3

解决 TS 问题的最好办法就是多练,这次解读 type-challenges Medium 难度 63~68 题。

精读

Unique

实现 Unique<T>,对 T 去重:

type Res = Unique<[1, 1, 2, 2, 3, 3]> // expected to be [1, 2, 3]
type Res1 = Unique<[1, 2, 3, 4, 4, 5, 6, 7]> // expected to be [1, 2, 3, 4, 5, 6, 7]
type Res2 = Unique<[1, 'a', 2, 'b', 2, 'a']> // expected to be [1, "a", 2, "b"]
type Res3 = Unique<[string, number, 1, 'a', 1, string, 2, 'b', 2, number]> // expected to be [string, number, 1, "a", 2, "b"]
type Res4 = Unique<[unknown, unknown, any, any, never, never]> // expected to be [unknown, any, never]

去重需要不断递归产生去重后结果,因此需要一个辅助变量 R 配合,并把 Tinfer 逐一拆解,判断第一个字符是否在结果数组里,如果不在就塞进去:

type Unique<T, R extends any[] = []> = T extends [infer F, ...infer Rest]
  ? Includes<R, F> extends true
    ? Unique<Rest, R>
    : Unique<Rest, [...R, F]>
  : R

那么剩下的问题就是,如何判断一个对象是否出现在数组中,使用递归可以轻松完成:

type Includes<Arr, Value> = Arr extends [infer F, ...infer Rest]
  ? Equal<F, Value> extends true
    ? true
    : Includes<Rest, Value>
  : false

每次取首项,如果等于 Value 直接返回 true,否则继续递归,如果数组递归结束(不构成 Arr extends [xxx] 的形式)说明递归完了还没有找到相等值,直接返回 false

把这两个函数组合一下就能轻松解决本题:

// 本题答案
type Unique<T, R extends any[] = []> = T extends [infer F, ...infer Rest]
  ? Includes<R, F> extends true
    ? Unique<Rest, R>
    : Unique<Rest, [...R, F]>
  : R

type Includes<Arr, Value> = Arr extends [infer F, ...infer Rest]
  ? Equal<F, Value> extends true
    ? true
    : Includes<Rest, Value>
  : false

MapTypes

实现 MapTypes<T, R>,根据对象 R 的描述来替换类型:

type StringToNumber = {
  mapFrom: string; // value of key which value is string
  mapTo: number; // will be transformed for number
}
MapTypes<{iWillBeANumberOneDay: string}, StringToNumber> // gives { iWillBeANumberOneDay: number; }

因为要返回一个新对象,所以我们使用 { [K in keyof T]: ... } 的形式描述结果对象。然后就要对 Value 类型进行判断了,为了防止 never 的作用,我们包一层数组进行判断:

type MapTypes<T, R extends { mapFrom: any; mapTo: any }> = {
  [K in keyof T]: [T[K]] extends [R['mapFrom']] ? R['mapTo'] : T[K]
}

但这个解答还有一个 case 无法通过:

MapTypes<{iWillBeNumberOrDate: string}, StringToDate | StringToNumber> // gives { iWillBeNumberOrDate: number | Date; }

我们需要考虑到 Union 分发机制以及每次都要重新匹配一次是否命中 mapFrom,因此需要抽一个函数:

type Transform<R extends { mapFrom: any; mapTo: any }, T> = R extends any
  ? T extends R['mapFrom']
    ? R['mapTo']
    : never
  : never

为什么要 R extends any 看似无意义的写法呢?原因是 R 是联合类型,这样可以触发分发机制,让每一个类型独立判断。所以最终答案就是:

// 本题答案
type MapTypes<T, R extends { mapFrom: any; mapTo: any }> = {
  [K in keyof T]: [T[K]] extends [R['mapFrom']] ? Transform<R, T[K]> : T[K]
}

type Transform<R extends { mapFrom: any; mapTo: any }, T> = R extends any
  ? T extends R['mapFrom']
    ? R['mapTo']
    : never
  : never

Construct Tuple

生成指定长度的 Tuple:

type result = ConstructTuple<2> // expect to be [unknown, unkonwn]

比较容易想到的办法是利用下标递归:

type ConstructTuple<
  L extends number,
  I extends number[] = []
> = I['length'] extends L ? [] : [unknown, ...ConstructTuple<L, [1, ...I]>]

但在如下测试用例会遇到递归长度过深的问题:

ConstructTuple<999> // Type instantiation is excessively deep and possibly infinite

一种解法是利用 minusOne 提到的 CountTo 方法快捷生成指定长度数组,把 1 替换为 unknown 即可:

// 本题答案
type ConstructTuple<L extends number> = CountTo<`${L}`>

type CountTo<
  T extends string,
  Count extends unknown[] = []
> = T extends `${infer First}${infer Rest}`
  ? CountTo<Rest, N<Count>[keyof N & First]>
  : Count

type N<T extends unknown[] = []> = {
  '0': [...T, ...T, ...T, ...T, ...T, ...T, ...T, ...T, ...T, ...T]
  '1': [...T, ...T, ...T, ...T, ...T, ...T, ...T, ...T, ...T, ...T, unknown]
  '2': [
    ...T,
    ...T,
    ...T,
    ...T,
    ...T,
    ...T,
    ...T,
    ...T,
    ...T,
    ...T,
    unknown,
    unknown
  ]
  '3': [
    ...T,
    ...T,
    ...T,
    ...T,
    ...T,
    ...T,
    ...T,
    ...T,
    ...T,
    ...T,
    unknown,
    unknown,
    unknown
  ]
  '4': [
    ...T,
    ...T,
    ...T,
    ...T,
    ...T,
    ...T,
    ...T,
    ...T,
    ...T,
    ...T,
    unknown,
    unknown,
    unknown,
    unknown
  ]
  '5': [
    ...T,
    ...T,
    ...T,
    ...T,
    ...T,
    ...T,
    ...T,
    ...T,
    ...T,
    ...T,
    unknown,
    unknown,
    unknown,
    unknown,
    unknown
  ]
  '6': [
    ...T,
    ...T,
    ...T,
    ...T,
    ...T,
    ...T,
    ...T,
    ...T,
    ...T,
    ...T,
    unknown,
    unknown,
    unknown,
    unknown,
    unknown,
    unknown
  ]
  '7': [
    ...T,
    ...T,
    ...T,
    ...T,
    ...T,
    ...T,
    ...T,
    ...T,
    ...T,
    ...T,
    unknown,
    unknown,
    unknown,
    unknown,
    unknown,
    unknown,
    unknown
  ]
  '8': [
    ...T,
    ...T,
    ...T,
    ...T,
    ...T,
    ...T,
    ...T,
    ...T,
    ...T,
    ...T,
    unknown,
    unknown,
    unknown,
    unknown,
    unknown,
    unknown,
    unknown,
    unknown
  ]
  '9': [
    ...T,
    ...T,
    ...T,
    ...T,
    ...T,
    ...T,
    ...T,
    ...T,
    ...T,
    ...T,
    unknown,
    unknown,
    unknown,
    unknown,
    unknown,
    unknown,
    unknown,
    unknown,
    unknown
  ]
}

Number Range

实现 NumberRange<T, P>,生成数字为从 TP 的联合类型:

type result = NumberRange<2, 9> //  | 2 | 3 | 4 | 5 | 6 | 7 | 8 | 9 

NumberRange<2, 9> 为例,我们需要实现 29 的递增递归,因此需要一个数组长度从 2 递增到 9 的辅助变量 U,以及一个存储结果的辅助变量 R:

type NumberRange<T, P, U extends any[] = 长度为 T 的数组, R>

所以我们先实现 LengthTo 函数,传入长度 N,返回一个长度为 N 的数组:

type LengthTo<N extends number, R extends any[] = []> =
  R['length'] extends N ? R : LengthTo<N, [0, ...R]>

然后就是递归了:

// 本题答案
type NumberRange<T extends number, P extends number, U extends any[] = LengthTo<T>, R extends number = never> =
  U['length'] extends P ? (
    R | U['length']
  ) : (
    NumberRange<T, P, [0, ...U], R | U['length']>
  )

R 的默认值为 never 非常重要,否则默认值为 any,最终类型就会被放大为 any

Combination

实现 Combination<T>:

// expected to be `"foo" | "bar" | "baz" | "foo bar" | "foo bar baz" | "foo baz" | "foo baz bar" | "bar foo" | "bar foo baz" | "bar baz" | "bar baz foo" | "baz foo" | "baz foo bar" | "baz bar" | "baz bar foo"`
type Keys = Combination<['foo', 'bar', 'baz']>

本题和 AllCombination 类似:

type AllCombinations_ABC = AllCombinations<'ABC'>
// should be '' | 'A' | 'B' | 'C' | 'AB' | 'AC' | 'BA' | 'BC' | 'CA' | 'CB' | 'ABC' | 'ACB' | 'BAC' | 'BCA' | 'CAB' | 'CBA'

还记得这题吗?我们要将字符串变成联合类型:

type StrToUnion<S> = S extends `${infer F}${infer R}`
  ? F | StrToUnion<R>
  : never

而本题 Combination 更简单,把数组转换为联合类型只需要 T[number]。所以本题第一种组合解法是,将 AllCombinations 稍微改造下,再利用 ExcludeTrimRight 删除多余的空格:

// 本题答案
type AllCombinations<T extends string[], U extends string = T[number]> = [
  U
] extends [never]
  ? ''
  : '' | { [K in U]: `${K} ${AllCombinations<never, Exclude<U, K>>}` }[U]

type TrimRight<T extends string> = T extends `${infer R} ` ? TrimRight<R> : T

type Combination<T extends string[]> = TrimRight<Exclude<AllCombinations<T>, ''>>

还有一种非常精彩的答案在此分析一下:

// 本题答案
type Combination<T extends string[], U = T[number], A = U> = U extends infer U extends string
  ? `${U} ${Combination<T, Exclude<A, U>>}` | U
  : never;

依然利用 T[number] 的特性将数组转成联合类型,再利用联合类型 extends 会分组的特性递归出结果。

之所以不会出现结尾出现多余的空格,是因为 U extends infer U extends string 这段判断已经杜绝了 U 消耗完的情况,如果消耗完会及时返回 never,所以无需用 TrimRight 处理右侧多余的空格。

至于为什么要定义 A = U,在前面章节已经介绍过了,因为联合类型 extends 过程中会进行分组,此时访问的 U 已经是具体类型了,但此时访问 A 还是原始的联合类型 U

Subsequence

实现 Subsequence<T> 输出所有可能的子序列:

type A = Subsequence<[1, 2]> // [] | [1] | [2] | [1, 2]

因为是返回数组的全排列,只要每次取第一项,与剩余项的递归构造出结果,| 上剩余项本身递归的结果就可以了:

// 本题答案
type Subsequence<T extends number[]> = T extends [infer F, ...infer R extends number[]] ? (
  Subsequence<R> | [F, ...Subsequence<R>]
) : T

总结

对全排列问题有两种经典解法:

  • 利用辅助变量方式递归,注意联合类型与字符串、数组之间转换的技巧。
  • 直接递归,不借助辅助变量,一般在题目返回类型容易构造时选择。
讨论地址是:精读《Unique, MapTypes, Construct Tuple...》· Issue #434 · dt-fe/weekly

如果你想参与讨论,请 点击这里,每周都有新的主题,周末或周一发布。前端精读 - 帮你筛选靠谱的内容。

关注 前端精读微信公众号

<img width=200 src="https://img.alicdn.com/tfs/TB165W0MCzqK1RjSZFLXXcn2XXa-258-258.jpg">

版权声明:自由转载-非商用-非衍生-保持署名(创意共享 3.0 许可证

黄子毅
7k 声望9.5k 粉丝