Haskell中的“浮动”和“分数”有什么区别？

Question

Prelude> let c=[1.0,2.0]
Prelude> :t c
c :: Fractional t => [t]

I'd expect "c" to be a list of either Num or Float. Why Fractional? Is there any implicit type conversion going on here in Haskell?

Answer 1

You can make it happen:

> :set -XNumDecimals
> let c=[1.0,2.0]
> :t c
c :: Num t => [t]

The answer to "why Fractional " is basically "because the Report says so":

 float → decimal . decimal [exponent] | decimal exponent 

A floating literal stands for an application of fromRational to a value of type Rational (that is, Ratio Integer ). Given the typings:

 fromInteger :: (Num a) => Integer -> a fromRational :: (Fractional a) => Rational -> a 

integer and floating literals have the typings (Num a) => a and (Fractional a) => a , respectively.

I would guess the Report says so because it is a nice simple rule to explain: no dot/exponent, it's Num -polymorphic, yes dot/exponent, it's Fractional -polymorphic.

Two relevant bits are:
https://www.haskell.org/onlinereport/haskell2010/haskellch2.html#x7-190002.5
https://www.haskell.org/onlinereport/haskell2010/haskellch6.html#x13-1360006.4.1

Answer 2

There is no implicit type conversion in Haskell. Numeric literals have polymorphic types.

> :t 3
3 :: Num a => a

> :t 3.5
3.5 :: Fractional a => a

Standard Haskell treats any number with a decimal point as fractional, even if the fractional digits are all 0. It also treats anything written in scientific notation as fractional. There's a GHC extension you can use to be more polymorphic with such representations if you like.

As for the question in your title, Float is a concrete type ( Float :: * ), while Fractional is a type class ( Fractional :: * -> Constraint ). You can write lots of functions that will work with all sorts of Fractional or RealFrac or RealFloat types without having to worry about the exact representation.

Side note: Float is a special-purpose type, for specialized algorithms, formats, etc., and certain cases where compact representation of arrays is more important than precision or numerical stability. When you want floating point, the one you usually want is Double .