Continuous Learning

Avoid arrays in your domain and create iterable classes instead

PHP's array is a powerful datatype. You can add and remove values of any type at any time. This comes at a cost though: You have no control over what is allowed and what not. This can become a problem in your domain code where you need to define very strict rules about your domain. If you pass an array of User objects into a method, is it allowed to contain users who are inactive? If you want the array to be sorted (the user with the latest registration date for example), you cannot assume that the calling code has already done that.

But there is a solution to this problem: You can create a class which implements the \Traversable interface of PHP. The class needs to implement a getIterator(): Traversable method now, so you have full control over how it's elements are handled by external code. Encapsulating your iterables into their own classes might seem to bloat your code at first. But while this approach indeed comes with additional code, it has some major advantages.

Advantages

Internal type safety

When you encapsulate the concept of a list of User objects into it's own class, you can easily make sure to only allow User objects to be part of the list. Throw an exception if a different type is passed into it. Now you never have to do any type checks in the code that uses your iterable object.

Constrained access options

You have full control over which methods you add to your class. You don't have to worry that elements of your object are modified in a way you did not intend. Yes, you need to write those methods by yourself, but this gives you the power to decide for each individual case what is needed and what not.

External type safety

If your code gets an array passed in, you don't really know what you get. You need to look at the calling code and hope that it does not change. But if you get a custom UserList for example, you know exactly what it's possibilities and restrictions are.

Encapsulation of business logic

Often you need to perform the same operation on a list of objects multiple times, for example filter down a list of all users to only those with an active account. This code tends to go into service classes or - even worse - into controllers and other application layer code. Encapsulating these business rules into your iterable classes makes them easy to test and reuse. The calling code is also very straightforward to read: $users->filterOnlyActive().

Immutability

Writing your own methods enables you to always return a new object instead of modifying the existing one. This way you don't have to worry to break existing code when working with an iterable object.

Sorting

In an array you don't have information about it's sorting. While you can sort an array you need to do it directly in the code that uses it to make sure it wasn't modified elsewhere. By creating a dedicated class for a sorted list (e.g. SortedUserByAgeList) which makes sure the elements are always in the correct order you don't need to worry about it in the calling code any more.

Disadvantages

More Code

This one is pretty straightforward: You need to write those classes. If you don't have a complex domain model and you don't intend your application to be maintained for a long time, this might be overkill. But as soon as these classes exist, the improved readability and safety kicks in.

Performance

This one is more critical. The additional type checks and method calls make the execution of the code slower than using simple arrays. If this is a problem or not simply comes down to the use case and needs some profiling. Don't just dismiss the idea because of premature optimization.

Summary

I've been using this approach in multiple projects over the last few years and it made my code much more readable and safe. I use it mainly in my domain layer where safety is a must and I found that I could simplify the calling code in many places by moving business logic into my iterable classes.

In my first approaches I used a base class that provided a type check in the constructor and all the common methods for filtering, mapping, counting, etc. But this caused a huge inheritance tree which in turn caused many methods to exist only for a few subclasses and made adjustments very hard. Then I shifted the concept to a simple class that provides the common functionality but is instead used by the custom classes. This way I can decide for every custom class which methods to expose and just forward the method calls to the underlying generic one. Very specific methods are then handled by the custom classes on their own.

Now that I've come to a point where the structure of the generic classes do not change that much any more I extracted them into their own library: https://packagist.org/packages/cunningsoft/generic-list.

Author image
About Daniel Mecke
You've successfully subscribed to Continuous Learning
Great! Next, complete checkout for full access to Continuous Learning
Welcome back! You've successfully signed in.
Unable to sign you in. Please try again.
Success! Your account is fully activated, you now have access to all content.
Error! Stripe checkout failed.
Success! Your billing info is updated.
Error! Billing info update failed.