skip to navigation
skip to content

yoyo-migrations 4.2.4

Database schema migration tool using SQL and DB-API

Latest Version: 5.0.5

Yoyo is a database schema migration tool using plain SQL and python’s builtin DB-API.

What does yoyo-migrations do?

As database applications evolve, changes to the database schema are often required. These can usually be written as one-off SQL scripts containing CREATE/ALTER table statements (although any SQL or python script may be used with yoyo).

Yoyo provides a command line tool for reading a directory of such scripts and applying them to your database as required.


Install from the PyPI with the command:

pip install yoyo-migrations

Database support

PostgreSQL, MySQL and SQLite databases are supported.


Yoyo is usually invoked as a command line script.


Read all migrations from directory migrations and apply them to a PostgreSQL database:

yoyo-migrate apply ./migrations/ postgres://user:password@localhost/database

Rollback migrations previously applied to a MySQL database:

yoyo-migrate rollback ./migrations/ mysql://user:password@localhost/database

Reapply (ie rollback then apply again) migrations to a SQLite database at location /home/sheila/important-data.db:

yoyo-migrate reapply ./migrations/ sqlite:////home/sheila/important-data.db

By default, yoyo-migrations starts in an interactive mode, prompting you for each migration file before applying it, making it easy to choose which migrations to apply and rollback.

The migrations directory should contain a series of migration scripts. Each migration script is a python file (.py) containing a series of steps. Each step should comprise a migration query and (optionally) a rollback query. For example:

# file: migrations/
from yoyo import step
    "CREATE TABLE foo (id INT, bar VARCHAR(20), PRIMARY KEY (id))",
    "DROP TABLE foo",

The filename of each file (without the .py extension) is used as the identifier for each migration. Migrations are applied in filename order, so it’s useful to name your files using a date (eg ‘’) or some other incrementing number.

yoyo-migrate creates a table in your target database, _yoyo_migration, to track which migrations have been applied.

Steps may also take an optional argument ignore_errors, which must be one of apply, rollback, or all. If in the previous example the table foo might have already been created by another means, we could add ignore_errors='apply' to the step to allow the migrations to continue regardless:

# file: migrations/
from yoyo import step
    "CREATE TABLE foo (id INT, bar VARCHAR(20), PRIMARY KEY (id))",
    "DROP TABLE foo",

Steps can also be python callable objects that take a database connection as their single argument. For example:

# file: migrations/
from yoyo import step
def do_step(conn):
    cursor = conn.cursor()
        "INSERT INTO sysinfo "
        " (osname, hostname, release, version, arch)"
        " VALUES (%s, %s, %s, %s, %s %s)",



By default each step is run in its own transaction. You can run multiple steps within a single transaction by wrapping them in a transaction call, like so:

# file: migrations/
from yoyo import step, transaction
    "CREATE TABLE foo (id INT, bar VARCHAR(20), PRIMARY KEY (id))",
    "DROP TABLE foo",
  step("INSERT INTO foo (1, 'baz')"),

If this is the case setting ignore_errors on individual steps makes no sense: database errors will always cause the entire transaction to be rolled back. The outer transaction can however have ignore_errors set.

Post-apply hook

It can be useful to have a script that’s run after successful migrations. For example you could use this to update database permissions or re-create views. To do this, create a migration file called This file should have the same format as any other migration file.

Password security

You normally specify your database username and password as part of the database connection string on the command line. On a multi-user machine, other users could view your database password in the process list.

The -p or --prompt-password flag causes yoyo-migrate to prompt for a password, ignoring any password specified in the connection string. This password will not be available to other users via the system’s process list.

Connection string caching

The first time you run yoyo-migrate on a new set of migrations, you will be asked if you want to cache the database connection string in a file called .yoyo-migrate in the migrations directory.

This cache is local to the migrations directory, so subsequent runs on the same migration set do not need the database connection string to be specified.

This saves typing, avoids your database username and password showing in process listings and lessens the risk of accidentally running yoyo-migrate on the wrong database (ie by re-running an earlier yoyo-migrate entry in your command history when you have moved to a different directory).

If you do not want this cache file to be used, add the --no-cache parameter to the command line options.

Using yoyo from python code

The following example shows how to apply migrations from inside python code:

from yoyo import read_migrations
from yoyo.connections import connect

conn, paramstyle = connect('postgres://myuser@localhost/mydatabase')
migrations = read_migrations(conn, paramstyle, 'path/to/migrations'))


Version 4.2.4

  • Fix for mismanaged 4.2.3 release

Version 4.2.3

  • Migrations are now datestamped with a UTC date (thanks to robi wan)
  • Fixes for installation and use under python 3

Version 4.2.2

  • Migration scripts can start with from yoyo import step, transaction. This prevents linters (eg flake8) throwing errors over undefined names.
  • Bugfix: functions declared in a migration file can access the script’s global namespace

Version 4.2.1

  • Bugfix for previous release, which omitted critical files

Version 4.2.0

  • Removed yoyo.migrate namespace package. Any code that uses the yoyo api directly needs have any imports modified, eg this:

    from yoyo.migrate import read_migrations
    from yoyo.migrate.connections import connect

    Should be changed to this:

    from yoyo import read_migrations
    from yoyo.connections import connect
  • Migrated from darcs to mercurial. Code is now hosted at

  • Bugfix: the migration_table option was not being passed to read_migrations, causing the value to be ignored

Version 4.1.6

  • Added windows support (thanks to Peter Shinners)

Version 4.1.5

  • Configure logging handlers so that the -v switch causes output to go to the console (thanks to Andrew Nelis).
  • -v command line switch no longer takes an argument but may be specified multiple times instead (ie use -vvv instead of -v3). --verbosity retains the old behaviour.

Version 4.1.4

  • Bugfix for post apply hooks

Version 4.1.3

  • Changed default migration table name back to ‘_yoyo_migration’

Version 4.1.2

  • Bugfix for error when running in interactive mode

Version 4.1.1

  • Introduced configuration option for migration table name

Version 4.1.0

  • Introduced ability to run steps within a transaction (thanks to Ryan Williams for suggesting this functionality along with assorted bug fixes.)
  • “post-apply” migrations can be run after every successful upward migration
  • Other minor bugfixes and improvements
  • Switched to <major>.<minor> version numbering convention

Version 4

  • Fixed problem installing due to missing manifest entry

Version 3

  • Use the console_scripts entry_point in preference to scripts=[] in, this provides better interoperability with buildout

Version 2

  • Fixed error when reading dburi from config file

Version 1

  • Initial release
File Type Py Version Uploaded on Size
yoyo-migrations-4.2.4.tar.gz (md5) Source 2014-02-02 16KB