11

我有以下 Haskell 代码：

-- Problem 69

import ProjectEuler

phi :: Integer -> Integer
phi n = n * product [p - 1 | p <- primeDivisors n] `div` product [p | p <- primeDivisors n]
-- primeDivisors n is a list of the prime divisors of n

maxRatio :: (Int, Int, Double) -> (Int, Int, Double) -> (Int, Int, Double)
maxRatio t1@(_, _, x) t2@(_, _, y)
  | x > y = t1
  | otherwise = t2

main = print (foldl
                maxRatio
                (0, 0, 0.0)
                [(n, phi n, ratio) | n <- [2..max], let ratio = fromIntegral n / (fromIntegral (phi n))]
              )
    where max = 1000

这给出了以下错误：

Couldn't match expected type `Int' with actual type `Integer'
In the expression: n
In the expression: (n, phi n, ratio)
In the third argument of `foldl', namely
  `[(n, phi n, ratio) |
      n <- [2 .. max],
      let ratio = fromIntegral n / (fromIntegral (phi n))]'

我怀疑三元组(0, 0, 0.0)中的 0 是 type Int。0在这种情况下，始终是类型Int还是 ghci 推断类型Int？如果是后者，我该如何强制它输入Integer呢？还是有其他原因导致此错误？

4

4 回答 4

18

Haskell can generally infer the type of numeric literals such as 0 as whatever appropriate type you need them to be. This is because it knows what functions you pass them to; if I have a function phi :: Integer -> Integer, and I call phi 0, Haskell knows that that particular 0 must have been an Integer. It's also fine if I call a function pho :: Int -> Int with pho 0; that particular 0 is inferred to be an Int.

However Int and Integer are different types, and there's no way one particular 0 can be passed to both phi and pho.

Your issue is simply that the tuples that maxRatio deals with are typed (by you) (Int, Int, Double), but that one such tuple is constructed as (n, phi n, ratio). Since phi takes and returns Integer, the n in that expression has to be an Integer. But then that doesn't work for maxRatio, so you get the error.

Depending on which type you actually wanted (Int or Integer), all you need to do is change the type signature of phi or maxRatio so that they're working with the same kind of number. Haskell will decide that your literally written 0s are whatever numeric type is necessary to make that work, provided there is one that can make it work!

Note that the error messaged specifically told you that it was n in (n, phi n, ratio) that was expected to be an Int and was actually an Integer. The (0, 0, 0.0) tuple is never mentioned. Often type errors originate somewhere other than where the compiler points you (since all the compiler can do is spot that different chains of inference produce inconsistent requirements on the type of something, with no way to know which part of the whole process is "wrong"), but in this case it did pretty well.

Haskell gets a (fairly justified) bad rep for inscrutable error messages, but it can help a lot to start from what the compiler is telling you is the problem and try to figure out why the facts it's complaining about arise from your code. This will be painful at first, but you'll quickly develop a basic literacy in Haskell's error messages (at least the more straightforward ones) that will help you spot these kinds of errors really quickly, which makes the compiler a very powerful error-detection system for you.

于 2012-09-05T02:32:39.477 回答
5

n被推断为Int由于的类型maxRatio，而类型phi表示它应该是Integer。最简单的解决方法是更改maxRatio​​要使用的类型，Integer甚至只是a因为它不触及这些值。

于 2012-09-05T02:10:13.077 回答
3

它正在被推断，因此您只需更改maxRatio. 不过，如果您需要将显式更改Int为Integer，请使用toInteger :: (Integral a) => a -> Integer

于 2012-09-05T02:25:43.063 回答
1

您的类型签名不一致 - 全部替换Int为Integer。

于 2012-09-05T15:02:10.640 回答