package com.mojang.math;

import org.apache.commons.lang3.tuple.Triple;
import org.joml.Math;
import org.joml.Matrix3f;
import org.joml.Matrix4f;
import org.joml.Quaternionf;
import org.joml.Vector3f;

public class MatrixUtil {
	private static final float G = 3.0F + 2.0F * Math.sqrt(2.0F);
	private static final GivensParameters PI_4 = GivensParameters.fromPositiveAngle((float) (java.lang.Math.PI / 4));

	private MatrixUtil() {
	}

	public static Matrix4f mulComponentWise(Matrix4f matrix, float scalar) {
		return matrix.set(
			matrix.m00() * scalar,
			matrix.m01() * scalar,
			matrix.m02() * scalar,
			matrix.m03() * scalar,
			matrix.m10() * scalar,
			matrix.m11() * scalar,
			matrix.m12() * scalar,
			matrix.m13() * scalar,
			matrix.m20() * scalar,
			matrix.m21() * scalar,
			matrix.m22() * scalar,
			matrix.m23() * scalar,
			matrix.m30() * scalar,
			matrix.m31() * scalar,
			matrix.m32() * scalar,
			matrix.m33() * scalar
		);
	}

	private static GivensParameters approxGivensQuat(float topCorner, float oppositeDiagonalAverage, float bottomCorner) {
		float f = 2.0F * (topCorner - bottomCorner);
		return G * oppositeDiagonalAverage * oppositeDiagonalAverage < f * f ? GivensParameters.fromUnnormalized(oppositeDiagonalAverage, f) : PI_4;
	}

	private static GivensParameters qrGivensQuat(float input1, float input2) {
		float f = (float)java.lang.Math.hypot(input1, input2);
		float g = f > 1.0E-6F ? input2 : 0.0F;
		float h = Math.abs(input1) + Math.max(f, 1.0E-6F);
		if (input1 < 0.0F) {
			float i = g;
			g = h;
			h = i;
		}

		return GivensParameters.fromUnnormalized(g, h);
	}

	private static void similarityTransform(Matrix3f input, Matrix3f tempStorage) {
		input.mul(tempStorage);
		tempStorage.transpose();
		tempStorage.mul(input);
		input.set(tempStorage);
	}

	private static void stepJacobi(Matrix3f input, Matrix3f tempStorage, Quaternionf resultEigenvector, Quaternionf resultEigenvalue) {
		if (input.m01 * input.m01 + input.m10 * input.m10 > 1.0E-6F) {
			GivensParameters givensParameters = approxGivensQuat(input.m00, 0.5F * (input.m01 + input.m10), input.m11);
			Quaternionf quaternionf = givensParameters.aroundZ(resultEigenvector);
			resultEigenvalue.mul(quaternionf);
			givensParameters.aroundZ(tempStorage);
			similarityTransform(input, tempStorage);
		}

		if (input.m02 * input.m02 + input.m20 * input.m20 > 1.0E-6F) {
			GivensParameters givensParameters = approxGivensQuat(input.m00, 0.5F * (input.m02 + input.m20), input.m22).inverse();
			Quaternionf quaternionf = givensParameters.aroundY(resultEigenvector);
			resultEigenvalue.mul(quaternionf);
			givensParameters.aroundY(tempStorage);
			similarityTransform(input, tempStorage);
		}

		if (input.m12 * input.m12 + input.m21 * input.m21 > 1.0E-6F) {
			GivensParameters givensParameters = approxGivensQuat(input.m11, 0.5F * (input.m12 + input.m21), input.m22);
			Quaternionf quaternionf = givensParameters.aroundX(resultEigenvector);
			resultEigenvalue.mul(quaternionf);
			givensParameters.aroundX(tempStorage);
			similarityTransform(input, tempStorage);
		}
	}

	public static Quaternionf eigenvalueJacobi(Matrix3f input, int iterations) {
		Quaternionf quaternionf = new Quaternionf();
		Matrix3f matrix3f = new Matrix3f();
		Quaternionf quaternionf2 = new Quaternionf();

		for (int i = 0; i < iterations; i++) {
			stepJacobi(input, matrix3f, quaternionf2, quaternionf);
		}

		quaternionf.normalize();
		return quaternionf;
	}

	public static Triple<Quaternionf, Vector3f, Quaternionf> svdDecompose(Matrix3f matrix) {
		Matrix3f matrix3f = new Matrix3f(matrix);
		matrix3f.transpose();
		matrix3f.mul(matrix);
		Quaternionf quaternionf = eigenvalueJacobi(matrix3f, 5);
		float f = matrix3f.m00;
		float g = matrix3f.m11;
		boolean bl = f < 1.0E-6;
		boolean bl2 = g < 1.0E-6;
		Matrix3f matrix3f3 = matrix.rotate(quaternionf);
		Quaternionf quaternionf2 = new Quaternionf();
		Quaternionf quaternionf3 = new Quaternionf();
		GivensParameters givensParameters;
		if (bl) {
			givensParameters = qrGivensQuat(matrix3f3.m11, -matrix3f3.m10);
		} else {
			givensParameters = qrGivensQuat(matrix3f3.m00, matrix3f3.m01);
		}

		Quaternionf quaternionf4 = givensParameters.aroundZ(quaternionf3);
		Matrix3f matrix3f4 = givensParameters.aroundZ(matrix3f);
		quaternionf2.mul(quaternionf4);
		matrix3f4.transpose().mul(matrix3f3);
		if (bl) {
			givensParameters = qrGivensQuat(matrix3f4.m22, -matrix3f4.m20);
		} else {
			givensParameters = qrGivensQuat(matrix3f4.m00, matrix3f4.m02);
		}

		givensParameters = givensParameters.inverse();
		Quaternionf quaternionf5 = givensParameters.aroundY(quaternionf3);
		Matrix3f matrix3f5 = givensParameters.aroundY(matrix3f3);
		quaternionf2.mul(quaternionf5);
		matrix3f5.transpose().mul(matrix3f4);
		if (bl2) {
			givensParameters = qrGivensQuat(matrix3f5.m22, -matrix3f5.m21);
		} else {
			givensParameters = qrGivensQuat(matrix3f5.m11, matrix3f5.m12);
		}

		Quaternionf quaternionf6 = givensParameters.aroundX(quaternionf3);
		Matrix3f matrix3f6 = givensParameters.aroundX(matrix3f4);
		quaternionf2.mul(quaternionf6);
		matrix3f6.transpose().mul(matrix3f5);
		Vector3f vector3f = new Vector3f(matrix3f6.m00, matrix3f6.m11, matrix3f6.m22);
		return Triple.of(quaternionf2, vector3f, quaternionf.conjugate());
	}

	public static boolean isPureTranslation(Matrix4f matrix) {
		return (matrix.properties() & 8) != 0;
	}

	public static boolean isOrthonormal(Matrix4f matrix) {
		return (matrix.properties() & 16) != 0;
	}
}