using System; #if REAL_T_IS_DOUBLE using real_t = System.Double; #else using real_t = System.Single; #endif namespace Godot { public static partial class Mathf { // Define constants with Decimal precision and cast down to double or float. public const real_t Tau = (real_t) 6.2831853071795864769252867666M; // 6.2831855f and 6.28318530717959 public const real_t Pi = (real_t) 3.1415926535897932384626433833M; // 3.1415927f and 3.14159265358979 public const real_t Inf = real_t.PositiveInfinity; public const real_t NaN = real_t.NaN; private const real_t Deg2RadConst = (real_t) 0.0174532925199432957692369077M; // 0.0174532924f and 0.0174532925199433 private const real_t Rad2DegConst = (real_t) 57.295779513082320876798154814M; // 57.29578f and 57.2957795130823 public static real_t Abs(real_t s) { return Math.Abs(s); } public static int Abs(int s) { return Math.Abs(s); } public static real_t Acos(real_t s) { return (real_t)Math.Acos(s); } public static real_t Asin(real_t s) { return (real_t)Math.Asin(s); } public static real_t Atan(real_t s) { return (real_t)Math.Atan(s); } public static real_t Atan2(real_t x, real_t y) { return (real_t)Math.Atan2(x, y); } public static Vector2 Cartesian2Polar(real_t x, real_t y) { return new Vector2(Sqrt(x * x + y * y), Atan2(y, x)); } public static real_t Ceil(real_t s) { return (real_t)Math.Ceiling(s); } public static int Clamp(int value, int min, int max) { return value < min ? min : value > max ? max : value; } public static real_t Clamp(real_t value, real_t min, real_t max) { return value < min ? min : value > max ? max : value; } public static real_t Cos(real_t s) { return (real_t)Math.Cos(s); } public static real_t Cosh(real_t s) { return (real_t)Math.Cosh(s); } public static int StepDecimals(real_t step) { double[] sd = new double[] { 0.9999, 0.09999, 0.009999, 0.0009999, 0.00009999, 0.000009999, 0.0000009999, 0.00000009999, 0.000000009999, }; double abs = Mathf.Abs(step); double decs = abs - (int)abs; // Strip away integer part for (int i = 0; i < sd.Length; i++) { if (decs >= sd[i]) { return i; } } return 0; } public static real_t Deg2Rad(real_t deg) { return deg * Deg2RadConst; } public static real_t Ease(real_t s, real_t curve) { if (s < 0f) { s = 0f; } else if (s > 1.0f) { s = 1.0f; } if (curve > 0f) { if (curve < 1.0f) { return 1.0f - Pow(1.0f - s, 1.0f / curve); } return Pow(s, curve); } if (curve < 0f) { if (s < 0.5f) { return Pow(s * 2.0f, -curve) * 0.5f; } return (1.0f - Pow(1.0f - (s - 0.5f) * 2.0f, -curve)) * 0.5f + 0.5f; } return 0f; } public static real_t Exp(real_t s) { return (real_t)Math.Exp(s); } public static real_t Floor(real_t s) { return (real_t)Math.Floor(s); } public static real_t InverseLerp(real_t from, real_t to, real_t weight) { return (weight - from) / (to - from); } public static bool IsEqualApprox(real_t a, real_t b) { real_t tolerance = Epsilon * Abs(a); if (tolerance < Epsilon) { tolerance = Epsilon; } return Abs(a - b) < tolerance; } public static bool IsInf(real_t s) { return real_t.IsInfinity(s); } public static bool IsNaN(real_t s) { return real_t.IsNaN(s); } public static bool IsZeroApprox(real_t s) { return Abs(s) < Epsilon; } public static real_t Lerp(real_t from, real_t to, real_t weight) { return from + (to - from) * weight; } public static real_t Log(real_t s) { return (real_t)Math.Log(s); } public static int Max(int a, int b) { return a > b ? a : b; } public static real_t Max(real_t a, real_t b) { return a > b ? a : b; } public static int Min(int a, int b) { return a < b ? a : b; } public static real_t Min(real_t a, real_t b) { return a < b ? a : b; } public static int NearestPo2(int value) { value--; value |= value >> 1; value |= value >> 2; value |= value >> 4; value |= value >> 8; value |= value >> 16; value++; return value; } public static Vector2 Polar2Cartesian(real_t r, real_t th) { return new Vector2(r * Cos(th), r * Sin(th)); } /// /// Performs a canonical Modulus operation, where the output is on the range [0, b). /// public static real_t PosMod(real_t a, real_t b) { real_t c = a % b; if ((c < 0 && b > 0) || (c > 0 && b < 0)) { c += b; } return c; } /// /// Performs a canonical Modulus operation, where the output is on the range [0, b). /// public static int PosMod(int a, int b) { int c = a % b; if ((c < 0 && b > 0) || (c > 0 && b < 0)) { c += b; } return c; } public static real_t Pow(real_t x, real_t y) { return (real_t)Math.Pow(x, y); } public static real_t Rad2Deg(real_t rad) { return rad * Rad2DegConst; } public static real_t Round(real_t s) { return (real_t)Math.Round(s); } public static int Sign(int s) { return s < 0 ? -1 : 1; } public static real_t Sign(real_t s) { return s < 0f ? -1f : 1f; } public static real_t Sin(real_t s) { return (real_t)Math.Sin(s); } public static real_t Sinh(real_t s) { return (real_t)Math.Sinh(s); } public static real_t SmoothStep(real_t from, real_t to, real_t weight) { if (IsEqualApprox(from, to)) { return from; } real_t x = Clamp((weight - from) / (to - from), (real_t)0.0, (real_t)1.0); return x * x * (3 - 2 * x); } public static real_t Sqrt(real_t s) { return (real_t)Math.Sqrt(s); } public static real_t Stepify(real_t s, real_t step) { if (step != 0f) { s = Floor(s / step + 0.5f) * step; } return s; } public static real_t Tan(real_t s) { return (real_t)Math.Tan(s); } public static real_t Tanh(real_t s) { return (real_t)Math.Tanh(s); } public static int Wrap(int value, int min, int max) { int rng = max - min; return rng != 0 ? min + ((value - min) % rng + rng) % rng : min; } public static real_t Wrap(real_t value, real_t min, real_t max) { real_t rng = max - min; return !IsEqualApprox(rng, default(real_t)) ? min + ((value - min) % rng + rng) % rng : min; } } }