Struct MetricSnapshot 

pub struct MetricSnapshot {

    pub metric: String,
    pub baseline: f64,
    pub unit: String,
    pub n: u64,
    pub window: String,
    pub subject: Option<String>,
    pub namespace: Option<String>,
    pub relation: Option<String>,
    pub query: String,
    pub review_after_ms: i64,
    pub horizons_ms: Vec<i64>,
    pub checkpoints: Vec<Checkpoint>,
    pub higher_is_better: bool,
}

A reproducible metric snapshot; powers outcome review.

Fields

metric: String

Metric kind — the engine knows how to re-measure a fixed set (e.g. tool_error_recurrence); unknown kinds are skipped, not faked.

baseline: f64

unit: String

n: u64

window: String

subject: Option<String>

The subject the metric is about (e.g. the tool name), used by the engine’s typed re-measurement.

namespace: Option<String>

Namespace scope for the re-measurement, normalized (case-fold/trim). None = all namespaces. Additive: older snapshots deserialize to None and existing metric kinds ignore it.

relation: Option<String>

Relation scope for fact-shaped metrics (e.g. which functional relation a contradiction was resolved under), normalized. Additive like namespace.

query: String

CAL that recomputes the metric at verify time (reproducibility / documentation; the engine re-measures with typed reads).

review_after_ms: i64

How long after apply to re-measure, in epoch-ms delta (the first / only checkpoint when horizons_ms is empty).

horizons_ms: Vec<i64>

A schedule of checkpoints (ms after apply) to re-measure at. An outcome that held at an early checkpoint can regress at a later one, so a verdict is never final until the last horizon. Empty → [review_after_ms].

checkpoints: Vec<Checkpoint>

The schedule in the deployment’s own unit (Checkpoint). When set it is THE schedule and horizons_ms/review_after_ms are ignored; empty (every snapshot written before it existed) falls back to them.

higher_is_better: bool

Which direction is an improvement. The built-in metrics are all recurrence counts, where lower is better — so this defaults to false and every existing snapshot deserializes unchanged. An evalset accuracy is the opposite, and getting it wrong does not merely misreport: the Verify gate would read a rule that IMPROVED accuracy as a regression and propose reverting it.

Implementations

impl MetricSnapshot

pub fn schedule(&self) -> Vec<Checkpoint>

The measurement schedule, sorted and deduplicated: checkpoints when set, else horizons_ms, else the single review_after_ms — each older spelling read as time, so a snapshot from before Checkpoint existed measures exactly as it did.

pub fn horizons(&self) -> Vec<i64>

The time checkpoints of the schedule, in ms — the pre-Checkpoint view, kept for callers that only understand time.

Trait Implementations

impl Clone for MetricSnapshot

fn clone(&self) -> Self

Returns a duplicate of the value.

fn clone_from(&mut self, source: &Self)

Performs copy-assignment from source.

impl Debug for MetricSnapshot

fn fmt(&self, f: &mut Formatter<'_>) -> Result

Formats the value using the given formatter.

impl<'de> Deserialize<'de> for MetricSnapshot

fn deserialize<__D>(__deserializer: __D) -> Result<Self, __D::Error>

where __D: Deserializer<'de>,

Deserialize this value from the given Serde deserializer.

impl PartialEq for MetricSnapshot

fn eq(&self, other: &Self) -> bool

Equality operator ==.

fn ne(&self, other: &Rhs) -> bool

Inequality operator !=.

impl Serialize for MetricSnapshot

fn serialize<__S>(&self, __serializer: __S) -> Result<__S::Ok, __S::Error>

where __S: Serializer,

Serialize this value into the given Serde serializer.

impl StructuralPartialEq for MetricSnapshot