This article is an excerpt from my book Advanced Web Application Architecture. It contains a couple of sections from the conclusion of Part I: Decoupling from infrastructure.

This chapter covers:

• A deeper discussion on the distinction between core and
  infrastructure code
• A summary of the strategy for pushing infrastructure to the sides
• A recommendation for using a domain- and test-first approach to
  software development
• A closer look at the concept of “pure” object-oriented programming

Core code and infrastructure code

In Chapter 1 we’ve looked
at definitions for the terms core code and infrastructure code. What
I personally find useful about these definitions is that you can look at
a piece of code and find out if the definitions apply to it. You can
then decide if it’s either core or infrastructure code. But there are
other ways of applying these terms to software. One way is to consider
the bigger picture of the application and its interactions with
actors. You’ll find the term actor in books about user stories and use
cases by authors like Ivar Jacobson and Alistair Cockburn, who make a
distinction between:

1. Primary actors, which act upon our system
2. Secondary or supporting actors, upon which our system acts

As an example, a primary actor could be a person using their web
browser to send an HTTP POST request to our application. A
supporting actor could be the relational database that our application
sends an SQL INSERT query to. Communicating with both actors requires
many infrastructural elements to be in place. The web server should be
up an running, and it should be accessible from the internet. The server
needs to pass incoming requests to the application, which likely uses a
web framework to process the HTTP messages and dispatch them to the
right controllers. On the other end of the application some data may
have to be stored in the database. PHP needs to have a PDO driver
installed before it can connect to and communicate with the database.
Most likely you’ll need a lot of supporting code as well to do the
mapping from domain objects to database records. All of the code
involved in this process, including a lot of third-party libraries and
frameworks, as well as software that isn’t maintained by yourself (like
the web server), should be considered infrastructure code.

Most of the time between the primary actor sending an HTTP request to
your server, and the database storing the modified data, will be spent
by running infrastructure code and most of this code can be found in PHP
extensions, frameworks, and libraries. But somewhere between ingoing and
outgoing communication the server will call some of your own code, the
so-called user code.

User code is what makes your application special: what things can you
do with your application? You can order an e-book. You can pay for it.
What kind of things can you learn from your application? You can see
what e-books are available. And once you’ve bought one, you can download
it. Frameworks, libraries, and PHP extensions could never help you with
this kind of code, because it’s domain-specific: it’s your business
logic.

The following figure shows that user
code is in the middle of a lot of infrastructure code:

Even if we try to
ignore most of the surrounding infrastructure while working on and
testing user code, we’ll often find that this code is hard to work with.
That’s because the code still contains many infrastructural details. A
use case may be inseparable from the web controller that invokes it. The
use of service locators and the likes prevents code from running in
isolation, or in a different context. Calls to external services require
the external service to be available when we want to locally test our
code. And so on…

If that’s the case, user code consists of a mix of infrastructure code
and core code.
The following figure shows what this looks like:

When I look at this diagram, I immediately feel
the urge to push the bits of infrastructure co

Truncated by Planet PHP, read more at the original (another 12319 bytes)

A lot can happen in 9 years. Back then I was still advocating that you should unit-test your controllers and that setter injection is very helpful when replacing controller dependencies with test doubles. I’ve changed my mind: constructor injection is the right way for any service object, including controllers. And controllers shouldn’t be unit tested, because:

• Those unit tests tend to be a one-to-one copy of the controller code itself. There is no healthy distance between the test and the implementation.
• Controllers need some form of integrated testing, because by zooming in on the class-level, you don’t know if the controller will behave well when the application is actually used. Is the routing configuration correct? Can the framework resolve all of the controller’s arguments? Will dependencies be injected properly? And so on.

The alternative I mentioned in 2012 is to write functional tests for your controller. But this is not preferable in the end. These tests are slow and fragile, because you end up invoking much more code than just the domain logic.

Ports and adapters

If you’re using a decoupled approach, you can already test your domain logic using fast, stable, coarse-grained unit tests. So you particularly don’t want to let your controller tests also invoke domain logic. You only want to verify that the controller correctly invokes your domain logic. We’ve seen one approach in Talk review: Thomas Pierrain at DDD Africa, where Thomas explained how he includes controller logic in his coarse-grained unit tests. He also mentioned that it’s not “by the book”, so here I’d like to take the time to explain what by the book would look like.

Hexagonal architecture prescribes that all application ports should be interfaces. That’s because right-side ports should potentially have more than one adapter. The port, being an interface, allows you to define a contract for communicating with external services. On the left side, the ports should be interfaces too, because this allows you to replace the port with a mock when testing the left-side adapter.

Following the example from my previous post, this is a schema of the use case “purchasing an e-book”:

PurchaseEbookController and PurchaseRepositoryUsingSql are adapter classes (which need supporting code from frameworks, libraries, etc.). On the left side, the adapter takes the request data and uses it to determine how to call the PurchaseEbookService, which represents the port. On the right side there is another port: the PurchaseRepository. One of its adapters is PurchaseRepositoryUsingSql. The following diagram shows how this setup allows you to invoke the left-side port from a test runner, without any problem, while you can replace the right-side port adapter with a fake repository:

Left-side adapter tests

Since the test case replaces the controller, the controller remains untested. Even though the controller should be only a few lines of code, there may be problems hiding there that will only be uncovered by exploratory (browser) testing or once the application has been deployed to production.

In this case it would help to create an adapter test for the controller. This is not a unit test, but an integrated test. We don’t invoke the controller object directly, but travel the normal route from browser request to response, through the web server, the framework, and back. This ensures that we got everything right, and leave (almost) no room for mistakes in interpreting framework configuration (routing, security, dependency injection, etc.).

We have already established that you wouldn’t want to test domain logic again through a web request. Two things we need to solve for adapter tests then:

1. The controller shouldn’t invoke the domain logic. Instead, we should only verify that it calls the port (which is an interface) in the right way. The pattern for this is called mocking: we need a test double that records the calls made to it and makes assertions about those calls (the number of times, and the arguments provided).
2. We need a way to inject this mock into the controller as a constructor argument. Thi

Truncated by Planet PHP, read more at the original (another 7582 bytes)

Truncated by Planet PHP, read more at the original (another 22440 bytes)

Where should it go?

If you’re one of those people who make a separation between an application and a domain layer in their code base (like I do), then a question you’ll often have is: does this service go in the application or in the domain layer? It sometimes makes you wonder if the distinction between these layers is superficial after all. I’m not going to write again about what the layers mean, but here is how I decide if a service goes into Application or Domain:

Is it going to be used in the Infrastructure layer? Then it belongs in the Application layer.

Dependency rule V2

I like to follow the Dependency rule: layers can have only inward dependencies. This ensures that the layers are decoupled. The Infrastructure layer can use services from the Application layer. The Application layer can use Domain layer services. In theory, Infrastructure could use services from Domain, but I’d rather not allow that. I want the Application layer to define a programming interface/API that can be used by the Infrastructure layer. This makes the Domain layer, including the Domain model, an implementation detail of the Application layer. Which I think is rather cool. No need to do anything with aggregates, or domain events; as long as everything can be hidden behind the Application-as-an-interface.

Making this a bit more concrete, consider the use case “purchasing an e-book”. In code it will be represented as a PurchaseEbookController, living in Infrastructure, which creates a PurchaseEbook command object, which it passes to the PurchaseEbookService. This service is an application service, living in the Application layer. The service creates a Purchase entity and saves it using the PurchaseRepository, living in the Domain layer. At runtime, a PurchaseRepositoryUsingSql will be used, living in Infrastructure, which implements the PurchaseRepository interface.

Why is the PurchaseRepository interface in Domain? Because it won’t and shouldn’t be used directly from Infrastructure (e.g. the controller). The same goes for the Purchase entity. It should only be created or manipulated in controlled ways by application services. But from the standpoint of the Infrastructure layer, we don’t care if the application service uses an entity, a repository interface, or any other design pattern. As long as it does its job in a decoupled way. That is, it’s not coupled to specific infrastructure, neither by code nor by the need for it to be available at runtime.

Why is the application service in the Application layer? Because it’s called directly from the controller, which is Infrastructure. The application service itself is part of the API defined by the Application layer.

Application-as-an-interface or ApplicationInterface?

This gets us to an interesting possibility, which is somewhat experimental: we can define the API that the Application layer offers to its surrounding infrastructure as an actual interface. E.g.

namespace Application; interface ApplicationInterface
{ public function purchaseEbook(PurchaseEbook $command): void; /** * @return EbookForList[] */ public function listAvailableEbooks(): array;
}

That second method, listAvailableEbooks() is an example of a view model that could be made accessible via the ApplicationInterface as well.

I think the-application-as-an-interface is a nice design trick to force Infrastructure to be decoupled from Application and Domain code. Infrastructure, like controllers, can only invoke Application behavior via this interface. Another advantage is that creating acceptance tests for the application becomes really easy. You only need the ApplicationInterface in your test and you can run commands and queries against it, to prove that it behaves correctly. You can also create better integration tests for your left-side, or input adapters, because you can replace the entire core of your application by mocking a single interface. I’ll leave a discussion of the options for another article.

This is an excerpt from my book PHP for the Web. It’s a book for people who want to learn to build web applications with PHP. It doesn’t focus on PHP programming, but shows how PHP can be used to serve dynamic web pages. HTTP requests and responses, forms, cookies, and sessions. Use it all to build a CRUD interface and an authentication system for your very first web application.

Chapter 11: Error handling

As soon as we started using a PHP server to serve .php scripts in Chapter 2 we had to worry about errors and showing them in the browser.
I mentioned back then that you need to make a distinction between the website as it is still running on your own computer and the website as it is running on a publicly accessible server.
You may find that people talk about this distinction in different ways.
When you’re working on your website on your own computer you’re running it “locally” or on your “development server”.
When it runs on a publicly accessible server it has been “deployed” to the “production server”.
We use different words here because these are different contexts or environments and there will be some differences in server configuration and behavior of the website depending on whether it runs locally or on the production server.
In this chapter we’ll improve the way our website handles errors and we’ll make this dependent on the environment in which the website runs.

Producing an error

Before we can improve error handling, let’s create a file that produces an error, so we can see how our website handles it.
Create a new script in pages/ called oops.php.
Also add it to the $urlMap in index.php so we can open the page in the browser:

$urlMap = [ '/oops' => 'oops.php', // ...
];

This isn’t going to be a real page, and we should remove it later, but we just need a place where we can freely produce errors.
The first type of error we have to deal with is an exception.
You can use an exception to indicate that the script can’t do what it was asked to do.
We already saw one back in Chapter 9 where the function load_tour_data().
The function “throws” an exception when it is asked to load data for a tour that doesn’t exist:

function load_tour_data(int $id): array
{ $toursData = load_all_tours_data(); foreach ($toursData as $tourData) { if ($tourData['id'] === $id) { return $tourData; } } throw new RuntimeException('Could not find tour with ID ' . $id);
}

In oops.php we’ll also throw an exception to see what that looks like for a user:

<?php throw new RuntimeException('Something went wrong');

Start the PHP server if it isn’t already running:

php -S 0.0.0.0:8000 -t public/ -c php.ini

Then go to http://localhost:8000/oops.
You should see the following:

The reason the error shows up on the page is because we have set the PHP setting display_errors to On in our custom php.ini file.
We loaded this file using the -c command-line option.

Seeing error messages on the screen is very useful for a developer like yourself: it will help you fix issues quickly.
But it would be quite embarrassing if this showed up in the browser of an actual visitor of the website.

Using different configuration settings in production

Once the website has been deployed to a production environment, the display_errors setting should be Off.
While developing locally you can simulate this by using multiple .ini files.
The php.ini file we’ve been using so far will be the development version.
We’ll create another .ini file called php-prod.ini that will contain settings that mimic the production environment.

; Show no errors in the response body
display_errors = Off

When you supply multiple php.ini files

Truncated by Planet PHP, read more at the original (another 18479 bytes)

As a rather unusual pastime for the Saturday night I attended the third Domain-Driven Design Africa online meetup. Thomas Pierrain a.k.a. use case driven spoke about his adaptation of Hexagonal architecture. “It’s not by the book,” as he said, but it solves a lot of the issues he encountered over the years. I’ll try to summarize his approach here, but I recommend watching the full talk as well.

Hexagonal architecture

Hexagonal architecture makes a distinction between the use cases of an application, and how they are connected to their surrounding infrastructure. Domain logic is represented by pure code (in the FP sense of the word), surrounded by a set of adapters that expose the use cases of the application to actual users and connect the application to databases, messages queues, and so on.

The strict separation guarantees that the domain logic can be tested with isolated tests (“unit tests”, or “acceptance tests”, which run without needing any IO). The adapter code will be tested separately from the domain logic, with adapter tests (“integration tests”). Finally, when connecting domain logic and adapters, the complete running application can be tested with end-to-end tests or exploratory tests.

Pragmatic hexagonal architecture

Thomas notes that in practice, developers like to write unit tests and acceptance tests, because they are fast, and domain-oriented. Adapter tests are boring (or hard) to write, and so they are often neglected. Thomas noticed that the adapters are where most of the bugs reside. So he stretches acceptance tests to also include part of the left-side adapters and part of the right-side adapters. Only the actual IO gets skipped, because the penalty is too high: it would make the test slow, and unpredictable.

I think it’s somewhat liberating to consider this an option. I’ve also experimented with tests that leave out the left-side adapter but include several right-side adapters. I always felt somewhat bad about it; it wasn’t ideal. Indeed, it may still not be ideal in some situations, but at least it gives you some options when you’re working on a feature.

I find I often don’t test left-side adapters on their own, so they are a nice place for mistakes to hide until deployment. Being able to make them part of an acceptance test is certainly a way to get rid of those mistakes. However, the standard arguments against doing this still hold up. Your acceptance tests become tied to the delivery mechanism. By invoking your web controller in an acceptance test, you’re coupling it to framework-specific and web-specific classes. This is going to be a long-term maintenance issue.

The same goes for the right-side adapters. If we are going to test part of those adapters in an acceptance test, the test will in fact end up being coupled to implementation logic, or a specific database technology. Thomas mentions that only the “last mile”, the actual IO, will be skipped. I took this to mean that your test may, for instance, use the real repository, but not provide the real database connection to it. This, again, seems like a very valuable technique. It saves some time creating a separate adapter test for the repository. However, this also comes at the price of increased coupling. The acceptance test verifies that a certain query will be sent to the database, but this will only be useful as long as we’re using this particular database.

Thomas explains that we can reduce the coupling issue by making assertions at a higher abstraction level, but even then, the acceptance tests being tied to specific technologies like that greatly reduces the power that came with hexagonal architecture: the ability to swap adapters, or experiment with alternative adapters, while leaving the tests intact. On the other hand, it is cool to have the option to write fewer types of tests and cover more or less the same ground.

Concerns

Although the concept of an acceptance test gets stretched a bit, it still doesn’t invoke any IO, which means it’s still mostly follows the hexagonal architecture approach, where we should be able to replace left-side adapters with our test runner, and replace right-side adapters with some fake adapters. However, when an ac

Truncated by Planet PHP, read more at the original (another 1759 bytes)

You’ve picked a good refactoring goal. You are prepared to stop the project at anytime. Now how to determine the steps that lead to the goal?

Bottom-up development

There is an interesting similarity between refactoring projects, and regular projects, where the goal is to add some new feature to the application. When working on a feature, I’m always happy to jump right in and think about what value objects, entities, controllers, etc. I need to build. Once I’ve written all that code and I’m ready to connect the dots, I often realize that I have created building blocks that I don’t even need, or that don’t offer a convenient API. This is the downside of what’s commonly called “bottom-up development”. Starting to build the low-level stuff, you can’t be certain if you’re contributing to the higher-level goal you have in mind.

Refactoring projects often suffer from the same problem. When you start at the bottom, you’ll imagine some basic tasks you need to perform. I find this kind of work very rewarding. I feel I’m capable of creating a value object. I’m capable of cleaning up a bit of old code. But does it bring me any closer to the goal I set? I can’t say for sure.

Top-down development

A great way to improve feature development is to turn the process around: start with defining the feature at a higher level, e.g. as a scenario that describes the desired behavior of the application at large, and at the same time tests if the application exposes this behavior (see Behavior-Driven Development). When you take this top-down approach, you’ll have less rework, because you’ll be constantly working towards the higher-level goal, formulated as a scenario.

The Mikado Method

For refactoring projects the same approach should be taken. Formulate what you want to achieve, and start your project from there. This has been described in great detail in the book The Mikado Method, by Ola Ellnestam and Daniel Brolund. I read it a few years ago, so I might not be completely faithful to the actual method here. What I’ve taken from it is that you have to start at the end of the refactoring project. I’ll give an example from my current project, where the team has decided they want to get rid of Doctrine ORM as a dependency. This seems like a daunting task. But Mikado can certainly help here.

The first thing to do is to actually remove the Doctrine ORM dependency: composer remove doctrine/orm. Commit this change, run the tests and of course you’ll get an error: could not find class Doctrine\ORM\... in file xxx.php on line y. So now you know that before you can remove the doctrine/orm package you have to ensure that file xxx.php does not use that class from the Doctrine\ORM namespace anymore. This is called a prerequisite; something we need to do first, before we can even think about doing the big change. We now have to revert the commit, because it’s not a commit we should keep; we broke the application.

This may seem like a stupid exercise, but it’s still very powerful because it’s an empirical method. Instead of guessing what needs to be done to achieve the end goal, you are now absolutely sure what needs to be done. In this example, it may be quite obvious that removing a package means you can no longer use classes from that package, but in other situations it’s less obvious.

The next step is to rewrite file xxx.php in such a way that it no longer uses those classes from Doctrine\ORM. This may require a bit of work, like rewriting the mapping logic. You can define all those tasks as prerequisites too.

When you’re done with any of the prerequisites, and the tests pass, you can commit and merge your work to the main branch. Everything is good. Of course, you still have all those other classes that use Doctrine\ORM classes, but at least there’s one less. You are closer to your end goal.

You can stop at any time

Being able to commit and merge smaller bits of work multiple times a day means that Mikado is compatible with the rule that you should be able to stop at any time. You can (and should) use short-lived branches with Mikado. Everything you do is going to get you closer to your goal, but also doesn’t break the project in any way.

With Mikado it actually feels like with every commit you’re gaining XP so you can unlock new levels for your application

Truncated by Planet PHP, read more at the original (another 1515 bytes)

In this blog, we look at the new features in Zend Server 2020 for IBM i, including improvements in packaging and performance.

Truncated by Planet PHP, read more at the original (another 13687 bytes)

Refactoring is often mentioned in the context of working with legacy code. Maybe you like to define legacy code as code without tests, or code you don’t understand, or even as code you didn’t write. Very often, legacy code is code you just don’t like, whether you wrote it, or someone else did. Since the code was written the team has introduced new and better ways of doing things. Unfortunately, half of the code base still uses the old and deprecated way…

The need to be consistent is very strong in me, and from what I see, in many developers. And so we get started and improve every bit of code everywhere. We become frustrated if we don’t “get” time for this work from our managers (“they don’t get it”), or if we can’t finish it; look at all this code, it’s so outdated!

Consistency is a bad refactoring goal

I’ve had this experience many times, and it made me give up on consistency as a refactoring goal. Not because consistency isn’t good. It is, because the uniformity of a code base makes it easier to contribute to it. Fewer surprises means you’ll make fewer mistakes. The problem is consistency as a refactoring goal:

1. It’s an all-or-nothing goal. Either the code base is consistent, or it isn’t. When you start the project, you have to finish it, or you’ll feel unsatisfied. This conflicts with what we established in the previous post: prepare to stop at any time.
2. It doesn’t serve a goal for other stakeholders. Which means it will be very easy for managers to pull the plug on the refactoring project, if they find out it’s costing them valuable development hours.

Higher refactoring goals

When it comes to refactoring projects, the development team (or sometimes just one developer) is often the primary stakeholder. Some common refactoring goals that developers have are:

• Being able to upgrade to PHP 7
• Being able to run tests without an actual database
• Being able to rely on static analysis for support during development

When discussing these goals in a meeting with business stakeholders, they will downplay them because they don’t seem relevant for their own cause. They aren’t right, of course, because what’s great about these “developer-oriented” goals is that they actually serve higher goals, goals that also serve business stakeholders:

• Being able to keep the software running for years to come
• Being able to develop new features faster
• Being able to release fewer mistakes to production

Alignment

Very often when I find my refactoring projects being postponed or blocked, I realize it’s because I didn’t explain the higher goals. What is often really useful is to talk to other stakeholders or decision makers (and pardon the management speak):

• What are the benefits they’ll notice, e.g. being able to work faster, release fewer mistakes, those are really good selling points for a refactoring project.
• Explain that you can work on this for just a few hours each week and still get those benefits. This can take away the fear that this may be one of those endless projects that have to get cancelled in the end. It forces you to think about refactoring steps, and being able to stop (and start) at any time.
• What are things you’ll unlock for the company, e.g. by deploying your application as a Docker image, deployments will be a single-step process, which finishes in a matter of seconds.

Conclusion

In summary, for a successful refactoring project you need to be able to stop and continue at any time. Establish refactoring goals that serve higher goals. Explain to business stakeholders that your development goals have benefits for them too.

Once the refactoring project gets the green light from the team, the next task is to determine refactoring steps. To be continued!