Buy

You bet your sweet App

You now know all the important things about Behat and Mink and how they work together. In this chapter, we’ll talk about some specific challenges that you’ll likely face when using Behat in your application, like how to handle data fixtures.

I’ve created a miniature application using Silex for us to play with. Don’t worry if you’re not using Silex - we’ll keep the instructions here as generic, but useful as possible. You’ll still have a bit of homework to do for your specific application, but we’ll guide you.

Note

The Silex application we’re using here is available in this screencast’s code download. (FTW!)

Bootstrapping your Application in FeatureContext

To test our application, we’ll make real HTTP requests as an outsider: surfing to pages, filling out forms, and checking how everything looks afterwards. To do this, we don’t need access to our project’s code, nor do we even need to be on the same server!

But when you are on the same server, it can be very useful to bootstrap your application inside FeatureContext. This allows you access to all the functions and tools that you use to do things like connect to the database, clear cache, or anything else. As we’ll talk about in a second, having access to your normal tools might make clearing and preparing data pretty easy.

Exactly how you do this will be different for every application and framework, but it will follow the same pattern. Usually it involves requiring some core bootstrap file and possibly calling some sort of initialize function.

Tip

If you’re using Symfony2, install the Symfony2Extension to bootstrap Symfony for you and give you access to the Kernel. The container is then available as $kernel->getContainer().

Since we only need to bootstrap our application once per test suite, we can take advantage of the beforeSuite hook. Your setup - and even the hook to use - may be different:

private static $app;

/**
 * @BeforeSuite
 */
static public function bootstrapSilex()
{
    if (!self::$app) {
        self::$app = require __DIR__.'/../../app/bootstrap.php';
    }

    return self::$app;
}

For Silex, the $app object is the key to any functionality we may need. In your app, you may need a different variable, or you might just need to statically call a function that magically gives you access to all of your classes and functions. The best way to find out how your code should look is to google to see if others have bootstrapped your framework for testing, or to look at the code in your front controller.

In the next sections, you’ll see how having access to the built-in functions

Preparing Data

One of the most complex issues with testing is dealing with and controlling the data you test with. If you’re testing against an external server, you may not be able to control your data at all. In this case, you’ll have to be a bit more careful and clever about how you write your tests. Fortunately, since we’re on the same server as our application, we’ll have full control over our data.

Let’s start with a scenario I already prepared, which tests the list page of a product admin section.

Feature: Product admin
  In order to manage the content on my site
  As an admin
  I need to be able to add, edit and delete products

  Scenario: Seeing a list of existing products
    Given I am logged in as an admin
    And there are 5 products
    And I am on "/admin"
    When I follow "Products"
    Then I should see 5 rows in the table

In order for this scenario to work, we need to guarantee that there is an admin user in the database and we need the ability to add 5 products. If you’ve bootstrapped your application in FeatureContext, this should be possible. Execute bin/behat so that the missing definitions are added to our FeatureContext.

I’ll fill in the code that my system needs to create my admin user. Notice that we didn’t say Given there is an admin user called "admin" in our scenario. We don’t really care about that detail, so we skip it and just make sure the user exists inside this definition. Next, use metasteps to actually login with this user. The exact wording will vary for your app:

/**
 * @Given /^I am logged in as an admin$/
 */
public function iAmLoggedInAsAnAdmin()
{
    self::$app['user_repository']->createAdminUser(
        'admin',
        'adminpass'
    );

    return array(
        new Given('I am on "/login"'),
        new Given('I fill in "Username" with "admin"'),
        new Given('I fill in "Password" with "adminpass"'),
        new Given('I press "Login"'),
    );
}

I’ll also add the code to insert 5 products. Notice that we’re not testing the actual creation of products in this scenario. We may do that later in another scenario, but for now we want to insert them as quickly as possible to test that the user sees them:

/**
 * @Given /^there are (\d+) products$/
 */
public function thereAreProducts($num)
{
    for ($i = 0; $i < $num; $i++) {
        self::$app['product_repository']->createProduct(
            'Sickle-shaped Claw'.$i,
            9.99+$i
        );
    }
}

Finally, write some custom Mink code for the final step. This step is purposefully generic, so that we can re-use it on other pages. Let’s use the find method to find a single table, then use it again to return an array of all of the tr elements. Use the PHPUnit assert functions to make sure we have the right number of rows:

/**
 * @Then /^I should see (\d+) rows in the table$/
 */
public function iShouldSeeRowsInTheTable($rows)
{
    $table = $this->getPage()->find('css', '.main-content table');
    assertNotNull($table, 'Cannot find a table!');

    assertCount(intval($rows), $table->findAll('css', 'tbody tr'));
}

Great! Execute Behat. It passes! There’s a lot going on behind the scenes, but the actual scenario is described with simple language.

Use Data Fixtures?

Another approach to loading data is using some sort of fixture, which inserts a whole set of default data. This may sound easy, but it’s not a great approach for two reasons. First, loading the extra data makes your tests run a bit slower. Second - and more importantly - loading a set of data can make your scenarios less readable. In our scenario, we’re being very specific about what data we have. If we loaded fixtures beforehand that contained 5 products, we might remove the And there are 5 products line. But now our scenario is a bit confusing - why are we expecting 5 products? Where did these products come from?

For those reasons, do your best to avoid loading fixtures. The one exception might be if you have lookup tables that contain data that never changes, and is important to your application. An example might be a table called product_status, with entries like Published, Draft, Archived. Since this data is static, it just needs to be there, so loading it before your tests is probably a good idea.

Cleaning out Data

Run the test again. This time, it fails spectacularly. When we try to insert a second user with the same username, a unique constraint in the database fails. As important as it is to add the data you need in a scenario, you also need to clean out data. At the beginning of each scenario, you should be able to assume that there is no data in the database. This prevents us from needing to say things like Given there are no users before saying Given I am logged in as an admin. The fact that we need to empty the user table before inserting a user is an implementation detail - not part of the feature’s description.

Exactly how you handle this depends on your application, but it almost always involves another BeforeScenario hook. Create a new function called clearData and tag it with BeforeScenario. In here, your goal is to empty the data in the database. In reality, you can do this, or just empty the tables that you know should be cleared before each test. For now, let’s clear the user and product tables:

/**
 * @BeforeScenario
 */
public function clearData()
{
    self::$app['user_repository']->emptyTable();
    self::$app['product_repository']->emptyTable();
}

We’re using BeforeScenario here because each scenario should be independent of every other scenario. In other words, data built in one scenario shouldn’t be used in the next. By clearing out the data before each one, we’re helping to guarantee that independence.

Re-run the test to see that things pass once again. Regardless of how you clear it, make sure to always think about what data you have and what you’re adding so that you’re testing against the exact data you want.

Using the Current User and other Objects

Sometimes you’ll write a scenario where you refer to something that was just created in the background. For example, what if we want the five new articles to be authored by us, the admin user? Lets change our scenario to reflect this:

Scenario: Seeing a list of existing products
  Given I am logged in as an admin
  And I author 5 products
  # ...

Run behat to generate this new step definition. We already know how to create products, but how can we set our user as the author? The trick is to set the current user on a private property when we login:

private $currentUser;

// ...

/**
 * @Given /^I am logged in as an admin$/
 */
public function iAmLoggedInAsAnAdmin()
{
    $this->currentUser = self::$app['user_repository']->createAdminUser(
        'admin',
        'adminpass'
    );

    return array(
        // ...
    );
}

/**
 * @Given /^I author (\d+) products$/
 */
public function iAuthorProducts($num)
{
    for ($i = 0; $i < $num; $i++) {
        $product = self::$app['product_repository']->createProduct(
            'Sickle-shaped Claw'.$i,
            9.99+$i
        );

        $product->author = $this->currentUser;
        self::$app['product_repository']->update($product);
    }
}

Once we’ve done this, we can use it in this definition or any other in the future. This is one of the pro tips to using Behat, and we saw it once before during the ls scenarios. If you need access to something between steps, just store it on a property. Scenarios should be completely independent of each other, but the steps in a scenario can be totally dependent.

Customizing behat.yml on each Machine

Finally, let’s cover one more obstacle to using Behat in your project. The behat.yml file holds the base_url configuration, and based on our virtualhost configuration, this value may be different on your machine versus my machine. But how can we commit this to our repository without everyone needing to modify this file and try not to commit those changes?

There are a few ways to handle this, but one of them is with the imports configuration. In behat.yml, import a separate file called behat.local.yml and then move the base_url into it.

# behat.yml
default:
  extensions:
    Behat\MinkExtension\Extension:
      goutte:    ~
      selenium2: ~

imports:
  - behat.local.yml
# behat.local.yml
default:
  extensions:
    Behat\MinkExtension\Extension:
      base_url:  http://store.l

The point of this is that we’ll commit behat.yml, but add behat.local.yml to our .gitignore file. When someone sets up the project for the first time, they’ll just create this file and customize it however they need.

To make this easier, copy this file to behat.local.yml.dist.

cp behat.local.yml behat.local.yml.dist

This new file has no functional purpose, but you can use it to create the local file when you setup the project.

Leave a comment!