sublinear 0.3.3

//! Identity Continuity Tracking for Temporal Consciousness
//!
//! This module tracks and preserves identity continuity across temporal boundaries
//! to ensure consciousness coherence. It monitors identity state, detects breaks
//! in continuity, and provides mechanisms for identity preservation.

use super::{TemporalError, TemporalResult, TscTimestamp};
use std::collections::{HashMap, VecDeque};

/// Metrics for identity continuity analysis
#[derive(Debug, Clone, Default)]
pub struct ContinuityMetrics {
    pub continuity_score: f64,
    pub identity_stability: f64,
    pub continuity_breaks: u64,
    pub average_gap_duration_ns: f64,
    pub max_gap_duration_ns: u64,
    pub identity_coherence: f64,
    pub temporal_consistency: f64,
    pub preservation_efficiency: f64,
}

/// Identity state snapshot at a specific time
#[derive(Debug, Clone)]
struct IdentitySnapshot {
    timestamp: TscTimestamp,
    state_hash: u64,
    feature_vector: Vec<f64>,
    coherence_score: f64,
    stability_metric: f64,
    memory_fingerprint: Vec<u8>,
}

impl IdentitySnapshot {
    /// Create a new identity snapshot
    fn new(timestamp: TscTimestamp, state: &[u8]) -> Self {
        let feature_vector = Self::extract_features(state);
        let state_hash = Self::compute_hash(state);
        let coherence_score = Self::calculate_coherence(&feature_vector);

        Self {
            timestamp,
            state_hash,
            feature_vector,
            coherence_score,
            stability_metric: 1.0, // Will be updated during tracking
            memory_fingerprint: state.to_vec(),
        }
    }

    /// Extract feature vector from state data
    fn extract_features(state: &[u8]) -> Vec<f64> {
        if state.is_empty() {
            return vec![0.0; 16]; // Default feature size
        }

        let mut features = Vec::with_capacity(16);

        // Statistical features
        let mean = state.iter().map(|&x| x as f64).sum::<f64>() / state.len() as f64;
        features.push(mean);

        let variance = state
            .iter()
            .map(|&x| (x as f64 - mean).powi(2))
            .sum::<f64>()
            / state.len() as f64;
        features.push(variance.sqrt());

        // Entropy-like measure
        let mut byte_counts = [0u32; 256];
        for &byte in state {
            byte_counts[byte as usize] += 1;
        }

        let entropy = byte_counts
            .iter()
            .filter(|&&count| count > 0)
            .map(|&count| {
                let p = count as f64 / state.len() as f64;
                -p * p.ln()
            })
            .sum::<f64>();
        features.push(entropy);

        // Spectral features (simple FFT-like)
        for i in 0..8 {
            let freq_component = state
                .iter()
                .enumerate()
                .map(|(j, &x)| {
                    let phase =
                        2.0 * std::f64::consts::PI * (i + 1) as f64 * j as f64 / state.len() as f64;
                    x as f64 * phase.cos()
                })
                .sum::<f64>();
            features.push(freq_component / state.len() as f64);
        }

        // Compression ratio estimate
        let complexity = Self::estimate_complexity(state);
        features.push(complexity);

        // Pattern density
        let pattern_density = Self::calculate_pattern_density(state);
        features.push(pattern_density);

        // Autocorrelation at lag 1
        let autocorr = Self::calculate_autocorrelation(state, 1);
        features.push(autocorr);

        // Trend measure
        let trend = Self::calculate_trend(state);
        features.push(trend);

        // Normalize features
        for feature in &mut features {
            *feature = feature.tanh(); // Bound between -1 and 1
        }

        features
    }

    /// Compute hash of state data
    fn compute_hash(state: &[u8]) -> u64 {
        use std::collections::hash_map::DefaultHasher;
        use std::hash::{Hash, Hasher};

        let mut hasher = DefaultHasher::new();
        state.hash(&mut hasher);
        hasher.finish()
    }

    /// Calculate coherence score from feature vector
    fn calculate_coherence(features: &[f64]) -> f64 {
        if features.is_empty() {
            return 0.0;
        }

        // Coherence based on feature consistency
        let mean = features.iter().sum::<f64>() / features.len() as f64;
        let variance =
            features.iter().map(|x| (x - mean).powi(2)).sum::<f64>() / features.len() as f64;

        // Lower variance = higher coherence
        (-variance).exp().min(1.0)
    }

    /// Calculate similarity with another snapshot.
    ///
    /// Pure cosine similarity is scale-invariant and the feature
    /// extractor tanh-normalises every component, so a small state and
    /// a 10× larger one with the same shape both saturate near
    /// `(1, 1, 1, …)` and look identical (similarity ≈ 1.0). That
    /// silently swallows the "very different states" case the
    /// continuity tracker is supposed to flag (test_continuity_break_detection).
    ///
    /// Combine cosine with a mean-L1 dissimilarity so component-wise
    /// differences register even after saturation, then weight to keep
    /// behavior similar for genuinely-similar states.
    fn calculate_similarity(&self, other: &IdentitySnapshot) -> f64 {
        if self.feature_vector.len() != other.feature_vector.len() || self.feature_vector.is_empty()
        {
            return 0.0;
        }

        let dot_product: f64 = self
            .feature_vector
            .iter()
            .zip(other.feature_vector.iter())
            .map(|(a, b)| a * b)
            .sum();
        let magnitude_self: f64 = self
            .feature_vector
            .iter()
            .map(|x| x * x)
            .sum::<f64>()
            .sqrt();
        let magnitude_other: f64 = other
            .feature_vector
            .iter()
            .map(|x| x * x)
            .sum::<f64>()
            .sqrt();

        let cosine = if magnitude_self > 0.0 && magnitude_other > 0.0 {
            dot_product / (magnitude_self * magnitude_other)
        } else {
            0.0
        };

        // Mean L1 + Chebyshev (max-per-component) over tanh-bounded
        // features. Mean L1 alone isn't strong enough under tanh
        // saturation: when one state is a 10× rescale of another, most
        // features look identical near ±1 and the mean stays small.
        // Chebyshev fires on the *worst* single component, which gives
        // a clean break signal when any spectral/statistical feature
        // genuinely differs. Both metrics are bounded to [0, 1].
        let mut mean_l1 = 0.0;
        let mut max_diff: f64 = 0.0;
        for (a, b) in self.feature_vector.iter().zip(other.feature_vector.iter()) {
            let d = (a - b).abs();
            mean_l1 += d;
            if d > max_diff {
                max_diff = d;
            }
        }
        mean_l1 /= self.feature_vector.len() as f64;
        let l1_similarity = (1.0 - mean_l1 / 2.0).max(0.0);
        // Features are tanh-bounded in [-1, 1] so max_diff is in [0, 2].
        let chebyshev_similarity = (1.0 - max_diff / 2.0).max(0.0);

        // Weighted blend: cosine 30%, mean-L1 30%, Chebyshev 40%. Identical
        // states score 1.0; substantially different states (one large
        // Chebyshev diff) drop well below 0.9 without breaking the
        // "slightly different" test_continuity_tracker case.
        0.3 * cosine + 0.3 * l1_similarity + 0.4 * chebyshev_similarity
    }

    // Helper methods for feature extraction

    fn estimate_complexity(data: &[u8]) -> f64 {
        if data.len() < 2 {
            return 0.0;
        }

        // Simple complexity estimate based on run lengths
        let mut runs = 0;
        let mut current_byte = data[0];

        for &byte in &data[1..] {
            if byte != current_byte {
                runs += 1;
                current_byte = byte;
            }
        }

        runs as f64 / data.len() as f64
    }

    fn calculate_pattern_density(data: &[u8]) -> f64 {
        if data.len() < 4 {
            return 0.0;
        }

        let mut patterns = HashMap::new();

        // Count 2-byte patterns
        for window in data.windows(2) {
            *patterns.entry((window[0], window[1])).or_insert(0) += 1;
        }

        patterns.len() as f64 / (data.len() - 1) as f64
    }

    fn calculate_autocorrelation(data: &[u8], lag: usize) -> f64 {
        if data.len() <= lag {
            return 0.0;
        }

        let mean = data.iter().map(|&x| x as f64).sum::<f64>() / data.len() as f64;

        let numerator: f64 = data
            .iter()
            .take(data.len() - lag)
            .zip(data.iter().skip(lag))
            .map(|(&x, &y)| (x as f64 - mean) * (y as f64 - mean))
            .sum();

        let denominator: f64 = data.iter().map(|&x| (x as f64 - mean).powi(2)).sum();

        if denominator > 0.0 {
            numerator / denominator
        } else {
            0.0
        }
    }

    fn calculate_trend(data: &[u8]) -> f64 {
        if data.len() < 2 {
            return 0.0;
        }

        let n = data.len() as f64;
        let sum_x = (0..data.len()).sum::<usize>() as f64;
        let sum_y = data.iter().map(|&x| x as f64).sum::<f64>();
        let sum_xy = data
            .iter()
            .enumerate()
            .map(|(i, &y)| i as f64 * y as f64)
            .sum::<f64>();
        let sum_x2 = (0..data.len()).map(|i| (i as f64).powi(2)).sum::<f64>();

        let denominator = n * sum_x2 - sum_x * sum_x;
        if denominator > 0.0 {
            (n * sum_xy - sum_x * sum_y) / denominator
        } else {
            0.0
        }
    }
}

/// Identity continuity tracker
pub struct IdentityContinuityTracker {
    snapshots: VecDeque<IdentitySnapshot>,
    metrics: ContinuityMetrics,
    max_snapshots: usize,
    continuity_threshold: f64,
    gap_tolerance_ns: u64,
    identity_baseline: Option<IdentitySnapshot>,
    last_validation_time: Option<TscTimestamp>,
}

impl IdentityContinuityTracker {
    /// Create a new identity continuity tracker
    pub fn new() -> Self {
        Self {
            snapshots: VecDeque::new(),
            metrics: ContinuityMetrics::default(),
            max_snapshots: 1000,
            continuity_threshold: 0.7,   // 70% similarity threshold
            gap_tolerance_ns: 1_000_000, // 1ms tolerance
            identity_baseline: None,
            last_validation_time: None,
        }
    }

    /// Track identity continuity at a specific timestamp
    pub fn track_continuity(
        &mut self,
        timestamp: TscTimestamp,
        state: &[u8],
    ) -> TemporalResult<()> {
        let snapshot = IdentitySnapshot::new(timestamp, state);

        // Establish baseline if this is the first snapshot
        if self.identity_baseline.is_none() {
            self.identity_baseline = Some(snapshot.clone());
        }

        // Check for continuity breaks
        let continuity_break_info = if let Some(prev_snapshot) = self.snapshots.back() {
            Some((snapshot.clone(), prev_snapshot.clone()))
        } else {
            None
        };

        if let Some((current, previous)) = continuity_break_info {
            self.check_continuity_break(&current, &previous)?;
        }

        // Update stability metrics
        self.update_stability_metrics(&snapshot);

        // Store the snapshot
        self.store_snapshot(snapshot);

        // Update overall metrics
        self.update_metrics();

        self.last_validation_time = Some(timestamp);

        Ok(())
    }

    /// Validate current identity continuity.
    ///
    /// Returns Ok with too little history to judge (fewer than 2 snapshots) —
    /// otherwise the very first scheduler tick that runs an
    /// `IdentityPreservation { continuity_check: true }` task would fail
    /// because `continuity_score` is still at its `0.0` initial value,
    /// below any sensible threshold.
    pub fn validate_continuity(&self) -> TemporalResult<()> {
        if self.snapshots.len() < 2 {
            return Ok(());
        }
        if self.metrics.continuity_score < self.continuity_threshold {
            return Err(TemporalError::IdentityContinuityBreak {
                gap_ns: self.metrics.max_gap_duration_ns,
            });
        }

        Ok(())
    }

    /// Get current continuity metrics
    pub fn get_metrics(&self) -> TemporalResult<ContinuityMetrics> {
        Ok(self.metrics.clone())
    }

    /// Get identity stability score
    pub fn get_identity_stability(&self) -> f64 {
        self.metrics.identity_stability
    }

    /// Get continuity score
    pub fn get_continuity_score(&self) -> f64 {
        self.metrics.continuity_score
    }

    /// Reset tracking state
    pub fn reset(&mut self) {
        self.snapshots.clear();
        self.metrics = ContinuityMetrics::default();
        self.identity_baseline = None;
        self.last_validation_time = None;
    }

    /// Set continuity threshold
    pub fn set_continuity_threshold(&mut self, threshold: f64) {
        self.continuity_threshold = threshold.clamp(0.0, 1.0);
    }

    /// Get recent identity trajectory
    pub fn get_identity_trajectory(&self, window_size: usize) -> Vec<f64> {
        self.snapshots
            .iter()
            .rev()
            .take(window_size)
            .map(|s| s.coherence_score)
            .collect()
    }

    /// Calculate identity drift over time
    pub fn calculate_identity_drift(&self) -> f64 {
        if let Some(baseline) = &self.identity_baseline {
            if let Some(current) = self.snapshots.back() {
                return 1.0 - baseline.calculate_similarity(current);
            }
        }
        0.0
    }

    // Private helper methods

    fn check_continuity_break(
        &mut self,
        current: &IdentitySnapshot,
        previous: &IdentitySnapshot,
    ) -> TemporalResult<()> {
        // Check temporal gap
        let gap_ns = current
            .timestamp
            .nanos_since(previous.timestamp, 3_000_000_000);
        if gap_ns > self.gap_tolerance_ns {
            self.metrics.continuity_breaks += 1;
            self.metrics.max_gap_duration_ns = self.metrics.max_gap_duration_ns.max(gap_ns);
        }

        // Check similarity
        let similarity = current.calculate_similarity(previous);
        if similarity < self.continuity_threshold {
            self.metrics.continuity_breaks += 1;
        }

        Ok(())
    }

    fn update_stability_metrics(&mut self, snapshot: &IdentitySnapshot) {
        if self.snapshots.len() < 2 {
            return;
        }

        // Calculate stability based on recent snapshots
        let recent_snapshots: Vec<_> = self.snapshots.iter().rev().take(10).collect();

        if recent_snapshots.len() >= 2 {
            let mut similarities = Vec::new();

            for i in 0..recent_snapshots.len() - 1 {
                let sim = recent_snapshots[i].calculate_similarity(recent_snapshots[i + 1]);
                similarities.push(sim);
            }

            // Current similarity with latest snapshot
            let current_sim = snapshot.calculate_similarity(recent_snapshots[0]);
            similarities.push(current_sim);

            // Stability is average similarity over recent window
            let avg_similarity = similarities.iter().sum::<f64>() / similarities.len() as f64;

            // Update stability metric with exponential moving average
            let alpha = 0.1;
            self.metrics.identity_stability =
                (1.0 - alpha) * self.metrics.identity_stability + alpha * avg_similarity;
        }
    }

    fn store_snapshot(&mut self, snapshot: IdentitySnapshot) {
        self.snapshots.push_back(snapshot);

        // Keep history bounded
        while self.snapshots.len() > self.max_snapshots {
            self.snapshots.pop_front();
        }
    }

    fn update_metrics(&mut self) {
        if self.snapshots.len() < 2 {
            return;
        }

        // Calculate continuity score
        let mut total_similarity = 0.0;
        let mut similarity_count = 0;

        for window in self.snapshots.iter().collect::<Vec<_>>().windows(2) {
            let sim = window[1].calculate_similarity(window[0]);
            total_similarity += sim;
            similarity_count += 1;
        }

        if similarity_count > 0 {
            self.metrics.continuity_score = total_similarity / similarity_count as f64;
        }

        // Calculate coherence
        let coherence_scores: Vec<f64> = self.snapshots.iter().map(|s| s.coherence_score).collect();

        if !coherence_scores.is_empty() {
            self.metrics.identity_coherence =
                coherence_scores.iter().sum::<f64>() / coherence_scores.len() as f64;
        }

        // Calculate temporal consistency
        self.calculate_temporal_consistency();

        // Calculate preservation efficiency
        self.calculate_preservation_efficiency();

        // Update gap metrics
        self.update_gap_metrics();
    }

    fn calculate_temporal_consistency(&mut self) {
        if self.snapshots.len() < 3 {
            return;
        }

        // Measure how consistently identity features change over time
        let mut consistency_scores = Vec::new();

        for i in 2..self.snapshots.len() {
            let s1 = &self.snapshots[i - 2];
            let s2 = &self.snapshots[i - 1];
            let s3 = &self.snapshots[i];

            // Calculate velocity vectors
            let vel1: Vec<f64> = s2
                .feature_vector
                .iter()
                .zip(s1.feature_vector.iter())
                .map(|(a, b)| a - b)
                .collect();

            let vel2: Vec<f64> = s3
                .feature_vector
                .iter()
                .zip(s2.feature_vector.iter())
                .map(|(a, b)| a - b)
                .collect();

            // Calculate consistency as cosine similarity of velocity vectors
            let dot_product: f64 = vel1.iter().zip(vel2.iter()).map(|(a, b)| a * b).sum();
            let mag1: f64 = vel1.iter().map(|x| x * x).sum::<f64>().sqrt();
            let mag2: f64 = vel2.iter().map(|x| x * x).sum::<f64>().sqrt();

            if mag1 > 0.0 && mag2 > 0.0 {
                consistency_scores.push(dot_product / (mag1 * mag2));
            }
        }

        if !consistency_scores.is_empty() {
            self.metrics.temporal_consistency =
                consistency_scores.iter().sum::<f64>() / consistency_scores.len() as f64;
        }
    }

    fn calculate_preservation_efficiency(&mut self) {
        if let Some(baseline) = &self.identity_baseline {
            if let Some(current) = self.snapshots.back() {
                // Efficiency based on how well identity is preserved relative to baseline
                let similarity_to_baseline = current.calculate_similarity(baseline);
                let time_factor = 1.0 / (1.0 + self.snapshots.len() as f64 / 1000.0); // Decay over time

                self.metrics.preservation_efficiency = similarity_to_baseline * time_factor;
            }
        }
    }

    fn update_gap_metrics(&mut self) {
        if self.snapshots.len() < 2 {
            return;
        }

        let mut gaps = Vec::new();

        for window in self.snapshots.iter().collect::<Vec<_>>().windows(2) {
            let gap_ns = window[1]
                .timestamp
                .nanos_since(window[0].timestamp, 3_000_000_000);
            gaps.push(gap_ns);
        }

        if !gaps.is_empty() {
            self.metrics.average_gap_duration_ns =
                gaps.iter().sum::<u64>() as f64 / gaps.len() as f64;
            self.metrics.max_gap_duration_ns = *gaps.iter().max().unwrap_or(&0);
        }
    }
}

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

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

    #[test]
    fn test_identity_snapshot_creation() {
        let timestamp = TscTimestamp::now();
        let state = vec![1, 2, 3, 4, 5];

        let snapshot = IdentitySnapshot::new(timestamp, &state);
        assert_eq!(snapshot.timestamp, timestamp);
        assert!(!snapshot.feature_vector.is_empty());
        assert!(snapshot.coherence_score >= 0.0 && snapshot.coherence_score <= 1.0);
    }

    #[test]
    fn test_feature_extraction() {
        let state = vec![1, 2, 3, 4, 5, 4, 3, 2, 1];
        let features = IdentitySnapshot::extract_features(&state);

        assert!(!features.is_empty());
        // Features should be normalized
        for &feature in &features {
            assert!(feature >= -1.0 && feature <= 1.0);
        }
    }

    #[test]
    fn test_similarity_calculation() {
        let timestamp = TscTimestamp::now();
        let state1 = vec![1, 2, 3, 4, 5];
        let state2 = vec![1, 2, 3, 4, 5]; // Identical
        let state3 = vec![5, 4, 3, 2, 1]; // Different

        let snapshot1 = IdentitySnapshot::new(timestamp, &state1);
        let snapshot2 = IdentitySnapshot::new(timestamp, &state2);
        let snapshot3 = IdentitySnapshot::new(timestamp, &state3);

        let sim12 = snapshot1.calculate_similarity(&snapshot2);
        let sim13 = snapshot1.calculate_similarity(&snapshot3);

        assert!(sim12 > sim13); // Identical states should be more similar
        assert!(sim12 >= 0.0 && sim12 <= 1.0);
        assert!(sim13 >= 0.0 && sim13 <= 1.0);
    }

    #[test]
    fn test_continuity_tracker() {
        let mut tracker = IdentityContinuityTracker::new();
        let timestamp = TscTimestamp::now();
        let state = vec![1, 2, 3, 4, 5];

        tracker.track_continuity(timestamp, &state).unwrap();
        assert_eq!(tracker.snapshots.len(), 1);

        // Track another similar state
        let timestamp2 = timestamp.add_nanos(1000, 3_000_000_000);
        let state2 = vec![1, 2, 3, 4, 6]; // Slightly different

        tracker.track_continuity(timestamp2, &state2).unwrap();
        assert_eq!(tracker.snapshots.len(), 2);

        let metrics = tracker.get_metrics().unwrap();
        assert!(metrics.continuity_score > 0.0);
    }

    #[test]
    fn test_continuity_break_detection() {
        let mut tracker = IdentityContinuityTracker::new();
        tracker.set_continuity_threshold(0.9); // High threshold

        let timestamp = TscTimestamp::now();
        let state1 = vec![1, 2, 3, 4, 5];
        let state2 = vec![10, 20, 30, 40, 50]; // Very different

        tracker.track_continuity(timestamp, &state1).unwrap();

        let timestamp2 = timestamp.add_nanos(1000, 3_000_000_000);
        tracker.track_continuity(timestamp2, &state2).unwrap();

        // Should detect continuity break
        assert!(tracker.metrics.continuity_breaks > 0);
    }

    #[test]
    fn test_identity_drift_calculation() {
        let mut tracker = IdentityContinuityTracker::new();
        let timestamp = TscTimestamp::now();

        // Start with baseline
        let baseline_state = vec![1, 2, 3, 4, 5];
        tracker
            .track_continuity(timestamp, &baseline_state)
            .unwrap();

        // Gradual drift
        for i in 1..=10 {
            let timestamp_i = timestamp.add_nanos(i * 1000, 3_000_000_000);
            let drifted_state = vec![1 + i as u8, 2, 3, 4, 5];
            tracker
                .track_continuity(timestamp_i, &drifted_state)
                .unwrap();
        }

        let drift = tracker.calculate_identity_drift();
        assert!(drift > 0.0); // Should detect some drift
        assert!(drift <= 1.0); // Should be bounded
    }

    #[test]
    fn test_stability_metrics() {
        let mut tracker = IdentityContinuityTracker::new();
        let timestamp = TscTimestamp::now();

        // Track several stable states
        for i in 0..10 {
            let timestamp_i = timestamp.add_nanos(i * 1000, 3_000_000_000);
            let stable_state = vec![1, 2, 3, 4, 5]; // Same state
            tracker
                .track_continuity(timestamp_i, &stable_state)
                .unwrap();
        }

        assert!(tracker.get_identity_stability() > 0.5); // Should be stable
    }
}