Got API's but no time to put automated tests around it? Already have automated tests but are hard to maintain? Here is an effective automated solution that will prevent regression and maintain itself on demand.
Major:
Minor:
Patch:
composer require genesis/behat-api-spec
behat.yml file
default:
suites:
default:
contexts:
- Genesis\BehatApiSpec\Context\ApiSpecContext
extensions:
Genesis\BehatApiSpec\Extension:
baseUrl: <Your API Url>
specMappings:
endpoint: <Namespace to folder to autoload - leading backslash>
You can generate an endpoint file using the --endpoint-generate option. Simply run the following command for an interactive shell:
./vendor/bin/behat --endpoint-generate
Manually create an endpoint file
<?php
namespace ...;
use Genesis\BehatApiSpec\Contracts\Endpoint;
class User implements Endpoint
{
public static function getEndpoint(): string
{
return '/users';
}
public static function getRequestHeaders(): array
{
return [
'accept-language' => 'en',
'accept' => 'text/html',
];
}
}
Add step definition to feature file
Scenario: 200 user response
When I make a POST request to the "User" endpoint
Then I expect a 500 "User" response
And the response should match the snapshot
And I expect the following content in the JSON response:
| key1.subkey1 | value1 |
| key1.subkey2 | value2 |
The When I make a POST request to "User" endpoint
will initially auto scaffold schema using the response and insert it into the endpoint file you've declared above. On subsequent calls this schema will be used to validate the response, providing protection against regression. A sample schema can be as follows for the response of a GET request with 200 response {"success": false, "error": "Something went wrong."}
:
<?php
namespace ...;
use Genesis\BehatApiSpec\Contracts\Endpoint;
class User implements Endpoint
{
...
public function getResponseSchema(): array
{
'GET' => [
500 => [
'headers' => [
'Host' => [
'value' => 'localhost:8090',
'type' => self::TYPE_STRING,
],
'Connection' => [
'value' => 'close',
'type' => self::TYPE_STRING,
],
'X-Powered-By' => [
'value' => 'PHP/7.2.26-1+ubuntu18.04.1+deb.sury.org+1',
'type' => self::TYPE_STRING,
],
'content-type' => [
'value' => 'application/json',
'type' => self::TYPE_STRING,
],
],
'body' => [
'success' => [
'type' => self::TYPE_BOOLEAN,
'optional' => false,
],
'error' => [
'type' => self::TYPE_STRING,
'optional' => false,
'pattern' => null,
],
],
]
]
}
}
Adjust accordingly.
Following on from this the And the response should match the snapshot
will generate a snapshot automatically storing the response against the scenario title. This will be stored in the same directory as the test. This file should be committed with the code to allow it to be peer reviewed. Upon subsequent requests, the response will be matched with this snapshot, any difference will generate a failure. You have either the option to update the snapshot automatically using the --update-snapshots
, -u
flag or fix the issue in the API. Any out of date snapshots will be identified and updated with the flag appropriately. Example snapshot:
<?php return [
'500 user response' =>
'{"success":false,"error":"Something went wrong."}',
];
All requests go call on the PlaceHolderService::resolveInString method with the body and url to replace any placeholders (format - {{placeholder_name}}
) you may have set using the default preRequestCallable hook which is overridable (See hooks section). To add placeholders, you can use the PlaceholderService like so:
public function ...
{
$value = ...;
PlaceholderService::add($name, $value);
PlaceHolderService::getValue($name); // returns $value;
}
Placeholders are reset after every scenario to prevent test session bleed. Example usage in feature file:
Scenario: 200 user response
When I make a POST request to the "User" endpoint with body:
"""
{"status": "{{status_failed}}"}
"""
In the above example if you've set PlaceHolderService::add('status_failure', -1)
then expect {"status": "-1"}
to be sent as the body. Note values have to scalar to be part of the body.
You can set the version of the API to be used from the feature files or by creating a new endpoint file. To set it from the feature file:
...
Given I use version "1" of the API
...
This will allow you to retrieve the version set through the ApiSpecContext::getVersion()
method in any file. For example setting it in the Endpoint getRequestHeaders method. The method also accepts a default API version if none is set. The version is also available as a placeholder {{API_VERSION}}
placeholder.
Pre request and post request hooks can be configured per context configuration in the behat.yml file like so:
#behat.yml file
default:
suites:
default:
contexts:
- Genesis\BehatApiSpec\Context\ApiSpecContext:
preRequestCallable: 'MyClass::preRequestStaticCallable'
postRequestCallable: 'MyClass::postRequestStaticCallable'
If you use the step definition When I make a POST request to "User" endpoint
to send requests to the API, you can use the --sample-request=<format>
flag to generate sample requests to execute quickly through the command line. An example would be:
vendor/bin/behat --sample-request=curl
Scenario: 200 user response
Given I set the following headers:
| content-type | application/json |
| accept | en |
When I make a GET request to "User" endpoint
│ curl -X GET --header 'content-type: application/json' --header 'accept: text/html' --header 'accept-language: en' 'http://localhost:8090/index.php/users'
Note the curl command generated below the step definition.
More step definitions are provided as part of the context file for validation of the API response. Find out using vendor/bin/behat -dl
.