Schrödinger Testing Framework

Last modified 12 Mar 2021 10:22 +01:00

Schrödinger is a testing framework designed to test midPoint user interface in a user-friendly and maintainable way. The idea is to hide the details of GUI look and feel design from the tester. The tester should focus on the presentation and application logic, rather than concern themselves with specific location of particular button on the screen.

Schrödinger provides fluent API to build the tests. Code of a Schrödinger test is meant to follow the path that an ordinary user would follow in midPoint GUI:

    UserPage user = basicPage.newUser();
                .addAttributeValue("name", "jdoe007")
                .addAttributeValue(UserType.F_GIVEN_NAME, "John")
                .addAttributeValue(UserType.F_FAMILY_NAME, "Doe")

Schrödinger is built to support two basic modes of operation:

  • Development tests. The tests start from empty midPoint. The test initializes the test and environment and runs test cases. The environment is torn down after the test. The goal is to have repeatable testing process that can be fully automated. Development tests are usually executed automatically as part of continuous integration environment.

  • Deployment tests. The test starts with initialized, configured and populated midPoint instance. The test runs on real data, usually in customer testing environment. Test cases start with an initialized environment, and the environment stays operational after the test. The goal is to semi-automate the tests with real configuration and real resources. Deployment tests are usually started manually by a deployment engineer.

Schrödinger is developed by Evolveum. It is designed specifically to test midPoint. It is not directly reusable for other software. However, the concepts of Schrödinger may be applicable for other testing framework.

Why Schrödinger?
The name "Schrödinger" is based on Erwin Schrödinger, an Austrian-Irish physicist. Erwin Schrödinger is famous, among other works, for "Schrödinger’s cat" thought-experiment, in which a cat in a box may be simultaneously dead and alive, util the box is opened. Similarly to Schrödinger’s cat, the system is both working and not working until it is tested.


Source code of the framework are in tools/schrodinger directory of midPoint source code. The tests are in testing/schrodingertest directory.

The proper name of the platform is "Schrödinger". However, in the 7bit ASCII world we use its non-diacritic version: schrodinger. The German transliteration (Schroedinger) is not used.

Schrödinger is built on top of Selenide framework. Schrödinger has a small server-side part, that adds special markers to HTML elements. The markers make it easier for the tests to recognize important GUI elements, such as tabs, form fields and buttons. The markers remain the same, even if the GUI look and feel changes. This helps to make the tests maintainable.

Majority of Schrödinger code is on client side. There are libraries that hide all the details about the markers and HTML structure from the tester. Obviously, there need to be a class or a method for every type of GUI element that is used in midPoint user interface. That makes maintenance of Schrödinger a bit hard, but it makes maintenance of tests much easier. We have only one framework to maintain, but there are a lot of tests to write, therefore the benefit is quite clear.

Schrödinger expects that midPoint runs as a separate process. This allows use of Schrödinger for deployment tests. For development testing, the test driver must start up clean midPoint instance.

The markers in HTML elements that Schrödinger generates are not supposed to be globally-unique. E.g. a marker for name input field may appear on a page several times, if it is used in several GUI components. All the fields will have the same marker. However, the enclosing element that represents the GUI component should provide sufficient context information for the test to reliably locate the correct input field. As of today, Schrödinger API should be prepared for this, however, the implementation may not be complete.

Schrödinger server-side code is not enabled by default, to avoid polluting the HTML code in production. The server side part of Schrödinger needs to be explicitly enabled.

Current State

Schrödinger is still in development. It is not complete yet. It is not officially "stable". There may be some changes to the API, or to any other part of the system.

Despite that, Schrödinger may be still useful for testing, even for midPoint deployments. However, you should be prepared to maintain the tests in case the framework changes. Of course, contributions to the framework or the tests is more than welcome.


Our plan is to maintain Schrödinger as the primary GUI testing framework for midPoint. It would be also a recommended way for semi-automated testing of midPoint deployments, once the framework is production-ready.

Planned Improvements

We have some specific plans for short-term and mid-term development of Schrödinger:

  • Improve test initialization, especially for development tests. This will most likely involve integration of REST client for REST-based initialization.

  • Improve handling of HTML element markers, e.g. better support for repeated fields.

  • Performance testing involving Schrödinger tests, most likely involving hybrid load/functionality tests.

  • Better coverage of GUI functionality.

  • Improve Schrödinger documentation.

See design notes for the Schrödinger development in midScale project.

Long-Term Future

We have an ambition to extend/integrate Schrödinger with end-to-end testing capabilities. We would like to involve real-world resources in Schrödinger tests, such as databases, CSV files, directory servers and so on. For that to happen, we will need an efficient way to set up, configure and prepare the resources, as well as assert their state in the tests.

The result should be a complete "midCredible" end-to-end testing platform.

However, there are no specific plans or dates at the moment.