Code Museums as Functional Tests for Static Analyses

Growing your software guided by tests [2] has the benefit of thoroughly tested implementations of the right functionality. If you are developing static analyses your “test data” consist of code that has a few lines to a few classes. How can tests based on this “data” be kept expressive and maintainable? After exploring a variety of different other approaches we decided to embed the expectations into the data instead of embedding the data into test cases. Figure 1 presents an overview of our tool set. The functional tests are within Java code, the tested static analyses are implemented as logic programs.