func roundFloat64(x exact.Value) exact.Value {
f, _ := exact.Float64Val(x)
if !math.IsInf(f, 0) {
return exact.MakeFloat64(f)
}
return nil
}
func fitsFloat32(x exact.Value) bool {
f, _ := exact.Float64Val(x)
// spec: "In all non-constant conversions involving floating-point
// or complex values, if the result type cannot represent the value
// the conversion succeeds but the result value is implementation-
// dependent."
//
// We assume that float32(f) returns an Inf if f is too large for
// a float32, or if f is an Inf; and that it returns 0 for values
// with too small a magnitude.
return !math.IsInf(float64(float32(f)), 0)
}
func fitsFloat64(x exact.Value) bool {
f, _ := exact.Float64Val(x)
return !math.IsInf(f, 0)
}