From out-place transformation evolution to in-place model patching

Abstract

Model transformation is a key technique to automate software engineering tasks. Like any other software, transformations are not resilient to change. As changes to transformations can invalidate previously produced models, these changes need to be reflected on existing models. Currently, revised out-place transformations are re-executed entirely to achieve this co-evolution task. However, this induces an unnecessary overhead, particularly when computation-intensive transformations are marginally revised, and if existing models have undergone updates prior the re-execution, these updates get discarded in the newly produced models.To overcome this co-evolution challenge, our idea is to infer from evolved out-place transformations patch transformations that propagate changes to existing models by re-executing solely the affected parts based on an in-place execution strategy. Thereby, existing models are only updated by a patch instead of newly produced. In this paper, we present the conceptual foundation of our approach and report on its evaluation in a real-world case study.

Towards systematic mutations for and with ATL model transformations

“…Our framework also provides means for efficiently executing mutated transformations by exploiting incremental transformation execution instead of forcing a complete re-execution of the transformation. Thereby, only the transformation rules affected by the mutation are executed, which generally results in an improved runtime performance of the mutated transformations, as shown in our previous work [20].…”

“…However, the re-execution of the mutated transformations in order to obtain the evolved models induces an unnecessary overhead, particularly when computation-intensive transformations are marginally revised. To tackle the challenge of co-evolving output models with mutations in transformations, we have proposed to infer inplace patch transformations from evolved out-place transformations for existing output models [20]. Such approach is highly complementary to the one we have described in this paper, and can be used for executing the obtained mutated transformations for obtaining the new output models.…”

“…For this, we will consider the application of single mutation operators in order to produce mutants, as well as the combination of several mutation operators for producing a mutant. We also want to apply the approach described in this paper and presented in our previous work [20] in order to optimize the time to produce mutants and to execute tests.…”

“…Our contribution is threefold: (i) we have derived a systematic set of mutation operators dedicated to ATL by proposing a general language-centric synthesis approach for mutation operators, (ii) we have automated the generation of model transformation mutants by realizing a framework that exploits the concept of Higher-Order Transformations (HOTs) [19], and (iii) we have studied the possibility to integrate into our framework means to reduce the computational costs of executing model transformation mutants by relying on techniques for incremental model transformation execution, which we have proposed in previous work [20].…”

Towards language-to-language transformation

“…4. The resulting tool provides functionalities commonly known from (model-based) IDEs: a modeling canvas (1), a palette from which available elements can be drag&dropped to the canvas (2), a property view (3) and a project explorer (4). The generated tool is easily extensible, e.g., by custom validation rules which are shown in the 'Model Validation' view (5).…”