Struct FeatureExtractor 

pub struct FeatureExtractor { /* private fields */ }

Composable feature engineering pipeline.

Holds an ordered list of FeatureSpecs. Each spec names one input feature and chains transforms that convert it to one or more f64 outputs.

Thread-safety

FeatureExtractor is not Send/Sync by default because of its internal AtomicU64 counter — if you need concurrent access, wrap it in a Mutex.

Implementations

impl FeatureExtractor

pub fn new() -> Self

Create an empty extractor with no specs.

pub fn add_spec(&mut self, spec: FeatureSpec) -> &mut Self

Append a spec to the pipeline (builder-style).

pub fn output_dim(&self) -> usize

Compute the total output dimension from spec metadata alone.

For OneHotEncode { categories } the contribution is categories.len(). For PolynomialFeatures { degree } it is degree as usize. All other specs contribute 1.

pub fn feature_names(&self) -> Vec<String>

List all output feature names (expanded for OneHot and Polynomial).

pub fn extract( &self, input: &HashMap<String, FeatureValue>, ) -> Result<ExtractedFeatures, FeatureError>

Extract features from a single input map.

Returns an error if a required feature is missing (no imputer configured) or a type mismatch prevents a transform from running.

pub fn extract_batch( &self, inputs: &[HashMap<String, FeatureValue>], ) -> Result<Vec<ExtractedFeatures>, FeatureError>

Extract features from a batch of input maps.

Returns Err(FeatureError::EmptyInput) when inputs is empty.

pub fn stats(&self) -> FePipelineStats

Return a snapshot of accumulated statistics.

Trait Implementations

impl Default for FeatureExtractor

fn default() -> Self

Returns the “default value” for a type.