> = rx_r.try_iter().collect(); let (total_pips, max_drawdown) = rx_eq.try_iter().next() .map(|(_, row)| (row[0].as_f64(), row[1].as_f64())).unwrap_or((0.0, 0.0)); let bias_sign_flips = rx_ex.try_iter().next().map(|(_, row)| row[2].as_i64() as u64).unwrap_or(0); let mut m = RunMetrics { total_pips, max_drawdown, bias_sign_flips, r: None }; m.r = Some(summarize_r(&r_rows, &[])); RunReport { manifest, metrics: m } }) .map(|(fam, lat)| (fam, Some(lat))) .expect("the r-sma named grid matches the r-sma param-space") } /// Run the chosen r-sma params over an OOS window; return the recorded pip-equity /// segment (for stitching) and the OOS RunReport whose `metrics.r` carries both the /// R metrics and the per-trade `trade_rs`. Non-reduce (raw recorders): one window's /// curve is bounded, and `stitch` needs the full pip-equity series. fn run_oos_r( params: &[Cell], from: Timestamp, to: Timestamp, trace: Option<&str>, data: &DataSource, env: &project::Env, ) -> (Vec<(Timestamp, f64)>, RunReport) { let pip = data.pip_size(); let (tx_eq, rx_eq) = mpsc::channel(); let (tx_ex, rx_ex) = mpsc::channel(); let (tx_r, rx_r) = mpsc::channel(); let (tx_req, rx_req) = mpsc::channel(); let bp = r_sma_graph(tx_eq, tx_ex, tx_r, tx_req, None, None, true, false, None); let space = bp.param_space(); let mut h = bp.bootstrap_with_cells(params) .expect("chosen params pre-validated by the in-sample GridSpace::new"); let sources = data.windowed_sources(from, to, env); let window = window_of(&sources).expect("non-empty out-of-sample window"); h.run(sources); let eq_rows = rx_eq.try_iter().collect::>(); let ex_rows = rx_ex.try_iter().collect::>(); let r_rows: Vec<(Timestamp, Vec)> = rx_r.try_iter().collect(); let req_rows: Vec<(Timestamp, Vec)> = rx_req.try_iter().collect(); let mut named: Vec<(String, Scalar)> = zip_params(&space, params).into_iter() .map(|(n, v)| (r_sma_friendly_name(&n), v)).collect(); named.push(("bias_scale".to_string(), Scalar::f64(0.5))); let mut manifest = sim_optimal_manifest(named, window, 0, pip); manifest.broker = r_sma_broker_label(pip); if let Some(name) = trace { persist_traces_r(&format!("{name}/oos{}", from.0), &manifest, &eq_rows, &ex_rows, &req_rows, &[], env); } let equity = f64_field(&eq_rows, 0); let exposure = f64_field(&ex_rows, 0); let mut metrics = summarize(&equity, &exposure); metrics.r = Some(summarize_r(&r_rows, &[])); (equity, RunReport { manifest, metrics }) } /// `aura sweep --strategy r-breakout`: sweep the breakout harness over a channel × /// stop grid. One `channel` length drives BOTH rolling nodes (parameter-ganging, #61), /// so the family iterates the cartesian product MANUALLY with a fully-bound graph per /// point (compile_with_params(&[]) + Harness::bootstrap, like run_r_sma) rather than /// the open .axis/bootstrap_with_cells path. Each member folds the dense R-record via /// summarize_r, so the family is rankable by sqn / expectancy_r / ... (parity with /// r-sma). With --trace, raw recorders persist the per-cycle streams. fn r_breakout_sweep_family(trace: Option<&str>, data: &DataSource, grid: &RGrid, env: &project::Env) -> SweepFamily { let pip = data.pip_size(); let window = data.full_window(env); let mut varying: HashSet = HashSet::new(); if grid.channel.len() > 1 { varying.insert("channel".to_string()); } if grid.stop_length.len() > 1 { varying.insert("stop_length".to_string()); } if grid.stop_k.len() > 1 { varying.insert("stop_k".to_string()); } let mut points = Vec::new(); for &c in &grid.channel { for &sl in &grid.stop_length { for &sk in &grid.stop_k { let (tx_eq, rx_eq) = mpsc::channel(); let (tx_ex, rx_ex) = mpsc::channel(); let (tx_r, rx_r) = mpsc::channel(); let (tx_req, rx_req) = mpsc::channel(); let reduce = trace.is_none(); let flat = r_breakout_graph(tx_eq, tx_ex, tx_r, tx_req, Some(c), sl, sk, reduce) .compile_with_params(&[]) .expect("valid r-breakout blueprint"); let mut h = Harness::bootstrap(flat).expect("valid r-breakout harness"); h.run(data.run_sources(env)); let named: Vec<(String, Scalar)> = vec![ ("channel".to_string(), Scalar::i64(c)), ("stop_length".to_string(), Scalar::i64(sl)), ("stop_k".to_string(), Scalar::f64(sk)), ]; let key = member_key(&named, &varying); let mut manifest = sim_optimal_manifest(named, window, 0, pip); manifest.broker = r_sma_broker_label(pip); let metrics = if reduce { let r_rows: Vec<(Timestamp, Vec)> = rx_r.try_iter().collect(); let (total_pips, max_drawdown) = rx_eq .try_iter() .next() .map(|(_, row)| (row[0].as_f64(), row[1].as_f64())) .unwrap_or((0.0, 0.0)); let bias_sign_flips = rx_ex.try_iter().next().map(|(_, row)| row[2].as_i64() as u64).unwrap_or(0); let mut m = RunMetrics { total_pips, max_drawdown, bias_sign_flips, r: None }; m.r = Some(summarize_r(&r_rows, &[])); m } else { let eq_rows: Vec<(Timestamp, Vec)> = rx_eq.try_iter().collect(); let ex_rows: Vec<(Timestamp, Vec)> = rx_ex.try_iter().collect(); let r_rows: Vec<(Timestamp, Vec)> = rx_r.try_iter().collect(); let req_rows: Vec<(Timestamp, Vec)> = rx_req.try_iter().collect(); if let Some(name) = trace { persist_traces_r(&format!("{name}/{key}"), &manifest, &eq_rows, &ex_rows, &req_rows, &[], env); } let mut m = summarize(&f64_field(&eq_rows, 0), &f64_field(&ex_rows, 0)); m.r = Some(summarize_r(&r_rows, &[])); m }; points.push(SweepPoint { params: vec![], report: RunReport { manifest, metrics } }); } } } SweepFamily { space: vec![], points } } fn r_meanrev_sweep_family(trace: Option<&str>, data: &DataSource, grid: &RGrid, env: &project::Env) -> SweepFamily { let pip = data.pip_size(); let window = data.full_window(env); let mut varying: HashSet = HashSet::new(); if grid.window.len() > 1 { varying.insert("window".to_string()); } if grid.band_k.len() > 1 { varying.insert("band_k".to_string()); } if grid.stop_length.len() > 1 { varying.insert("stop_length".to_string()); } if grid.stop_k.len() > 1 { varying.insert("stop_k".to_string()); } let mut points = Vec::new(); for &n in &grid.window { for &bk in &grid.band_k { for &sl in &grid.stop_length { for &sk in &grid.stop_k { let (tx_eq, rx_eq) = mpsc::channel(); let (tx_ex, rx_ex) = mpsc::channel(); let (tx_r, rx_r) = mpsc::channel(); let (tx_req, rx_req) = mpsc::channel(); let reduce = trace.is_none(); let flat = r_meanrev_graph(tx_eq, tx_ex, tx_r, tx_req, Some(n), bk, sl, sk, reduce) .compile_with_params(&[]) .expect("valid r-meanrev blueprint"); let mut h = Harness::bootstrap(flat).expect("valid r-meanrev harness"); h.run(data.run_sources(env)); let named: Vec<(String, Scalar)> = vec![ ("window".to_string(), Scalar::i64(n)), ("band_k".to_string(), Scalar::f64(bk)), ("stop_length".to_string(), Scalar::i64(sl)), ("stop_k".to_string(), Scalar::f64(sk)), ]; let key = member_key(&named, &varying); let mut manifest = sim_optimal_manifest(named, window, 0, pip); manifest.broker = r_sma_broker_label(pip); let metrics = if reduce { let r_rows: Vec<(Timestamp, Vec)> = rx_r.try_iter().collect(); let (total_pips, max_drawdown) = rx_eq .try_iter() .next() .map(|(_, row)| (row[0].as_f64(), row[1].as_f64())) .unwrap_or((0.0, 0.0)); let bias_sign_flips = rx_ex.try_iter().next().map(|(_, row)| row[2].as_i64() as u64).unwrap_or(0); let mut m = RunMetrics { total_pips, max_drawdown, bias_sign_flips, r: None }; m.r = Some(summarize_r(&r_rows, &[])); m } else { let eq_rows: Vec<(Timestamp, Vec)> = rx_eq.try_iter().collect(); let ex_rows: Vec<(Timestamp, Vec)> = rx_ex.try_iter().collect(); let r_rows: Vec<(Timestamp, Vec)> = rx_r.try_iter().collect(); let req_rows: Vec<(Timestamp, Vec)> = rx_req.try_iter().collect(); if let Some(name) = trace { persist_traces_r(&format!("{name}/{key}"), &manifest, &eq_rows, &ex_rows, &req_rows, &[], env); } let mut m = summarize(&f64_field(&eq_rows, 0), &f64_field(&ex_rows, 0)); m.r = Some(summarize_r(&r_rows, &[])); m }; points.push(SweepPoint { params: vec![], report: RunReport { manifest, metrics } }); } } } } SweepFamily { space: vec![], points } } /// Render a sweep family as one `RunReport` JSON line per point. Test helper: /// production (`run_sweep`) renders *and* persists per point. #[cfg(test)] fn sweep_report() -> String { let mut out = String::new(); for pt in &sweep_family(None, &DataSource::Synthetic, &project::Env::std()).points { out.push_str(&pt.report.to_json()); out.push('\n'); } out } /// Which built-in strategy `aura sweep` runs. Default (today's behaviour) is the /// SMA-cross sample; `momentum` is the bool-param demo. #[derive(Clone, Copy, PartialEq, Debug)] enum Strategy { SmaCross, Momentum, RSma, RBreakout, RMeanRev, } /// In-sample winner-selection objective for walk-forward (cycle 0077). `Argmax` is /// the bare-best pick deflated for trials (#144, the default); `Plateau` argmaxes /// the neighbourhood-smoothed surface instead (opt-in via `--select`). #[derive(Clone, Copy)] enum Selection { Argmax, Plateau(PlateauMode), } /// Parse a `--select` token: `argmax` | `plateau:mean` | `plateau:worst`. Unknown /// tokens are a usage error (the caller maps `Err(())` to exit 2). fn parse_select(s: &str) -> Result { match s { "argmax" => Ok(Selection::Argmax), "plateau:mean" => Ok(Selection::Plateau(PlateauMode::Mean)), "plateau:worst" => Ok(Selection::Plateau(PlateauMode::Worst)), _ => Err(()), } } impl Strategy { /// The CLI `--strategy` token this variant parses from — the inverse of the /// `parse_*_args` match arms. Used to echo the offending strategy in error /// messages so they name the actual input. fn cli_token(self) -> &'static str { match self { Strategy::SmaCross => "sma", Strategy::Momentum => "momentum", Strategy::RSma => "r-sma", Strategy::RBreakout => "r-breakout", Strategy::RMeanRev => "r-meanrev", } } } /// Parse a comma-separated list of `T` (each item parsed via `FromStr`), rejecting /// any item that fails to parse — the shared validator for the r-sma grid flags /// (#137). The caller folds the `Err` into the subcommand `usage()` so a malformed /// `--fast 2,x` is the strict usage path, not a downstream panic. `str::split(',')` /// always yields at least one item, and an empty item (`""`, or a trailing comma) /// fails its per-item `parse`, so a non-empty result needs no separate empty-list /// check — the empty/blank inputs are rejected by the per-item parse itself. fn parse_csv_list(s: &str) -> Result, ()> { let items: Vec<&str> = s.split(',').collect(); let mut out = Vec::with_capacity(items.len()); for item in items { out.push(item.parse::().map_err(|_| ())?); } Ok(out) } /// Render a family-member stdout line: the assigned `family_id` plus the embedded /// `RunReport`. The report is emitted in its own declaration key order (manifest /// leads with `commit`, C18) so the line is byte-identical to the stored /// `families.jsonl`. `serde_json::json!` would route the report through a /// `serde_json::Value` and re-alphabetize the manifest keys (broker-first), /// diverging from the store — hence the report is spliced in pre-serialized (#99). fn family_member_line(id: &str, report: &RunReport) -> String { format!( r#"{{"family_id":{},"report":{}}}"#, serde_json::to_string(id).expect("a string id always serializes"), report.to_json() ) } /// Monte-Carlo variant of [`family_member_line`]: the per-draw line also carries the /// realization `seed` (between `family_id` and `report`), matching `run_mc`'s shape. fn mc_member_line(id: &str, seed: u64, report: &RunReport) -> String { format!( r#"{{"family_id":{},"seed":{},"report":{}}}"#, serde_json::to_string(id).expect("a string id always serializes"), seed, report.to_json() ) } /// `aura sweep [--strategy ] [--name |--trace ]`: run the /// selected built-in sweep, persist it as a *family* (related records sharing one /// `family_id`, C18/C21) via `append_family`, and print each point's record line /// carrying the assigned id. With `--trace`, every strategy /// (`sma`/`momentum`/`r-sma`) persists each member's streams under /// `runs/traces///` (opt-in); the `r-sma` member also carries /// the `r_equity` tap (via `persist_traces_r`). fn run_sweep( strategy: Strategy, name: &str, persist: bool, data: DataSource, grid: &RGrid, env: &project::Env, ) { if persist && let Err(e) = env.trace_store().ensure_name_free(name, WriteKind::Family) { eprintln!("aura: {e}"); std::process::exit(1); } let reg = env.registry(); let family = match strategy { Strategy::SmaCross => sweep_family(persist.then_some(name), &data, env), Strategy::Momentum => momentum_sweep_family(persist.then_some(name), &data, env), Strategy::RSma => r_sma_sweep_family(persist.then_some(name), &data, grid, env), Strategy::RBreakout => r_breakout_sweep_family(persist.then_some(name), &data, grid, env), Strategy::RMeanRev => r_meanrev_sweep_family(persist.then_some(name), &data, grid, env), }; let id = match reg.append_family(name, FamilyKind::Sweep, &sweep_member_reports(&family)) { Ok(id) => id, Err(e) => { eprintln!("aura: {e}"); std::process::exit(1); } }; for pt in &family.points { println!("{}", family_member_line(&id, &pt.report)); } } /// `aura generalize --real --fast --slow --stop-length /// --stop-k `: grade one r-sma candidate (a single-cell grid) across an /// instrument list. Pre-checks the R-metric data-free (refuse a non-R / unknown /// metric before any run), runs the candidate per instrument (stamping each report's /// manifest `instrument`), reduces to the worst-case floor + sign-agreement + /// per-instrument breakdown (`generalization`, C9/C10), prints the aggregate, and /// persists the per-instrument members as a `CrossInstrument` family (C12/C18). fn run_generalize( name: &str, symbols: &[String], grid: &RGrid, metric: &str, from_ms: Option, to_ms: Option, env: &project::Env, ) { // data-free metric pre-check: refuse a non-R / unknown metric before any run. if let Err(e) = check_r_metric(metric) { eprintln!("aura: {e}"); std::process::exit(2); } let mut members: Vec = Vec::new(); for symbol in symbols { let choice = DataChoice::Real { symbol: symbol.clone(), from_ms, to_ms }; let data = DataSource::from_choice(choice, env); // per-instrument pip_or_refuse + has_symbol let family = r_sma_sweep_family(None, &data, grid, env); // single-cell grid -> 1 member let mut report = family.points[0].report.clone(); report.manifest.instrument = Some(symbol.clone()); members.push(report); } let pairs: Vec<(String, &RunReport)> = symbols.iter().cloned().zip(members.iter()).collect(); let agg = match generalization(&pairs, metric) { Ok(a) => a, Err(e) => { eprintln!("aura: {e}"); std::process::exit(2); } }; println!("{}", generalize_json(&agg)); let reg = env.registry(); match reg.append_family(name, FamilyKind::CrossInstrument, &members) { Ok(id) => println!("{{\"family_id\":\"{id}\"}}"), Err(e) => { eprintln!("aura: failed to persist family: {e}"); std::process::exit(1); } } } /// `aura walkforward [--name |--trace ]`: run a built-in rolling walk-forward /// over the sample blueprint + a synthetic windowed source. Per window: sweep the /// built-in grid on the in-sample slice, optimize by total_pips (axis 2 inside axis 3, /// where aura-cli bridges engine + registry), run the chosen params out-of-sample. /// Persist the per-window OOS reports as a *family* (C18/C21) via `append_family`, /// print each carrying the assigned id, then the stitched summary line. With /// `--trace`, also persist each OOS member's streams under /// `runs/traces//oos/` (opt-in). Deterministic (C1). fn run_walkforward( strategy: Strategy, name: &str, persist: bool, data: DataSource, grid: &RGrid, select: Selection, env: &project::Env, ) { if persist && let Err(e) = env.trace_store().ensure_name_free(name, WriteKind::Family) { eprintln!("aura: {e}"); std::process::exit(1); } let reg = env.registry(); let result = walkforward_family(strategy, persist.then_some(name), &data, grid, select, env); let id = match reg.append_family(name, FamilyKind::WalkForward, &walkforward_member_reports(&result)) { Ok(id) => id, Err(e) => { eprintln!("aura: {e}"); std::process::exit(1); } }; for w in &result.windows { println!("{}", family_member_line(&id, &w.run.oos_report)); } println!("{}", walkforward_summary_json(&result)); } /// Resolve the in-sample winner under the chosen selection objective. `Argmax` /// defers to the trials-deflation pick (#144). `Plateau` argmaxes the smoothed grid /// surface — it needs the grid lattice, so a sweep with no lattice (a future random /// walk-forward producer) is refused rather than silently argmaxed. The metric is /// always known at the call sites, so a metric error is unreachable (`expect`); the /// only fallible outcome is the plateau-without-lattice refusal, returned as /// `Err(message)` for the caller to print and exit 2. fn select_winner( family: &SweepFamily, metric: &str, select: Selection, lattice: Option<&[usize]>, ) -> Result<(SweepPoint, FamilySelection), String> { match select { Selection::Argmax => Ok(optimize_deflated( family, metric, DEFLATION_N_RESAMPLES, DEFLATION_BLOCK_LEN, DEFLATION_SEED, ).expect("walk-forward metrics are known")), Selection::Plateau(mode) => match lattice { Some(lens) => Ok(optimize_plateau(family, lens, metric, mode) .expect("walk-forward metrics are known")), None => Err( "--select plateau requires a grid sweep; a random sweep has no parameter lattice" .to_string(), ), }, } } /// The built-in rolling walk-forward: 24-bar in-sample, 12-bar out-of-sample, /// stepping 12 (contiguous OOS tiling), over the 60-bar synthetic span -> 3 /// windows. Each window sweeps a grid in-sample, optimizes by a metric, and runs /// the chosen params out-of-sample — both strategy-dispatched: the `SmaCross` arm /// sweeps the SMA sample grid and optimizes by `total_pips` (axis 2); the /// `RSma` arm sweeps the r-sma grid and optimizes by `sqn_normalized`. /// Other strategies have no walk-forward form yet (exit 2). fn walkforward_family( strategy: Strategy, trace: Option<&str>, data: &DataSource, grid: &RGrid, select: Selection, env: &project::Env, ) -> WalkForwardResult { let span = data.wf_full_span(env); let (is_len, oos_len, step) = data.wf_window_sizes(); let roller = match WindowRoller::new(span, is_len, oos_len, step, RollMode::Rolling) { Ok(r) => r, Err(e) => { eprintln!("aura: walk-forward window too short for one IS+OOS span: {e:?}"); std::process::exit(2); } }; match strategy { Strategy::SmaCross => { let space = sample_blueprint_with_sinks(data.pip_size()).0.param_space(); walk_forward(roller, space, |w: WindowBounds| { let (is_family, lattice) = sweep_over(w.is.0, w.is.1, data, env); let (best, selection) = match select_winner(&is_family, "total_pips", select, lattice.as_deref()) { Ok(v) => v, Err(msg) => { eprintln!("aura: {msg}"); std::process::exit(2); } }; let (oos_equity, mut oos_report) = run_oos(&best.params, w.oos.0, w.oos.1, trace, data, env); oos_report.manifest.selection = Some(selection); WindowRun { // The tag-free sweep winner is the chosen point; its kinds live on // WalkForwardResult.space (computed once above from the same blueprint). chosen_params: best.params, oos_equity, oos_report, } }) } Strategy::RSma => { let space = r_sma_space(); walk_forward(roller, space, |w: WindowBounds| { let (is_family, lattice) = r_sma_sweep_over(w.is.0, w.is.1, data, grid, env); let (best, selection) = match select_winner(&is_family, "sqn_normalized", select, lattice.as_deref()) { Ok(v) => v, Err(msg) => { eprintln!("aura: {msg}"); std::process::exit(2); } }; let (oos_equity, mut oos_report) = run_oos_r(&best.params, w.oos.0, w.oos.1, trace, data, env); oos_report.manifest.selection = Some(selection); WindowRun { chosen_params: best.params, oos_equity, oos_report } }) } other => { eprintln!( "aura: walkforward has no form for strategy '{}'", other.cli_token() ); std::process::exit(2); } } } /// Sweep the built-in named grid over an in-sample window, sourcing the in-memory /// windowed stream. Mirrors `sweep_family`, but windowed by `[from, to]`. fn sweep_over( from: Timestamp, to: Timestamp, data: &DataSource, env: &project::Env, ) -> (SweepFamily, Option>) { let pip = data.pip_size(); let bp = sample_blueprint_with_sinks(pip).0; let space = bp.param_space(); let (tf, ts, (mf, ms, msig)) = data.strategy_lengths(); bp.axis("signals.trend.fast.length", tf) .axis("signals.trend.slow.length", ts) .axis("signals.momentum.fast.length", [mf]) .axis("signals.momentum.slow.length", [ms]) .axis("signals.momentum.signal.length", [msig]) .axis("signals.blend.weights[0]", [1.0]) .axis("signals.blend.weights[1]", [1.0]) .axis("bias.scale", [0.5]) .sweep_with_lattice(|point| { let (bp, rx_eq, rx_ex) = sample_blueprint_with_sinks(pip); let mut h = bp .bootstrap_with_cells(point) .expect("grid points are kind-checked against param_space"); let sources = data.windowed_sources(from, to, env); let window = window_of(&sources).expect("non-empty in-sample window"); h.run(sources); let equity = f64_field(&rx_eq.try_iter().collect::>(), 0); let exposure = f64_field(&rx_ex.try_iter().collect::>(), 0); RunReport { manifest: sim_optimal_manifest(zip_params(&space, point), window, 0, pip), metrics: summarize(&equity, &exposure), } }) .map(|(fam, lat)| (fam, Some(lat))) .expect("the built-in named grid matches the sample param-space") } /// Run the chosen params over an out-of-sample window; return the recorded /// pip-equity segment (for stitching) and the OOS RunReport (the C18 record). fn run_oos( params: &[Cell], from: Timestamp, to: Timestamp, trace: Option<&str>, data: &DataSource, env: &project::Env, ) -> (Vec<(Timestamp, f64)>, RunReport) { let pip = data.pip_size(); let (bp, rx_eq, rx_ex) = sample_blueprint_with_sinks(pip); let space = bp.param_space(); let mut h = bp .bootstrap_with_cells(params) .expect("chosen params pre-validated by the in-sample GridSpace::new"); let sources = data.windowed_sources(from, to, env); let window = window_of(&sources).expect("non-empty out-of-sample window"); h.run(sources); let eq_rows = rx_eq.try_iter().collect::>(); let ex_rows = rx_ex.try_iter().collect::>(); let manifest = sim_optimal_manifest(zip_params(&space, params), window, 0, pip); if let Some(name) = trace { persist_traces(&format!("{name}/oos{}", from.0), &manifest, &eq_rows, &ex_rows, env); } let equity = f64_field(&eq_rows, 0); let exposure = f64_field(&ex_rows, 0); let report = RunReport { manifest, metrics: summarize(&equity, &exposure) }; (equity, report) } /// Pool every OOS window's per-trade R series into one flat vector, in roll order /// (window order, then within-window trade order). Windows with no `r` block /// contribute nothing. The single home of the pooling-in-roll-order semantics — /// both the walk-forward `oos_r` summary and the `mc` R-bootstrap reduce this. fn pooled_oos_trade_rs(result: &WalkForwardResult) -> Vec { result .windows .iter() .flat_map(|w| w.run.oos_report.metrics.r.as_ref().map(|r| r.trade_rs.clone()).unwrap_or_default()) .collect() } /// The walk-forward summary line: window count, stitched OOS total pips (the last /// stitched-curve value), and the on-demand per-param stability. Canonical JSON /// (C14). fn walkforward_summary_json(result: &WalkForwardResult) -> String { let total = result.stitched_oos_equity.last().map(|&(_, v)| v).unwrap_or(0.0); let pooled_rs = pooled_oos_trade_rs(result); let mut obj = serde_json::json!({ "windows": result.windows.len(), "stitched_total_pips": total, "param_stability": param_stability(result), }); if result.windows.iter().any(|w| w.run.oos_report.metrics.r.is_some()) { // RMetrics serializes its scalar fields (trade_rs is serde-skipped, so the // oos_r block is the clean R-metric summary of the pooled series). obj["oos_r"] = serde_json::to_value(r_metrics_from_rs(&pooled_rs)) .expect("RMetrics serializes"); } serde_json::json!({ "walkforward": obj }).to_string() } /// The cross-instrument generalization line: the chosen metric, instrument count, /// worst-case floor, sign-agreement count, and the per-instrument breakdown. Canonical /// JSON (C14), mirroring `walkforward_summary_json`'s `{"generalize": obj}` shape. fn generalize_json(agg: &Generalization) -> String { let per: Vec = agg .per_instrument .iter() .map(|(sym, v)| serde_json::json!([sym, v])) .collect(); let obj = serde_json::json!({ "metric": agg.selection_metric, "n_instruments": agg.n_instruments, "worst_case": agg.worst_case, "sign_agreement": agg.sign_agreement, "per_instrument": per, }); serde_json::json!({ "generalize": obj }).to_string() } /// A longer deterministic stream than `showcase_prices` — enough for several /// IS/OOS windows with SMA warm-up. Seed-determined via `SyntheticSpec` (C1). fn walkforward_prices() -> Vec<(Timestamp, Scalar)> { let spec = SyntheticSpec { start: 1.0, len: 60, step: 1 }; let mut src = spec.source(7); let mut out = Vec::new(); while let Some(item) = aura_engine::Source::next(&mut src) { out.push(item); } out } /// The in-memory windowed source the built-in demo uses (the firewall mapping to /// `DataServer::stream_m1_windowed` is the real-data path; the demo stays in-memory, /// mirroring `run_sweep`'s `showcase_prices`). Inclusive `[from, to]`. fn walkforward_window_source(from: Timestamp, to: Timestamp) -> VecSource { VecSource::new( walkforward_prices() .into_iter() .filter(|&(t, _)| t >= from && t <= to) .collect(), ) } /// Render the built-in walk-forward as the per-window OOS RunReport lines plus the /// summary line — the `run_walkforward` shape minus registry persistence. Test /// helper (mirrors `sweep_report`). #[cfg(test)] fn walkforward_report() -> String { let result = walkforward_family( Strategy::SmaCross, None, &DataSource::Synthetic, &RGrid::default(), Selection::Argmax, &project::Env::std(), ); let mut out = String::new(); for w in &result.windows { out.push_str(&w.run.oos_report.to_json()); out.push('\n'); } out.push_str(&walkforward_summary_json(&result)); out.push('\n'); out } /// The built-in Monte-Carlo family: the sample harness over a fixed (empty) base /// point, re-seeded across a built-in seed set — each seed a disjoint C1 /// realization of a synthetic price walk (C12 axis 4). Mirrors `sweep_family`, /// varying the *seed* rather than a tuning param. The seed -> `Source` /// construction lives inside the per-draw closure (eager-agnostic, #71). fn mc_family(trace: Option<&str>, env: &project::Env) -> McFamily { let base_point: Vec = Vec::new(); monte_carlo(&base_point, &[1, 2, 3], |seed, _base| { let (mut h, rx_eq, rx_ex) = sample_harness(SYNTHETIC_PIP_SIZE); let spec = SyntheticSpec { start: 1.0, len: 32, step: 1 }; let sources: Vec> = vec![Box::new(spec.source(seed))]; let window = window_of(&sources).expect("non-empty synthetic stream"); h.run(sources); let eq_rows = rx_eq.try_iter().collect::>(); let ex_rows = rx_ex.try_iter().collect::>(); let manifest = sim_optimal_manifest( vec![ ("sma_fast".to_string(), Scalar::i64(2)), ("sma_slow".to_string(), Scalar::i64(4)), ("bias_scale".to_string(), Scalar::f64(0.5)), ], window, seed, SYNTHETIC_PIP_SIZE, ); if let Some(name) = trace { persist_traces(&format!("{name}/seed{seed}"), &manifest, &eq_rows, &ex_rows, env); } let equity = f64_field(&eq_rows, 0); let exposure = f64_field(&ex_rows, 0); RunReport { manifest, metrics: summarize(&equity, &exposure) } }) } /// Render an `McAggregate` as one canonical JSON line. `McAggregate` itself is not /// `Serialize` (only its `MetricStats` fields are), so the line is built from the /// three per-metric stat blocks. fn mc_aggregate_json(agg: &McAggregate) -> String { serde_json::json!({ "mc_aggregate": { "total_pips": agg.total_pips, "max_drawdown": agg.max_drawdown, "bias_sign_flips": agg.bias_sign_flips, } }) .to_string() } /// `aura mc [--name |--trace ]`: run the built-in Monte-Carlo family, persist /// it to the family store via `append_family` (C18/C21), print each draw's record /// line (carrying the assigned `family_id`) plus the aggregate line. With `--trace`, /// also persist each draw's streams under `runs/traces//seed/` (opt-in). fn run_mc(name: &str, persist: bool, env: &project::Env) { if persist && let Err(e) = env.trace_store().ensure_name_free(name, WriteKind::Family) { eprintln!("aura: {e}"); std::process::exit(1); } let reg = env.registry(); let family = mc_family(persist.then_some(name), env); let id = match reg.append_family(name, FamilyKind::MonteCarlo, &mc_member_reports(&family)) { Ok(id) => id, Err(e) => { eprintln!("aura: {e}"); std::process::exit(1); } }; for draw in &family.draws { println!("{}", mc_member_line(&id, draw.seed, &draw.report)); } println!("{}", mc_aggregate_json(&family.aggregate)); } /// Parsed form of the `mc` tail: either the synthetic seed-resweep (today's path, /// preserved byte-for-byte) or the real-candidate R-bootstrap. The R path accepts /// ONLY `--strategy r-sma` (the sole R-reporting walk-forward strategy); a bare /// `--real` without it is a usage error (the synthetic seed-resweep is undefined /// over real bars). // Parsed once at the dispatch boundary and immediately destructured into the run // call; never stored or collected, so the inter-variant size gap is free here. #[allow(clippy::large_enum_variant)] #[derive(Clone, Debug, PartialEq)] enum McArgs { Synthetic { name: String, persist: bool }, RealR { choice: DataChoice, grid: RGrid, block_len: usize, n_resamples: usize, seed: u64 }, } /// `aura mc --strategy r-sma [--real ]`: run the r-sma walk-forward, pool /// every OOS window's per-trade R series in roll order, and print one moving-block /// bootstrap `mc_r_bootstrap` line (`E[R]` distribution + P(`E[R]` <= 0)). Frictionless /// gross R (no costs); deterministic given `seed` (C1). fn run_mc_r_bootstrap( data: DataSource, grid: &RGrid, block_len: usize, n_resamples: usize, seed: u64, env: &project::Env, ) { println!("{}", mc_r_bootstrap_report(&data, grid, block_len, n_resamples, seed, env)); } /// Assemble the `mc` R-bootstrap line: run the r-sma walk-forward, pool the OOS /// per-trade R series in roll order, bootstrap `E[R]`, and render the `mc_r_bootstrap` /// line. The body of `run_mc_r_bootstrap` minus the `println!`, so the full real-R /// wiring (walk-forward -> non-empty pooling -> bootstrap -> render) is reachable /// over synthetic data in a `#[cfg(test)]` unit, mirroring `mc_report` / /// `walkforward_report` / `sweep_report`. fn mc_r_bootstrap_report( data: &DataSource, grid: &RGrid, block_len: usize, n_resamples: usize, seed: u64, env: &project::Env, ) -> String { let result = walkforward_family(Strategy::RSma, None, data, grid, Selection::Argmax, env); let pooled = pooled_oos_trade_rs(&result); let boot = r_bootstrap(&pooled, n_resamples, block_len, seed); mc_r_bootstrap_json(&boot) } /// Render an `RBootstrap` as one canonical JSON line (`MetricStats` serializes; the /// scalar fields are spliced in), mirroring `mc_aggregate_json`. fn mc_r_bootstrap_json(b: &RBootstrap) -> String { serde_json::json!({ "mc_r_bootstrap": { "n_trades": b.n_trades, "block_len": b.block_len, "n_resamples": b.n_resamples, "e_r": b.e_r, "prob_le_zero": b.prob_le_zero, } }) .to_string() } /// Render the built-in Monte-Carlo family as the per-draw `RunReport` lines plus /// the aggregate line — the `run_mc` shape minus registry persistence (no /// `family_id`, which is store-assigned). Test helper, mirroring `sweep_report` / /// `walkforward_report`: it carries the C1-determinism test of the family /// computation, separate from the store-dependent id. #[cfg(test)] fn mc_report() -> String { let family = mc_family(None, &project::Env::std()); let mut out = String::new(); for draw in &family.draws { out.push_str(&draw.report.to_json()); out.push('\n'); } out.push_str(&mc_aggregate_json(&family.aggregate)); out.push('\n'); out } /// `aura runs families`: one header line per stored family (id, kind, member /// count), in first-seen store order. fn runs_families(env: &project::Env) { let reg = env.registry(); let members = match reg.load_family_members() { Ok(m) => m, Err(e) => { eprintln!("aura: {e}"); std::process::exit(1); } }; for fam in group_families(members) { println!( "{}", serde_json::json!({ "family_id": fam.id, "kind": fam.kind, "members": fam.members.len() }) ); } } /// `aura runs family [rank ]`: list one family's member reports in /// ordinal order, or best-first by `metric`. An unknown id is an empty family /// (prints nothing, exit 0); an unknown metric is a usage error (stderr + exit 2). fn runs_family(id: &str, rank: Option<&str>, env: &project::Env) { let reg = env.registry(); let members = match reg.load_family_members() { Ok(m) => m, Err(e) => { eprintln!("aura: {e}"); std::process::exit(1); } }; let Some(family) = group_families(members).into_iter().find(|f| f.id == id) else { return; // unknown family id: empty, exit 0 }; let reports: Vec = family.members.iter().map(|m| m.report.clone()).collect(); let ordered = match rank { Some(metric) => match rank_by(reports, metric) { Ok(r) => r, Err(e) => { eprintln!("aura: {e}"); std::process::exit(2); } }, None => reports, }; for report in &ordered { println!("{}", report.to_json()); if let Some(sel) = &report.manifest.selection { match sel.mode { // `deflated_score` is `None` only on an Argmax record with no // deflation run (report.rs); guard it, symmetric with the // plateau branch, so a from-disk record cannot panic here. When // present (the sole producer always stamps it), the bytes are // unchanged. SelectionMode::Argmax => if let Some(deflated) = sel.deflated_score { match sel.overfit_probability { Some(p) => println!(" deflated={deflated:.4} P(overfit)={p:.4}"), None => println!(" deflated={deflated:.4}"), } }, SelectionMode::PlateauMean | SelectionMode::PlateauWorst => { let label = if matches!(sel.mode, SelectionMode::PlateauMean) { "mean" } else { "worst" }; if let (Some(score), Some(n)) = (sel.neighbourhood_score, sel.n_neighbours) { println!(" plateau({label})={score:.4} over {n} cells"); } } } } } } /// The outcome of reproducing one persisted family: per member, whether its re-run /// metrics are bit-identical to the stored metrics (C1). struct ReproduceReport { outcomes: Vec<(String, bool)>, } /// Reconstruct a member's bootstrap point from its recorded named params — the inverse /// of `zip_params(space, point)`. Walks the reloaded signal's `param_space` in order /// (deterministic for the same blueprint) and reads each knob's value from the manifest. fn point_from_params(space: &[ParamSpec], params: &[(String, Scalar)]) -> Vec { space .iter() .map(|ps| { params .iter() .find(|(n, _)| n == &ps.name) .map(|(_, s)| s.cell()) .unwrap_or_else(|| { // A manifest missing a param the reloaded space expects is // corrupted-on-disk data — exit cleanly (`aura:` + exit 1) like // every other persisted-data failure on the reproduce path, not // a panic. eprintln!("aura: manifest is missing param {}", ps.name); std::process::exit(1); }) }) .collect() } /// Re-derive every member of a persisted sweep family from the content-addressed store /// and compare to the stored result, against an explicit registry (testable seam). fn reproduce_family_in( reg: &Registry, id: &str, data: &DataSource, env: &project::Env, ) -> ReproduceReport { let members = reg.load_family_members().unwrap_or_else(|e| { eprintln!("aura: {e}"); std::process::exit(1); }); let Some(family) = group_families(members).into_iter().find(|f| f.id == id) else { // reproduce is an action, not a lookup: an unknown id is a hard error (distinct // from `runs family `'s treat-as-empty exit 0). eprintln!("aura: no such family '{id}'"); std::process::exit(1); }; let pip = data.pip_size(); let window = data.full_window(env); let mut outcomes = Vec::new(); for member in &family.members { let stored = &member.report; let hash = stored.manifest.topology_hash.clone().unwrap_or_else(|| { eprintln!("aura: family member has no topology_hash; not a generated run"); std::process::exit(1); }); let doc = reg .get_blueprint(&hash) .unwrap_or_else(|e| { eprintln!("aura: {e}"); std::process::exit(1); }) .unwrap_or_else(|| { eprintln!("aura: blueprint {hash} missing from store"); std::process::exit(1); }); // Reload the stored blueprint per use: a Composite is !Clone, and both the // param-space probe (below) and the re-run each consume one. The doc was // canonical-serialized at store time, so every reload is infallible. let reload = || { blueprint_from_json(&doc, &|t| env.resolve(t)).unwrap_or_else(|e| { eprintln!("aura: stored blueprint {hash} does not parse: {e:?}"); std::process::exit(1); }) }; // The param_space of the WRAPPED signal — its knobs carry the `r_sma` // wrapper's `sma_signal.` node-path prefix, exactly the names the manifest // recorded at write time. Mirrors `blueprint_sweep_family`'s probe so the // reproduce-side space name-matches the stored params (raw `signal.param_space()` // would drop the prefix and `point_from_params` could not find the knobs). let (tx_eq, _) = mpsc::channel(); let (tx_ex, _) = mpsc::channel(); let (tx_r, _) = mpsc::channel(); let (tx_req, _) = mpsc::channel(); let space = wrap_r(reload(), tx_eq, tx_ex, tx_r, tx_req, false, true, None).param_space(); let point = point_from_params(&space, &stored.manifest.params); // A MonteCarlo member carries no tuning params (the params-join is empty), so its // reproduce line would print a BLANK member label; the seed IS its realization // identity, so surface `seed=` instead. Sweep / walk-forward members echo their // tuning params (the params-join), unchanged. let label = match family.kind { FamilyKind::MonteCarlo => format!("seed={}", stored.manifest.seed), _ => stored .manifest .params .iter() .map(|(n, v)| format!("{n}={}", render_value(v))) .collect::>() .join(", "), }; // Realization-aware: a MonteCarlo member ran over a seed-driven synthetic walk, // not the showcase — reconstruct it from manifest.seed so the re-run matches (C1). // The seed is manifest-carried and identical across realizations; only the sources // and window vary by kind. let seed = stored.manifest.seed; let (sources, member_window) = match family.kind { FamilyKind::MonteCarlo => { let s = synthetic_walk_sources(seed); let w = window_of(&s).expect("non-empty synthetic walk"); (s, w) } FamilyKind::WalkForward => { // each member is one OOS window: rebuild its windowed slice from the // stored window bounds; the winner params come from the shared // manifest->cells recovery below (as Sweep members do). let (from, to) = stored.manifest.window; let s = data.windowed_sources(from, to, env); let w = window_of(&s).expect("non-empty OOS window"); (s, w) } _ => (data.run_sources(env), window), }; let rerun = run_blueprint_member( reload(), &point, &space, sources, member_window, seed, pip, &hash, env, ); outcomes.push((label, rerun.metrics == stored.metrics)); } ReproduceReport { outcomes } } /// `aura reproduce `: re-derive a persisted sweep family from the /// content-addressed blueprint store and verify each member reproduces bit-identically /// (C18 "re-derives full results on demand"). Synthetic data this cycle (deterministic); /// recorded-dataset reproduction rides the DataServer seam (#124). fn reproduce_family(id: &str, env: &project::Env) { let rep = reproduce_family_in(&env.registry(), id, &DataSource::Synthetic, env); let total = rep.outcomes.len(); let ok = rep.outcomes.iter().filter(|(_, b)| *b).count(); for (label, identical) in &rep.outcomes { let verdict = if *identical { "bit-identical" } else { "DIVERGED" }; println!("{id} member {label} reproduced: {verdict}"); } println!("reproduced {ok}/{total} members bit-identically"); if ok != total { std::process::exit(1); } } // --- MACD proof-of-concept (a richer, nested indicator + strategy) ----------- /// The MACD signal as a named composite: price → fast/slow `Ema` → the MACD line /// (their spread) → a signal `Ema` of that line → the histogram (line − signal). /// The composite exposes all **three MACD lines** as a named output record /// (`macd`, `signal`, `histogram`); the strategy trades the histogram by reading /// `from_field: 2`. A richer fixture than `sma_cross`: a nested EMA-of-EMA chain /// with interior fan-out (the MACD line feeds *both* the signal EMA and the /// histogram). Three `length` knobs (fast, slow, signal) are injected at compile /// in node order; value-empty here. fn macd(name: &str) -> Composite { let mut g = GraphBuilder::new(name); let fast = g.add(Ema::builder().named("fast")); // fast EMA let slow = g.add(Ema::builder().named("slow")); // slow EMA let line = g.add(Sub::builder()); // MACD line = fast − slow let signal = g.add(Ema::builder().named("signal")); // signal EMA of the MACD line let hist = g.add(Sub::builder()); // histogram = MACD line − signal let price = g.input_role("price"); g.feed(price, [fast.input("series"), slow.input("series")]); g.connect(fast.output("value"), line.input("lhs")); // fast → line g.connect(slow.output("value"), line.input("rhs")); // slow → line g.connect(line.output("value"), signal.input("series")); // line → signal EMA g.connect(line.output("value"), hist.input("lhs")); // line → histogram g.connect(signal.output("value"), hist.input("rhs")); // signal → histogram g.expose(line.output("value"), "macd"); // the MACD line g.expose(signal.output("value"), "signal"); // the signal line g.expose(hist.output("value"), "histogram"); // the histogram g.build().expect("sample macd wiring resolves") } /// The MACD strategy blueprint (value-empty): the `macd` histogram → `Bias` → /// `SimBroker` → recording sinks. Channels are threaded so a run can drain the /// sinks; `macd_blueprint` drops the receivers for the structural render. fn macd_strategy_blueprint( tx_eq: mpsc::Sender<(Timestamp, Vec)>, tx_ex: mpsc::Sender<(Timestamp, Vec)>, ) -> Composite { let mut g = GraphBuilder::new("macd_strategy"); let macd_node = g.add(macd("macd")); let exposure = g.add(Bias::builder().named("bias")); let broker = g.add(SimBroker::builder(SYNTHETIC_PIP_SIZE)); 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 price = g.source_role("price", ScalarKind::F64); g.feed(price, [macd_node.input("price"), broker.input("price")]); g.connect(macd_node.output("histogram"), exposure.input("signal")); // histogram → Bias g.connect(exposure.output("bias"), broker.input("exposure")); // bias → broker slot 0 g.connect(broker.output("equity"), eq.input("col[0]")); // equity → sink g.connect(exposure.output("bias"), ex.input("col[0]")); // bias → sink g.build().expect("macd_strategy wiring resolves") } /// The MACD strategy blueprint as a param-space fixture (receivers dropped, since /// `param_space()` reads structure only, never runs the graph). #[cfg(test)] fn macd_blueprint() -> Composite { let (tx_eq, _rx_eq) = mpsc::channel(); let (tx_ex, _rx_ex) = mpsc::channel(); macd_strategy_blueprint(tx_eq, tx_ex) } /// The point vector for the MACD strategy, in `param_space()` slot order: /// `[fast EMA length, slow EMA length, signal EMA length, exposure scale]`. Short /// windows so the 7-tick synthetic stream still produces a non-trivial trace /// (conventional MACD is 12/26/9, meaningless on 7 points). fn macd_point() -> Vec { vec![Scalar::i64(2), Scalar::i64(4), Scalar::i64(3), Scalar::f64(0.5)] } /// A longer synthetic stream than the SMA sample's 7 ticks: MACD's EMAs each warm /// up over their `length`, so the stream rises, falls, then rises again to give the /// histogram room to flip sign more than once *after* warm-up. Deterministic (C1). fn macd_prices() -> Vec<(Timestamp, Scalar)> { [ 1.0000_f64, 1.0008, 1.0021, 1.0039, 1.0062, 1.0090, 1.0083, 1.0061, 1.0034, 1.0012, 0.9998, 1.0006, 1.0024, 1.0047, 1.0069, 1.0086, 1.0097, 1.0092, ] .iter() .enumerate() .map(|(i, &p)| (Timestamp(i as i64 + 1), Scalar::f64(p))) .collect() } /// Run the MACD strategy: compile the nested composite blueprint to a flat harness /// (the same bootstrap path the SMA sample's compiled view uses), drive it on the /// synthetic stream, and fold both sinks into a `RunReport`. Pure and /// deterministic (C1). fn run_macd(trace: Option<&str>, env: &project::Env) -> RunReport { if let Some(n) = trace && let Err(e) = env.trace_store().ensure_name_free(n, WriteKind::Run) { eprintln!("aura: {e}"); std::process::exit(1); } let (tx_eq, rx_eq) = mpsc::channel(); let (tx_ex, rx_ex) = mpsc::channel(); let flat = macd_strategy_blueprint(tx_eq, tx_ex) .compile_with_params(&macd_point()) .expect("valid macd blueprint"); let mut h = Harness::bootstrap(flat).expect("valid macd harness"); let sources: Vec> = vec![Box::new(VecSource::new(macd_prices()))]; let window = window_of(&sources).expect("non-empty macd stream"); h.run(sources); let eq_rows: Vec<(Timestamp, Vec)> = rx_eq.try_iter().collect(); let ex_rows: Vec<(Timestamp, Vec)> = rx_ex.try_iter().collect(); let manifest = sim_optimal_manifest( vec![ ("ema_fast".to_string(), Scalar::i64(2)), ("ema_slow".to_string(), Scalar::i64(4)), ("ema_signal".to_string(), Scalar::i64(3)), ("bias_scale".to_string(), Scalar::f64(0.5)), ], window, 0, SYNTHETIC_PIP_SIZE, ); if let Some(name) = trace { persist_traces(name, &manifest, &eq_rows, &ex_rows, env); } let metrics = summarize(&f64_field(&eq_rows, 0), &f64_field(&ex_rows, 0)); RunReport { manifest, metrics } } // --- r-sma harness (the SMA-cross signal scored in R) -------------------- /// The r-sma vol-stop EWMA length (cycles). Single source for the `StopRule::Vol` /// the blueprint embeds and the `stop` param the manifest records — kept honest by one /// constant instead of a hand-synced literal across the function boundary. const R_SMA_STOP_LENGTH: i64 = 3; /// The r-sma vol-stop multiplier (1R = `k`·σ). Single source for the embedded /// `StopRule::Vol` and its manifest record, like [`R_SMA_STOP_LENGTH`]. const R_SMA_STOP_K: f64 = 2.0; /// The r-sma demo signal SMA lengths (fast/slow), bound at single-run build time. /// Single source for the graph that runs (`r_sma_graph`), the signal that is hashed /// (`sma_signal` → `topology_hash`), and the recorded manifest params — so the /// hashed topology cannot drift from the executed one across the function boundary. const R_SMA_FAST: i64 = 2; const R_SMA_SLOW: i64 = 4; /// The r-sma demo `Bias` scale (conviction magnitude). Single source for the /// `Bias` node bound in `sma_signal` (the hashed + executed signal) and the /// recorded manifest param, so the hashed topology cannot drift from the recorded /// `bias_scale` across the function boundary — the same guard as the SMA lengths. const R_SMA_BIAS_SCALE: f64 = 0.5; /// Short-horizon realized-range window for vol-scaled slippage. Deliberately /// distinct from `R_SMA_STOP_LENGTH` (3): scaling slippage by the stop's own /// vol would collapse cost-in-R to a constant (spec 0082). Short enough to warm /// within the synthetic smoke fixture so the run path exercises non-zero slippage. const SLIP_VOL_LENGTH: i64 = 5; /// Which cost nodes the run-path cost graph builds. At least one field is `Some` /// (the carrier `Option` is `None` when no cost flag was given). struct CostConfig { const_cost: Option, // --cost-per-trade slip_vol_mult: Option, // --slip-vol-mult carry_per_cycle: Option, // --carry-per-cycle } /// Which data a `run` drives a harness on: the built-in synthetic stream, or real M1 /// close bars for a vetted symbol over an optional window. #[derive(Debug)] enum RunData { Synthetic, Real { symbol: String, from: Option, to: Option }, } /// A rise-fall-rise synthetic stream for the r-sma smoke run: long enough to warm /// the `vol_stop(length=3)` and flip the SMA(2)/SMA(4) cross at least once, so the /// RiskExecutor opens and closes at least one trade. Deterministic (C1). fn r_sma_prices() -> Vec<(Timestamp, Scalar)> { [ 1.0000_f64, 1.0008, 1.0021, 1.0039, 1.0062, 1.0090, 1.0083, 1.0061, 1.0034, 1.0012, 0.9998, 1.0006, 1.0024, 1.0047, 1.0069, 1.0086, 1.0097, 1.0092, ] .iter() .enumerate() .map(|(i, &p)| (Timestamp(i as i64 + 1), Scalar::f64(p))) .collect() } /// Interned `col[i]` recorder-port names, built once. The `GraphBuilder::input` API /// wants `&'static str`; interning here (instead of `format!(...).leak()` per field) /// means reusing the r-sma harness in a sweep / Monte-Carlo loop reuses these /// strings rather than leaking 14 fresh ones per build (#132). static COL_PORTS: LazyLock> = LazyLock::new(|| (0..PM_FIELD_NAMES.len()).map(|i| format!("col[{i}]")).collect()); /// The r-sma signal leg as a standalone, serializable Composite: SMA-cross → /// Bias. Exposes a `price` input-role and a single `bias` output — the boundary /// shape a serialized blueprint round-trips (the `blueprint_serde_e2e.rs` /// template). The two SMA knobs are bound when `Some` (byte-identical to the old /// build-time bind) and left open as `fast.length` / `slow.length` when `None`. fn sma_signal(fast_len: Option, slow_len: Option) -> Composite { let mut g = GraphBuilder::new("sma_signal"); let mut fast_b = Sma::builder().named("fast"); if let Some(l) = fast_len { fast_b = fast_b.bind("length", Scalar::i64(l)); } let fast = g.add(fast_b); let mut slow_b = Sma::builder().named("slow"); if let Some(l) = slow_len { slow_b = slow_b.bind("length", Scalar::i64(l)); } let slow = g.add(slow_b); let spread = g.add(Sub::builder()); let exposure = g.add( Bias::builder() .named("bias") .bind("scale", Scalar::f64(R_SMA_BIAS_SCALE)), ); let price = g.source_role("price", ScalarKind::F64); g.feed(price, vec![fast.input("series"), slow.input("series")]); g.connect(fast.output("value"), spread.input("lhs")); g.connect(slow.output("value"), spread.input("rhs")); g.connect(spread.output("value"), exposure.input("signal")); g.expose(exposure.output("bias"), "bias"); g.build().expect("sma_signal wiring resolves") } /// SHA256 (hex) of a canonical (#164) blueprint JSON string — the content id (#158). /// The single hashing primitive, shared by [`topology_hash`] (from a live `Composite`) /// and the op-script `graph introspect --content-id` path (`crate::content_id`), so the /// two surfaces agree by construction over the same canonical bytes. Research-side /// (aura-cli), off the frozen engine (invariant 8). fn content_id(canonical_json: &str) -> String { aura_research::content_id_of(canonical_json) } /// SHA256 (hex) of the canonical (#164, no-trailing-newline) serialization of a /// signal blueprint — the run's `topology_hash` (#158). fn topology_hash(signal: &Composite) -> String { content_id(&blueprint_to_json(signal).expect("a buildable signal serializes")) } /// The r-sma harness topology, shared by the single run and the sweep. The two /// signal knobs are bound when `Some` (single run — identical to the old /// build-time bind, output byte-unchanged) and left free when `None` (sweep — they /// now nest in the `sma_signal` composite, so they land in `param_space` as /// `sma_signal.fast.length` / `sma_signal.slow.length`, rendered back to the /// bare `fast.length` / `slow.length` manifest names by `r_sma_friendly_name`). /// The vol-stop knobs are /// bound to the pinned constants when `stop_open` is `false` (single run + the /// default sweep, output byte-unchanged) and left free when `true` (a gridded sweep /// — they land in `param_space` under the `.vol_stop.stop_length.length` / /// `.vol_stop.stop_k.weights[0]` suffixes). Four taps: equity (SimBroker), exposure /// (Bias), the 14-column R-record (→ summarize_r), and r_equity = cum_realized_r + /// unrealized_r. /// /// The seven-arg signature is a conscious keep, not an oversight: the four `tx_*` /// are the per-tap recorder channels (one Recorder edge each — equity, exposure, /// the R-record, r_equity), `fast_len`/`slow_len` are the two floatable signal /// knobs, and `stop_open` selects the bound-vs-open vol-stop arm. A sender bundle /// would only rename the same four channels into one struct field without removing /// an edge, trading the lint for indirection; the `too_many_arguments` allow is /// preferred over that churn. #[allow(clippy::type_complexity, clippy::too_many_arguments)] fn r_sma_graph( tx_eq: mpsc::Sender<(Timestamp, Vec)>, tx_ex: mpsc::Sender<(Timestamp, Vec)>, tx_r: mpsc::Sender<(Timestamp, Vec)>, tx_req: mpsc::Sender<(Timestamp, Vec)>, fast_len: Option, slow_len: Option, stop_open: bool, reduce: bool, cost: Option<(CostConfig, mpsc::Sender<(Timestamp, Vec)>, mpsc::Sender<(Timestamp, Vec)>)>, ) -> Composite { wrap_r( sma_signal(fast_len, slow_len), tx_eq, tx_ex, tx_r, tx_req, stop_open, reduce, cost, ) } /// Wrap a `signal` composite (a `price`→`bias` leg) in the R run /// scaffolding: pip broker, the per-tap recorders, the vol-stop RiskExecutor, the /// r_equity / cost legs. The signal is nested and its `price`/`bias` boundary is /// wired across; everything else is verbatim from the old `r_sma_graph` body, so /// a serialized signal loaded via `blueprint_from_json` runs through exactly the /// scaffolding the Rust-built signal does. #[allow(clippy::type_complexity, clippy::too_many_arguments)] fn wrap_r( signal: Composite, tx_eq: mpsc::Sender<(Timestamp, Vec)>, tx_ex: mpsc::Sender<(Timestamp, Vec)>, tx_r: mpsc::Sender<(Timestamp, Vec)>, tx_req: mpsc::Sender<(Timestamp, Vec)>, stop_open: bool, reduce: bool, cost: Option<( CostConfig, mpsc::Sender<(Timestamp, Vec)>, mpsc::Sender<(Timestamp, Vec)>, )>, ) -> Composite { let mut g = GraphBuilder::new("r_sma"); // SMA-cross signal → Bias, nested as a serializable `price`→`bias` leg. let sig = g.add(BlueprintNode::Composite(signal)); // pip branch (verbatim from sample_blueprint_with_sinks). let broker = g.add(SimBroker::builder(SYNTHETIC_PIP_SIZE)); // R branch: bias + price → RiskExecutor(vol_stop) → dense R-record. The stop knobs // are bound to the pinned constants (default) or left open as sweep axes (`stop_open`). // In `reduce` mode the per-cycle taps fold online: SeriesReducer folds the eq/ex // f64 series to one summary row, GatedRecorder retains only the gated R rows — the // O(cycles)→O(trades) memory win. The raw `Recorder`s (and the r_equity tap) are the // `--trace` path, where the full per-cycle series is persisted. let gate_col = PM_FIELD_NAMES .iter() .position(|&n| n == "closed_this_cycle") .expect("PM record has a closed_this_cycle column"); let eq = if reduce { g.add(SeriesReducer::builder(Firing::Any, tx_eq)) } else { g.add(Recorder::builder(vec![ScalarKind::F64], Firing::Any, tx_eq)) }; let ex = if reduce { g.add(SeriesReducer::builder(Firing::Any, tx_ex)) } else { g.add(Recorder::builder(vec![ScalarKind::F64], Firing::Any, tx_ex)) }; let exec = g.add(if stop_open { risk_executor_vol_open(1.0) } else { risk_executor(StopRule::Vol { length: R_SMA_STOP_LENGTH, k: R_SMA_STOP_K }, 1.0) }); let rrec = if reduce { g.add(GatedRecorder::builder(PM_RECORD_KINDS.to_vec(), gate_col, Firing::Any, tx_r)) } else { g.add(Recorder::builder(PM_RECORD_KINDS.to_vec(), Firing::Any, tx_r)) }; // Hoisted above the single main feed: the short-horizon vol proxy iff a // vol-slippage cost is actually wired (run path, non-reduce), so its `price` // inputs join the one `price_targets` array (no second feed call). let vol_proxy = match &cost { Some((cfg, _, _)) if !reduce && cfg.slip_vol_mult.is_some() => { let vhi = g.add(RollingMax::builder().named("slip_vol_hi").bind("length", Scalar::i64(SLIP_VOL_LENGTH))); let vlo = g.add(RollingMin::builder().named("slip_vol_lo").bind("length", Scalar::i64(SLIP_VOL_LENGTH))); let vrange = g.add(Sub::builder().named("slip_vol_range")); g.connect(vhi.output("value"), vrange.input("lhs")); g.connect(vlo.output("value"), vrange.input("rhs")); Some((vhi, vlo, vrange)) } _ => None, }; let price = g.source_role("price", ScalarKind::F64); let mut price_targets = vec![ sig.input("price"), broker.input("price"), exec.input("price"), ]; if let Some((vhi, vlo, _)) = vol_proxy { price_targets.push(vhi.input("series")); price_targets.push(vlo.input("series")); } g.feed(price, price_targets); g.connect(sig.output("bias"), broker.input("exposure")); g.connect(sig.output("bias"), ex.input("col[0]")); g.connect(sig.output("bias"), exec.input("bias")); g.connect(broker.output("equity"), eq.input("col[0]")); for (i, field) in PM_FIELD_NAMES.iter().enumerate() { g.connect(exec.output(field), rrec.input(COL_PORTS[i].as_str())); } if !reduce { // r_equity = cum_realized_r + unrealized_r — one tapped series for charting. let r_equity = g.add( LinComb::builder(2) .bind("weights[0]", Scalar::f64(1.0)) .bind("weights[1]", Scalar::f64(1.0)), ); let req = g.add(Recorder::builder(vec![ScalarKind::F64], Firing::Any, tx_req)); g.connect(exec.output("cum_realized_r"), r_equity.input("term[0]")); g.connect(exec.output("unrealized_r"), r_equity.input("term[1]")); g.connect(r_equity.output("value"), req.input("col[0]")); if let Some((cfg, tx_net, tx_cost)) = cost { // Build the active cost nodes (same conditional order), tracking the vol // node's index so its `cost[k].volatility` role can be fed below. let mut cost_nodes = Vec::new(); let mut vol_slot = None; if let Some(cpt) = cfg.const_cost { cost_nodes.push(ConstantCost::builder().bind("cost_per_trade", Scalar::f64(cpt))); } if let Some(svm) = cfg.slip_vol_mult { vol_slot = Some(cost_nodes.len()); cost_nodes.push(VolSlippageCost::builder().bind("slip_vol_mult", Scalar::f64(svm))); } if let Some(cpc) = cfg.carry_per_cycle { cost_nodes.push(CarryCost::builder().bind("carry_per_cycle", Scalar::f64(cpc))); } // A single cost_graph composite fans the shared PM-geometry into the active // cost nodes and sums their per-field charges into the run's cost streams. let cg = g.add(cost_graph(cost_nodes)); g.connect(exec.output("closed_this_cycle"), cg.input("closed")); g.connect(exec.output("open"), cg.input("open")); g.connect(exec.output("entry_price"), cg.input("entry_price")); g.connect(exec.output("stop_price"), cg.input("stop_price")); if let Some(k) = vol_slot { let (_, _, vrange) = vol_proxy.expect("vol proxy is built whenever slip_vol_mult is set"); let role: &'static str = format!("cost[{k}].volatility").leak(); g.connect(vrange.output("value"), cg.input(role)); } // net_r_equity = cum_realized_r + unrealized_r - Σcum_cost_in_r - Σopen_cost_in_r let net_eq = g.add( LinComb::builder(4) .bind("weights[0]", Scalar::f64(1.0)) .bind("weights[1]", Scalar::f64(1.0)) .bind("weights[2]", Scalar::f64(-1.0)) .bind("weights[3]", Scalar::f64(-1.0)), ); g.connect(exec.output("cum_realized_r"), net_eq.input("term[0]")); g.connect(exec.output("unrealized_r"), net_eq.input("term[1]")); g.connect(cg.output("cum_cost_in_r"), net_eq.input("term[2]")); g.connect(cg.output("open_cost_in_r"), net_eq.input("term[3]")); let net_rec = g.add(Recorder::builder(vec![ScalarKind::F64], Firing::Any, tx_net)); g.connect(net_eq.output("value"), net_rec.input("col[0]")); // The aggregate cost record summarize_r folds (col 0 per-close, col 2 window-end). let cost_rec = g.add(Recorder::builder( vec![ScalarKind::F64, ScalarKind::F64, ScalarKind::F64], Firing::Any, tx_cost, )); g.connect(cg.output("cost_in_r"), cost_rec.input("col[0]")); g.connect(cg.output("cum_cost_in_r"), cost_rec.input("col[1]")); g.connect(cg.output("open_cost_in_r"), cost_rec.input("col[2]")); } } g.build().expect("r_sma wiring resolves") } /// Resolve a `RunData` selector to the `(sources, window, pip_size)` triple the /// r-sma run paths feed to the harness: the built-in synthetic R stream, /// or a lazily-streamed real M1 close source (with its sidecar pip + probed window). /// One definition shared by `run_r_sma` and `run_signal_r` so the /// source/window/pip wiring cannot drift between the two paths. #[allow(clippy::type_complexity)] fn resolve_run_data( data: &RunData, env: &project::Env, ) -> ( Vec>, (Timestamp, Timestamp), f64, ) { match data { RunData::Synthetic => { let sources: Vec> = vec![Box::new(VecSource::new(r_sma_prices()))]; let window = window_of(&sources).expect("non-empty synthetic stream"); (sources, window, SYNTHETIC_PIP_SIZE) } RunData::Real { symbol, from, to } => { let (source, window, pip_size) = open_real_source(symbol, *from, *to, env); (vec![source], window, pip_size) } } } /// Run a signal blueprint through the R scaffolding: hash the signal, /// wrap it (broker + equity/exposure/R sinks), compile with `params`, bootstrap, /// run over `data`, and build the RunReport (manifest carries topology_hash). /// The single construction+run path shared by the `aura run ` CLI /// arm and its bit-identical test. fn run_signal_r( signal: Composite, params: &[Scalar], data: RunData, seed: u64, env: &project::Env, ) -> RunReport { let topo = topology_hash(&signal); // before signal is consumed let names: Vec = signal .param_space() .iter() .map(|p| p.name.clone()) .collect(); let (tx_eq, rx_eq) = mpsc::channel(); let (tx_ex, rx_ex) = mpsc::channel(); let (tx_r, rx_r) = mpsc::channel(); // The req tap (r_equity recorder) is wired but not persisted on this path; keep the // receiver alive so the sink's sends do not fail, but do not drain it. let (tx_req, _rx_req) = mpsc::channel(); let wrapped = wrap_r(signal, tx_eq, tx_ex, tx_r, tx_req, false, false, None); let flat = wrapped .compile_with_params(params) .expect("signal binds + wraps to a valid harness"); let mut h = Harness::bootstrap(flat).expect("valid r-sma harness"); let (sources, window, pip_size) = resolve_run_data(&data, env); h.run(sources); let eq_rows: Vec<(Timestamp, Vec)> = rx_eq.try_iter().collect(); let ex_rows: Vec<(Timestamp, Vec)> = rx_ex.try_iter().collect(); let r_rows: Vec<(Timestamp, Vec)> = rx_r.try_iter().collect(); let named_params: Vec<(String, Scalar)> = names.into_iter().zip(params.iter().copied()).collect(); let mut manifest = sim_optimal_manifest(named_params, window, seed, pip_size); manifest.broker = r_sma_broker_label(pip_size); manifest.topology_hash = Some(topo); manifest.project = env.provenance(); let mut metrics = summarize(&f64_field(&eq_rows, 0), &f64_field(&ex_rows, 0)); metrics.r = Some(summarize_r(&r_rows, &[])); RunReport { manifest, metrics } } /// Run one bootstrapped member of a loaded-signal sweep: the reduce-mode path /// (`SeriesReducer` folds eq/ex, `GatedRecorder` retains the gated R rows — the /// O(cycles)→O(trades) fold), shared by the live sweep AND reproduction so a /// reproduced member re-derives bit-identically (C1). `signal` is a freshly /// reloaded blueprint (`Composite` is `!Clone`); `point` is the member's bound /// cells; `space` gives the by-name manifest params; `topo` the shared signal hash. #[allow(clippy::too_many_arguments)] fn run_blueprint_member( signal: Composite, point: &[Cell], space: &[ParamSpec], sources: Vec>, window: (Timestamp, Timestamp), seed: u64, pip: f64, topo: &str, env: &project::Env, ) -> RunReport { let (tx_eq, rx_eq) = mpsc::channel(); let (tx_ex, rx_ex) = mpsc::channel(); let (tx_r, rx_r) = mpsc::channel(); let (tx_req, _rx_req) = mpsc::channel(); let mut h = wrap_r(signal, tx_eq, tx_ex, tx_r, tx_req, false, true, None) .bootstrap_with_cells(point) .expect("member bootstraps (point kind-checked against param_space)"); h.run(sources); let named = zip_params(space, point); // by-name params for the manifest record let mut manifest = sim_optimal_manifest(named, window, seed, pip); manifest.broker = r_sma_broker_label(pip); manifest.topology_hash = Some(topo.to_string()); manifest.project = env.provenance(); let r_rows: Vec<(Timestamp, Vec)> = rx_r.try_iter().collect(); let (total_pips, max_drawdown) = rx_eq .try_iter() .next() .map(|(_, row)| (row[0].as_f64(), row[1].as_f64())) .unwrap_or((0.0, 0.0)); let bias_sign_flips = rx_ex.try_iter().next().map(|(_, row)| row[2].as_i64() as u64).unwrap_or(0); let mut m = RunMetrics { total_pips, max_drawdown, bias_sign_flips, r: None }; m.r = Some(summarize_r(&r_rows, &[])); RunReport { manifest, metrics: m } } /// The exact wrapped probe the loaded-blueprint sweep resolves its axes /// against: the loaded signal wrapped in the r-sma scaffolding (stop bound, /// reduce, no cost), taps discarded. `param_space()` on it is the axis /// namespace `--axis` binds; `.axis()` consumes it to seed a sweep. Single /// source for the sweep terminal, the MC closed-check, AND `--list-axes`, so /// the listed names track the swept names by construction (incl. across #159's /// harness retirement). The reload is infallible under the SAME /// dispatch-boundary contract the callers already rely on: the doc is /// `blueprint_from_json`-validated at the `["sweep", ..]` / `["mc", ..]` /// boundary before this runs, so a malformed doc has already exited 2 and /// never reaches the `.expect`. fn blueprint_axis_probe(doc: &str, env: &project::Env) -> Composite { let signal = blueprint_from_json(doc, &|t| env.resolve(t)) .expect("doc parse-validated at the dispatch boundary; reload is infallible"); let (tx_eq, _) = mpsc::channel(); let (tx_ex, _) = mpsc::channel(); let (tx_r, _) = mpsc::channel(); let (tx_req, _) = mpsc::channel(); wrap_r(signal, tx_eq, tx_ex, tx_r, tx_req, false, true, None) } /// `aura sweep --list-axes`: one `:` line per open /// sweepable knob, in `param_space()` order; a closed blueprint prints nothing. /// Names are exactly what `--axis` binds (same probe as the sweep terminal). fn list_blueprint_axes(doc: &str, env: &project::Env) { for p in blueprint_axis_probe(doc, env).param_space() { println!("{}:{:?}", p.name, p.kind); // ScalarKind Debug -> I64/F64/Bool/Timestamp } } /// Sweep a serialized signal `doc` over user-named param-space axes — the structural /// twin of [`r_sma_sweep_family`], with three deviations. (1) The signal source is /// `wrap_r(blueprint_from_json(doc))` — a loaded blueprint, not the Rust-built /// r-sma graph. (2) The signal is RE-loaded from `doc` per member (a `Composite` is /// `!Clone`, so the throwaway param-space probe and each grid point each reload). (3) /// The axes are taken verbatim BY NAME (not the four suffix-resolved r-sma knobs): /// each `(name, vals)` is fed straight to the `SweepBinder`, so an unknown name or a /// kind mismatch surfaces as the sweep terminal's [`BindError`], rendered to a message /// string — a named error, never a panic. A FULLY BOUND (closed) blueprint has an empty /// `param_space` — nothing to sweep — and is refused up front with a clear message /// (the symmetric inverse of [`blueprint_mc_family`]'s closed-blueprint requirement), /// pre-empting the misleading `UnknownKnob()` the per-axis resolve would emit for a /// knob that is not unknown but bound out. Every member manifest carries the shared /// `topology_hash` of the loaded signal; reduce-mode fold, identical to the mirror's /// default (no-trace) arm. fn blueprint_sweep_family( doc: &str, axes: &[(String, Vec)], data: &DataSource, env: &project::Env, ) -> Result { // The doc is parse-validated at the dispatch boundary (with file-path context), // so every reload here is infallible: the builder has a single error contract — // the `BindError` returned by the sweep terminal — and no hidden process exit. let reload = |d: &str| { blueprint_from_json(d, &|t| env.resolve(t)) .expect("doc parse-validated at the dispatch boundary; reload is infallible") }; let topo = topology_hash(&reload(doc)); let pip = data.pip_size(); let window = data.full_window(env); // a single throwaway floated build, only to resolve param_space (borrow) then seed // the named axes (move) — its taps are never drained. The wrapped probe is the one // `blueprint_axis_probe` single-sources (identical `false, true, None` wrap). let probe = blueprint_axis_probe(doc, env); let space = probe.param_space(); // A fully bound blueprint has no open knob — there is nothing to sweep. Refuse it up // front (mirroring blueprint_mc_family's closed-blueprint guard, inverted) BEFORE the // per-axis resolve, which would otherwise return `UnknownKnob()` — misleading, as // the named knob is not unknown but bound out. Exit-free: the message is returned for // the IO wrapper to render + exit 2, so the precondition is unit-testable. if space.is_empty() { return Err( "this blueprint is fully bound; nothing to sweep (use `aura sweep --list-axes` \ to confirm there are no open knobs)" .to_string(), ); } // seed the named axes verbatim: the first via Composite::axis (consumes the probe), // the rest via SweepBinder::axis. resolve_axes name- and kind-checks them at the // sweep terminal, so an UnknownKnob / KindMismatch is returned, not panicked. let mut iter = axes.iter(); let (first_name, first_vals) = iter.next().expect("a blueprint sweep declares >= 1 axis"); let mut binder = probe.axis(first_name, first_vals.clone()); for (n, vals) in iter { binder = binder.axis(n, vals.clone()); } binder .sweep(|point| { // fresh per-member graph (Composite is !Clone, reload per member) run through // the shared reduce-mode member path — the same fn reproduction re-runs. run_blueprint_member(reload(doc), point, &space, data.run_sources(env), window, 0, pip, &topo, env) }) // render the sweep terminal's BindError to a message (the fn's String error contract), // so `UnknownKnob("nope")` still surfaces verbatim at the IO wrapper. .map_err(|e| format!("{e:?}")) } /// Sweep the LOADED blueprint over the user `--axis` grid on an in-sample window /// `[from,to]` — the windowed, lattice-carrying twin of `r_sma_sweep_over` and /// `blueprint_sweep_family`. `sweep_with_lattice` gives the grid lattice `--select /// plateau` needs. An unknown/kind-mismatched axis surfaces as `BindError` at the /// sweep terminal (no panic, no hidden exit) for the caller to render. fn blueprint_sweep_over( doc: &str, axes: &[(String, Vec)], from: Timestamp, to: Timestamp, data: &DataSource, env: &project::Env, ) -> Result<(SweepFamily, Vec), BindError> { let reload = |d: &str| { blueprint_from_json(d, &|t| env.resolve(t)) .expect("doc parse-validated at the dispatch boundary; reload is infallible") }; let pip = data.pip_size(); let topo = topology_hash(&reload(doc)); let probe = blueprint_axis_probe(doc, env); let space = probe.param_space(); let mut iter = axes.iter(); let (first_name, first_vals) = iter.next().expect("a blueprint walk-forward declares >= 1 axis"); let mut binder = probe.axis(first_name, first_vals.clone()); for (n, vals) in iter { binder = binder.axis(n, vals.clone()); } binder.sweep_with_lattice(|point| { let sources = data.windowed_sources(from, to, env); let window = window_of(&sources).expect("non-empty in-sample window"); run_blueprint_member(reload(doc), point, &space, sources, window, 0, pip, &topo, env) }) } /// Run the winner params over an out-of-sample window `[from,to]` on the loaded /// blueprint — the loaded-member analog of `run_oos_r`. The reduce-mode member /// (`run_blueprint_member`) retains R-metrics, not a raw pip curve, so the stitching /// segment is empty (an empty segment leaves the stitched curve unbroken). #[allow(clippy::too_many_arguments)] fn run_oos_blueprint( doc: &str, params: &[Cell], space: &[ParamSpec], from: Timestamp, to: Timestamp, topo: &str, data: &DataSource, env: &project::Env, ) -> (Vec<(Timestamp, f64)>, RunReport) { let reload = blueprint_from_json(doc, &|t| env.resolve(t)) .expect("doc parse-validated at the dispatch boundary; reload is infallible"); let pip = data.pip_size(); let sources = data.windowed_sources(from, to, env); let window = window_of(&sources).expect("non-empty out-of-sample window"); let report = run_blueprint_member(reload, params, space, sources, window, 0, pip, topo, env); (Vec::new(), report) } /// The loaded-blueprint IS-refit walk-forward: per IS window, re-optimize the /// blueprint over the user `--axis` grid, select by `sqn_normalized`, run the /// winner OOS. Structural twin of `walkforward_family`'s `RSma` arm — the /// generic `walk_forward` driver + `select_winner` reused; only the per-window /// sweep/OOS source the loaded blueprint. In-closure errors (a bad `--axis`) /// `exit(2)` with the sweep terminal's message, as the hard-wired arm does. fn blueprint_walkforward_family( doc: &str, axes: &[(String, Vec)], data: &DataSource, select: Selection, env: &project::Env, ) -> WalkForwardResult { let span = data.wf_full_span(env); let (is_len, oos_len, step) = data.wf_window_sizes(); let roller = match WindowRoller::new(span, is_len, oos_len, step, RollMode::Rolling) { Ok(r) => r, Err(e) => { eprintln!("aura: walk-forward window too short for one IS+OOS span: {e:?}"); std::process::exit(2); } }; let space = blueprint_axis_probe(doc, env).param_space(); let topo = topology_hash(&blueprint_from_json(doc, &|t| env.resolve(t)) .expect("doc parse-validated at the dispatch boundary; reload is infallible")); // Validate the `--axis` grid ONCE at the dispatch boundary, mirroring `aura sweep` // (which resolves its axes a single time before any member runs). `walk_forward` fans // the per-window closure out across the windows in parallel, so a `BindError` raised // *inside* the closure would `eprintln!`+`exit(2)` from several windows before any one // exit lands — a racy, duplicated rejection (#177). Axis resolution is window-agnostic, // so pre-flighting the first IS window here surfaces the error exactly once (it fails in // `resolve_axes`, before any member runs); a per-window resolve then cannot re-raise. let first_is = WindowRoller::new(span, is_len, oos_len, step, RollMode::Rolling) .expect("roller config validated just above") .next() .expect("roller yields >= 1 window (validated above)") .is; if let Err(e) = blueprint_sweep_over(doc, axes, first_is.0, first_is.1, data, env) { eprintln!("aura: {e:?}"); std::process::exit(2); } walk_forward(roller, space.clone(), |w: WindowBounds| { let (is_family, lattice) = blueprint_sweep_over(doc, axes, w.is.0, w.is.1, data, env) .expect("axes validated in the dispatch-boundary pre-flight"); let (best, selection) = match select_winner(&is_family, "sqn_normalized", select, Some(&lattice)) { Ok(v) => v, Err(msg) => { eprintln!("aura: {msg}"); std::process::exit(2); } }; let (oos_equity, mut oos_report) = run_oos_blueprint(doc, &best.params, &space, w.oos.0, w.oos.1, &topo, data, env); oos_report.manifest.selection = Some(selection); WindowRun { chosen_params: best.params, oos_equity, oos_report } }) } /// A fresh seeded synthetic price walk for one Monte-Carlo draw — the `mc_family` /// pattern (a distinct realization per seed). A FIXED `SyntheticSpec` shared by the /// `aura mc ` persist path AND the reproduce MonteCarlo branch, so the /// seed->walk reconstruction is bit-exact (C1). Length 60 comfortably warms the loaded /// r-sma graph (SMA slow=4 + the len-3 vol stop) so draws produce differing trades. fn synthetic_walk_sources(seed: u64) -> Vec> { let spec = SyntheticSpec { start: 1.0, len: 60, step: 1 }; vec![Box::new(spec.source(seed))] } /// Build a Monte-Carlo family from a loaded CLOSED signal blueprint: run the fixed /// blueprint across `n_seeds` seeds, each seed drawing a distinct synthetic walk. The /// blueprint must be CLOSED (empty wrapped `param_space`) — MC binds no axis, so a free /// knob has no binder; an OPEN blueprint yields a named `Err` (exit-free like the sibling /// [`blueprint_sweep_family`]: the IO wrapper [`run_blueprint_mc`] renders it to stderr + /// exit 2) before any run, pre-empting the `compile_with_params` arity panic. Each draw /// runs the shared reduce-mode member path (`run_blueprint_member`, the same fn reproduce /// re-runs), so reproduction is bit-identical (C1); every member carries the shared /// `topology_hash`. fn blueprint_mc_family( doc: &str, n_seeds: u64, data: &DataSource, env: &project::Env, ) -> Result { let reload = |d: &str| { blueprint_from_json(d, &|t| env.resolve(t)) .expect("doc parse-validated at the dispatch boundary; reload is infallible") }; let topo = topology_hash(&reload(doc)); let pip = data.pip_size(); // probe the wrapped param_space (the same probe the sweep resolves against); // MC needs it empty. `blueprint_axis_probe` is the single source of that wrap. let space = blueprint_axis_probe(doc, env).param_space(); if !space.is_empty() { // Exit-free like blueprint_sweep_family: the builder's single error contract is this // returned message (no hidden process exit), so the rejection is unit-testable; the IO // wrapper run_blueprint_mc renders it to stderr + exit 2 at the boundary. return Err(format!( "mc requires a closed blueprint (no free parameters); {} free knob(s) — \ bind them or use `aura sweep --axis`", space.len() )); } // Closed blueprint -> an empty base point (as `aura run `); the MC // draws vary the SEED, not a tuning param (C12 axis 4). Delegate the disjoint C1 draws // to the shared `monte_carlo` helper — it runs them in parallel across sims (invariant 1), // deterministic in seed-input order — exactly as the sibling `mc_family` does. Each draw // re-runs the shared reduce-mode member path over its own seeded synthetic walk. let seeds: Vec = (1..=n_seeds).collect(); let base_point: Vec = Vec::new(); let family = monte_carlo(&base_point, &seeds, |seed, _base| { let sources = synthetic_walk_sources(seed); let window = window_of(&sources).expect("non-empty synthetic walk"); run_blueprint_member(reload(doc), &[], &space, sources, window, seed, pip, &topo, env) }); // Silent-vacuous MC guard (refuse-don't-guess, C10): with >= 2 seeds, if every draw's // metrics are bit-identical to the first, no seed reached a distinguishable realization — // the strategy never warmed over the fixed synthetic walk (e.g. a lookback as deep as the // walk is long), so the "distribution" is a single point masquerading as a family: a wrong // result with no error. Compare `metrics`, not the whole `RunReport` — the manifest's // `seed` differs per draw by construction, so a whole-report compare could never detect the // collapse; the metrics are the realization the seed is meant to move. A single-draw MC // (n == 1) is trivially "all identical" and is NOT this cross-seed condition, so it passes. if family.draws.len() >= 2 && family .draws .iter() .all(|d| d.report.metrics == family.draws[0].report.metrics) { return Err( "mc is vacuous: every seed produced an identical result — the strategy never warmed \ over the synthetic walk, so no seed reached a distinguishable realization; use a \ shallower-lookback blueprint or a longer walk" .to_string(), ); } Ok(family) } /// `aura sweep --axis = …`: sweep a loaded signal over its /// named param-space axes (the cycle-2 World/C21 verb). Builds the family via /// [`blueprint_sweep_family`], surfaces an unknown / kind-mismatched axis as a named /// error (stderr + exit 2, never a panic), ALWAYS records it as a `FamilyKind::Sweep` /// family (C18/C21 lineage, exactly as the other family verbs), and prints each member /// carrying the assigned `family_id` via [`family_member_line`] — so a printed member is /// linkable back to its stored family, like `run_sweep` / `run_walkforward`. /// /// Divergence from `run_sweep` (the one place it does less): the blueprint sweep is /// reduce-only this cycle — [`blueprint_sweep_family`] writes no per-member traces — so /// `persist`/`--trace` neither writes trace files nor reserves a trace-store name (that /// reservation would guard a write that never happens, and could spuriously reject a /// valid sweep on a name collision). `persist` is therefore not yet load-bearing here; /// it is retained for the deferred per-member trace path. The family record itself is /// written unconditionally, so lineage (C18/C21) holds whether or not `--trace` is given. fn run_blueprint_sweep( doc: &str, axes: &[(String, Vec)], name: &str, persist: bool, data: DataSource, env: &project::Env, ) { let _ = persist; // reserved for the deferred per-member trace path; the family record below is unconditional let family = blueprint_sweep_family(doc, axes, &data, env).unwrap_or_else(|e| { eprintln!("aura: {e}"); std::process::exit(2); }); let reg = env.registry(); // Store the canonical blueprint ONCE, keyed by the family's shared topology_hash — // exactly the bytes whose SHA256 the members carry (#164 byte-canonical, round-trip // idempotent). One stored topology per family (C18/C11/C12). let topo = family.points[0] .report .manifest .topology_hash .clone() .expect("a blueprint sweep stamps every member's topology_hash"); let canonical = blueprint_to_json( &blueprint_from_json(doc, &|t| env.resolve(t)) .expect("doc parse-validated at the dispatch boundary"), ) .expect("a loaded blueprint re-serializes"); reg.put_blueprint(&topo, &canonical).unwrap_or_else(|e| { eprintln!("aura: {e}"); std::process::exit(1); }); // Record the family unconditionally (C18/C21 lineage), exactly like `run_sweep`. let id = match reg.append_family(name, FamilyKind::Sweep, &sweep_member_reports(&family)) { Ok(id) => id, Err(e) => { eprintln!("aura: {e}"); std::process::exit(1); } }; for pt in &family.points { println!("{}", family_member_line(&id, &pt.report)); } } /// `aura walkforward --axis …`: build the loaded-blueprint /// IS-refit walk-forward, store the canonical blueprint ONCE keyed by the shared /// `topology_hash` (the C18 hook, so `aura reproduce` re-derives it), record it as /// a `FamilyKind::WalkForward` family, and print each OOS member line + the summary. /// Mirrors `run_blueprint_sweep` (content-addressed family verb; no ensure_name_free). fn run_blueprint_walkforward( doc: &str, axes: &[(String, Vec)], name: &str, data: DataSource, select: Selection, env: &project::Env, ) { let result = blueprint_walkforward_family(doc, axes, &data, select, env); let reg = env.registry(); let topo = result.windows[0] .run .oos_report .manifest .topology_hash .clone() .expect("a blueprint walk-forward stamps every member's topology_hash"); let canonical = blueprint_to_json( &blueprint_from_json(doc, &|t| env.resolve(t)).expect("doc parse-validated at the dispatch boundary"), ) .expect("a loaded blueprint re-serializes"); reg.put_blueprint(&topo, &canonical).unwrap_or_else(|e| { eprintln!("aura: {e}"); std::process::exit(1); }); let id = match reg.append_family(name, FamilyKind::WalkForward, &walkforward_member_reports(&result)) { Ok(id) => id, Err(e) => { eprintln!("aura: {e}"); std::process::exit(1); } }; for w in &result.windows { println!("{}", family_member_line(&id, &w.run.oos_report)); } println!("{}", walkforward_summary_json(&result)); } /// `aura mc --seeds N`: build a Monte-Carlo family from a loaded CLOSED /// blueprint (the World/C21 verb), store the canonical blueprint ONCE keyed by the shared /// `topology_hash` (the 0094 hook, so `aura reproduce` re-derives it), record it as a /// `FamilyKind::MonteCarlo` family (C18/C21 lineage), and print each draw's member line /// (carrying the seed) plus the aggregate — mirroring `run_mc` / `run_blueprint_sweep`. fn run_blueprint_mc(doc: &str, n_seeds: u64, name: &str, data: DataSource, env: &project::Env) { let family = blueprint_mc_family(doc, n_seeds, &data, env).unwrap_or_else(|e| { eprintln!("aura: {e}"); std::process::exit(2); }); let reg = env.registry(); // Store the canonical blueprint ONCE, keyed by the family's shared topology_hash. let topo = family.draws[0] .report .manifest .topology_hash .clone() .expect("a blueprint mc stamps every member's topology_hash"); let canonical = blueprint_to_json( &blueprint_from_json(doc, &|t| env.resolve(t)) .expect("doc parse-validated at the dispatch boundary"), ) .expect("a loaded blueprint re-serializes"); reg.put_blueprint(&topo, &canonical).unwrap_or_else(|e| { eprintln!("aura: {e}"); std::process::exit(1); }); let id = match reg.append_family(name, FamilyKind::MonteCarlo, &mc_member_reports(&family)) { Ok(id) => id, Err(e) => { eprintln!("aura: {e}"); std::process::exit(1); } }; for draw in &family.draws { println!("{}", mc_member_line(&id, draw.seed, &draw.report)); } println!("{}", mc_aggregate_json(&family.aggregate)); } /// The r-sma harness with its signal leg swapped for a Donchian channel breakout: /// `close -> Delay(1) -> {RollingMax,RollingMin}(channel) -> {Gt,Gt} -> {Latch,Latch} /// -> Sub = bias in {-1,0,+1}`. The one Delay(1) on close feeds both rolling nodes, so /// each channel covers `close[t-N..t-1]` (the C2 guard: the current bar is excluded). /// `channel = None` leaves the lengths open (unused today; the family binds them); the /// stop is bound (defines R), identical downstream to r_sma_graph. #[allow(clippy::type_complexity, clippy::too_many_arguments)] fn r_breakout_graph( tx_eq: mpsc::Sender<(Timestamp, Vec)>, tx_ex: mpsc::Sender<(Timestamp, Vec)>, tx_r: mpsc::Sender<(Timestamp, Vec)>, tx_req: mpsc::Sender<(Timestamp, Vec)>, channel: Option, stop_length: i64, stop_k: f64, reduce: bool, ) -> Composite { let mut g = GraphBuilder::new("r_breakout"); // Donchian breakout signal leg. let delay = g.add(Delay::builder().bind("lag", Scalar::i64(1))); let mut mx_b = RollingMax::builder().named("channel_hi"); let mut mn_b = RollingMin::builder().named("channel_lo"); if let Some(n) = channel { mx_b = mx_b.bind("length", Scalar::i64(n)); mn_b = mn_b.bind("length", Scalar::i64(n)); } let mx = g.add(mx_b); let mn = g.add(mn_b); let gt_up = g.add(Gt::builder()); let gt_down = g.add(Gt::builder()); let up_latch = g.add(Latch::builder()); let down_latch = g.add(Latch::builder()); let exposure = g.add(Sub::builder()); // up_latch - down_latch -> bias in {-1,0,+1} // pip branch (verbatim from r_sma_graph). let broker = g.add(SimBroker::builder(SYNTHETIC_PIP_SIZE)); let gate_col = PM_FIELD_NAMES .iter() .position(|&n| n == "closed_this_cycle") .expect("PM record has a closed_this_cycle column"); let eq = if reduce { g.add(SeriesReducer::builder(Firing::Any, tx_eq)) } else { g.add(Recorder::builder(vec![ScalarKind::F64], Firing::Any, tx_eq)) }; let ex = if reduce { g.add(SeriesReducer::builder(Firing::Any, tx_ex)) } else { g.add(Recorder::builder(vec![ScalarKind::F64], Firing::Any, tx_ex)) }; let exec = g.add(risk_executor(StopRule::Vol { length: stop_length, k: stop_k }, 1.0)); let rrec = if reduce { g.add(GatedRecorder::builder(PM_RECORD_KINDS.to_vec(), gate_col, Firing::Any, tx_r)) } else { g.add(Recorder::builder(PM_RECORD_KINDS.to_vec(), Firing::Any, tx_r)) }; let price = g.source_role("price", ScalarKind::F64); g.feed( price, [ delay.input("series"), gt_up.input("a"), gt_down.input("b"), broker.input("price"), exec.input("price"), ], ); g.connect(delay.output("value"), mx.input("series")); g.connect(delay.output("value"), mn.input("series")); g.connect(mx.output("value"), gt_up.input("b")); g.connect(mn.output("value"), gt_down.input("a")); g.connect(gt_up.output("value"), up_latch.input("set")); g.connect(gt_down.output("value"), up_latch.input("reset")); g.connect(gt_down.output("value"), down_latch.input("set")); g.connect(gt_up.output("value"), down_latch.input("reset")); g.connect(up_latch.output("value"), exposure.input("lhs")); g.connect(down_latch.output("value"), exposure.input("rhs")); g.connect(exposure.output("value"), broker.input("exposure")); g.connect(exposure.output("value"), ex.input("col[0]")); g.connect(exposure.output("value"), exec.input("bias")); g.connect(broker.output("equity"), eq.input("col[0]")); for (i, field) in PM_FIELD_NAMES.iter().enumerate() { g.connect(exec.output(field), rrec.input(COL_PORTS[i].as_str())); } if !reduce { let r_equity = g.add( LinComb::builder(2) .bind("weights[0]", Scalar::f64(1.0)) .bind("weights[1]", Scalar::f64(1.0)), ); let req = g.add(Recorder::builder(vec![ScalarKind::F64], Firing::Any, tx_req)); g.connect(exec.output("cum_realized_r"), r_equity.input("term[0]")); g.connect(exec.output("unrealized_r"), r_equity.input("term[1]")); g.connect(r_equity.output("value"), req.input("col[0]")); } g.build().expect("r_breakout wiring resolves") } /// The EWMA Bollinger-band mean-reversion candidate, mirroring /// `r_breakout_graph` but swapping the signal leg: fade deviation from a /// rolling mean (price above `mean + k*sigma` -> short; below `mean - k*sigma` -> /// long), latched +-1. sigma = `Sqrt(Ema((price-mean)^2))` (deviation squared /// then smoothed, the vol_stop shape — no catastrophic cancellation, no NaN). /// No Delay: the current bar legitimately belongs to its own band (causal, C2). /// `window` gangs the mean Ema and the variance Ema (one Bollinger window); /// `band_k` is the band half-width in sigma. Everything below the signal leg is /// byte-identical to `r_breakout_graph`. #[allow(clippy::type_complexity, clippy::too_many_arguments)] fn r_meanrev_graph( tx_eq: mpsc::Sender<(Timestamp, Vec)>, tx_ex: mpsc::Sender<(Timestamp, Vec)>, tx_r: mpsc::Sender<(Timestamp, Vec)>, tx_req: mpsc::Sender<(Timestamp, Vec)>, window: Option, band_k: f64, stop_length: i64, stop_k: f64, reduce: bool, ) -> Composite { let mut g = GraphBuilder::new("r_meanrev"); // EWMA Bollinger-band mean-reversion signal leg (the ONLY change vs breakout). let (mut mean_b, mut var_b) = (Ema::builder().named("mean_window"), Ema::builder().named("var_window")); if let Some(n) = window { mean_b = mean_b.bind("length", Scalar::i64(n)); var_b = var_b.bind("length", Scalar::i64(n)); } let mean = g.add(mean_b); let dev = g.add(Sub::builder()); // price - mean let sq = g.add(Mul::builder()); // dev * dev let var = g.add(var_b); // EWMA variance let sigma = g.add(Sqrt::builder()); // sigma (price units) let band = g.add(LinComb::builder(1).bind("weights[0]", Scalar::f64(band_k))); // k*sigma let upper = g.add(Add::builder()); // mean + k*sigma let lower = g.add(Sub::builder()); // mean - k*sigma let gt_hi = g.add(Gt::builder()); // price > upper -> overextended up -> fade short let gt_lo = g.add(Gt::builder()); // lower > price -> overextended down -> fade long let short_latch = g.add(Latch::builder()); let long_latch = g.add(Latch::builder()); let exposure = g.add(Sub::builder()); // long_latch - short_latch -> bias in {-1,0,+1} // pip branch (VERBATIM from r_breakout_graph). let broker = g.add(SimBroker::builder(SYNTHETIC_PIP_SIZE)); let gate_col = PM_FIELD_NAMES .iter() .position(|&n| n == "closed_this_cycle") .expect("PM record has a closed_this_cycle column"); let eq = if reduce { g.add(SeriesReducer::builder(Firing::Any, tx_eq)) } else { g.add(Recorder::builder(vec![ScalarKind::F64], Firing::Any, tx_eq)) }; let ex = if reduce { g.add(SeriesReducer::builder(Firing::Any, tx_ex)) } else { g.add(Recorder::builder(vec![ScalarKind::F64], Firing::Any, tx_ex)) }; let exec = g.add(risk_executor(StopRule::Vol { length: stop_length, k: stop_k }, 1.0)); let rrec = if reduce { g.add(GatedRecorder::builder(PM_RECORD_KINDS.to_vec(), gate_col, Firing::Any, tx_r)) } else { g.add(Recorder::builder(PM_RECORD_KINDS.to_vec(), Firing::Any, tx_r)) }; let price = g.source_role("price", ScalarKind::F64); g.feed( price, [ mean.input("series"), dev.input("lhs"), gt_hi.input("a"), gt_lo.input("b"), broker.input("price"), exec.input("price"), ], ); g.connect(mean.output("value"), dev.input("rhs")); g.connect(dev.output("value"), sq.input("lhs")); g.connect(dev.output("value"), sq.input("rhs")); // square: feed dev to both legs g.connect(sq.output("value"), var.input("series")); g.connect(var.output("value"), sigma.input("value")); g.connect(sigma.output("value"), band.input("term[0]")); g.connect(mean.output("value"), upper.input("lhs")); g.connect(band.output("value"), upper.input("rhs")); // upper = mean + k*sigma g.connect(mean.output("value"), lower.input("lhs")); g.connect(band.output("value"), lower.input("rhs")); // lower = mean - k*sigma g.connect(upper.output("value"), gt_hi.input("b")); g.connect(lower.output("value"), gt_lo.input("a")); g.connect(gt_hi.output("value"), short_latch.input("set")); g.connect(gt_lo.output("value"), short_latch.input("reset")); g.connect(gt_lo.output("value"), long_latch.input("set")); g.connect(gt_hi.output("value"), long_latch.input("reset")); g.connect(long_latch.output("value"), exposure.input("lhs")); g.connect(short_latch.output("value"), exposure.input("rhs")); g.connect(exposure.output("value"), broker.input("exposure")); g.connect(exposure.output("value"), ex.input("col[0]")); g.connect(exposure.output("value"), exec.input("bias")); g.connect(broker.output("equity"), eq.input("col[0]")); for (i, field) in PM_FIELD_NAMES.iter().enumerate() { g.connect(exec.output(field), rrec.input(COL_PORTS[i].as_str())); } if !reduce { let r_equity = g.add( LinComb::builder(2) .bind("weights[0]", Scalar::f64(1.0)) .bind("weights[1]", Scalar::f64(1.0)), ); let req = g.add(Recorder::builder(vec![ScalarKind::F64], Firing::Any, tx_req)); g.connect(exec.output("cum_realized_r"), r_equity.input("term[0]")); g.connect(exec.output("unrealized_r"), r_equity.input("term[1]")); g.connect(r_equity.output("value"), req.input("col[0]")); } g.build().expect("r_meanrev wiring resolves") } /// `aura run --harness r-sma [--real [--from][--to]] [--trace ]`: build the /// dual-tap r-sma harness, run it on synthetic or real M1 data, fold the pip taps via /// `summarize` and the dense R-record via `summarize_r`, and attach the R block as /// `RunMetrics.r = Some(..)`. Pure/deterministic (C1). `cost` is the optional flat /// round-trip cost per trade (price units): `Some(c)` wires the `ConstantCost` node and /// the `net_r_equity` tap and folds the cost stream into `net_expectancy_r`; `None` is the /// frictionless gross-R baseline (net == gross, no extra tap, byte-unchanged). fn run_r_sma( data: RunData, trace: Option<&str>, const_cost: Option, slip_vol_mult: Option, carry_per_cycle: Option, env: &project::Env, ) -> RunReport { if let Some(n) = trace && let Err(e) = env.trace_store().ensure_name_free(n, WriteKind::Run) { eprintln!("aura: {e}"); std::process::exit(1); } let (tx_eq, rx_eq) = mpsc::channel(); let (tx_ex, rx_ex) = mpsc::channel(); let (tx_r, rx_r) = mpsc::channel(); let (tx_req, rx_req) = mpsc::channel(); let (tx_net, rx_net) = mpsc::channel(); let (tx_cost, rx_cost) = mpsc::channel(); let cost_bundle = if const_cost.is_some() || slip_vol_mult.is_some() || carry_per_cycle.is_some() { Some(( CostConfig { const_cost, slip_vol_mult, carry_per_cycle, }, tx_net, tx_cost, )) } else { None }; let flat = r_sma_graph( tx_eq, tx_ex, tx_r, tx_req, Some(R_SMA_FAST), Some(R_SMA_SLOW), false, false, cost_bundle, ) .compile_with_params(&[]) .expect("valid r-sma blueprint"); let mut h = Harness::bootstrap(flat).expect("valid r-sma harness"); let (sources, window, pip_size) = resolve_run_data(&data, env); h.run(sources); let eq_rows: Vec<(Timestamp, Vec)> = rx_eq.try_iter().collect(); let ex_rows: Vec<(Timestamp, Vec)> = rx_ex.try_iter().collect(); let r_rows: Vec<(Timestamp, Vec)> = rx_r.try_iter().collect(); let req_rows: Vec<(Timestamp, Vec)> = rx_req.try_iter().collect(); // The cost taps drain empty when `cost` is `None` (no cost node wired): an empty // `net_rows` suppresses the `net_r_equity` trace and an empty `cost_rows` folds to // net == gross — the byte-unchanged frictionless baseline. let net_rows: Vec<(Timestamp, Vec)> = rx_net.try_iter().collect(); let cost_rows: Vec<(Timestamp, Vec)> = rx_cost.try_iter().collect(); let mut manifest = sim_optimal_manifest( vec![ ("sma_fast".to_string(), Scalar::i64(R_SMA_FAST)), ("sma_slow".to_string(), Scalar::i64(R_SMA_SLOW)), ("bias_scale".to_string(), Scalar::f64(R_SMA_BIAS_SCALE)), ("stop_length".to_string(), Scalar::i64(R_SMA_STOP_LENGTH)), // vol_stop EWMA length ("stop_k".to_string(), Scalar::f64(R_SMA_STOP_K)), // vol_stop multiplier (1R = k·σ) ], window, 0, pip_size, ); manifest.broker = r_sma_broker_label(pip_size); manifest.topology_hash = Some(topology_hash(&sma_signal( Some(R_SMA_FAST), Some(R_SMA_SLOW), ))); // Project provenance is paired with `topology_hash` — the same #158 // reproducibility-anchor group, exactly the three spots that already stamp // `topology_hash` today (this fn, `run_signal_r`, `run_blueprint_member`). manifest.project = env.provenance(); if let Some(name) = trace { persist_traces_r(name, &manifest, &eq_rows, &ex_rows, &req_rows, &net_rows, env); } let mut metrics = summarize(&f64_field(&eq_rows, 0), &f64_field(&ex_rows, 0)); metrics.r = Some(summarize_r(&r_rows, &cost_rows)); RunReport { manifest, metrics } } /// Persist a r-sma run's taps: equity (off the SimBroker), exposure (off the Bias), /// r_equity = cum_realized_r + unrealized_r (off the RiskExecutor), and — only on a cost /// run — net_r_equity (gross r_equity minus the cost-in-R taps). Separate from the two-tap /// `persist_traces` so the pip handlers stay byte-unchanged on disk; the `net_r_equity` tap /// is emitted only when `net_rows` is non-empty, so a no-cost run's on-disk trace set is /// byte-unchanged too. fn persist_traces_r( name: &str, manifest: &RunManifest, eq_rows: &[(Timestamp, Vec)], ex_rows: &[(Timestamp, Vec)], req_rows: &[(Timestamp, Vec)], net_rows: &[(Timestamp, Vec)], env: &project::Env, ) { let mut taps = vec![ ColumnarTrace::from_rows("equity", &[ScalarKind::F64], eq_rows), ColumnarTrace::from_rows("exposure", &[ScalarKind::F64], ex_rows), ColumnarTrace::from_rows("r_equity", &[ScalarKind::F64], req_rows), ]; if !net_rows.is_empty() { taps.push(ColumnarTrace::from_rows("net_r_equity", &[ScalarKind::F64], net_rows)); } if let Err(e) = env.trace_store().write(name, manifest, &taps) { eprintln!("aura: trace persist failed: {e}"); std::process::exit(1); } } /// Which built-in harness `aura run` drives. A fixed compile-time enumeration over /// Rust-authored harnesses — NOT a runtime node registry and NOT a DSL (C9/C17): the /// CLI *runs* an authored harness, it does not wire one. #[derive(Debug)] enum HarnessKind { Sma, Macd, RSma, } /// The parsed `aura run` invocation: a harness, a data source, an optional trace name, and /// an optional flat round-trip cost per trade (`--cost-per-trade`, r-sma only this cycle). #[derive(Debug)] struct RunArgs { harness: HarnessKind, data: RunData, trace: Option, cost: Option, slip_vol_mult: Option, carry_per_cycle: Option, } /// Parse the `--params` value: a JSON array of externally-tagged `Scalar` cells /// (`[{"I64":2},{"F64":0.5}]`, the #155 wire form `Scalar` derives via serde). The cells /// bind positionally against the loaded signal's `param_space`. A malformed array is /// refused with the flag named, never silently coerced to an empty param vector /// (refuse-don't-guess, C10) — a dropped param would bootstrap a *different* graph. fn parse_param_cells(json: &str) -> Result, String> { serde_json::from_str(json).map_err(|e| format!("--params: {e}")) } /// Lex a `--axis` CSV into typed Scalars by shape: an integer-shaped token is i64, /// otherwise f64. `resolve_axes` kind-checks each value against the param's declared /// kind afterwards (a mismatch is a named error, not a panic). fn parse_scalar_csv(csv: &str) -> Option> { csv.split(',').map(|t| { let t = t.trim(); if t.is_empty() { return None; } match t.parse::() { Ok(i) => Some(Scalar::i64(i)), Err(_) => t.parse::().ok().map(Scalar::f64), } }).collect() } /// Route a parsed `run` invocation to its harness handler. The compile-time `match` IS /// the selector — the harnesses are Rust-authored built-ins, picked by name (C9/C17). fn run_dispatch(args: RunArgs, env: &project::Env) -> Result { let trace = args.trace.as_deref(); Ok(match (args.harness, args.data) { (HarnessKind::Sma, RunData::Synthetic) => run_sample(trace, env), (HarnessKind::Macd, RunData::Synthetic) => run_macd(trace, env), (HarnessKind::RSma, data) => { run_r_sma(data, trace, args.cost, args.slip_vol_mult, args.carry_per_cycle, env) } (HarnessKind::Sma, RunData::Real { symbol, from, to }) => { run_sample_real(&symbol, from, to, trace, env) } (HarnessKind::Macd, RunData::Real { .. }) => { return Err("the macd harness has no --real form".to_string()) } }) } // ============================== clap parser surface ============================== // The declarative argument grammar. clap owns argv tokenizing, scoped `--help`, // `--version`, `--flag=value`, `--`, and long-option abbreviation; the `dispatch_*` // handlers below convert each `*Cmd` into the argument shapes the existing execution // fns accept, reusing the value helpers (`Strategy`, `RGrid`, `Selection`, // `DataSource::from_choice`, `parse_scalar_csv`, `parse_csv_list`, `parse_select`, // `parse_param_cells`). The four dual-grammar subcommands carry an optional // `[blueprint]` positional; a first-positional that names an existing `.json` file // (`is_blueprint_file`) selects the loaded-blueprint branch, otherwise the built-in // grammar. Usage errors (clap parse + argv-applicability guards) exit 2; runtime failures exit 1. /// The `aura` root parser. `#[command(version)]` reads `CARGO_PKG_VERSION` /// (the workspace `0.1.0`), so `aura --version` prints `aura 0.1.0`. #[derive(Parser)] #[command(name = "aura", version, about = "Author, backtest, and validate trading strategies — research CLI", infer_long_args = true)] struct Cli { #[command(subcommand)] command: Command, /// Load the project dylib from target/release instead of target/debug. #[arg(long, global = true)] release: bool, } #[derive(Subcommand)] enum Command { /// Run a single backtest (built-in harness or a loaded blueprint). Run(RunCmd), /// Render a recorded run's trace to an HTML chart. Chart(ChartCmd), /// Emit / construct / introspect a graph. Graph(GraphCmd), /// Sweep a parameter grid over a strategy or a loaded blueprint. Sweep(SweepCmd), /// Walk-forward validation over a strategy or a loaded blueprint. Walkforward(WalkforwardCmd), /// Grade one candidate across multiple instruments. Generalize(GeneralizeCmd), /// Monte-Carlo over synthetic draws, an R-bootstrap, or a loaded blueprint. Mc(McCmd), /// List or inspect recorded run families. Runs(RunsCmd), /// Reproduce a recorded family by content id. Reproduce(ReproduceCmd), /// Scaffold a new research project crate. New(NewCmd), /// Validate, introspect, and register process documents (methodology). Process(research_docs::ProcessCmd), /// Validate, introspect, and register campaign documents (experiment intent). Campaign(research_docs::CampaignCmd), } #[derive(Args)] struct ChartCmd { /// The recorded run or family name to chart. name: String, /// Chart only the given tap. #[arg(long)] tap: Option, /// Render stacked panels instead of an overlay. #[arg(long)] panels: bool, } #[derive(Args)] struct NewCmd { /// Directory / crate name to create. name: String, /// Engine checkout the project depends on (path deps). Default: the /// engine root this binary was built from. #[arg(long)] engine_path: Option, /// Vocabulary namespace (default: the name with dashes as underscores). #[arg(long)] namespace: Option, } #[derive(Args)] struct GraphCmd { #[command(subcommand)] sub: Option, } #[derive(Subcommand)] enum GraphSub { /// Construct a graph from a stdin op-list. Build, /// Introspect a graph. Introspect(GraphIntrospectCmd), } #[derive(Args)] struct GraphIntrospectCmd { /// List the closed node vocabulary (one node type per line). #[arg(long)] vocabulary: bool, /// Describe one node type's ports by name. #[arg(long)] node: Option, /// List the graph's unwired (unbound) ports. #[arg(long)] unwired: bool, /// Print the graph's content id (topology hash). #[arg(long)] content_id: bool, /// Print the graph's topology-identity id (debug names stripped). #[arg(long)] identity_id: bool, } #[derive(Args)] struct GeneralizeCmd { /// The candidate strategy; must be r-sma (the candidate must produce R). #[arg(long)] strategy: Option, /// Comma-separated instrument list (>=2 distinct, required). #[arg(long)] real: Option, /// Candidate fast-MA length (single value; required). #[arg(long)] fast: Option, /// Candidate slow-MA length (single value; required). #[arg(long)] slow: Option, /// Candidate stop length (single value; required). #[arg(long)] stop_length: Option, /// Candidate stop-k multiple (single value; required). #[arg(long)] stop_k: Option, /// Window start (Unix ms, inclusive). #[arg(long)] from: Option, /// Window end (Unix ms, inclusive). #[arg(long)] to: Option, /// Grading metric (default expectancy_r). #[arg(long)] metric: Option, /// Family name (default generalize). #[arg(long)] name: Option, } #[derive(Args)] struct RunsCmd { #[command(subcommand)] sub: RunsSub, } #[derive(Subcommand)] enum RunsSub { /// List recorded families. Families, /// Inspect one family, optionally ranked by a metric. Family { id: String, /// The literal keyword `rank`, if a ranking metric follows. rank_kw: Option, /// The metric to rank by (only valid after `rank`). metric: Option, }, } #[derive(Args)] struct ReproduceCmd { /// The family content id to reproduce. id: String, } #[derive(Args)] struct RunCmd { /// A serialized signal blueprint (.json). An existing file selects the /// loaded-blueprint grammar; otherwise the built-in harness grammar. blueprint: Option, /// Built-in harness: sma | macd | r-sma (default sma). #[arg(long)] harness: Option, /// Blueprint params (JSON scalar-cell array; .json mode). #[arg(long)] params: Option, /// Blueprint seed (.json mode). #[arg(long)] seed: Option, /// Real instrument symbol to backtest over (recorded data); omit for the synthetic stream. #[arg(long)] real: Option, /// Window start (Unix ms, inclusive); requires --real. #[arg(long)] from: Option, /// Window end (Unix ms, inclusive); requires --real. #[arg(long)] to: Option, /// Persist the run's taps under `runs/traces/` and still print the report. #[arg(long)] trace: Option, // `allow_hyphen_values`: the cost rates accept a negative value (e.g. `-0.5`) as // the flag value so the non-negativity guard (`run_args_from`) can surface the // named `must be non-negative` refusal, rather than clap rejecting `-0.5` as an // unknown option before the value is parsed. #[arg(long, allow_hyphen_values = true, long_help = "per-trade cost in price units. r-sma only; charged in R as cost/|entry-stop|.")] cost_per_trade: Option, #[arg(long, allow_hyphen_values = true, long_help = "slippage multiplier on realized vol. r-sma only.")] slip_vol_mult: Option, #[arg(long, allow_hyphen_values = true, long_help = "carry in price units per ENGINE cycle (not per day, not an overnight swap). r-sma only.")] carry_per_cycle: Option, } #[derive(Args)] struct SweepCmd { /// A loaded blueprint (.json); omit for the built-in --strategy grammar. blueprint: Option, /// Built-in strategy: sma | momentum | r-sma | r-breakout | r-meanrev (default sma). #[arg(long)] strategy: Option, /// Real instrument symbol to sweep over (recorded data); omit for the synthetic stream. #[arg(long)] real: Option, /// Window start (Unix ms, inclusive); requires --real. #[arg(long)] from: Option, /// Window end (Unix ms, inclusive); requires --real. #[arg(long)] to: Option, /// Family name (records to the registry without persisting per-member traces). #[arg(long)] name: Option, /// Family name that also persists each member's taps (mutually exclusive with --name). #[arg(long)] trace: Option, /// Fast-MA grid axis (comma-separated values). #[arg(long)] fast: Option, /// Slow-MA grid axis (comma-separated values). #[arg(long)] slow: Option, /// Stop-length grid axis (comma-separated values). #[arg(long)] stop_length: Option, /// Stop-k grid axis (comma-separated values). #[arg(long)] stop_k: Option, /// Breakout-channel grid axis (comma-separated values; r-breakout). #[arg(long)] channel: Option, /// Mean-reversion window grid axis (comma-separated values; r-meanrev). #[arg(long)] window: Option, /// Mean-reversion band-k grid axis (comma-separated values; r-meanrev). #[arg(long)] band_k: Option, /// Blueprint sweep axis `=` (repeatable; .json mode). #[arg(long)] axis: Vec, /// List a loaded blueprint's sweepable axes and exit (.json mode, stands alone). #[arg(long)] list_axes: bool, } #[derive(Args)] struct WalkforwardCmd { /// A loaded blueprint (.json); omit for the built-in --strategy grammar. blueprint: Option, /// Built-in strategy: sma | momentum | r-sma | r-breakout | r-meanrev (default sma). #[arg(long)] strategy: Option, /// Real instrument symbol to validate over (recorded data); omit for the synthetic stream. #[arg(long)] real: Option, /// Window start (Unix ms, inclusive); requires --real. #[arg(long)] from: Option, /// Window end (Unix ms, inclusive); requires --real. #[arg(long)] to: Option, /// Family name (records to the registry without persisting per-member traces). #[arg(long)] name: Option, /// Family name that also persists each OOS window's taps (mutually exclusive with --name). #[arg(long)] trace: Option, /// Fast-MA IS-refit grid axis (comma-separated values). #[arg(long)] fast: Option, /// Slow-MA IS-refit grid axis (comma-separated values). #[arg(long)] slow: Option, /// Stop-length IS-refit grid axis (comma-separated values). #[arg(long)] stop_length: Option, /// Stop-k IS-refit grid axis (comma-separated values). #[arg(long)] stop_k: Option, /// In-sample winner selection: argmax | plateau:mean | plateau:worst (default argmax). #[arg(long)] select: Option, /// Blueprint IS-refit axis `=` (repeatable, >=1 required; .json mode). #[arg(long)] axis: Vec, } #[derive(Args)] struct McCmd { /// A loaded blueprint (.json); omit for the built-in grammar. blueprint: Option, /// Built-in strategy: r-sma selects the R-bootstrap path (else the synthetic seed-resweep). #[arg(long)] strategy: Option, /// Real instrument symbol for the R-bootstrap (recorded data); requires --strategy r-sma. #[arg(long)] real: Option, /// Window start (Unix ms, inclusive); requires --real. #[arg(long)] from: Option, /// Window end (Unix ms, inclusive); requires --real. #[arg(long)] to: Option, /// Family name for the synthetic seed-resweep (records without persisting traces). #[arg(long)] name: Option, /// Family name that also persists each seed's taps (mutually exclusive with --name). #[arg(long)] trace: Option, /// Fast-MA grid axis for the R-bootstrap (comma-separated values). #[arg(long)] fast: Option, /// Slow-MA grid axis for the R-bootstrap (comma-separated values). #[arg(long)] slow: Option, /// Stop-length grid axis for the R-bootstrap (comma-separated values). #[arg(long)] stop_length: Option, /// Stop-k grid axis for the R-bootstrap (comma-separated values). #[arg(long)] stop_k: Option, /// Moving-block bootstrap block length (R-bootstrap; default 1). #[arg(long)] block_len: Option, /// Number of bootstrap resamples (R-bootstrap; default 1000). #[arg(long)] resamples: Option, /// Bootstrap RNG seed (R-bootstrap; default 1). #[arg(long)] seed: Option, /// Number of synthetic draws (required; .json mode). #[arg(long)] seeds: Option, } /// The dual-grammar discriminator: a first-positional that names an existing /// `.json` file selects the loaded-blueprint branch. Single-sourced so the /// four dual-grammar subcommands stay in lockstep. fn is_blueprint_file(arg: &Option) -> Option<&str> { arg.as_deref() .filter(|a| a.ends_with(".json") && std::path::Path::new(a).is_file()) } /// Resolve a `[blueprint].json`-branch `--real`/`--from`/`--to` into a `RunData`, /// mirroring the old `parse_blueprint_run_args` window guard (`--from`/`--to` /// require `--real`; empty symbol rejected). Refuses in place (stderr + exit 2). fn run_data_from(real: Option<&str>, from: Option, to: Option) -> RunData { let usage = "Usage: aura run [--params ] [--seed ] [--real [--from ] [--to ]]"; match real { Some(s) if !s.is_empty() => RunData::Real { symbol: s.to_string(), from, to }, Some(_) => { eprintln!("aura: {usage}"); std::process::exit(2); } None if from.is_some() || to.is_some() => { eprintln!("aura: {usage}"); std::process::exit(2); } None => RunData::Synthetic, } } /// Build the existing `RunArgs` (the type `run_dispatch` consumes) from the /// built-in-branch `RunCmd` fields — the old `parse_run_args` body minus the argv /// tokenizing clap now owns. A stray positional (a non-`.json`-file `[blueprint]`) /// is an unexpected token; the harness-enum map, the cost-flags-require-R-harness /// guard, and the non-negative-rate checks reuse the existing message strings. fn run_args_from(a: &RunCmd) -> Result { let usage = || "Usage: aura run [--harness ] [--real [--from ] [--to ]] [--trace ] [--cost-per-trade ] [--slip-vol-mult ] [--carry-per-cycle ]".to_string(); // A positional that is not an existing `.json` blueprint is an unexpected token // (the built-in run grammar takes only flags) — the #16 strict reading. if a.blueprint.is_some() { return Err(usage()); } let harness = match a.harness.as_deref() { None | Some("sma") => HarnessKind::Sma, Some("macd") => HarnessKind::Macd, Some("r-sma") => HarnessKind::RSma, Some(_) => return Err(usage()), }; // A parsed-but-negative rate is a named refusal (a sign typo distinguished from a // mistyped flag), matching the old `parse_nonneg_rate` message. Checked before the // R-harness guard so the precedence matches the old per-flag parse order. let nonneg = |flag: &str, v: Option| -> Result, String> { match v { Some(x) if x < 0.0 => Err(format!("{flag} must be non-negative, got {x}")), other => Ok(other), } }; let cost = nonneg("--cost-per-trade", a.cost_per_trade)?; let slip_vol_mult = nonneg("--slip-vol-mult", a.slip_vol_mult)?; let carry_per_cycle = nonneg("--carry-per-cycle", a.carry_per_cycle)?; if !matches!(harness, HarnessKind::RSma) && (cost.is_some() || slip_vol_mult.is_some() || carry_per_cycle.is_some()) { return Err( "cost flags require an R-evaluator harness (r-sma); \ --harness sma/macd produces no R to charge against" .to_string(), ); } if a.real.is_none() && (a.from.is_some() || a.to.is_some()) { return Err(usage()); } let data = match a.real.as_deref() { Some(s) if !s.is_empty() => RunData::Real { symbol: s.to_string(), from: a.from, to: a.to }, Some(_) => return Err(usage()), None => RunData::Synthetic, }; Ok(RunArgs { harness, data, trace: a.trace.clone(), cost, slip_vol_mult, carry_per_cycle }) } /// The old `parse_generalize_args` body minus tokenizing: convert `GeneralizeCmd` /// into the `run_generalize` argument shape. The candidate is a single cell (clap /// already types `--fast`/etc. as one `i64`/`f64`), all four knobs required, `--real` /// a `>=2`-distinct comma list; every refusal reuses the old message string. #[allow(clippy::type_complexity)] fn generalize_args_from( a: &GeneralizeCmd, ) -> Result<(String, Vec, RGrid, String, Option, Option), String> { if let Some(s) = a.strategy.as_deref() && s != "r-sma" { return Err("generalize requires --strategy r-sma (the candidate must produce R)".to_string()); } let symbols: Vec = match a.real.as_deref() { None => return Err("generalize requires --real — a comma list of two or more instruments".to_string()), Some(v) => { let parts: Vec = v.split(',').map(|s| s.to_string()).collect(); if parts.iter().any(|s| s.is_empty()) { return Err("generalize: --real takes a comma list of non-empty symbols (e.g. GER40,USDJPY)".to_string()); } parts } }; if symbols.len() < 2 { return Err(format!( "generalize needs at least two instruments to compare across; got {} (--real takes a comma list of >=2 symbols)", symbols.len() )); } let mut seen = HashSet::new(); if !symbols.iter().all(|s| seen.insert(s.clone())) { return Err("generalize: each instrument may appear once; --real has a duplicate symbol".to_string()); } let knobs = || "generalize requires all four candidate knobs: --fast --slow --stop-length --stop-k, each a single value".to_string(); let grid = RGrid { fast: vec![a.fast.ok_or_else(knobs)?], slow: vec![a.slow.ok_or_else(knobs)?], stop_length: vec![a.stop_length.ok_or_else(knobs)?], stop_k: vec![a.stop_k.ok_or_else(knobs)?], ..RGrid::default() }; let metric = a.metric.clone().unwrap_or_else(|| "expectancy_r".to_string()); let name = a.name.clone().unwrap_or_else(|| "generalize".to_string()); Ok((name, symbols, grid, metric, a.from, a.to)) } /// The shared `--real`/`--from`/`--to` resolution for the family subcommands: a /// non-empty symbol yields `DataChoice::Real`, a window flag without `--real` is a /// usage error, absence is synthetic. Single-sourced (the old `RealWindowGrammar` /// finish logic) so the family subcommands agree. fn data_choice_from( real: Option<&str>, from: Option, to: Option, usage: &impl Fn() -> String, ) -> Result { match real { Some("") => Err(usage()), Some(s) => Ok(DataChoice::Real { symbol: s.to_string(), from_ms: from, to_ms: to }), None if from.is_some() || to.is_some() => Err(usage()), None => Ok(DataChoice::Synthetic), } } /// Map a built-in `--strategy` token to a `Strategy` (default sma), reusing the old /// five-way parse arms; an unknown token is a usage error. fn strategy_from(s: Option<&str>, usage: &impl Fn() -> String) -> Result { Ok(match s { None | Some("sma") => Strategy::SmaCross, Some("momentum") => Strategy::Momentum, Some("r-sma") => Strategy::RSma, Some("r-breakout") => Strategy::RBreakout, Some("r-meanrev") => Strategy::RMeanRev, Some(_) => return Err(usage()), }) } /// Resolve `--name`/`--trace` (mutually exclusive) into `(family_name, persist)`, /// defaulting the name when neither is given. fn name_persist( name: Option<&str>, trace: Option<&str>, default: &str, usage: &impl Fn() -> String, ) -> Result<(String, bool), String> { match (name, trace) { (Some(_), Some(_)) => Err(usage()), (Some(n), None) => Ok((n.to_string(), false)), (None, Some(t)) => Ok((t.to_string(), true)), (None, None) => Ok((default.to_string(), false)), } } /// Parse the repeatable `--axis =` list into by-name grid axes, mirroring /// the old blueprint-sweep axis grammar: an empty/duplicate name or a malformed csv /// is a usage error. fn parse_axes( raw: &[String], usage: &impl Fn() -> String, ) -> Result)>, String> { let mut axes: Vec<(String, Vec)> = Vec::new(); for item in raw { let (n, csv) = item.split_once('=').ok_or_else(usage)?; if n.is_empty() || axes.iter().any(|(a, _)| a == n) { return Err(usage()); } let vals = parse_scalar_csv(csv).ok_or_else(usage)?; axes.push((n.to_string(), vals)); } Ok(axes) } /// `aura run`: the loaded-blueprint branch (an existing `.json` first-positional) or /// the built-in harness-kind dispatch. fn dispatch_run(a: RunCmd, env: &project::Env) { match is_blueprint_file(&a.blueprint) { Some(path) => { if a.harness.is_some() { eprintln!("aura: --harness is not valid with a blueprint file"); std::process::exit(2); } // The loaded-blueprint grammar takes only --params/--seed/--real/--from/--to; // the built-in-only flags are rejected here (exit 2), never silently dropped — // mirroring the sweep/mc blueprint branches, which reject their non-branch flags // exhaustively (refuse-don't-guess). clap's optional `[blueprint]` positional // makes these structurally parseable, so the guard is re-asserted at dispatch. if a.trace.is_some() || a.cost_per_trade.is_some() || a.slip_vol_mult.is_some() || a.carry_per_cycle.is_some() { eprintln!("aura: Usage: aura run [--params ] [--seed ] [--real [--from ] [--to ]]"); std::process::exit(2); } let doc = std::fs::read_to_string(path).unwrap_or_else(|e| { eprintln!("aura: {path}: {e}"); std::process::exit(2); }); let signal = blueprint_from_json(&doc, &|t| env.resolve(t)).unwrap_or_else(|e| { eprintln!("aura: {path}: {e:?}"); std::process::exit(2); }); // Refuse an open (free-knob) blueprint at the dispatch boundary, mirroring // `blueprint_mc_family`'s closed-guard: `run` bootstraps over the EMPTY point, so a // free knob would panic in `compile_with_params` — reject it clean instead (#176). let free = blueprint_axis_probe(&doc, env).param_space(); if !free.is_empty() { eprintln!( "aura: run requires a closed blueprint (no free parameters); {} free knob(s) — \ bind them or use `aura sweep --axis`", free.len() ); std::process::exit(2); } let params = match a.params.as_deref() { Some(j) => parse_param_cells(j).unwrap_or_else(|m| { eprintln!("aura: {m}"); std::process::exit(2); }), None => Vec::new(), }; let data = run_data_from(a.real.as_deref(), a.from, a.to); let report = run_signal_r(signal, ¶ms, data, a.seed.unwrap_or(0), env); println!("{}", report.to_json()); } None => { if a.params.is_some() || a.seed.is_some() { eprintln!("aura: --params/--seed require a blueprint file"); std::process::exit(2); } let run_args = run_args_from(&a).unwrap_or_else(|m| { eprintln!("aura: {m}"); std::process::exit(2); }); match run_dispatch(run_args, env) { Ok(report) => println!("{}", report.to_json()), Err(msg) => { eprintln!("aura: {msg}"); std::process::exit(2); } } } } } fn dispatch_chart(a: ChartCmd, env: &project::Env) { emit_chart( &a.name, a.tap.as_deref(), if a.panels { ChartMode::Panels } else { ChartMode::Overlay }, env, ); } fn dispatch_graph(a: GraphCmd, env: &project::Env) { match a.sub { None => print!("{}", render::render_html(&sample_blueprint())), Some(GraphSub::Build) => graph_construct::build_cmd(env), Some(GraphSub::Introspect(i)) => graph_construct::introspect_cmd(i, env), } } fn dispatch_generalize(a: GeneralizeCmd, env: &project::Env) { let (name, symbols, grid, metric, from_ms, to_ms) = generalize_args_from(&a).unwrap_or_else(|m| { eprintln!("aura: {m}"); std::process::exit(2); }); run_generalize(&name, &symbols, &grid, &metric, from_ms, to_ms, env); } fn dispatch_runs(a: RunsCmd, env: &project::Env) { match a.sub { RunsSub::Families => runs_families(env), RunsSub::Family { id, rank_kw, metric } => match (rank_kw.as_deref(), metric) { (None, None) => runs_family(&id, None, env), (Some("rank"), Some(m)) => runs_family(&id, Some(&m), env), _ => { eprintln!("aura: Usage: aura runs family [rank ]"); std::process::exit(2); } }, } } fn dispatch_reproduce(a: ReproduceCmd, env: &project::Env) { reproduce_family(&a.id, env); } fn dispatch_new(a: NewCmd, _env: &project::Env) { let cwd = std::env::current_dir().unwrap_or_else(|e| { eprintln!("aura: {e}"); std::process::exit(1); }); let spec = scaffold::scaffold_spec( &a.name, a.engine_path.as_deref(), a.namespace.as_deref(), &cwd, ) .unwrap_or_else(|m| { eprintln!("aura: {m}"); std::process::exit(2); }); scaffold::scaffold(&spec).unwrap_or_else(|m| { eprintln!("aura: {m}"); std::process::exit(1); }); println!( "created project `{}` (namespace `{}`)", spec.name, spec.namespace ); } /// `aura sweep`: loaded-blueprint by-name axis sweep (or `--list-axes` probe) when the /// first-positional is an existing `.json`, else the built-in `--strategy` grid sweep. fn dispatch_sweep(a: SweepCmd, env: &project::Env) { match is_blueprint_file(&a.blueprint) { Some(path) => { let usage = || "Usage: aura sweep --axis = [--axis …] [--name | --trace ] [--real [--from ] [--to ]]".to_string(); let doc = std::fs::read_to_string(path).unwrap_or_else(|e| { eprintln!("aura: {path}: {e}"); std::process::exit(2); }); // Parse-validate the blueprint once at the boundary (with file-path context). if let Err(e) = blueprint_from_json(&doc, &|t| env.resolve(t)) { eprintln!("aura: {path}: {e:?}"); std::process::exit(2); } // A built-in-only flag with a blueprint file is not in this grammar. if a.strategy.is_some() || a.fast.is_some() || a.slow.is_some() || a.stop_length.is_some() || a.stop_k.is_some() || a.channel.is_some() || a.window.is_some() || a.band_k.is_some() { eprintln!("aura: {}", usage()); std::process::exit(2); } if a.list_axes { // A query, not a sweep: it must stand alone. if !a.axis.is_empty() || a.name.is_some() || a.trace.is_some() || a.real.is_some() || a.from.is_some() || a.to.is_some() { eprintln!("aura: --list-axes lists axes and takes no other flags"); std::process::exit(2); } list_blueprint_axes(&doc, env); return; } let axes = parse_axes(&a.axis, &usage).unwrap_or_else(|m| { eprintln!("aura: {m}"); std::process::exit(2); }); if axes.is_empty() { eprintln!("aura: {}", usage()); std::process::exit(2); } let (name, persist) = name_persist(a.name.as_deref(), a.trace.as_deref(), "sweep", &usage) .unwrap_or_else(|m| { eprintln!("aura: {m}"); std::process::exit(2); }); let data = data_choice_from(a.real.as_deref(), a.from, a.to, &usage).unwrap_or_else(|m| { eprintln!("aura: {m}"); std::process::exit(2); }); run_blueprint_sweep(&doc, &axes, &name, persist, DataSource::from_choice(data, env), env); } None => { let usage = || "Usage: aura sweep [--strategy ] [--real [--from ] [--to ]] [--name | --trace ] [--fast ] [--slow ] [--stop-length ] [--stop-k ] [--channel ] [--window ] [--band-k ]".to_string(); if a.blueprint.is_some() || !a.axis.is_empty() || a.list_axes { eprintln!("aura: {}", usage()); std::process::exit(2); } let strategy = strategy_from(a.strategy.as_deref(), &usage).unwrap_or_else(|m| { eprintln!("aura: {m}"); std::process::exit(2); }); let mut grid = RGrid::default(); let bad = |m: String| -> ! { eprintln!("aura: {m}"); std::process::exit(2) }; if let Some(v) = a.fast.as_deref() { grid.fast = parse_csv_list(v).unwrap_or_else(|()| bad(usage())); } if let Some(v) = a.slow.as_deref() { grid.slow = parse_csv_list(v).unwrap_or_else(|()| bad(usage())); } if let Some(v) = a.stop_length.as_deref() { grid.stop_length = parse_csv_list(v).unwrap_or_else(|()| bad(usage())); } if let Some(v) = a.stop_k.as_deref() { grid.stop_k = parse_csv_list(v).unwrap_or_else(|()| bad(usage())); } if let Some(v) = a.channel.as_deref() { grid.channel = parse_csv_list(v).unwrap_or_else(|()| bad(usage())); } if let Some(v) = a.window.as_deref() { grid.window = parse_csv_list(v).unwrap_or_else(|()| bad(usage())); } if let Some(v) = a.band_k.as_deref() { grid.band_k = parse_csv_list(v).unwrap_or_else(|()| bad(usage())); } let (name, persist) = name_persist(a.name.as_deref(), a.trace.as_deref(), "sweep", &usage) .unwrap_or_else(|m| { eprintln!("aura: {m}"); std::process::exit(2); }); let data = data_choice_from(a.real.as_deref(), a.from, a.to, &usage).unwrap_or_else(|m| { eprintln!("aura: {m}"); std::process::exit(2); }); run_sweep(strategy, &name, persist, DataSource::from_choice(data, env), &grid, env); } } } /// `aura walkforward`: IS-refit walk-forward over a loaded blueprint (an existing /// `.json` first-positional) or the built-in `--strategy` walk-forward. fn dispatch_walkforward(a: WalkforwardCmd, env: &project::Env) { match is_blueprint_file(&a.blueprint) { Some(path) => { let usage = || "Usage: aura walkforward --axis = [--axis …] [--select ] [--name ]".to_string(); let doc = std::fs::read_to_string(path).unwrap_or_else(|e| { eprintln!("aura: {path}: {e}"); std::process::exit(2); }); if let Err(e) = blueprint_from_json(&doc, &|t| env.resolve(t)) { eprintln!("aura: {path}: {e:?}"); std::process::exit(2); } if a.strategy.is_some() || a.fast.is_some() || a.slow.is_some() || a.stop_length.is_some() || a.stop_k.is_some() || a.trace.is_some() { eprintln!("aura: {}", usage()); std::process::exit(2); } let axes = parse_axes(&a.axis, &usage).unwrap_or_else(|m| { eprintln!("aura: {m}"); std::process::exit(2); }); if axes.is_empty() { eprintln!("aura: walkforward requires >= 1 --axis to re-fit per window"); std::process::exit(2); } let select = match a.select.as_deref() { Some(s) => parse_select(s).unwrap_or_else(|()| { eprintln!("aura: {}", usage()); std::process::exit(2); }), None => Selection::Argmax, }; let name = a.name.clone().unwrap_or_else(|| "walkforward".to_string()); run_blueprint_walkforward(&doc, &axes, &name, DataSource::Synthetic, select, env); } None => { let usage = || "Usage: aura walkforward [--strategy ] [--real [--from ] [--to ]] [--name | --trace ] [--fast ] [--slow ] [--stop-length ] [--stop-k ] [--select ]".to_string(); if a.blueprint.is_some() || !a.axis.is_empty() { eprintln!("aura: {}", usage()); std::process::exit(2); } let strategy = strategy_from(a.strategy.as_deref(), &usage).unwrap_or_else(|m| { eprintln!("aura: {m}"); std::process::exit(2); }); let mut grid = RGrid::default(); let bad = |m: String| -> ! { eprintln!("aura: {m}"); std::process::exit(2) }; if let Some(v) = a.fast.as_deref() { grid.fast = parse_csv_list(v).unwrap_or_else(|()| bad(usage())); } if let Some(v) = a.slow.as_deref() { grid.slow = parse_csv_list(v).unwrap_or_else(|()| bad(usage())); } if let Some(v) = a.stop_length.as_deref() { grid.stop_length = parse_csv_list(v).unwrap_or_else(|()| bad(usage())); } if let Some(v) = a.stop_k.as_deref() { grid.stop_k = parse_csv_list(v).unwrap_or_else(|()| bad(usage())); } let select = match a.select.as_deref() { Some(s) => parse_select(s).unwrap_or_else(|()| bad(usage())), None => Selection::Argmax, }; let (name, persist) = name_persist(a.name.as_deref(), a.trace.as_deref(), "walkforward", &usage) .unwrap_or_else(|m| { eprintln!("aura: {m}"); std::process::exit(2); }); let data = data_choice_from(a.real.as_deref(), a.from, a.to, &usage).unwrap_or_else(|m| { eprintln!("aura: {m}"); std::process::exit(2); }); run_walkforward(strategy, &name, persist, DataSource::from_choice(data, env), &grid, select, env); } } } /// `aura mc`: loaded-blueprint Monte-Carlo (an existing `.json` first-positional), or /// the built-in synthetic seed-resweep / r-sma R-bootstrap split. fn dispatch_mc(a: McCmd, env: &project::Env) { match is_blueprint_file(&a.blueprint) { Some(path) => { let usage = "Usage: aura mc --seeds [--name ]"; let doc = std::fs::read_to_string(path).unwrap_or_else(|e| { eprintln!("aura: {path}: {e}"); std::process::exit(2); }); if let Err(e) = blueprint_from_json(&doc, &|t| env.resolve(t)) { eprintln!("aura: {path}: {e:?}"); std::process::exit(2); } // A built-in-only flag with a blueprint file is not in this grammar // (MC over a loaded blueprint is synthetic-only this cycle). if a.strategy.is_some() || a.real.is_some() || a.from.is_some() || a.to.is_some() || a.fast.is_some() || a.slow.is_some() || a.stop_length.is_some() || a.stop_k.is_some() || a.block_len.is_some() || a.resamples.is_some() || a.seed.is_some() || a.trace.is_some() { eprintln!("aura: {usage}"); std::process::exit(2); } let n_seeds = match a.seeds { Some(n) if n > 0 => n, _ => { eprintln!("aura: {usage}"); std::process::exit(2); } }; let name = a.name.clone().unwrap_or_else(|| "mc".to_string()); run_blueprint_mc(&doc, n_seeds, &name, DataSource::Synthetic, env); } None => { let usage = || "Usage: aura mc [--name |--trace ] | aura mc --strategy r-sma [--real [--from ] [--to ]] [--fast ] [--slow ] [--stop-length ] [--stop-k ] [--block-len ] [--resamples ] [--seed ]".to_string(); if a.blueprint.is_some() || a.seeds.is_some() { eprintln!("aura: {}", usage()); std::process::exit(2); } // The R-bootstrap path is selected by any r-sma knob; otherwise the // synthetic seed-resweep family. Name flags are invalid on the R path. let r_path = a.strategy.is_some() || a.real.is_some() || a.from.is_some() || a.to.is_some() || a.fast.is_some() || a.slow.is_some() || a.stop_length.is_some() || a.stop_k.is_some() || a.block_len.is_some() || a.resamples.is_some() || a.seed.is_some(); let mc_args = if r_path { if a.strategy.as_deref() != Some("r-sma") || a.name.is_some() || a.trace.is_some() { eprintln!("aura: {}", usage()); std::process::exit(2); } let mut grid = RGrid::default(); let bad = |m: String| -> ! { eprintln!("aura: {m}"); std::process::exit(2) }; if let Some(v) = a.fast.as_deref() { grid.fast = parse_csv_list(v).unwrap_or_else(|()| bad(usage())); } if let Some(v) = a.slow.as_deref() { grid.slow = parse_csv_list(v).unwrap_or_else(|()| bad(usage())); } if let Some(v) = a.stop_length.as_deref() { grid.stop_length = parse_csv_list(v).unwrap_or_else(|()| bad(usage())); } if let Some(v) = a.stop_k.as_deref() { grid.stop_k = parse_csv_list(v).unwrap_or_else(|()| bad(usage())); } let choice = data_choice_from(a.real.as_deref(), a.from, a.to, &usage).unwrap_or_else(|m| bad(m)); McArgs::RealR { choice, grid, block_len: a.block_len.unwrap_or(1), n_resamples: a.resamples.unwrap_or(1000), seed: a.seed.unwrap_or(1), } } else { let (name, persist) = name_persist(a.name.as_deref(), a.trace.as_deref(), "mc", &usage) .unwrap_or_else(|m| { eprintln!("aura: {m}"); std::process::exit(2); }); McArgs::Synthetic { name, persist } }; match mc_args { McArgs::Synthetic { name, persist } => run_mc(&name, persist, env), McArgs::RealR { choice, grid, block_len, n_resamples, seed } => run_mc_r_bootstrap( DataSource::from_choice(choice, env), &grid, block_len, n_resamples, seed, env, ), } } } } fn main() { // Restore the default SIGPIPE disposition. Rust's runtime sets SIGPIPE to SIG_IGN // at startup, so a write to a closed stdout pipe (`aura sweep | head`, a closed UI // pane) returns EPIPE and panics in `println!` instead of terminating quietly on // SIGPIPE — the conventional Unix CLI behaviour. One reset covers every // family-emitting subcommand at once. #[cfg(unix)] unsafe { libc::signal(libc::SIGPIPE, libc::SIG_DFL); } let cli = Cli::parse(); // `aura new` scaffolds; it must not require a loadable project even // when invoked inside one (e.g. an unbuilt tree). let env = if matches!(cli.command, Command::New(_)) { project::Env::std() } else { match std::env::current_dir() .ok() .and_then(|d| project::discover_from(&d)) { Some(root) => match project::load(&root, cli.release) { Ok(p) => project::Env::with_project(p), Err(e) => { eprintln!("aura: {e}"); std::process::exit(1); } }, None => project::Env::std(), } }; match cli.command { Command::Run(a) => dispatch_run(a, &env), Command::Chart(a) => dispatch_chart(a, &env), Command::Graph(a) => dispatch_graph(a, &env), Command::Sweep(a) => dispatch_sweep(a, &env), Command::Walkforward(a) => dispatch_walkforward(a, &env), Command::Generalize(a) => dispatch_generalize(a, &env), Command::Mc(a) => dispatch_mc(a, &env), Command::Runs(a) => dispatch_runs(a, &env), Command::Reproduce(a) => dispatch_reproduce(a, &env), Command::New(a) => dispatch_new(a, &env), Command::Process(a) => research_docs::process_cmd(a, &env), Command::Campaign(a) => research_docs::campaign_cmd(a, &env), } } #[cfg(test)] mod tests { use super::*; #[test] fn select_winner_refuses_plateau_without_a_lattice() { // A plateau request with no lattice (a random sweep would yield None) is // refused, never silently argmaxed. The refuse short-circuits before the // family is read, so an empty family is fine here. let fam = SweepFamily { space: vec![], points: vec![] }; let err = select_winner(&fam, "total_pips", Selection::Plateau(PlateauMode::Mean), None) .unwrap_err(); assert!(err.contains("requires a grid sweep"), "refuse message: {err}"); } /// The `--select` token grammar (`parse_select`) maps argmax / plateau:mean / /// plateau:worst and rejects an unknown token — the pure selector clap's /// `--select` value feeds on both the built-in and blueprint walk-forward paths. #[test] fn parse_select_token_grammar() { assert!(matches!(parse_select("argmax").unwrap(), Selection::Argmax), "default is argmax"); assert!(matches!(parse_select("plateau:mean").unwrap(), Selection::Plateau(PlateauMode::Mean))); assert!(matches!(parse_select("plateau:worst").unwrap(), Selection::Plateau(PlateauMode::Worst))); assert!(parse_select("bogus").is_err(), "unknown --select token is a usage error"); } fn cmp_member(key: &str, ts: &[i64], vals: &[f64]) -> FamilyMember { cmp_member_win(key, ts, vals, (0, 0)) } /// Like [`cmp_member`] but with an explicit manifest `window` so a test can /// model walk-forward members (disjoint per-member OOS windows) and assert the /// family window spans them. fn cmp_member_win(key: &str, ts: &[i64], vals: &[f64], window: (i64, i64)) -> FamilyMember { let rows: Vec<(Timestamp, Vec)> = ts.iter().zip(vals).map(|(&t, &v)| (Timestamp(t), vec![Scalar::f64(v)])).collect(); let tap = ColumnarTrace::from_rows("equity", &[ScalarKind::F64], &rows); FamilyMember { key: key.to_string(), traces: RunTraces { manifest: sim_optimal_manifest( vec![], (Timestamp(window.0), Timestamp(window.1)), 0, 1.0, ), taps: vec![tap], }, } } #[test] fn comparison_overlays_one_shared_scale_series_per_member() { let members = vec![ cmp_member("a", &[1, 2, 3], &[10.0, 11.0, 12.0]), cmp_member("b", &[1, 2, 3], &[20.0, 21.0, 22.0]), ]; let data = build_comparison_chart_data("fam", &members, "equity").expect("builds"); assert_eq!(data.xs, vec![1, 2, 3]); assert_eq!(data.series.len(), 2); assert_eq!(data.series[0].name, "a"); assert_eq!(data.series[1].name, "b"); // ONE shared y-scale across members (same quantity). assert_eq!(data.series[0].y_scale_id, data.series[1].y_scale_id); // shared ts -> dense, no nulls. assert!(data.series[0].points.iter().all(Option::is_some)); // #102 meta wiring: a family carries kind/name/member-count + the one // compared tap, and never the per-member params (those are the labels). assert_eq!(data.meta.kind, "family"); assert_eq!(data.meta.name, "fam"); assert_eq!(data.meta.members, Some(2)); assert_eq!(data.meta.taps, vec!["equity".to_string()]); assert!(data.meta.params.is_empty(), "family meta must not repeat per-member params"); } #[test] fn comparison_disjoint_members_are_null_complementary() { let members = vec![ cmp_member("oos1", &[1, 2], &[10.0, 11.0]), cmp_member("oos2", &[3, 4], &[20.0, 21.0]), ]; let data = build_comparison_chart_data("fam", &members, "equity").expect("builds"); assert_eq!(data.xs, vec![1, 2, 3, 4]); assert_eq!(data.series[0].points, vec![Some(10.0), Some(11.0), None, None]); assert_eq!(data.series[1].points, vec![None, None, Some(20.0), Some(21.0)]); } /// #102 family-window semantics: the header's `window` for a family is the /// SPAN across all members — `(min member.from, max member.to)` — not the first /// member's window. The distinction is load-bearing for a walk-forward family, /// whose members are DISJOINT OOS windows (commit 4c64feb): labelling such a /// family with `members[0]`'s window mislabels the family's true coverage. The /// span reading is correct for all three kinds (sweep/MC members share a window, /// so their span collapses to that shared window). #[test] fn comparison_window_spans_disjoint_walk_forward_members() { let members = vec![ cmp_member_win("oos1", &[10, 20], &[1.0, 2.0], (10, 20)), cmp_member_win("oos2", &[30, 40], &[3.0, 4.0], (30, 40)), cmp_member_win("oos3", &[50, 60], &[5.0, 6.0], (50, 60)), ]; let data = build_comparison_chart_data("wf", &members, "equity").expect("builds"); // SPAN of all OOS windows (10..60), NOT members[0]'s window (10..20). assert_eq!(data.meta.window, (10, 60)); } #[test] fn comparison_errors_when_no_member_has_the_tap() { let members = vec![cmp_member("a", &[1], &[1.0])]; assert!(build_comparison_chart_data("fam", &members, "nosuch").is_err()); } #[test] fn decimate_bounds_the_spine_to_twice_the_bucket_count() { let n = 10_000usize; let xs: Vec = (0..n as i64).collect(); let points: Vec> = (0..n).map(|i| Some(i as f64)).collect(); let data = ChartData { xs, series: vec![Series { name: "equity".into(), y_scale_id: "y_0".into(), points, reduce: ReduceKind::MinMax }], meta: ChartMeta::default(), }; let out = decimate(data, 2000); assert!(out.xs.len() <= 4000, "spine not bounded: {}", out.xs.len()); assert_eq!(out.xs.len(), out.series[0].points.len(), "xs and points must stay aligned"); } #[test] fn decimate_preserves_per_bucket_min_and_max() { // 10 points, 2 buckets -> bucket 0 = idx 0..5 (a spike), bucket 1 = idx 5..10 (a trough). let xs: Vec = (0..10).collect(); let mut pv = vec![1.0_f64; 10]; pv[3] = 999.0; pv[7] = -50.0; let points: Vec> = pv.into_iter().map(Some).collect(); let data = ChartData { xs, series: vec![Series { name: "equity".into(), y_scale_id: "y_0".into(), points, reduce: ReduceKind::MinMax }], meta: ChartMeta::default(), }; let out = decimate(data, 2); let got = out.series[0].points.clone(); assert!(got.contains(&Some(999.0)), "bucket max (spike) dropped: {got:?}"); assert!(got.contains(&Some(-50.0)), "bucket min (trough) dropped: {got:?}"); } #[test] fn decimate_keeps_an_all_null_bucket_null() { let xs: Vec = (0..10).collect(); let mut points: Vec> = (0..5).map(|i| Some(i as f64)).collect(); points.extend(std::iter::repeat_n(None, 5)); let data = ChartData { xs, series: vec![Series { name: "equity".into(), y_scale_id: "y_0".into(), points, reduce: ReduceKind::MinMax }], meta: ChartMeta::default(), }; let out = decimate(data, 2); assert_eq!(*out.series[0].points.last().unwrap(), None, "all-null bucket must stay null"); } #[test] fn decimate_is_a_noop_within_budget() { let data = ChartData { xs: vec![1, 2, 3], series: vec![Series { name: "equity".into(), y_scale_id: "y_0".into(), points: vec![Some(1.0), Some(2.0), Some(3.0)], reduce: ReduceKind::MinMax }], meta: ChartMeta::default(), }; let out = decimate(data, 2000); assert_eq!(out.xs, vec![1, 2, 3], "within-budget data must pass through unchanged"); assert_eq!(out.series[0].points, vec![Some(1.0), Some(2.0), Some(3.0)]); } #[test] fn decimate_passes_meta_through_and_keeps_xs_monotonic() { let n = 10_000usize; let xs: Vec = (0..n as i64).collect(); let points: Vec> = (0..n).map(|i| Some(i as f64)).collect(); let meta = ChartMeta { name: "keep-me".into(), ..Default::default() }; let data = ChartData { xs, series: vec![Series { name: "equity".into(), y_scale_id: "y_0".into(), points, reduce: ReduceKind::MinMax }], meta }; let out = decimate(data, 2000); assert_eq!(out.meta.name, "keep-me", "meta must pass through decimation"); assert!(out.xs.windows(2).all(|w| w[0] < w[1]), "decimated spine must stay strictly increasing"); } /// #111: a bounded *level* series with `reduce = Mean` decimates to each bucket's /// MEAN, not its min/max envelope — so a high-flip bipolar exposure shows its /// net/duty-cycle level instead of collapsing to a -1..+1 band. RED under the /// shipped min/max-only decimation (any bucket holding a +1 emits +1); GREEN once /// `decimate` honours `ReduceKind::Mean`. #[test] fn decimate_mean_reduces_a_bipolar_series_to_its_bucket_level() { // 10 points, 2 buckets. Bucket 0 (idx 0..5) = [+1,+1,-1,+1,+1] -> mean +0.6; // bucket 1 (idx 5..10) = all -1 -> mean -1.0. let xs: Vec = (0..10).collect(); let pv = vec![1.0, 1.0, -1.0, 1.0, 1.0, -1.0, -1.0, -1.0, -1.0, -1.0]; let points: Vec> = pv.into_iter().map(Some).collect(); let data = ChartData { xs, series: vec![Series { name: "exposure".into(), y_scale_id: "y_0".into(), points, reduce: ReduceKind::Mean }], meta: ChartMeta::default(), }; let out = decimate(data, 2); let got = out.series[0].points.clone(); // No -1..+1 envelope: bucket 0 is its mean (+0.6), not a min/max pair. assert!(!got.contains(&Some(1.0)), "mean reduce must not emit a +1 envelope point: {got:?}"); assert!(got.contains(&Some(0.6)), "bucket-0 duty-cycle mean (+0.6) missing: {got:?}"); // bucket 0 spans two slots, both = the mean (a flat step, not a -1->+1 ramp). assert_eq!(got[0], Some(0.6), "first slot must be the bucket mean"); assert_eq!(got[1], Some(0.6), "second slot must also be the bucket mean"); } /// #102 single-run meta wiring: `build_chart_data` maps the `RunManifest` into /// `ChartData.meta` — kind "run", the name arg, the manifest window/broker, the /// charted taps, and the bound params stringified (each typed `Scalar` rendered /// via `render_value`, preserving its lexical form: `i64` decimal, `f64` /// shortest round-trip). A single run carries no member count. #[test] fn build_chart_data_threads_run_manifest_into_meta() { let eq_rows: Vec<(Timestamp, Vec)> = [1i64, 2, 3].iter().map(|&t| (Timestamp(t), vec![Scalar::f64(t as f64)])).collect(); let traces = RunTraces { manifest: sim_optimal_manifest( vec![("len".into(), Scalar::i64(10)), ("scale".into(), Scalar::f64(0.5))], (Timestamp(1), Timestamp(3)), 7, 1.0, ), taps: vec![ColumnarTrace::from_rows("equity", &[ScalarKind::F64], &eq_rows)], }; let data = build_chart_data("demo", traces); let meta = &data.meta; assert_eq!(meta.kind, "run"); assert_eq!(meta.name, "demo"); assert_eq!(meta.window, (1, 3)); assert_eq!(meta.broker, "sim-optimal(pip_size=1)"); assert_eq!(meta.seed, 7); assert_eq!(meta.taps, vec!["equity".to_string()]); assert_eq!(meta.members, None); // params stringified via render_value: typed Scalars keep their lexical form. assert_eq!( meta.params, vec![("len".to_string(), "10".to_string()), ("scale".to_string(), "0.5".to_string())] ); } /// #99: a sweep/walk-forward family-member stdout line embeds the `RunReport` in /// its own declaration key order (manifest leads with `commit`), byte-matching the /// stored `families.jsonl` — never `serde_json::Value`'s alphabetical order (which /// would lead the manifest with `broker`). #[test] fn family_member_line_keeps_report_in_store_key_order() { let report = RunReport { manifest: sim_optimal_manifest(vec![], (Timestamp(0), Timestamp(0)), 0, 1.0), metrics: summarize(&[], &[]), }; let line = family_member_line("demo-1", &report); assert!( line.starts_with(r#"{"family_id":"demo-1","report":{"manifest":{"commit":"#), "got: {line}" ); assert!( !line.contains(r#""manifest":{"broker":"#), "manifest re-alphabetized (broker-first), should be commit-first: {line}" ); } /// #99: the Monte-Carlo per-draw line carries the `seed` between `family_id` and /// `report`, and the embedded report stays in store (commit-first) key order. #[test] fn mc_member_line_keeps_report_in_store_key_order_with_seed() { let report = RunReport { manifest: sim_optimal_manifest(vec![], (Timestamp(0), Timestamp(0)), 7, 1.0), metrics: summarize(&[], &[]), }; let line = mc_member_line("mc-1", 7, &report); assert!( line.starts_with(r#"{"family_id":"mc-1","seed":7,"report":{"manifest":{"commit":"#), "got: {line}" ); assert!( !line.contains(r#""manifest":{"broker":"#), "manifest re-alphabetized (broker-first), should be commit-first: {line}" ); } // The vetted GER40 real-data window: the whole of September 2024 (UTC, // inclusive), the same calendar month the gated ingest `ger40_breakout_real` // test drives. Expressed in Unix-ms (`run_sample_real`'s window currency): // `[2024-09-01T00:00:00Z, 2024-10-01T00:00:00Z - 1ms]`. Both gated GER40 // tests bound their runs to this window so the C1-determinism check stays // fast — an unbounded `None, None` run drains the full archive every call. const GER40_SEP2024_FROM_MS: i64 = 1_725_148_800_000; const GER40_SEP2024_TO_MS: i64 = 1_727_740_799_999; #[test] fn data_source_synthetic_pip_and_window_match_the_built_ins() { let env = project::Env::std(); let d = DataSource::Synthetic; assert_eq!(d.pip_size(), SYNTHETIC_PIP_SIZE); assert!(!d.run_sources(&env).is_empty()); assert_eq!(d.wf_window_sizes(), (24, 12, 12)); // full_window equals window_of over the showcase stream (byte-unchanged source) let s: Vec> = vec![Box::new(VecSource::new(showcase_prices()))]; assert_eq!(d.full_window(&env), window_of(&s).unwrap()); } #[test] fn wf_real_roller_sizes_are_90_30_30_days_in_ns() { // Independent expected value: a day reconstructed from its time units // (24 h * 60 min * 60 s * 1e9 ns), not the constant's own `86_400_000_000_000` // literal — so the test fails if either the literal or the day-count is wrong. let day_ns: i64 = 24 * 60 * 60 * 1_000_000_000; assert_eq!(WF_REAL_IS_NS, 90 * day_ns); assert_eq!(WF_REAL_OOS_NS, 30 * day_ns); assert_eq!(WF_REAL_STEP_NS, 30 * day_ns); } #[test] fn strategy_lengths_are_short_for_synthetic_realistic_for_real() { // Synthetic keeps the demo-stream lengths (byte-unchanged: the 18/60-bar // built-in streams cannot warm a long MA). assert_eq!(DataSource::Synthetic.strategy_lengths(), ([2, 3], [4, 5], (2, 4, 3))); // Real uses realistic M1 lengths — no 2-5-bar noise over tens of thousands // of bars. Constructing Real needs a server, but strategy_lengths matches on // the variant only (no data access). let real = DataSource::Real { server: std::sync::Arc::new(data_server::DataServer::new(data_server::DEFAULT_DATA_PATH)), symbol: "EURUSD".into(), from_ms: None, to_ms: None, pip: 0.0001, }; let (tf, ts, macd) = real.strategy_lengths(); assert_eq!((tf, ts, macd), ([50, 100], [200, 400], (12, 26, 9))); // every trend-fast < every trend-slow (a valid SMA cross, both variants). assert!(tf.iter().max().unwrap() < ts.iter().min().unwrap()); let (stf, sts, _) = DataSource::Synthetic.strategy_lengths(); assert!(stf.iter().max().unwrap() < sts.iter().min().unwrap()); } #[test] fn walkforward_report_is_deterministic() { // The built-in WFO render is byte-identical across two // calls (C1). assert_eq!(walkforward_report(), walkforward_report()); } #[test] fn walkforward_report_has_one_oos_line_per_window_plus_summary() { // N per-window OOS RunReport lines + one summary line. let out = walkforward_report(); let lines: Vec<&str> = out.lines().collect(); assert_eq!(lines.len(), 4); // built-in roll = 3 windows + 1 summary assert!(lines[3].contains(r#""walkforward""#), "summary line: {}", lines[3]); for line in &lines[..3] { assert!( line.contains(r#""manifest""#) && line.contains(r#""metrics""#), "expected an OOS RunReport line, got: {line}", ); } } /// The drained sink trace of a seeded run — the recorded rows of the equity /// and exposure sinks. Compared row-for-row so the C1 seed-determinism /// property is tested at the trace level (strictly stronger than the folded /// 3-field metrics). `PartialEq` not `Eq`: `Scalar` carries `f64`. #[derive(Debug, PartialEq)] struct SeededTrace { equity: Vec<(Timestamp, Vec)>, exposure: Vec<(Timestamp, Vec)>, } /// A seeded run of the sample harness: the synthetic stream is generated /// from `seed`, that same seed is recorded into the manifest, and the /// drained sink trace is returned alongside the report. Every byte of both /// is a function of `seed`. fn run_sample_seeded(seed: u64) -> (RunReport, SeededTrace) { let (mut h, rx_eq, rx_ex) = sample_harness(SYNTHETIC_PIP_SIZE); let spec = SyntheticSpec { start: 1.0, len: 64, step: 1 }; let window = (Timestamp(1), Timestamp((spec.len as i64 - 1) * spec.step + 1)); h.run(vec![Box::new(spec.source(seed))]); let eq_rows: Vec<(Timestamp, Vec)> = rx_eq.try_iter().collect(); let ex_rows: Vec<(Timestamp, Vec)> = rx_ex.try_iter().collect(); let metrics = summarize(&f64_field(&eq_rows, 0), &f64_field(&ex_rows, 0)); let report = RunReport { manifest: sim_optimal_manifest( vec![ ("sma_fast".to_string(), Scalar::i64(2)), ("sma_slow".to_string(), Scalar::i64(4)), ("bias_scale".to_string(), Scalar::f64(0.5)), ], window, seed, SYNTHETIC_PIP_SIZE, ), metrics, }; (report, SeededTrace { equity: eq_rows, exposure: ex_rows }) } #[test] fn same_seed_bit_identical_trace() { // Bit-identical sink trace for a fixed seed (acceptance bullet 1, C1). let (_, trace_a) = run_sample_seeded(42); let (_, trace_b) = run_sample_seeded(42); assert_eq!(trace_a, trace_b); } #[test] fn different_seed_different_trace() { // Different seeds perturb the trace (acceptance bullet 2). let (a, _) = run_sample_seeded(1); let (b, _) = run_sample_seeded(2); assert_ne!(a.metrics, b.metrics); } #[test] fn seed_recorded_in_manifest() { // The seed that drove the run is recorded (acceptance bullet 3). let (report, _) = run_sample_seeded(7); assert_eq!(report.manifest.seed, 7); } #[test] fn sample_blueprint_with_sinks_bootstraps_runs_and_drains() { // the factory returns the two Recorder receivers (build_sample drops them), // so a caller can bootstrap one point, run it, and drain both sinks. let (bp, rx_eq, rx_ex) = sample_blueprint_with_sinks(SYNTHETIC_PIP_SIZE); let mut h = bp .with("signals.trend.fast.length", 2) .with("signals.trend.slow.length", 4) .with("signals.momentum.fast.length", 2) .with("signals.momentum.slow.length", 4) .with("signals.momentum.signal.length", 3) .with("signals.blend.weights[0]", 1.0) .with("signals.blend.weights[1]", 1.0) .with("bias.scale", 0.5) .bootstrap() .expect("sample blueprint compiles under a valid point"); h.run(vec![Box::new(VecSource::new(showcase_prices()))]); assert!(!rx_eq.try_iter().collect::>().is_empty(), "equity sink drained empty"); assert!(!rx_ex.try_iter().collect::>().is_empty(), "exposure sink drained empty"); } #[test] fn sweep_report_renders_four_points_in_odometer_order() { let out = sweep_report(); let lines: Vec<&str> = out.lines().collect(); assert_eq!(lines.len(), 4, "one JSON line per grid point; got: {out:?}"); // each line is a full RunReport; the commit is the real git HEAD // (volatile), so pin the per-point manifest params (odometer order, last // axis fastest) + the metric keys, not the commit value. for line in &lines { assert!(line.starts_with(r#"{"manifest":{"commit":""#), "not a RunReport: {line}"); } assert!(lines[0].contains(r#""params":[["signals.trend.fast.length",{"I64":2}],["signals.trend.slow.length",{"I64":4}],["signals.momentum.fast.length",{"I64":2}],["signals.momentum.slow.length",{"I64":4}],["signals.momentum.signal.length",{"I64":3}],["signals.blend.weights[0]",{"F64":1.0}],["signals.blend.weights[1]",{"F64":1.0}],["bias.scale",{"F64":0.5}]]"#), "line0: {}", lines[0]); assert!(lines[1].contains(r#""params":[["signals.trend.fast.length",{"I64":2}],["signals.trend.slow.length",{"I64":5}],["signals.momentum.fast.length",{"I64":2}],["signals.momentum.slow.length",{"I64":4}],["signals.momentum.signal.length",{"I64":3}],["signals.blend.weights[0]",{"F64":1.0}],["signals.blend.weights[1]",{"F64":1.0}],["bias.scale",{"F64":0.5}]]"#), "line1: {}", lines[1]); assert!(lines[2].contains(r#""params":[["signals.trend.fast.length",{"I64":3}],["signals.trend.slow.length",{"I64":4}],["signals.momentum.fast.length",{"I64":2}],["signals.momentum.slow.length",{"I64":4}],["signals.momentum.signal.length",{"I64":3}],["signals.blend.weights[0]",{"F64":1.0}],["signals.blend.weights[1]",{"F64":1.0}],["bias.scale",{"F64":0.5}]]"#), "line2: {}", lines[2]); assert!(lines[3].contains(r#""params":[["signals.trend.fast.length",{"I64":3}],["signals.trend.slow.length",{"I64":5}],["signals.momentum.fast.length",{"I64":2}],["signals.momentum.slow.length",{"I64":4}],["signals.momentum.signal.length",{"I64":3}],["signals.blend.weights[0]",{"F64":1.0}],["signals.blend.weights[1]",{"F64":1.0}],["bias.scale",{"F64":0.5}]]"#), "line3: {}", lines[3]); for line in &lines { assert!(line.contains(r#""total_pips":"#), "missing total_pips: {line}"); assert!(line.contains(r#""max_drawdown":"#), "missing max_drawdown: {line}"); assert!(line.contains(r#""bias_sign_flips":"#), "missing flips: {line}"); assert!(line.ends_with('}'), "line not closed: {line}"); } } #[test] fn sweep_report_is_deterministic() { // C1 at the CLI edge: the same build yields a bit-identical report. assert_eq!(sweep_report(), sweep_report()); } #[test] fn mc_report_is_deterministic_and_one_line_per_seed() { // C1 at the CLI edge: the family computation renders bit-identically. assert_eq!(mc_report(), mc_report()); let out = mc_report(); let lines: Vec<&str> = out.lines().collect(); // three seeds -> three member lines + one aggregate line assert_eq!(lines.len(), 4, "expected 3 members + 1 aggregate: {out}"); for line in &lines[..3] { assert!(line.contains(r#""total_pips":"#), "member line missing metrics: {line}"); } assert!(lines[3].contains(r#""mc_aggregate":"#), "missing aggregate line: {}", lines[3]); } #[test] fn cli_families_persist_and_round_trip_per_kind() { use aura_registry::{ group_families, mc_member_reports, sweep_member_reports, walkforward_member_reports, FamilyKind, Registry, }; let env = project::Env::std(); let dir = std::env::temp_dir().join(format!("aura-cli-fam-{}", std::process::id())); let _ = std::fs::remove_dir_all(&dir); std::fs::create_dir_all(&dir).expect("temp dir"); let reg = Registry::open(dir.join("runs.jsonl")); // the exact persist chain `run_sweep`/`run_mc`/`run_walkforward` use, against // a fresh temp store (the run_* fns themselves bind `env.registry()`): // engine family -> per-kind extractor -> append_family. let sid = reg .append_family( "sweep", FamilyKind::Sweep, &sweep_member_reports(&sweep_family(None, &DataSource::Synthetic, &env)), ) .expect("sweep family"); let mid = reg .append_family("mc", FamilyKind::MonteCarlo, &mc_member_reports(&mc_family(None, &env))) .expect("mc family"); let wid = reg .append_family( "walkforward", FamilyKind::WalkForward, &walkforward_member_reports(&walkforward_family( Strategy::SmaCross, None, &DataSource::Synthetic, &RGrid::default(), Selection::Argmax, &env, )), ) .expect("walkforward family"); assert_eq!((sid.as_str(), mid.as_str(), wid.as_str()), ("sweep-0", "mc-0", "walkforward-0")); let families = group_families(reg.load_family_members().expect("load")); assert_eq!(families.len(), 3); let by_id = |id: &str| families.iter().find(|f| f.id == id).expect("family present"); assert_eq!(by_id("sweep-0").kind, FamilyKind::Sweep); assert_eq!(by_id("mc-0").kind, FamilyKind::MonteCarlo); assert_eq!(by_id("mc-0").members.len(), 3); // 3 seeds assert_eq!(by_id("walkforward-0").kind, FamilyKind::WalkForward); assert_eq!(by_id("walkforward-0").members.len(), 3); // 3 windows let _ = std::fs::remove_dir_all(&dir); } #[test] fn run_macd_compiles_from_nested_composite_and_is_deterministic() { // the MACD strategy authors a nested EMA-of-EMA composite, compiles it to a // flat runnable harness (the call not panicking proves the compile+bootstrap // path), and runs it. C1 determinism: two runs are bit-identical. let env = project::Env::std(); let r1 = run_macd(None, &env); let r2 = run_macd(None, &env); assert_eq!(r1.metrics, r2.metrics); assert_eq!(r1.to_json(), r2.to_json()); // the synthetic stream is carried end-to-end and the trace is well-formed. let (from, to) = r1.manifest.window; assert_eq!((from.0, to.0), (1, 18)); assert!(r1.metrics.total_pips.is_finite(), "macd pips must be finite: {:?}", r1.metrics); assert!(r1.metrics.max_drawdown >= 0.0, "drawdown is non-negative: {:?}", r1.metrics); // after warm-up the EMA-of-EMA histogram crosses zero, so the strategy // reverses exposure at least once — a genuinely non-trivial trace. assert!( r1.metrics.bias_sign_flips >= 1, "macd trace should flip exposure: {:?}", r1.metrics ); } /// E2E acceptance (#41, the worked example): the real MACD strategy /// blueprint's swept param surface qualifies the three otherwise-indistinguishable /// EMA `length` slots by node name to `macd.fast.length` / `macd.slow.length` / /// `macd.signal.length` — the named composite boundary visible end-to-end through /// `param_space()`, with the slot count and order unchanged (C23 — node names are /// non-load-bearing: every interior slot stays sweepable, the `bias.scale` /// knob is unaffected). #[test] fn macd_param_space_surfaces_the_three_named_legs() { let names: Vec = macd_blueprint().param_space().into_iter().map(|p| p.name).collect(); // three named composite-interior slots, in declared (fast, slow, signal) // order, then the strategy-level Bias `scale` (a root-level leaf). assert_eq!( names, vec![ "macd.fast.length".to_string(), "macd.slow.length".to_string(), "macd.signal.length".to_string(), "bias.scale".to_string(), ], "MACD param surface must expose the three named EMA lengths + scale", ); } /// `aura run --real ` dogfoods the #71 streaming Source seam: the same /// built-in sample signal-quality harness, but fed real M1 **close** bars /// streamed lazily through `aura_ingest::M1FieldSource` (a `Box`), /// not synthetic `VecSource` prices. The property: over the verified bounded /// Sept-2024 GER40 window, `run_sample_real` yields a `RunReport` whose /// `total_pips` is finite and is C1-deterministic — two runs of the same /// window are bit-identical JSON. Bounding the window (vs the full unbounded /// archive) keeps the determinism check fast. /// /// Gated like the ingest `streaming_seam` test: skip (early return) when the /// local Pepperstone archive is absent, so the test never fails on a machine /// without the data. Uses `GER40` — a *vetted* symbol (pip 1.0): the /// per-instrument-pip refusal makes `run_sample_real` reject an un-specced /// symbol before any data access (`std::process::exit(1)`), so this CLI-level /// test must drive a symbol in the instrument table. The bounded-window AAPL.US /// streaming property still lives in the ingest `streaming_seam` test, which /// builds its source literally without the spec lookup. #[test] fn run_sample_real_streams_real_close_bars_deterministically() { // GER40 is in the vetted instrument table (index pip 1.0); the un-specced // AAPL.US would now refuse at the spec lookup before any data access. const SYMBOL: &str = "GER40"; // Mirror skip_if_no_data: never fail where the local archive is absent. let server = std::sync::Arc::new(data_server::DataServer::new( data_server::DEFAULT_DATA_PATH, )); if !server.has_symbol(SYMBOL) { eprintln!( "skip: no local data at {} (symbol {SYMBOL} absent)", data_server::DEFAULT_DATA_PATH ); return; } // The headline: a real-data run over the bounded Sept-2024 window yields // a finite, C1-deterministic RunReport. let env = project::Env::std(); let r1 = run_sample_real(SYMBOL, Some(GER40_SEP2024_FROM_MS), Some(GER40_SEP2024_TO_MS), None, &env); let r2 = run_sample_real(SYMBOL, Some(GER40_SEP2024_FROM_MS), Some(GER40_SEP2024_TO_MS), None, &env); assert!( r1.metrics.total_pips.is_finite(), "real-data run must yield finite pips: {:?}", r1.metrics ); // C1 at the CLI edge: the same real window streamed twice is bit-identical. assert_eq!( r1.to_json(), r2.to_json(), "two real-data runs of the same window must be bit-identical (C1)" ); } #[test] fn sim_optimal_manifest_renders_per_instrument_pip() { let m = sim_optimal_manifest(vec![], (Timestamp(1), Timestamp(2)), 0, 1.0); assert_eq!(m.broker, "sim-optimal(pip_size=1)"); let m2 = sim_optimal_manifest(vec![], (Timestamp(1), Timestamp(2)), 0, 0.0001); assert_eq!(m2.broker, "sim-optimal(pip_size=0.0001)"); } #[test] fn run_real_ger40_uses_index_pip() { // Gated: needs local GER40 data. Mirrors the existing real-path test's skip. let server = data_server::DataServer::new(data_server::DEFAULT_DATA_PATH); if !server.has_symbol("GER40") { eprintln!("skip: no local GER40 data at {}", data_server::DEFAULT_DATA_PATH); return; } // Bounded to the vetted Sept-2024 window so the pip-label + determinism // checks run fast (not the full unbounded archive). let env = project::Env::std(); let report = run_sample_real("GER40", Some(GER40_SEP2024_FROM_MS), Some(GER40_SEP2024_TO_MS), None, &env); // The looked-up index pip (1.0) reaches the manifest — not the FX 0.0001. assert_eq!(report.manifest.broker, "sim-optimal(pip_size=1)"); // Deterministic (C1): a second run yields the same report. let again = run_sample_real("GER40", Some(GER40_SEP2024_FROM_MS), Some(GER40_SEP2024_TO_MS), None, &env); assert_eq!(report.manifest.broker, again.manifest.broker); assert_eq!(report.metrics.total_pips, again.metrics.total_pips); } #[test] fn run_sample_is_deterministic_and_non_trivial() { let env = project::Env::std(); let r1 = run_sample(None, &env); let r2 = run_sample(None, &env); // C1 determinism: two runs are bit-identical (metrics + rendered JSON). assert_eq!(r1.metrics, r2.metrics); assert_eq!(r1.to_json(), r2.to_json()); let m = &r1.metrics; // exactly one bias sign flip in the demo trace (rises then reverses). assert_eq!(m.bias_sign_flips, 1); // a non-trivial, populated trace: a real drawdown. assert!(m.max_drawdown > 0.0); // hand-computed magnitudes for the chosen stream (float tolerance; the // computation's dust is ~1e-15). assert!( (m.max_drawdown - 0.17).abs() < 1e-9, "max_drawdown = {}", m.max_drawdown ); assert!( (m.total_pips - (-0.13)).abs() < 1e-9, "total_pips = {}", m.total_pips ); // manifest carries the sample's known configuration. let (from, to) = r1.manifest.window; assert_eq!((from.0, to.0), (1, 7)); // commit is the build's git identity (or the no-git "unknown" fallback); // either way it is non-empty and fixed at compile time, so it is stable // across runs of the same build (C1 determinism, already asserted above // via `to_json()`). assert!(!r1.manifest.commit.is_empty()); assert_eq!(r1.manifest.commit, r2.manifest.commit); } fn pair(name: &str, v: Scalar) -> (String, Scalar) { (name.to_string(), v) } #[test] fn member_key_renders_varying_axes_portably() { let named = vec![ pair("ema.length", Scalar::i64(5)), pair("bias.scale", Scalar::f64(0.5)), pair("longonly.enabled", Scalar::bool(true)), ]; let varying: std::collections::HashSet = named.iter().map(|(n, _)| n.clone()).collect(); let key = member_key(&named, &varying); assert_eq!(key, "ema.length-5_bias.scale-0.5_longonly.enabled-true"); assert!(key.chars().all(|c| c.is_ascii_alphanumeric() || matches!(c, '.' | '_' | '-'))); } #[test] fn member_key_omits_pinned_axes() { let named = vec![ pair("ema.length", Scalar::i64(5)), pair("bias.scale", Scalar::f64(0.5)), pair("longonly.enabled", Scalar::bool(false)), ]; let mut varying = std::collections::HashSet::new(); varying.insert("longonly.enabled".to_string()); assert_eq!(member_key(&named, &varying), "longonly.enabled-false"); } #[test] fn member_key_handles_negative_float_and_sanitises_names() { let named = vec![pair("bias.scale", Scalar::f64(-0.5))]; let varying: std::collections::HashSet = ["bias.scale".to_string()].into_iter().collect(); assert_eq!(member_key(&named, &varying), "bias.scale--0.5"); let named2 = vec![pair("weird key!", Scalar::f64(1.0))]; let varying2: std::collections::HashSet = ["weird key!".to_string()].into_iter().collect(); assert_eq!(member_key(&named2, &varying2), "weird_key_-1"); } #[test] fn member_key_is_m_when_no_axis_varies() { let named = vec![pair("bias.scale", Scalar::f64(0.5))]; let varying = std::collections::HashSet::new(); assert_eq!(member_key(&named, &varying), "m"); } #[test] fn member_key_caps_length_with_conformant_hash_fallback() { let named: Vec<(String, Scalar)> = (0..40) .map(|i| pair(&format!("some.long.axis.path.number.{i}"), Scalar::i64(i))) .collect(); let varying: std::collections::HashSet = named.iter().map(|(n, _)| n.clone()).collect(); let key = member_key(&named, &varying); assert!(key.len() <= MAX_KEY, "over-cap key not bounded: {} bytes", key.len()); assert!(key.chars().all(|c| c.is_ascii_alphanumeric() || matches!(c, '.' | '_' | '-'))); let mut named2 = named.clone(); named2[0].1 = Scalar::i64(999); assert_ne!(key, member_key(&named2, &varying), "distinct over-cap inputs must differ"); } #[test] fn member_key_collision_free_over_a_non_trend_axis_set() { // #105 regression: vary bias.scale + a bool (NOT the old hardcoded // trend.fast/slow). Distinct points -> distinct keys; the old f…s… key // would have collapsed them all to one dir. let varying: std::collections::HashSet = ["bias.scale".to_string(), "longonly.enabled".to_string()].into_iter().collect(); let p = |s: f64, b: bool| { vec![ pair("ema.length", Scalar::i64(5)), // pinned -> omitted from key pair("bias.scale", Scalar::f64(s)), pair("longonly.enabled", Scalar::bool(b)), ] }; let keys: Vec = [(0.5, true), (0.5, false), (1.0, true), (1.0, false)] .iter() .map(|&(s, b)| member_key(&p(s, b), &varying)) .collect(); let unique: std::collections::HashSet<&String> = keys.iter().collect(); assert_eq!(unique.len(), 4, "distinct points must yield distinct keys: {keys:?}"); } #[test] fn momentum_param_space_is_ema_exposure_longonly() { // pins the default node-name path segments (ema / exposure / longonly) and // the param order/kinds the member key + sweep depend on. let names: Vec = momentum_blueprint_with_sinks(SYNTHETIC_PIP_SIZE).0.param_space().into_iter().map(|p| p.name).collect(); assert_eq!( names, vec![ "ema.length".to_string(), "bias.scale".to_string(), "longonly.enabled".to_string(), ], ); } #[test] fn momentum_sweep_is_deterministic_and_has_eight_points() { let env = project::Env::std(); let a = momentum_sweep_family(None, &DataSource::Synthetic, &env); let b = momentum_sweep_family(None, &DataSource::Synthetic, &env); assert_eq!(a.points.len(), 8, "2x2x2 grid = 8 points"); assert_eq!(a, b, "C1: the momentum family is a pure function of the build"); } /// Property: the *shipped* `r_meanrev_graph` (the CLI compile unit, not /// the hand-rebuilt subgraph in `r_meanrev_e2e.rs`) FADES against the /// move — its exposure tap reads short (-1) above the band and long (+1) /// below, the sign-inverted-vs-breakout polarity that defines mean-reversion. /// A latch-polarity copy-paste from `r_breakout_graph` (swapping the /// `set`/`reset` legs) would leave the fold-vs-raw and window-grid CLI tests /// green yet silently invert the signal; only an observable read of this /// function's bias catches it. `k = 0` collapses the band to the lagging EWMA /// mean, isolating direction + latch from the sigma threshold; window 3 /// (alpha = 0.5) lags the level clearly. The exposure Recorder (`tx_ex`, /// `reduce = false`) carries the bias in col[0]. #[test] fn r_meanrev_graph_fades_short_above_the_band_and_long_below() { let (tx_eq, _rx_eq) = mpsc::channel(); let (tx_ex, rx_ex) = mpsc::channel(); let (tx_r, _rx_r) = mpsc::channel(); let (tx_req, _rx_req) = mpsc::channel(); let flat = r_meanrev_graph(tx_eq, tx_ex, tx_r, tx_req, Some(3), 0.0, 3, 2.0, false) .compile_with_params(&[]) .expect("r-meanrev blueprint compiles"); let mut h = Harness::bootstrap(flat).expect("r-meanrev harness bootstraps"); // calm (price == lagging mean -> no fade) | sustained UP (price > mean -> // fade SHORT) | sustained DOWN (price < mean -> fade LONG). let closes = [ 100.0, 100.0, 100.0, 100.0, 100.0, 100.0, 130.0, 130.0, 130.0, 70.0, 70.0, 70.0, 70.0, ]; let prices: Vec<(Timestamp, Scalar)> = closes.iter().enumerate().map(|(i, &c)| (Timestamp(i as i64), Scalar::f64(c))).collect(); let src: Vec> = vec![Box::new(VecSource::new(prices))]; h.run(src); let bias: Vec = rx_ex.try_iter().map(|(_, row): (Timestamp, Vec)| row[0].as_f64()).collect(); assert!(!bias.is_empty(), "the meanrev exposure tap must emit once warmed up"); assert_eq!(*bias.first().unwrap(), 0.0, "calm bars (price == mean) must not fade: {bias:?}"); let first_short = bias.iter().position(|&b| b == -1.0).expect("an up-move must fade SHORT (-1)"); let first_long = bias.iter().position(|&b| b == 1.0).expect("a down-move must fade LONG (+1)"); assert!(first_short < first_long, "short (up-fade) must precede long (down-fade): {bias:?}"); assert_eq!(*bias.last().unwrap(), 1.0, "the down-fade long must hold to the end: {bias:?}"); } /// Property: a `blueprint_sweep_family` member built from a serialized signal is /// the SAME trading result as the cycle-1 single run of that signal at the same /// params — the loaded-blueprint sweep reuses the identical `wrap_r` run path /// (the keystone). Every member of one family carries the SAME `topology_hash` (the /// loaded signal's, the deviation from the Rust-built mirror), and distinct grid /// points key to distinct `member_key`s (members are distinguishable). #[test] fn blueprint_sweep_member_equals_single_run_and_shares_topology_hash() { // An OPEN signal (both SMA knobs free) so the sweep can bind them by name; the // serialized doc round-trips to the topology the single run hashes. let env = project::Env::std(); let open = sma_signal(None, None); let doc = blueprint_to_json(&open).expect("serializes"); let data = DataSource::Synthetic; // fast pinned at 2, slow varied over {4, 6}: a 2x1 grid, slow the varying axis. let axes = vec![ ("sma_signal.fast.length".to_string(), vec![Scalar::i64(2)]), ("sma_signal.slow.length".to_string(), vec![Scalar::i64(4), Scalar::i64(6)]), ]; let family = blueprint_sweep_family(&doc, &axes, &data, &env).expect("named axes resolve"); assert_eq!(family.points.len(), 2, "2x1 grid -> 2 members"); // (b) every member carries the shared topology_hash of the loaded signal. let topo = topology_hash(&open); assert_eq!(topo.len(), 64, "topology_hash is a 64-hex SHA256"); for pt in &family.points { assert_eq!(pt.report.manifest.topology_hash.as_deref(), Some(topo.as_str())); } // (a) the slow=4 member reproduces the cycle-1 single run at fast=2, slow=4 — // same equity/exposure stream (total_pips/max_drawdown/bias_sign_flips) and the // same topology_hash, proving the loaded blueprint runs through the identical path. let single = run_signal_r( sma_signal(None, None), &[Scalar::i64(2), Scalar::i64(4)], RunData::Synthetic, 0, &env, ); let member4 = &family.points[0].report; // slow=4 is the first odometer point assert_eq!(member4.metrics, single.metrics, "loaded sweep member == single run"); assert_eq!(member4.manifest.topology_hash, single.manifest.topology_hash); // (c) the two members' keys differ (member_key over the varying slow.length axis). let varying: HashSet = ["sma_signal.slow.length".to_string()].into_iter().collect(); let k4 = member_key(&family.points[0].report.manifest.params, &varying); let k6 = member_key(&family.points[1].report.manifest.params, &varying); assert_ne!(k4, k6, "distinct grid points key distinctly"); } #[test] fn blueprint_axis_probe_lists_prefixed_open_knobs() { // The open fixture's two SMA lengths are the sweepable knobs; the probe // wraps the signal (name "sma_signal") so the names are prefixed — // exactly what `--axis` binds. let env = project::Env::std(); let open = include_str!("../tests/fixtures/sma_signal_open.json"); let space = blueprint_axis_probe(open, &env).param_space(); let names: Vec<&str> = space.iter().map(|p| p.name.as_str()).collect(); assert_eq!(names, ["sma_signal.fast.length", "sma_signal.slow.length"]); assert!(space.iter().all(|p| matches!(p.kind, ScalarKind::I64))); // A closed blueprint (both lengths bound) has no open axes. let closed = include_str!("../tests/fixtures/sma_signal.json"); assert!(blueprint_axis_probe(closed, &env).param_space().is_empty()); } #[test] fn blueprint_walkforward_family_refits_each_window() { // The open fixture's two SMA lengths are re-fit per IS window over a 2x2 grid. let env = project::Env::std(); let doc = include_str!("../tests/fixtures/sma_signal_open.json"); let axes = vec![ ("sma_signal.fast.length".to_string(), vec![Scalar::i64(2), Scalar::i64(3)]), ("sma_signal.slow.length".to_string(), vec![Scalar::i64(4), Scalar::i64(6)]), ]; let result = blueprint_walkforward_family(doc, &axes, &DataSource::Synthetic, Selection::Argmax, &env); // 24/12/12 over the 60-bar synthetic span -> 3 rolling windows. assert_eq!(result.windows.len(), 3, "three rolling IS/OOS windows"); for w in &result.windows { assert_eq!(w.run.chosen_params.len(), 2, "both axes re-fit each window"); assert!(w.run.oos_report.metrics.r.is_some(), "OOS record is R-metrics"); } // reduce-mode retains no raw pip curve -> the stitched pip-equity is empty. assert!(result.stitched_oos_equity.is_empty(), "no raw pip curve in reduce-mode"); } #[test] fn blueprint_mc_family_seeds_differ() { // MC over a CLOSED signal (both SMA knobs bound): 3 seeds -> 3 draws, one shared // topology_hash, and DIFFERING metrics — the seed reaches the DATA (a distinct // synthetic walk per draw), not just the manifest label. The anti-degenerate guard: // a regression to seed-as-label-only would make the three draws identical. let env = project::Env::std(); let closed = sma_signal(Some(2), Some(4)); let doc = blueprint_to_json(&closed).expect("serializes"); let family = blueprint_mc_family(&doc, 3, &DataSource::Synthetic, &env).expect("closed blueprint"); assert_eq!(family.draws.len(), 3, "one draw per seed"); assert_eq!(family.draws.iter().map(|d| d.seed).collect::>(), vec![1, 2, 3]); let topo = family.draws[0].report.manifest.topology_hash.clone(); assert!(topo.is_some(), "members carry a topology_hash"); assert!( family.draws.iter().all(|d| d.report.manifest.topology_hash == topo), "all members share one topology_hash" ); let m: Vec<_> = family.draws.iter().map(|d| &d.report.metrics).collect(); assert!(m[0] != m[1] || m[1] != m[2], "seeds must yield differing realizations"); } #[test] fn blueprint_mc_family_rejects_vacuous_deep_lookback() { // Property: the mc family builder REFUSES a silent-vacuous Monte-Carlo — one where // every per-seed draw collapses to a bit-identical realization — by RETURNING a named // error rather than an `Ok` family that looks like a real (but indistinguishable) // distribution. A CLOSED deep-lookback signal whose slow SMA length (60) equals the // fixed 60-bar synthetic walk never warms, so every seed yields zero trades and thus // identical metrics; that is a wrong result with no error (C10 refuse-don't-guess). let env = project::Env::std(); let deep = sma_signal(Some(2), Some(60)); // slow len == walk len -> never warms let doc = blueprint_to_json(&deep).expect("serializes"); let err = blueprint_mc_family(&doc, 3, &DataSource::Synthetic, &env) .expect_err("a vacuous (all-identical) Monte-Carlo is rejected, not returned"); assert!( err.contains("vacuous") || err.contains("identical"), "names the vacuous/degenerate condition: {err}" ); } #[test] fn blueprint_mc_family_rejects_an_open_blueprint() { // Property: the mc family builder REFUSES an open blueprint (free knobs) by RETURNING // a named error — never a hidden process exit — so the closed-blueprint precondition // is unit-testable (the IO wrapper renders it to stderr + exit 2, mirroring the sibling // blueprint_sweep_family). MC binds no axis, so a free knob would have no binder; the // rejection pre-empts the downstream compile_with_params arity panic. let env = project::Env::std(); let open = sma_signal(None, None); // both SMA knobs free -> non-empty param_space let doc = blueprint_to_json(&open).expect("serializes"); let err = blueprint_mc_family(&doc, 4, &DataSource::Synthetic, &env) .expect_err("an open blueprint is rejected, not run"); assert!(err.contains("closed blueprint"), "names the closed-blueprint requirement: {err}"); } #[test] fn reproduce_family_re_derives_every_member_bit_identically() { // a unique temp runs store so the on-disk family + blueprint store do not collide. let dir = std::env::temp_dir().join(format!("aura-repro-{}", std::process::id())); let _ = std::fs::remove_dir_all(&dir); std::fs::create_dir_all(&dir).expect("temp dir"); let reg = Registry::open(dir.join("runs.jsonl")); let env = project::Env::std(); let open = sma_signal(None, None); let doc = blueprint_to_json(&open).expect("serializes"); let data = DataSource::Synthetic; // 2x grid over slow.length {4,6} at fast=2 — slow=4 is the open-at-end member. let axes = vec![ ("sma_signal.fast.length".to_string(), vec![Scalar::i64(2)]), ("sma_signal.slow.length".to_string(), vec![Scalar::i64(4), Scalar::i64(6)]), ]; let family = blueprint_sweep_family(&doc, &axes, &data, &env).expect("axes resolve"); // persist exactly as run_blueprint_sweep does: store the blueprint, append the family. let topo = family.points[0].report.manifest.topology_hash.clone().expect("topo"); let canonical = blueprint_to_json(&blueprint_from_json(&doc, &|t| std_vocabulary(t)).unwrap()).unwrap(); reg.put_blueprint(&topo, &canonical).expect("store blueprint"); let id = reg .append_family("repro", FamilyKind::Sweep, &sweep_member_reports(&family)) .expect("append"); // reproduce: every member re-derives bit-identically (incl the open-at-end member). let rep = reproduce_family_in(®, &id, &data, &env); assert_eq!(rep.outcomes.len(), 2, "two members reproduced"); assert!( rep.outcomes.iter().all(|(_, ok)| *ok), "every member re-derives bit-identically: {:?}", rep.outcomes ); let _ = std::fs::remove_dir_all(&dir); } #[test] fn reproduce_family_re_derives_every_mc_member_bit_identically() { let dir = std::env::temp_dir().join(format!("aura-repro-mc-{}", std::process::id())); let _ = std::fs::remove_dir_all(&dir); std::fs::create_dir_all(&dir).expect("temp dir"); let reg = Registry::open(dir.join("runs.jsonl")); // a CLOSED signal (both SMA knobs bound) — MC binds no axis. let env = project::Env::std(); let closed = sma_signal(Some(2), Some(4)); let doc = blueprint_to_json(&closed).expect("serializes"); let data = DataSource::Synthetic; let family = blueprint_mc_family(&doc, 3, &data, &env).expect("closed blueprint"); // persist exactly as run_blueprint_mc does: store the blueprint, append the MC family. let topo = family.draws[0].report.manifest.topology_hash.clone().expect("topo"); let canonical = blueprint_to_json(&blueprint_from_json(&doc, &|t| std_vocabulary(t)).unwrap()).unwrap(); reg.put_blueprint(&topo, &canonical).expect("store blueprint"); let id = reg .append_family("mcrepro", FamilyKind::MonteCarlo, &mc_member_reports(&family)) .expect("append"); // reproduce: every MC member re-derives bit-identically (its seed-driven walk is // reconstructed from manifest.seed — the realization branch). let rep = reproduce_family_in(®, &id, &data, &env); assert_eq!(rep.outcomes.len(), 3, "three MC members reproduced"); assert!( rep.outcomes.iter().all(|(_, ok)| *ok), "every MC member re-derives bit-identically: {:?}", rep.outcomes ); let _ = std::fs::remove_dir_all(&dir); } /// Property: sweeping a FULLY BOUND (closed) blueprint is refused up front with a /// clear "fully bound / nothing to sweep" message — NOT the terse `UnknownKnob()` /// the per-axis resolve would otherwise emit (misleading: the named knob is not unknown, /// it is bound out, so there is simply nothing to sweep). Symmetric inverse of /// `blueprint_mc_family_rejects_an_open_blueprint` (mc requires a closed blueprint; a /// sweep requires >= 1 open knob). Type-agnostic on the error (renders it via Debug), so /// the RED phase compiles whether the builder returns a `BindError` or a `String`. #[test] fn blueprint_sweep_family_rejects_a_fully_bound_blueprint() { let env = project::Env::std(); let closed = sma_signal(Some(2), Some(4)); // both SMA knobs bound -> empty param_space let doc = blueprint_to_json(&closed).expect("serializes"); // an axis naming a bound-out knob: the pre-fix path returns UnknownKnob for it. let axes = vec![( "sma_signal.slow.length".to_string(), vec![Scalar::i64(4), Scalar::i64(6)], )]; let err = match blueprint_sweep_family(&doc, &axes, &DataSource::Synthetic, &env) { Ok(_) => panic!("a fully-bound blueprint has nothing to sweep"), Err(e) => format!("{e:?}"), }; assert!( err.contains("fully bound") || err.contains("nothing to sweep"), "names the fully-bound / nothing-to-sweep condition: {err}" ); assert!( !err.contains("UnknownKnob"), "must not leak the misleading UnknownKnob: {err}" ); } /// Property: an MC family's `aura reproduce` lines carry the member's own `seed=` /// label. MC members hold no tuning params (the params-join is empty), so the line would /// otherwise print a BLANK member label; the seed is each draw's realization identity and /// must show. Sweep / walk-forward labels still echo their params (covered elsewhere). #[test] fn reproduce_mc_member_labels_carry_the_seed() { let dir = std::env::temp_dir().join(format!("aura-repro-mc-seed-{}", std::process::id())); let _ = std::fs::remove_dir_all(&dir); std::fs::create_dir_all(&dir).expect("temp dir"); let reg = Registry::open(dir.join("runs.jsonl")); let env = project::Env::std(); let closed = sma_signal(Some(2), Some(4)); // MC binds no axis -> closed blueprint let doc = blueprint_to_json(&closed).expect("serializes"); let data = DataSource::Synthetic; let family = blueprint_mc_family(&doc, 3, &data, &env).expect("closed blueprint"); let topo = family.draws[0].report.manifest.topology_hash.clone().expect("topo"); let canonical = blueprint_to_json(&blueprint_from_json(&doc, &|t| std_vocabulary(t)).unwrap()).unwrap(); reg.put_blueprint(&topo, &canonical).expect("store blueprint"); let id = reg .append_family("mcseed", FamilyKind::MonteCarlo, &mc_member_reports(&family)) .expect("append"); let rep = reproduce_family_in(®, &id, &data, &env); assert_eq!(rep.outcomes.len(), 3, "three MC members"); for (label, _) in &rep.outcomes { assert!( label.starts_with("seed="), "an MC reproduce label carries the seed, not a blank params-join: {label:?}" ); } let seen: HashSet<&str> = rep.outcomes.iter().map(|(l, _)| l.as_str()).collect(); assert!( seen.contains("seed=1") && seen.contains("seed=2") && seen.contains("seed=3"), "each MC draw's own seed appears: {seen:?}" ); let _ = std::fs::remove_dir_all(&dir); } /// Property: an f64 blueprint param survives the content-addressed store's /// serialize -> parse -> re-serialize round-trip **bit-identically**, so `aura /// reproduce` re-derives an f64-bearing member without DIVERGED. This is exactly /// what workspace `serde_json/float_roundtrip` buys: the constant below is a /// full-precision f64 (`0.12387080150408619`) that the DEFAULT serde_json parser /// mis-parses by 1 ULP — with `float_roundtrip` on it parses back exactly, so the /// canonical bytes are stable and the re-run reproduces. The i64-axis reproduce /// tests exercise a `scale=0.5` blueprint (exactly representable), so this is the /// only test that actually depends on the feature. #[test] fn f64_blueprint_param_survives_store_round_trip_bit_identically() { // 1-ULP canary for serde_json float_roundtrip: parses back off-by-one without it. const HARD_SCALE: f64 = 0.12387080150408619; // A one-node signal whose only knob is the non-short-decimal f64 scale. let mut g = GraphBuilder::new("scale_probe"); let bias = g.add(Bias::builder().named("bias").bind("scale", Scalar::f64(HARD_SCALE))); let signal = g.source_role("signal", ScalarKind::F64); g.feed(signal, vec![bias.input("signal")]); g.expose(bias.output("bias"), "bias"); let sig = g.build().expect("one-node bias signal wiring resolves"); // the store keeps canonical bytes verbatim (dumb bytes-by-key), so the f64 // fidelity lives entirely in this serialize -> parse -> re-serialize hop. let doc = blueprint_to_json(&sig).expect("serializes"); assert!( doc.contains("0.12387080150408619"), "canonical JSON carries the full-precision f64: {doc}" ); let reloaded = blueprint_from_json(&doc, &|t| std_vocabulary(t)).expect("loads"); let doc2 = blueprint_to_json(&reloaded).expect("re-serializes"); // Without float_roundtrip the reparsed scale drifts 1 ULP and ryu re-serializes // it to a different string, so these canonical bytes would differ. assert_eq!(doc, doc2, "f64 blueprint param survives the store round-trip bit-identically"); } #[test] fn run_r_sma_synthetic_folds_an_r_block() { // the r-sma harness scores the SMA-cross signal in R: one shell-callable run // yields a RunReport whose metrics.r is Some with a finite SQN and >= 1 trade. let report = run_r_sma(RunData::Synthetic, None, None, None, None, &project::Env::std()); let r = report.metrics.r.as_ref().expect("r-sma run must populate metrics.r"); assert!(r.n_trades >= 1, "expected >= 1 trade, got {}", r.n_trades); assert!(r.sqn.is_finite(), "SQN must be finite, got {}", r.sqn); assert!(r.expectancy_r.is_finite(), "E[R] must be finite, got {}", r.expectancy_r); // #132: the dual-tap harness runs a RiskExecutor branch alongside the // SimBroker, so its manifest carries a dedicated broker label. assert!( report.manifest.broker.contains("risk-executor"), "r-sma manifest should carry a dedicated broker label, got: {}", report.manifest.broker ); } #[test] fn parse_param_cells_decodes_typed_cells_in_order_and_refuses_malformed() { // The property: `--params` round-trips the externally-tagged Scalar wire form in // array order, and a malformed array is refused with the flag named — never // silently dropped (a dropped cell would compile-bind a different graph). let cells = parse_param_cells("[{\"I64\":2},{\"F64\":0.5}]").expect("valid cell array"); assert_eq!(cells, vec![Scalar::I64(2), Scalar::F64(0.5)]); assert!(parse_param_cells("[]").expect("empty array").is_empty()); let err = parse_param_cells("{not an array}").unwrap_err(); assert!(err.contains("--params"), "a malformed --params value names the flag: {err}"); } /// Regenerates the committed demo signal blueprint the `aura run ` /// E2E loads. Ignored by default; run with `--ignored` after a signal change. #[test] #[ignore = "regenerates the committed demo signal blueprint fixture"] fn emit_demo_signal_fixture() { let json = blueprint_to_json(&sma_signal(Some(2), Some(4))).expect("serializes"); std::fs::write("tests/fixtures/sma_signal.json", json).expect("write fixture"); } /// Regenerates the committed OPEN demo signal blueprint the `aura sweep /// ` E2E sweeps — fast/slow left unbound, so its param_space exposes /// `sma_signal.fast.length` / `.slow.length` axes to grid. Ignored by default. #[test] #[ignore = "regenerates the committed open demo signal blueprint fixture"] fn emit_demo_signal_open_fixture() { let json = blueprint_to_json(&sma_signal(None, None)).expect("serializes"); std::fs::write("tests/fixtures/sma_signal_open.json", json).expect("write fixture"); } /// The cycle-1 keystone (C1): a signal serialized → loaded back through the /// public `blueprint_from_json` path runs bit-identically to its Rust-built /// twin — same metrics, same traces, same topology_hash. #[test] fn loaded_signal_runs_bit_identical_to_rust_built() { let env = project::Env::std(); let json = blueprint_to_json(&sma_signal(Some(2), Some(4))).expect("serializes"); let loaded = blueprint_from_json(&json, &|t| std_vocabulary(t)).expect("loads"); let a = run_signal_r(sma_signal(Some(2), Some(4)), &[], RunData::Synthetic, 0, &env); let b = run_signal_r(loaded, &[], RunData::Synthetic, 0, &env); assert_eq!(a.to_json(), b.to_json(), "loaded run is bit-identical incl. topology_hash"); assert_eq!(a.manifest.topology_hash.as_deref().map(str::len), Some(64), "64-hex sha256 present"); } /// `topology_hash` is deterministic per signal and distinguishes topologies — /// the #158 reproducibility-anchor property. #[test] fn topology_hash_is_stable_and_distinguishes() { let h = topology_hash(&sma_signal(Some(2), Some(4))); assert_eq!(h, topology_hash(&sma_signal(Some(2), Some(4))), "same signal -> same hash"); assert_ne!(h, topology_hash(&sma_signal(Some(3), Some(4))), "different topology -> different hash"); } /// #158 acc 1 (content-id stability across the store round-trip): a blueprint's content /// id survives serialize -> reload -> re-serialize (#164 idempotence) — the property /// content-addressed reproduction rests on (the stored bytes re-hash to the members' /// topology_hash). `content_id` is the shared primitive `topology_hash` uses. #[test] fn content_id_is_stable_across_the_store_round_trip() { let json = blueprint_to_json(&sma_signal(Some(2), Some(4))).expect("serializes"); let id = content_id(&json); assert_eq!(id.len(), 64, "a 64-hex sha256"); let reloaded = blueprint_from_json(&json, &|t| std_vocabulary(t)).expect("reloads"); assert_eq!( content_id(&blueprint_to_json(&reloaded).expect("re-serializes")), id, "content id survives serialize -> reload -> re-serialize (#164)" ); } /// #158 acc 3 (Tier-1 format addition leaves the content id unchanged): a Tier-1 /// optional field the blueprint does not use is tolerated by the loader (#156) and /// absent from the canonical omit-defaults form, so re-serializing yields the same /// bytes and thus the same content id. #[test] fn content_id_is_stable_across_a_tolerated_tier1_field() { let base = blueprint_to_json(&sma_signal(Some(2), Some(4))).expect("serializes"); // a future Tier-1 optional field injected at the top level (the blueprint does not use it). let with_extra = base.replacen("{\"format_version\":1,", "{\"format_version\":1,\"future_optional\":123,", 1); assert_ne!(base, with_extra, "the doc actually carries the extra field"); let reparsed = blueprint_from_json(&with_extra, &|t| std_vocabulary(t)) .expect("loader tolerates an unknown Tier-1 field (#156)"); assert_eq!( content_id(&blueprint_to_json(&reparsed).expect("re-serializes")), content_id(&base), "a Tier-1 optional the blueprint does not use leaves the content id unchanged" ); } /// #158 acc 1 (cross-surface agreement): `topology_hash` (from a live `Composite`, the /// run/sweep path) and the op-script `graph introspect --content-id` surface are the /// SAME hash of the SAME canonical bytes — both go through the one `content_id` /// primitive. Pins that the two command paths cannot silently drift apart. #[test] fn topology_hash_is_the_content_id_of_the_canonical_form() { let sig = sma_signal(Some(2), Some(4)); assert_eq!(topology_hash(&sig), content_id(&blueprint_to_json(&sig).expect("serializes"))); } /// The op-script twin of `sma_signal(Some(2), Some(4))`: same topology — /// SMA(2)/SMA(4) over one `price` role, spread, Bias with `scale` BOUND to /// 0.5 (= `R_SMA_BIAS_SCALE`; boundness is identity-bearing) — differing /// only in debug names (composite "graph" vs "sma_signal", unnamed Bias vs /// `.named("bias")`). const IDENTITY_TWIN_DOC: &str = r#"[ {"op":"source","role":"price","kind":"F64"}, {"op":"add","type":"SMA","name":"fast","bind":{"length":{"I64":2}}}, {"op":"add","type":"SMA","name":"slow","bind":{"length":{"I64":4}}}, {"op":"add","type":"Sub"}, {"op":"add","type":"Bias","bind":{"scale":{"F64":0.5}}}, {"op":"feed","role":"price","into":["fast.series","slow.series"]}, {"op":"connect","from":"fast.value","to":"sub.lhs"}, {"op":"connect","from":"slow.value","to":"sub.rhs"}, {"op":"connect","from":"sub.value","to":"bias.signal"}, {"op":"expose","from":"bias.bias","as":"bias"} ]"#; /// #171 acc 1 (cross-path identity): the Rust `sma_signal` builder and its /// op-script twin differ in canonical bytes (debug names) — distinct content /// ids — but project to the same identity JSON, hence one identity id across /// authoring paths. #[test] fn identity_id_bridges_the_rust_builder_and_op_script_paths() { let rust_built = sma_signal(Some(2), Some(4)); let env = project::Env::std(); let json = crate::graph_construct::build_from_str(IDENTITY_TWIN_DOC, &env) .expect("op-script twin builds"); assert_ne!( content_id(&json), topology_hash(&rust_built), "authoring paths keep distinct content ids" ); let loaded = blueprint_from_json(&json, &|t| std_vocabulary(t)).expect("twin reloads"); let identity = |c: &Composite| { content_id(&aura_engine::blueprint_identity_json(c).expect("identity-serializes")) }; assert_eq!( identity(&loaded), identity(&rust_built), "one topology -> one identity id across authoring paths" ); } #[test] fn mc_r_bootstrap_json_carries_every_bootstrap_field_under_the_mc_r_bootstrap_key() { // Property: the `mc_r_bootstrap` output line is the full RBootstrap shape — // each field is named and value-faithful, so a renamed/dropped field here // breaks a test instead of silently shipping. Pins the user-visible wire // shape of the mc R-bootstrap render (the parser + engine primitive are // covered elsewhere; this is the output-shape layer). let boot = r_bootstrap(&[1.0, -0.5, 2.0, -1.0], 64, 2, 7); let line = mc_r_bootstrap_json(&boot); let v: serde_json::Value = serde_json::from_str(&line).expect("canonical json line"); let obj = &v["mc_r_bootstrap"]; assert_eq!(obj["n_trades"], serde_json::json!(boot.n_trades)); assert_eq!(obj["block_len"], serde_json::json!(boot.block_len)); assert_eq!(obj["n_resamples"], serde_json::json!(boot.n_resamples)); assert_eq!(obj["prob_le_zero"], serde_json::json!(boot.prob_le_zero)); // e_r is the nested MetricStats block (mean + quantiles), not a flat scalar. assert_eq!(obj["e_r"], serde_json::to_value(&boot.e_r).expect("MetricStats serializes")); assert!(obj["e_r"]["mean"].is_number(), "e_r should nest the MetricStats block: {line}"); } #[test] fn mc_r_bootstrap_report_pools_a_non_empty_oos_r_series_over_synthetic() { // Property: the full real-R assembly path — walkforward_family(RSma) -> // pooled_oos_trade_rs -> r_bootstrap -> mc_r_bootstrap_json — wires up and // reduces a NON-EMPTY pooled OOS R series (the synthetic r-sma walk-forward // closes >= 1 trade across its windows). Guards the wiring + the non-empty // pooling branch the parser/primitive unit tests cannot reach; mirrors // `mc_report` / `walkforward_report`. Deterministic (C1). let env = project::Env::std(); let result = walkforward_family( Strategy::RSma, None, &DataSource::Synthetic, &RGrid::default(), Selection::Argmax, &env, ); let pooled = pooled_oos_trade_rs(&result); assert!(!pooled.is_empty(), "synthetic r-sma walk-forward must pool >= 1 OOS trade R"); let line = mc_r_bootstrap_report(&DataSource::Synthetic, &RGrid::default(), 1, 256, 1, &env); let v: serde_json::Value = serde_json::from_str(&line).expect("canonical json line"); let obj = &v["mc_r_bootstrap"]; // n_trades is the pooled-series length the bootstrap actually saw — non-zero // proves the assembled pooling fed the primitive, not an empty fallback. assert_eq!(obj["n_trades"], serde_json::json!(pooled.len())); assert!(obj["n_trades"].as_u64().expect("n_trades is an integer") >= 1); assert_eq!(obj["n_resamples"], serde_json::json!(256)); assert!(obj["e_r"]["mean"].as_f64().expect("e_r.mean is a number").is_finite()); // C1 at the CLI edge: the assembled real-R line is byte-identical on re-run. assert_eq!( line, mc_r_bootstrap_report(&DataSource::Synthetic, &RGrid::default(), 1, 256, 1, &env), ); } }