The index function in Data.ByteString throws an exception if the given index is negative or too large. This makes it cumbersome to deal with possibly-invalid indices: you have to either catch the exception, which doesn't compose well with other Haskell machinery, or replicate the range check in your own code. It would be useful to have a non-throwing version of index of type ByteString -> Int -> Maybe Word8, which would enable more idiomatic handling of bad indices. This would be in line with the existing bytestring functions uncons and unsnoc, which return a Maybe value to deal with empty byte strings.

The same goes for Data.ByteString.Lazy and Data.ByteString.Short, since these also have unsafe index functions.