Memoization in JavaScript

1. 基本概念

在一個CPU密集型應用中,咱們可使用Memoization來進行優化,其主要用於經過存儲昂貴的函數調用的結果來加速程序,並在再次發生相同的輸入時返回緩存的結果。
例如一個簡單的求平方根的函數:javascript

const sqrt = Math.sqrt;
//使用cache緩存
const sqrt = (arg)=>{
    if(!sqrt.cache){
        sqrt.cache = {};
    }
    if(!sqrt.cache[arg]){
        sqrt.cache[arg] = Math.sqrt(arg)
    }
    return sqrt.cache[arg]
}
//簡單的運行時間對比
//第一次運行:
console.time('start1')
sqrt(779)
console.timeEnd('start1')
VM516:3 start1: 0.01806640625ms
//第二次運行:
console.time('start1')
sqrt(779)
console.timeEnd('start1')
VM521:3 start1: 0.005859375ms

2. 簡單通用實現

咱們實現一個通用的memoize函數,用它來包裝任意純函數,並緩存其計算結果。html

function memoize(fn){
        return function(){
            var args = Array.prototype.slice.call(arguments);
            fn.cache = fn.cache || {};
            return fn.cache[args] ?
                 fn.cache[args] :(fn.cache[args] = fn.apply(this,args))
        }
}

必須注意的是,memoize的原理是在發生相同的輸入時返回緩存的結果,這意味着其包裝的函數應當是一個純函數,當函數不純時,相同的輸入在不一樣時間可能返回不一樣的結果,此時再使用memoize明顯不合時宜。前端

3. 常見的memoize庫

3.1 lodash庫中的memoize

在該庫中,使用Map緩存計算結果,支持傳入resolver函數將傳入的參數轉換爲存入Map的鍵名(默認鍵名是第一個參數,當鍵名是引用類型時,引用不變,緩存不會更新)。java

function memoize(func, resolver) {
  if (typeof func != 'function' || (resolver != null && typeof resolver != 'function')) {
    throw new TypeError('Expected a function')
  }
  const memoized = function(...args) {
    const key = resolver ? resolver.apply(this, args) : args[0]
    const cache = memoized.cache

    if (cache.has(key)) {
      return cache.get(key)
    }
    const result = func.apply(this, args)
    memoized.cache = cache.set(key, result) || cache
    return result
  }
  memoized.cache = new (memoize.Cache || Map)
  return memoized
}

memoize.Cache = Map

3.2 memoize-one

與lodash不一樣,memoize-one僅僅保存上一次調用時的結果,若是下次參數變化,則更新緩存。所以沒必要擔憂因爲maxAge, maxSize, exclusions等致使的內存泄漏。react

//源碼
export default function<ResultFn: (...any[]) => mixed>(
  resultFn: ResultFn,
  isEqual?: EqualityFn = areInputsEqual,
): ResultFn {
  let lastThis: mixed;
  let lastArgs: mixed[] = [];
  let lastResult: mixed;
  let calledOnce: boolean = false;
//函數調用過,this沒有變化,參數isEqual時返回緩存值。
  const result = function(...newArgs: mixed[]) {
    if (calledOnce && lastThis === this && isEqual(newArgs, lastArgs)) {
      return lastResult;
    }
    lastResult = resultFn.apply(this, newArgs);
    calledOnce = true;
    lastThis = this;
    lastArgs = newArgs;
    return lastResult;
  };

  return (result: any);
}

能夠看到,能夠經過第二個參數自定義參數是否相同,默認是areInputsEqual函數。git

//areInputsEqual實現
export default function areInputsEqual(
  newInputs: mixed[],
  lastInputs: mixed[],
) {
//先進行參數長度比較
  if (newInputs.length !== lastInputs.length) {
    return false;
  }
 //參數淺比較
  for (let i = 0; i < newInputs.length; i++) {
    if (newInputs[i] !== lastInputs[i]) {
      return false;
    }
  }
  return true;
}

能夠自定義參數比較函數進行深比較github

import memoizeOne from 'memoize-one';
import isDeepEqual from 'lodash.isequal';

const identity = x => x;

const shallowMemoized = memoizeOne(identity);
const deepMemoized = memoizeOne(identity, isDeepEqual);

const result1 = shallowMemoized({ foo: 'bar' });
const result2 = shallowMemoized({ foo: 'bar' });

result1 === result2; // false - difference reference

const result3 = deepMemoized({ foo: 'bar' });
const result4 = deepMemoized({ foo: 'bar' });

result3 === result4; // true - arguments are deep equal

4. memoize-one在React中的應用

在React中有這樣一個應用場景,當部分props變化須要改變派生state時,能夠在getDerivedStateFromProps中經過if判斷是否須要從新計算,但它比它須要的更復雜,由於它必須單獨跟蹤和檢測每一個props和state的變化,當props不少時,這種判斷方式會變得糾纏不清。redux

class Example extends Component {
  state = {
    filterText: "",
  };
  static getDerivedStateFromProps(props, state) {
    if (
      props.list !== state.prevPropsList ||
      state.prevFilterText !== state.filterText
    ) {
      return {
        prevPropsList: props.list,
        prevFilterText: state.filterText,
        filteredList: props.list.filter(item => item.text.includes(state.filterText))
      };
    }
    return null;
  }
  handleChange = event => {
    this.setState({ filterText: event.target.value });
  };
  render() {
    return (
      <Fragment>
        <input onChange={this.handleChange} value={this.state.filterText} />
        <ul>{this.state.filteredList.map(item => <li key={item.id}>{item.text}</li>)}</ul>
      </Fragment>
    );
  }
}

而經過Memoization,能夠把上一次的計算結果保存下來,而避免重複計算。例如如下經過memoize-one庫實現的記憶,可讓咱們不用手動判斷list, filterText是否變化,是否須要從新計算。segmentfault

import memoize from "memoize-one";
class Example extends Component {
  // State only needs to hold the current filter text value:
  state = { filterText: "" };
  // Re-run the filter whenever the list array or filter text changes:
  filter = memoize(
    (list, filterText) => list.filter(item => item.text.includes(filterText))
  );

  handleChange = event => {
    this.setState({ filterText: event.target.value });
  };

  render() {
    const filteredList = this.filter(this.props.list, this.state.filterText);
    return (
      <Fragment>
        <input onChange={this.handleChange} value={this.state.filterText} />
        <ul>{filteredList.map(item => <li key={item.id}>{item.text}</li>)}</ul>
      </Fragment>
    );
  }
}

5. 配合Redux使用的memoize庫——reselect

在react中,每當組件從新渲染,計算派生狀態的邏輯就會執行一遍(無論這些邏輯是放在render或者是放在getDerivedStateFromProps中,若是沒有采用不少if限制的話)。上節介紹了使用memoize-one來緩存來避免重複計算,當咱們使用redux時,一般在mapStateToProps中計算派生狀態,每當store中的任意state更新時,都會觸發mapStateToProps中的計算,然而,每每派生state一般只依賴部分state,沒必要每次都計算。
Reselect是一個十分貼近redux的memoize庫,其和memoize-one同樣只緩存前一次的計算值,並支持自定義memoize函數,自定義參數比較函數等;其輸入參數由inputSelectors functions 產生,生成的的依然是inputSelectors functions,這意味的與memoize相比,這能夠很容易的組合。
另外,mapStateProps的第二個參數ownProps也能夠傳入selector中。數組

import { createSelector } from 'reselect'

fSelector = createSelector(
    a => state.a,
    b => state.b,
    (a, b) => f(a, b)
)
hSelector = createSelector(
    b => state.b,
    c => state.c,
    (b, c) => h(b, c)
)
gSelector =  createSelector(
    a => state.a,
    c => state.c,
    (a, c) => g(a, c)
)
uSelector = createSelector(
    a => state.a,
    b => state.b,
    c => state.c,
    (a, b, c) => u(a, b, c)
)

...
function mapStateToProps(state) {
    const { a, b, c } = state
    return {
        a,
        b,
        c,
        fab: fSelector(state),
        hbc: hSelector(state),
        gac: gSelector(state),
        uabc: uSelector(state)
    }
}

6. react中memoize原生支持——React.memo

若是你的函數組件在給定相同的props的狀況下呈現相同的結果,你能夠React.memo經過記憶結果將它包裝在一些調用中以提升性能。這意味着React將跳過渲染組件,並重用最後渲染的結果。

默認狀況下,它只會淺顯比較props對象中的複雜對象。若是要控制比較,還能夠提供自定義比較函數做爲第二個參數。

function MyComponent(props) {
  /* render using props */
}
function areEqual(prevProps, nextProps) {
  /*
  return true if passing nextProps to render would return
  the same result as passing prevProps to render,
  otherwise return false
  */
}
export default React.memo(MyComponent, areEqual);

You Probably Don’t Need Derived State - React Blog
Understanding Memoization in JavaScript to Improve Performance
性能優化:memoization | Taobao FED | 淘寶前端團隊
lodash/memoize.js at master · lodash/lodash · GitHub
GitHub - alexreardon/memoize-one: A memoization library which only remembers the latest invocation
爲何咱們須要reselect - 從0開始實現react技術棧 - SegmentFault 思否
React Hooks: Memoization – Sandro Dolidze – Medium

相關文章
相關標籤/搜索