Create a reusable bundle¶
Here is a very early version of this section, we’ll keep enriching it in the upcoming weeks.
You are an integrator and you want to share reusable parts of your project with the awesome Akeneo PIM community? Let’s read our pieces of advice to write a great and documented bundle!
What is the purpose of a reusable bundle?
Compatibility: to fit Akeneo PIM features
Maintainability: to make your code flexible and testable (and tested)
Extensibility: to allow integrators to use your bundle to fit with their specific needs
Scalability: to make your application works with high volume of data and document its behavior in such cases
Your bundle won’t be reusable if you forget some cases. That is no more an issue if you document compatibilities of your bundle. Follow these cookbooks for more details: “How to Manipulate Products” and “How to Manipulate Non-Product Objects”.
There are two existing distributions: Community (CE) and Enterprise (EE) which comes with several extra features dedicated to advanced needs.
Many attribute types are managed (14 in CE and one more in EE). Make sure you do not forget any use cases for an attribute type.
Try your bundle with other catalogs. Don’t forget use cases with many locales, many channels and various catalog volumes.
Keep in mind the five SOLID principles of Object Oriented Programming and especially the Single Responsibility principle.
You can use PHP Mess Detector to help make your code simple, flexible and easily testable.
In Akeneo PIM, we use PhpSpec as unit testing tool and Behat for behavior-oriented tests.
Use a Continuous Integration (CI) tool to deploy your application, test your code and improve its quality. On our open-source projects, we use free CI tools Travis for unit tests. You can find some configuration examples on our Akeneo-Labs projects (travis.yml).
Another important point about maintainability is to rely on interfaces instead of concrete classes. It will facilitate your future migrations of Akeneo PIM.
Always prefer composition over inheritance. Indeed, class constructors can change from a minor version to another, while an interface will never change.
To give integrators a clean way to rely on your code, you can also create your own interfaces.
Follow Akeneo PIM dependencies¶
You MUST follow Akeneo PIM dependencies at all costs. Don’t change versions of the existing ones. Akeneo PIM has been tested with these dependencies. Changing them could impact the whole application performance. You may add new dependencies needed for your customisations.
Registries are also a good extension point when you have to deal with many cases (different attribute types, entities, etc.). They are used to gather related classes (i.e: filters, sorters). Simply declare it as a service and tag it.
Learn more about registry pattern.
You can also follow the localizers’ cookbook example: How to Use Localizers
Plug on events (datagrid, savers, remover, etc.)¶
It is the best extensibility point you have. Unfortunately, it is not always possible as Akeneo PIM cannot cover all cases where you would plug your code. In this case, you can still decorate or extend the desired class and add the extension point you need.
By the way, feel free to contribute by opening an issue to discuss with our team on our github repository. Your feedback is very valuable for us so we can improve our application.
Avoid to override Akeneo PIM classes/services¶
The easiest way to add a new behavior on your services is to override it. But wait! That’s not a good thing! First, each override leads to potential compatibility problems. Let’s assume that you overrode the pim_catalog.saver.product service and that a final customer wants to use your bundle and another one. If both of them override this service, the last bundle listed in the app/AppKernel.php will determine the one that will be used.
Another consequence is that it will have an impact on the whole application. That being said, it might be what you want to do.
You should avoid to extend models (entities and documents) in order to add properties. You would face the same problem as the one you encounter when you override classes/services. There are unfortunately no great solutions for these cases.
There is not a lot of changes on the entity you overrode. Document what is missing and let future integrators of your bundle handle this.
Set up a oneToOne unidirectional association. Here is an example with the Category entity where we want to add a description field. You could also create a brand new entity named MyCategoryDescription with an id, a description and a relation to the Category entity. As a side effect, you won’t have access to the description from the Category object (opposite is possible).
For your own model classes, create your class and its interface. Then you can rely on your interface and use the Akeneo target resolver which is based on the Doctrine target entity resolver.
Doctrine does not allow more than one repository per entity. For this reason, you can’t declare them as such. Nevertheless, you can create a service, inject the ObjectManager in it and the class you want to work on. Then you will have access to the query builder as in a doctrine repository and you can prepare the query you want.
Do you know the limitations of your application? It does not matter if you can’t handle millions of products but you have to document what are the limitations of your application.
Keep in mind that some users will use your bundle with a data volume you never thought could be possible (true story!).
You can test your bundle with the representative catalogs we provide: https://github.com/akeneo/catalogs
Remember never to use the findAll() method from a repository as you don’t know how many entities will be retrieved.
On batch processes, don’t forget to detach your objects from the Doctrine UnitOfWork and check the memory usage. You can use blackfire and php-meminfo to help you track memory leaks.
Found a typo or a hole in the documentation and feel like contributing?
Join us on Github!