Struct SpatialResult 

pub struct SpatialResult {

    pub density_maps: Vec<Array2<f64>>,
    pub uncertainty_maps: Vec<Array2<f64>>,
    pub chi_squared_map: Array2<f64>,
    pub deviance_per_dof_map: Option<Array2<f64>>,
    pub converged_map: Array2<bool>,
    pub temperature_map: Option<Array2<f64>>,
    pub temperature_uncertainty_map: Option<Array2<f64>>,
    pub isotope_labels: Vec<String>,
    pub anorm_map: Option<Array2<f64>>,
    pub background_maps: Option<[Array2<f64>; 3]>,
    pub back_d_map: Option<Array2<f64>>,
    pub back_f_map: Option<Array2<f64>>,
    pub t0_us_map: Option<Array2<f64>>,
    pub l_scale_map: Option<Array2<f64>>,
    pub n_converged: usize,
    pub n_total: usize,
    pub n_failed: usize,
}

Result of spatial mapping over a 2D image.

NaN-on-failure contract (issue #458 B1/B2): every per-pixel parameter map (density_maps, uncertainty_maps, chi_squared_map, deviance_per_dof_map, temperature_map, temperature_uncertainty_map, anorm_map, background_maps, back_d_map, back_f_map, t0_us_map, l_scale_map) contains NaN at every pixel where converged_map is false. The only map written unconditionally is converged_map itself — it is how callers discover that a pixel failed. Callers rendering numeric values should gate on converged_map (or check value.is_finite()) to avoid displaying the placeholder NaN.

Fields

density_maps: Vec<Array2<f64>>

Fitted areal density maps, one per isotope. Each Array2 has shape (height, width). NaN at pixels where converged_map is false.

uncertainty_maps: Vec<Array2<f64>>

Uncertainty maps, one per isotope. NaN at pixels where converged_map is false.

chi_squared_map: Array2<f64>

Reduced chi-squared map. For the counts-KL dispatch (joint-Poisson deviance per memo 35 §P1.2) this is back-compat-mirrored to D/(n−k); the semantically-correct per-pixel value is also exposed as Self::deviance_per_dof_map. NaN at pixels where converged_map is false.

deviance_per_dof_map: Option<Array2<f64>>

Per-pixel conditional binomial deviance D/(n−k) map. Some when the effective per-pixel solver is the counts-KL dispatch (joint-Poisson); None for LM-only runs and transmission+PoissonKL where Pearson χ²/dof is the GOF. NaN at pixels where converged_map is false.

converged_map: Array2<bool>

Convergence map (true = converged).

temperature_map: Option<Array2<f64>>

Fitted temperature map (K). Some when config.fit_temperature() is true. NaN at pixels where converged_map is false.

temperature_uncertainty_map: Option<Array2<f64>>

Per-pixel temperature uncertainty map (K, 1-sigma). Some when config.fit_temperature() is true. Entries are NaN where uncertainty was unavailable for that pixel.

isotope_labels: Vec<String>

Isotope labels captured at compute time, one per density map. Ensures display labels stay in sync with density data even if the user modifies the isotope list after fitting.

anorm_map: Option<Array2<f64>>

Per-pixel SAMMY Anorm map (when background fitting is enabled). NaN at pixels where converged_map is false.

background_maps: Option<[Array2<f64>; 3]>

Per-pixel SAMMY background polynomial coefficient maps — only the first three coefficients [BackA, BackB, BackC] of the SAMMY 6-term form bg(E) = BackA + BackB/√E + BackC·√E + BackD·exp(-BackF/√E). Both LM-transmission and counts-KL paths use these semantics (legacy alpha-fitting [b0, b1, alpha_2] layout was retired with fit_counts_poisson in PR #450).

The exponential BackD/BackF terms are surfaced separately in Self::back_d_map / Self::back_f_map — both None for counts-KL runs (the joint-Poisson dispatch never fits the exponential tail) and for LM transmission runs that left fit_back_d / fit_back_f at their default false.

NaN at pixels where converged_map is false.

back_d_map: Option<Array2<f64>>

Per-pixel fitted SAMMY exponential background amplitude BackD. Some only when the LM transmission background path was active AND fit_back_d=true; None otherwise (counts-KL runs, LM runs without a background model, and LM runs that fit the polynomial terms but left the exponential tail at its initial value). NaN at pixels where converged_map is false.

back_f_map: Option<Array2<f64>>

Per-pixel fitted SAMMY exponential background decay constant BackF. Some only when the LM transmission background path was active AND fit_back_f=true; None otherwise. Mirrors Self::back_d_map’s gating because BackD and BackF are required to fit together (see validate_transmission_background in crate::pipeline). NaN at pixels where converged_map is false.

t0_us_map: Option<Array2<f64>>

Per-pixel fitted SAMMY TZERO offset (µs) map. Some when config.fit_energy_scale is true; None otherwise. NaN at pixels where converged_map is false.

l_scale_map: Option<Array2<f64>>

Per-pixel fitted SAMMY TZERO flight-path scale factor. Some when config.fit_energy_scale is true; None otherwise. NaN at pixels where converged_map is false.

n_converged: usize

Number of pixels that converged.

n_total: usize

Total number of pixels fitted.

n_failed: usize

Number of pixels where the fitter returned an error (not just non-convergence — a hard failure like invalid parameters or NaN model output). These pixels have NaN density and false convergence.

Trait Implementations

impl Debug for SpatialResult

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

Formats the value using the given formatter.