Depends what you mean by "reasonable" I'm guessing. Manually encoding a semantic analysis would have a very high overhead, and you might be able to do it via dependent types instead (not sure though, they tend to hurt my brain when I try to actually understand them), but I don't think they're considered practical (yet).