@bradleyayers/node-pg-migrate

pg-migrate

Node.js database migration management built exclusively for postgres.

Installation

$ npm install node-pg-migrate

Installing this module adds a runnable file into your node_modules/.bin directory. If installed globally (with the -g option), you can run pg-migrate and if not, you can run ./node_modules/.bin/pg-migrate

Usage

You can specify your database connection information using config.

// config/default.json
{
  "db": "postgres://postgres:password@localhost:5432/name"
}

or

// config/default.json
{
  "db": {
    "user": "postgres",
    "password": "",
    "host": "localhost",
    "port": 5432,
    "name": "name"
  }
}

You could also specify your database url by setting the environment variable DATABASE_URL.

DATABASE_URL=postgres://postgres@localhost/name node-pg-migrate

You can specify custom JSON file with config (format is same as for db entry of config file)

If a .env file exists, it will be loaded using dotenv (if installed) when running the pg-migrate binary.

Depending on your project's setup, it may make sense to write some custom grunt tasks that set this env var and run your migration commands. More on that below.

The following are the available commands:

• pg-migrate create {migration-name} - creates a new migration file with the name you give it. Spaces and underscores will be replaced by dashes and a timestamp is prepended to your file name.
• pg-migrate up - runs all up migrations from the current state.
• pg-migrate up {N} - runs N up migrations from the current state.
• pg-migrate down - runs a single down migration.
• pg-migrate down {N} - runs N down migrations from the current state.
• pg-migrate unlock - unlocks migrations (if previous up/down migration failed and was not automatically unlocked).

Configuration

You can adjust defaults by passing arguments to pg-migrate:

• config-file (f) - The file with migration JSON config (defaults to undefined)

• schema (s) - The schema on which migration will be run (defaults to public)

• database-url-var (d) - Name of env variable with database url string (defaults to DATABASE_URL)

• migrations-dir (m) - The directory containing your migration files (defaults to migrations)

• migrations-schema - The schema storing table which migrations have been run (defaults to same value as schema)

• migrations-table (t) - The table storing which migrations have been run (defaults to pgmigrations)

• check-order - Check order of migrations before running them (defaults to true, to switch it off supply --no-check-order on command line). (There should be no migration with timestamp lesser than last run migration.)

See all by running pg-migrate --help.

Most of configuration options can be also specified in config file.

For SSL connection to DB you can set PGSSLMODE environment variable to value from list other then disable. e.g. PGSSLMODE=require pg-migrate up (pg will take it into account)

JSON Configuration

You can use config or your own json file with configuration (config-file command line option).

Available options are:

• migrations-dir, migrations-schema, migrations-table, check-order - same as above

• either url or [user], [password], host (defaults to localhost), port (defaults to 5432), name - for connection details

• type-shorthands - for column type shorthands

  You can specify custom types which will be expanded to column definition (e.g. for module.exports = { "type-shorthands": { id: { type: 'uuid', primaryKey: true }, createdAt: { type: 'timestamp', notNull: true, default: new require('node-pg-migrate').PgLiteral('current_timestamp') } } } it will in pgm.createTable('test', { id: 'id', createdAt: 'createdAt' }); produce SQL CREATE TABLE "test" ("id" uuid PRIMARY KEY, "createdAt" timestamp DEFAULT current_timestamp NOT NULL);).

Locking

pg-migrate automatically checks if no other migration is running. To do so, it locks the migration table and enters comment there. There are other options how to do it, but I choose this one (see #88). In some circumstances it is possible that lock will not be released (Error message - Error: Unable to fetch migrations: Error: Another migration is already running). In that case you need to run pg-migrate unlock to release the lock again.

Defining Migrations

When you run pg-migrate create a new migration file is created that looks like this:

exports.up = function up(pgm) {

}
exports.down = function down(pgm) {

}

pgm is a helper object that provides migration operations and run is the callback to call when you are done.

IMPORTANT Calling the migration functions on pgm doesn't actually migrate your database. These functions just add sql commands to a stack that is run.

Automatic Down Migrations

If exports.down is not present in a migration, pg-migrate will try to automatically infer the operations that make up the down migration by reversing the operations of the up migration. Only some operations have automatically inferrable equivalents (details below on each operation). Sometimes, migrations are destructive and cannot be rolled back. In this case, you can set exports.down = false to tell pg-migrate that the down migration is impossible.

Async Migrations

In some cases, you may want to perform some async operation during a migration, for example fetching some information from an external server, or inserting some data into the database. To make a migration block operate in async mode, just add another callback argument to the function signature. However, be aware that NONE of the pgm operations will be executed until run() is called. Here's an example:

exports.up = function up(pgm, run) {
  doSomethingAsync(function() {
    run();
  });
}

Another way how to perform some async operation is to return Promise from up or down function. Example:

exports.up = function(pgm) {
  return new Promise(resolve => {
    // doSomethingAsync
    resolve();
  });
}

Migration methods

The pgm object that is passed to each up/down block has many different operations available. Each operation is simply a function that generates some sql and stores it in the current pgm context.

By default, each migration will be run in a transaction. To disable transactions for a specific migration, call pgm.noTransaction() This is required for some SQL operations that cannot be run within a transaction. It should be used carefully.

Creating & Altering Tables / Columns

pgm.createTable( tablename, columns, options )

Create a new table - postgres docs

Arguments:

• tablename [string] - name for the new table
• columns [object] - column names / options -- see column definitions section
• options [object] - table options (optional)
  • inherits [string] - table to inherit from

Reverse Operation: dropTable

---

pgm.dropTable( tablename )

Drop existing table - postgres docs

Arguments:

• tablename [string] - name of the table to drop

---

pgm.renameTable( tablename, new_tablename )

Rename a table - postgres docs

Arguments:

• tablename [string] - name of the table to rename
• new_table [object] - new name of the table

Reverse Operation: same operation in opposite direction

---

pgm.addColumns( tablename, new_columns )

Add columns to an existing table - postgres docs

Arguments:

• tablename [string] - name of the table to alter
• new_columns [object] - column names / options -- see column definitions section

Aliases: addColumn Reverse Operation: dropColumns

---

pgm.dropColumns( tablename, columns )

Drop columns from a table - postgres docs

Arguments:

• tablename [string] - name of the table to alter
• columns [array of strings or object] - columns to drop (if object, uses keys)

Aliases: dropColumn

---

pgm.renameColumn( tablename, old_column_name, new_column_name )

Rename a column - postgres docs

Arguments:

• tablename [string] - name of the table to alter
• old_column_name [string] - current column name
• new_column_name [string] - new column name

Reverse Operation: same operation in opposite direction

---

pgm.alterColumn( tablename, column_name, column_options )

Alter a column (default value, type, allow null) - postgres docs

Arguments:

• tablename [string] - name of the table to alter
• column_name [string] - column to alter
• column_options [object] - optional new column options
  • default [string or null] - null, string
  • type [string] - new datatype
  • notNull [boolean] - sets NOT NULL if true

---

pgm.addConstraint( tablename, constraint_name, expression )

Add a named column constraint - postgres docs

Arguments:

• tablename [string] - name of the table to alter
• constraint_name [string] - name for the constraint
• expression [string] - constraint expression (raw sql)

Aliases: createConstraint Reverse Operation: dropConstraint

---

pgm.dropConstraint( tablename, constraint_name )

Drop a named column constraint - postgres docs

Arguments:

• tablename [string] - name of the table to alter
• constraint_name [string] - name for the constraint

---

pgm.createIndex( tablename, columns, options )

Create a new index - postgres docs

Arguments:

• tablename [string] - name of the table to alter
• columns [string or array of strings] - columns to add to the index
• options [index options] - optional options:
  • name [string] - name for the index (one will be inferred from table/columns if undefined)
  • unique [boolean] - set to true if this is a unique index
  • where [string] - raw sql for where clause of index
  • concurrently [boolean] - create this index concurrently
  • method [string] - btree | hash | gist | spgist | gin

Aliases: addIndex Reverse Operation: dropIndex

---

pgm.dropIndex( tablename, columns, options )

Drop an index - postgres docs

Arguments:

• tablename [string] - name of the table to alter
• columns [string or array of strings] - column names, used only to infer an index name
• options [index options] - optional options:
  • name [string] - name of the index to drop

---

Miscellaneous Operations

pgm.sql( sql )

Run raw sql -- with some optional very basic mustache templating

Arguments:

• sql [string] - SQL query to run
• args [object] - (optional) key/val of arguments to replace

---

pgm.func( sql )

Inserts raw string, which is not escaped

e.g. pgm.func('CURRENT_TIMESTAMP') to use in default option for column definition

Arguments:

• sql [string] - string to not be escaped

---

pgm.createExtension( extension )

Install postgres extension(s) - postgres docs

Arguments:

• extension [string or array of strings] - name(s) of extensions to install

Aliases: addExtension Reverse Operation: dropExtension

---

pgm.dropExtension( extension )

Un-install postgres extension(s) - postgres docs

Arguments:

• extension [string or array of strings] - name(s) of extensions to install

---

pgm.createType( type_name, values )

Create a new data type - postgres docs

Arguments:

• type_name [string] - name of the new type
• values [array of strings or object] if an array the contents are possible values for an enum type, if an object names and types for a composite type

Aliases: addType Reverse Operation: dropType

---

pgm.dropType( type_name )

Drop a custom data type - postgres docs

Arguments:

• type_name [string] - name of the new type

---

Column Definitions

The createTable and addColumns methods both take a columns argument that specifies column names and options. It is a object (key/value) where each key is the name of the column, and the value is another object that defines the options for the column.

• type [string] - data type (use normal postgres types)
• unique [boolean] - set to true to add a unique constraint on this column
• primaryKey [boolean] - set to true to make this column the primary key
• notNull [boolean] - set to true to make this column not null
• check [string] - sql for a check constraint for this column
• references [string] - a table name that this column is a foreign key to
• onDelete [string] - adds ON DELETE constraint for a reference column
• onUpdate [string] - adds ON UPDATE constraint for a reference column

Data types & Convenience Shorthand

Data type strings will be passed through directly to postgres, so write types as you would if you were writing the queries by hand.

There are some aliases on types to make things more foolproof: (int, string, float, double, datetime, bool)

There is a shorthand to pass only the type instead of an options object: pgm.addColumns('myTable', { age: 'integer' }); is equivalent to pgm.addColumns('myTable', { age: { type: 'integer' } });

There is a shorthand for normal auto-increment IDs: pgm.addColumns('myTable', { id: 'id' }); is equivalent to pgm.addColumns('myTable', { id: { type: 'serial', primaryKey: true } });

Using schemas

Instead of passing string as name to pgm functions, you can pass an object with keys schema and name. E.g.

pgm.createTable( {schema: 'my_schema', name: 'my_table_name'}, {id: 'serial'});

will generate

CREATE TABLE "my_schema"."my_table_name" (
 "id" serial
);

Explanation & Goals

Why only Postgres? - By writing this migration tool specifically for postgres instead of accommadating many databases, we can actually provide a full featured tool that is much simpler to use and maintain. I was tired of using crippled database tools just in case one day we switch our database.

Async / Sync - Everything is async in node, and that's great, but a migration tool should really just be a fancy wrapper that generates SQL. Most other migration tools force you to bring in control flow libraries or wrap everything in callbacks as soon as you want to do more than a single operation in a migration. Plus by building up a stack of operations, we can automatically infer down migrations (sometimes) to save even more time.

Naming / Raw Sql - Many tools force you to use their constants to do things like specify data types. Again, this tool should be a fancy wrapper that generates SQL, so whenever possible, it should just pass through user values directly to the SQL. The hard part is remembering the syntax of the specific operation, not remembering how to type "timestamp"!