jonathannewman and yogthos Update README and CHANGES for 1.0.8 release. (#142)
Updates changelog from 1.0.5 to 1.0.8, updates docstring for
core migration function, and adds to README for the same function.
Latest commit dd6a9aa Jul 17, 2018





A general migration framework, with implementations for migrations as SQL scripts or general Clojure code.

Designed to be compatible with a git based work flow where multiple topic branches may exist simultaneously, and be merged into a master branch in unpredictable order.

This is accomplished two ways:

  1. Migration ids are not assumed to be incremented integers. It is recommended that they be timestamps (e.g. ‘20111202091200’).
  2. Migrations are considered for completion independently.

Using a 14 digit timestamp will accommodate migrations granularity to a second, reducing the chance of collisions for a distributed team.

In contrast, using a single global version for a store and incremented integers for migration versions, it is possible for a higher numbered migration to get merged to master and deployed before a lower numbered migration, in which case the lower numbered migration would never get run, unless it is renumbered.

Migratus does not use a single global version for a store. It considers each migration independently, and runs all uncompleted migrations in sorted order.

Quick Start

  • add the Migratus dependency:


  • Add the following code to resources/migrations/20111206154000-create-foo-table.up.sql
  • Add the following code to resources/migrations/20111206154000-create-foo-table.down.sql

Multiple Statements

If you would like to run multiple statements in your migration, then separate them with --;;. For example:

CREATE TABLE IF NOT EXISTS quux(id bigint, name varchar(255));
CREATE INDEX quux_name on quux(name);

This is necessary because JDBC does not have a method that allows you to send multiple SQL commands for execution. Migratus will split your commands, and attempt to execute them inside of a transaction.

Note that some databases, such as MySQL, do not support transactional DDL commands. If you’re working with such a database then it will not be able to rollback all the DDL statements that were applied in case a statement fails.

Disabling transactions

Migratus attempts to run migrations within a transaction by default. However, some databases do not support transactional DDL statements. Transactions can be disabled by adding the following line at the start of the migration file:

-- :disable-transaction

Running Functions in Migrations

Functions inside migrations may need to be additionally wrapped, a PostgreSQL example would look as follows:

DO $func$
  PERFORM schema_name.function_name('foo', 10);

Supporting “use” statements

To run migrations against several different databases (in MySQL, or “schemas” in Postgres, etc.), with embedded “use” statements in your migrations, specify the database in your migration-table-name in the connections, i.e. “database_name.table_name” not “table_name”


  • Add Migratus as a dependency to your project.clj
    :dependencies [[migratus <VERSION>]]

There are hidden dependencies on slf4j inside migratus, so to avoid errors or silent failures you’ll need to also add

[org.slf4j/slf4j-log4j12 <VERSION>]

or if you’re using Timbre

[com.fzakaria/slf4j-timbre <VERSION>]

Next, create a namespace to manage the migrations:

(ns my-migrations
 (:require [migratus.core :as migratus]))
(def config {:store                :database
             :migration-dir        "migrations/"
             :init-script          "init.sql"
             ;defaults to true, some databases do not support
             ;schema initialization in a transaction
             :init-in-transaction? false
             :migration-table-name "foo_bar"
             :db {:classname   "org.h2.Driver"
                  :subprotocol "h2"
                  :subname     "site.db"}})
;initialize the database using the 'init.sql' script
(migratus/init config)
;apply pending migrations
(migratus/migrate config)
;rollback the last migration applied
(migratus/rollback config)
;bring up migrations matching the ids
(migratus/up config 20111206154000)
;bring down migrations matching the ids
(migratus/down config 20111206154000)

Generate migration files

Migratus also provides a convenience function for creating migration files:

(migratus/create config "create-user")

This will result with up/down migration files being created prefixed with the current timestamp, e.g:


Code-based Migrations

Application developers often encounter situations where migrations cannot be easily expressed as a SQL script. For instance:

  • Executing programmatically-generated DDL statements (e.g. updating the schema of a dynamically-sharded table).
  • Transferring data between database servers.
  • Backfilling existing records with information that must be retrieved from an external system.

A common approach in these scenarios is to write one-off scripts which an admin must manually apply for each instance of the application, but issues arise if a script is not run or run multiple times.

Migratus addresses this problem by providing support for code-based migrations. You can write a migration as a Clojure function, and Migratus will ensure that it’s run exactly once for each instance of the application.

Defining a code-based migration

Create a code-based migration by adding a .edn file to your migrations directory that contains the namespace and up/down functions to run, e.g. resources/migrations/20170331141500-import-users.edn:

{:ns app.migrations.import-users
 :up-fn migrate-up
 :down-fn migrate-down}

Then, in src/app/migrations/import_users.clj:

(ns app.migrations.import-users)


(defn migrate-up [config]
  ;; do stuff here


(defn migrate-down [config]
  ;; maybe undo stuff here
  • The up and down migration functions should both accept a single parameter, which is the config map passed to Migratus (so your migrations can be configurable).
  • You can omit the up or down migration by setting :up-fn or down-fn to nil in the EDN file.

Generate code-based migration files

The migratus.core/create function accepts an optional type parameter, which you can pass as :edn to create a new migration file.

(migratus/create config "import-users" :edn)

Mixing SQL and code-based migrations

You can include both SQL and code-based migrations in the same migrations directory, in which case they will be run intermixed in the order defined by their timestamps and their status stored in the same table in the migrations database. This way if there are dependencies between your SQL and code-based migrations, you can be assured that they’ll run in the correct order.

Quick Start (Leiningen 2.x)

Migratus provides a Leiningen plugin:

  • Add migratus-lein as a plugin in addition to the Migratus dependency:


  • Add the following key and value to your project.clj:
    :migratus {:store :database
               :migration-dir "migrations"
               :db {:classname "com.mysql.jdbc.Driver"
                    :subprotocol "mysql"
                    :subname "//localhost/migratus"
                    :user "root"
                    :password ""}}

To apply pending migrations:

  • Run lein migratus migrate

To rollback the last migration that was applied run:

  • Run lein migratus rollback

Then follow the rest of the above instructions.


Migratus is configured via a configuration map that you pass in as its first parameter. The :store key describes the type of store against which migrations should be run. All other keys/values in the configuration map are store specific.


To run migrations against a database use a :store of :database, and specify the database connection configuration in the :db key of the configuration map. This connection information is passed directly to clojure.java.jdbc. For example:

{:store :database
 :migration-dir "migrations"
 :db {:classname "com.mysql.jdbc.Driver"
      :subprotocol "mysql"
      :subname "//localhost/migratus"
      :user "root"
      :password ""}}


{:store :database
 :migration-dir "migrations"
 :db ~(get (System/getenv) "DATABASE_URL")}

The :migration-dir key specifies the directory on the classpath in which to find SQL migration files. Each file should be named with the following pattern “[id]-[name].[direction].sql” where id is a unique integer id (ideally it should be a timestamp) for the migration, name is some human readable description of the migration, and direction is either ‘up’ or ‘down’.

If Migratus is trying to run either the up or down migration and it does not exist, then an Exception will be thrown.

See test/migrations in this repository for an example of how database migrations work.

Modify sql fn

If you want to do some processing of the sql before it gets executed, you can provide a `:modify-sql-fn` in the config data structure to do so. This is intended for use with http://2ndquadrant.com/en/resources/pglogical/ and similar systems, where DDL statements need to be executed via an extension-provided function.


Migratus can be used programmatically by calling one of the following functions:

migratus.core/initRuns a script to nitialize the database, e.g: create a new schema.
migratus.core/createCreate a new migration with the current date.
migratus.core/migrateRun ‘up’ for any migrations that have not been run. Returns nil if successful, :ignore if the table is reserved. Supports thread cancellation.
migratus.core/rollbackRun ‘down’ for the last migration that was run.
migratus.core/upRun ‘up’ for the specified migration ids. Will skip any migration that is already up.
migratus.core/downRun ‘down’ for the specified migration ids. Will skip any migration that is already down.
migratus.core/pending-listReturns a list of pending migrations.
migratus.core/migrate-until-just-beforeRun ‘up’ for for any pending migrations which precede the given migration id (good for testing migrations).

See the docstrings of each function for more details.

Migratus can also be used from leiningen if you add it as a plugin dependency.

:plugins [[migratus-lein <VERSION>]]

And add a configuration :migratus key to your project.clj.

:migratus {:store :database
           :migration-dir "migrations"
           :db {:classname "com.mysql.jdbc.Driver"
                :subprotocol "mysql"
                :subname "//localhost/migratus"
                :user "root"
                :password ""}}

You can then run the following tasks:

lein migratus create <name>Create a new migration with the current date.
lein migratus migrateRun ‘up’ for any migrations that have not been run.
lein migratus rollbackRun ‘down’ for the last migration that was run.
lein migratus up & idsRun ‘up’ for the specified migration ids. Will skip any migration that is already up.
lein migratus down & idsRun ‘down’ for the specified migration ids. Will skip any migration that is already down.
lein migratus resetRun ‘down’ for all migrations that have been run, and ‘up’ for all migrations.
lein migratus pendingRun ‘pending-list’ to get all pending migrations.


Copyright © 2016 Paul Stadig, Dmitri Sotnikov


Licensed under the Apache License, Version 2.0.