aboutsummaryrefslogtreecommitdiff
path: root/src/raymath.h
diff options
context:
space:
mode:
authorvictorfisac <victorfisac@gmail.com>2016-01-20 19:26:52 +0100
committervictorfisac <victorfisac@gmail.com>2016-01-20 19:26:52 +0100
commit8ae0da1fa1a9d462f95ee7cf02b0fc336c4a77df (patch)
tree751d2c7303dc92780b027b122f26b09a4930e88b /src/raymath.h
parent25f5995be0f52c756052a9fff276bd11e95fc895 (diff)
parentcf6c6fefd76135df0247339c9adc2712da6c2691 (diff)
downloadraylib-8ae0da1fa1a9d462f95ee7cf02b0fc336c4a77df.tar.gz
raylib-8ae0da1fa1a9d462f95ee7cf02b0fc336c4a77df.zip
Merge remote-tracking branch 'refs/remotes/raysan5/develop' into develop
Diffstat (limited to 'src/raymath.h')
-rw-r--r--src/raymath.h414
1 files changed, 226 insertions, 188 deletions
diff --git a/src/raymath.h b/src/raymath.h
index f5912795..8ad32528 100644
--- a/src/raymath.h
+++ b/src/raymath.h
@@ -1,9 +1,23 @@
/**********************************************************************************************
*
-* raymath
+* raymath (header only file)
*
* Some useful functions to work with Vector3, Matrix and Quaternions
*
+* You must:
+* #define RAYMATH_IMPLEMENTATION
+* before you include this file in *only one* C or C++ file to create the implementation.
+*
+* Example:
+* #define RAYMATH_IMPLEMENTATION
+* #include "raymath.h"
+*
+* You can also use:
+* #define RAYMATH_EXTERN_INLINE // Inlines all functions code, so it runs faster.
+* // This requires lots of memory on system.
+* #define RAYMATH_STANDALONE // Not dependent on raylib.h structs: Vector3, Matrix.
+*
+*
* Copyright (c) 2015 Ramon Santamaria (@raysan5)
*
* This software is provided "as-is", without any express or implied warranty. In no event
@@ -22,37 +36,21 @@
* 3. This notice may not be removed or altered from any source distribution.
*
**********************************************************************************************/
-//============================================================================
-// YOU MUST
-//
-// #define RAYMATH_DEFINE
-//
-// Like:
-//
-// #define RAYMATH_DEFINE
-// #include "raymath.h"
-//
-// YOU CAN:
-// #define RAYMATH_INLINE //inlines all code, so it runs faster. This requires lots of memory on system.
-// AND
-// #define RAYMATH_STANDALONE //not dependent on outside libs
-//
-// This needs to be done for every library/source file.
-//============================================================================
-
-#ifdef RAYMATH_INLINE
- #define RMDEF static inline
-#else
- #define RMDEF static
-#endif
#ifndef RAYMATH_H
#define RAYMATH_H
-//#define RAYMATH_STANDALONE // NOTE: To use raymath as standalone lib, just uncomment this line
+//#define RAYMATH_STANDALONE // NOTE: To use raymath as standalone lib, just uncomment this line
+//#define RAYMATH_EXTERN_INLINE // NOTE: To compile functions as static inline, uncomment this line
#ifndef RAYMATH_STANDALONE
- #include "raylib.h" // Required for typedef: Vector3
+ #include "raylib.h" // Required for structs: Vector3, Matrix
+#endif
+
+#if defined(RAYMATH_EXTERN_INLINE)
+ #define RMDEF extern inline
+#else
+ #define RMDEF extern
#endif
//----------------------------------------------------------------------------------
@@ -63,18 +61,18 @@
#endif
#ifndef DEG2RAD
- #define DEG2RAD (PI / 180.0f)
+ #define DEG2RAD (PI/180.0f)
#endif
#ifndef RAD2DEG
- #define RAD2DEG (180.0f / PI)
+ #define RAD2DEG (180.0f/PI)
#endif
//----------------------------------------------------------------------------------
// Types and Structures Definition
//----------------------------------------------------------------------------------
-#ifdef RAYMATH_STANDALONE
+#if defined(RAYMATH_STANDALONE)
// Vector2 type
typedef struct Vector2 {
float x;
@@ -105,7 +103,77 @@ typedef struct Quaternion {
float w;
} Quaternion;
-#ifdef RAYMATH_DEFINE
+#ifndef RAYMATH_EXTERN_INLINE
+
+#ifdef __cplusplus
+extern "C" {
+#endif
+
+//------------------------------------------------------------------------------------
+// Functions Declaration to work with Vector3
+//------------------------------------------------------------------------------------
+RMDEF Vector3 VectorAdd(Vector3 v1, Vector3 v2); // Add two vectors
+RMDEF Vector3 VectorSubtract(Vector3 v1, Vector3 v2); // Substract two vectors
+RMDEF Vector3 VectorCrossProduct(Vector3 v1, Vector3 v2); // Calculate two vectors cross product
+RMDEF Vector3 VectorPerpendicular(Vector3 v); // Calculate one vector perpendicular vector
+RMDEF float VectorDotProduct(Vector3 v1, Vector3 v2); // Calculate two vectors dot product
+RMDEF float VectorLength(const Vector3 v); // Calculate vector lenght
+RMDEF void VectorScale(Vector3 *v, float scale); // Scale provided vector
+RMDEF void VectorNegate(Vector3 *v); // Negate provided vector (invert direction)
+RMDEF void VectorNormalize(Vector3 *v); // Normalize provided vector
+RMDEF float VectorDistance(Vector3 v1, Vector3 v2); // Calculate distance between two points
+RMDEF Vector3 VectorLerp(Vector3 v1, Vector3 v2, float amount); // Calculate linear interpolation between two vectors
+RMDEF Vector3 VectorReflect(Vector3 vector, Vector3 normal); // Calculate reflected vector to normal
+RMDEF void VectorTransform(Vector3 *v, Matrix mat); // Transforms a Vector3 by a given Matrix
+RMDEF Vector3 VectorZero(void); // Return a Vector3 init to zero
+
+//------------------------------------------------------------------------------------
+// Functions Declaration to work with Matrix
+//------------------------------------------------------------------------------------
+RMDEF float MatrixDeterminant(Matrix mat); // Compute matrix determinant
+RMDEF float MatrixTrace(Matrix mat); // Returns the trace of the matrix (sum of the values along the diagonal)
+RMDEF void MatrixTranspose(Matrix *mat); // Transposes provided matrix
+RMDEF void MatrixInvert(Matrix *mat); // Invert provided matrix
+RMDEF void MatrixNormalize(Matrix *mat); // Normalize provided matrix
+RMDEF Matrix MatrixIdentity(void); // Returns identity matrix
+RMDEF Matrix MatrixAdd(Matrix left, Matrix right); // Add two matrices
+RMDEF Matrix MatrixSubstract(Matrix left, Matrix right); // Substract two matrices (left - right)
+RMDEF Matrix MatrixTranslate(float x, float y, float z); // Returns translation matrix
+RMDEF Matrix MatrixRotate(float angle, Vector3 axis); // Returns rotation matrix for an angle around an specified axis (angle in radians)
+RMDEF Matrix MatrixRotateX(float angle); // Returns x-rotation matrix (angle in radians)
+RMDEF Matrix MatrixRotateY(float angle); // Returns y-rotation matrix (angle in radians)
+RMDEF Matrix MatrixRotateZ(float angle); // Returns z-rotation matrix (angle in radians)
+RMDEF Matrix MatrixScale(float x, float y, float z); // Returns scaling matrix
+RMDEF Matrix MatrixMultiply(Matrix left, Matrix right); // Returns two matrix multiplication
+RMDEF Matrix MatrixFrustum(double left, double right, double bottom, double top, double near, double far); // Returns perspective projection matrix
+RMDEF Matrix MatrixPerspective(double fovy, double aspect, double near, double far); // Returns perspective projection matrix
+RMDEF Matrix MatrixOrtho(double left, double right, double bottom, double top, double near, double far); // Returns orthographic projection matrix
+RMDEF Matrix MatrixLookAt(Vector3 position, Vector3 target, Vector3 up); // Returns camera look-at matrix (view matrix)
+RMDEF void PrintMatrix(Matrix m); // Print matrix utility
+
+//------------------------------------------------------------------------------------
+// Functions Declaration to work with Quaternions
+//------------------------------------------------------------------------------------
+RMDEF float QuaternionLength(Quaternion quat); // Compute the length of a quaternion
+RMDEF void QuaternionNormalize(Quaternion *q); // Normalize provided quaternion
+RMDEF Quaternion QuaternionMultiply(Quaternion q1, Quaternion q2); // Calculate two quaternion multiplication
+RMDEF Quaternion QuaternionSlerp(Quaternion q1, Quaternion q2, float slerp); // Calculates spherical linear interpolation between two quaternions
+RMDEF Quaternion QuaternionFromMatrix(Matrix matrix); // Returns a quaternion for a given rotation matrix
+RMDEF Matrix QuaternionToMatrix(Quaternion q); // Returns a matrix for a given quaternion
+RMDEF Quaternion QuaternionFromAxisAngle(float angle, Vector3 axis); // Returns rotation quaternion for an angle and axis
+RMDEF void QuaternionToAxisAngle(Quaternion q, float *outAngle, Vector3 *outAxis); // Returns the rotation angle and axis for a given quaternion
+RMDEF void QuaternionTransform(Quaternion *q, Matrix mat); // Transform a quaternion given a transformation matrix
+
+#ifdef __cplusplus
+}
+#endif
+
+#endif // notdef RAYMATH_EXTERN_INLINE
+
+//////////////////////////////////////////////////////////////////// end of header file
+
+#if defined(RAYMATH_IMPLEMENTATION) || defined(RAYMATH_EXTERN_INLINE)
+
#include <stdio.h> // Used only on PrintMatrix()
#include <math.h> // Standard math libary: sin(), cos(), tan()...
#include <stdlib.h> // Used for abs()
@@ -114,18 +182,6 @@ typedef struct Quaternion {
// Module Functions Definition - Vector3 math
//----------------------------------------------------------------------------------
-// Converts Vector3 to float array
-RMDEF float *VectorToFloat(Vector3 vec)
-{
- static float buffer[3];
-
- buffer[0] = vec.x;
- buffer[1] = vec.y;
- buffer[2] = vec.z;
-
- return buffer;
-}
-
// Add two vectors
RMDEF Vector3 VectorAdd(Vector3 v1, Vector3 v2)
{
@@ -229,9 +285,9 @@ RMDEF void VectorNormalize(Vector3 *v)
length = VectorLength(*v);
- if (length == 0) length = 1;
+ if (length == 0) length = 1.0f;
- ilength = 1.0/length;
+ ilength = 1.0f/length;
v->x *= ilength;
v->y *= ilength;
@@ -257,9 +313,9 @@ RMDEF Vector3 VectorLerp(Vector3 v1, Vector3 v2, float amount)
{
Vector3 result;
- result.x = v1.x + amount * (v2.x - v1.x);
- result.y = v1.y + amount * (v2.y - v1.y);
- result.z = v1.z + amount * (v2.z - v1.z);
+ result.x = v1.x + amount*(v2.x - v1.x);
+ result.y = v1.y + amount*(v2.y - v1.y);
+ result.z = v1.z + amount*(v2.z - v1.z);
return result;
}
@@ -269,15 +325,15 @@ RMDEF Vector3 VectorReflect(Vector3 vector, Vector3 normal)
{
// I is the original vector
// N is the normal of the incident plane
- // R = I - (2 * N * ( DotProduct[ I,N] ))
+ // R = I - (2*N*( DotProduct[ I,N] ))
Vector3 result;
float dotProduct = VectorDotProduct(vector, normal);
- result.x = vector.x - (2.0 * normal.x) * dotProduct;
- result.y = vector.y - (2.0 * normal.y) * dotProduct;
- result.z = vector.z - (2.0 * normal.z) * dotProduct;
+ result.x = vector.x - (2.0f*normal.x)*dotProduct;
+ result.y = vector.y - (2.0f*normal.y)*dotProduct;
+ result.z = vector.z - (2.0f*normal.z)*dotProduct;
return result;
}
@@ -308,34 +364,6 @@ RMDEF Vector3 VectorZero(void)
// Module Functions Definition - Matrix math
//----------------------------------------------------------------------------------
-// Converts Matrix to float array
-// NOTE: Returned vector is a transposed version of the Matrix struct,
-// it should be this way because, despite raymath use OpenGL column-major convention,
-// Matrix struct memory alignment and variables naming are not coherent
-RMDEF float *MatrixToFloat(Matrix mat)
-{
- static float buffer[16];
-
- buffer[0] = mat.m0;
- buffer[1] = mat.m4;
- buffer[2] = mat.m8;
- buffer[3] = mat.m12;
- buffer[4] = mat.m1;
- buffer[5] = mat.m5;
- buffer[6] = mat.m9;
- buffer[7] = mat.m13;
- buffer[8] = mat.m2;
- buffer[9] = mat.m6;
- buffer[10] = mat.m10;
- buffer[11] = mat.m14;
- buffer[12] = mat.m3;
- buffer[13] = mat.m7;
- buffer[14] = mat.m11;
- buffer[15] = mat.m15;
-
- return buffer;
-}
-
// Compute matrix determinant
RMDEF float MatrixDeterminant(Matrix mat)
{
@@ -413,7 +441,7 @@ RMDEF void MatrixInvert(Matrix *mat)
float b11 = a22*a33 - a23*a32;
// Calculate the invert determinant (inlined to avoid double-caching)
- float invDet = 1/(b00*b11 - b01*b10 + b02*b09 + b03*b08 - b04*b07 + b05*b06);
+ float invDet = 1.0f/(b00*b11 - b01*b10 + b02*b09 + b03*b08 - b04*b07 + b05*b06);
temp.m0 = (a11*b11 - a12*b10 + a13*b09)*invDet;
temp.m1 = (-a01*b11 + a02*b10 - a03*b09)*invDet;
@@ -461,7 +489,10 @@ RMDEF void MatrixNormalize(Matrix *mat)
// Returns identity matrix
RMDEF Matrix MatrixIdentity(void)
{
- Matrix result = { 1, 0, 0, 0, 0, 1, 0, 0, 0, 0, 1, 0, 0, 0, 0, 1 };
+ Matrix result = { 1.0f, 0.0f, 0.0f, 0.0f,
+ 0.0f, 1.0f, 0.0f, 0.0f,
+ 0.0f, 0.0f, 1.0f, 0.0f,
+ 0.0f, 0.0f, 0.0f, 1.0f };
return result;
}
@@ -519,7 +550,10 @@ RMDEF Matrix MatrixSubstract(Matrix left, Matrix right)
// Returns translation matrix
RMDEF Matrix MatrixTranslate(float x, float y, float z)
{
- Matrix result = { 1, 0, 0, 0, 0, 1, 0, 0, 0, 0, 1, 0, x, y, z, 1 };
+ Matrix result = { 1.0f, 0.0f, 0.0f, 0.0f,
+ 0.0f, 1.0f, 0.0f, 0.0f,
+ 0.0f, 0.0f, 1.0f, 0.0f,
+ x, y, z, 1.0f };
return result;
}
@@ -536,9 +570,9 @@ RMDEF Matrix MatrixRotate(float angle, Vector3 axis)
float length = sqrt(x*x + y*y + z*z);
- if ((length != 1) && (length != 0))
+ if ((length != 1.0f) && (length != 0.0f))
{
- length = 1/length;
+ length = 1.0f/length;
x *= length;
y *= length;
z *= length;
@@ -594,15 +628,15 @@ RMDEF Matrix MatrixRotate(float angle, float x, float y, float z)
m2 = result.m2, m6 = result.m6, m10 = result.m10, m14 = result.m14;
// build rotation matrix
- float r0 = x * x * c1 + c;
- float r1 = x * y * c1 + z * s;
- float r2 = x * z * c1 - y * s;
- float r4 = x * y * c1 - z * s;
- float r5 = y * y * c1 + c;
- float r6 = y * z * c1 + x * s;
- float r8 = x * z * c1 + y * s;
- float r9 = y * z * c1 - x * s;
- float r10= z * z * c1 + c;
+ float r0 = x*x*c1 + c;
+ float r1 = x*y*c1 + z*s;
+ float r2 = x*z*c1 - y*s;
+ float r4 = x*y*c1 - z*s;
+ float r5 = y*y*c1 + c;
+ float r6 = y*z*c1 + x*s;
+ float r8 = x*z*c1 + y*s;
+ float r9 = y*z*c1 - x*s;
+ float r10= z*z*c1 + c;
// multiply rotation matrix
result.m0 = r0*m0 + r4*m1 + r8*m2;
@@ -673,7 +707,10 @@ RMDEF Matrix MatrixRotateZ(float angle)
// Returns scaling matrix
RMDEF Matrix MatrixScale(float x, float y, float z)
{
- Matrix result = { x, 0, 0, 0, 0, y, 0, 0, 0, 0, z, 0, 0, 0, 0, 1 };
+ Matrix result = { x, 0.0f, 0.0f, 0.0f,
+ 0.0f, y, 0.0f, 0.0f,
+ 0.0f, 0.0f, z, 0.0f,
+ 0.0f, 0.0f, 0.0f, 1.0f };
return result;
}
@@ -713,25 +750,25 @@ RMDEF Matrix MatrixFrustum(double left, double right, double bottom, double top,
float tb = (top - bottom);
float fn = (far - near);
- result.m0 = (near*2.0f) / rl;
- result.m1 = 0;
- result.m2 = 0;
- result.m3 = 0;
+ result.m0 = (near*2.0f)/rl;
+ result.m1 = 0.0f;
+ result.m2 = 0.0f;
+ result.m3 = 0.0f;
- result.m4 = 0;
- result.m5 = (near*2.0f) / tb;
- result.m6 = 0;
- result.m7 = 0;
+ result.m4 = 0.0f;
+ result.m5 = (near*2.0f)/tb;
+ result.m6 = 0.0f;
+ result.m7 = 0.0f;
- result.m8 = (right + left) / rl;
- result.m9 = (top + bottom) / tb;
- result.m10 = -(far + near) / fn;
+ result.m8 = (right + left)/rl;
+ result.m9 = (top + bottom)/tb;
+ result.m10 = -(far + near)/fn;
result.m11 = -1.0f;
- result.m12 = 0;
- result.m13 = 0;
- result.m14 = -(far*near*2.0f) / fn;
- result.m15 = 0;
+ result.m12 = 0.0f;
+ result.m13 = 0.0f;
+ result.m14 = -(far*near*2.0f)/fn;
+ result.m15 = 0.0f;
return result;
}
@@ -739,7 +776,7 @@ RMDEF Matrix MatrixFrustum(double left, double right, double bottom, double top,
// Returns perspective projection matrix
RMDEF Matrix MatrixPerspective(double fovy, double aspect, double near, double far)
{
- double top = near*tanf(fovy*PI / 360.0f);
+ double top = near*tanf(fovy*PI/360.0f);
double right = top*aspect;
return MatrixFrustum(-right, right, -top, top, near, far);
@@ -754,22 +791,22 @@ RMDEF Matrix MatrixOrtho(double left, double right, double bottom, double top, d
float tb = (top - bottom);
float fn = (far - near);
- result.m0 = 2 / rl;
- result.m1 = 0;
- result.m2 = 0;
- result.m3 = 0;
- result.m4 = 0;
- result.m5 = 2 / tb;
- result.m6 = 0;
- result.m7 = 0;
- result.m8 = 0;
- result.m9 = 0;
- result.m10 = -2 / fn;
- result.m11 = 0;
- result.m12 = -(left + right) / rl;
- result.m13 = -(top + bottom) / tb;
- result.m14 = -(far + near) / fn;
- result.m15 = 1;
+ result.m0 = 2.0f/rl;
+ result.m1 = 0.0f;
+ result.m2 = 0.0f;
+ result.m3 = 0.0f;
+ result.m4 = 0.0f;
+ result.m5 = 2.0f/tb;
+ result.m6 = 0.0f;
+ result.m7 = 0.0f;
+ result.m8 = 0.0f;
+ result.m9 = 0.0f;
+ result.m10 = -2.0f/fn;
+ result.m11 = 0.0f;
+ result.m12 = -(left + right)/rl;
+ result.m13 = -(top + bottom)/tb;
+ result.m14 = -(far + near)/fn;
+ result.m15 = 1.0f;
return result;
}
@@ -789,19 +826,19 @@ RMDEF Matrix MatrixLookAt(Vector3 eye, Vector3 target, Vector3 up)
result.m0 = x.x;
result.m1 = x.y;
result.m2 = x.z;
- result.m3 = -((x.x * eye.x) + (x.y * eye.y) + (x.z * eye.z));
+ result.m3 = -((x.x*eye.x) + (x.y*eye.y) + (x.z*eye.z));
result.m4 = y.x;
result.m5 = y.y;
result.m6 = y.z;
- result.m7 = -((y.x * eye.x) + (y.y * eye.y) + (y.z * eye.z));
+ result.m7 = -((y.x*eye.x) + (y.y*eye.y) + (y.z*eye.z));
result.m8 = z.x;
result.m9 = z.y;
result.m10 = z.z;
- result.m11 = -((z.x * eye.x) + (z.y * eye.y) + (z.z * eye.z));
- result.m12 = 0;
- result.m13 = 0;
- result.m14 = 0;
- result.m15 = 1;
+ result.m11 = -((z.x*eye.x) + (z.y*eye.y) + (z.z*eye.z));
+ result.m12 = 0.0f;
+ result.m13 = 0.0f;
+ result.m14 = 0.0f;
+ result.m15 = 1.0f;
return result;
}
@@ -834,9 +871,9 @@ RMDEF void QuaternionNormalize(Quaternion *q)
length = QuaternionLength(*q);
- if (length == 0) length = 1;
+ if (length == 0.0f) length = 1.0f;
- ilength = 1.0/length;
+ ilength = 1.0f/length;
q->x *= ilength;
q->y *= ilength;
@@ -882,8 +919,8 @@ RMDEF Quaternion QuaternionSlerp(Quaternion q1, Quaternion q2, float amount)
}
else
{
- float ratioA = sin((1 - amount)*halfTheta) / sinHalfTheta;
- float ratioB = sin(amount*halfTheta) / sinHalfTheta;
+ float ratioA = sin((1 - amount)*halfTheta)/sinHalfTheta;
+ float ratioB = sin(amount*halfTheta)/sinHalfTheta;
result.x = (q1.x*ratioA + q2.x*ratioB);
result.y = (q1.y*ratioA + q2.y*ratioB);
@@ -902,15 +939,15 @@ RMDEF Quaternion QuaternionFromMatrix(Matrix matrix)
float trace = MatrixTrace(matrix);
- if (trace > 0)
+ if (trace > 0.0f)
{
- float s = (float)sqrt(trace + 1) * 2;
- float invS = 1 / s;
+ float s = (float)sqrt(trace + 1)*2.0f;
+ float invS = 1.0f/s;
- result.w = s * 0.25;
- result.x = (matrix.m6 - matrix.m9) * invS;
- result.y = (matrix.m8 - matrix.m2) * invS;
- result.z = (matrix.m1 - matrix.m4) * invS;
+ result.w = s*0.25f;
+ result.x = (matrix.m6 - matrix.m9)*invS;
+ result.y = (matrix.m8 - matrix.m2)*invS;
+ result.z = (matrix.m1 - matrix.m4)*invS;
}
else
{
@@ -918,33 +955,33 @@ RMDEF Quaternion QuaternionFromMatrix(Matrix matrix)
if (m00 > m11 && m00 > m22)
{
- float s = (float)sqrt(1 + m00 - m11 - m22) * 2;
- float invS = 1 / s;
+ float s = (float)sqrt(1.0f + m00 - m11 - m22)*2.0f;
+ float invS = 1.0f/s;
- result.w = (matrix.m6 - matrix.m9) * invS;
- result.x = s * 0.25;
- result.y = (matrix.m4 + matrix.m1) * invS;
- result.z = (matrix.m8 + matrix.m2) * invS;
+ result.w = (matrix.m6 - matrix.m9)*invS;
+ result.x = s*0.25f;
+ result.y = (matrix.m4 + matrix.m1)*invS;
+ result.z = (matrix.m8 + matrix.m2)*invS;
}
else if (m11 > m22)
{
- float s = (float)sqrt(1 + m11 - m00 - m22) * 2;
- float invS = 1 / s;
+ float s = (float)sqrt(1.0f + m11 - m00 - m22)*2.0f;
+ float invS = 1.0f/s;
- result.w = (matrix.m8 - matrix.m2) * invS;
- result.x = (matrix.m4 + matrix.m1) * invS;
- result.y = s * 0.25;
- result.z = (matrix.m9 + matrix.m6) * invS;
+ result.w = (matrix.m8 - matrix.m2)*invS;
+ result.x = (matrix.m4 + matrix.m1)*invS;
+ result.y = s*0.25f;
+ result.z = (matrix.m9 + matrix.m6)*invS;
}
else
{
- float s = (float)sqrt(1 + m22 - m00 - m11) * 2;
- float invS = 1 / s;
+ float s = (float)sqrt(1.0f + m22 - m00 - m11)*2.0f;
+ float invS = 1.0f/s;
- result.w = (matrix.m1 - matrix.m4) * invS;
- result.x = (matrix.m8 + matrix.m2) * invS;
- result.y = (matrix.m9 + matrix.m6) * invS;
- result.z = s * 0.25;
+ result.w = (matrix.m1 - matrix.m4)*invS;
+ result.x = (matrix.m8 + matrix.m2)*invS;
+ result.y = (matrix.m9 + matrix.m6)*invS;
+ result.z = s*0.25f;
}
}
@@ -974,22 +1011,22 @@ RMDEF Matrix QuaternionToMatrix(Quaternion q)
float wy = w*y2;
float wz = w*z2;
- result.m0 = 1 - (yy + zz);
+ result.m0 = 1.0f - (yy + zz);
result.m1 = xy - wz;
result.m2 = xz + wy;
- result.m3 = 0;
+ result.m3 = 0.0f;
result.m4 = xy + wz;
- result.m5 = 1 - (xx + zz);
+ result.m5 = 1.0f - (xx + zz);
result.m6 = yz - wx;
- result.m7 = 0;
+ result.m7 = 0.0f;
result.m8 = xz - wy;
result.m9 = yz + wx;
- result.m10 = 1 - (xx + yy);
- result.m11 = 0;
- result.m12 = 0;
- result.m13 = 0;
- result.m14 = 0;
- result.m15 = 1;
+ result.m10 = 1.0f - (xx + yy);
+ result.m11 = 0.0f;
+ result.m12 = 0.0f;
+ result.m13 = 0.0f;
+ result.m14 = 0.0f;
+ result.m15 = 1.0f;
return result;
}
@@ -998,17 +1035,17 @@ RMDEF Matrix QuaternionToMatrix(Quaternion q)
// NOTE: angle must be provided in radians
RMDEF Quaternion QuaternionFromAxisAngle(float angle, Vector3 axis)
{
- Quaternion result = { 0, 0, 0, 1 };
+ Quaternion result = { 0.0f, 0.0f, 0.0f, 1.0f };
- if (VectorLength(axis) != 0.0)
+ if (VectorLength(axis) != 0.0f)
- angle *= 0.5;
+ angle *= 0.5f;
VectorNormalize(&axis);
- result.x = axis.x * (float)sin(angle);
- result.y = axis.y * (float)sin(angle);
- result.z = axis.z * (float)sin(angle);
+ result.x = axis.x*(float)sin(angle);
+ result.y = axis.y*(float)sin(angle);
+ result.z = axis.z*(float)sin(angle);
result.w = (float)cos(angle);
QuaternionNormalize(&result);
@@ -1021,23 +1058,23 @@ RMDEF void QuaternionToAxisAngle(Quaternion q, float *outAngle, Vector3 *outAxis
{
if (fabs(q.w) > 1.0f) QuaternionNormalize(&q);
- Vector3 resAxis = { 0, 0, 0 };
- float resAngle = 0;
+ Vector3 resAxis = { 0.0f, 0.0f, 0.0f };
+ float resAngle = 0.0f;
- resAngle = 2.0f * (float)acos(q.w);
- float den = (float)sqrt(1.0 - q.w * q.w);
+ resAngle = 2.0f*(float)acos(q.w);
+ float den = (float)sqrt(1.0f - q.w*q.w);
if (den > 0.0001f)
{
- resAxis.x = q.x / den;
- resAxis.y = q.y / den;
- resAxis.z = q.z / den;
+ resAxis.x = q.x/den;
+ resAxis.y = q.y/den;
+ resAxis.z = q.z/den;
}
else
{
// This occurs when the angle is zero.
// Not a problem: just set an arbitrary normalized axis.
- resAxis.x = 1.0;
+ resAxis.x = 1.0f;
}
*outAxis = resAxis;
@@ -1058,5 +1095,6 @@ RMDEF void QuaternionTransform(Quaternion *q, Matrix mat)
q->w = mat.m3*x + mat.m7*y + mat.m11*z + mat.m15*w;
}
-#endif // RAYMATH_DEFINE
-#endif // RAYMATH_H \ No newline at end of file
+#endif // RAYMATH_IMPLEMENTATION
+
+#endif // RAYMATH_H \ No newline at end of file