As a software developer, I write code (a lot !).
When working with the UI5 framework, the recommendation for testing is to use either QUnit or OPA.
These two frameworks are used to cover the three lower stacks of the test pyramid : unit, component and integration testing. This is usually the location where the testing is focused. Everything is already written down in the book Testing SAPUI5 Applications.
This article presents the point of view of the developer working on the UI side. To be complete and cover more stages of the test pyramid, one also needs to consider end-to-end, application and scenario testing. To achieve these tests, UIVeri5 can be used.
The beauty of QUnit and OPA is that they only require a browser to execute the tests. Hence the minimal work environment of an UI5 developer is composed of an editor, a browser and a web server. Regarding the last point, UI5 provides command line tools to serve the application.
OPA tests results
In case you missed it, and because the rest of the article assumes you already know it, I documented an alternate solution on how REserve could be used to serve an Open UI5 application.
Once the code is tested, linted and reviewed, it goes to the continuous integration pipeline.
In this context, the game is different.
Indeed, to execute the tests, the pipeline needs a tooling that is capable of serving the application, opening a browser, piloting the test execution and consolidating the results. Furthermore, this process is usually the place where the code coverage is measured.
Last but not least, a report is generated to let the developer know the execution status. Actually, the report is crucial when the tests fail : the developer needs a maximum of information to fix the problems.
By default, the UI5 tooling proposes a solution based on karma. In the current implementation, all the tests are executed sequentially in the same browser window.
UI5 karma runner in action
Unfortunately, on very big projects this model appears to not scale properly. Indeed, the combination of iframes in the OPA tests and the extra memory needed to collect the code coverage generates leaks that accumulate over time.
This may lead the browser to crash.
We are talking of complex projects with a huge test suite that takes more than 45 minutes to execute.
So I wanted to investigate a different approach and enable the tests execution in an environment where :
- By splitting the QUnit tests and the different OPA journeys, the tests would be no more sequential but parallel.
- By using one browser window per test, it limits the consequences of memory leaks and reduces the risk of crash.
A new test runner
In this serie of articles, we will detail step by step the solution that was prototyped to run all the tests contained in an UI5 application and also take coverage measurement.
To properly generate the report, we have to make sure that the runner collects all the information about the different tests (such as the test names, success, failure and execution) as well as individual coverage information.
This little experiment actually demonstrates some interesting features of REserve as it shows how complex problems can be solved with very little code.
Disclaimer : these articles provide details about a proof of concept. The code is far from being perfect and it can surely be improved.
- Building a platform : in this first article, we setup the runner by building a configurable platform that serves the web application and offers basic services.
- Probing tests : in this second article, we fetch the list of test pages by triggering a specific URL that references all the tests to execute. This requires the use of script substitution as well as offering an endpoint to receive the collected tests.
- Executing tests : in this third article, the runner is improved to enable the execution of the tests (qUnit and OPA). The web server is modified to inject hooking scripts and new endpoints are provided to receive the tests results. Also, a basic execution queue is implemented so that we can control the number of instances that are executed simultaneously.
- Measuring code coverage : in this last article, we explain how nyc is used to instrument the sources and the runner is modified to handle code coverage. The web server switches between instrumented sources and the original ones (in case one does not want to measure the coverage of specific files). Because of the way OPA tests are designed (and the use of IFrames), the instrumented files are altered on the fly to update their scope. Once every individual coverage information is extracted, nyc is called again to merge the coverage and generate a report.