On the VGTech blog today Christer Edvartsen looks at a method of generating the code coverage of Behat tests using a special bit of code and the $_SERVER superglobal to detect what parts of the code is being executed.
Yes, I know, it sounds silly, but bear with me. The nature of acceptance tests is not really to tests units of code, but to assure that the behavior of your application meets a certain set of criteria (Behat Scenarios). When your applications grow over time, code coverage can be a nice tool to help you pinpoint where you need to add more tests. In a perfect world tests are added while implementing new features so that your applications are always fully tested, but that isn't always as easy as it sounds.
The first chunk of code is something that would go in the router of your application, capturing the "collect coverage" and "enable coverage" input variables. The other part of the code creates a unique ID in the set up and uses it in a call (via Guzzle) to the server to start the tracking. In the tear down method, it sends the call to get the coverage results and process them through the PHP_CodeCoverage tool from Sebastian Bergmann.