import 'dart:io';
import 'dart:math';
import 'dart:typed_data';

import 'package:flutter/material.dart';
import 'package:flutter_liveness_check/flutter_liveness_check.dart';
import 'package:google_mlkit_face_detection/google_mlkit_face_detection.dart';

/// Result from a face liveness check.
class FaceLivenessResult {
  final Uint8List imageBytes;
  final String? imagePath;
  FaceLivenessResult({required this.imageBytes, this.imagePath});
}

/// Run face liveness detection on mobile using flutter_liveness_check.
/// Navigates to the LivenessCheckScreen and returns the captured photo.
Future<FaceLivenessResult?> runFaceLiveness(
  BuildContext context, {
  int requiredBlinks = 3,
}) async {
  String? capturedPath;

  await Navigator.of(context).push(
    MaterialPageRoute(
      builder: (ctx) => LivenessCheckScreen(
        config: LivenessCheckConfig(
          callbacks: LivenessCheckCallbacks(
            onPhotoTaken: (path) {
              capturedPath = path;
              // Package never calls onSuccess in v1.0.3 — pop here
              // so the screen doesn't hang after photo capture.
              Navigator.of(ctx).pop();
            },
            // Don't pop in onCancel/onError — the package's AppBar
            // already calls Navigator.pop() after invoking these.
          ),
          settings: LivenessCheckSettings(
            requiredBlinkCount: requiredBlinks,
            requireSmile: false,
            autoNavigateOnSuccess: false,
          ),
        ),
      ),
    ),
  );

  if (capturedPath == null) return null;

  final file = File(capturedPath!);
  if (!await file.exists()) return null;

  final bytes = await file.readAsBytes();
  return FaceLivenessResult(imageBytes: bytes, imagePath: capturedPath);
}

/// Compare a captured face photo with enrolled face photo bytes.
/// Uses Google ML Kit face contour comparison.
/// Returns similarity score 0.0 (no match) to 1.0 (perfect match).
Future<double> compareFaces(
  Uint8List capturedBytes,
  Uint8List enrolledBytes,
) async {
  final detector = FaceDetector(
    options: FaceDetectorOptions(
      enableContours: true,
      performanceMode: FaceDetectorMode.accurate,
    ),
  );

  try {
    // Save both to temp files for ML Kit
    final tempDir = Directory.systemTemp;
    final capturedFile = File('${tempDir.path}/face_captured_temp.jpg');
    await capturedFile.writeAsBytes(capturedBytes);

    final enrolledFile = File('${tempDir.path}/face_enrolled_temp.jpg');
    await enrolledFile.writeAsBytes(enrolledBytes);

    // Process both images
    final capturedInput = InputImage.fromFilePath(capturedFile.path);
    final enrolledInput = InputImage.fromFilePath(enrolledFile.path);

    final capturedFaces = await detector.processImage(capturedInput);
    final enrolledFaces = await detector.processImage(enrolledInput);

    // Cleanup temp files
    await capturedFile.delete().catchError((_) => capturedFile);
    await enrolledFile.delete().catchError((_) => enrolledFile);

    if (capturedFaces.isEmpty || enrolledFaces.isEmpty) return 0.0;

    return _compareContours(capturedFaces.first, enrolledFaces.first);
  } catch (_) {
    return 0.0;
  } finally {
    await detector.close();
  }
}

double _compareContours(Face face1, Face face2) {
  const contourTypes = [
    FaceContourType.face,
    FaceContourType.leftEye,
    FaceContourType.rightEye,
    FaceContourType.noseBridge,
    FaceContourType.noseBottom,
    FaceContourType.upperLipTop,
    FaceContourType.lowerLipBottom,
  ];

  double totalScore = 0;
  int validComparisons = 0;

  for (final type in contourTypes) {
    final c1 = face1.contours[type];
    final c2 = face2.contours[type];

    if (c1 != null &&
        c2 != null &&
        c1.points.isNotEmpty &&
        c2.points.isNotEmpty) {
      final score = _comparePointSets(c1.points, c2.points);
      totalScore += score;
      validComparisons++;
    }
  }

  if (validComparisons == 0) return 0.0;
  return totalScore / validComparisons;
}

double _comparePointSets(List<Point<int>> points1, List<Point<int>> points2) {
  final norm1 = _normalizePoints(points1);
  final norm2 = _normalizePoints(points2);

  final n = min(norm1.length, norm2.length);
  if (n == 0) return 0.0;

  double totalDist = 0;
  for (int i = 0; i < n; i++) {
    final dx = norm1[i].x - norm2[i].x;
    final dy = norm1[i].y - norm2[i].y;
    totalDist += sqrt(dx * dx + dy * dy);
  }

  final avgDist = totalDist / n;
  // Convert distance to similarity: 0 distance → 1.0 score
  return max(0.0, 1.0 - avgDist * 2.5);
}

List<Point<double>> _normalizePoints(List<Point<int>> points) {
  if (points.isEmpty) return [];

  double minX = double.infinity, minY = double.infinity;
  double maxX = double.negativeInfinity, maxY = double.negativeInfinity;

  for (final p in points) {
    minX = min(minX, p.x.toDouble());
    minY = min(minY, p.y.toDouble());
    maxX = max(maxX, p.x.toDouble());
    maxY = max(maxY, p.y.toDouble());
  }

  final w = maxX - minX;
  final h = maxY - minY;
  if (w == 0 || h == 0) return [];

  return points
      .map((p) => Point<double>((p.x - minX) / w, (p.y - minY) / h))
      .toList();
}