feathers-sequelize

Caution: When you're using feathers v4 and want to upgrade to feathers v5, please make sure to read the migration guide.

NOTE: This is the version for Feathers v5. For Feathers v4 use feathers-sequelize v6

A Feathers database adapter for Sequelize, an ORM for Node.js. It supports PostgreSQL, MySQL, MariaDB, SQLite and MSSQL and features transaction support, relations, read replication and more.

API
Caveats
- Sequelize raw queries
- Working with MSSQL
Example
Associations
- Embrace the ORM
- Setting params.sequelize.include
Querying
- Querying a nested column
Working with Sequelize Model instances
Validation
Errors
Testing sequelize queries in isolation
- 1. Build a test file
- 2. Integrate the query using a "before" hook
Migrations
License
Migrating to feathers v5

Very Important: Before using this adapter you have to be familiar with both, the Feathers Basics and general use of Sequelize. For associations and relations see the associations section. This adapter may not cover all use cases but they can still be implemented using Sequelize models directly in a Custom Feathers service.

npm install --save feathers-sequelize@pre

And one of the following:

npm install --save pg pg-hstore
npm install --save mysql2 // For both mysql and mariadb dialects
npm install --save sqlite3
npm install --save tedious // MSSQL

Important: feathers-sequelize implements the Feathers Common database adapter API and querying syntax. For more information about models and general Sequelize usage, follow up in the Sequelize documentation.

API

`new SequelizeService(options)`

Returns a new service instance initialized with the given options.

const Model = require('./models/mymodel');
const { SequelizeService } = require('feathers-sequelize');

app.use('/messages', new SequelizeService({ Model }));
app.use('/messages', new SequelizeService({ Model, id, events, paginate }));

Options:

Model (required) - The Sequelize model definition
id (optional, default: primary key of the model) - The name of the id field property. Will use the first property with primaryKey: true by default.
raw (optional, default: true) - Runs queries faster by returning plain objects instead of Sequelize models.
Sequelize (optional, default: Model.sequelize.Sequelize) - The Sequelize instance
events (optional) - A list of custom service events sent by this service
paginate (optional) - A pagination object containing a default and max page size
multi (optional) - Allow create with arrays and update and remove with id null to change multiple items. Can be true for all methods or an array of allowed methods (e.g. [ 'remove', 'create' ])
operatorMap (optional) - A mapping from query syntax property names to to Sequelize secure operators
operators (optional) - An array of additional query operators to allow (e..g [ '$regex', '$geoNear' ]). Default is the supported operators
filters (optional) - An object of additional query parameters to allow (e..g { '$post.id$': true }).`

params.sequelize

When making a service method call, params can contain an sequelize property which allows to pass additional Sequelize options. This can e.g. be used to retrieve associations. Normally this wil be set in a before hook:

app.service('messages').hooks({
  before: {
    find(context) {
      // Get the Sequelize instance. In the generated application via:
      const sequelize = context.app.get('sequelizeClient');
      const { User } = sequelize.models;

      context.params.sequelize = {
        include: [ User ]
      }

      return context;
    }
  }
});

Other options that params.sequelize allows you to pass can be found in Sequelize querying docs. Beware that when setting a top-level where property (usually for querying based on a column on an associated model), the where in params.sequelize will overwrite your query.

This library offers some additional functionality when using sequelize.returning in services that support multi. The multi option allows you to create, patch, and remove multiple records at once. When using sequelize.returning with multi, the sequelize.returning is used to indicate if the method should return any results. This is helpful when updating large numbers of records and you do not need the API (or events) to be bogged down with results.

operatorMap

Sequelize deprecated string based operators a while ago for security reasons. Starting at version 4.0.0 feathers-sequelize converts queries securely, so you can still use string based operators listed below. If you want to support additional Sequelize operators, the operatorMap service option can contain a mapping from query parameter name to Sequelize operator. By default supported are:

'$eq',
'$ne',
'$gte',
'$gt',
'$lte',
'$lt',
'$in',
'$nin',
'$like',
'$notLike',
'$iLike',
'$notILike',
'$or',
'$and'

// Find all users with name similar to Dav
app.service('users').find({
  query: {
    name: {
      $like: 'Dav%'
    }
  }
});

GET /users?name[$like]=Dav%

Caveats

Sequelize `raw` queries

By default, all feathers-sequelize operations will return raw data (using raw: true when querying the database). This results in faster execution and allows feathers-sequelize to interoperate with feathers-common hooks and other 3rd party integrations. However, this will bypass some of the "goodness" you get when using Sequelize as an ORM:

custom getters/setters will be bypassed
model-level validations are bypassed
associated data loads a bit differently
...and several other issues that one might not expect

Don't worry! The solution is easy. Please read the guides about working with model instances. You can also pass { raw: true/false} in params.sequelize to change the behavior per service call.

Working with MSSQL

When using MSSQL as the database, a default sort order always has to be applied, otherwise the adapter will throw an Invalid usage of the option NEXT in the FETCH statement. error. This can be done in your model with:

model.beforeFind(model => model.order.push(['id', 'ASC']))

Or in a hook like this:

export default function (options = {}) {
  return async context => {
    const { query = {} } = context.params;
    // Sort by id field ascending (or any other property you want)
    // See https://docs.feathersjs.com/api/databases/querying.html#sort
    const $sort = { id: 1 };

    context.params.query = {
      $sort: {

      },
      ...query
    }

    return context;
  }
}

Primary keys

All tables used by a feathers-sequelize service require a primary key. Although it is common practice for many-to-many tables to not have a primary key, this service will break if the table does not have a primary key. This is because most service methods require an ID and because of how feathers maps services to URLs.

Example

Here is an example of a Feathers server with a messages SQLite Sequelize Model:

$ npm install @feathersjs/feathers @feathersjs/errors @feathersjs/express @feathersjs/socketio sequelize feathers-sequelize sqlite3

In app.js:

import path from 'path';
import { feathers } from '@feathersjs/feathers';
import express from '@feathersjs/express';
import socketio from '@feathersjs/socketio';

import Sequelize from 'sequelize';
import SequelizeService from 'feathers-sequelize';

const sequelize = new Sequelize('sequelize', '', '', {
  dialect: 'sqlite',
  storage: path.join(__dirname, 'db.sqlite'),
  logging: false
});

const Message = sequelize.define('message', {
  text: {
    type: Sequelize.STRING,
    allowNull: false
  }
}, {
  freezeTableName: true
});

// Create an Express compatible Feathers application instance.
const app = express(feathers());

// Turn on JSON parser for REST services
app.use(express.json());
// Turn on URL-encoded parser for REST services
app.use(express.urlencoded({ extended: true }));
// Enable REST services
app.configure(express.rest());
// Enable Socket.io services
app.configure(socketio());
// Create an in-memory Feathers service with a default page size of 2 items
// and a maximum size of 4
app.use('/messages', new SequelizeService({
  Model: Message,
  paginate: {
    default: 2,
    max: 4
  }
}));
app.use(express.errorHandler());

Message.sync({ force: true }).then(() => {
  // Create a dummy Message
  app.service('messages').create({
    text: 'Message created on server'
  }).then(message => console.log('Created message', message));
});

// Start the server
const port = 3030;

app.listen(port, () => {
  console.log(`Feathers server listening on port ${port}`);
});

Run the example with node app and go to localhost:3030/messages.

Associations

Embrace the ORM

The documentation on Sequelize associations and relations is essential to implementing associations with this adapter and one of the steepest parts of the Sequelize learning curve. If you have never used an ORM, let it do a lot of the heavy lifting for you!

Setting `params.sequelize.include`

Once you understand how the include option works with Sequelize, you will want to set that option from a before hook in Feathers. Feathers will pass the value of context.params.sequelize as the options parameter for all Sequelize method calls. This is what your hook might look like:

// GET /my-service?name=John&include=1
function (context) {
  const { include, ...query } = context.params.query;

   if (include) {
      const AssociatedModel = context.app.services.fooservice.Model;
      context.params.sequelize = {
        include: [{ model: AssociatedModel }]
      };
      // Update the query to not include `include`
      context.params.query = query;
   }

   return context;
}

Underneath the hood, feathers will call your models find method sort of like this:

// YourModel is a sequelize model
const options = Object.assign({ where: { name: 'John' }}, context.params.sequelize);
YourModel.findAndCount(options);

For more information, follow up up in the Sequelize documentation for associations and this issue.

Querying

Additionally to the common querying mechanism this adapter also supports all Sequelize query operators.

Querying a nested column

To query based on a column in an associated model, you can use Sequelize's nested column syntax in a query. The nested column syntax is considered a filter by Feathers, and so each such usage has to be whitelisted.

Example:

// Find a user with post.id == 120
app.service('users').find({
  query: {
    '$post.id$': 120,
    include: {
      model: posts
    }
  }
});

For this case to work, you'll need to add '$post.id$' to the service options' 'filters' property.

Working with Sequelize Model instances

It is highly recommended to use raw queries, which is the default. However, there are times when you will want to take advantage of Sequelize Instance methods. There are two ways to tell feathers to return Sequelize instances:

Set { raw: false } in a "before" hook:

const rawFalse = () => (context) => {
  if (!context.params.sequelize) context.params.sequelize = {};
  Object.assign(context.params.sequelize, { raw: false });
  return context;
}

export default {
  after: {
    // ...
    find: [rawFalse()]
    // ...
  },
  // ...
};

Use the hydrate hook in the "after" phase:

import { hydrate } from 'feathers-sequelize';

export default {
  after: {
    // ...
    find: [hydrate()]
    // ...
  },
  // ...
};

// Or, if you need to include associated models, you can do the following:
const includeAssociated = () => (context) => hydrate({
  include: [{ model: context.app.services.fooservice.Model }]
});

export default {
  after: {
    // ...
    find: [includeAssociated()]
    // ...
  },
  // ...
};

For a more complete example see this gist.

Important: When working with Sequelize Instances, most of the feathers-hooks-common will no longer work. If you need to use a common hook or other 3rd party hooks, you should use the "dehydrate" hook to convert data back to a plain object:
import { dehydrate, hydrate } from 'feathers-sequelize';
import { populate } = from 'feathers-hooks-common';

export default {
  after: {
    // ...
    find: [hydrate(), doSomethingCustom(), dehydrate(), populate()]
    // ...
  },
  // ...
};

Validation

Sequelize by default gives you the ability to add validations at the model level. Using an error handler like the one that comes with Feathers your validation errors will be formatted nicely right out of the box!

Errors

Errors do not contain Sequelize specific information. The original Sequelize error can be retrieved on the server via:

import { ERROR } = from 'feathers-sequelize';

try {
  await sequelizeService.doSomething();
} catch(error) {
  // error is a FeathersError
  // Safely retrieve the Sequelize error
  const sequelizeError = error[ERROR];
}

Testing sequelize queries in isolation

If you wish to use some of the more advanced features of sequelize, you should first test your queries in isolation (without feathers). Once your query is working, you can integrate it into your feathers app.

1. Build a test file

Create a temporary file in your project root like this:

// test.js
import app from from './src/app';
// run setup to initialize relations
app.setup();

const seqClient = app.get('sequelizeClient');
const SomeModel = seqClient.models['some-model'];
const log = console.log.bind(console);

SomeModel.findAll({
  /*
  * Build your custom query here. We will use this object later.
  */
}).then(log).catch(log);

And then run this file like this:

node test.js

Continue updating the file and running it until you are satisfied with the results.

2. Integrate the query using a "before" hook

Once your have your custom query working to your satisfaction, you will want to integrate it into your feathers app. Take the guts of the findAll operation above and create a "before" hook:

function buildCustomQuery(context) {
	context.params.sequelize = {
       /*
        * This is the same object you passed to "findAll" above.
        * This object is *shallow merged* onto the underlying query object
        * generated by feathers-sequelize (it is *not* a deep merge!).
        * The underlying data will already contain the following:
        *   - "where" condition based on query paramters
        *   - "limit" and "offset" based on pagination settings
        *   - "order" based $sort query parameter
        * You can override any/all of the underlying data by setting it here.
        * This gives you full control over the query object passed to sequelize!
        */
	};
}

someService.hooks({
	before: {
		find: [buildCustomQuery]
	}
});

Migrations

Migrations with feathers and sequelize are quite simple. This guide will walk you through creating the recommended file structure, but you are free to rearrange things as you see fit. The following assumes you have a migrations folder in the root of your app.

Initial Setup: one-time tasks

Install the sequelize CLI:

npm install sequelize-cli --save -g

Create a .sequelizerc file in your project root with the following content:

const path = require('path');

module.exports = {
  'config': path.resolve('migrations/config.js'),
  'migrations-path': path.resolve('migrations/scripts'),
  'seeders-path': path.resolve('migrations/seeders'),
  'models-path': path.resolve('migrations/models.js')
};

Create the migrations config in migrations/config.js:

const app = require('../src/app');
const env = process.env.NODE_ENV || 'development';
const dialect = 'postgres'; // Or your dialect name

module.exports = {
  [env]: {
    dialect,
    url: app.get(dialect),
    migrationStorageTableName: '_migrations'
  }
};

Define your models config in migrations/models.js:

const Sequelize = require('sequelize');
const app = require('../src/app');
const sequelize = app.get('sequelizeClient');
const models = sequelize.models;

// The export object must be a dictionary of model names -> models
// It must also include sequelize (instance) and Sequelize (constructor) properties
module.exports = Object.assign({
  Sequelize,
  sequelize
}, models);

Migrations workflow

The migration commands will load your application and it is therefore required that you define the same environment variables as when running your application. For example, many applications will define the database connection string in the startup command:

DATABASE_URL=postgres://user:pass@host:port/dbname npm start

All of the following commands assume that you have defined the same environment variables used by your application.

ProTip: To save typing, you can export environment variables for your current bash/terminal session:

export DATABASE_URL=postgres://user:pass@host:port/db

Create a new migration

To create a new migration file, run the following command and provide a meaningful name:

sequelize migration:create --name="meaningful-name"

This will create a new file in the migrations/scripts folder. All migration file names will be prefixed with a sortable data/time string: 20160421135254-meaningful-name.js. This prefix is crucial for making sure your migrations are executed in the proper order.

NOTE: The order of your migrations is determined by the alphabetical order of the migration scripts in the file system. The file names generated by the CLI tools will always ensure that the most recent migration comes last.

Add the up/down scripts:

Open the newly created migration file and write the code to both apply and undo the migration. Please refer to the sequelize migration functions for available operations. Do not be lazy - write the down script too and test! Here is an example of converting a NOT NULL column accept null values:

'use strict';

module.exports = {
  up: function (queryInterface, Sequelize) {
    return queryInterface.changeColumn('tableName', 'columnName', {
      type: Sequelize.STRING,
      allowNull: true
    });
  },

  down: function (queryInterface, Sequelize) {
    return queryInterface.changeColumn('tableName', 'columnName', {
      type: Sequelize.STRING,
      allowNull: false
    });
  }
};

ProTip: As of this writing, if you use the changeColumn method you must always specify the type, even if the type is not changing.

ProTip: Down scripts are typically easy to create and should be nearly identical to the up script except with inverted logic and inverse method calls.

Keeping your app code in sync with migrations

The application code should always be up to date with the migrations. This allows the app to be freshly installed with everything up-to-date without running the migration scripts. Your migrations should also never break a freshly installed app. This often times requires that you perform any necessary checks before executing a task. For example, if you update a model to include a new field, your migration should first check to make sure that new field does not exist:

'use strict';

module.exports = {
  up: function (queryInterface, Sequelize) {
    return queryInterface.describeTable('tableName').then(attributes => {
      if ( !attributes.columnName ) {
        return queryInterface.addColumn('tableName', 'columnName', {
          type: Sequelize.INTEGER,
          defaultValue: 0
        });
      }
    })
  },

  down: function (queryInterface, Sequelize) {
    return queryInterface.describeTable('tableName').then(attributes => {
      if ( attributes.columnName ) {
        return queryInterface.removeColumn('tableName', 'columnName');
      }
    });
  }
};

Apply a migration

The CLI tools will always run your migrations in the correct order and will keep track of which migrations have been applied and which have not. This data is stored in the database under the _migrations table. To ensure you are up to date, simply run the following:

sequelize db:migrate

ProTip: You can add the migrations script to your application startup command to ensure that all migrations have run every time your app is started. Try updating your package.json scripts attribute and run npm start:

scripts: {
    start: "sequelize db:migrate && node src/"
}

Undo the previous migration

To undo the last migration, run the following command:

sequelize db:migrate:undo

Continue running the command to undo each migration one at a time - the migrations will be undone in the proper order.

Note: - You shouldn't really have to undo a migration unless you are the one developing a new migration and you want to test that it works. Applications rarely have to revert to a previous state, but when they do you will be glad you took the time to write and test your down scripts!

Reverting your app to a previous state

In the unfortunate case where you must revert your app to a previous state, it is important to take your time and plan your method of attack. Every application is different and there is no one-size-fits-all strategy for rewinding an application. However, most applications should be able to follow these steps (order is important):

Stop your application (kill the process)
Find the last stable version of your app
Count the number of migrations which have been added since that version
Undo your migrations one at a time until the db is in the correct state
Revert your code back to the previous state
Start your app

License

Licensed under the MIT license.

whitelist

The whitelist property is no longer, you should use filters instead. Checkout the migration guide below.

Feathers v5 introduces a convention for options.operators and options.filters. The way feathers-sequelize worked in previous version is not compatible with these conventions. Please read https://dove.feathersjs.com/guides/migrating.html#custom-filters-operators.

Migrate to Feathers v5 (dove)

There are several breaking changes for feathers-sequelize in Feathers v5. This guide will help you to migrate your existing Feathers v4 application to Feathers v5.

Named export

The default export of feathers-sequelize has been removed. You now have to import the SequelizeService class directly:

import { SequelizeService } from 'feathers-sequelize';

app.use('/messages', new SequelizeService({ ... }));

This follows conventions from feathers v5.

operators / operatorMap

Feathers v5 introduces a convention for options.operators and options.filters. The way feathers-sequelize worked in previous version is not compatible with these conventions. Please read https://dove.feathersjs.com/guides/migrating.html#custom-filters-operators first.

The old options.operators object is renamed to options.operatorMap:

import { SequelizeService } from 'feathers-sequelize';
import { Op } from 'sequelize';

app.use('/messages', new SequelizeService({
  Model,
  // operators is now operatorMap:
  operatorMap: {
    $between: Op.between
  }
}));

filters