Question

I can do the following:

Prelude> reads "1 2 3" :: [(Int, String)]
[(1," 2 3")]
Prelude> reads "(1,2) (3,4)" :: [((Int, Int), String)]
[((1,2)," (3,4)")]
Prelude> reads "(1,2)(3,4)" :: [((Int, Int), String)]
[((1,2),"(3,4)")]
Prelude> reads "(1,2)\n(3,4)" :: [((Int, Int), String)]
[((1,2),"\n(3,4)")]
Prelude> reads "(1,2)    (3,4)" :: [((Int, Int), String)]
[((1,2),"    (3,4)")]

I can derive Read and get reads to read those too. But I’ve never gotten reads to return more than one tuple in the list. Why does reads return a list?

Answer 1

None of the standard instances do so, but it’s intended for ambiguous parses; since this is not really very useful, and parsers that use this functionality would be very inefficient, reads‘s return value is for all practical purposes a Maybe masquerading as a [].

The Report’s definition of read reveals the intended meaning of multiple parses:

read    :: (Read a) => String -> a  
read s  =  case [x | (x,t) <- reads s, ("","") <- lex t] of  
              [x] -> x  
              []  -> error "PreludeText.read: no parse"  
              _   -> error "PreludeText.read: ambiguous parse"

So: historical reasons, basically.