Umzug

Umzug is a framework-agnostic migration tool for Node. It provides a clean API for running and rolling back tasks.

Highlights

• Written in TypeScript
  • Built-in typings
  • Auto-completion right in your IDE
  • Documentation right in your IDE
• Programmatic API for migrations
• Built-in CLI
• Database agnostic
• Supports logging of migration process
• Supports multiple storages for migration data
• Usage examples

Documentation

Note: these are the docs for the latest version of umzug, which has several breaking changes from v2.x. See the upgrading section for a migration guide. For the previous stable version, please refer to the v2.x branch.

Minimal Example

The following example uses a Sqlite database through sequelize and persists the migration data in the database itself through the sequelize storage. There are several more involved examples covering a few different scenarios in the examples folder. Note that although this uses Sequelize, Umzug isn't coupled to Sequelize, it's just one of the (most commonly-used) supported storages.

// index.js
const {Sequelize} = require('sequelize')
const {Umzug, SequelizeStorage} = require('umzug')

const sequelize = new Sequelize({dialect: 'sqlite', storage: './db.sqlite'})

const umzug = new Umzug({
  migrations: {glob: 'migrations/*.js'},
  context: sequelize.getQueryInterface(),
  storage: new SequelizeStorage({sequelize}),
  logger: console,
})

// Checks migrations and run them if they are not already applied. To keep
// track of the executed migrations, a table (and sequelize model) called SequelizeMeta
// will be automatically created (if it doesn't exist already) and parsed.
await umzug.up()

// migrations/00_initial.js

const {Sequelize} = require('sequelize')

async function up({context: queryInterface}) {
  await queryInterface.createTable('users', {
    id: {
      type: Sequelize.INTEGER,
      allowNull: false,
      primaryKey: true,
    },
    name: {
      type: Sequelize.STRING,
      allowNull: false,
    },
    createdAt: {
      type: Sequelize.DATE,
      allowNull: false,
    },
    updatedAt: {
      type: Sequelize.DATE,
      allowNull: false,
    },
  })
}

async function down({context: queryInterface}) {
  await queryInterface.dropTable('users')
}

module.exports = {up, down}

Note that we renamed the context argument to queryInterface for clarity. The context is whatever we specified when creating the Umzug instance in index.js.

// index.ts
require('ts-node/register')

import { Sequelize } from 'sequelize';
import { Umzug, SequelizeStorage } from 'umzug';

const sequelize = new Sequelize({ dialect: 'sqlite', storage: './db.sqlite' });

const umzug = new Umzug({
  migrations: { glob: 'migrations/*.ts' },
  context: sequelize.getQueryInterface(),
  storage: new SequelizeStorage({ sequelize }),
  logger: console,
});

// export the type helper exposed by umzug, which will have the `context` argument typed correctly
export type Migration = typeof umzug._types.migration;

(async () => {
  await umzug.up();
})();

// migrations/00_initial.ts
import type { Migration } from '..';

// types will now be available for `queryInterface`
export const up: Migration = ({ context: queryInterface }) => queryInterface.createTable(...)
export const down: Migration = ({ context: queryInterface }) => queryInterface.dropTable(...)

See these tests for more examples of Umzug usage, including:

• passing ignore and cwd parameters to the glob instructions
• customising migrations ordering
• finding migrations from multiple different directories
• using non-js file extensions via a custom resolver, e.g. .sql

Usage

Installation

Umzug is available on npm by specifying the correct tag:

npm install umzug

Umzug instance

It is possible to configure an Umzug instance by passing an object to the constructor.

const {Umzug} = require('umzug')

const umzug = new Umzug({
  /* ... options ... */
})

Detailed documentation for these options are in the UmzugOptions TypeScript interface, which can be found in src/types.ts.

Getting all pending migrations

You can get a list of pending (i.e. not yet executed) migrations with the pending() method:

const migrations = await umzug.pending()
// returns an array of all pending migrations.

Getting all executed migrations

You can get a list of already executed migrations with the executed() method:

const migrations = await umzug.executed()
// returns an array of all already executed migrations

Executing pending migrations

The up method can be used to execute all pending migrations.

const migrations = await umzug.up()
// returns an array of all executed migrations

It is also possible to pass the name of a migration in order to just run the migrations from the current state to the passed migration name (inclusive).

await umzug.up({to: '20141101203500-task'})

To limit the number of migrations that are run, step can be used:

// This will run the next two migrations
await umzug.up({step: 2})

Running specific migrations while ignoring the right order, can be done like this:

await umzug.up({migrations: ['20141101203500-task', '20141101203501-task-2']})

Reverting executed migration

The down method can be used to revert the last executed migration.

const migration = await umzug.down()
// reverts the last migration and returns it.

To revert more than one migration, you can use step:

// This will revert the last two migrations
await umzug.down({step: 2})

It is possible to pass the name of a migration until which (inclusive) the migrations should be reverted. This allows the reverting of multiple migrations at once.

const migrations = await umzug.down({to: '20141031080000-task'})
// returns an array of all reverted migrations.

To revert all migrations, you can pass 0 as the to parameter:

await umzug.down({to: 0})

Reverting specific migrations while ignoring the right order, can be done like this:

await umzug.down({migrations: ['20141101203500-task', '20141101203501-task-2']})

Migrations

There are two ways to specify migrations: via files or directly via an array of migrations.

Migration files

A migration file ideally exposes an up and a down async functions. They will perform the task of upgrading or downgrading the database.

module.exports = {
  async up() {
    /* ... */
  },
  async down() {
    /* ... */
  },
}

Migration files can be located anywhere - they will typically be loaded according to a glob pattern provided to the Umzug constructor.

Direct migrations list

You can also specify directly a list of migrations to the Umzug constructor:

const {Umzug} = require('umzug')

const umzug = new Umzug({
  migrations: [
    {
      // the name of the migration is mandatory
      name: '00-first-migration',
      async up({context}) {
        /* ... */
      },
      async down({context}) {
        /* ... */
      },
    },
    {
      name: '01-foo-bar-migration',
      async up({context}) {
        /* ... */
      },
      async down({context}) {
        /* ... */
      },
    },
  ],
  context: sequelize.getQueryInterface(),
  logger: console,
})

Modifying the parameters passed to your migration methods

Sometimes it's necessary to modify the parameters umzug will pass to your migration methods when the library calls the up and down methods for each migration. This is the case when using migrations currently generated using sequelize-cli. In this case you can use the resolve fuction during migration configuration to determine which parameters will be passed to the relevant method

import {Sequelize} from 'sequelize'
import {Umzug, SequelizeStorage} from 'umzug'

const sequelize = new Sequelize(/* ... */)

const umzug = new Umzug({
  migrations: {
    glob: 'migrations/*.js',
    resolve: ({name, path, context}) => {
      const migration = require(path)
      return {
        // adjust the parameters Umzug will
        // pass to migration methods when called
        name,
        up: async () => migration.up(context, Sequelize),
        down: async () => migration.down(context, Sequelize),
      }
    },
  },
  context: sequelize.getQueryInterface(),
  storage: new SequelizeStorage({sequelize}),
  logger: console,
})

Additional migration configuration options

To load migrations in another format, you can use the resolve function:

const fs = require('fs')
const {Sequelize} = require('sequelize')
const {Umzug} = require('umzug')

const umzug = new Umzug({
  migrations: {
    glob: 'migrations/*.up.sql',
    resolve: ({name, path, context: sequelize}) => ({
      name,
      up: async () => {
        const sql = fs.readFileSync(path).toString()
        return sequelize.query(sql)
      },
      down: async () => {
        // Get the corresponding `.down.sql` file to undo this migration
        const sql = fs
          .readFileSync(path.replace('.up.sql', '.down.sql'))
          .toString()
        return sequelize.query(sql)
      },
    }),
  },
  context: new Sequelize(/* ... */),
  logger: console,
})

You can support mixed migration file types, and use umzug's default resolver for javascript/typescript:

const fs = require('fs')
const {Sequelize} = require('sequelize')
const {Umzug} = require('umzug')

const umzug = new Umzug({
  migrations: {
    glob: 'migrations/*.{js,ts,up.sql}',
    resolve: params => {
      if (!params.path.endsWith('.sql')) {
        return Umzug.defaultResolver(params)
      }
      const {context: sequelize} = params
      return {
        name: params.name,
        up: async () => {
          const sql = fs.readFileSync(params.path).toString()
          return sequelize.query(sql)
        },
        down: async () => {
          // Get the corresponding `.down.sql` file to undo this migration
          const sql = fs
            .readFileSync(params.path.replace('.up.sql', '.down.sql'))
            .toString()
          return sequelize.query(sql)
        },
      }
    },
  },
  logger: console,
  context: new Sequelize(/* ... */),
})

The glob syntax allows loading migrations from multiple locations:

const {Sequelize} = require('sequelize')
const {Umzug} = require('umzug')

const umzug = new Umzug({
  migrations: {
    glob: '{first-folder/*.js,second-folder-with-different-naming-convention/*.js}',
  },
  context: new Sequelize(/* ... */),
  logger: console,
})

Note on migration file sorting:

• file matches, found using fast-glob, will be lexicographically sorted based on their paths
  • so if your migrations are one/m1.js, two/m2.js, three/m3.js, the resultant order will be one/m1.js, three/m3.js, two/m2.js
  • similarly, if your migrations are called m1.js, m2.js, ... m10.js, m11.js, the resultant ordering will be m1.js, m10.js, m11.js, ... m2.js
• The easiest way to deal with this is to ensure your migrations appear in a single folder, and their paths match lexicographically with the order they should run in
• If this isn't possible, the ordering can be customised using a new instance (previously, in the beta release for v3, this could be done with .extend(...) - see below for example using a new instance)

Upgrading from v2.x

The Umzug class should be imported as a named import, i.e. import { Umzug } from 'umzug'.

The MigrationMeta type, which is returned by umzug.executed() and umzug.pending(), no longer has a file property - it has a name and optional path - since migrations are not necessarily bound to files on the file system.

The migrations.glob parameter replaces path, pattern and traverseDirectories. It can be used, in combination with cwd and ignore to do much more flexible file lookups. See https://npmjs.com/package/fast-glob for more information on the syntax.

The migrations.resolve parameter replaces customResolver. Explicit support for wrap and nameFormatter has been removed - these can be easily implemented in a resolve function.

The constructor option logging is replaced by logger to allow for warn and error messages in future. NodeJS's global console object can be passed to this. To disable logging, replace logging: false with logger: undefined.

Events have moved from the default nodejs EventEmitter to emittery. It has better design for async code, a less bloated API surface and strong types. But, it doesn't allow passing multiple arguments to callbacks, so listeners have to change slightly, as well as .addListener(...) and .removeListener(...) no longer being supported (.on(...) and .off(...) should now be used):

Before:

umzug.on('migrating', (name, m) => console.log({name, path: m.path}))

After:

umzug.on('migrating', ev => console.log({name: ev.name, path: ev.path}))

The Umzug#execute method is removed. Use Umzug#up or Umzug#down.

The options for Umguz#up and Umzug#down have changed:

• umzug.up({ to: 'some-name' }) and umzug.down({ to: 'some-name' }) are still valid.
• umzug.up({ from: '...' }) and umzug.down({ from: '...' }) are no longer supported. To run migrations out-of-order (which is not generally recommended), you can explicitly use umzug.up({ migrations: ['...'] }) and umzug.down({ migrations: ['...'] }).
• name matches must be exact. umzug.up({ to: 'some-n' }) will no longer match a migration called some-name.
• umzug.down({ to: 0 }) is still valid but umzug.up({ to: 0 }) is not.
• umzug.up({ migrations: ['m1', 'm2'] }) is still valid but the shorthand umzug.up(['m1', 'm2']) has been removed.
• umzug.down({ migrations: ['m1', 'm2'] }) is still valid but the shorthand umzug.down(['m1', 'm2']) has been removed.
• umzug.up({ migrations: ['m1', 'already-run'] }) will throw an error, if already-run is not found in the list of pending migrations.
• umzug.down({ migrations: ['m1', 'has-not-been-run'] }) will throw an error, if has-not-been-run is not found in the list of executed migrations.
• umzug.up({ migrations: ['m1', 'm2'], rerun: 'ALLOW' }) will re-apply migrations m1 and m2 even if they've already been run.
• umzug.up({ migrations: ['m1', 'm2'], rerun: 'SKIP' }) will skip migrations m1 and m2 if they've already been run.
• umzug.down({ migrations: ['m1', 'm2'], rerun: 'ALLOW' }) will "revert" migrations m1 and m2 even if they've never been run.
• umzug.down({ migrations: ['m1', 'm2'], rerun: 'SKIP' }) will skip reverting migrations m1 and m2 if they haven't been run or are already reverted.
• umzug.up({ migrations: ['m1', 'does-not-exist', 'm2'] }) will throw an error if the migration name is not found. Note that the error will be thrown and no migrations run unless all migration names are found - whether or not rerun: 'ALLOW' is added.

The context parameter replaces params, and is passed in as a property to migration functions as an options object, alongs side name and path. This means the signature for migrations, which in v2 was (context) => Promise, has changed slightly in v3, to ({ name, path, context }) => Promise.

Handling existing v2-format migrations

The resolve function can also be used to upgrade your umzug version to v3 when you have existing v2-compatible migrations:

const {Umzug} = require('umzug')

const umzug = new Umzug({
  migrations: {
    glob: 'migrations/umzug-v2-format/*.js',
    resolve: ({name, path, context}) => {
      // Adjust the migration from the new signature to the v2 signature, making easier to upgrade to v3
      const migration = require(path)
      return {
        name,
        up: async () => migration.up(context),
        down: async () => migration.down(context),
      }
    },
  },
  context: sequelize.getQueryInterface(),
  logger: console,
})

Similarly, you no longer need migrationSorting, you can instantiate a new Umzug instance to manipulate migration lists directly:

const {Umzug} = require('umzug')

const parent = new Umzug({
  migrations: {glob: 'migrations/**/*.js'},
  context: sequelize.getQueryInterface(),
})

const umzug = new Umzug({
  ...parent.options,
  migrations: async ctx =>
    (await parent.migrations()).sort((a, b) => b.path.localeCompare(a.path)),
})

Storages

Storages define where the migration data is stored.

JSON Storage

Using JSONStorage will create a JSON file which will contain an array with all the executed migrations. You can specify the path to the file. The default for that is umzug.json in the working directory of the process.

Detailed documentation for the options it can take are in the JSONStorageConstructorOptions TypeScript interface, which can be found in src/storage/json.ts.

Memory Storage

Using memoryStorage will store migrations with an in-memory array. This can be useful for proof-of-concepts or tests, since it doesn't interact with databases or filesystems.

It doesn't take any options, just import the memoryStorage function and call it to return a storage instance:

import { Umzug, memoryStorage } from 'umzug'

const umzug = new Umzug({
  migrations: ...,
  storage: memoryStorage(),
  logger: console,
})

Sequelize Storage

Using SequelizeStorage will create a table in your SQL database called SequelizeMeta containing an entry for each executed migration. You will have to pass a configured instance of Sequelize or an existing Sequelize model. Optionally you can specify the model name, table name, or column name. All major Sequelize versions are supported.

Detailed documentation for the options it can take are in the _SequelizeStorageConstructorOptions TypeScript interface, which can be found in src/storage/sequelize.ts.

This library has been tested with sequelize v6. It may or may not work with lower versions - use at your own risk.

MongoDB Storage

Using MongoDBStorage will create a collection in your MongoDB database called migrations containing an entry for each executed migration. You will have either to pass a MongoDB Driver Collection as collection property. Alternatively you can pass a established MongoDB Driver connection and a collection name.

Detailed documentation for the options it can take are in the MongoDBStorageConstructorOptions TypeScript interface, which can be found in src/storage/mongodb.ts.

Custom

In order to use a custom storage, you can pass your storage instance to Umzug constructor.

class CustomStorage {
  constructor() {}
  logMigration() {}
  unlogMigration() {}
  executed() {}
}

const umzug = new Umzug({
  storage: new CustomStorage(/* ... */),
  logger: console,
})

Your instance must adhere to the UmzugStorage interface. If you're using TypeScript you can ensure this at compile time, and get IDE type hints by importing it:

import { UmzugStorage } from 'umzug'

class CustomStorage implements UmzugStorage {
  /* ... */
}

Events

Umzug is an emittery event emitter. Each of the following events will be called with migration parameters as its payload (with context, name, and nullable path properties). Events are a convenient place to implement application-specific logic that must run around each migration:

• migrating - A migration is about to be executed.
• migrated - A migration has successfully been executed.
• reverting - A migration is about to be reverted.
• reverted - A migration has successfully been reverted.

These events run at the beginning and end of up and down calls. They'll receive an object containing a context property:

• beforeCommand - Before any command ('up' | 'down' | 'executed' | 'pending') is run.
• afterCommand - After any command ('up' | 'down' | 'executed' | 'pending') is run. Note: this will always run, even if the command throws an error.

The FileLocker class uses beforeAll and afterAll to implement a simple filesystem-based locking mechanism.

All events are type-safe, so IDEs will prevent typos and supply strong types for the event payloads.

Errors

When a migration throws an error, it will be wrapped in a MigrationError which captures the migration metadata (name, path etc.) as well as the original error message, and will be rethrown. In most cases, this is expected behaviour, and doesn't require any special handling beyond standard error logging setups.

If you expect failures and want to try to recover from them, you will need to try-catch the call to umzug.up(). You can access the original error from the .cause property if necessary:

try {
  await umzug.up()
} catch (e) {
  if (e instanceof MigrationError) {
    const original = e.cause
    // do something with the original error here
  }
  throw e
}

Under the hood, verror is used to wrap errors.

CLI

🚧🚧🚧 The CLI is new to Umzug v3. Feedback on it is welcome in discussions 🚧🚧🚧

Umzug instances provide a .runAsCLI() method. When called, this method will automatically cause your program to become a complete CLI, with help text and such:

// migrator.js
const {Umzug} = require('umzug')

const umzug = new Umzug({
  /* ... */
})

exports.umzug = umzug

if (require.main === module) {
  umzug.runAsCLI()
}

Note that this uses the @rushstack/ts-command-line package, which shows only the top-level message of any errors throw by default. See here for how you can see a full stack trace.

CLI Usage

A script like the one above is now a runnable CLI program. You can run node migrator.js --help to see how to use it. It will print something like:

usage: