SEMI-AUTOMATED TEST MODEL GENERATION

Appropriate test models that can satisfy complex constraints are required for testing
model management programs in order to build confidence in their correctness. Models have inherently complex structures and are often required to satisfy non-trivial constraints which makes them time consuming, labour intensive and error prone to construct manually. Automated capabilities are therefore required, however, existing fully-automated model generation tools cannot generate models that satisfy arbitrarily complex constraints. This thesis addresses this problem by proposing a semi-automated approach towards the generation of such models. A new framework named Epsilon Model Generator (EMG) that implements this approach is presented. The framework supports the development of model generators that can produce random and reproducible test models that satisfy complex constraints.