Facilitates atomic database migrations in SilverStripe 3.x. Inspired by Laravel's migration capability, this relatively simple module was built to work as an augmentation to the existing dev/build
that is already commonplace in SilverStripe.
SilverStripe's database schema is declarative, which means that the code defines what the state of the database currently should be and therefore the system will do what it needs to do in order to change the current schema to match that declared structure at any given moment (by proxy of dev/build
). In contrast, database migrations offer a method to imperatively define how that structure (as well as the data) should change progressively over time. The advantage to this is that it makes it easier for you to rename columns (while retaining data), combine multiple columns or even change the format of data over time without leaving behind legacy code which should no longer exist, helping keep things tidy.
- Run
composer require "patricknelson/silverstripe-migrations:dev-master"
- Run
sake dev/build
from the command line to ensure it is properly loaded into SilverStripe.
- Download the latest repository zip file.
- Extract the folder
silverstripe-migrations-master
and rename itmigrations
. - Copy that folder to the root of your website.
- Run
sake dev/build
from the command line to ensure it is properly loaded into SilverStripe.
This module sets up a task called MigrateTask
which is available from the command line only via either:
sake dev/tasks/MigrateTask [option]
php framework/cli-script.php dev/tasks/MigrateTask [option]
Using this task, you can do the following:
- Run migrations (i.e. "up").
sake dev/tasks/MigrateTask up
- Reverse previous migrations (i.e. "down").
sake dev/tasks/MigrateTask down
- Make a new migration file for you with boilerplate code.
sake dev/tasks/MigrateTask make:migration_name
- Note: This file will be automatically placed in your project directory in the path
<project>/code/migrations
. You can customize this location by defining aMIGRATIONS_PATH
constant which should be the absolute path to the desired directory (either in your_ss_environment.php
or_config.php
files). Also, migration files that are automatically generated will be pseudo-namespaced with aMigration_
prefix to help reduce possible class name collisions.
Up
Each time you run an up
migration, this task will look through all migration files that have been setup in your <project>/code/migrations
folder (which you can customize) and then compare that to a list of migrations that it has already run in the DatabaseMigrations
table. If it finds a new migration that has not yet been run, it will execute the ->up()
method on that migration file and keep a record of it in that table. Also, it will make sure to only run migrations in alphanumeric order based on their file name.
Down
When multiple migrations are run together, they are considered a single batch and can be rolled back (or reversed) all together as well by using the down
option. When down
migrations are performed, they are done in reverse order one batch at a time to help ensure consistency.
You can easily generate migration files by running the task with the make:migration_name
option (switching out migration_name
with a concise description of your migration using only underscores, letters and numbers).
Example:
sake dev/tasks/MigrateTask make:change_serialize_to_json
This will generate a file following the format YYYY_MM_DD_HHMMSS_change_serialize_to_json.php
, using the current date to name the file (to ensure it is executed in order) and containing the class Migration_ChangeSerializeToJson
. It should look like this:
<?php
class Migration_ChangeSerializeToJson extends Migration {
/**
* Run the migrations.
*
* @return void
*/
public function up() {
// Go through each DataObject and convert the column format from serialized to JSON.
foreach(MyDataObject::get() as $instance) {
$instance->EncodedField = json_encode(unserialize($instance->EncodedField));
}
}
/**
* Reverse the migration (this does the opposite of the method above).
*
* @return void
*/
public function down() {
// Go through each DataObject and convert the column format back from JSON to serialized again.
foreach(MyDataObject::get() as $instance) {
$instance->EncodedField = serialize(json_decode($instance->EncodedField));
}
}
}
A collection of helper methods are also available on the Migration
class that perform common database interactions, such as dropping columns from a table or retrieving a value from a column that is no longer available through the ORM but still exists in the database table.
Documentation for helper methods can be found here.
It's important that before you ever migrate up
or down
that you make sure you run the SilverStripe dev/build
task first. For example, you could do the following:
sake dev/build
sake dev/tasks/MigrateTask up
This will ensure that both the migration classes are available (in the class map) and that the new fields you've declared in your DataObject
's are accessible to your migration.
Due to the fact that the existing dev/build
process runs independently from these migrations (instead of being based already on migrations), it is possible that you might end up running migrations that are no longer applicable to the given declared state in your current DataObject ::$db
static definitions. To help avoid issues in more complex websites, you can either perform checks within the migrations themselves or setup new migrations to coexist with code that is bundled into release branches and deploy them discretely (e.g. release-1.2.0
or hotfix-1.2.1
). The primary goal is to ensure that data/content in existing environments can be retained and changed selectively without losing a column, a table or replacing an entire table or database each time your schema has to change.
- Add ability to "prune" database fields and tables, which not only removes defunct fields/tables which are no longer used, but also re-orders them. Would be a highly destructive feature but potentially very helpful if used regularly (and responsibly).
- Method to obtain a hash as a signature for the current schema being defined by
DataObject
child classes. Will be helpful in creating migrations that should only be run if the current schema/state matches a certain hash. - Setup a
--pretend
option to allow the ability to preview all queries that will be executed by migrations (both up and down).