Skip to content

Latest commit

 

History

History
218 lines (182 loc) · 9.98 KB

README.md

File metadata and controls

218 lines (182 loc) · 9.98 KB

This code was written by Andrew Otto for CouchSurfing.org and the Kohana 2 PHP framework. It has been extracted from CouchSurfing's codebase and put here for reference. This code will not work by itself, but could be adapted into a more general PHP MySQL migrations library.

Also note that the use of timestamped migrations, rather than incremental migration numbers was hacked on as an after the fact modification and not part of the original design. I would change some of the logic and structure dealing with timestamped migrations if I had to do this again :)

What are Migrations?

Migrations are a way to keep database schema changes in a version control system. Often, migrations are coded in a database-agnostic manner. Meaning that the DDL statements are usually abstracted so that any database schema may be recreated on any RDBMS.

This implementation does not do this. Each migration file specifies the exact MySQL statements that need to be run in order for the schema to be brought up to date. By default, migration files are stored in modules/database/migrations. These files are ordered by timestamp.
The file name must match the class name inside, and each migration class must extend class Migration. Migration is an abstract class representing one migration. It has two abstract methods: the 'up' method and the 'down' method. Your migration subclass must implement both of these methods. The 'up' method is used for running migrations forward, and 'down' is used to migrate backwards.

Example

Example File: 20120104192233_AlterMyTable.php

class AlterMyTable extends Migration
{
  public function up()
  {
    $this->statement('main', 
      'ALATER TABLE `my_table` 
       ADD COLUMN `newcol` INT, data VARCHAR(100);'
    );
  }
  
  public function down()
  {
    $this->statement('main', 'ALTER TABLE `my_table` DROP COLUMN `newcol`;');
  }
}

This migrations 'up' method adds a statement to alter the table 'my_table' on the database connection 'main'. The Migration#statement() method is the really important one here. It is how you indicate what statements need to be run for a migration. You may call $this->statement() as many times as you need to. Statements will be run in the order that they are added.

The migrate script (located in bin/migrate) can generate skeleton migration files for you. See the help info by running 'migrate --help'.

Schema Migration

The schema_migration table is used to keep track of which schema version your database is currently at. When you update your codebase, you could possibly then get a new migration file from another developer. When you run migrate, the code will notice that there is a new migration file and add a record to the schema_migration table. If you tell it to do so, the migrate script will automatically run this new migration's up() method. See the migrate script's help documentation by running 'migrate --help'.

Use Case 1

Doodeedoo, here I am, a happy developer working on code. I'll be a good developer and run svn update before continuing to work. Oh! Look! Someone checked in a new migration! I will run the migrate script to make sure my database is up to date as well:

$ bin/migrate
Migrating up to version 20120104192233
Migrating up through schema version 20120104192233: 'create_boogers_table'...

--
-- SQL Statements for migrating up through schema version 20120104192233: create_boogers_table
--
-- Statements for database connection main 
USE `cs_core`; 
CREATE TABLE boogers (id INT, data VARCHAR(100));

Success!

Yay! Look now I have that new boogers table. That boogers table was the best idea ever. Great. Now let me get back to coding.

2 hours later...

ACK! ACK! That boogers table needs an index! Doh! It will never run in production like that! First, I will generate a new migration:

$ bin/migrate generate boogers_add_data_index Generating new migration file at .../modules/migration/migrations/20120104212233_boogers_add_data_index.php.

Now I will go edit that file and add the index. Oh yeah, and I can't forget to add proper instructions on how to migrate back down as well. Ah, there we go, now it looks like this:

class boogers_add_data_index extends Migration
{
	public function up()
	{
		$this->statement('main', "ALTER TABLE `booger` ADD INDEX `data` (data);");
	}

	public function down()
	{
		$this->statement('main', "ALTER TABLE `booger` DROP INDEX `data`;");
	}
}		

Beautiful. Now I can run the migrate script to get the index, and check this file in so that other developers can get the index on their databases as well.

Use Case 2: Database Creation

Very rarely you will have to create or drop an actual database. Migrations supports this, but it is best if you do not use the statement() method to run direct CREATE and DROP DATABASE statements. Instead, use the methods create_database() and drop_database(). To create a new database:

class create_cs_yay_database extends Migration
{
	public function up()
	{
		$this->create_database('cs_yay');
	}

	public function down()
	{
		$this->drop_database('cs_yay');
	}
}		

Notice that neither of these methods actually take any SQL or a database name.
The database name to create or dropped will be inferred from the database configs for the 'cs_yay' connection. Note that this requires that the 'cs_yay' connection is actually defined in the config/database.php file.

But what about production databases?

Migrations are not intended to be run directly on anything but the developer databases. They do two important things.

  • Keep track of schema changes.
  • Automatically sync these schema changes to developer databases.

If you are a sysadmin looking for a way to run a migration on a real production database, then you should use the migrate tool to print out the migrations that you need.

bin/migrate --output

This will print out only the migration SQL suitable for running on a MySQL instance. You may save this output in a file, or copy and paste it to run it.

bin/migrate script

usage help

$ bin/migrate --help
Database Migration Command Line Interface

Usage:
  bin/migrate [options] <version>
  bin/migrate [options] <command> [options] [args]

Options:
  -n, --dry-run                    If specified, no migrations will be
                                   executed.
  -o, --output-only                If specified, no migrations will be
                                   executed, and no log messages will be
                                   printed.  Instead, executable SQL will
                                   be printed to STDOUT.
  -a, --all                        If specified, then ALL migrations in the
                                   given range will be attempted to be run,
                                   regardless of their migrated_at status. 
                                   This is useful with the --output-only
                                   flag, so that you may print out all
                                   migrations in given range.
  -m PATH, --migrations-path=PATH  Path to directory where migrations are
                                   stored.  If this is not specified the
                                   default path from Kohana
                                   config/migrations.php will be used.
  -s, --skip-schema-version        Don't update the schema_version table
                                   after running migrations.
  -h, --help                       show this help message and exit

Arguments:
  version  Version to migrate to.   This may be a single version (e.g. 5),
           or a version range (e.g. 20120101000039:20120104192233).  A
           version range of X:Y will migrate from X (exclusive) to Y
           (inclusive).  You may also use X: to migrate from X to the
           latest schema version.  Optional.  If not specified, then this
           will migrate all un migrated versions.  You may also add a minus
           or plus (- or +) at the beginning of a version number, to
           indicate that you would like to run that single migration down
           or up. (e.g. -20120101000039 or +20120101000039).  Note that to
           use a minus sign, you must add -- before your version argument
           to indicate that the version argument is not a command line
           option. (e.g.  bin/migrate -- -20120101000039).

Commands:
  status    Prints out database migration status, including the current
            database version.
  generate  Generates a new skeleton migration file

status command

$ bin/migrate status

Database Schema Migration Status
  Migrations last ran at:           2012-02-28 21:02:29
+---------------------------------------------------------------+----------------+---------------------+
| Migration Name                                                | Version        | migrated_at         |
+---------------------------------------------------------------+----------------+---------------------+
| base_schema                                                   | 0              | 2012-01-17 20:27:49 |
| Survey_info_create_table                                      | 20120101000001 | 2012-01-17 20:27:49 |
| campaign_schema                                               | 20120101000002 | 2012-01-17 20:27:49 |
| Survey_info_and_research_survey                               | 20120101000003 | 2012-01-17 20:27:49 |
| activity_sphinx_index                                         | 20120101000004 | 2012-01-17 20:27:49 |
| tr_i18n_archive                                               | 20120101000005 | 2012-01-17 20:27:49 |
| drop_org_goal_tables                                          | 20120101000006 | 2012-01-17 20:27:49 |
...