//! Microstructure Features for HFT ML Models
//!
//! This module implements high-frequency microstructure features for real-time trading:
//! - Amihud Illiquidity Ratio: Measures price impact per unit of volume
//! - Roll Measure: Estimates effective bid-ask spread (Agent A9)
//! - Corwin-Schultz: High-low spread estimator (Agent A10)
//!
//! ## Performance Targets
//! - Latency: <8μs per feature update (Amihud), <5μs (Roll, Corwin-Schultz)
//! - Memory: ≤72 bytes per symbol per feature
//! - Data: OHLCV-only (no Level-2 order book required)
//!
//! ## Integration
//! These features are part of the 256-dimension training feature vector:
//! - Features 115-164: Microstructure proxies (50 features)
//!
//! ## References
//! - Amihud (2002): "Illiquidity and Stock Returns"
//! - Roll (1984): "A Simple Implicit Measure of the Effective Bid-Ask Spread"
//! - Corwin & Schultz (2012): "A Simple Way to Estimate Bid-Ask Spreads"
//! - Hudson & Thames `MLFinLab`: Research-backed implementations

/// Trait for microstructure features with normalization for ML models
pub trait MicrostructureFeatures {
    /// Returns the feature name for logging/debugging
    fn feature_name(&self) -> &'static str;

    /// Returns the raw feature value
    fn value(&self) -> f64;

    /// Returns normalized feature value for ML training (typically [-1, 1])
    fn get_normalized(&self) -> f64;

    /// Resets internal state (useful for backtesting)
    fn reset(&mut self);
}

// ============================================================================
// Amihud Illiquidity Ratio (Agent A8)
// ============================================================================

/// Amihud Illiquidity Ratio: Measures price impact per unit of trading volume
///
/// ## Formula
/// ```text
/// Illiquidity_t = |return_t| / dollar_volume_t
/// ```
///
/// Where:
/// - `return_t` = (`price_t` - price_{t-1}) / price_{t-1}
/// - `dollar_volume_t` = `price_t` * `volume_t`
///
/// ## Interpretation
/// - **High illiquidity** (>1e-6): Large price impact per dollar traded (illiquid market)
/// - **Low illiquidity** (<1e-9): Small price impact (liquid market)
/// - Used for transaction cost estimation and position sizing
///
/// ## Implementation
/// Uses Exponential Moving Average (EMA) for smoothing:
/// ```text
/// EMA_illiquidity_t = α * instant_illiquidity_t + (1-α) * EMA_illiquidity_{t-1}
/// ```
///
/// ## Performance
/// - **Latency**: 3-8μs per update (O(1) complexity)
/// - **Memory**: 24 bytes (3 f64 fields)
/// - **Data**: OHLCV only (no tick data required)
///
/// ## Example
/// ```rust
/// use ml::features::microstructure::AmihudIlliquidity;
///
/// let mut amihud = AmihudIlliquidity::new(0.05); // α=0.05 for 20-bar window
///
/// // Feed OHLCV bars
/// amihud.update(100.0, 10000.0); // price, volume
/// amihud.update(101.0, 12000.0);
///
/// let illiquidity = amihud.value();
/// let normalized = amihud.get_normalized(); // For ML training
/// ```
#[derive(Debug, Clone)]
pub struct AmihudIlliquidity {
    /// EMA smoothing factor: α ∈ (0, 1]
    /// - α = 0.05 → effective window ≈ 20 bars
    /// - α = 0.1 → effective window ≈ 10 bars
    alpha: f64,

    /// Exponentially weighted average of illiquidity
    ema_illiq: f64,

    /// Previous price for return calculation
    prev_price: f64,
}

impl AmihudIlliquidity {
    /// Creates a new Amihud Illiquidity calculator
    ///
    /// ## Arguments
    /// - `alpha`: EMA smoothing factor ∈ (0, 1]
    ///   - Smaller α = more smoothing (longer effective window)
    ///   - Larger α = more responsive (shorter effective window)
    ///   - Recommended: 0.05 (20-bar window) for stable estimates
    ///
    /// ## Panics
    /// Panics if `alpha` ≤ 0 or `alpha` > 1
    pub fn new(alpha: f64) -> Self {
        assert!(
            alpha > 0.0 && alpha <= 1.0,
            "Alpha must be in (0, 1], got: {}",
            alpha
        );

        Self {
            alpha,
            ema_illiq: 0.0,
            prev_price: 0.0,
        }
    }

    /// Creates a new Amihud Illiquidity calculator with default alpha (0.05)
    pub fn default() -> Self {
        Self::new(0.05)
    }

    /// Updates the illiquidity estimate with a new OHLCV bar
    ///
    /// ## Arguments
    /// - `price`: Current close price
    /// - `volume`: Current bar volume
    ///
    /// ## Returns
    /// Updated EMA illiquidity value
    ///
    /// ## Behavior
    /// - First update: Returns 0.0 (no previous price for return calculation)
    /// - Second update: Returns instantaneous illiquidity (initializes EMA)
    /// - Subsequent updates: Returns EMA-smoothed illiquidity
    /// - Zero volume: Returns 0.0 (no measurable illiquidity)
    /// - Zero price: Handled gracefully (returns 0.0)
    ///
    /// ## Performance
    /// - O(1) complexity: 3 multiplications, 2 divisions, 1 absolute value
    /// - Expected latency: 3-8μs
    pub fn update(&mut self, price: f64, volume: f64) -> f64 {
        // Calculate return
        let ret = if self.prev_price > 0.0 {
            (price - self.prev_price) / self.prev_price
        } else {
            // First update: no return yet
            self.prev_price = price;
            return 0.0;
        };

        // Calculate instantaneous illiquidity
        let dollar_volume = price * volume;
        let instant_illiq = if dollar_volume > 0.0 {
            ret.abs() / dollar_volume
        } else {
            // Zero volume: no measurable illiquidity
            0.0
        };

        // Update EMA: if this is the first real measurement (ema_illiq == 0.0),
        // initialize with instant_illiq. Otherwise, apply EMA smoothing.
        self.ema_illiq = if self.ema_illiq == 0.0 {
            instant_illiq
        } else {
            self.alpha * instant_illiq + (1.0 - self.alpha) * self.ema_illiq
        };

        // Update state
        self.prev_price = price;

        self.ema_illiq
    }

    /// Returns the current illiquidity value (alias for compute for compatibility)
    pub const fn compute(&self) -> f64 {
        self.ema_illiq
    }

    /// Returns the EMA smoothing factor
    pub const fn alpha(&self) -> f64 {
        self.alpha
    }

    /// Returns the current EMA illiquidity value
    pub const fn ema_illiquidity(&self) -> f64 {
        self.ema_illiq
    }

    /// Returns the previous price used for return calculation
    pub const fn prev_price(&self) -> f64 {
        self.prev_price
    }
}

impl MicrostructureFeatures for AmihudIlliquidity {
    fn feature_name(&self) -> &'static str {
        "amihud_illiquidity"
    }

    fn value(&self) -> f64 {
        self.ema_illiq
    }

    fn get_normalized(&self) -> f64 {
        // Amihud is unbounded and highly skewed: typical range 1e-9 to 1e-5
        // Apply log-transform + clipping for ML models

        if self.ema_illiq <= 0.0 {
            return -5.0; // Map zero/negative to minimum
        }

        // Scale to [ln(0.01), ln(1000)] ≈ [-4.6, 6.9]
        let log_illiq = (self.ema_illiq * 1e8).ln();

        // Clip outliers to [-5, 5]
        let clamped = log_illiq.clamp(-5.0, 5.0);

        // Map to [-1, 1]
        clamped / 5.0
    }

    fn reset(&mut self) {
        self.ema_illiq = 0.0;
        self.prev_price = 0.0;
    }
}

// ============================================================================
// Roll Measure (Agent A9)
// ============================================================================

/// Roll Measure: Estimates effective bid-ask spread from serial covariance
///
/// ## Formula
/// ```text
/// Roll Spread = 2 * sqrt(-cov(Δp_t, Δp_{t-1}))
/// ```
///
/// Where:
/// - `Δp_t` = `p_t` - p_{t-1} (price change at time t)
/// - `cov()` = covariance between consecutive price changes
///
/// ## Intuition
/// Bid-ask bounce creates negative serial correlation in transaction prices.
/// Roll (1984) showed this covariance relates to the effective spread.
///
/// ## Implementation Details
/// - Uses rolling window of 20 price changes for stability
/// - Handles negative covariance (take sqrt of absolute value)
/// - O(1) amortized update using `VecDeque`
/// - Returns 0.0 for insufficient data (<3 prices)
///
/// ## Performance
/// - Update: O(1) amortized (`VecDeque` push/pop)
/// - Compute: O(n) where n=20 (window size)
/// - Memory: 72 bytes (8B per price × 20 + overhead)
/// - Latency: <2μs per update+compute
///
/// ## References
/// - Roll (1984): "A Simple Implicit Measure of the Effective Bid-Ask Spread"
#[derive(Debug, Clone)]
pub struct RollMeasure {
    /// Rolling window of prices (max 21 for 20 price changes)
    prices: std::collections::VecDeque<f64>,
    /// Window size for covariance calculation
    window_size: usize,
}

impl RollMeasure {
    /// Create new Roll Measure estimator
    pub fn new() -> Self {
        Self {
            prices: std::collections::VecDeque::with_capacity(21),
            window_size: 20,
        }
    }

    /// Update with new price observation
    ///
    /// ## Arguments
    /// - `price`: Transaction price (e.g., close price)
    ///
    /// ## Performance
    /// - O(1) amortized (`VecDeque` push/pop)
    /// - <500ns typical latency
    pub fn update(&mut self, price: f64) {
        if !price.is_finite() {
            return; // Skip invalid prices
        }

        self.prices.push_back(price);

        // Keep window_size + 1 prices (for window_size price changes)
        if self.prices.len() > self.window_size + 1 {
            self.prices.pop_front();
        }
    }

    /// Compute Roll spread estimate
    ///
    /// ## Returns
    /// - Effective spread estimate in price units
    /// - Returns 0.0 if insufficient data (<3 prices)
    ///
    /// ## Performance
    /// - O(n) where `n=window_size` (20)
    /// - <2μs typical latency
    pub fn compute(&self) -> f64 {
        // Need at least 3 prices for 2 price changes
        if self.prices.len() < 3 {
            return 0.0;
        }

        // Compute price changes Δp_t = p_t - p_{t-1}
        let price_changes: Vec<f64> = self
            .prices
            .iter()
            .zip(self.prices.iter().skip(1))
            .map(|(prev, curr)| curr - prev)
            .collect();

        if price_changes.len() < 2 {
            return 0.0;
        }

        // Compute covariance between Δp_t and Δp_{t-1}
        let cov = self.compute_serial_covariance(&price_changes);

        // Roll spread = 2 * sqrt(-cov)
        // Handle negative covariance case: take sqrt of absolute value
        if cov >= 0.0 {
            // Positive covariance (trending) => no bid-ask bounce
            // Return small spread estimate
            return 0.0;
        }

        // Negative covariance (mean-reverting) => bid-ask bounce present
        let spread = 2.0 * (-cov).sqrt();

        // Sanity check: cap at 100 (unrealistic spread)
        spread.min(100.0)
    }

    /// Compute serial covariance: `cov(Δp_t`, Δp_{t-1})
    ///
    /// ## Formula
    /// ```text
    /// cov(X, Y) = E[(X - μ_X)(Y - μ_Y)]
    /// ```
    ///
    /// ## Performance
    /// - O(n) where n = length of `price_changes`
    /// - <1μs for n=20
    fn compute_serial_covariance(&self, price_changes: &[f64]) -> f64 {
        if price_changes.len() < 2 {
            return 0.0;
        }

        let n = price_changes.len() - 1; // Number of overlapping pairs

        // Compute means
        let mean_t: f64 = price_changes.iter().skip(1).sum::<f64>() / n as f64;
        let mean_t_minus_1: f64 = price_changes.iter().take(n).sum::<f64>() / n as f64;

        // Compute covariance
        let mut cov_sum = 0.0;
        for i in 0..n {
            let x = price_changes[i] - mean_t_minus_1;
            let y = price_changes[i + 1] - mean_t;
            cov_sum += x * y;
        }

        cov_sum / n as f64
    }
}

impl Default for RollMeasure {
    fn default() -> Self {
        Self::new()
    }
}

// ============================================================================
// Normalization Helper Functions (for extraction.rs integration)
// ============================================================================

/// Normalize Roll spread for ML training
///
/// Converts absolute spread (in price units) to normalized value [0, 1]
///
/// ## Arguments
/// - `spread`: Raw Roll spread estimate (typically 0.01 - 10.0 for ES.FUT)
/// - `max_spread`: Maximum expected spread for normalization (default: 10.0)
///
/// ## Returns
/// Normalized value in [0, 1] where:
/// - 0.0 = No spread (perfect liquidity)
/// - 1.0 = Maximum spread (illiquid market)
pub fn normalize_roll_spread(spread: f64, max_spread: f64) -> f64 {
    if spread <= 0.0 || !spread.is_finite() {
        return 0.0;
    }
    (spread / max_spread).clamp(0.0, 1.0)
}

/// Normalize Amihud illiquidity for ML training
///
/// Converts absolute illiquidity to normalized value [0, 1]
///
/// ## Arguments
/// - `illiquidity`: Raw Amihud illiquidity (typically 1e-9 to 1e-5)
/// - `max_illiq`: Maximum expected illiquidity for normalization (default: 1e-5)
///
/// ## Returns
/// Normalized value in [0, 1] where:
/// - 0.0 = Perfect liquidity
/// - 1.0 = Maximum illiquidity
pub fn normalize_amihud_illiquidity(illiquidity: f64, max_illiq: f64) -> f64 {
    if illiquidity <= 0.0 || !illiquidity.is_finite() {
        return 0.0;
    }
    (illiquidity / max_illiq).clamp(0.0, 1.0)
}

// ============================================================================
// Corwin-Schultz Spread Estimator (Agent A10)
// ============================================================================

/// Corwin-Schultz Spread: High-low volatility decomposition estimator
///
/// ## Formula
/// ```text
/// Spread = 2 * (e^α - 1) / (1 + e^α)
/// α = [(√(2β₁) + √(2β₂)) - √γ] / (3 - 2√2)
/// β = [ln(H_t/L_t)]²  (single-period high-low variance)
/// γ = [ln(max(H_t,H_{t-1}) / min(L_t,L_{t-1}))]²  (two-period variance)
/// ```
///
/// ## Intuition
/// The high-low range contains both fundamental volatility and bid-ask spread.
/// By comparing single-period and two-period ranges, we decompose the spread
/// component from the volatility component.
///
/// ## Performance
/// - Latency: <15μs per update
/// - Memory: 72 bytes
/// - Data: OHLC only (no Level-2 required)
///
/// ## References
/// - Corwin & Schultz (2012): "A Simple Way to Estimate Bid-Ask Spreads from Daily High and Low Prices"
#[derive(Debug, Clone)]
pub struct CorwinSchultzSpread {
    /// Rolling window of (high, low, close) tuples
    bars: std::collections::VecDeque<(f64, f64, f64)>,
    /// Window size for averaging spread estimates
    window_size: usize,
}

impl CorwinSchultzSpread {
    /// Create new Corwin-Schultz spread estimator
    pub fn new() -> Self {
        Self {
            bars: std::collections::VecDeque::with_capacity(21),
            window_size: 20,
        }
    }

    /// Update with new OHLC bar
    pub fn update(&mut self, high: f64, low: f64, close: f64) {
        if !high.is_finite() || !low.is_finite() || !close.is_finite() {
            return;
        }
        if high < low || close < low || close > high || high <= 0.0 || low <= 0.0 {
            return;
        }

        self.bars.push_back((high, low, close));
        if self.bars.len() > self.window_size + 1 {
            self.bars.pop_front();
        }
    }

    /// Compute Corwin-Schultz spread estimate
    pub fn compute(&self) -> f64 {
        if self.bars.len() < 2 {
            return 0.0;
        }

        let mut spread_estimates = Vec::with_capacity(self.bars.len() - 1);

        for i in 0..self.bars.len() - 1 {
            let (high_prev, low_prev, _) = self.bars[i];
            let (high_curr, low_curr, _) = self.bars[i + 1];

            if let Some(spread) =
                self.compute_two_bar_spread(high_prev, low_prev, high_curr, low_curr)
            {
                spread_estimates.push(spread);
            }
        }

        if spread_estimates.is_empty() {
            0.0
        } else {
            let avg = spread_estimates.iter().sum::<f64>() / spread_estimates.len() as f64;
            avg.min(0.5)
        }
    }

    fn compute_two_bar_spread(
        &self,
        high_prev: f64,
        low_prev: f64,
        high_curr: f64,
        low_curr: f64,
    ) -> Option<f64> {
        if high_prev <= low_prev || high_curr <= low_curr {
            return None;
        }

        let beta_prev = (high_prev / low_prev).ln().powi(2);
        let beta_curr = (high_curr / low_curr).ln().powi(2);
        let max_high = high_prev.max(high_curr);
        let min_low = low_prev.min(low_curr);
        let gamma = (max_high / min_low).ln().powi(2);

        if !beta_prev.is_finite() || !beta_curr.is_finite() || !gamma.is_finite() {
            return None;
        }

        let sqrt_2 = 2.0_f64.sqrt();
        let denominator = 3.0 - 2.0 * sqrt_2;
        let numerator = (sqrt_2 * beta_prev).sqrt() + (sqrt_2 * beta_curr).sqrt() - gamma.sqrt();
        let alpha = numerator / denominator;

        if !alpha.is_finite() || alpha < 0.0 {
            return None;
        }

        let e_alpha = alpha.exp();
        let spread = 2.0 * (e_alpha - 1.0) / (1.0 + e_alpha);

        (spread.is_finite() && spread >= 0.0).then_some(spread)
    }
}

impl Default for CorwinSchultzSpread {
    fn default() -> Self {
        Self::new()
    }
}

/// Normalize Corwin-Schultz spread to [0, 1] for ML features
pub fn normalize_corwin_schultz_spread(spread: f64, max_spread: f64) -> f64 {
    if !spread.is_finite() || spread < 0.0 {
        return 0.0;
    }
    (spread / max_spread).min(1.0)
}

// ============================================================================
// Unit Tests
// ============================================================================

#[cfg(test)]
#[allow(clippy::manual_range_contains)]
mod tests {
    use super::*;

    #[test]
    fn test_amihud_initialization() {
        let amihud = AmihudIlliquidity::new(0.05);
        assert_eq!(amihud.alpha(), 0.05);
        assert_eq!(amihud.ema_illiquidity(), 0.0);
        assert_eq!(amihud.prev_price(), 0.0);
    }

    #[test]
    #[should_panic(expected = "Alpha must be in (0, 1]")]
    fn test_amihud_invalid_alpha_zero() {
        let _ = AmihudIlliquidity::new(0.0);
    }

    #[test]
    #[should_panic(expected = "Alpha must be in (0, 1]")]
    fn test_amihud_invalid_alpha_negative() {
        let _ = AmihudIlliquidity::new(-0.1);
    }

    #[test]
    #[should_panic(expected = "Alpha must be in (0, 1]")]
    fn test_amihud_invalid_alpha_too_large() {
        let _ = AmihudIlliquidity::new(1.5);
    }

    #[test]
    fn test_amihud_first_update() {
        let mut amihud = AmihudIlliquidity::new(0.05);
        let illiq = amihud.update(100.0, 10000.0);

        assert_eq!(illiq, 0.0, "First update should return 0.0");
        assert_eq!(amihud.prev_price(), 100.0);
        assert_eq!(amihud.ema_illiquidity(), 0.0);
    }

    #[test]
    fn test_amihud_high_volume_low_illiquidity() {
        let mut amihud = AmihudIlliquidity::new(0.05);

        amihud.update(100.0, 100000.0);
        let illiq = amihud.update(101.0, 100000.0);

        // Illiquidity = |0.01| / (101 * 100000) ≈ 9.9e-10
        let expected = 0.01 / (101.0 * 100000.0);
        let tolerance = expected * 0.01;

        assert!(
            (illiq - expected).abs() < tolerance,
            "Expected: {}, Got: {}",
            expected,
            illiq
        );
    }

    #[test]
    fn test_amihud_low_volume_high_illiquidity() {
        let mut amihud = AmihudIlliquidity::new(0.05);

        amihud.update(100.0, 1000.0);
        let illiq = amihud.update(101.0, 100.0);

        // Illiquidity = |0.01| / (101 * 100) ≈ 9.9e-7
        let expected = 0.01 / (101.0 * 100.0);
        let tolerance = expected * 0.01;

        assert!(
            (illiq - expected).abs() < tolerance,
            "Expected: {}, Got: {}",
            expected,
            illiq
        );
        assert!(illiq > 1e-8, "Low volume should yield high illiquidity");
    }

    #[test]
    fn test_amihud_zero_volume() {
        let mut amihud = AmihudIlliquidity::new(0.05);

        amihud.update(100.0, 1000.0);
        let illiq = amihud.update(101.0, 0.0);

        assert_eq!(illiq, 0.0, "Zero volume should yield zero illiquidity");
        assert_eq!(amihud.prev_price(), 101.0);
    }

    #[test]
    fn test_amihud_zero_price() {
        let mut amihud = AmihudIlliquidity::new(0.05);

        amihud.update(100.0, 1000.0);
        let illiq = amihud.update(0.0, 1000.0);

        // Should handle gracefully
        assert!(illiq.is_finite());
        assert_eq!(amihud.prev_price(), 0.0);
    }

    #[test]
    fn test_amihud_negative_return() {
        let mut amihud = AmihudIlliquidity::new(0.05);

        amihud.update(100.0, 10000.0);
        let illiq = amihud.update(99.0, 10000.0);

        // Illiquidity = |-0.01| / (99 * 10000) ≈ 1.01e-8
        let expected = 0.01 / (99.0 * 10000.0);
        let tolerance = expected * 0.01;

        assert!(
            (illiq - expected).abs() < tolerance,
            "Negative return should use abs value"
        );
    }

    #[test]
    fn test_amihud_ema_smoothing() {
        let mut amihud = AmihudIlliquidity::new(0.05);

        amihud.update(100.0, 10000.0);
        let illiq1 = amihud.update(101.0, 10000.0);
        let illiq2 = amihud.update(102.0, 10000.0);

        // EMA should smooth values
        assert_ne!(illiq1, illiq2);
        assert!(amihud.ema_illiquidity() > 0.0);
    }

    #[test]
    fn test_amihud_trait_methods() {
        let mut amihud = AmihudIlliquidity::new(0.05);

        assert_eq!(amihud.feature_name(), "amihud_illiquidity");

        amihud.update(100.0, 10000.0);
        amihud.update(101.0, 10000.0);

        let value = amihud.value();
        let normalized = amihud.get_normalized();

        assert!(value > 0.0);
        assert!(normalized.is_finite());
        assert!(normalized >= -5.0 && normalized <= 5.0);
    }

    #[test]
    fn test_amihud_reset() {
        let mut amihud = AmihudIlliquidity::new(0.05);

        amihud.update(100.0, 10000.0);
        amihud.update(101.0, 10000.0);

        assert!(amihud.ema_illiquidity() > 0.0);
        assert!(amihud.prev_price() > 0.0);

        amihud.reset();

        assert_eq!(amihud.ema_illiquidity(), 0.0);
        assert_eq!(amihud.prev_price(), 0.0);
    }

    #[test]
    fn test_amihud_memory_size() {
        use std::mem::size_of;

        let size = size_of::<AmihudIlliquidity>();
        assert!(size <= 72, "Memory {} bytes exceeds 72-byte target", size);
    }

    #[test]
    fn test_amihud_latency_benchmark() {
        use std::time::Instant;

        let mut amihud = AmihudIlliquidity::new(0.05);
        amihud.update(100.0, 10000.0);

        let iterations = 10000;
        let start = Instant::now();

        for i in 0..iterations {
            let price = 100.0 + (i as f64 * 0.01);
            amihud.update(price, 10000.0);
        }

        let elapsed = start.elapsed();
        let avg_latency_us = elapsed.as_micros() as f64 / iterations as f64;

        assert!(
            avg_latency_us < 8.0,
            "Average latency {:.2}μs exceeds 8μs target",
            avg_latency_us
        );
    }

    #[test]
    fn test_amihud_numerical_stability() {
        let mut amihud = AmihudIlliquidity::new(0.05);

        // Test extreme values
        let test_cases = vec![(1e-6, 1e-6), (1e6, 1e6), (100.0, 1e-6), (1e-6, 1e6)];

        amihud.update(100.0, 10000.0);

        for (price, volume) in test_cases {
            let illiq = amihud.update(price, volume);
            assert!(illiq.is_finite(), "Illiquidity must be finite");
            assert!(illiq >= 0.0, "Illiquidity must be non-negative");
        }
    }

    #[test]
    fn test_normalization_functions() {
        // Test Roll spread normalization
        assert_eq!(normalize_roll_spread(0.0, 10.0), 0.0);
        assert_eq!(normalize_roll_spread(5.0, 10.0), 0.5);
        assert_eq!(normalize_roll_spread(10.0, 10.0), 1.0);
        assert_eq!(normalize_roll_spread(20.0, 10.0), 1.0); // Clipped

        // Test Amihud illiquidity normalization
        assert_eq!(normalize_amihud_illiquidity(0.0, 1e-5), 0.0);
        assert_eq!(normalize_amihud_illiquidity(5e-6, 1e-5), 0.5);
        assert_eq!(normalize_amihud_illiquidity(1e-5, 1e-5), 1.0);
        assert_eq!(normalize_amihud_illiquidity(2e-5, 1e-5), 1.0); // Clipped
    }
}