Existing support for regular expressions in automated test generation or verification tools is lacking. Common aspects of regular expression engines found in mainstream programming languages, such as backreferences or greedy matching, are commonly ignored or imprecisely approximated, leading to poor test coverage or failed proofs. In this paper, we present the first complete strategy to faithfully reason about regular expressions in the context of symbolic execution, focusing on the operators found in JavaScript. We model regular expression operations using string constraints and classical regular expressions and use a refinement scheme to address the problem of matching precedence and greediness. Our survey of over 400,000 JavaScript packages from the NPM software repository shows that one fifth make use of complex regular expressions features. We implemented our model in a dynamic symbolic execution engine for JavaScript and evaluated it on over 1,000 Node.js packages containing regular expressions, demonstrating that the strategy is effective and can increase line coverage of programs by up to 30%

This arXiv version (v4) contains fixes for some typographical errors of the PLDI'19 version (the numbering of indices in Section 4.1 and the example in Section 4.3)