Regular expression matching asks whether a given regular expression of length $m$ matches a given string of length $n$, and its time complexity is fairly well understood. Meanwhile, regular expressions have been extended to support various extensions for both theoretical and practical reasons, which can substantially change the time complexity of the matching problem. In this work, we consider four well-known extensions to regular expressions called intersection, squaring, complement and backreference, and we prove a number of novel conditional time complexity lower bounds for the regular expression matching problems with these extensions under the Orthogonal Vectors Conjecture, the $k$-Orthogonal Vectors Conjecture or the $k$-Clique hypotheses. Our lower bounds significantly narrow the gaps to the known upper bounds, with some even being (conditionally) optimal in that the dependence on $n$ cannot be polynomially reduced while keeping the dependence on $m$ polynomial (or even elementary, for complement). Additionally, to obtain our results for intersection, squaring and complement in a unified manner, we introduce novel problems called Dyck Selection, Orthogonal Dyck Selection and Generalized Dyck Selection. They are arguably natural formal language and stringology problems that are based on Greibach's hardest context-free language and capture the essence of why the regular expression matching problems with these extensions are hard.

翻译：暂无翻译