Postgrator

A Node.js SQL migration library using a directory of plain SQL scripts. Supports Postgres, MySQL, and SQL Server.

Available as a cli tool: https://www.npmjs.com/package/postgrator-cli.

Installation

As of postgrator 4 Node.js version 10 or greater is required

npm install postgrator
# install necessary db engine(s) if not installed yet
npm install pg@8
npm install mysql@2
npm install mysql2@2
npm install mssql@6

Supported DB Drivers

Using a package version other than the below may not work.

pg 6.x.x
pg 7.x.x
pg 8.x.x
mysql 2.x.x
mysql2 1.x.x
mysql2 2.x.x
mssql 4.x.x
mssql 5.x.x
mssql 6.x.x

Usage

Create a directory and stick some SQL scripts in there that change your database in some way. It might look like:

migrations/
  |- 001.do.sql
  |- 001.undo.sql
  |- 002.do.optional-description-of-script.sql
  |- 002.undo.optional-description-of-script.sql
  |- 003.do.sql
  |- 003.undo.sql
  |- 004.do.js
  |- 004.undo.js
  |- ... and so on

The files must follow the convention [version].[action].[optional-description].sql or [version].[action].[optional-description].js

Version must be a number, but you may start and increment the numbers in any way you'd like. If you choose to use a purely sequential numbering scheme instead of something based off a timestamp, you will find it helpful to start with 000s or some large number for file organization purposes.

Action must be either "do" or "undo". Do implements the version, and undo undoes it.

Optional-description can be a label or tag to help keep track of what happens inside the script. Descriptions should not contain periods.

SQL or JS You have a choice of either using a plain SQL file or you can also generate your SQL via a javascript module. The javascript module should export a function called generateSql() that returns back a string representing the SQL. For example:

module.exports.generateSql = function () {
  return (
    "CREATE USER transaction_user WITH PASSWORD '" +
    process.env.TRANSACTION_USER_PASSWORD +
    "'"
  )
}

You might want to choose the JS file approach, in order to make use (secret) environment variables such as the above.

Support for asynchronous functions is provided, in the event you need to retrieve data from a external source, for example:

const axios = require('axios')

module.exports.generateSql = async () => {
  const response = await axios({
    method: 'get',
    url: 'https://api.example.org/person/1',
  })
  return `INSERT INTO person (name, age) VALUES ('${response.data.name}', ${response.data.age});`
}

To run your sql migrations with Postgrator, write a Node.js script or integrate postgrator with your application.

When first run against your database, Postgrator will create the table specified by config.schemaTable. Postgrator relies on this table to track what version the database is at.

Postgrator automatically determines whether it needs to go "up" or "down", and will update the schemaTable accordingly. If the database is already at the version specified to migrate to, Postgrator does nothing. After running migrations, postgrator will close its connection created.

const Postgrator = require('postgrator')

const postgrator = new Postgrator({
  // Directory containing migration files
  migrationDirectory: __dirname + '/migrations',
  // or a glob pattern to files
  migrationPattern: __dirname + '/some/pattern/*',
  // Driver: must be pg, mysql, mysql2 or mssql
  driver: 'pg',
  // Database connection config
  host: '127.0.0.1',
  port: 5432,
  database: 'databasename',
  username: 'username',
  password: 'password',
  // Schema table name. Optional. Default is schemaversion
  // If using Postgres, schema may be specified using . separator
  // For example, { schemaTable: 'schema_name.table_name' }
  schemaTable: 'schemaversion',
})

// Migrate to specific version
postgrator
  .migrate('002')
  .then((appliedMigrations) => console.log(appliedMigrations))
  .catch((error) => {
    console.log(error)
    // Because migrations prior to the migration with error would have run
    // error object is decorated with appliedMigrations
    console.log(error.appliedMigrations) // array of migration objects
  })

// Migrate to max version (optionally can provide 'max')
postgrator
  .migrate()
  .then((appliedMigrations) => console.log(appliedMigrations))
  .catch((error) => console.log(error))

Postgres options:

Postgres supports connection string url as well as simple ssl config:

const postgrator = new Postgrator({
  connectionString: 'tcp://username:password@hosturl/databasename',
  ssl: true, // passed directly to node.TLSSocket, supports all tls.connect options
  currentSchema: 'my-schema-name', // migrations will only run against this schema
})

MySQL options:

MySQL also supports ssl config option

const postgrator = new Postgrator({
  // common options plus
  // ssl should be object as mysql expects (https://www.npmjs.com/package/mysql#ssl-options)
  ssl,
})

SQL Server options:

For SQL Server, you may optionally provide an additional options configuration. This may be necessary if requiring a secure connection for Azure.

const postgrator = new Postgrator({
  requestTimeout: 1000 * 60 * 60, // Default 1 hour
  connectionTimeout: 30000, // override mssql 15 second default
  options: {
    encrypt: true, // for azure
    trustServerCertificate: false, // change to true for local dev / self-signed certs
  },
})

Reference options for mssql for more details: https://www.npmjs.com/package/mssql

Checksum validation

By default Postgrator will generate an md5 checksum for each migration file, and save the value to the schema table after a successful migration.

Prior to applying migrations to a database, for any existing migration in the migration directory already run Postgrator will validate the md5 checksum to ensure the contents of the script have not changed. If a change is detected, migration will stop, reporting an error.

Because line endings may differ between environments/editors, an option is available to force a specific line ending prior to generating the checksum.

const postgrator = new Postgrator({
  validateChecksums: true, // Set to false to skip validation
  newline: 'CRLF', // Force using 'CRLF' (windows) or 'LF' (unix/mac)
})

Migration object

Postgrator will often return a migration object or array of migrations. The format of a migration object is

{
  version: versionNumber,
  action: 'do',
  name: 'first-table',
  filename: '0001.up.first-table.sql',
  md5: 'checksumvalue',
  getSql: () => {} // sync function to get sql from file
}

Logging

As of v3 nothing is logged to the console, and the option to toggle that has been removed. Instead postgrator is an event emiter, allowing you to log however you want to log. There are no events for error or finish

const postgrator = new Postgrator(options)
postgrator.on('validation-started', (migration) => console.log(migration))
postgrator.on('validation-finished', (migration) => console.log(migration))
postgrator.on('migration-started', (migration) => console.log(migration))
postgrator.on('migration-finished', (migration) => console.log(migration))

Migration errors

If postgrator.migrate() fails running multiple migrations, Postgrator will stop running any further migrations. Migrations successfully run prior to the migration with the error will remain implemented however.

If you need to migration back down to the version the database was at prior to running migrate(), that is up to you to implement. Instead of doing this however, consider writing your application in a way that is compatible with any version of a future release.

In the event of an error during migration, the error object will be decorated with an array of migrations that run successfully (error.appliedMIgrations).

Keep in mind how you write your SQL - You may (or may not) want to write your SQL defensively (ie, check for pre-existing objects before you create new ones).

Preventing partial migrations

Depending on your database and database configuration, consider wrapping each migration in a transaction or BEGIN/END block. By default Postgres and SQL Server consider multiple statements run in one execution part of one implicit transaction. MySQL however will implement up to the failure.

If using SQL Server, do not write a migration containing multiple statements using the GO keyword. Instead break statements between the GO keyword into multiple migration files, ensuring that you do not end up with partial migrations implemented but no record of that happening.

Utility methods

Some of postgrator's methods may come in useful performing other migration tasks

// To get max version available from filesystem
// version returned as number, not string
postgrator
  .getMaxVersion()
  .then((version) => console.log(version))
  .catch((error) => console.error(error))

// "current" database schema version
// version returned as number, not string
postgrator
  .getDatabaseVersion()
  .then((version) => console.log(version))
  .catch((error) => console.error(error))

// To get all migrations from directory and parse metadata
postgrator
  .getMigrations()
  .then((migrations) => console.log(migrations))
  .catch((error) => console.error(error))

// Run arbitrary SQL query against database
// Connection is established, query is run, then connection is ended
// `results.rows` will be an array of row objects, with column names as keys
// `results` object may have other properties depending on db driver
postgrator
  .runQuery('SELECT * FROM sometable')
  .then((results) => console.log(results))
  .catch((error) => console.error(error))

Tests

A docker-compose file is provided with postgres and mysql (mariadb) containers configured for the tests. SQL Server tests also exist, but are commented out since the requirements are quite high to run them.

To run postgrator tests locally, you'll need Docker installed.

# In one terminal window
docker-compose up
# In another terminal once databases are up
npm test
# After tests, in docker session
# control/command-c to quit docker-compose and remove containers
docker-compose rm --force

License

MIT

terbooter/postgrator