OpenSceneGraph/include/osg/Math
2006-05-09 09:35:12 +00:00

206 lines
6.1 KiB
C++

/* -*-c++-*- OpenSceneGraph - Copyright (C) 1998-2004 Robert Osfield
*
* This library is open source and may be redistributed and/or modified under
* the terms of the OpenSceneGraph Public License (OSGPL) version 0.0 or
* (at your option) any later version. The full license is in LICENSE file
* included with this distribution, and on the openscenegraph.org website.
*
* This library is distributed in the hope that it will be useful,
* but WITHOUT ANY WARRANTY; without even the implied warranty of
* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
* OpenSceneGraph Public License for more details.
*/
#ifndef __OSG_MATH
#define __OSG_MATH
#include <math.h>
//certain math functions were not defined until 10.2
//so this code checks the version so it can add in workarounds for older versions.
#ifdef __APPLE__
#include <AvailabilityMacros.h>
#if !defined(MAC_OS_X_VERSION_10_2) || (MAC_OS_X_VERSION_MIN_REQUIRED < MAC_OS_X_VERSION_10_2)
#define APPLE_PRE_10_2
#endif
#endif
#if defined(_MSC_VER)
#include <float.h>
#endif
#if defined (sun) || defined (__APPLE__)
#include <float.h>
#ifndef acosf
inline float acosf(float value) { return static_cast<float>(acos(value)); }
#endif
#ifndef asinf
inline float asinf(float value) { return static_cast<float>(asin(value)); }
#endif
#ifndef cosf
inline float cosf(float value) { return static_cast<float>(cos(value)); }
#endif
#ifndef sinf
inline float sinf(float value) { return static_cast<float>(sin(value)); }
#endif
#ifndef logf
inline float logf(float value) { return static_cast<float>(log(value)); }
#endif
#ifndef powf
inline float powf(float value1,float value2) { return static_cast<float>(pow(value1,value2)); }
#endif
#ifndef sqrtf
inline float sqrtf(float value) { return static_cast<float>(sqrt(value)); }
#endif
#ifndef fabsf
inline float fabsf(float value) { return static_cast<float>(fabs(value)); }
#endif
#ifndef atan2f
inline float atan2f(float value1, float value2) { return static_cast<float>(atan2(value1,value2)); }
#endif
#endif
#if defined (sun) || \
defined (__hpux) || \
defined (APPLE_PRE_10_2)
#ifndef floorf
inline float floorf(float value) { return static_cast<float>(floor(value)); }
#endif
#endif
namespace osg {
// define the standard trig values
#ifdef PI
#undef PI
#undef PI_2
#undef PI_4
#endif
const double PI = 3.14159265358979323846;
const double PI_2 = 1.57079632679489661923;
const double PI_4 = 0.78539816339744830962;
/** return the minimum of two values, equivalent to std::min.
* std::min not used because of STL implementation under IRIX not
* containing std::min.
*/
template<typename T>
inline T absolute(T v) { return v<(T)0?-v:v; }
/** return true if float lhs and rhs are equivalent,
* meaning that the difference between them is less than an epsilon value
* which defaults to 1e-6.
*/
inline float equivalent(float lhs,float rhs,float epsilon=1e-6)
{ float delta = rhs-lhs; return delta<0.0f?delta>=-epsilon:delta<=epsilon; }
/** return true if double lhs and rhs are equivalent,
* meaning that the difference between them is less than an epsilon value
* which defaults to 1e-6.
*/
inline double equivalent(double lhs,double rhs,double epsilon=1e-6)
{ double delta = rhs-lhs; return delta<0.0?delta>=-epsilon:delta<=epsilon; }
/** return the minimum of two values, equivilent to std::min.
* std::min not used because of STL implementation under IRIX not containing
* std::min.
*/
template<typename T>
inline T minimum(T lhs,T rhs) { return lhs<rhs?lhs:rhs; }
/** return the maximum of two values, equivilent to std::max.
* std::max not used because of STL implementation under IRIX not containing
* std::max.
*/
template<typename T>
inline T maximum(T lhs,T rhs) { return lhs>rhs?lhs:rhs; }
template<typename T>
inline T clampTo(T v,T minimum,T maximum)
{ return v<minimum?minimum:v>maximum?maximum:v; }
template<typename T>
inline T clampAbove(T v,T minimum) { return v<minimum?minimum:v; }
template<typename T>
inline T clampBelow(T v,T maximum) { return v>maximum?maximum:v; }
template<typename T>
inline T clampBetween(T v,T minimum, T maximum)
{ return clampBelow(clampAbove(v,minimum),maximum); }
template<typename T>
inline T sign(T v) { return v<(T)0?(T)-1:(T)1; }
template<typename T>
inline T square(T v) { return v*v; }
template<typename T>
inline T signedSquare(T v) { return v<(T)0?-v*v:v*v;; }
inline float inDegrees(float angle) { return angle*(float)PI/180.0f; }
inline double inDegrees(double angle) { return angle*PI/180.0; }
template<typename T>
inline T inRadians(T angle) { return angle; }
inline float DegreesToRadians(float angle) { return angle*(float)PI/180.0f; }
inline double DegreesToRadians(double angle) { return angle*PI/180.0; }
inline float RadiansToDegrees(float angle) { return angle*180.0f/(float)PI; }
inline double RadiansToDegrees(double angle) { return angle*180.0/PI; }
inline float round(float v) { return v>=0.0f?floorf(v+0.5f):ceilf(v-0.5f); }
inline double round(double v) { return v>=0.0?floor(v+0.5):ceil(v-0.5); }
#if defined(WIN32) && !defined(__CYGWIN__) && !defined(__MWERKS__)
inline bool isNaN(float v) { return _isnan(v)!=0; }
inline bool isNaN(double v) { return _isnan(v)!=0; }
#else
# if defined(__APPLE__) && !defined (APPLE_PRE_10_2)
inline bool isNaN(float v) { return __isnanf(v); }
inline bool isNaN(double v) { return __isnand(v); }
#else
inline bool isNaN(float v) { return isnan(v); }
inline bool isNaN(double v) { return isnan(v); }
#endif
#endif
/** compute the volume of a tetrahedron. */
template<typename T>
inline float computeVolume(const T& a,const T& b,const T& c,const T& d)
{
return fabsf(((b-c)^(a-b))*(d-b));
}
/** compute the volume of a prism. */
template<typename T>
inline float computeVolume(const T& f1,const T& f2,const T& f3,
const T& b1,const T& b2,const T& b3)
{
return computeVolume(f1,f2,f3,b1)+
computeVolume(b1,b2,b3,f2)+
computeVolume(b1,b3,f2,f3);
}
}
#endif // __OSG_MATH