Model-driven engineering is a software engineering technique which relies heavily on the use of models. They are not just used as documentation, but actually define the system. Transformations on the models are described using model transformation languages. There are a lot of different model transformation languages available, all having a different approach to model transformation.

We show that these languages are actually not that different at all. Based on the analysis of a varied selection of transformation languages we define a small number of primitive transformation operation that can be used to describe model transformations written in any transformation language. We define our own primitive transformation language using only these operations and verify our analysis by implementing a few well-known transformation languages in our own language, including a language not considered in the initial analysis. We show that the execution of model transformations with our interpreter is on par with their original interpreter.