Signature inference for functional property discovery
Property discovery is the process of discovering properties of code via
automated testing. It has the potential to be a great tool for a practical
approach to software correctness. Unfortunately current methods remain
infeasible because of the immense search space. We contribute a new
approach to taming the complexity of a state-of-the-art generate and test
algorithm. Our approach runs this algorithm several times with carefully
chosen inputs of smaller size: signature inference. We implement this
approach in a new tool called EasySpec. The results suggest that this
approach is several orders of magnitude faster, fast enough to be practical,
and produces similar results.
Outline/Structure of the Talk
1. The problem with testing
2. Quickspec as a failed solution
3. Signature inference as the solution
Learning Outcome
Participants will be familiar with the most recent research in property discovery
Target Audience
Anyone interested in practical correctness
Prerequisites for Attendees
Property testing
Links
https://cs-syd.eu/thesis