/** * @file v2math.h * @brief LLVector2 class header file. * * $LicenseInfo:firstyear=2000&license=viewergpl$ * * Copyright (c) 2000-2007, Linden Research, Inc. * * Second Life Viewer Source Code * The source code in this file ("Source Code") is provided by Linden Lab * to you under the terms of the GNU General Public License, version 2.0 * ("GPL"), unless you have obtained a separate licensing agreement * ("Other License"), formally executed by you and Linden Lab. Terms of * the GPL can be found in doc/GPL-license.txt in this distribution, or * online at http://secondlife.com/developers/opensource/gplv2 * * There are special exceptions to the terms and conditions of the GPL as * it is applied to this Source Code. View the full text of the exception * in the file doc/FLOSS-exception.txt in this software distribution, or * online at http://secondlife.com/developers/opensource/flossexception * * By copying, modifying or distributing this software, you acknowledge * that you have read and understood your obligations described above, * and agree to abide by those obligations. * * ALL LINDEN LAB SOURCE CODE IS PROVIDED "AS IS." LINDEN LAB MAKES NO * WARRANTIES, EXPRESS, IMPLIED OR OTHERWISE, REGARDING ITS ACCURACY, * COMPLETENESS OR PERFORMANCE. * $/LicenseInfo$ */ #ifndef LL_V2MATH_H #define LL_V2MATH_H #include "llmath.h" class LLVector4; class LLMatrix3; class LLQuaternion; // Llvector2 = |x y z w| static const U32 LENGTHOFVECTOR2 = 2; class LLVector2 { public: F32 mV[LENGTHOFVECTOR2]; static LLVector2 zero; LLVector2(); // Initializes LLVector2 to (0, 0) LLVector2(F32 x, F32 y); // Initializes LLVector2 to (x. y) LLVector2(const F32 *vec); // Initializes LLVector2 to (vec[0]. vec[1]) // Clears LLVector2 to (0, 0). DEPRECATED - prefer zeroVec. void clear(); void setZero(); void clearVec(); // deprecated void zeroVec(); // deprecated void set(F32 x, F32 y); // Sets LLVector2 to (x, y) void set(const LLVector2 &vec); // Sets LLVector2 to vec void set(const F32 *vec); // Sets LLVector2 to vec void setVec(F32 x, F32 y); // deprecated void setVec(const LLVector2 &vec); // deprecated void setVec(const F32 *vec); // deprecated F32 length() const; // Returns magnitude of LLVector2 F32 lengthSquared() const; // Returns magnitude squared of LLVector2 F32 normalize(); // Normalizes and returns the magnitude of LLVector2 F32 magVec() const; // deprecated F32 magVecSquared() const; // deprecated F32 normVec(); // deprecated BOOL abs(); // sets all values to absolute value of original value (first octant), returns TRUE if changed const LLVector2& scaleVec(const LLVector2& vec); // scales per component by vec BOOL isNull(); // Returns TRUE if vector has a _very_small_ length BOOL isExactlyZero() const { return !mV[VX] && !mV[VY]; } F32 operator[](int idx) const { return mV[idx]; } F32 &operator[](int idx) { return mV[idx]; } friend bool operator<(const LLVector2 &a, const LLVector2 &b); // For sorting. x is "more significant" than y friend LLVector2 operator+(const LLVector2 &a, const LLVector2 &b); // Return vector a + b friend LLVector2 operator-(const LLVector2 &a, const LLVector2 &b); // Return vector a minus b friend F32 operator*(const LLVector2 &a, const LLVector2 &b); // Return a dot b friend LLVector2 operator%(const LLVector2 &a, const LLVector2 &b); // Return a cross b friend LLVector2 operator/(const LLVector2 &a, F32 k); // Return a divided by scaler k friend LLVector2 operator*(const LLVector2 &a, F32 k); // Return a times scaler k friend LLVector2 operator*(F32 k, const LLVector2 &a); // Return a times scaler k friend bool operator==(const LLVector2 &a, const LLVector2 &b); // Return a == b friend bool operator!=(const LLVector2 &a, const LLVector2 &b); // Return a != b friend const LLVector2& operator+=(LLVector2 &a, const LLVector2 &b); // Return vector a + b friend const LLVector2& operator-=(LLVector2 &a, const LLVector2 &b); // Return vector a minus b friend const LLVector2& operator%=(LLVector2 &a, const LLVector2 &b); // Return a cross b friend const LLVector2& operator*=(LLVector2 &a, F32 k); // Return a times scaler k friend const LLVector2& operator/=(LLVector2 &a, F32 k); // Return a divided by scaler k friend LLVector2 operator-(const LLVector2 &a); // Return vector -a friend std::ostream& operator<<(std::ostream& s, const LLVector2 &a); // Stream a }; // Non-member functions F32 angle_between(const LLVector2 &a, const LLVector2 &b); // Returns angle (radians) between a and b BOOL are_parallel(const LLVector2 &a, const LLVector2 &b, F32 epsilon=F_APPROXIMATELY_ZERO); // Returns TRUE if a and b are very close to parallel F32 dist_vec(const LLVector2 &a, const LLVector2 &b); // Returns distance between a and b F32 dist_vec_squared(const LLVector2 &a, const LLVector2 &b);// Returns distance sqaured between a and b F32 dist_vec_squared2D(const LLVector2 &a, const LLVector2 &b);// Returns distance sqaured between a and b ignoring Z component LLVector2 lerp(const LLVector2 &a, const LLVector2 &b, F32 u); // Returns a vector that is a linear interpolation between a and b // Constructors inline LLVector2::LLVector2(void) { mV[VX] = 0.f; mV[VY] = 0.f; } inline LLVector2::LLVector2(F32 x, F32 y) { mV[VX] = x; mV[VY] = y; } inline LLVector2::LLVector2(const F32 *vec) { mV[VX] = vec[VX]; mV[VY] = vec[VY]; } // Clear and Assignment Functions inline void LLVector2::clear(void) { mV[VX] = 0.f; mV[VY] = 0.f; } inline void LLVector2::setZero(void) { mV[VX] = 0.f; mV[VY] = 0.f; } // deprecated inline void LLVector2::clearVec(void) { mV[VX] = 0.f; mV[VY] = 0.f; } // deprecated inline void LLVector2::zeroVec(void) { mV[VX] = 0.f; mV[VY] = 0.f; } inline void LLVector2::set(F32 x, F32 y) { mV[VX] = x; mV[VY] = y; } inline void LLVector2::set(const LLVector2 &vec) { mV[VX] = vec.mV[VX]; mV[VY] = vec.mV[VY]; } inline void LLVector2::set(const F32 *vec) { mV[VX] = vec[VX]; mV[VY] = vec[VY]; } // deprecated inline void LLVector2::setVec(F32 x, F32 y) { mV[VX] = x; mV[VY] = y; } // deprecated inline void LLVector2::setVec(const LLVector2 &vec) { mV[VX] = vec.mV[VX]; mV[VY] = vec.mV[VY]; } // deprecated inline void LLVector2::setVec(const F32 *vec) { mV[VX] = vec[VX]; mV[VY] = vec[VY]; } // LLVector2 Magnitude and Normalization Functions inline F32 LLVector2::length(void) const { return fsqrtf(mV[0]*mV[0] + mV[1]*mV[1]); } inline F32 LLVector2::lengthSquared(void) const { return mV[0]*mV[0] + mV[1]*mV[1]; } inline F32 LLVector2::normalize(void) { F32 mag = fsqrtf(mV[0]*mV[0] + mV[1]*mV[1]); F32 oomag; if (mag > FP_MAG_THRESHOLD) { oomag = 1.f/mag; mV[0] *= oomag; mV[1] *= oomag; } else { mV[0] = 0.f; mV[1] = 0.f; mag = 0; } return (mag); } // deprecated inline F32 LLVector2::magVec(void) const { return fsqrtf(mV[0]*mV[0] + mV[1]*mV[1]); } // deprecated inline F32 LLVector2::magVecSquared(void) const { return mV[0]*mV[0] + mV[1]*mV[1]; } // deprecated inline F32 LLVector2::normVec(void) { F32 mag = fsqrtf(mV[0]*mV[0] + mV[1]*mV[1]); F32 oomag; if (mag > FP_MAG_THRESHOLD) { oomag = 1.f/mag; mV[0] *= oomag; mV[1] *= oomag; } else { mV[0] = 0.f; mV[1] = 0.f; mag = 0; } return (mag); } inline const LLVector2& LLVector2::scaleVec(const LLVector2& vec) { mV[VX] *= vec.mV[VX]; mV[VY] *= vec.mV[VY]; return *this; } inline BOOL LLVector2::isNull() { if ( F_APPROXIMATELY_ZERO > mV[VX]*mV[VX] + mV[VY]*mV[VY] ) { return TRUE; } return FALSE; } // LLVector2 Operators // For sorting. By convention, x is "more significant" than y. inline bool operator<(const LLVector2 &a, const LLVector2 &b) { if( a.mV[VX] == b.mV[VX] ) { return a.mV[VY] < b.mV[VY]; } else { return a.mV[VX] < b.mV[VX]; } } inline LLVector2 operator+(const LLVector2 &a, const LLVector2 &b) { LLVector2 c(a); return c += b; } inline LLVector2 operator-(const LLVector2 &a, const LLVector2 &b) { LLVector2 c(a); return c -= b; } inline F32 operator*(const LLVector2 &a, const LLVector2 &b) { return (a.mV[0]*b.mV[0] + a.mV[1]*b.mV[1]); } inline LLVector2 operator%(const LLVector2 &a, const LLVector2 &b) { return LLVector2(a.mV[0]*b.mV[1] - b.mV[0]*a.mV[1], a.mV[1]*b.mV[0] - b.mV[1]*a.mV[0]); } inline LLVector2 operator/(const LLVector2 &a, F32 k) { F32 t = 1.f / k; return LLVector2( a.mV[0] * t, a.mV[1] * t ); } inline LLVector2 operator*(const LLVector2 &a, F32 k) { return LLVector2( a.mV[0] * k, a.mV[1] * k ); } inline LLVector2 operator*(F32 k, const LLVector2 &a) { return LLVector2( a.mV[0] * k, a.mV[1] * k ); } inline bool operator==(const LLVector2 &a, const LLVector2 &b) { return ( (a.mV[0] == b.mV[0]) &&(a.mV[1] == b.mV[1])); } inline bool operator!=(const LLVector2 &a, const LLVector2 &b) { return ( (a.mV[0] != b.mV[0]) ||(a.mV[1] != b.mV[1])); } inline const LLVector2& operator+=(LLVector2 &a, const LLVector2 &b) { a.mV[0] += b.mV[0]; a.mV[1] += b.mV[1]; return a; } inline const LLVector2& operator-=(LLVector2 &a, const LLVector2 &b) { a.mV[0] -= b.mV[0]; a.mV[1] -= b.mV[1]; return a; } inline const LLVector2& operator%=(LLVector2 &a, const LLVector2 &b) { LLVector2 ret(a.mV[0]*b.mV[1] - b.mV[0]*a.mV[1], a.mV[1]*b.mV[0] - b.mV[1]*a.mV[0]); a = ret; return a; } inline const LLVector2& operator*=(LLVector2 &a, F32 k) { a.mV[0] *= k; a.mV[1] *= k; return a; } inline const LLVector2& operator/=(LLVector2 &a, F32 k) { F32 t = 1.f / k; a.mV[0] *= t; a.mV[1] *= t; return a; } inline LLVector2 operator-(const LLVector2 &a) { return LLVector2( -a.mV[0], -a.mV[1] ); } inline void update_min_max(LLVector2& min, LLVector2& max, const LLVector2& pos) { for (U32 i = 0; i < 2; i++) { if (min.mV[i] > pos.mV[i]) { min.mV[i] = pos.mV[i]; } if (max.mV[i] < pos.mV[i]) { max.mV[i] = pos.mV[i]; } } } inline std::ostream& operator<<(std::ostream& s, const LLVector2 &a) { s << "{ " << a.mV[VX] << ", " << a.mV[VY] << " }"; return s; } #endif