/** * Name: wx/math.h * Purpose: Declarations/definitions of common math functions * Author: John Labenski and others * Modified by: * Created: 02/02/03 * RCS-ID: $Id$ * Copyright: (c) John Labenski * Licence: wxWindows licence */ /* THIS IS A C FILE, DON'T USE C++ FEATURES (IN PARTICULAR COMMENTS) IN IT */ #ifndef _WX_MATH_H_ #define _WX_MATH_H_ #include "wx/defs.h" #include #ifndef M_PI #define M_PI 3.1415926535897932384626433832795 #endif /* Scaling factors for various unit conversions: 1 inch = 2.54 cm */ #ifndef METRIC_CONVERSION_CONSTANT #define METRIC_CONVERSION_CONSTANT (1./254) #endif #ifndef mm2inches #define mm2inches (METRIC_CONVERSION_CONSTANT) #endif #ifndef inches2mm #define inches2mm (1/(mm2inches)) #endif #ifndef mm2twips #define mm2twips (METRIC_CONVERSION_CONSTANT*1440) #endif #ifndef twips2mm #define twips2mm (1/(mm2twips)) #endif #ifndef mm2pt #define mm2pt (METRIC_CONVERSION_CONSTANT*72) #endif #ifndef pt2mm #define pt2mm (1/(mm2pt)) #endif /* unknown __VISAGECC__, __SYMANTECCC__ */ #if defined(__VISUALC__) || defined(__BORLANDC__) || defined(__WATCOMC__) #include #define wxFinite(x) _finite(x) #elif defined(__GNUG__)||defined(__GNUWIN32__)||defined(__DJGPP__)|| \ defined(__SGI_CC__)||defined(__SUNCC__)||defined(__XLC__)|| \ defined(__HPUX__)||defined(__MWERKS__) #define wxFinite(x) finite(x) #else #define wxFinite(x) ((x) == (x)) #endif #if defined(__VISUALC__)||defined(__BORLAND__) #define wxIsNaN(x) _isnan(x) #elif defined(__GNUG__)||defined(__GNUWIN32__)||defined(__DJGPP__)|| \ defined(__SGI_CC__)||defined(__SUNCC__)||defined(__XLC__)|| \ defined(__HPUX__)||defined(__MWERKS__) #define wxIsNaN(x) isnan(x) #else #define wxIsNaN(x) ((x) != (x)) #endif #ifdef __cplusplus #ifdef __INTELC__ inline bool wxIsSameDouble(double x, double y) { // VZ: this warning, given for operators==() and !=() is not wrong, as == // shouldn't be used with doubles, but we get too many of them and // removing these operators is probably not a good idea // // Maybe we should alway compare doubles up to some "epsilon" precision #pragma warning(push) // floating-point equality and inequality comparisons are unreliable #pragma warning(disable: 1572) return x == y; #pragma warning(pop) } #else /* !__INTELC__ */ inline bool wxIsSameDouble(double x, double y) { return x == y; } #endif /* __INTELC__/!__INTELC__ */ inline bool wxIsNullDouble(double x) { return wxIsSameDouble(x, 0.); } inline int wxRound(double x) { wxASSERT_MSG( x > INT_MIN - 0.5 && x < INT_MAX + 0.5, _T("argument out of supported range") ); #if defined(HAVE_ROUND) return int(round(x)); #else return (int)(x < 0 ? x - 0.5 : x + 0.5); #endif } #endif /* __cplusplus */ #if defined(__WXMSW__) && !defined(__WXWINCE__) #define wxMulDivInt32( a , b , c ) ::MulDiv( a , b , c ) #elif defined( __WXMAC__ ) #define wxMulDivInt32( a , b , c ) ( (wxInt32) ( ( (wxInt64)(a) * (wxInt64)(b) ) / (wxInt64)(c) ) ) #else #define wxMulDivInt32( a , b , c ) ((wxInt32)((a)*(((wxDouble)b)/((wxDouble)c)))) #endif #if wxUSE_APPLE_IEEE #ifdef __cplusplus extern "C" { #endif /* functions from common/extended.c */ WXDLLIMPEXP_BASE wxFloat64 wxConvertFromIeeeExtended(const wxInt8 *bytes); WXDLLIMPEXP_BASE void wxConvertToIeeeExtended(wxFloat64 num, wxInt8 *bytes); /* use wxConvertFromIeeeExtended() and wxConvertToIeeeExtended() instead */ #if WXWIN_COMPATIBILITY_2_8 wxDEPRECATED( WXDLLIMPEXP_BASE wxFloat64 ConvertFromIeeeExtended(const wxInt8 *bytes) ); wxDEPRECATED( WXDLLIMPEXP_BASE void ConvertToIeeeExtended(wxFloat64 num, wxInt8 *bytes) ); #endif // WXWIN_COMPATIBILITY_2_8 #ifdef __cplusplus } #endif #endif /* wxUSE_APPLE_IEEE */ #endif /* _WX_MATH_H_ */