Now that we have apple/swift-numerics officially, you should switch to ElementaryFunctions protocol of RealModule of Numerics.
Protocol-Oriented Math Functions for Swift
This module provides a protocol FloatingPointMath. Conforming protocols MUST implemnt:
public protocol FloatingPointMath {
init (_:Double) // BinaryFloatingPoint already has one
var asDouble:Double { get } // you have to add it yourself
}
And SHOULD implement (though default implementations are provided):
static func acos (_ x:Self)->Self
static func acosh(_ x:Self)->Self
static func asin (_ x:Self)->Self
static func asinh(_ x:Self)->Self
static func atan (_ x:Self)->Self
static func atanh(_ x:Self)->Self
static func cbrt (_ x:Self)->Self
static func cos (_ x:Self)->Self
static func cosh (_ x:Self)->Self
static func exp (_ x:Self)->Self
static func exp2 (_ x:Self)->Self
static func expm1(_ x:Self)->Self
static func log (_ x:Self)->Self
static func log2 (_ x:Self)->Self
static func log10(_ x:Self)->Self
static func log1p(_ x:Self)->Self
static func sin (_ x:Self)->Self
static func sinh (_ x:Self)->Self
static func sqrt (_ x:Self)->Self
static func tan (_ x:Self)->Self
static func tanh (_ x:Self)->Self
static func atan2(_ y:Self, _ x:Self)->Self
static func hypot(_ x:Self, _ y:Self)->Self
static func pow (_ x:Self, _ y:Self)->Self
static func erf (_ x:Self)->Self
static func erfc (_ x:Self)->Self
static func lgamma(_ x:Self)->Self
static func tgamma(_ x:Self)->Self
In short, conforming types are guaranteed to have math functions as (static|class) methods. For convenience, Double and Float are made FloatingPointMath when you import.
Unlike Foundation (or Glibc or Darwin), this module:
- does nothing further. Just math functions found in
libm. - imports these functions as (static|class) methods to avoid namespace collision.
import Foundation atan2(0.0, -1.0) == M_PI // imported in the global namespace with lots of other symbols
import FloatingPointMath Double.atan2(0.0, -1.0) == Double.pi // explicitly under `Double`.
If the type already conform to BinaryFloatingPoint, all you have to do is add .asDouble as follows:
extension CGFloat : FloatingPointMath {
public var asDouble:Double { return Double(self) }
}Default implementations just convert to Double, do the calculation and convert the result back like the following.
public static func acos (_ x:Self)->Self {
return Self(Darwin.acos (x.asDouble))
}
So the result is only as accurate as Double. For more accurate floating-point types (say, Float128 or BigFloat), you should implement your own.
Add the following to the Package.swift file of your project.
-
dependencies:.package(url: "https://github.com/dankogai/swift-floatingpointmath.git", from: "0.0.7")
-
.targetintargets:dependencies:["FloatingPointMath"]