Repairing Regular Expressions for Extraction

While synthesizing and repairing regular expressions (regexes) based on Programming-by-Examples (PBE) methods have seen rapid progress in recent years, all existing works only support or regexes for membership testing, the extraction is still an open problem. This paper fills void by proposing first PBE-based method extraction. Our work supports that real-world extensions such as backreferences lookarounds. The significantly affect synthesis repair In fact, we show there are unsolvable instances of problem if synthesized not allowed to use extensions, i.e., no regex without correctly classify given set examples, whereas every instance solvable allowed. stark contrast case where guaranteed a solution expressible pure extensions. main contribution algorithm solve builds enumerative search algorithms with SMT constraint solving. However, significant needed because constraints previous non-deterministic semantics regexes. Non-deterministic sound but extraction, which substrings extracted depends deterministic behavior actual engines. To address issue, propose new generation respects For this, define novel formal engine big-step operational semantics, it basis design method. key idea simulate determinism consider continuations matching them disambiguation. We also two space pruning techniques called approximation-by-pure-regex approximation-by-backreferences make information examples. implemented tool R3 (Repairing Regex extRaction) evaluated 50 contain evaluation shows effectiveness our substantially prune space.