We describe a new tool, named REgen, that generates regular expressions (RE) to be used as test cases, and that generates also synthetic benchmarks for exercising and measuring the performance of RE-based software libraries and applications. Each group of REs is randomly generated and satisfies a user-specified set of constraints, such as length, nesting depth, operator arity, repetition depth, and syntax tree balancing. In addition to such parameters, other features are chosen by the tool. An RE group may include REs that are ambiguous, or that define the same regular language but differ with respect to their syntactic structure. A benchmark is a collection of RE groups that have a user-specified numerosity and distribution, together with a representative sample of texts for each RE in the collection. We present two generation algorithms for RE groups and for benchmarks. Experimental results are reported for a large benchmark we used to compare the performance of different RE parsing algorithms. The tool REgen and the RE benchmark are publicly available and fill a gap in supporting tools for the development and evaluation of RE applications.
|Titolo:||A Benchmark Production Tool for Regular Expressions|
|Data di pubblicazione:||2019|
|Appare nelle tipologie:||04.1 Contributo in Atti di convegno|
File in questo prodotto:
|banchmark_production_tool_for_regexp.pdf||Articolo principale||Publisher’s version||Accesso riservato|