我想介绍一个相当简单的ES5替代方案。该函数获得2个参数——目标和源，必须为“object”类型。Target将是结果对象。Target保留其所有原始属性，但它们的值可能会被修改。

function deepMerge(target, source) {
if(typeof target !== 'object' || typeof source !== 'object') return false; // target or source or both ain't objects, merging doesn't make sense
for(var prop in source) {
  if(!source.hasOwnProperty(prop)) continue; // take into consideration only object's own properties.
  if(prop in target) { // handling merging of two properties with equal names
    if(typeof target[prop] !== 'object') {
      target[prop] = source[prop];
    } else {
      if(typeof source[prop] !== 'object') {
        target[prop] = source[prop];
      } else {
        if(target[prop].concat && source[prop].concat) { // two arrays get concatenated
          target[prop] = target[prop].concat(source[prop]);
        } else { // two objects get merged recursively
          target[prop] = deepMerge(target[prop], source[prop]); 
        } 
      }  
    }
  } else { // new properties get added to target
    target[prop] = source[prop]; 
  }
}
return target;
}

例:

如果target没有source属性，则target获取source属性; 如果目标有source属性，而target & source没有 两个对象(4个中的3个)，目标的属性被覆盖; 如果target确实有一个source属性，并且它们都是对象/数组(剩余1种情况)，那么递归发生合并两个对象(或两个数组的连接);

还要考虑以下几点:

Array + obj = Array Obj + array = Obj Obj + Obj = Obj(递归合并) Array + Array = Array (concat)

它是可预测的，支持基本类型以及数组和对象。我们可以合并两个对象，我认为我们可以通过reduce函数合并两个以上的对象。

看一个例子(如果你想的话，也可以玩一下):

var a = { "a_prop": 1, "arr_prop": [4, 5, 6], "obj": { "a_prop": { "t_prop": 'test' }, "b_prop": 2 } }; var b = { "a_prop": 5, "arr_prop": [7, 8, 9], "b_prop": 15, "obj": { "a_prop": { "u_prop": false }, "b_prop": { "s_prop": null } } }; function deepMerge(target, source) { if(typeof target !== 'object' || typeof source !== 'object') return false; for(var prop in source) { if(!source.hasOwnProperty(prop)) continue; if(prop in target) { if(typeof target[prop] !== 'object') { target[prop] = source[prop]; } else { if(typeof source[prop] !== 'object') { target[prop] = source[prop]; } else { if(target[prop].concat && source[prop].concat) { target[prop] = target[prop].concat(source[prop]); } else { target[prop] = deepMerge(target[prop], source[prop]); } } } } else { target[prop] = source[prop]; } } return target; } console.log(deepMerge(a, b));

有一个限制-浏览器的调用堆栈长度。现代浏览器会在一些真正深层的递归中抛出错误(想想成千上万的嵌套调用)。此外，您还可以自由地处理像数组+对象等情况，因为您希望添加新的条件和类型检查。

当涉及到宿主对象或比值包更复杂的任何类型的对象时，这个问题就不那么简单了

do you invoke a getter to obtain a value or do you copy over the property descriptor? what if the merge target has a setter (either own property or in its prototype chain)? Do you consider the value as already-present or call the setter to update the current value? do you invoke own-property functions or copy them over? What if they're bound functions or arrow functions depending on something in their scope chain at the time they were defined? what if it's something like a DOM node? You certainly don't want to treat it as simple object and just deep-merge all its properties over into how to deal with "simple" structures like arrays or maps or sets? Consider them already-present or merge them too? how to deal with non-enumerable own properties? what about new subtrees? Simply assign by reference or deep clone? how to deal with frozen/sealed/non-extensible objects?

另一件需要记住的事情是:包含循环的对象图。这通常不难处理——简单地保留一组已经访问过的源对象——但经常被遗忘。

您可能应该编写一个深度合并函数，它只期望原始值和简单对象(结构化克隆算法最多可以处理的那些类型)作为合并源。如果遇到它不能处理的东西，或者只是通过引用而不是深度合并进行赋值，则抛出。

换句话说，没有一种适合所有人的算法，您要么必须使用自己的算法，要么寻找恰好涵盖您的用例的库方法。

下面是TypeScript的实现:

export const mergeObjects = <T extends object = object>(target: T, ...sources: T[]): T  => {
  if (!sources.length) {
    return target;
  }
  const source = sources.shift();
  if (source === undefined) {
    return target;
  }

  if (isMergebleObject(target) && isMergebleObject(source)) {
    Object.keys(source).forEach(function(key: string) {
      if (isMergebleObject(source[key])) {
        if (!target[key]) {
          target[key] = {};
        }
        mergeObjects(target[key], source[key]);
      } else {
        target[key] = source[key];
      }
    });
  }

  return mergeObjects(target, ...sources);
};

const isObject = (item: any): boolean => {
  return item !== null && typeof item === 'object';
};

const isMergebleObject = (item): boolean => {
  return isObject(item) && !Array.isArray(item);
};

和单元测试:

describe('merge', () => {
  it('should merge Objects and all nested Ones', () => {
    const obj1 = { a: { a1: 'A1'}, c: 'C', d: {} };
    const obj2 = { a: { a2: 'A2'}, b: { b1: 'B1'}, d: null };
    const obj3 = { a: { a1: 'A1', a2: 'A2'}, b: { b1: 'B1'}, c: 'C', d: null};
    expect(mergeObjects({}, obj1, obj2)).toEqual(obj3);
  });
  it('should behave like Object.assign on the top level', () => {
    const obj1 = { a: { a1: 'A1'}, c: 'C'};
    const obj2 = { a: undefined, b: { b1: 'B1'}};
    expect(mergeObjects({}, obj1, obj2)).toEqual(Object.assign({}, obj1, obj2));
  });
  it('should not merge array values, just override', () => {
    const obj1 = {a: ['A', 'B']};
    const obj2 = {a: ['C'], b: ['D']};
    expect(mergeObjects({}, obj1, obj2)).toEqual({a: ['C'], b: ['D']});
  });
  it('typed merge', () => {
    expect(mergeObjects<TestPosition>(new TestPosition(0, 0), new TestPosition(1, 1)))
      .toEqual(new TestPosition(1, 1));
  });
});

class TestPosition {
  constructor(public x: number = 0, public y: number = 0) {/*empty*/}
}

(本机解决方案)如果你知道你想要深度合并的属性，那么

const x = { a: { a: 1 } }
const y = { a: { b: 1 } }
Object.assign(y.a, x.a);
Object.assign(x, y);
// output: a: {b: 1, a: 1}

有一个lodash包专门处理对象的深度克隆。这样做的好处是不需要包含整个lodash库。

它叫lodash.clonedeep

在nodejs中，这种用法是这样的

var cloneDeep = require('lodash.clonedeep');
 
const newObject = cloneDeep(oldObject);

在ReactJS中，用法是

import cloneDeep from 'lodash/cloneDeep';

const newObject = cloneDeep(oldObject);

查看这里的文档。如果您对它的工作原理感兴趣，请查看这里的源文件

我在加载缓存redux状态时遇到了这个问题。如果我只是加载缓存的状态，我会遇到错误的新应用程序版本与更新的状态结构。

前面已经提到过，lodash提供了merge函数，我使用了这个函数:

const currentInitialState = configureState().getState();
const mergedState = _.merge({}, currentInitialState, cachedState);
const store = configureState(mergedState);