Loïc Faugeron Technical Blog

Tests: Tools overview 12/02/2014

This article is part of a series on Tests in general and on how to practice them:

  1. Introduction
  2. Tools overview
  3. Test Driven Development
  4. TDD: just do it!
  5. spec BDD
  6. phpspec: a quick tour
  7. Behavior Driven Development: story BDD
  8. Behat: a quick tour
  9. Conclusion

This article can be read by any regular developer (no special level required), we'll put some code on the principles we've previously seen and we'll have a quick look on existing tools. Here's what we'll cover:

  1. test frameworks
  2. PHPUnit
  3. unit tests with test doubles (mocks and stubs)
  4. functional tests
  5. user interface tests

TL;DR: jump to the conclusion.

Test frameworks

In order to automate your tests (whether they're unitary, functionnal or anything else), you'll need to use some tools. These would be libraries helping you doing assertions, or libraries helping you creating test doubles without writing any new classes, or even frameworks which group those libraries together.

Most of the test frameworks follow the xUnit convention, which have been (accidently?) created by Kent Beck. Those are composed of:

Those are, among a ton of others:

You could aslo find Behavior Driven Development (BDD) style test frameworks:

I won't talk about these, as it will be the subject of a future article (there's so much to say about them).

Full stack frameworks isn't the only thing around here to help you write tests, there also are some libraries:

Choose your weapon wisely!


I mainly code in PHP, and in this language PHPUnit is the most popular test framework. It's been there for so long (version 1.0.0 released in July 2006) that almost any libraries and frameworks are tested with it. So it'll be our tool for the next examples.

You can install it using Composer:

curl -sS https://getcomposer.org/installer | php # Download composer
composer install "phpunit/phpunit:~3.7"
php vendor/bin/phpunit -h

Note: if you don't know Composer, let's just say that it makes your life easier by downloading for you the libraries you told him to (it takes care of selecting the good versions and can update them to get bug fixes). It also autoloads your classes so you don't have to require them.

Now that you have the latest stable version, you'll need to configure it:

<?xml version="1.0" encoding="UTF-8"?>
<!-- File: phpunit.xml -->
        <testsuite name="Fossil Test Suite">
            <directory suffix="Test.php">./test/</directory>

This configuration tells PHPUnit to look (recursively) in the test directory for files ending in Test.php. Those will be your test suites.

You'll need a bootstrap file in order to use composer's autoloader:


$loader = require __DIR__.'/../vendor/autoload.php';
$loader->add('Gnugat\\Fossil\\Test', __DIR__);

And voilà! You can now write your tests in test.

Unit tests

The common understanding of unit test is a symmetry between classes and tests: when you have a class with 2 methods, you need to have a test class for it which will test these two methods.

It looks like this wasn't the real meaning of the term unit, which should have meant making tests which can be run in any order without ruinning them, and as many times as wanted. Fortunately a new kind of test was created to fix this misunderstanding, so let's stick with the common one.

Imagine you have the following class, which creates a file if it doesn't already exist:

// File: src/MarkdownFile/DocumentationWriter.php

namespace Gnugat\Fossil\MarkdownFile;

use Symfony\Component\Filesystem\Filesystem;

class DocumentationWriter
    private $filesystem;

    public function __construct(Filesystem $filesystem)
        $this->filesystem = $filesystem;

    public function write($absolutePathname, $content)
        if (!$this->filesystem->exists($absolutePathname)) {
            $this->filesystem->dumpFile($absolutePathname, $content);

Note: once again the code samples are taken from fossil, have a look at this article to discover what it is.

In order for it to be autoloaded, you'll need to edit your composer.json file:

    "require": {
        "phpunit/phpunit": "~3.7"
    "autoload": {
        "psr-4": {
            "Gnugat\\Fossil": "src/"

To test it, we could run it and check if the file has been created with the given content, but that would be testing Symfony2's Filesystem which happens to be already tested.

So what does our class adds to it? Well it calls Filesystem to check if the file exists, and if not it calls again the Filesystem to create it. We could check if those calls are made using stubs and mocks.

Reminder: stubs are a substitute of an object which forces it to return a given value. When a System Under Test (SUT, the class you're testing) has collaborators (classes used by the SUT, also called dependencies), we can stub them so their behavior is completly controlled.

Reminder: mocks are a substitute of an object which checks if its methods have been called. When a System Under Test (SUT, the class you're testing) has collaborators (classes used by the SUT, also called dependencies), we can mock them to monitor their use.

PHPUnit allows us to create stubs and mocks without having to write a class which extends the colaborator and overwrites its methods:

// File: test/MarkdownFile/DocumentationWriterTest.php

namespace Gnugat\Fossil\Test\MarkdownFile;

use Gnugat\Fossil\MarkdownFile\DocumentationWriter;

class DocumentationWriterTest extends \PHPUnit_Framework_TestCase
    const FIXTURE_ABSOLUTE_PATHNAME = '/tmp/example.txt';
    const FIXTURE_CONTENT = "Hello world\n";

    public function setUp()
        $this->filesystem = $this->getMock('Symfony\\Component\\Filesystem\\Filesystem');
        $this->documentationWriter = new DocumentationWriter($this->filesystem);

    public function testWriteNewFile()
        // Stub

        // Mock

        // Call

In the stub: for every call (expects($this->any())) of the method exists with the parameter self::FIXTURE_ABSOLUTE_PATHNAME, force the colaborator to return true.

In the mock: a unique call must be made (expects($this->once())) of the method dumpFile with the two parameters self::FIXTURE_ABSOLUTE_PATHNAME and self::FIXTURE_CONTENT.

You can run the test using php vendor/bin/phpunit and see that they pass. As an exercise, write a second method in this test suite testDoesNotWriteExistingFile, with a stub returning false and a mock checking that dumpFile is never called (hint).

I advise you to watch extract till you drop, a nice refactoring session by Mathias Verraes: in order to improve his code, he writes tests which give him enough confidence to proceed. He uses PHPUnit with assertions, mocks and stubs, so you can really see their use.

Functional tests

Unit tests are used in order to make sure each unit of code works as expected. But applications aren't just about code, they're also about interactions between these units. This is what functional tests are for: they use the entry point of the system and check the final status code.

To illustrate this, we'll still use PHPUnit: even though there's the word unit in its name, this test framework allows us to write many types of tests.

The Symfony2 web framework is all about the HTTP protocol: it takes a HTTP Request and returns a HTTP Response. It also provides a convenient client which simulates HTTP Requests, allowing us to write easily functional tests:

// File: src/Acme/DemoBundle/Tests/Controller/DemoControllerTest.php

namespace Acme\DemoBundle\Tests\Controller;

use Symfony\Bundle\FrameworkBundle\Test\WebTestCase;
use Symfony\Component\HttpFoundation\Response;

class DemoControllerTest extends WebTestCase
    public function testIndex()
        $client = static::createClient();

        $client->request('GET', '/demo/hello/Fabien');


The assertSame line is an assertion: it compares the expected value (first argument) with the actual one (second one). PHPUnit provides many assertions:

Different approaches exist with assertions, for example jasmine uses the expect method to set the actual value, chained with a matcher like toBe which takes the expected value:

describe("A suite", function() {
  it("contains spec with an expectation", function() {
    var expectedValue = true;
    var actualValue = true;


In this case, assertions are splitted into expectations and matchers.

Back to our sheeps. Here's an example of functional test for a Symfony2 command:

// File: src/Acme/DemoBundle/Tests/Command/GreetCommandTest.php

namespace Acme/DemoBundle/Tests/Command;

use Symfony\Component\Console\Tester\CommandTester;
use Symfony\Bundle\FrameworkBundle\Console\Application;
use Symfony\Bundle\FrameworkBundle\Test\WebTestCase;
use Acme\DemoBundle\Command\GreetCommand;

class ListCommandTest extends WebTestCase
    public function testExecute()
        $kernel = $this->createKernel();

        $application = new Application($kernel);
        $application->add(new GreetCommand());

        $command = $application->find('demo:greet');
        $commandTester = new CommandTester($command);
                'name' => 'Fabien',
                '--yell'  => true,

        $exitSuccess = 0;

        $this->assertSame($exitSuccess, $commandTester->getStatusCode());

Just like with a controller where we check the HTTP Response's status code, in a command we check the exit status code.

User Interfact tests

Up until now, we've been testing that the code worked and that interractions between all those units go well. But what about the thing the user actually sees and interacts with?

The User Interface (UI) turns out to be tricky to test, but not impossible. You can click on buttons, or load pages, or run CLI tasks programmatically and you can inspect the standard output or the HTTP Response's content.

The main problem with this is that you'll tie your tests to the UI, which changes a lot.

We'll see briefly the tools available and then better explain the flaws of such a technique.


Selenium allows you to open a browser and simulates interractions with the page. Technically it means having a Selenium server running in the background, and using a webdriver library in your test to send messages to it. In PHP, you can find those webdrivers:

You can find code samples in the respective documentation of each library, for example here's the doc describing how to click on a button with Alexandre's one.

If you're curious, here's how to use Selenium with another test framework: Codeception.


You don't like the idea of running a server in order to run your tests? You don't want a browser to be openned and you find it too slow?

Then you might want to try a headless website testing solution like CasperJs. It's headless because it won't open a browser to make the tests.

Here's the get started documentation, so you can have a quick look on how to use it.


Simulating a browser is too much for you? Making a curl request and parsing its response would be sufficient for your needs? Then have a look at the Goutte web scrapper.

This one also allows you to click on links.


One Tool to rule them all, One Tool to find them, One Tool to bring them all and in the webness bind them

Mink can use either Goutte or Selenium as a driver to interract with the UI. It's goal is to provide a unified API.

What's the point?

The thing in common with all these tools is that they rely on the HTML rendered in the HTTP Response. In order to check if a message appears, you'll have to crawl the page, find the div using its class or id in a CSS selector, or even worse using its xpath.

When the page will change (and it will) your tests will be broken. There's some best practices out there, mainly making your tests rely on ID's on one hand and making your code and stylesheets rely on classes on the other hand, but in the end it still is a risky business.

Well that's my opinion and this section might be more a rant than an objective description. To counter balance this, here's a nice article on writing reliable locators for Selenium and WebDriver tests.


In PHP, PHPUnit is the most popular test framework and it allows you to write unit, functional and every other kinds of tests.

Unit tests allow you to make sure each functions return the expected output when given a set of fixtures, and functional tests are used to check the status code (think HTTP response, or a command's exit status).

Stubs are used to force a colaborator's method to return a wanted value, and mocks are used to check if a colaborator's method have been called.

Selenium is the most popular tool to test the User Interface, which is done by interracting with the HTML rendered in the HTML's reponse.

I hope this article helped you to see how each kinds of tests are written. The next one will be on Test Driven Development, a practice where tests are written before the code: TDD.

If there's something bothering you in this post, or if you have a question, or if you want to give me your opinion, be sure to do so on Twitter ;) .