Struct EncodedDataset 

pub struct EncodedDataset {
    pub headers: Vec<String>,
    pub values: Array2<f64>,
    pub schema: DataSchema,
    pub column_kinds: Vec<ColumnKindTag>,
}

Fields

headers: Vec<String>

values: Array2<f64>

schema: DataSchema

column_kinds: Vec<ColumnKindTag>

Implementations

impl EncodedDataset

pub fn validate_fit_boundary(&self) -> Result<(), DataError>

Validate the table at the common formula-fit boundary.

This deliberately lives below formula materialization: library, CLI, and Python fits all carry an EncodedDataset through that seam, so a degenerate design can never acquire frontend-specific behaviour.

Constancy is NOT a boundary rule: a constant column is legitimate input for many designs (an all-zero left-truncation entry time, an event indicator, a scalar term the model prunes) and the layers that judge it carry the specific message (an all-zero count response, #2255; a constant calibrated score column in a marginal-slope fit).

pub fn column_map(&self) -> HashMap<String, usize>

pub fn feature_ranges(&self) -> Vec<(f64, f64)>

Per-column finite (min, max) of the training values, parallel to headers. Columns with no finite values default to (0.0, 0.0) so that downstream clipping is a no-op for them. Used to populate training_feature_ranges so prediction can clip out-of-hull inputs to the training bounding box.

Trait Implementations

impl Clone for EncodedDataset

fn clone(&self) -> EncodedDataset

Returns a duplicate of the value.

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

Performs copy-assignment from source.

impl Debug for EncodedDataset

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

Formats the value using the given formatter.