Test-data generation for Xtext

We describe a method and a corresponding tool for grammarbased test-data generation (GBTG). The basic generation principle is to enumerate test data based on grammatical choices. However, generation is broken down into two phases to deal with context-sensitive properties in an efficient and convenient manner. The first phase enumerates test data (i.e., parse trees) with placeholders. The second phase instantiates the placeholders through post-processors. A DSL for grammar transformation is used to customize a given grammar, meant for parsing, to be more suitable for test-data generation. Post-processors are derived from a corresponding object-oriented framework. The actual tool, Xtextgen, extends the Xtext technology for language development.