normalizr

Normalizes deeply nested JSON API responses according to a schema for Flux and Redux apps.
Kudos to Jing Chen for suggesting this approach.

Installation

npm install --save normalizr

Sample App

Flux

See flux-react-router-example.

Redux

See redux/examples/real-world.

The Problem

• You have a JSON API that returns deeply nested objects;
• You want to port your app to Flux or Redux;
• You noticed it's hard for Stores (or Reducers) to consume data from nested API responses.

Normalizr takes JSON and a schema and replaces nested entities with their IDs, gathering all entities in dictionaries.

For example,

[{
  id: 1,
  title: 'Some Article',
  author: {
    id: 1,
    name: 'Dan'
  }
}, {
  id: 2,
  title: 'Other Article',
  author: {
    id: 1,
    name: 'Dan'
  }
}]

can be normalized to

{
  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'
      }
    }
  }
}

Note the flat structure (all nesting is gone).

Features

• Entities can be nested inside other entities, objects and arrays;
• Combine entity schemas to express any kind of API response;
• Entities with same IDs are automatically merged (with a warning if they differ);
• Allows using a custom ID attribute (e.g. slug).

Usage

import { normalize, Schema, arrayOf } from 'normalizr';

First, define a schema for our entities:

const article = new Schema('articles')
const user = new Schema('users');
const collection = new Schema('collections');

Then we define nesting rules:

article.define({
  author: user,
  collections: arrayOf(collection)
});

collection.define({
  curator: user
});

Now we can use this schema in our API response handlers:

const ServerActionCreators = {

  // These are two different XHR endpoints with different response schemas.
  // We can use the schema objects defined earlier to express both of them:

  receiveArticles(response) {
  
    // Passing { articles: arrayOf(article) } as second parameter to normalize()
    // lets it correctly traverse the response tree and gather all entities:
    
    // BEFORE
    // {
    //   articles: [{
    //     id: 1,
    //     title: 'Some Article',
    //     author: {
    //       id: 7,
    //       name: 'Dan',
    //     }
    //   }, ...]
    // }
    //
    // AFTER:
    // {
    //   result: {
    //    articles: [1, 2, ...] // <--- Note how object array turned into ID array
    //   },
    //   entities: {
    //     articles: {
    //       1: { author: 7, ... }, // <--- Same happens for references to other entities in the schema
    //       2: { ... },
    //       ...
    //     },
    //     users: {
    //       7: { ... },
    //       ..
    //     }
    //   }
    
    response = normalize(response, {
      articles: arrayOf(article)
    });

    AppDispatcher.handleServerAction({
      type: ActionTypes.RECEIVE_ARTICLES,
      response: 
    });
  },
  
  // Though this is a different API endpoint, we can describe it just as well
  // with our normalizr schema objects:

  receiveUsers(response) {

    // Passing { users: arrayOf(user) } as second parameter to normalize()
    // lets it correctly traverse the response tree and gather all entities:
    
    // BEFORE
    // {
    //   users: [{
    //     id: 7,
    //     name: 'Dan',
    //     ...
    //   }, ...]
    // }
    //
    // AFTER:
    // {
    //   result: {
    //    users: [7, ...] // <--- Note how object array turned into ID array
    //   },
    //   entities: {
    //     users: {
    //       7: { ... },
    //       ..
    //     }
    //   }
    

    response = normalize(response, {
      users: arrayOf(user)
    });

    AppDispatcher.handleServerAction({
      type: ActionTypes.RECEIVE_USERS,
      response
    });
  }
}

Finally, different Stores can tune in to listen to all API responses and grab entity lists from action.response.entities:

AppDispatcher.register((payload) => {
  const { action } = payload;

  if (action.response && action.response.entities && action.response.entities.users) {
    mergeUsers(action.response.entities.users);
    UserStore.emitChange();
    break;
  }
});

API Reference

new Schema(key, [options])

Schema lets you define a type of entity returned by your API.
This should correspond to model in your server code.

The key parameter lets you specify the name of the dictionary for this kind of entity.

const article = new Schema('articles');

// You can use a custom id attribute
const article = new Schema('articles', { idAttribute: 'slug' });

Schema.prototype.define(nestedSchema)

Lets you specify relationships between different entities.

const article = new Schema('articles');
const user = new Schema('users');

article.define({
  author: user
});

arrayOf(schema)

Describes an array of the schema passed as argument.

const article = new Schema('articles');
const user = new Schema('users');

article.define({
  author: user,
  contributors: arrayOf(user)
});

normalize(obj, schema, [options])

Normalizes object according to schema.
Passed schema should be a nested object reflecting the structure of API response.

You may optionally specify a custom assignEntity function in options. This is useful if your backend emits additional fields, such as separate ID fields, you'd like to delete in the normalized entity. See the test and the discussion for a usage example.

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));

Explanation by Example

Say, you have /articles API with the following schema:

articles: article*

article: {
  author: user,
  likers: user*
  primary_collection: collection?
  collections: collection*
}

collection: {
  curator: user
}

Without normalizr, your Stores would need to know too much about API response schema.
For example, UserStore would include a lot of boilerplate to extract fresh user info when articles are fetched:

// Without normalizr, you'd have to do this in every store:

AppDispatcher.register((payload) => {
  const { action } = payload;

  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;
  }
});

Normalizr solves the problem by converting API responses to a flat form where nested entities are replaced with IDs:

{
  result: [12, 10, 3, ...],
  entities: {
    articles: {
      12: {
        authorId: 3,
        likers: [2, 1, 4],
        primaryCollection: 12,
        collections: [12, 11]
      },
      ...
    },
    users: {
      3: {
        name: 'Dan'
      },
      2: ...,
      4: ....
    },
    collections: {
      12: {
        curator: 2,
        name: 'Stuff'
      },
      ...
    }
  }
}

Then UserStore code can be rewritten as:

// With normalizr, users are always in action.entities.users

AppDispatcher.register((payload) => {
  const { action } = payload;

  if (action.response && action.response.entities && action.response.entities.users) {
    mergeUsers(action.response.entities.users);
    UserStore.emitChange();
    break;
  }
});

Dependencies

• lodash for isObject and isEqual

Running Tests

git clone https://github.com/gaearon/normalizr.git
cd normalizr
npm install
npm test # run tests once
npm run test:watch # run test watcher