Is there a simple generalization of Boyer Moore to regular expressions? Can something be said about the optimality of string searching, for example by assuming a probability distribution of input texts?