Add a wxRound() overload for float

This commit is contained in:
Paul Cornett
2020-10-14 12:03:02 -07:00
parent 0825b59553
commit c7122a8c0d

View File

@@ -62,14 +62,6 @@
#define wxFinite(x) std::isfinite(x) #define wxFinite(x) std::isfinite(x)
#define wxIsNaN(x) std::isnan(x) #define wxIsNaN(x) std::isnan(x)
inline int wxRound(double x)
{
wxASSERT_MSG(x > (double)INT_MIN - 0.5 && x < (double)INT_MAX + 0.5,
wxT("argument out of supported range"));
return (int)std::lround(x);
}
#else /* C++98 */ #else /* C++98 */
#if defined(__VISUALC__) || defined(__BORLANDC__) #if defined(__VISUALC__) || defined(__BORLANDC__)
@@ -112,18 +104,6 @@
#define wxIsNaN(x) ((x) != (x)) #define wxIsNaN(x) ((x) != (x))
#endif #endif
inline int wxRound(double x)
{
wxASSERT_MSG(x > (double)INT_MIN - 0.5 && x < (double)INT_MAX + 0.5,
wxT("argument out of supported range"));
#if defined(HAVE_ROUND) || wxCHECK_VISUALC_VERSION(12)
return int(lround(x));
#else
return (int)(x < 0 ? x - 0.5 : x + 0.5);
#endif
}
#endif /* C++11/C++98 */ #endif /* C++11/C++98 */
#ifdef __INTELC__ #ifdef __INTELC__
@@ -154,6 +134,34 @@
inline bool wxIsNullDouble(double x) { return wxIsSameDouble(x, 0.); } inline bool wxIsNullDouble(double x) { return wxIsSameDouble(x, 0.); }
inline int wxRound(double x)
{
wxASSERT_MSG(x > double(INT_MIN) - 0.5 && x < double(INT_MAX) + 0.5,
"argument out of supported range");
#if __cplusplus >= 201103
return int(std::lround(x));
#elif defined(HAVE_ROUND) || wxCHECK_VISUALC_VERSION(12)
return int(lround(x));
#else
return int(x < 0 ? x - 0.5 : x + 0.5);
#endif
}
inline int wxRound(float x)
{
wxASSERT_MSG(x > float(INT_MIN) && x < float(INT_MAX),
"argument out of supported range");
#if __cplusplus >= 201103
return int(std::lround(x));
#elif defined(HAVE_ROUND) || wxCHECK_VISUALC_VERSION(12)
return int(lroundf(x));
#else
return int(x < 0.0f ? x - 0.5f : x + 0.5f);
#endif
}
// Convert between degrees and radians. // Convert between degrees and radians.
inline double wxDegToRad(double deg) { return (deg * M_PI) / 180.0; } inline double wxDegToRad(double deg) { return (deg * M_PI) / 180.0; }
inline double wxRadToDeg(double rad) { return (rad * 180.0) / M_PI; } inline double wxRadToDeg(double rad) { return (rad * 180.0) / M_PI; }