On Improving Run-time Checking in Dynamic Languages

In order to detect incorrect program behaviors, a number of approaches have been proposed, which include a combination of language-level constructs (procedure-level annotations such as assertions/contracts, gradual types, etc.) and associated tools (such as static code analyzers and run-time verification frameworks). However, it is often the case that these constructs and tools are not used to their full extent in practice due to a number of limitations such as excessive run-time overhead and/or limited expressiveness. The issue is especially prominent in the context of dynamic languages without an underlying strong type system, such as Prolog. In our work we propose several practical solutions for minimizing the run-time overhead associated with assertion-based verification while keeping the correctness guarantees provided by run-time checks. We present the solutions in the context of the Ciao system, where a combination of an abstract interpretation-based static analyzer and run-time verification framework is available, although our proposals can be straightforwardly adapted to any other similar system