Uniform Random Expressions Lack Expressivity

Abstract : In this article, we question the relevance of uniform random models for algorithms that use expressions as inputs. Using a general framework to describe expressions, we prove that if there is a subexpression that is absorbing for a given operator, then, after repeatedly applying the induced simplification to a uniform random expression of size n, we obtain an equivalent expression of constant expected size. This proves that uniform random expressions lack expressivity, as soon as there is an absorbing pattern. For instance, (a + b) is absorbing for the union for regular expressions on {a, b}, hence random regular expressions can be drastically reduced using the induced simplification.