Monarch

MigratiON manAgeR with a bad aCronym, duH

Over the last several projects I've worked, one of the main problems I've encountered is managing databases throughout different environments. There are several migration managers out there, but all of them are overly bloated as most of them rely on ORMs, and were an overkill for what I really needed:

Track the state of the DB in different environments.
A simple way to transition from one DB state to another one.
A sane way to manage in what order the migrations should be applied (other than the usual file-naming by numbers).

Hence, Monarch was created.

Why should I use Monarch instead of <insert name of migration tool>?

DB engine and language independent: A lot of migration tools rely on you using a specific language, or a specific DB or even an ORM. Monarch is a single executable file that can work with any project independently of the language of the DB used and doesn't expect any dependency.
Work with plain SQL: When dealing with the state of a DB, the most natural way of handling it through plain SQL. This gives flexibility and detaches the DB handling from any framework (at least at a migration level).
Tracking of the DB state: When working on a project it's common to have multiple environments, and it can get really cumbersome to remember what the state of each of these environments is. Monarch solves that for you.
Use of environment variables: No more creating separate config files for every different environment. Monarch takes all its configuration from environment variables (and optionally a .env file).
No reliance on file-naming for dependencies: Usually, migration tools use file naming conventions to sort which migration should be applied first (usually numbered migrations i.e. 1_first_migrations.sql). This poses a problem when submitting the migrations to version control and make said migrations in a collaborative environment. The way Monarch works, dependent migrations are not handled as a single stream, but rather as branching succesions. This means that two developers can work simultaneously on changing different tables at the same time without stepping on each others way, provided said changes are not related to each other.
Migrations are stored independently from the target database itself: The state of the DB needs to be stored into a DB itself, but what if you have restricted access to the DB you need to make changes to? Monarch's approach gives more flexibility on this end, since you can store and even check a DB's state without actually accessing it (saving it as a sqlite file for instance).
Granular dependency control: Assume you have 3 migrations: 1_create_person.sql, 2_change_address.sql, 3_change_person.sql. If you wanted to only make changes to the Person table but preserve the state of Address, you'd have to cherry pick the needed migrations in <insert name of migration tool>, while in Monarch, you could only run the last Person-related migration and the program will apply only the strictly needed migration to make that work.

If any of this sounds appealing, maybe give Monarch a try.

Installation

Linux and MacOS

# Download the binary
$ sudo curl -fsSL -o /usr/local/bin/monarch https://github.com/aropie/monarch/releases/download/v1.0/monarch

# Make it executable
$ sudo chmod +x /usr/local/bin/monarch

Windows

Not available yet :(

Migrations

Migrations are single files that represent an atomic change to the DB's state.

Migrations are always additive. This means that once a migration is commited into Monarch, it shouldn't be modified, and changes can only be reverted through a new migration.
Migrations can't contain BEGIN TRANSACTION/COMMIT statements. Monarch takes care of this and all the migrations are run as a single transaction.
Dependencies must be declared at the header. If a migration is not self-contained, i.e. it assumes a pre-existing state of the DB, all dependencies must be declared at the header. More on this header later.

Migrations are looked for in the migrations dir, migrations by default. This can be changed with the -d flag.

Usage

# Run a single migration
$ ./monarch -m some-pending-migrations.sql

# Or run all available migrations
$ ./monarch -a

It's important to note that Monarch runs all the migrations requested (be it a single migration and its dependencies or a whole schema) as a single transaction. This means that if you are applying migrations 1 through 5 and no. 3 breaks the DB state, all changes are rolled back and none of the 5 migrations are applied. To override this behavior, run the migration with -t. With the -t flag, if migration 3 breaks the DB state, migrations 1 and 2 will already be applied and saved.

Monarch saves the current DB state through a migrations table, in order to know which migrations are needed to move from one state to another. You can make Monarch ignore previously applied migrations with -ignore-applied, and avoid new migrations being registered into the DB with -skip-register, or only register migrations without actually applying them with --fake.

Database connection

Monarch functions on two different databases: an internal one to save the applied migrations, and a target database, which is the one where the migrations will actually be applied.

One of the objectives of Monarch is to be database-agnostic. Internally, Monarch uses SQLAlchemy to handle the connections with the databases. This means that all the databases supported by SQLAlchemy are supported by Monarch as well, which covers all of the most common DB choices (MySQL, PostgreSQL, SQLite, Microsoft SQL Server).

To configure the database connections, a database url is used,which is defined through environment variables. Simply set INTERNAL_DB_URL and TARGET_DB_URL and you're ready to go:

# Both of these could point to the same db.
$ export INTERNAL_DB_URL='sqlite:///internal.db'
$ export TARGET_DB_URL='postgresql://postgres:@localhost'

To allow for different development environments, a .env can also be supplied to set the corresponding environment variables.

Header

The header is where Monarch figures out how to handle the migration. A valid Monarch header starts with --! and then the declaration which is json-formatted in the following manner: {command1: value, command2: value, ...}.

The existing commands are all that follow:

depends_on Expects a list of srings, where each element is a migration that the current one depends on.

And that's all.

Example migration

--! {"depends_on": ["start_up_db.sql", "delete_a_table.sql"]}

CREATE TABLE person(
   id INTEGER,
   name VARCHAR(80),
   PRIMARY KEY(id)
);

Dependencies

Monarch solves the declared dependencies by creating a dependency tree and follow it recursively. Suppose you have first_migration.sql and another migration second_migration.sql which depends on first_migration.sql. The header for second_migration.sql would look something like:

--! {"depends_on": ["first_migration.sql"]}

If later you have a third migration my_third_migration.sql which depends on second_migration.sql (and therefore also on first_migration.sql) its header would look like:

--! {"depends_on": ["second_migration.sql"]}

Note that we're not declaring first_migration.sql as a dependency here because second_migration.sql already has it listed, so Monarch will traverse the dependency tree and apply all necessary migrations when migrating third_migration.sql.

pecalleja/monarch