The replacement of kaliop/ezmigrationbundle for Ibexa 4.
This bundle makes it easy to programmatically deploy changes to Ibexa database structure and contents.
It is inspired by the DoctrineMigrationsBundle
Many thanks to Wizhippo for having started this version, and kindly donated the code!
-
PHP 7.4 or later.
-
Ibexa 4.0 (aka. Ibexa DXP).
For Ibexa DXP 3, head on to tanoconsulting/ezmigrationbundle2. For eZPlatform 1-2 and earlier, to kaliop-uk/ezmigrationbundle.
In either require
or require-dev
at the end of the bundle list in the composer.json file add:
composer require 'tanoconsulting/ibexa-migration-bundle:^1.0'
Save it and run
composer update --dev tanoconsulting/ibexa-migration-bundle
This will install the bundle and all its dependencies.
Then make sure that the bundle is active, generally by editing config/bundles.php
If you run php bin/console
you should see the following new commands in the list:
kaliop
kaliop:migration:generate
kaliop:migration:mass_migrate
kaliop:migration:migrate
kaliop:migration:migration
kaliop:migration:resume
kaliop:migration:status
This indicates that the bundle has been installed and registered correctly.
To get the latest version, you can update the bundle to the latest available version by using composer
composer update tanoconsulting/ibexa-migration-bundle
For upgrades from kaliop/ezmigrationbundle, see the instructions in ezmigrationbundle_to_ibexamigrationbundle.md.
All commands accept the standard Symfony/eZPlatform options, although some of them might not have any effect on the command's execution.
The bundle provides a command to easily generate a new blank migration definition file.
For example:
php bin/console kaliop:migration:generate --format=yml
The above command will place a new yml skeleton file in the MigrationsDefinitions
directory in ./src/
.
If the directory does not exist then the command will create it for you.
If the command is successful it will create a new yml file named with the following pattern: YYYYMMDDHHMMSS_placeholder.yml
.
You are encouraged to rename the file and change the placeholder
part to something more meaningful, but please keep
the timestamp part and underscore, as well as the extension.
(the contents of the skeleton Yaml file are stored as twig template)
To see all the migrations definitions available in the system and whether they have been applied or not simply run the status command in your eZPlatform root directory:
php bin/console kaliop:migration:status
The list of migrations which have been already applied is stored in the database, in a table named kaliop_migrations
.
The bundle will automatically create the table if needed.
In case you need to use a different name for that table, you can change the Symfony parameter ez_migration_bundle.table_name
.
To apply all available migrations run the migrate command in your eZPlatform root directory:
php bin/console kaliop:migration:migrate
NB: if you just executed the above command and got an error message because the migration definition file that you had just generated is invalid, do not worry - that is by design. Head on to the next paragraph...
NB: migrations execute by the default as the admin user with ID 14. Without this user account in the database, you must
specify the use of another admin account by passing the -a
flag.
To apply a single migration run the migrate command passing it the path to its definition, as follows:
php bin/console kaliop:migration:migrate --path=src/MyNamespace/MyBundle/MigrationVersions/20160803193400_a_migration.yml
NB: you can specify as well a folder with the --path
flag, in which case all the migration definitions contained in that
folder will be executed.
So far so good, but what kind of actions can be actually done using a migration?
Each migration definition consists of a series of steps, where each step defines an action.
A simple example of a migration to create a 'folder' content is:
-
mode: create
type: content
content_type: folder
parent_location: 2
attributes:
name: hello world
In a Yaml migration, you can execute the following types of actions:
- creation, update and deletion of Contents
- creation, update and deletion of ContentTypes
- creation, update and deletion of ContentTypeGroups
- deletion of Content Versions
- creation and deletion of Languages
- creation, update and deletion of Locations
- creation, update and deletion of ObjectStates
- creation, update and deletion of ObjectStateGroups
- creation, update and deletion of Roles
- creation, update and deletion of Sections
- creation and deletion of Tags (from the Netgen Tags Bundle)
- creation and deletion of URL aliases and wildcards
- creation, update and deletion of Users
- creation, update and deletion of UserGroups
- purging and recovering Contents from the Trash
- creation, appending, copy, renaming and deletion of files
- execution of SQL queries
- execution of command-line scripts
- execution of methods of Symfony services](Resources/doc/DSL/Services.yml)
- execution of php functions and static methods
- execution of http calls
- sending of email
- looping over arrays and executing one of the above actions on each element
- canceling, snoozing or suspending the migration itself
- generating and saving new migration definitions
More details on all supported migration-language features are in the DSL Language description
For more specific needs, you can also use 2 other types of migrations:
- SQL migrations
- PHP migrations
Example command to generate an SQL migration definition:
php bin/console kaliop:migration:generate create-new-table --format=sql
This will create the following file, which you are free to edit:
./src/MigrationsDefinitions/2021XXYYHHMMSS_mysql_create-new-table.sql
NB if you rename the sql file, keep in mind that the type of database to which it is supposed to apply is the part of the filename between the first and second underscore characters. If you later try to execute that migration on an eZPublish installation running on, say, PostgreSQL, the migration will fail. You are of course free to create a specific SQL migration for a different database type.
The Migration bundle itself imposes no limitations on the type of databases supported, but as it is based on the Doctrine DBAL, it will only work on the databases that Doctrine supports.
NB you can also save the SQL statement to execute in a yml-formatted migration file. This gives you access to more options, such as setting and resolving references. Yml-formatted migration files do not need to have the db type in their name.
NB if the SQL statement (or statements) in your migration is too long, the migration might fail or be only partially applied, in some cases (such as when using MySQL) without even reporting an error. If you need to execute multiple, long queries, you are better off splitting them, either in many .sql migrations, or in a single .yml migration with sql steps.
If the type of manipulation that you need to do is too complex for either YML or SQL, you can use a php class as migration definition. To generate a PHP migration definition, execute:
php bin/console kaliop:migration:generate AMigrationClass --format=php
This will create the following file, which you are free to edit:
./src/MigrationsDefinitions/2021XXYYHHMMSS_AMigrationClass.php
As you can see in the generated definition, the php class to be used for a migration needs to implement a specific interface. The Symfony DIC container is passed to the migration class so that it can access from it all the services, parameters and other thing that it needs.
For a more detailed example of a migration definition done in PHP, look in the MigrationVersions folder of this very bundle.
NB if you rename the php file, keep in mind that the filename and the name of the class it contains are tied - the standard autoloading mechanism of the application does not apply when loading the migration definition. This is also the reason why the php classes used as migrations should not use namespaces.
NB it is also possible to run any method of any existing Symfony service just by declaring it as migration step in a yaml migration. See the relevant DSL for details.
NB it is also possible to run any existing php function or static class method just by declaring it as migration step in a yaml migration. See the relevant DSL for details.
The easiest way to re-execute a migration in 'failed' status, is to remove it from the migrations table:
php bin/console kaliop:migration:migration migration_name --delete
After removing the information about the migration from the migrations table, running the migrate
command will execute it again.
By default the bundle runs each migration in a database transaction. This means that if a step fails, all the previous steps get rolled back, and the database is left in its original state. This is a safety feature built in by design;
- if you prefer the migration steps to be executed in separate transactions the easiest way is to create a separate migration file for each step
- you can use the command-line flag
-u
to disable usage of transactions by the migrate command
Note also that by default the migrate
command stops on the 1st failed migration, but it can be executed with a flag
to allow it to continue and execute all available migrations even in case of failures.
As for rolling back changes: given the nature of the eZPublish API, rolling back changes to Content is not an easy feat. As such, the bundle does not provide built-in support for rolling back the database to the version it had before applying a given migration. We recommend always taking a database snapshot before applying migrations, and use it in case you need to roll back your changes. Another approach consists in writing a separate migration to undo the changes.
An easy way to hook up custom logic to the execution of migrations - without having to implement your own customized action executors - is to use Event Listeners.
Two events are fired for each step during execution of migrations:
* ez_migration.before_execution => listeners receive a BeforeStepExecutionEvent event instance
* ez_migration.step_executed => listeners receive a StepExecutedEvent event instance
An event is fired only in case a migration fails because a step throws a specific migration abort exception:
* ez_migration.migration_aborted => listeners receive a MigrationAbortedEvent event instance
An event is fired when a migration is being generated using the kaliop:migration:generate
command, allowing to alter
the data that will be serialized as migration steps:
* ez_migration.migration_generated => listeners receive a MigrationGeneratedEvent event instance
In order to act on those events, you will need to declare tagged services, such as for ex:
my.step_executed_listener:
class: my\helper\StepExecutedListener
tags:
- { name: kernel.event_listener, event: ez_migration.step_executed, method: onStepExecuted }
and the corresponding php class:
use Kaliop\eZMigrationBundle\API\Event\StepExecutedEvent;
class StepExecutedListener
{
public function onStepExecuted(StepExecutedEvent $event)
{
// do something...
}
}
Event Subscribers are supported as an alternative to Event Listeners, as is standard with Symfony projects.
-
unlike the Doctrine Migrations Bundle, this bundle does not support rollback of changes. Read above for the reason why.
-
if you are using the Doctrine Migrations Bundle to manage your schema, you will get spurious sql created to handle the database tables belonging to Kaliop Migrations Bundle. For the moment, the best workaround is to use the
filter-expression
parameter on the command-line when runningdoctrine:migrations:diff
and friends, with a value ofkaliop_migrations_*
-
fatal errors when running a migration can manifest themselves when you are using the Solr Search Engine Bundle. In this case the problem is compounded the fact that, even if a node or object is sent to Solr from within a database transaction, the Solr search index might be configured to only commit received data within a short time delay. A known workaround involve:
- separate your migration steps into separate migrations
- running the migrations each in its own transaction (and process) by using the
-p
flag to themigrate
command - adding
sleep
migration steps to migrations 2 .. N - and/or configuring Solr to always commit changes to the index immediately (eg. disable
commitwithin
)
-
when using SOLR in multi-core configurations and getting a
java.lang.NegativeArraySizeException
error, you will have to set a lower value than the default 2147483647 for parameterez_migration_bundle.query_limit
-
if you get fatal errors without any error message when running a migration which involves a lot of content changes, such as f.e. altering a contentType with many contents, it might be that you are running out of memory for your php process. Known workarounds involve:
- increase the maximum amount of memory allowed for the php script by running it with option '-d memory_limit=-1'
- execute the migration command using a Symfony environment which has reduced logging and kernel debug disabled:
the default configuration for the
dev
environment is known to leak memory - transform the migration, where possible, into one which loads and modifies contents one by one in a loop, instead of modifying them all in a single action. See an example of using loops here.
-
if you get fatal errors with the message 'You cannot create a service ("request") of an inactive scope ("request")', take a look at the following issue for a possible explanation and ideas for workarounds: https://jira.ez.no/browse/EZP-24691
-
when updating a Role, you have to specify in the migration all the policies for it. Any existing policies that are not in the yml file will be removed. To make it easy to create a migration for updating a role, please use the
migration:generate
command using the--type=role
flag -
take care when creating content types: the eZPublish API, used internally by the migration bundle, will let you use dash characters in the content type identifiers, even if the resulting content types will then be unusable, eg.
Example of an invalid definition:
type: ezstring name: Topbar-hover-color identifier: topbar-hover-color
-
when eZ is set up in cluster mode, if you are setting references to the path of a content field of type ezimage, ezbinaryfile or ezmedia, or generating a migration for creating/updating it, the value you will get for the path will not be the absolute path on disk, but the path relative to the 'nfsvar' directory, which makes it unsuitable for being used directly in eg. a content/create migration. Check out the example in the Cookbook for how to deal with this.
A: use the 'reference/set' migration step to define a reference for the desired content Id, and use a Symfony parameter to store a different value for each Symfony environment. For example:
-
type: reference
mode: set
identifier: content_id_ref
value: '%a.parameter.name%'
-
type: content
mode: update
match:
content_id: "reference:content_id_ref"
etc: ...
Note that there are many more solutions for this issue, sych as making sure your target Contents and Locations have
the same Remote_id in all environments, or passing values for references as options to the migrate
command-line.
When using a migration to update a Role, you must define ALL its policies. Any not defined will be removed. The safest and simplest way to make sure that you do not forget any of the existing policies is to first generate an update migration that has the complete specification of the role as it currently is defined, and then edit manually.
Example command to create such a migration:
php bin/console kaliop:migration:generate --type=role --mode=update --match-type=identifier --match-value=Anonymous bundleName
When dumping a Content into a yml migration via the generate
command, the list of attributes is empty
A: this is most likely due to using a bad language configuration
A: yes, please take a look in folder Resources/doc/Cookbook/
A: sure. Take a look at the relevant dsl and cookbook example for details.
The bundle has been designed to be easily extended in many ways, such as:
- adding support for custom/complex field-types
- adding support for completely new actions in the Yml definitions
- adding support for a new file format for storing migration definitions
- adding support for new resolvers for the custom references in the migration definitions
- taking over the way that the migrations definitions are loaded from the filesystem or stored in the database
- etc...
Following Symfony best practices, for the first 4 options in the list above all you need to do is to create a service and give it an appropriate tag (the class implementing service should of course implement an appropriate interface).
To find out the names of the tags that you need to implement, as well as for all the other services which you can override, take a look at the services.yml file.
It is also possible to define custom event listeners/subscribers to expand migration execution logic. See the dedicated paragraphs above for more details.
The bundle uses PHPUnit to run functional tests.
To run the tests:
export KERNEL_CLASS=App\Kernel (or whatever you renamed it to)
export APP_ENV=behat (or whatever your environment is)
bin/phpunit --stderr -c vendor/tanoconsulting/ibexa-migration-bundle/phpunit.xml.dist
NB the tests do not mock interaction with the database, but create/modify/delete many types of data in it. As such, there are good chances that running tests will leave stale/broken data. It is recommended to run the tests suite using a dedicated eZPublish installation or at least a dedicated database.
A safer choice to run the tests of the bundle is to set up a dedicated environment, similar to the one used when the test suite is run on GitHub Actions. The advantages are multiple: on one hand you can start with any version of eZPublish you want; on the other you will be more confident that any tests you add or modify will also pass on GitHub. The disadvantages are that you will need Docker and Docker-compose, and that the environment you will use will look quite unlike a standard eZPublish setup! Also, it will take a considerable amount of disk space and time to build.
Steps to set up a dedicated test environment and run the tests in it:
git clone --depth 1 https://github.com/tanoconsulting/euts.git teststack
# if you have a github auth token, it is a good idea to copy it now to teststack/docker/data/.composer/auth.json
# this config sets up a test environment with Ibexa 4.2 running on php 8.0 / ubuntu jammy
export TESTSTACK_CONFIG_FILE=Tests/environment/.euts.4.2.env
./teststack/teststack build
./teststack/teststack runtests
./teststack/teststack stop
You can also run a single test case:
./teststack/teststack runtests ./Tests/phpunit/01_CollectionsTest.php
Note: this will take some time the 1st time your run it, but it will be quicker on subsequent runs. Note: make sure to have enough disk space available.
In case you want to run manually commands, such as the symfony console:
./teststack/teststack console cache:clear
Or easily get to a database shell prompt:
./teststack/teststack dbconsole
Or command-line shell prompt to the Docker container where tests are run:
./teststack/teststack shell
The tests in the Docker container run using the version of debian/php/eZPlatform kernel specified in the file
Tests/environment/.euts.4.2.env
, as specified in env var TESTSTACK_CONFIG_FILE
.
If no value is set for that environment variable, a file named .euts.env
is looked for.
If no such file is present, some defaults are used, you can check the documentation in ./teststack/README.md to find out
what they are.
If you want to test against a different version of eZ/php/mysql/debian, feel free to:
- create the
.euts.env
file, if it does not exist - add to it any required var (see file
teststack/.euts.env.example
as guidance) - rebuild the test stack
- run tests the usual way
You can even keep multiple test stacks available in parallel, by using different env files, eg:
- create a file
.euts.env.local
and add to it any required env var, starting with a uniqueCOMPOSE_PROJECT_NAME
- build the new test stack via
./teststack/teststack. -e .euts.env.local build
- run the tests via:
./teststack/teststack -e .euts.env.local runtests
This bundle adheres to Semantic Versioning principles.
However, backward compatibility comes in many different flavors. In fact, almost every change can potentially break an application. For example, if we add a new method to a class, this will break an application which extended that class and added the same method, but with a different method signature.
This section is dedicated to explain in detail which guidelines inform the choice of incrementing the major/minor/patch number for every new release of the bundle.
Adherence to Semantic versioning is implemented via the following:
- existing migrations will continue to work across all minor and patch version increases
- new migration steps and new supported elements for existing migration steps might be introduced in minor versions, but not in patch versions
- any migration steps or step element targeted for removal will first be deprecated for at least one minor version. When
that happens, said element will be immediately removed from the DSL documentation, and
@deprecated
orBC
comments added to the codebase next to where it is handled - the syntax for cli commands will continue to work across all minor and patch version increases
- new options for existing cli commands, and new commands, might be introduced in minor versions, but not in patch versions
- the textual output of existing cli commands might be altered in any version - please do not rely on it having a fixed format to parse it
- events emitted by the bundle will continue to work across all minor and patch version increases
- new events might be introduced in minor versions, but not in patch versions
Things are a bit more sketchy in terms of the internals of the bundle, ie. PHP classes and Symfony services. Although great care is taken to avoid breaking user code which extends existing classes or redefines existing services, this is not always possible.
- any change which might have a potential impact on developers extending or modifying the bundle should not happen in patch versions, but it might happen in minor versions. Any such change will only be made if there is a compelling reason to do so (ie. mostly because it is required to fix a bug or to make possible implementation of important new functionality) and be documented in the NEWS file