<?php

declare(strict_types=1);

namespace App\Framework\Waf\MachineLearning;

use App\Framework\Core\ValueObjects\Duration;
use App\Framework\Core\ValueObjects\Percentage;
use App\Framework\DateTime\Clock;
use App\Framework\MachineLearning\Core\AnomalyDetectorInterface;
use App\Framework\MachineLearning\Core\FeatureExtractorMetadata;
use App\Framework\MachineLearning\ValueObjects\AnomalyDetection;
use App\Framework\MachineLearning\ValueObjects\Baseline;
use App\Framework\MachineLearning\ValueObjects\Feature;
use App\Framework\MachineLearning\ValueObjects\FeatureType;
use App\Framework\Waf\Analysis\ValueObjects\RequestAnalysisData;
use App\Framework\Waf\MachineLearning\WafFeatureExtractor;

/**
 * Main machine learning engine for behavioral analysis and anomaly detection
 */
final class MachineLearningEngine
{
    /**
     * @param WafFeatureExtractor[] $extractors
     * @param AnomalyDetectorInterface[] $detectors
     */
    public function __construct(
        private readonly bool $enabled,
        private readonly array $extractors,
        private readonly array $detectors,
        private readonly Clock $clock,
        private readonly Duration $analysisTimeout,
        private readonly Percentage $confidenceThreshold,
        private readonly bool $enableParallelProcessing = false,
        private readonly bool $enableFeatureCaching = true,
        private readonly int $maxFeaturesPerRequest = 100,
        private array $featureCache = [],
        private array $baselineCache = [],
        private array $performanceMetrics = []
    ) {
    }

    /**
     * Analyze request for behavioral anomalies
     */
    public function analyzeRequest(RequestAnalysisData $requestData, array $context = []): MachineLearningResult
    {
        $startTime = $this->clock->time();

        if (! $this->enabled) {
            return new MachineLearningResult(
                features: [],
                anomalies: [],
                confidence: Percentage::from(0.0),
                processingTime: Duration::zero(),
                enabled: false
            );
        }

        try {
            // Extract behavioral features
            $features = $this->extractFeatures($requestData, $context);

            // Get relevant baselines
            $baselines = $this->getBaselines($features);

            // Detect anomalies
            $anomalies = $this->detectAnomalies($features, $baselines);

            // Calculate overall confidence
            $confidence = $this->calculateOverallConfidence($anomalies);

            // Update models with new data
            $this->updateModels($features);

            $processingTime = $startTime->diff($this->clock->time());

            // Record performance metrics
            $this->recordPerformanceMetrics($processingTime, count($features), count($anomalies));

            return new MachineLearningResult(
                features: $features,
                anomalies: $anomalies,
                confidence: $confidence,
                processingTime: $processingTime,
                enabled: true,
                extractorResults: $this->getExtractorResults($features),
                detectorResults: $this->getDetectorResults($anomalies),
                baselineStats: $this->getBaselineStats($baselines)
            );

        } catch (\Throwable $e) {
            $processingTime = $startTime->diff($this->clock->time());

            return new MachineLearningResult(
                features: [],
                anomalies: [],
                confidence: Percentage::from(0.0),
                processingTime: $processingTime,
                enabled: true,
                error: $e->getMessage()
            );
        }
    }

    /**
     * Extract behavioral features from request data
     */
    private function extractFeatures(RequestAnalysisData $requestData, array $context): array
    {
        $allFeatures = [];
        $extractorResults = [];

        // Check cache first
        $cacheKey = $this->generateFeatureCacheKey($requestData);
        if ($this->enableFeatureCaching && isset($this->featureCache[$cacheKey])) {
            return $this->featureCache[$cacheKey];
        }

        // Sort extractors by priority
        $sortedExtractors = $this->extractors;
        usort($sortedExtractors, fn ($a, $b) => $b->getPriority() <=> $a->getPriority());

        foreach ($sortedExtractors as $extractor) {
            if (! $extractor->isEnabled() || ! $extractor->canExtract($requestData)) {
                continue;
            }

            try {
                $extractorStart = $this->clock->time();

                // Check timeout
                if ($extractorStart->diff($this->clock->time())->toMilliseconds() > $this->analysisTimeout->toMilliseconds()) {
                    break;
                }

                $features = $extractor->extractFeatures($requestData, $context);

                $extractorTime = $extractorStart->diff($this->clock->time());

                // Validate and filter features
                $validFeatures = $this->validateFeatures($features);
                $allFeatures = array_merge($allFeatures, $validFeatures);

                $extractorResults[] = [
                    'extractor' => get_class($extractor),
                    'feature_type' => $extractor->getFeatureType()->value,
                    'feature_count' => count($validFeatures),
                    'processing_time' => $extractorTime->toMilliseconds(),
                    'success' => true,
                ];

                // Check feature limit
                if (count($allFeatures) >= $this->maxFeaturesPerRequest) {
                    break;
                }

            } catch (\Throwable $e) {
                $extractorResults[] = [
                    'extractor' => get_class($extractor),
                    'feature_type' => $extractor->getFeatureType()->value,
                    'feature_count' => 0,
                    'processing_time' => 0,
                    'success' => false,
                    'error' => $e->getMessage(),
                ];
            }
        }

        // Cache results
        if ($this->enableFeatureCaching) {
            $this->featureCache[$cacheKey] = $allFeatures;

            // Limit cache size
            if (count($this->featureCache) > 100) {
                array_shift($this->featureCache);
            }
        }

        return $allFeatures;
    }

    /**
     * Detect anomalies in extracted features
     */
    private function detectAnomalies(array $features, array $baselines): array
    {
        $allAnomalies = [];
        $detectorResults = [];

        foreach ($this->detectors as $detector) {
            if (! $detector->isEnabled() || ! $detector->canAnalyze($features)) {
                continue;
            }

            try {
                $detectorStart = $this->clock->time();

                // Get relevant baseline for this detector
                $relevantBaseline = $this->getRelevantBaseline($detector, $baselines);

                $anomalies = $detector->detectAnomalies($features, $relevantBaseline);

                $detectorTime = $detectorStart->diff($this->clock->time());

                // Filter by confidence threshold
                $validAnomalies = array_filter(
                    $anomalies,
                    fn (AnomalyDetection $anomaly) => $anomaly->confidence->getValue() >= $this->confidenceThreshold->getValue()
                );

                $allAnomalies = array_merge($allAnomalies, $validAnomalies);

                $detectorResults[] = [
                    'detector' => get_class($detector),
                    'detector_name' => $detector->getName(),
                    'anomaly_count' => count($validAnomalies),
                    'processing_time' => $detectorTime->toMilliseconds(),
                    'success' => true,
                ];

            } catch (\Throwable $e) {
                $detectorResults[] = [
                    'detector' => get_class($detector),
                    'detector_name' => $detector->getName(),
                    'anomaly_count' => 0,
                    'processing_time' => 0,
                    'success' => false,
                    'error' => $e->getMessage(),
                ];
            }
        }

        // Remove duplicate anomalies and rank by severity
        return $this->deduplicateAndRankAnomalies($allAnomalies);
    }

    /**
     * Get behavioral baselines for analysis
     */
    private function getBaselines(array $features): array
    {
        $baselines = [];

        // Group features by feature type
        $featureGroups = [];
        foreach ($features as $feature) {
            if ($feature instanceof Feature) {
                $typeKey = $feature->type->value;
                if (! isset($featureGroups[$typeKey])) {
                    $featureGroups[$typeKey] = [];
                }
                $featureGroups[$typeKey][] = $feature;
            }
        }

        // Get or create baselines for each feature type
        foreach ($featureGroups as $featureType => $groupFeatures) {
            $cacheKey = "baseline:{$featureType}";

            if (isset($this->baselineCache[$cacheKey])) {
                $baselines[$featureType] = $this->baselineCache[$cacheKey];
            } else {
                // Create new baseline from features
                $baseline = $this->createBaselineFromFeatures($groupFeatures, FeatureType::from($featureType));
                if ($baseline !== null) {
                    $baselines[$featureType] = $baseline;
                    $this->baselineCache[$cacheKey] = $baseline;
                }
            }
        }

        return $baselines;
    }

    /**
     * Create behavioral baseline from feature set
     */
    private function createBaselineFromFeatures(array $features, FeatureType $featureType): ?Baseline
    {
        if (empty($features)) {
            return null;
        }

        $values = array_map(fn (Feature $f) => $f->value, $features);

        if (empty($values)) {
            return null;
        }

        $mean = array_sum($values) / count($values);
        $variance = array_sum(array_map(fn ($v) => pow($v - $mean, 2), $values)) / count($values);
        $stdDev = sqrt($variance);

        sort($values);
        $p50 = $values[(int)(count($values) * 0.5)];
        $p95 = $values[(int)(count($values) * 0.95)];
        $p99 = $values[(int)(count($values) * 0.99)];

        return new Baseline(
            type: $featureType,
            identifier: 'dynamic-baseline',
            mean: $mean,
            standardDeviation: $stdDev,
            median: $p50,
            minimum: min($values),
            maximum: max($values),
            percentiles: [
                50 => $p50,
                95 => $p95,
                99 => $p99
            ],
            sampleCount: count($values),
            createdAt: $this->clock->time(),
            lastUpdated: $this->clock->time(),
            windowSize: Duration::fromMinutes(30),
            confidence: min(1.0, count($values) * 0.05) // 5% per sample, max 100%
        );
    }

    /**
     * Get relevant baseline for a detector
     */
    private function getRelevantBaseline(AnomalyDetectorInterface $detector, array $baselines): ?Baseline
    {
        $supportedTypes = $detector->getSupportedFeatureTypes();

        foreach ($supportedTypes as $featureType) {
            $typeKey = $featureType->value;
            if (isset($baselines[$typeKey])) {
                return $baselines[$typeKey];
            }
        }

        return null;
    }

    /**
     * Validate extracted features
     */
    private function validateFeatures(array $features): array
    {
        $validFeatures = [];

        foreach ($features as $feature) {
            if ($feature instanceof Feature) {
                // Validate feature values
                if (is_numeric($feature->value) &&
                    ! is_nan($feature->value) &&
                    ! is_infinite($feature->value)) {
                    $validFeatures[] = $feature;
                }
            }
        }

        return $validFeatures;
    }

    /**
     * Calculate overall confidence from anomaly detections
     */
    private function calculateOverallConfidence(array $anomalies): Percentage
    {
        if (empty($anomalies)) {
            return Percentage::from(0.0);
        }

        $confidenceSum = 0.0;
        $weightSum = 0.0;

        foreach ($anomalies as $anomaly) {
            if ($anomaly instanceof AnomalyDetection) {
                $weight = $anomaly->anomalyScore; // Use anomaly score as weight
                $confidenceSum += $anomaly->confidence->getValue() * $weight;
                $weightSum += $weight;
            }
        }

        $overallConfidence = $weightSum > 0 ? $confidenceSum / $weightSum : 0.0;

        return Percentage::from(min(100.0, $overallConfidence));
    }

    /**
     * Update machine learning models with new data
     */
    private function updateModels(array $features): void
    {
        foreach ($this->detectors as $detector) {
            if ($detector->isEnabled()) {
                try {
                    $detector->updateModel($features);
                } catch (\Throwable $e) {
                    // Log error but continue processing
                }
            }
        }
    }

    /**
     * Apply feedback-based adjustments to machine learning models
     *
     * @param array<string, ValueObjects\ModelAdjustment> $adjustments Adjustments to apply
     * @return array<string, mixed> Results of applying adjustments
     */
    public function applyFeedbackAdjustments(array $adjustments): array
    {
        if (empty($adjustments)) {
            return [
                'success' => true,
                'applied_count' => 0,
                'message' => 'No adjustments to apply',
            ];
        }

        $appliedCount = 0;
        $failedCount = 0;
        $results = [];

        foreach ($adjustments as $id => $adjustment) {
            try {
                // Find detectors that handle this category
                $applicableDetectors = $this->findDetectorsForCategory($adjustment->category);

                if (empty($applicableDetectors)) {
                    $results[$id] = [
                        'success' => false,
                        'message' => 'No applicable detectors found for category: ' . $adjustment->category->value,
                    ];
                    $failedCount++;

                    continue;
                }

                // Apply adjustments to each applicable detector
                $detectorResults = [];
                foreach ($applicableDetectors as $detector) {
                    $detectorResult = $this->applyAdjustmentToDetector($detector, $adjustment);
                    $detectorResults[$detector::class] = $detectorResult;
                }

                $results[$id] = [
                    'success' => true,
                    'detector_results' => $detectorResults,
                    'adjustment' => $adjustment->toArray(),
                ];

                $appliedCount++;
            } catch (\Throwable $e) {
                $results[$id] = [
                    'success' => false,
                    'message' => 'Error applying adjustment: ' . $e->getMessage(),
                    'adjustment' => $adjustment->toArray(),
                ];
                $failedCount++;
            }
        }

        return [
            'success' => $failedCount === 0,
            'applied_count' => $appliedCount,
            'failed_count' => $failedCount,
            'results' => $results,
        ];
    }

    /**
     * Find detectors that handle a specific category
     *
     * @param DetectionCategory $category The category to find detectors for
     * @return array<AnomalyDetectorInterface> Applicable detectors
     */
    private function findDetectorsForCategory(DetectionCategory $category): array
    {
        // In a real implementation, this would use detector metadata or capabilities
        // to determine which detectors can handle which categories

        // For now, we'll use a simplified approach based on detector class names
        $applicableDetectors = [];

        foreach ($this->detectors as $detector) {
            // Check if detector handles this category based on class name or metadata
            $detectorClass = get_class($detector);
            $categoryName = $category->value;

            // Simple heuristic: if detector class name contains category name or is generic
            if (
                stripos($detectorClass, $categoryName) !== false ||
                stripos($detectorClass, 'Generic') !== false ||
                stripos($detectorClass, 'Statistical') !== false ||
                stripos($detectorClass, 'Clustering') !== false
            ) {
                $applicableDetectors[] = $detector;
            }
        }

        return $applicableDetectors;
    }

    /**
     * Apply a model adjustment to a specific detector
     *
     * @param AnomalyDetectorInterface $detector The detector to adjust
     * @param ValueObjects\ModelAdjustment $adjustment The adjustment to apply
     * @return array<string, mixed> Result of applying the adjustment
     */
    private function applyAdjustmentToDetector(
        AnomalyDetectorInterface $detector,
        ValueObjects\ModelAdjustment $adjustment
    ): array {
        $result = [
            'threshold_adjusted' => false,
            'confidence_adjusted' => false,
            'features_adjusted' => 0,
        ];

        // Apply threshold adjustment if detector supports it
        if ($detector instanceof ThresholdAdjustableInterface && ! $adjustment->thresholdAdjustment->isZero()) {
            $detector->adjustThreshold($adjustment->thresholdAdjustment);
            $result['threshold_adjusted'] = true;
        }

        // Apply confidence adjustment if detector supports it
        if ($detector instanceof ConfidenceAdjustableInterface && ! $adjustment->confidenceAdjustment->isZero()) {
            $detector->adjustConfidence($adjustment->confidenceAdjustment);
            $result['confidence_adjusted'] = true;
        }

        // Apply feature weight adjustments if detector supports it
        if ($detector instanceof FeatureWeightAdjustableInterface && $adjustment->hasFeatureWeightAdjustments()) {
            $adjustedFeatures = $detector->adjustFeatureWeights($adjustment->featureWeightAdjustments);
            $result['features_adjusted'] = count($adjustedFeatures);
            $result['adjusted_features'] = $adjustedFeatures;
        }

        return $result;
    }

    /**
     * Deduplicate and rank anomalies by severity
     */
    private function deduplicateAndRankAnomalies(array $anomalies): array
    {
        // Remove duplicates based on type and feature type
        $seen = [];
        $unique = [];

        foreach ($anomalies as $anomaly) {
            if ($anomaly instanceof AnomalyDetection) {
                $key = $anomaly->type->value . ':' . $anomaly->featureType->value;
                if (! isset($seen[$key]) || $anomaly->confidence->getValue() > $seen[$key]->confidence->getValue()) {
                    $seen[$key] = $anomaly;
                }
            }
        }

        $unique = array_values($seen);

        // Sort by anomaly score (descending)
        usort($unique, fn ($a, $b) => $b->anomalyScore <=> $a->anomalyScore);

        return $unique;
    }

    /**
     * Generate cache key for features
     */
    private function generateFeatureCacheKey(RequestAnalysisData $requestData): string
    {
        return md5(serialize([
            'path' => $requestData->path,
            'method' => $requestData->method,
            'params' => $requestData->getAllParameters(),
            'user_agent' => (string) $requestData->userAgent,
            'ip' => (string) $requestData->clientIp,
        ]));
    }

    /**
     * Get extractor results summary
     */
    private function getExtractorResults(array $features): array
    {
        $results = [];
        $featuresByType = [];

        foreach ($features as $feature) {
            if ($feature instanceof Feature) {
                $typeKey = $feature->type->value;
                if (! isset($featuresByType[$typeKey])) {
                    $featuresByType[$typeKey] = [];
                }
                $featuresByType[$typeKey][] = $feature;
            }
        }

        foreach ($featuresByType as $featureType => $typeFeatures) {
            $results[] = [
                'feature_type' => $featureType,
                'feature_count' => count($typeFeatures),
                'avg_value' => array_sum(array_map(fn ($f) => $f->value, $typeFeatures)) / count($typeFeatures),
                'feature_names' => array_unique(array_map(fn ($f) => $f->name, $typeFeatures)),
            ];
        }

        return $results;
    }

    /**
     * Get detector results summary
     */
    private function getDetectorResults(array $anomalies): array
    {
        $results = [];
        $anomaliesByDetector = [];

        foreach ($anomalies as $anomaly) {
            if ($anomaly instanceof AnomalyDetection) {
                $detectorKey = $anomaly->type->value;
                if (! isset($anomaliesByDetector[$detectorKey])) {
                    $anomaliesByDetector[$detectorKey] = [];
                }
                $anomaliesByDetector[$detectorKey][] = $anomaly;
            }
        }

        foreach ($anomaliesByDetector as $detectorType => $detectorAnomalies) {
            $results[] = [
                'detector_type' => $detectorType,
                'anomaly_count' => count($detectorAnomalies),
                'avg_confidence' => array_sum(array_map(fn ($a) => $a->confidence->getValue(), $detectorAnomalies)) / count($detectorAnomalies),
                'max_score' => max(array_map(fn ($a) => $a->anomalyScore, $detectorAnomalies)),
            ];
        }

        return $results;
    }

    /**
     * Get baseline statistics summary
     */
    private function getBaselineStats(array $baselines): array
    {
        $stats = [];

        foreach ($baselines as $featureType => $baseline) {
            if ($baseline instanceof Baseline) {
                $stats[] = [
                    'feature_type' => $featureType,
                    'sample_size' => $baseline->sampleCount,
                    'mean' => $baseline->mean,
                    'std_dev' => $baseline->standardDeviation,
                    'confidence' => $baseline->confidence,
                    'last_updated' => $baseline->lastUpdated->toIso8601String(),
                ];
            }
        }

        return $stats;
    }

    /**
     * Record performance metrics
     */
    private function recordPerformanceMetrics(Duration $processingTime, int $featureCount, int $anomalyCount): void
    {
        $this->performanceMetrics[] = [
            'timestamp' => $this->clock->time()->toTimestamp(),
            'processing_time_ms' => $processingTime->toMilliseconds(),
            'feature_count' => $featureCount,
            'anomaly_count' => $anomalyCount,
        ];

        // Limit metrics history
        if (count($this->performanceMetrics) > 1000) {
            array_shift($this->performanceMetrics);
        }
    }

    /**
     * Get performance statistics
     */
    public function getPerformanceStats(): array
    {
        if (empty($this->performanceMetrics)) {
            return [];
        }

        $processingTimes = array_column($this->performanceMetrics, 'processing_time_ms');
        $featureCounts = array_column($this->performanceMetrics, 'feature_count');
        $anomalyCounts = array_column($this->performanceMetrics, 'anomaly_count');

        return [
            'total_requests' => count($this->performanceMetrics),
            'avg_processing_time_ms' => array_sum($processingTimes) / count($processingTimes),
            'max_processing_time_ms' => max($processingTimes),
            'avg_feature_count' => array_sum($featureCounts) / count($featureCounts),
            'avg_anomaly_count' => array_sum($anomalyCounts) / count($anomalyCounts),
            'cache_hit_ratio' => $this->enableFeatureCaching ? count($this->featureCache) / max(count($this->performanceMetrics), 1) : 0.0,
        ];
    }

    /**
     * Get configuration
     */
    public function getConfiguration(): array
    {
        return [
            'enabled' => $this->enabled,
            'analysis_timeout_ms' => $this->analysisTimeout->toMilliseconds(),
            'confidence_threshold' => $this->confidenceThreshold->getValue(),
            'enable_parallel_processing' => $this->enableParallelProcessing,
            'enable_feature_caching' => $this->enableFeatureCaching,
            'max_features_per_request' => $this->maxFeaturesPerRequest,
            'extractor_count' => count($this->extractors),
            'detector_count' => count($this->detectors),
            'cache_size' => count($this->featureCache),
            'baseline_count' => count($this->baselineCache),
        ];
    }

    /**
     * Check if engine is enabled
     */
    public function isEnabled(): bool
    {
        return $this->enabled;
    }
}