loess-rs 0.9.0

//! Batch adapter for standard LOESS smoothing.
//!
//! This module provides the batch execution adapter for LOESS smoothing.
//! It handles complete datasets in memory with sequential processing, making
//! it suitable for small to medium-sized datasets.
//!
//! ## srrstats Compliance
//!
//! @srrstats {G1.3} Builder pattern for fluent, validated configuration.
//! @srrstats {G2.0} Comprehensive input validation via Validator before processing.

// Feature-gated imports
#[cfg(not(feature = "std"))]
use alloc::vec::Vec;
#[cfg(feature = "std")]
use std::vec::Vec;

// External dependencies
use core::fmt::Debug;

// Internal dependencies
use crate::adapters::defaults::*;
use crate::algorithms::defaults::*;
use crate::algorithms::regression::{PolynomialDegree, SolverLinalg, ZeroWeightFallback};
use crate::algorithms::robustness::RobustnessMethod;
use crate::engine::defaults::*;
use crate::engine::executor::{
    CVPassFn, FitPassFn, IntervalPassFn, KDTreeBuilderFn, LoessConfig, LoessExecutor, SmoothPassFn,
    SurfaceMode, VertexPassFn,
};
use crate::engine::output::LoessResult;
use crate::engine::validator::Validator;
use crate::evaluation::cv::CVKind;
use crate::evaluation::diagnostics::Diagnostics;
use crate::evaluation::intervals::IntervalMethod;
use crate::math::boundary::BoundaryPolicy;
use crate::math::defaults::*;
use crate::math::distance::{DistanceLinalg, DistanceMetric};
use crate::math::hat_matrix::HatMatrixStats;
use crate::math::kernel::WeightFunction;
use crate::math::linalg::FloatLinalg;
use crate::math::scaling::ScalingMethod;
use crate::primitives::backend::Backend;
use crate::primitives::errors::LoessError;

// Builder for batch LOESS processor.
#[derive(Debug, Clone)]
pub struct BatchLoessBuilder<T: FloatLinalg + DistanceLinalg + SolverLinalg> {
    // Smoothing fraction (span)
    pub fraction: T,

    // Number of robustness iterations
    pub iterations: usize,

    // Kernel weight function
    pub weight_function: WeightFunction,

    // Robustness method
    pub robustness_method: RobustnessMethod,

    // Residual scaling method
    pub scaling_method: ScalingMethod,

    // Confidence/Prediction interval configuration
    pub interval_type: Option<IntervalMethod<T>>,

    // Fractions for cross-validation
    pub cv_fractions: Option<Vec<T>>,

    // Cross-validation method kind
    pub cv_kind: Option<CVKind>,

    // Cross-validation seed
    pub cv_seed: Option<u64>,

    // Deferred error from adapter conversion
    pub deferred_error: Option<LoessError>,

    // Tolerance for auto-convergence
    pub auto_converge: Option<T>,

    // Whether to compute diagnostic statistics
    pub return_diagnostics: bool,

    // Whether to return residuals
    pub compute_residuals: bool,

    // Whether to return robustness weights
    pub return_robustness_weights: bool,

    // Policy for handling zero-weight neighborhoods
    pub zero_weight_fallback: ZeroWeightFallback,

    // Policy for handling data boundaries
    pub boundary_policy: BoundaryPolicy,

    // Polynomial degree for local regression
    pub polynomial_degree: PolynomialDegree,

    // Number of predictor dimensions (default: 1).
    pub dimensions: usize,

    // Distance metric for nD neighborhood computation.
    pub distance_metric: DistanceMetric<T>,

    // Cell size for interpolation subdivision (default: 0.2).
    pub cell: Option<f64>,

    // Maximum number of vertices for interpolation.
    pub interpolation_vertices: Option<usize>,

    // Evaluation mode (default: Interpolation)
    pub surface_mode: SurfaceMode,

    // Whether to reduce polynomial degree to Linear at boundary vertices during interpolation.
    // When `true` (default), vertices outside the tight data bounds use Linear fits.
    // Set to `false` to match R's loess behavior exactly.
    pub boundary_degree_fallback: bool,

    // Tracks if any parameter was set multiple times (for validation)
    #[doc(hidden)]
    pub(crate) duplicate_param: Option<&'static str>,

    // ++++++++++++++++++++++++++++++++++++++
    // +               DEV                  +
    // ++++++++++++++++++++++++++++++++++++++
    // Custom smooth pass function.
    #[doc(hidden)]
    pub custom_smooth_pass: Option<SmoothPassFn<T>>,

    // Custom cross-validation pass function.
    #[doc(hidden)]
    pub custom_cv_pass: Option<CVPassFn<T>>,

    // Custom interval estimation pass function.
    #[doc(hidden)]
    pub custom_interval_pass: Option<IntervalPassFn<T>>,

    // Custom fit pass function.
    #[doc(hidden)]
    pub custom_fit_pass: Option<FitPassFn<T>>,

    // Custom vertex pass function.
    #[doc(hidden)]
    pub custom_vertex_pass: Option<VertexPassFn<T>>,

    // Custom KD-tree builder function.
    #[doc(hidden)]
    pub custom_kdtree_builder: Option<KDTreeBuilderFn<T>>,

    // Execution backend hint.
    #[doc(hidden)]
    pub backend: Option<Backend>,

    // Parallel execution hint.
    #[doc(hidden)]
    pub parallel: Option<bool>,

    // User-defined case weights (one per observation).
    pub custom_weights: Option<Vec<T>>,
}

impl<T: FloatLinalg + DistanceLinalg + Debug + Send + Sync + SolverLinalg> Default
    for BatchLoessBuilder<T>
{
    fn default() -> Self {
        Self::new()
    }
}

impl<T: FloatLinalg + DistanceLinalg + Debug + Send + Sync + SolverLinalg> BatchLoessBuilder<T> {
    // Create a new batch LOESS builder with default parameters.
    fn new() -> Self {
        Self {
            fraction: T::from(DEFAULT_FRACTION).unwrap(),
            iterations: DEFAULT_ITERATIONS,
            weight_function: DEFAULT_WEIGHT_FUNCTION_ENUM,
            robustness_method: DEFAULT_ROBUSTNESS_METHOD_ENUM,
            scaling_method: DEFAULT_SCALING_METHOD_ENUM,
            interval_type: None,
            cv_fractions: None,
            cv_kind: None,
            cv_seed: DEFAULT_CV_SEED,
            deferred_error: None,
            auto_converge: default_auto_converge(),
            return_diagnostics: DEFAULT_RETURN_DIAGNOSTICS,
            compute_residuals: DEFAULT_RETURN_RESIDUALS,
            return_robustness_weights: DEFAULT_RETURN_ROBUSTNESS_WEIGHTS,
            zero_weight_fallback: DEFAULT_ZERO_WEIGHT_FALLBACK_ENUM,
            boundary_policy: DEFAULT_BOUNDARY_POLICY_ENUM,
            polynomial_degree: DEFAULT_POLYNOMIAL_DEGREE_ENUM,
            dimensions: DEFAULT_DIMENSIONS,
            distance_metric: default_distance_metric(),
            cell: None,
            interpolation_vertices: None,
            surface_mode: DEFAULT_SURFACE_MODE_ENUM,
            boundary_degree_fallback: DEFAULT_BOUNDARY_DEGREE_FALLBACK,
            duplicate_param: None,
            // ++++++++++++++++++++++++++++++++++++++
            // +               DEV                  +
            // ++++++++++++++++++++++++++++++++++++++
            custom_smooth_pass: None,
            custom_cv_pass: None,
            custom_interval_pass: None,
            custom_fit_pass: None,
            custom_vertex_pass: None,
            custom_kdtree_builder: None,
            parallel: None,
            backend: None,
            custom_weights: None,
        }
    }

    // Build the batch processor.
    pub fn build(self) -> Result<BatchLoess<T>, LoessError> {
        if let Some(err) = self.deferred_error {
            return Err(err);
        }

        // Check for duplicate parameter configuration
        Validator::validate_no_duplicates(self.duplicate_param)?;

        // Validate fraction
        Validator::validate_fraction(self.fraction)?;

        // Validate iterations
        Validator::validate_iterations(self.iterations)?;

        // Validate interval type
        if let Some(ref method) = self.interval_type {
            Validator::validate_interval_level(method.level)?;
        }

        // Validate CV fractions and method
        if let Some(ref fracs) = self.cv_fractions {
            Validator::validate_cv_fractions(fracs)?;
        }
        if let Some(CVKind::KFold(k)) = self.cv_kind {
            Validator::validate_kfold(k)?;
        }

        // Validate auto convergence tolerance
        if let Some(tol) = self.auto_converge {
            Validator::validate_tolerance(tol)?;
        }

        Ok(BatchLoess { config: self })
    }
}

// Batch LOESS processor.
#[derive(Clone)]
pub struct BatchLoess<T: FloatLinalg + DistanceLinalg + SolverLinalg> {
    config: BatchLoessBuilder<T>,
}

impl<T: FloatLinalg + DistanceLinalg + Debug + Send + Sync + 'static + SolverLinalg> BatchLoess<T> {
    // Perform LOESS smoothing on the provided data.
    pub fn fit(self, x: &[T], y: &[T]) -> Result<LoessResult<T>, LoessError> {
        Validator::validate_inputs(x, y, self.config.dimensions)?;

        // Validate custom_weights length if provided
        if let Some(ref uw) = self.config.custom_weights {
            Validator::validate_custom_weights(uw, y.len())?;
        }

        // KD-Tree handles unsorted data natively - no need to sort

        // Check grid resolution only for interpolation mode
        if self.config.surface_mode == SurfaceMode::Interpolation {
            let n = y.len();
            let cell_to_use = self.config.cell.unwrap_or(0.2);
            let limit = self.config.interpolation_vertices.unwrap_or(n);
            let cell_provided = self.config.cell.is_some();
            let limit_provided = self.config.interpolation_vertices.is_some();

            Validator::validate_interpolation_grid(
                T::from(cell_to_use).unwrap_or_else(|| T::from(0.2).unwrap()),
                self.config.fraction,
                self.config.dimensions,
                limit,
                cell_provided,
                limit_provided,
            )?;
        }

        // Configure batch execution
        let config = LoessConfig {
            fraction: Some(self.config.fraction),
            iterations: self.config.iterations,
            weight_function: self.config.weight_function,
            zero_weight_fallback: self.config.zero_weight_fallback,
            robustness_method: self.config.robustness_method,
            scaling_method: self.config.scaling_method,
            cv_fractions: self.config.cv_fractions,
            cv_kind: self.config.cv_kind,
            auto_converge: self.config.auto_converge,
            return_variance: self.config.interval_type,
            boundary_policy: self.config.boundary_policy,
            polynomial_degree: self.config.polynomial_degree,
            dimensions: self.config.dimensions,
            distance_metric: self.config.distance_metric.clone(),
            cv_seed: self.config.cv_seed,
            surface_mode: self.config.surface_mode,
            interpolation_vertices: self.config.interpolation_vertices,
            cell: self.config.cell,
            boundary_degree_fallback: self.config.boundary_degree_fallback,
            custom_weights: self.config.custom_weights,
            // ++++++++++++++++++++++++++++++++++++++
            // +               DEV                  +
            // ++++++++++++++++++++++++++++++++++++++
            custom_smooth_pass: self.config.custom_smooth_pass,
            custom_cv_pass: self.config.custom_cv_pass,
            custom_interval_pass: self.config.custom_interval_pass,
            custom_fit_pass: self.config.custom_fit_pass,
            custom_vertex_pass: self.config.custom_vertex_pass,
            custom_kdtree_builder: self.config.custom_kdtree_builder,
            parallel: self.config.parallel.unwrap_or(false),
            backend: self.config.backend,
        };

        // Execute unified LOESS (KD-Tree handles unsorted data)
        let result = LoessExecutor::run_with_config(x, y, config);

        let y_smooth = result.smoothed;
        let std_errors = result.std_errors;
        let iterations_used = result.iterations;
        let fraction_used = result.used_fraction;
        let cv_scores = result.cv_scores;

        // Calculate residuals (data is in original order, no unsorting needed)
        let residuals: Vec<T> = y
            .iter()
            .zip(y_smooth.iter())
            .map(|(&orig, &smoothed_val)| orig - smoothed_val)
            .collect();

        // Get robustness weights from executor result (final iteration weights)
        let rob_weights = if self.config.return_robustness_weights {
            result.robustness_weights
        } else {
            Vec::new()
        };

        // Compute diagnostic statistics if requested
        let diagnostics = if self.config.return_diagnostics {
            Some(Diagnostics::compute(
                y,
                &y_smooth,
                &residuals,
                std_errors.as_deref(),
            ))
        } else {
            None
        };

        // Compute hat matrix statistics from leverage if available
        // (Must happen before residuals is moved into residuals_out)
        let (enp, trace_hat, delta1, delta2, residual_scale, leverage_out) =
            if let Some(lev) = result.leverage {
                let stats = HatMatrixStats::from_leverage(lev);
                // Compute RSS (residual sum of squares)
                let rss = residuals.iter().fold(T::zero(), |acc, &r| acc + r * r);
                let res_scale = stats.compute_residual_scale(rss);
                (
                    Some(stats.trace),
                    Some(stats.trace),
                    Some(stats.delta1),
                    Some(stats.delta2),
                    Some(res_scale),
                    Some(stats.leverage),
                )
            } else {
                (None, None, None, None, None, None)
            };

        // Compute intervals
        let (conf_lower, conf_upper, pred_lower, pred_upper) =
            if let Some(method) = &self.config.interval_type {
                if let Some(se) = &std_errors {
                    method.compute_intervals(&y_smooth, se, &residuals, delta1, delta2)?
                } else {
                    (None, None, None, None)
                }
            } else {
                (None, None, None, None)
            };

        // Results are already in original order (no sorting/unsorting needed with KD-Tree)
        let residuals_out = if self.config.compute_residuals {
            Some(residuals)
        } else {
            None
        };
        let rob_weights_out = if self.config.return_robustness_weights {
            Some(rob_weights)
        } else {
            None
        };

        Ok(LoessResult {
            x: x.to_vec(),
            dimensions: self.config.dimensions,
            distance_metric: self.config.distance_metric.clone(),
            polynomial_degree: self.config.polynomial_degree,
            y: y_smooth,
            standard_errors: std_errors,
            confidence_lower: conf_lower,
            confidence_upper: conf_upper,
            prediction_lower: pred_lower,
            prediction_upper: pred_upper,
            residuals: residuals_out,
            robustness_weights: rob_weights_out,
            fraction_used,
            iterations_used,
            cv_scores,
            diagnostics,
            enp,
            trace_hat,
            delta1,
            delta2,
            residual_scale,
            leverage: leverage_out,
        })
    }
}