Dun*_*n C 6 sorting generics swift keypaths
似乎应该可以使用KeyPaths 数组作为排序键,以使用任意数量的排序键对Swift结构数组进行排序。从概念上讲,这很简单。您定义了一个通用对象的KeyPath数组,其中唯一的限制是keypath中的属性为Comparable。
只要所有KeyPath都指向相同类型的属性,就可以了。但是,一旦您尝试使用指向数组中不同类型元素的KeyPath,它就会停止工作。
请参见下面的代码。我创建了一个具有2个Int属性和一个double属性的简单结构。我为实现函数的Array创建扩展,sortedByKeypaths(_:)
该函数指定可比较的通用类型PROPERTY。它需要一个kepath数组到某个对象元素,该对象指定了PROPERTY类型的属性。(可比较的属性。)
只要您使用一组KeyPath来调用具有相同类型的属性的函数,它就可以完美地工作。
但是,如果您尝试将键路径数组传递给不同类型的属性,则会引发错误“无法将类型'[PartialKeyPath]'的值转换为预期参数类型'[KeyPath]'”
由于数组包含异构键路径,因此由于类型擦除,数组会演变为类型为[[PartialKeyPath]],并且您不能使用PartialKeyPath从数组中获取元素。
有解决这个问题的方法吗?无法使用异构的KeyPath数组似乎严重限制了Swift KeyPath的用途
import UIKit
struct Stuff {
let value: Int
let value2: Int
let doubleValue: Double
}
extension Array {
func sortedByKeypaths<PROPERTY: Comparable>(_ keypaths: [KeyPath<Element, PROPERTY>]) -> [Element] {
return self.sorted { lhs, rhs in
var keypaths = keypaths
while !keypaths.isEmpty {
let keypath = keypaths.removeFirst()
if lhs[keyPath: keypath] != rhs[keyPath: keypath] {
return lhs[keyPath: keypath] < rhs[keyPath: keypath]
}
}
return true
}
}
}
var stuff = [Stuff]()
for _ in 1...20 {
stuff.append(Stuff(value: Int(arc4random_uniform(5)),
value2: Int(arc4random_uniform(5)),
doubleValue: Double(arc4random_uniform(10))))
}
let sortedStuff = stuff.sortedByKeypaths([\Stuff.value, \Stuff.value2]) //This works
sortedStuff.forEach { print($0) }
let moreSortedStuff = stuff.sortedByKeypaths([\Stuff.value, \Stuff.doubleValue]) //This throws a compiler error
moreSortedStuff.forEach { print($0) }
Run Code Online (Sandbox Code Playgroud)
使用部分键路径数组的问题在于,您不能保证属性类型为Comparable。一种可能的解决方案是使用类型擦除包装器,以擦除键路径的值类型,同时确保它是Comparable:
struct PartialComparableKeyPath<Root> {
private let _isEqual: (Root, Root) -> Bool
private let _isLessThan: (Root, Root) -> Bool
init<Value : Comparable>(_ keyPath: KeyPath<Root, Value>) {
self._isEqual = { $0[keyPath: keyPath] == $1[keyPath: keyPath] }
self._isLessThan = { $0[keyPath: keyPath] < $1[keyPath: keyPath] }
}
func isEqual(_ lhs: Root, _ rhs: Root) -> Bool {
return _isEqual(lhs, rhs)
}
func isLessThan(_ lhs: Root, _ rhs: Root) -> Bool {
return _isLessThan(lhs, rhs)
}
}
Run Code Online (Sandbox Code Playgroud)
然后,您可以将排序功能实现为:
extension Sequence {
func sorted(by keyPaths: PartialComparableKeyPath<Element>...) -> [Element] {
return sorted { lhs, rhs in
for keyPath in keyPaths {
if !keyPath.isEqual(lhs, rhs) {
return keyPath.isLessThan(lhs, rhs)
}
}
return false
}
}
}
Run Code Online (Sandbox Code Playgroud)
然后像这样使用:
struct Stuff {
let value: Int
let value2: Int
let doubleValue: Double
}
var stuff = [Stuff]()
for _ in 1 ... 20 {
stuff.append(Stuff(value: Int(arc4random_uniform(5)),
value2: Int(arc4random_uniform(5)),
doubleValue: Double(arc4random_uniform(10))))
}
let sortedStuff = stuff.sorted(by: PartialComparableKeyPath(\.value),
PartialComparableKeyPath(\.value2))
sortedStuff.forEach { print($0) }
let moreSortedStuff = stuff.sorted(by: PartialComparableKeyPath(\.value),
PartialComparableKeyPath(\.doubleValue))
moreSortedStuff.forEach { print($0) }
Run Code Online (Sandbox Code Playgroud)
虽然很不幸,但这需要将每个单独的键路径包装在一个PartialComparableKeyPath值中,以便捕获和清除键路径的值类型,这并不是特别漂亮。
实际上,我们这里需要的功能是可变参数泛型,它可以让您在键路径的值类型的可变数量的泛型占位符上定义函数,每个占位符都限于Comparable。
在此之前,另一种选择是只写给定数量的重载,以比较不同数量的键路径:
extension Sequence {
func sorted<A : Comparable>(by keyPathA: KeyPath<Element, A>) -> [Element] {
return sorted { lhs, rhs in
lhs[keyPath: keyPathA] < rhs[keyPath: keyPathA]
}
}
func sorted<A : Comparable, B : Comparable>
(by keyPathA: KeyPath<Element, A>, _ keyPathB: KeyPath<Element, B>) -> [Element] {
return sorted { lhs, rhs in
(lhs[keyPath: keyPathA], lhs[keyPath: keyPathB]) <
(rhs[keyPath: keyPathA], rhs[keyPath: keyPathB])
}
}
func sorted<A : Comparable, B : Comparable, C : Comparable>
(by keyPathA: KeyPath<Element, A>, _ keyPathB: KeyPath<Element, B>, _ keyPathC: KeyPath<Element, C>) -> [Element] {
return sorted { lhs, rhs in
(lhs[keyPath: keyPathA], lhs[keyPath: keyPathB], lhs[keyPath: keyPathC]) <
(rhs[keyPath: keyPathA], rhs[keyPath: keyPathB], rhs[keyPath: keyPathC])
}
}
func sorted<A : Comparable, B : Comparable, C : Comparable, D : Comparable>
(by keyPathA: KeyPath<Element, A>, _ keyPathB: KeyPath<Element, B>, _ keyPathC: KeyPath<Element, C>, _ keyPathD: KeyPath<Element, D>) -> [Element] {
return sorted { lhs, rhs in
(lhs[keyPath: keyPathA], lhs[keyPath: keyPathB], lhs[keyPath: keyPathC], lhs[keyPath: keyPathD]) <
(rhs[keyPath: keyPathA], rhs[keyPath: keyPathB], rhs[keyPath: keyPathC], rhs[keyPath: keyPathD])
}
}
func sorted<A : Comparable, B : Comparable, C : Comparable, D : Comparable, E : Comparable>
(by keyPathA: KeyPath<Element, A>, _ keyPathB: KeyPath<Element, B>, _ keyPathC: KeyPath<Element, C>, _ keyPathD: KeyPath<Element, D>, _ keyPathE: KeyPath<Element, E>) -> [Element] {
return sorted { lhs, rhs in
(lhs[keyPath: keyPathA], lhs[keyPath: keyPathB], lhs[keyPath: keyPathC], lhs[keyPath: keyPathD], lhs[keyPath: keyPathE]) <
(rhs[keyPath: keyPathA], rhs[keyPath: keyPathB], rhs[keyPath: keyPathC], rhs[keyPath: keyPathD], rhs[keyPath: keyPathE])
}
}
func sorted<A : Comparable, B : Comparable, C : Comparable, D : Comparable, E : Comparable, F : Comparable>
(by keyPathA: KeyPath<Element, A>, _ keyPathB: KeyPath<Element, B>, _ keyPathC: KeyPath<Element, C>, _ keyPathD: KeyPath<Element, D>, _ keyPathE: KeyPath<Element, E>, _ keyPathF: KeyPath<Element, F>) -> [Element] {
return sorted { lhs, rhs in
(lhs[keyPath: keyPathA], lhs[keyPath: keyPathB], lhs[keyPath: keyPathC], lhs[keyPath: keyPathD], lhs[keyPath: keyPathE], lhs[keyPath: keyPathF]) <
(rhs[keyPath: keyPathA], rhs[keyPath: keyPathB], rhs[keyPath: keyPathC], rhs[keyPath: keyPathD], rhs[keyPath: keyPathE], rhs[keyPath: keyPathF])
}
}
}
Run Code Online (Sandbox Code Playgroud)
我已经为它们定义了最多6个键路径,对于大多数排序情况而言,这应该足够了。我们正在做的字典元组比较重载的优势<在这里,也表现出在这里。
尽管实现的效果并不理想,但呼叫站点现在看起来要好得多,因为它可以让您说:
let sortedStuff = stuff.sorted(by: \.value, \.value2)
sortedStuff.forEach { print($0) }
let moreSortedStuff = stuff.sorted(by: \.value, \.doubleValue)
moreSortedStuff.forEach { print($0) }
Run Code Online (Sandbox Code Playgroud)