Symbolic String Transformations with Regular Lookahead and Rollback

Implementing string transformation routines, such as encoders, decoders, and sanitizers, correctly and efficiently is a difficult and error prone task. Such routines are often used in security critical settings, process large amounts of data, and must work efficiently and correctly. We introduce a new declarative language called Bex that builds on elements of regular expressions, symbolic automata and transducers, and enables a compilation scheme into C, C# or JavaScript that avoids many of the potential sources of errors that arise when such routines are implemented directly. The approach allows correctness analysis using symbolic automata theory that is not possible at the level of the generated code. Moreover, the case studies show that the generated code consistently outperforms hand-optimized code.