The evaluation model of most of these parser combinator libraries is simple enough that you can just think directly about the call tree. It's difficult to get "surprised" like you can with PCREs. E.g. (x+x+)+y as a regex may blow up, but "many1 (many1 x >> many1 x) >> y" will just evaluate forwards a single time, greedily, and fail, at least with any haskell parser combinator library I can think of.