[译]JSON数据范式化（normalizr）
jopen 9年前
开发复杂的应用时，不可避免会有一些数据相互引用。建议你尽可能地把 state 范式化，不存在嵌套。把所有数据放到一个对象里，每个数据以 ID 为主键，不同数据相互引用时通过 ID 来查找。把 应用的 state 想像成数据库 。这种方法在 normalizr 文档里有详细阐述。
normalizr ：将嵌套的JSON格式扁平化，方便被Redux利用；
目标
我们的目标是将：
[{ id: 1, title: 'Some Article', author: { id: 1, name: 'Dan' } }, { id: 2, title: 'Other Article', author: { id: 1, name: 'Dan' } }]- 数组的每个对象都糅合了三个维度 文章 、 作者
- 按照数据范式，应当将这两个维度拆分出来，两者的联系通过id关联起来即可
我们描述上述的结构： - 返回的是一个数组（array） - 数组的对象中包含另外一个schema —— user
应该比较合理的，应该是转换成：
{ result: [1, 2], entities: { articles: { 1: { id: 1, title: 'Some Article', author: 1 }, 2: { id: 2, title: 'Other Article', author: 1 } }, users: { 1: { id: 1, name: 'Dan' } } } }如何使用
观看示例最好的，就是官方的测试文件： https://github.com/gaearon/normalizr/blob/master/test/index.js
先引入 normalizr ：
import { normalize, Schema, arrayOf } from 'normalizr'; 定义 schema ：
var article = new Schema('articles'), user = new Schema('users'), collection = new Schema('collections'), feedSchema, input;定义规则：
article.define({ author: user, collections: arrayOf(collection) }); collection.define({ curator: user }); feedSchema = { feed: arrayOf(article) };测试：
input = { feed: [{ id: 1, title: 'Some Article', author: { id: 3, name: 'Mike Persson' }, collections: [{ id: 1, title: 'Awesome Writing', curator: { id: 4, name: 'Andy Warhol' } }, { id: 7, title: 'Even Awesomer', curator: { id: 100, name: 'T.S. Eliot' } }] }, { id: 2, title: 'Other Article', collections: [{ id: 2, title: 'Neverhood', curator: { id: 120, name: 'Ada Lovelace' } }], author: { id: 2, name: 'Pete Hunt' } }] }; Object.freeze(input); normalize(input, feedSchema).should.eql({ result: { feed: [1, 2] }, entities: { articles: { 1: { id: 1, title: 'Some Article', author: 3, collections: [1, 7] }, 2: { id: 2, title: 'Other Article', author: 2, collections: [2] } }, collections: { 1: { id: 1, title: 'Awesome Writing', curator: 4 }, 2: { id: 2, title: 'Neverhood', curator: 120 }, 7: { id: 7, title: 'Even Awesomer', curator: 100 } }, users: { 2: { id: 2, name: 'Pete Hunt' }, 3: { id: 3, name: 'Mike Persson' }, 4: { id: 4, name: 'Andy Warhol' }, 100: { id: 100, name: 'T.S. Eliot' }, 120: { id: 120, name: 'Ada Lovelace' } } } });优势
假定请求 /articles 返回的数据的schema如下：
articles: article* article: { author: user, likers: user* primary_collection: collection? collections: collection* } collection: { curator: user } 如果不做范式化，store需要事先知道API的各种结构，比如 UserStore 会包含很多样板代码来获取新用户，诸如下面那样：
switch (action.type) { case ActionTypes.RECEIVE_USERS: mergeUsers(action.rawUsers); break; case ActionTypes.RECEIVE_ARTICLES: action.rawArticles.forEach(rawArticle => { mergeUsers([rawArticle.user]); mergeUsers(rawArticle.likers); mergeUsers([rawArticle.primaryCollection.curator]); rawArticle.collections.forEach(rawCollection => { mergeUsers(rawCollection.curator); }); }); UserStore.emitChange(); break; }store表示累觉不爱啊！！ 每个store都要对返回的 进行各种foreach 才能获取想要的数据。
来一个范式吧：
const article = new Schema('articles'); const user = new Schema('users'); article.define({ author: user, contributors: arrayOf(user), meta: { likes: arrayOf({ user: user }) } }); // ... const json = getArticleArray(); const normalized = normalize(json, arrayOf(article)); 经过范式整顿之后，你爱理或者不爱理，users对象总是在 action.entities.users 中：
const { action } = payload; if (action.response && action.response.entities && action.response.entities.users) { mergeUsers(action.response.entities.users); UserStore.emitChange(); break; }更多示例（来自测试文件）
规范化单个文件
var article = new Schema('articles'), input; input = { id: 1, title: 'Some Article', isFavorite: false }; Object.freeze(input); normalize(input, article).should.eql({ result: 1, entities: { articles: { 1: { id: 1, title: 'Some Article', isFavorite: false } } } });规范化内嵌对象，并删除额外key
有时候后端接口会返回很多额外的字段，甚至会有重复的字段；比如下方示例中 typeId 和 type.id 是重复的；注意方法中 形参 key 是经过 artcle.define 定义过的。
var article = new Schema('articles'), type = new Schema('types'), input; // 定义内嵌规则 article.define({ type: type }); input = { id: 1, title: 'Some Article', isFavorite: false, typeId: 1, type: { id: 1, } }; Object.freeze(input); // assignEntity删除后端返回额外数据的 var options = { assignEntity: function(obj, key, val) { obj[key] = val; delete obj[key + 'Id']; } }; normalize(input, article, options).should.eql({ result: 1, entities: { articles: { 1: { id: 1, title: 'Some Article', isFavorite: false, type: 1 } }, types: { 1: { id: 1 } } } });添加额外数据
和上个示例相反的是， mergeIntoEntity 用于将多份同质数据不同信息融合到一起，用于解决冲突。
下方示例中， author 和 reviewer 是同一个人，只是前者留下的联系方式是手机，后者留下的联系方式是邮箱，但无论如何都是同一个人；
此时就可以使用 mergeIntoEntity 将两份数据融合到一起；（注意这里是用 valueOf 规则 ）
var author = new Schema('authors'), input; input = { author: { id: 1, name: 'Ada Lovelace', contact: { phone: '555-0100' } }, reviewer: { id: 1, name: 'Ada Lovelace', contact: { email: 'ada@lovelace.com' } } } Object.freeze(input); var options = { mergeIntoEntity: function(entityA, entityB, entityKey) { var key; for (key in entityB) { if (!entityB.hasOwnProperty(key)) { continue; } if (!entityA.hasOwnProperty(key) || isEqual(entityA[key], entityB[key])) { entityA[key] = entityB[key]; continue; } if (isObject(entityA[key]) && isObject(entityB[key])) { merge(entityA[key], entityB[key]) continue; } console.warn('Unequal data!'); } } }; normalize(input, valuesOf(author), options).should.eql({ result: { author: 1, reviewer: 1 }, entities: { authors: { 1: { id: 1, name: 'Ada Lovelace', contact: { phone: '555-0100', email: 'ada@lovelace.com' } } } } });按指定的属性规范化
有时候对象没有 id 属性，或者我们并不想按 id 属性规范化，可以使用 idAttribute 指定；
下面的例子，就是使用 slug 作为规范化的key：
var article = new Schema('articles', { idAttribute: 'slug' }), input; input = { id: 1, slug: 'some-article', title: 'Some Article', isFavorite: false }; Object.freeze(input); normalize(input, article).should.eql({ result: 'some-article', entities: { articles: { 'some-article': { id: 1, slug: 'some-article', title: 'Some Article', isFavorite: false } } } });创建自定义的属性
有时候想自己创建一个key，虽然今天和去年创建的文章名称都是 Happy ，但明显是不一样的，为了按时间区分出来，可以 使用自定义函数生成想要的key 。
function makeSlug(article) { var posted = article.posted, title = article.title.toLowerCase().replace(' ', '-'); return [title, posted.year, posted.month, posted.day].join('-'); } var article = new Schema('articles', { idAttribute: makeSlug }), input; input = { id: 1, title: 'Some Article', isFavorite: false, posted: { day: 12, month: 3, year: 1983 } }; Object.freeze(input); normalize(input, article).should.eql({ result: 'some-article-1983-3-12', entities: { articles: { 'some-article-1983-3-12': { id: 1, title: 'Some Article', isFavorite: false, posted: { day: 12, month: 3, year: 1983 } } } } });规范化数组
后端返回的数据往往是一串数组居多，此时规范化起到很大的作用，规范化的同时将数据压缩了一遍；
var article = new Schema('articles'), input; input = [{ id: 1, title: 'Some Article' }, { id: 2, title: 'Other Article' }]; Object.freeze(input); normalize(input, arrayOf(article)).should.eql({ result: [1, 2], entities: { articles: { 1: { id: 1, title: 'Some Article' }, 2: { id: 2, title: 'Other Article' } } } });抽取多个schema
上面讲的情形比较简单，只涉及抽出结果是单个schema的情形；现实中，你往往想抽象出多个schema，比如下方，我想抽离出 tutorials （教程） 和 articles （文章）两个 schema，此时需要 通过 schemaAttribute 选项指定区分这两个 schema 的字段 ：
var article = new Schema('articles'), tutorial = new Schema('tutorials'), articleOrTutorial = { articles: article, tutorials: tutorial }, input; input = [{ id: 1, type: 'articles', title: 'Some Article' }, { id: 1, type: 'tutorials', title: 'Some Tutorial' }]; Object.freeze(input); normalize(input, arrayOf(articleOrTutorial, { schemaAttribute: 'type' })).should.eql({ result: [ {id: 1, schema: 'articles'}, {id: 1, schema: 'tutorials'} ], entities: { articles: { 1: { id: 1, type: 'articles', title: 'Some Article' } }, tutorials: { 1: { id: 1, type: 'tutorials', title: 'Some Tutorial' } } } });这个示例中，虽然文章的id都是1，但很明显它们是不同的文章，因为一篇是普通文章，一篇是教程文章；因此要按schema维度抽离数据；
这里的 arrayOf(articleOrTutorial) 中的 articleOrTutorial 是包含多个属性的对象，这表示 input 应该是 articleOrTutorial 中的一种情况；
有时候原始数据属性 和 我们定义的有些差别，此时可以将 schemaAttribute 的值设成函数，将原始属性经过适当加工；比如原始属性是 tutorial , 而抽离出的 schema 名字为 tutorials ，相差一个 s ：
function guessSchema(item) { return item.type + 's'; } var article = new Schema('articles'), tutorial = new Schema('tutorials'), articleOrTutorial = { articles: article, tutorials: tutorial }, input; input = [{ id: 1, type: 'article', title: 'Some Article' }, { id: 1, type: 'tutorial', title: 'Some Tutorial' }]; Object.freeze(input); normalize(input, arrayOf(articleOrTutorial, { schemaAttribute: guessSchema })).should.eql({ result: [ { id: 1, schema: 'articles' }, { id: 1, schema: 'tutorials' } ], entities: { articles: { 1: { id: 1, type: 'article', title: 'Some Article' } }, tutorials: { 1: { id: 1, type: 'tutorial', title: 'Some Tutorial' } } } });上述是数组情况，针对普通的对象也是可以的，将规则 改成 valueOf 即可：
var article = new Schema('articles'), tutorial = new Schema('tutorials'), articleOrTutorial = { articles: article, tutorials: tutorial }, input; input = { one: { id: 1, type: 'articles', title: 'Some Article' }, two: { id: 2, type: 'articles', title: 'Another Article' }, three: { id: 1, type: 'tutorials', title: 'Some Tutorial' } }; Object.freeze(input); normalize(input, valuesOf(articleOrTutorial, { schemaAttribute: 'type' })).should.eql({ result: { one: {id: 1, schema: 'articles'}, two: {id: 2, schema: 'articles'}, three: {id: 1, schema: 'tutorials'} }, entities: { articles: { 1: { id: 1, type: 'articles', title: 'Some Article' }, 2: { id: 2, type: 'articles', title: 'Another Article' } }, tutorials: { 1: { id: 1, type: 'tutorials', title: 'Some Tutorial' } } } });schemaAttribute 是函数的情况就不列举了，和上述一致；
规范化内嵌情形
上面的对象比较简单，原本就是扁平化的；如果对象格式稍微复杂一些，比如每篇文章有多个作者的情形。此时需要使用 define 事先声明 schema 之间的层级关系：
var article = new Schema('articles'), user = new Schema('users'), input; article.define({ author: user }); input = { id: 1, title: 'Some Article', author: { id: 3, name: 'Mike Persson' } }; Object.freeze(input); normalize(input, article).should.eql({ result: 1, entities: { articles: { 1: { id: 1, title: 'Some Article', author: 3 } }, users: { 3: { id: 3, name: 'Mike Persson' } } } }); 上面是不是觉得简单了？那么给你一个比较复杂的情形，万变不离其宗。我们最终想抽离出 articles 、 users 以及 collections 这三个 schema，所以只要定义这三个schema就行了，
然后使用 define 方法声明这三个schema之间千丝万缕的关系；
最外层的feed只是属性，并不需要定义；
var article = new Schema('articles'), user = new Schema('users'), collection = new Schema('collections'), feedSchema, input; article.define({ author: user, collections: arrayOf(collection) }); collection.define({ curator: user }); feedSchema = { feed: arrayOf(article) }; input = { feed: [{ id: 1, title: 'Some Article', author: { id: 3, name: 'Mike Persson' }, collections: [{ id: 1, title: 'Awesome Writing', curator: { id: 4, name: 'Andy Warhol' } }, { id: 7, title: 'Even Awesomer', curator: { id: 100, name: 'T.S. Eliot' } }] }, { id: 2, title: 'Other Article', collections: [{ id: 2, title: 'Neverhood', curator: { id: 120, name: 'Ada Lovelace' } }], author: { id: 2, name: 'Pete Hunt' } }] }; Object.freeze(input); normalize(input, feedSchema).should.eql({ result: { feed: [1, 2] }, entities: { articles: { 1: { id: 1, title: 'Some Article', author: 3, collections: [1, 7] }, 2: { id: 2, title: 'Other Article', author: 2, collections: [2] } }, collections: { 1: { id: 1, title: 'Awesome Writing', curator: 4 }, 2: { id: 2, title: 'Neverhood', curator: 120 }, 7: { id: 7, title: 'Even Awesomer', curator: 100 } }, users: { 2: { id: 2, name: 'Pete Hunt' }, 3: { id: 3, name: 'Mike Persson' }, 4: { id: 4, name: 'Andy Warhol' }, 100: { id: 100, name: 'T.S. Eliot' }, 120: { id: 120, name: 'Ada Lovelace' } } } });内嵌+数组倾斜
var article = new Schema('articles'), tutorial = new Schema('tutorials'), articleOrTutorial = { articles: article, tutorials: tutorial }, user = new Schema('users'), collection = new Schema('collections'), feedSchema, input; article.define({ author: user, collections: arrayOf(collection) }); tutorial.define({ author: user, collections: arrayOf(collection) }); collection.define({ curator: user }); feedSchema = { feed: arrayOf(articleOrTutorial, { schemaAttribute: 'type' }) }; input = { feed: [{ id: 1, type: 'articles', title: 'Some Article', author: { id: 3, name: 'Mike Persson' }, collections: [{ id: 1, title: 'Awesome Writing', curator: { id: 4, name: 'Andy Warhol' } }, { id: 7, title: 'Even Awesomer', curator: { id: 100, name: 'T.S. Eliot' } }] }, { id: 1, type: 'tutorials', title: 'Some Tutorial', collections: [{ id: 2, title: 'Neverhood', curator: { id: 120, name: 'Ada Lovelace' } }], author: { id: 2, name: 'Pete Hunt' } }] }; Object.freeze(input); normalize(input, feedSchema).should.eql({ result: { feed: [ { id: 1, schema: 'articles' }, { id: 1, schema: 'tutorials' } ] }, entities: { articles: { 1: { id: 1, type: 'articles', title: 'Some Article', author: 3, collections: [1, 7] } }, tutorials: { 1: { id: 1, type: 'tutorials', title: 'Some Tutorial', author: 2, collections: [2] } }, collections: { 1: { id: 1, title: 'Awesome Writing', curator: 4 }, 2: { id: 2, title: 'Neverhood', curator: 120 }, 7: { id: 7, title: 'Even Awesomer', curator: 100 } }, users: { 2: { id: 2, name: 'Pete Hunt' }, 3: { id: 3, name: 'Mike Persson' }, 4: { id: 4, name: 'Andy Warhol' }, 100: { id: 100, name: 'T.S. Eliot' }, 120: { id: 120, name: 'Ada Lovelace' } } } });内嵌 + 对象（再内嵌）
看到下面的 valuesOf(arrayOf(user)) 了没有，它表示该属性是一个对象，对象里面各个数组值是 User对象数组；
var article = new Schema('articles'), user = new Schema('users'), feedSchema, input; article.define({ collaborators: valuesOf(arrayOf(user)) }); feedSchema = { feed: arrayOf(article), suggestions: valuesOf(arrayOf(article)) }; input = { feed: [{ id: 1, title: 'Some Article', collaborators: { authors: [{ id: 3, name: 'Mike Persson' }], reviewers: [{ id: 2, name: 'Pete Hunt' }] } }, { id: 2, title: 'Other Article', collaborators: { authors: [{ id: 2, name: 'Pete Hunt' }] } }, { id: 3, title: 'Last Article' }], suggestions: { 1: [{ id: 2, title: 'Other Article', collaborators: { authors: [{ id: 2, name: 'Pete Hunt' }] } }, { id: 3, title: 'Last Article' }] } }; Object.freeze(input); normalize(input, feedSchema).should.eql({ result: { feed: [1, 2, 3], suggestions: { 1: [2, 3] } }, entities: { articles: { 1: { id: 1, title: 'Some Article', collaborators: { authors: [3], reviewers: [2] } }, 2: { id: 2, title: 'Other Article', collaborators: { authors: [2] } }, 3: { id: 3, title: 'Last Article' } }, users: { 2: { id: 2, name: 'Pete Hunt' }, 3: { id: 3, name: 'Mike Persson' } } } });还有更加复杂的，这次用上 valuesOf(userOrGroup, { schemaAttribute: 'type' }) 了：
var article = new Schema('articles'), user = new Schema('users'), group = new Schema('groups'), userOrGroup = { users: user, groups: group }, feedSchema, input; article.define({ collaborators: valuesOf(userOrGroup, { schemaAttribute: 'type' }) }); feedSchema = { feed: arrayOf(article), suggestions: valuesOf(arrayOf(article)) }; input = { feed: [{ id: 1, title: 'Some Article', collaborators: { author: { id: 3, type: 'users', name: 'Mike Persson' }, reviewer: { id: 2, type: 'groups', name: 'Reviewer Group' } } }, { id: 2, title: 'Other Article', collaborators: { author: { id: 2, type: 'users', name: 'Pete Hunt' } } }, { id: 3, title: 'Last Article' }], suggestions: { 1: [{ id: 2, title: 'Other Article' }, { id: 3, title: 'Last Article' }] } }; Object.freeze(input); normalize(input, feedSchema).should.eql({ result: { feed: [1, 2, 3], suggestions: { 1: [2, 3] } }, entities: { articles: { 1: { id: 1, title: 'Some Article', collaborators: { author: { id: 3, schema: 'users' }, reviewer: { id: 2, schema: 'groups' } } }, 2: { id: 2, title: 'Other Article', collaborators: { author: { id: 2, schema: 'users' } } }, 3: { id: 3, title: 'Last Article' } }, users: { 2: { id: 2, type: 'users', name: 'Pete Hunt' }, 3: { id: 3, type: 'users', name: 'Mike Persson' } }, groups: { 2: { id: 2, type: 'groups', name: 'Reviewer Group' } } } });递归调用
比如某某人关注了另外的人， 用户 写了一系列文章，该文章 被其他用户 订阅就是这种情况：
var article = new Schema('articles'), user = new Schema('users'), collection = new Schema('collections'), feedSchema, input; user.define({ articles: arrayOf(article) }); article.define({ collections: arrayOf(collection) }); collection.define({ subscribers: arrayOf(user) }); feedSchema = { feed: arrayOf(article) }; input = { feed: [{ id: 1, title: 'Some Article', collections: [{ id: 1, title: 'Awesome Writing', subscribers: [{ id: 4, name: 'Andy Warhol', articles: [{ id: 1, title: 'Some Article' }] }, { id: 100, name: 'T.S. Eliot', articles: [{ id: 1, title: 'Some Article' }] }] }, { id: 7, title: 'Even Awesomer', subscribers: [{ id: 100, name: 'T.S. Eliot', articles: [{ id: 1, title: 'Some Article' }] }] }] }] }; Object.freeze(input); normalize(input, feedSchema).should.eql({ result: { feed: [1] }, entities: { articles: { 1: { id: 1, title: 'Some Article', collections: [1, 7] } }, collections: { 1: { id: 1, title: 'Awesome Writing', subscribers: [4, 100] }, 7: { id: 7, title: 'Even Awesomer', subscribers: [100] } }, users: { 4: { id: 4, name: 'Andy Warhol', articles: [1] }, 100: { id: 100, name: 'T.S. Eliot', articles: [1] } } } });上面还算好的，有些schema直接就递归声明了，比如 儿女和父母 的关系：
var user = new Schema('users'), input; user.define({ parent: user }); input = { id: 1, name: 'Andy Warhol', parent: { id: 7, name: 'Tom Dale', parent: { id: 4, name: 'Pete Hunt' } } }; Object.freeze(input); normalize(input, user).should.eql({ result: 1, entities: { users: { 1: { id: 1, name: 'Andy Warhol', parent: 7 }, 7: { id: 7, name: 'Tom Dale', parent: 4 }, 4: { id: 4, name: 'Pete Hunt' } } } });自动merge属性
在一个数组里面，如果id属性一致，会自动抽取并合属性成一个：
var writer = new Schema('writers'), book = new Schema('books'), schema = arrayOf(writer), input; writer.define({ books: arrayOf(book) }); input = [{ id: 3, name: 'Jo Rowling', isBritish: true, location: { x: 100, y: 200, nested: ['hello', { world: true }] }, books: [{ id: 1, soldWell: true, name: 'Harry Potter' }] }, { id: 3, name: 'Jo Rowling', bio: 'writer', location: { x: 100, y: 200, nested: ['hello', { world: true }] }, books: [{ id: 1, isAwesome: true, name: 'Harry Potter' }] }]; normalize(input, schema).should.eql({ result: [3, 3], entities: { writers: { 3: { id: 3, isBritish: true, name: 'Jo Rowling', bio: 'writer', books: [1], location: { x: 100, y: 200, nested: ['hello', { world: true }] } } }, books: { 1: { id: 1, isAwesome: true, soldWell: true, name: 'Harry Potter' } } } });如果合并过程中有冲突会有提示，并自动剔除冲突的属性；比如下方同一个作者写的书，一个对象里描述“卖得好”，而在另外一个对象里却描述“卖得差”，明显是有问题的：
var writer = new Schema('writers'), book = new Schema('books'), schema = arrayOf(writer), input; writer.define({ books: arrayOf(book) }); input = [{ id: 3, name: 'Jo Rowling', books: [{ id: 1, soldWell: true, name: 'Harry Potter' }] }, { id: 3, name: 'Jo Rowling', books: [{ id: 1, soldWell: false, name: 'Harry Potter' }] }]; var warnCalled = false, realConsoleWarn; function mockWarn() { warnCalled = true; } realConsoleWarn = console.warn; console.warn = mockWarn; normalize(input, schema).should.eql({ result: [3, 3], entities: { writers: { 3: { id: 3, name: 'Jo Rowling', books: [1] } }, books: { 1: { id: 1, soldWell: true, name: 'Harry Potter' } } } }); warnCalled.should.eq(true); console.warn = realConsoleWarn;传入不存在的schema规范
如果应用的schma规范不存在，你还传入，就会创建一个新的父属性：
var writer = new Schema('writers'), schema = writer, input; input = { id: 'constructor', name: 'Constructor', isAwesome: true }; normalize(input, schema).should.eql({ result: 'constructor', entities: { writers: { constructor: { id: 'constructor', name: 'Constructor', isAwesome: true } } }