Myt - MySQL version control

Utilities to ease the maintenance of MySQL or MariaDB database versioning using a Git repository.

This project is just to bring an idea to life and is still in an early stage of development, so any help is welcomed! Feel free to contribute.

Requirements

• Git
• Docker (Optional, used to manage local server)

Installation

It's recommended to install the package globally.

# npm install -g @verdnatura/myt
$ myt <command>

You can also install locally and use the npx command to execute it.

$ npm install @verdnatura/myt
$ npx myt <command>

How to use

Execute myt with the desired command.

$ [npx] myt [-w|--workspace <string>] [-r|--remote <string>] [-d|--debug]
[-h|--help] <command> [<args>]

The default workspace directory is the current working directory and unless otherwise indicated, the default remote is local.

Database versioning commands:

• init: Initialize an empty workspace.
• pull: Incorporate database routine changes into workspace.
• push: Apply changes into database.
• version: Creates a new version.
• create: Creates a new routine file.
• clean: Cleans old versions.

Local server management commands:

• dump: Export database structure and fixtures.
• fixtures: Export local database fixtures.
• run: Build and start local database server container.
• start: Start local database server container.

Each command can have its own specific commandline options.

Basic information

First of all you have to initalize the workspace.

$ myt init

Now you can configure Myt using myt.config.yml file, located at the root of your workspace. This file should include the project codename and schemas/tables wich are exported when you use pull or dump commands.

Don't forget to initialize git (if it isn't initialized yet).

$ git init

Remotes

Create database connection configuration for each environment at remotes folder using standard MySQL ini configuration files. The convention remote names are local, production and test.

remotes/[remote].ini

Startup

Once the basic configuration is done, routines can be imported from the database into the project, it is recommended to use the production remote.

$ myt pull production

From now on, you can use the project as if it were a standard git repository (since it is). To apply changes to the database run the push command on the desired remote.

$ myt push [<remote>] [--commit]

Routines

Routines are placed inside routines folder. All objects that have PL/SQL code are considered routines. It includes events, functions, procedures, triggers and views with the following structure.

  routines
  `- schema
     |- events
     |  `- eventName.sql
     |- functions
     |  `- functionName.sql
     |- procedures
     |  `- procedureName.sql
     |- triggers
     |  `- triggerName.sql
     `- views
        `- viewName.sql

Versions

Versions are placed inside versions folder with the following structure. Don't place your PL/SQL objects here, use the routines folder!

  versions
  |- 00001-firstVersionCodeName
  |  |- 00-firstExecutedScript.sql
  |  |- 01-secondScript.sql
  |  `- 99-lastScript.sql
  `- 00002-secondVersion
     |- 00-firstExecutedScript.sql
     `- 00-sameNumbers.sql

Local server

The local server is created as a MariaDB Docker container using the base dump created with the dump command plus pushing local versions and changed routines.

Dumps

You can create your local fixture and structure files.

• dump/beforeDump.sql
• dump/afterDump.sql
• dump/fixtures.sql

Versioning commands

init

Initializes an empty workspace.

$ myt init

pull

Incorporates database routine changes into workspace.

$ myt pull [remote] [-f|--force] [-c|--checkout] [-u|--update] [-s|--sums]

When --checkout option is provided, it does the following before export:

1. Get the last database push commit (saved in versioning tables).
2. Creates and checkout to a new branch based in database commit.

push

Applies versions and routine changes into database.

$ myt push [<remote>] [-f|--force] [-c|--commit] [-s|--sums]

Commit is saved into database only if --commit option is provided, it prevents from accidentally saving local commits into shared servers, causing subsequent pushes from other clients to fail because they can't get that commit from the git tree in order to get differences.

version

Creates a new version folder, when name is not specified it generates a random name mixing a color with a plant name.

$ myt version [<name>]

create

Creates a new routine file with a default template.

$ myt create [-t <type>] <schema>.<name>

clean

Cleans all already applied versions older than maxOldVersions.

$ myt clean

Local server commands

dump

Exports database structure and fixtures from remote into hidden files located in dump folder. If no remote is specified production is used.

$ myt dump [<remote>]

fixtures

Exports local database fixtures into dump/fixtures.sql files. If no remote is specified local is used.

$ myt fixtures [<remote>]

run

Builds and starts local database server container. It only rebuilds the image when fixtures have been modified or when the day on which the image was built is different to today.

$ myt run [-c|--ci] [-r|--random]

start

Starts local database server container. It does the minium effort, if it doesn't exists calls the run command, if it is started does nothing. Keep in mind that when you do not rebuild the docker you may be using an outdated version of it.

$ myt start

Why

The main reason for starting this project it's because there are no fully free and open source migration tools available that allow versioning database routines with an standard CVS system as if they were normal application code.

Also, the existing tools are too complex and require too much knowledge to start a small project.

ToDo

• Undo changes when there is an error applying a version using "undo" files.
• Console logging via events.
• Lock version table row when pushing.
• Preserve all characteristics on pull: SQL mode, character set, algorithm...

Built With

• Git
• nodejs
• NodeGit
• docker