/* LIBGIMP - The GIMP Library
* Copyright (C) 1995-1997 Peter Mattis and Spencer Kimball
*
* gimpmath.h
*
* This library is free software: you can redistribute it and/or
* modify it under the terms of the GNU Lesser General Public
* License as published by the Free Software Foundation; either
* version 3 of the License, or (at your option) any later version.
*
* 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 GNU
* Lesser General Public License for more details.
*
* You should have received a copy of the GNU Lesser General Public
* License along with this library. If not, see
* .
*/
#ifndef __GIMP_MATH_H__
#define __GIMP_MATH_H__
#include
#ifdef HAVE_IEEEFP_H
#include
#endif
#ifdef G_OS_WIN32
#include
#endif
#define __GIMP_MATH_H_INSIDE__
#include
#include
#include
#undef __GIMP_MATH_H_INSIDE__
G_BEGIN_DECLS
/**
* SECTION: gimpmath
* @title: GimpMath
* @short_description: Mathematical definitions and macros.
*
* Mathematical definitions and macros for use both by the GIMP
* application and plug-ins. These macros should be used rather than
* the ones from <math.h> for enhanced portability.
**/
/**
* RINT:
* @x: the value to be rounded
*
* This macro rounds its argument @x to an integer value in floating
* point format. Use RINT() instead of rint().
**/
#if defined (HAVE_RINT) && 0
/* note: rint() depends on the current floating-point rounding mode. when the
* rounding mode is FE_TONEAREST, it, in parctice, breaks ties to even. this
* is different from 'floor (x + 0.5)', which breaks ties up. in other words
* 'rint (2.5) == 2.0', while 'floor (2.5 + 0.5) == 3.0'. this is asking for
* trouble, so let's just use the latter.
*/
#define RINT(x) rint(x)
#else
#define RINT(x) floor ((x) + 0.5)
#endif
/**
* ROUND:
* @x: the value to be rounded.
*
* This macro rounds its positive argument @x to the nearest integer.
**/
#define ROUND(x) ((int) ((x) + 0.5))
/**
* SIGNED_ROUND:
* @x: the value to be rounded.
*
* This macro rounds its argument @x to the nearest integer.
**/
#define SIGNED_ROUND(x) ((int) RINT (x))
/**
* SQR:
* @x: the value to be squared.
*
* This macro squares its argument @x.
**/
#define SQR(x) ((x) * (x))
/**
* MAX255:
* @a: the value to be limited.
*
* This macro limits it argument @a, an (0-511) int, to 255.
**/
#define MAX255(a) ((a) | (((a) & 256) - (((a) & 256) >> 8)))
/**
* CLAMP0255:
* @a: the value to be clamped.
*
* This macro clamps its argument @a, an int32-range int, between 0
* and 255 inclusive.
**/
#define CLAMP0255(a) CLAMP(a,0,255)
/**
* SAFE_CLAMP:
* @x: the value to be limited.
* @low: the lower limit.
* @high: the upper limit.
*
* Ensures that @x is between the limits set by @low and @high,
* even if @x is NaN. If @low is greater than @high, or if either
* of them is NaN, the result is undefined.
*
* Since: 2.10
**/
#define SAFE_CLAMP(x, low, high) ((x) > (low) ? (x) < (high) ? (x) : (high) : (low))
/**
* gimp_deg_to_rad:
* @angle: the angle to be converted.
*
* This macro converts its argument @angle from degree to radian.
**/
#define gimp_deg_to_rad(angle) ((angle) * (2.0 * G_PI) / 360.0)
/**
* gimp_rad_to_deg:
* @angle: the angle to be converted.
*
* This macro converts its argument @angle from radian to degree.
**/
#define gimp_rad_to_deg(angle) ((angle) * 360.0 / (2.0 * G_PI))
G_END_DECLS
#endif /* __GIMP_MATH_H__ */