Higher-Order Process Engineering
in the context of Active Continuous Quality Control
Johannes Neubauer
Computer Science, Chair for Programming Systems TU Dortmund, Germany
johannes.neubauer@cs.tu-dortmund.de
Abstract:In this talk we present howHigher-Order Process Engineering(HOPE) and Active Continuous Quality Control(ACQC) can be combined to drastically reduce the manual effort of risk-based regression testing. That is, integrating active automata learning, for automatically maintainingtest modelsof asystem under test(SUT), with a rigorous model-driven quality assurance process.
Today’s enterprise software systems are increasingly based on complex stacks of technolo- gies and the integration of heterogeneous third party components. Especially long-running multi-user systems like web applications possess this heterogenity, which makes it partic- ularly difficult if not impossible to control/predict the overall behavior at the system level and therefore calls for advanced system-level testing. In real-life, the situation is even more problematic, because the systems evolve at an increasing pace. Even worse, each of these changes typically requires substantial manual effort in order to have an up-to-date testing harness. These requirements are infeasible in most practical scenarios, both for cost and time. In order to tackle this issue,risk-based testing[FR14] has been introduced.
It focuses on detecting the most critical bugs. This is achieved by test priorization based on the results of a previousrisk analysis.
In this talk we present howHigher-Order Process Engineering(HOPE) [NS13] and the active automata learning basedActive Continuous Quality Control(ACQC) [WNS+13]
approach can be combined to drastically reduce the manual effort of risk-based regression testing. Automata learning [SHM11], sometimes referred to as test-based modeling, pro- vides the benefits of model-based testing, but does not require any a priori models. Our model-driven approach has several dimensions (cf. Fig. 1):
• Symbol Modelsreflect single user actions. They are generic in that they may have multiple input and output parameters. Symbol models are used to define generic learning alphabet symbols based ontest blocks. Test blocks provide a stable ab- straction[WNS+13] of the API of the SUT.
• Alphabet modelsselect, parameterize, and combine generic learning alphabet sym- bols in order to form a learn alphabet, fully instantiating/parameterizing the symbol models. They can easily be tuned according to a given risk profile via specific data- flows.
87
Figure 1: The modeling layers of the HOPE enhanced ACQC for risk-based testing [NWS14]
• Test case modelsrepresent amembership query[SHM11] in the learn process, which are generated fully automatically during the model inference process.
The potential of this approach becomes particularly apparent under a risk-based perspec- tive ofsystem migrationandfunctional evolution. Risk analysts are provided with a mod- eling level where they can build executable alphabet models that typically remain valid during the system lifecycle. The realization of these models is based on our hierarchi- cal higher-order process modeling approach, supporting asimplicity-orientedversion of higher-order process passingwith full-code generation capabilities. In fact, based on the higher-order concepts, services and processes can be selected, modified, constructed and then safely passed as if they were data. Though unlike data they may be plugged into activ- ities and executed (played) dynamically. Thisplug&playapproach allows one to add new services, components, and processes without the need to change the system or interrupt the running processes. This way, we are able to generate alphabet models to executable code and, since the control flow is still accessibleandmutable, the test case models may be created on the fly during learning. The benefit of this rigorous and highly automated model-driven approach is that it supports reuse, separation of concerns, and automation.
References
[FR14] Michael Felderer and Rudolf Ramler. A multiple case study on risk-based testing in industry.International Journal on Software Tools for Technology Transfer, pages 1–17, 2014.
[NS13] Johannes Neubauer and Bernhard Steffen. Plug-and-Play Higher-Order Process Inte- gration.Computer, 46(11):56–62, 2013.
[NWS14] Johannes Neubauer, Stephan Windm¨uller, and Bernhard Steffen. Risk-Based Testing via Active Continuous Quality Control. International Journal on Software Tools for Technology Transfer, 16(5):569–591, 2014.
[SHM11] Bernhard Steffen, Falk Howar, and Maik Merten. Introduction to Active Automata Learning from a Practical Perspective. InSFM, pages 256–296, 2011.
[WNS+13] Stephan Windm¨uller, Johannes Neubauer, Bernhard Steffen, Falk Howar, and Oliver Bauer. Active continuous quality control. InProceedings of the 16th International ACM Sigsoft symposium on Component-based software engineering, CBSE ’13, pages 111–120, New York, NY, USA, 2013. ACM.
88
