7.1 Desired Features of Transformations

In section 2.3 we defined a transformation as the generation of a target model from a source model. This means that transformations are purely processes. The process is described by a transformation definition , which consists of a number of transformation rules , and is executed by a transformation tool . In an MDA approach there are a number of features of the transformation process that are very desirable. We name them in order of importance:

1. Tunability, which means that although the general rule has been given in the transformation definition, an application of that rule can be tuned ; for example, when transforming a UML String to a VARCHAR in an entity-relationship model, you might want the length of the VARCHAR to differ for each occurrence of a UML String.

2. Traceability, which means that one can trace an element in the target model back to the element(s) in the source model from which it is generated.

3. Incremental consistency, which means that when target-specific information has been added to the target model and it is regenerated, the extra information persists.

4. Bidirectionality, which means that a transformation can be applied not only from source to target, but also back from target to source.

Each of these features puts demands on the transformations. In the next sections, we further investigate the desired features. Section 7.6 describes the implications of the features on transformations.