Swift 中 String 取下标及性能问题

Swift 中 String 取下标及性能问题

取下标

String

String 用 String.Index 取下标(subscript)得到 Character，String.Index 要从 String 中获取

let greeting = "Guten Tag!"
greeting[greeting.startIndex] // Character "G"
greeting[greeting.index(before: greeting.endIndex)] // Character "!"
greeting[greeting.index(after: greeting.startIndex)] // Character "u"
let index = greeting.index(greeting.startIndex, offsetBy: 7)
greeting[index] // Character "a"

String 用 Range<String.Index> 或 ClosedRange<String.Index> (以下 Range 和 ClosedRange 统称为 Range) 取下标得到 String

let str = "abc"
str[str.startIndex..<str.index(after: str.startIndex)] // String "a"
str[str.startIndex...str.index(after: str.startIndex)] // String "ab"

Character

String 通过 characters 属性获得 String.CharacterView，表示屏幕上显示的内容。String.CharacterView 通过 String.CharacterView.Index 取下标得到 Character，String.CharacterView.Index 要从 String.CharacterView 中获取

let str = "abc"
let characters = str.characters // String.CharacterView
characters[characters.startIndex] // Character "a"

注意，String.CharacterView 不遵循 RandomAccessCollection 协议，用 String.CharacterView.Index 取下标不可以随机访问。另外，String.CharacterView.Index 与 String.Index 是相同的类型，属于 Struct。String.Index 的文档在 String 文档下

typealias Index = String.CharacterView.Index

String.CharacterView 通过 Range<String.CharacterView.Index> 得到 String.CharacterView。用 Character 和 String.CharacterView 都可以生成 String

let str = "abc"
let characters = str.characters // String.CharacterView
let characters2 = characters[characters.startIndex..<characters.index(after: characters.startIndex)] // String.CharacterView
String(characters.first!) == String(characters2) // true. characters.first! is Character

用 String.CharacterView 生成 Array<Character>，可以用 Int、Range<Int> 取下标。用 Array<Character> 也可以生成 String

let str = "abc"
let arr = Array(str.characters) // Array<Character> ["a", "b", "c"]
arr[1] // Character "b"
arr[1...2] // ArraySlice<Character> ["b", "c"]
String(arr) // String "abc"

Character 可以直接与 "a" 比较

let str = "abc"
let a = str[str.startIndex] // Character "a"
let b = str[str.index(str.startIndex, offsetBy: 1)] // Character "b"
a == "a" // true
b > "a" // true

UTF-8

String 通过 utf8 属性获得 String.UTF8View，表示 UTF-8 编码的内容。String.UTF8View 通过 String.UTF8View.Index 取下标得到 UTF8.CodeUnit，实际上是 UInt8；通过 Range<String.UTF8View.Index> 取下标得到 String.UTF8View。String.UTF8View.Index 要从 String.UTF8View 中获取。String.UTF8View 不遵循 RandomAccessCollection 协议，用 String.UTF8View.Index 取下标不可以随机访问。用 String.UTF8View 生成 Array<UInt8>，可以用 Int、Range<Int> 取下标。用 String.UTF8View 可以生成 String。用 UInt8 或 Array<UInt8> 也可以生成 String，但内容表示数字或数字数组，不是数字的 UTF-8 编码内容。

let str = "abc"
let utf8 = str.utf8 // String.UTF8View
let n = utf8[utf8.startIndex] // UInt8 97
let a = utf8[utf8.startIndex..<utf8.index(after: utf8.startIndex)] // String.UTF8View "a"
let ab = utf8[utf8.startIndex...utf8.index(after: utf8.startIndex)] // String.UTF8View "ab"
String(n) // "97", NOT "a"
String(a) // "a"
String(ab) // "ab"

let arr = Array(utf8) // Array<UInt8> [97, 98, 99]
let n2 = arr[0] // UInt8 97
let arr2 = arr[0...1] // // ArraySlice<UInt8> [97, 98]

String 通过 utf8CString 属性获得 ContiguousArray<CChar>，实际上是 ContiguousArray<Int8>，表示 UTF-8 编码的内容并且末尾增加一个 0，所以长度比 utf8 属性的长度大 1。ContiguousArray<Int8> 可以用 Int、Range<Int> 取下标，分别得到 Int8 和 ArraySlice<Int8>。ContiguousArray 遵循 RandomAccessCollection 协议，用 Int 取下标可以随机访问。

let str = "abc"
let utf8 = str.utf8CString // ContiguousArray<Int8> [97, 98, 99, 0]
let a = utf8[0] // Int8 97
let ab = utf8[0...1] // ArraySlice<Int8> [97, 98]

UTF-16

String 通过 utf16 属性获得 String.UTF16View，表示 UTF-16 编码的内容。String.UTF16View 通过 String.UTF16View.Index 取下标得到 UTF16.CodeUnit，实际上是 UInt16；通过 Range<String.UTF16View.Index> 取下标得到 String.UTF16View。String.UTF16View.Index 要从 String.UTF16View 中获取。String.UTF16View 遵循 RandomAccessCollection 协议，用 String.UTF16View.Index 取下标可以随机访问。用 String.UTF16View 生成 Array<UInt16>，可以用 Int、Range<Int> 取下标。用 String.UTF16View 可以生成 String。用 UInt16 或 Array<UInt16> 也可以生成 String，但内容表示数字或数字数组，不是数字的 UTF-16 编码内容。

let str = "abc"
let utf16 = str.utf16 // String.UTF16View
let n = utf16[utf16.startIndex] // UInt16 97
let a = utf16[utf16.startIndex..<utf16.index(after: utf16.startIndex)] // String.UTF16View "a"
let ab = utf16[utf16.startIndex...utf16.index(after: utf16.startIndex)] // String.UTF16View "ab"
String(n) // "97", NOT "a"
String(a) // "a"
String(ab) // "ab"

let arr = Array(utf16) // Array<UInt16> [97, 98, 99]
let n2 = arr[0] // UInt16 97
let arr2 = arr[0...1] // // ArraySlice<UInt8> [97, 98]

性能对比

对 String、String.CharacterView、Array<Character>、String.UTF8View、Array<UInt8>、ContiguousArray<Int8>、String.UTF16View、Array<UInt16> 进行判空(isEmpty)、获取长度(count)、一个位置的取下标([index])、一段距离的取下标([range])测试，统计执行时间。

定义测试类型、打印和更新时间的方法、要测试的 String

import Foundation

enum TestType {
    case isEmpty
    case count
    case index
    case range
}

func printAndUpdateTime(_ date: inout Date) {
    let now = Date()
    print(now.timeIntervalSince(date))
    date = now
}

let s = "aasdfsdfsdfgfdsg vrutj7edbj7 ergcwhmkl5lknjklqawkrcqjljkljqjlqjhbrlqwfcbhafcpiluioufnlkqjvjakjn fnvjalgkhlkdkjlkasdfsdfsdfgfdsg vrutj7edbj7 ergcwhmkl5lknjklqawkrcqjliopjktyuljqjlqjhbrlqwfcbhafciluioufnlkjvjakjn fnvjalgkhlkdkjlkasdfsdfsdfgfdsg vrutj7edbj7 ergcwhmkl5lknjklqawkrcqjljkljqjlqjhbrlqwfcbhafciluioufnlkjvjakjn fnvjalgkhlkdkjlkasderwytwghfsdfsdfgfdsg vrutj7edbj7 fdgotuyoergcwhmkl5lknjklqawkyrcqjljkljqjlqjhbrlqwfcbhafci luioufnlkjvjakjn fnvjalgkhlkdkjlkasdfsdfsdfgfdsg vrutj7edbj7 ergcvcnvbwhmkl5lknjklqawkrcqjljkljqjlqjhbrlqwfcbhafciluioufnlkjvjakjn fnvjalgkhlkdkjlkasdfsdfsdfgfdsg vrutj7edbj7 ergcwhmkl5lknjklqawkrcqjljkljqjlqjhbrlqwfcbhafciluioufnlkjvjakjkn fnvjalgkhlkdkjlkasdfsdfsdfgfdsg vrutj7edbj7 ergcwhmkl5lknjklqawkrcqjljkljqjlqjhbrlqwfcbhafciluioufnlkjvjakjn fnvjalgkhlkdkjlkasdfsdfsdfgfdsg iopiouvrutj7edbj7 ergcwhmkl5lknjklqawkrcqjljkfghngdljqjlqjhbrlqwfcbhafciluioufnlkjvjakjn fnvjalgkhlkdkjlkasdfsdfsdfgfdsg vrutj7edbj7 ergcwhmbkl5lknjklqawkrcqjljkljqjlqjhbrlqwfcbhafciluioufnlkjvjakjn fnvjalgkhlkdkjlkasdfsdfsdfgfdsg vrutj7edbj7 ergcwhmkl5lknjklqawkrcqjljkljqjlqjhbrlqwfcbhafciluioufnlkjvjakjn fnvjalgkhlkdkjlkasdfsdfsdfgfdsg vrutj7edbj7 ergcwhmkl5lknjklqawkrcqjljkljqjlqjhbrlqwfcbhafciluioufnlkjvjakjn fnvjalgkhlkdkjlkasdfsdfsdfgfdsg vrutj7edbj7 ergcwhmkl5lknjklqawkrcqjljkljqjlqjhbrlqwfcbhafciluioufnlkjvjakjn fnvjalgkhlkdkjlkasdfsdfsdfgfdsg vrutj7edbj7 ergcwhmkl5lknjklqasdfsdwkrcqjljkljqjlqjhbrlqwfcbhafciluioufnlkjvjakjn fnvjalgkhlkdkjlkasdfsdfsdfgfdsg vrutj7edbj7 ergcwhmkl5lknjklqawkrcqjljkljqjlqjhbrlqwfcbhafciluioufnlkjvjakjn fnvjalgkhlkdkjlkasdfsdfsdfgfdsg vrutj7edbj7 ergcwhmkl5lknjklqawkrcqjljkljdqjlqjhbrlqwfcbhafciluioufnlkjvjakjn fnvjalgkhlkdkjlkasddfsdfsdfgfdsg vrutj7edbj7 ergcwhmkl5lknjklqawkrcqjljkljqjlqjhbrlqwfcbhafciluioufnlkjvjakjn fnvjalgkhlkdkjlkasdfsdfsdfgfdsg vrutj7edbj7 ergcwhmkl5lknjklqawkrcqjljkljqjlqjhbsdfdsrlqwfcbhafciluioufnlkjvjakjn fnvjalgkhlkdkjlkasdfsdfsdfgfdsg vrutj7edbj7 ergcwhmkl5lknjklqawkrcqjljkljqjlqjhbrlqwfcbhafciluioufnlkjvjakjn fnvjalgkhlkdkjlkasdfsdfsdfgfdsg vrutj7edbj7 ergcwhmkl5lknjklqawkrcqjljkljqjlqjhbrlqwfcbhafciluioufnlkjvjakjn fnvjalgkhlkdkjlkasdfsdfsdfgfdsg vrutj7edbj7 ergcwhmkl5lknjklqawkrcqjljkljqjlqjhbrlqwfcbhafciluioufnlkjvjakjn fnvjalgkhlkdkjlkasdfsdfsdfsadfsdgfdsg vrutj7edbj7 ergcwhmkl5lknjklqawkrcqjljkljqsdfasjlqjhbrlqwfcbhafciluioufnlkjvjakjn fnvjalgkhlkdkjlkasdfsdfsdfgfdsg vrutj7edbj7 ergcwhmkl5lknjklqawkrcqjljkljqjlqjhbrlqwfcbhafciluioufnlkjvjakjn fnvjalgkhlkdkjlkasdfsdfsdfgfdsg vrutj7edbj7 ergcwhmkl5lknjklqawkrcqjljkljqjlqjhbrlqwfcbhafciluioufnlkjvjakjn fnvjalgkhlkdkjlkasdfsdfsdafgfdsg vrutj7edbj7 ergcwhmkl5lknjklqawkrcqjljkljqjlqjhbrlqwfcbhafciluioufnlkjvjakjn fnvjalgkhlkdkjlkasdfsdfsdfgfdsg vrutj7edbj7 ergcwhmkl5lknjklqawkrcqjljkljqjlqjhbrlqwfcbhafciluioufnlkjvjakjn fnvjalgkhlkdkjlkasdfsdfsdfgfdsg vrutj7edbj7 ergcwhmkl5lknjklqawkrcqjljkljqjlqjhbrlqwfcbhafciluioufnlkjvjakjn fnvjalgkhlkdkjlkasdfsdfsdfgfdsg vrutj7edbj7 ergcwhmkl5lknjklqawkrcqjljkljqjlqjhbrlqwfcbhafciluioufnlkjvjakjn fnvjalgkhlkdkjlk"

测试代码

let loopCount = 10000
let index = s.characters.count / 2
let testType: TestType = .range
print(testType)
var date = Date()

forLoop: for _ in 0..<loopCount {
    switch testType {
    case .isEmpty:
        _ = s.isEmpty
    case .count:
        break forLoop
    case .index:
        _ = s[s.index(s.startIndex, offsetBy: index)]
    case .range:
        let endIndex = s.index(s.startIndex, offsetBy: index)
        _ = s[s.startIndex..<endIndex]
    }
}

if testType == .count {
    date = Date()
} else {
    print("String")
    printAndUpdateTime(&date)
}

let characters = s.characters
for _ in 0..<loopCount {
    switch testType {
    case .isEmpty:
        _ = characters.isEmpty
    case .count:
        _ = characters.count
    case .index:
        _ = characters[characters.index(characters.startIndex, offsetBy: index)]
    case .range:
        let endIndex = characters.index(characters.startIndex, offsetBy: index)
        _ = characters[characters.startIndex..<endIndex]
    }
}

print("Characters")
printAndUpdateTime(&date)

let characterArr = Array(characters)
for _ in 0..<loopCount {
    switch testType {
    case .isEmpty:
        _ = characterArr.isEmpty
    case .count:
        _ = characterArr.count
    case .index:
        _ = characterArr[index]
    case .range:
        _ = characterArr[0..<index]
    }
}

print("Characters array")
printAndUpdateTime(&date)

let utf8 = s.utf8
for _ in 0..<loopCount {
    switch testType {
    case .isEmpty:
        _ = utf8.isEmpty
    case .count:
        _ = utf8.count
    case .index:
        _ = utf8[utf8.index(utf8.startIndex, offsetBy: index)]
    case .range:
        let endIndex = utf8.index(utf8.startIndex, offsetBy: index)
        _ = utf8[utf8.startIndex..<endIndex]
    }
}

print("UTF-8")
printAndUpdateTime(&date)

let utf8Arr = Array(utf8)
for _ in 0..<loopCount {
    switch testType {
    case .isEmpty:
        _ = utf8Arr.isEmpty
    case .count:
        _ = utf8Arr.count
    case .index:
        _ = utf8Arr[index]
    case .range:
        _ = utf8Arr[0..<index]
    }
}

print("UTF-8 array")
printAndUpdateTime(&date)

let utf8CString = s.utf8CString
for _ in 0..<loopCount {
    switch testType {
    case .isEmpty:
        _ = utf8CString.isEmpty
    case .count:
        _ = utf8CString.count
    case .index:
        _ = utf8CString[index]
    case .range:
        _ = utf8CString[0..<index]
    }
}

print("UTF-8 C string")
printAndUpdateTime(&date)

let utf16 = s.utf16
for _ in 0..<loopCount {
    switch testType {
    case .isEmpty:
        _ = utf16.isEmpty
    case .count:
        _ = utf16.count
    case .index:
        _ = utf16[utf16.index(utf16.startIndex, offsetBy: index)]
    case .range:
        let endIndex = utf16.index(utf16.startIndex, offsetBy: index)
        _ = utf16[utf16.startIndex..<endIndex]
    }
}

print("UTF-16")
printAndUpdateTime(&date)

let utf16Arr = Array(utf16)
for _ in 0..<loopCount {
    switch testType {
    case .isEmpty:
        _ = utf16Arr.isEmpty
    case .count:
        _ = utf16Arr.count
    case .index:
        _ = utf16Arr[index]
    case .range:
        _ = utf16Arr[0..<index]
    }
}

print("UTF-16 array")
printAndUpdateTime(&date)

测试结果

判空

获取长度

一个位置的取下标

一段距离的取下标

以上比较中，判断 String 是否为空，访问 String 的 isEmpty 速度最快。对于其他操作，遵循 RandomAccessCollection 协议(ContiguousArray<Int8>、String.UTF16View 以及其他 Array)的类型效率较高。

进一步比较判空操作

let loopCount = 10000
var date = Date()

for _ in 0..<loopCount {
    _ = s.isEmpty
}

print("isEmpty")
printAndUpdateTime(&date)

for _ in 0..<loopCount {
    _ = s == ""
}

print("== \"\"")
printAndUpdateTime(&date)

与访问 String 的 isEmpty 相比，判断 String 是否等于空 String 速度更快！

注意到文档中，对 String.UTF8View 和 String.UTF16View 的 Range 取下标方法的说明

subscript(bounds: Range<String.UTF8View.Index>) -> String.UTF8View { get }
subscript(bounds: Range<String.UTF16View.Index>) -> String.UTF16View { get }

Complexity: O(n) if the underlying string is bridged from Objective-C, where n is the length of the string; otherwise, O(1).

如果 String 是从 Objective-C 的 NSString 桥接来的，时间复杂度为 O(n)，否则为 O(1)。这句话怎么理解呢？前面说了，String.UTF8View 不遵循 RandomAccessCollection 协议，而 String.UTF16View 遵循 RandomAccessCollection 协议，两者的时间复杂度应该不同。这里怎么说时间复杂度与 String 是否桥接自 NSString 有关？以下进一步探究。

let s2 = NSString(string: s) as String

let loopCount = 10000
let index = s.characters.count / 2
let index2 = s.characters.count - 1

func test(_ s: String) {
    var date = Date()

    let utf8 = s.utf8
    for _ in 0..<loopCount {
        _ = utf8[utf8.startIndex..<utf8.index(utf8.startIndex, offsetBy: index)]
    }

    print("UTF-8 index")
    printAndUpdateTime(&date)

    for _ in 0..<loopCount {
        _ = utf8[utf8.startIndex..<utf8.index(utf8.startIndex, offsetBy: index2)]
    }

    print("UTF-8 index2")
    printAndUpdateTime(&date)

    let utf16 = s.utf16
    for _ in 0..<loopCount {
        _ = utf16[utf16.startIndex..<utf16.index(utf16.startIndex, offsetBy: index)]
    }

    print("UTF-16 index")
    printAndUpdateTime(&date)

    for _ in 0..<loopCount {
        _ = utf16[utf16.startIndex..<utf16.index(utf16.startIndex, offsetBy: index2)]
    }

    print("UTF-16 index2")
    printAndUpdateTime(&date)
}

print("String")
test(s)
print("\nString bridged from NSString")
test(s2)

测试结果

对比 index 与 index2 的差异。测试参数 index2 约为 index 的 2 倍。UTF-8 index2 的耗时也约为 index 的 2 倍。UTF-16 的 index 和 index2 耗时相近。这与是否遵循 RandomAccessCollection 协议一致。

对比 String 与 NSString 的差异。桥接自 NSString 的 String 耗时比 String 要长，UTF-8 尤其明显。这应该就是文档说明的情况。用 Range 取下标，桥接自 NSString 的 String，比 String 多一些操作，多出 O(n) 级别的时间，而不是取下标的时间复杂度是 O(n)。

应用

具体应用时，选取哪种编码方式、取下标方式？首先，编码方式要看具体应用场景。编码方法不同，字符串的长度可能不同。如果字符串只含英文，比较好办。如果字符串含有中文或 Emoji，选择编码方式就要慎重。注意，NSString 的 length 属性获得的长度对应 UTF-16 编码。

let str = "abc"
str.characters.count // 3
str.unicodeScalars.count // 3
str.utf16.count // 3
(str as NSString).length // 3
str.utf8.count // 3
str.utf8CString.count - 1 // 3
strlen(str) // 3

let emojiStr = "