Struct SourceContributionV1 

pub struct SourceContributionV1 {
    pub axis: InputAxis,
    pub sigma: f64,
    pub d_drop_d_x: f64,
    pub d_windage_d_x: f64,
    pub scheme: DifferenceScheme,
    pub variance_share: f64,
    pub ellipse_area_reduction_m2: f64,
    pub p_hit_gain_if_perfect: Option<f64>,
}

One declared source’s contribution to impact variance at one range.

Constructed only for a source central_difference actually evaluated – an axis it refused is recorded in UnavailableSourceV1 instead, never here with a fabricated zero.

Fields

axis: InputAxis

sigma: f64

The caller-declared one-sigma uncertainty for this axis, in the axis’s own physical unit (crate::perturbation::axis_meta(axis).kind).

d_drop_d_x: f64

d_windage_d_x: f64

scheme: DifferenceScheme

Which finite-difference scheme produced the two derivatives above. A one-sided scheme (ForwardOneSided/BackwardOneSided) has larger truncation error than Central – see this module’s “Unavailable sources” doc section and DifferenceScheme’s own doc for when the kernel falls back to one. Still air (wind.speed_mps: 0.0) is the ordinary, ROUTINE case for WindSpeed, not an exotic one – wind-call uncertainty is this report’s flagship use case, so a non-Central scheme here is expected, not a warning sign by itself.

variance_share: f64

Fraction of total impact variance (sigma_drop_m^2 + sigma_windage_m^2) attributable to this source. Sums to 1.0 across a row’s sources whenever at least one source has a nonzero contribution.

ellipse_area_reduction_m2: f64

Reduction in the row’s 95% ellipse area if this source alone were measured perfectly (its sigma set to zero, every other source unchanged). Always >= 0.0.

Degenerate with two or fewer sources. The remaining covariance after removing any ONE of exactly two declared sources is an outer product (rank <= 1, area exactly 0 – see Symmetric2::largest_smallest_eigenvalues’s doc for why that is normal, not a bug), so with exactly two sources BOTH report a reduction equal to the FULL ellipse area, even when one dominates the variance share overwhelmingly (e.g. a 99%/1% split). That is literally true (perfecting either one alone does leave a zero-area ellipse) but reads, next to a ranking, as a tie where none exists on variance_share. A declaration of two sources (e.g. muzzle velocity and wind call, the most likely pair this ticket sees) always has this property; prefer variance_share to distinguish sources when sources.len() <= 2. With three or more sources the reduction is generically discriminating.

p_hit_gain_if_perfect: Option<f64>

The hit-probability gain over the row’s target if THIS source alone were measured perfectly (its sigma set to zero, every other source unchanged), i.e. p_hit(without this source) - p_hit(with every declared source). Some exactly when error_budget_with_target was given a target; None (never a fabricated 0.0) when no target was supplied, matching ErrorBudgetRowV1::p_hit.

Always >= 0.0 (see p_hit_bivariate’s doc: shrinking a target-centred impact covariance in the Loewner order cannot reduce the mass of a symmetric normal over a symmetric convex target – Anderson’s theorem). The unclamped value is checked against a small negative tolerance before being clamped to zero (a debug_assert! in error_budget_with_target), so a bug that made it SYSTEMATICALLY negative – as opposed to a few ULP of quadrature noise – fails a debug/test build loudly rather than being silently laundered into 0.0.

Does NOT share ellipse_area_reduction_m2’s two-source degeneracy above – it is one of the few fields on this type that DOES discriminate at sources.len() == 2, the muzzle-velocity/wind-call pair this ticket names as the most likely declaration. The ellipse-area field is degenerate there because AREA is exactly zero for ANY rank-1 covariance, blind to which specific rank-1 covariance it is – removing either of two sources leaves a different rank-1 covariance (one source’s own outer product, not the other’s), but both have zero area, so both report the same “full area” reduction regardless of the real variance split. Hit probability is not blind that way: it depends on the SHAPE of the remaining rank-1 covariance (which is oriented along the surviving source’s own (d_drop_d_x, d_windage_d_x) direction), not merely on whether it is singular. On this module’s own flagship two-source fixture (MuzzleVelocityMps + WindSpeed, resolved(), range 600 m, target 0.5 m x 0.75 m), measured p_hit_gain_if_perfect values are ~0.4416 (perfecting WindSpeed, which nearly dominates windage) and ~0.00008 (perfecting MuzzleVelocityMps) – a difference of three and a half orders of magnitude, not a tie. See p_hit_gain_if_perfect_discriminates_with_only_two_sources in this module’s tests.

Trait Implementations

impl Clone for SourceContributionV1

fn clone(&self) -> SourceContributionV1

Returns a duplicate of the value.

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

Performs copy-assignment from source.

impl Debug for SourceContributionV1

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

Formats the value using the given formatter.

impl<'de> Deserialize<'de> for SourceContributionV1

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

where __D: Deserializer<'de>,

Deserialize this value from the given Serde deserializer.

impl PartialEq for SourceContributionV1

fn eq(&self, other: &SourceContributionV1) -> bool

Equality operator ==.

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

Inequality operator !=.

impl Serialize for SourceContributionV1

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 SourceContributionV1