I just started learning Haskell and am having trouble with
using the hmatrix library. I want to write some simple code
to compute the eigenvalue using the power iteration method.
I start with:
c = fromList [4.0, 4.0, 4.0, 4.0]::Vector Double
n = norm2 c
Which creates a vector c and finds the 2-norm of the vector.
Multiplication with c:
c * 2 (Works)
c * 0.5 (Works)
c * pi (Works)
c * n (Error)
I checked that:
>:t pi
pi :: Floating a => a
>:t n
n :: Double
The problem is with the types but I do not know how to get
around it.
Would I need to define my own (/) function in this case?
Update:
The error I obtain from ghci:
Couldn't match expected type `Vector Double'
with actual type `Double'
In the second argument of `(*)', namely `n'
In the expression: c * n
In an equation for `it': it = c * n
You're doing the right thing by checking the types. If we're a bit more explicit, we can see what is going on.
Prelude Numeric.LinearAlgebra> :t let a = 2; b = c * a in a
let a = 2; b = c * a in a :: Vector Double
The problem is that the type of norm2 c is Double and thus cannot be made into a Vector Double
Let's see the value of that polymorphic 2 from earlier.
Prelude Numeric.LinearAlgebra> let a = 2; b = c * a in a
fromList [2.0]
So instead, n = fromList [norm2 c]
Edit: The very same library exposes functions scalar and scale which you should look into.
(*) assumes that both of its arguments have the same type:
(*) :: (Num a) => a -> a -> a
The reason that your first three multiplications worked was because in all three cases the right argument successfully type-checked as a Vector Double!
To see why, let's ask ghci what the types of those three arguments are:
> :t 2
2 :: Num a => a
> :t 0.5
0.5 :: Fractional a => a
> :t pi
pi :: Floating a => a
All three of those are valid Vector Doubles because hmatrix provides the following three instances:
instance Num (Vector Double) where ...
instance Fractional (Vector Double) where ...
instance Floating (Vector Double) where ...
In other words, Haskell will transform 2, 0.5, and pi into Vector Doubles automatically, thanks to those instances.
This explains why your last example does not type check. n has type Double, which means that there is no chance it can ever type-check also as a Vector Double.
Related
If I compile the following source file with ghc -Wall:
main = putStr . show $ squareOfSum 5
squareOfSum :: Integral a => a -> a
squareOfSum n = (^2) $ sum [1..n]
I get:
powerTypes.hs:4:18: warning: [-Wtype-defaults]
• Defaulting the following constraints to type ‘Integer’
(Integral b0) arising from a use of ‘^’ at powerTypes.hs:4:18-19
(Num b0) arising from the literal ‘2’ at powerTypes.hs:4:19
• In the expression: (^ 2)
In the expression: (^ 2) $ sum [1 .. n]
In an equation for ‘squareOfSum’:
squareOfSum n = (^ 2) $ sum [1 .. n]
|
4 | squareOfSum n = (^2) $ sum [1..n]
| ^^
I understand that the type of (^) is:
Prelude> :t (^)
(^) :: (Integral b, Num a) => a -> b -> a
which means it works for any a^b provided a is a Num and b is an Integral. I also understand the type hierarchy to be:
Num --> Integral --> Int or Integer
where --> denotes "includes" and the first two are typeclasses while the last two are types.
Why does ghc not conclusively infer that 2 is an Int, instead of "defaulting the constraints to Integer". Why is ghc defaulting anything? Is replacing 2 with 2 :: Int a good way to resolve this warning?
In Haskell, numeric literals have a polymorphic type
2 :: Num a => a
This means that the expression 2 can be used to generate a value in any numeric type. For instance, all these expression type-check:
2 :: Int
2 :: Integer
2 :: Float
2 :: Double
2 :: MyCustomTypeForWhichIDefinedANumInstance
Technically, each time we use 2 we would have to write 2 :: T to choose the actual numeric type T we want. Fortunately, this is often not needed since type inference can frequently deduce T from the context. E.g.,
foo :: Int -> Int
foo x = x + 2
Here, x is an Int because of the type annotation, and + requires both operands to have the same type, hence Haskell infers 2 :: Int. Technically, this is because (+) has type
(+) :: Num a => a -> a -> a
Sometimes, however, type inference can not deduce T from the context. Consider this example involving a custom type class:
class C a where bar :: a -> String
instance C Int where bar x = "Int: " ++ show x
instance C Integer where bar x = "Integer: " ++ show x
test :: String
test = bar 2
What is the value of test? Well, if 2 is an Int, then we have test = "Int: 2". If it is an Integer, then we have test = "Integer: 2". If it's another numeric type T, we can not find an instance for C T.
This code is inherently ambiguous. In such a case, Haskell mandates that numeric types that can not be deduced are defaulted to Integer (the programmer can change this default to another type, but it's not relevant now). Hence we have test = "Integer: 2".
While this mechanism makes our code type check, it might cause an unintended result: for all we know, the programmer might have wanted 2 :: Int instead. Because of this, GHC chooses the default, but warns about it.
In your code, (^) can work with any Integral type for the exponent. But, in principle, x ^ (2::Int) and x ^ (2::Integer) could lead to different results. We know this is not the case since we know the semantics of (^), but for the compiler (^) is only a random function with that type, which could behave differently on Int and Integer. Consider, e.g.,
a ^ n = if n + 3000000000 < 0 then 0 else 1
When n = 2, if we use n :: Int the if guard could be true on a 32 bit system. This is not the case when using n :: Integer which never overflows.
The standard solution, in these cases, is to resolve the warning using something like x ^ (2 :: Int).
I'm trying to work a problem where I need to calculate the "small" divisors of an integer. I'm just bruteforcing through all numbers up to the square root of the given number, so to get the divisors of 10 I'd write:
[k|k<-[1...floor(sqrt 10)],rem 10 k<1]
This seems to work well. But as soon as I plug this in a function
f n=[k|k<-[1...floor(sqrt n)],rem n k<1]
And actually call this function, I do get an error
f 10
No instance for (Floating t0) arising from a use of `it'
The type variable `t0' is ambiguous
Note: there are several potential instances:
instance Floating Double -- Defined in `GHC.Float'
instance Floating Float -- Defined in `GHC.Float'
In the first argument of `print', namely `it'
In a stmt of an interactive GHCi command: print it
As far as I undrestand the actual print function that prints the result to the console is causing trouble, but I cannot find out what is wrong. It says the type is ambiguous, but the function can clearly only return a list of integers. Then again I checked the type, and it the (inferred) type of f is
f :: (Floating t, Integral t, RealFrac t) => t -> [t]
I can understand that fshould be able to accept any real numerical value, but can anyone explain why the return type should be anything else than Integral or int?
[k|k<-[1...floor(sqrt 10)],rem 10 k<1]
this works because the first 10 is not the same as the latter one - to see this, we need the type signature of your functions:
sqrt :: Floating a => a -> a
rem :: Integral a => a -> a -> a
so the first one means that it works for stuff that have a floating point representation - a.k.a. Float, Double ..., and the second one works for Int, Integer (bigint), Word8 (unsigned 8bit integers)...
so for the 10 in sqrt 10 the compiler says - ahh this is a floating point number, null problemo, and for the 10 in rem 10 k, ahh this is an integer like number, null problemo as well.
But when you bundle them up in a function - you are saying n has to be a floating point and an integral number, the compiler knows no such thing and - complains.
So what do we do to fix that (and a side note ranges in haskell are indicated by .. not ...!). So let us start by taking a concrete solution and generalize it.
f :: Int -> [Int]
f n = [k|k <- [1..n'],rem n k < 1]
where n' = floor $ sqrt $ fromIntegral n
the neccessary part was converting the Int to a floating point number. But if you are putting that in a library all your users need to stick with using Int which is okay, but far from ideal - so how do we generalize (as promised)? We use GHCi to do that for us, using a lazy language we ourselves tend to be lazy as well.
We start by commenting out the type-signature
-- f :: Int -> [Int]
f n = [k|k <- [1..n'],rem n k < 1]
where n' = floor $ sqrt $ fromIntegral n
$> ghci MyLib.hs
....
MyLib > :type f
f :: Integral a => a -> [a]
then we can take this and put it into the library and if someone worked with Word8 or Integer that would work as well.
Another solution would be to use rem (floor n) k < 1 and have
f :: Floating a, Integral b => a -> [b]
as the type, but that would be kind of awkward.
f1 :: Mesh -> Matrix Double
f1 me = knx where
hx :: Double
(hx , _) = h me
a, knx :: Matrix Double
a = fromLists [[1,2], [3,4]]
knx = hx * a
-- knx = 2 * a
I don't get why in the above function, mutliplying by 2 works whereas multiplying by hx = 0.5 doesn't. OTOH, multiplying a Matrix Double with a Double outside a function works as it should.
Couldn't match expected type ‘Matrix Double’
with actual type ‘Double’
In the first argument of ‘(*)’, namely ‘hx’
In the expression: hx * a
Failed, modules loaded: none.
I am seriously puzzled. Any pointers are welcome!
In HMatrix, scale :: Container c e => e -> c e -> c e does what it says on the label (multiplies the e in a c e by the first e). Here are some usage examples here: https://hackage.haskell.org/package/hmatrix-0.16.1.4/docs/src/Data-Packed-Internal-Numeric.html
It should be noted that scale x constructs a Container type by considering x a singleton list, via fromList.
It would be really handy if at least the common arithmetic operations would be overloaded, so that formulas may resemble their mathematical counterpart. I'm not sure whether defining function synonyms (e.g. (.*) = scale ) would be a good idea or it would just add a layer of complexity. Any thoughts?
Let's say I have the following Haskell type description:
divide_by_hundred :: Integer -> IO()
divide_by_hundred n = print(n/100)
Why is it that when I attempt to run this through ghc I get:
No instance for (Fractional Integer) arising from a use of `/'
Possible fix: add an instance declaration for (Fractional Integer)
In the first argument of `print', namely `(n / 100)'
In the expression: print (n / 100)
In an equation for `divide_by_hundred':
divide_by_hundred n = print (n / 100)
By running :t (/)
I get:
(/) :: Fractional a => a -> a -> a
which, to me, suggests that the (/) can take any Num that can be expressed as fractional (which I was under the impression should include Integer, though I am unsure as how to verify this), as long as both inputs to / are of the same type.
This is clearly not accurate. Why? And how would I write a simple function to divide an Integer by 100?
Haskell likes to keep to the mathematically accepted meaning of operators. / should be the inverse of multiplication, but e.g. 5 / 4 * 4 couldn't possibly yield 5 for a Fractional Integer instance1.
So if you actually mean to do truncated integer division, the language forces you2 to make that explicit by using div or quot. OTOH, if you actually want the result as a fraction, you can use / fine, but you first need to convert to a type with a Fractional instance. For instance,
Prelude> let x = 5
Prelude> :t x
x :: Integer
Prelude> let y = fromIntegral x / 100
Prelude> y
5.0e-2
Prelude> :t y
y :: Double
Note that GHCi has selected the Double instance here because that's the simples default; you could also do
Prelude> let y' = fromIntegral x / 100 :: Rational
Prelude> y'
1 % 20
1Strictly speaking, this inverse identity doesn't quite hold for the Double instance either because of floating-point glitches, but there it's true at least approximately.
2Actually, not the language but the standard libraries. You could define
instance Fractional Integer where
(/) = div
yourself, then your original code would work just fine. Only, it's a bad idea!
You can use div for integer division:
div :: Integral a => a -> a -> a
Or you can convert your integers to fractionals using fromIntegral:
fromIntegral :: (Integral a, Num b) => a -> b
So in essence:
divide_by_hundred :: Integer -> IO()
divide_by_hundred n = print $ fromIntegral n / 100
Integers do not implement Fractional, which you can see in the manual.
I made this (what I thought to be) fairly straightforward code to calculate the third side of a triangle:
toRadians :: Int -> Double
toRadians d = let deg = mod d 360
in deg/180 * pi
lawOfCosines :: Int -> Int -> Int -> Double
lawOfCosines a b gamma = sqrt $ a*a + b*b - 2*a*b*(cos (toRadians gamma))
However, when I tried to load it into GHCi, I got the following errors:
[1 of 1] Compiling Main ( law_of_cosines.hs, interpreted )
law_of_cosines.hs:3:18:
Couldn't match expected type `Double' with actual type `Int'
In the first argument of `(/)', namely `deg'
In the first argument of `(*)', namely `deg / 180'
In the expression: deg / 180 * pi
law_of_cosines.hs:6:26:
No instance for (Floating Int)
arising from a use of `sqrt'
Possible fix: add an instance declaration for (Floating Int)
In the expression: sqrt
In the expression:
sqrt $ a * a + b * b - 2 * a * b * (cos (toRadians gamma))
In an equation for `lawOfCosines':
lawOfCosines a b gamma
= sqrt $ a * a + b * b - 2 * a * b * (cos (toRadians gamma))
law_of_cosines.hs:6:57:
Couldn't match expected type `Int' with actual type `Double'
In the return type of a call of `toRadians'
In the first argument of `cos', namely `(toRadians gamma)'
In the second argument of `(*)', namely `(cos (toRadians gamma))'
It turns out the fix was to remove my type signatures, upon which it worked fine.
toRadians d = let deg = mod d 360
in deg/180 * pi
lawOfCosines a b gamma = sqrt $ a*a + b*b - 2*a*b*(cos (toRadians gamma))
And when I query the type of toRadians and lawOfCosines:
*Main> :t toRadians
toRadians :: (Floating a, Integral a) => a -> a
*Main> :t lawOfCosines
lawOfCosines :: (Floating a, Integral a) => a -> a -> a -> a
*Main>
Can someone explain to me what's going on here? Why the "intuitive" type signatures I had written were in fact incorrect?
The problem is in toRadians: mod has the type Integral a => a -> a -> a, therefore, deg has the type Integral i => i (so either Int or Integer).
You then try and use / on deg, but / doesn't take integral numbers (divide integrals with div):
(/) :: Fractional a => a -> a -> a
The solution is to simply use fromIntegral :: (Integral a, Num b) => a -> b:
toRadians :: Int -> Double
toRadians d = let deg = mod d 360
in (fromIntegral deg)/180 * pi
Seeing Floating a and Integral a in a type signature together always sets off my internal alarm bells, as these classes are supposed to be mutually exclusive - at least, there are no standard numeric types that are instances of both classes. GHCi tells me (along with a lot of other stuff):
> :info Integral
...
instance Integral Integer -- Defined in `GHC.Real'
instance Integral Int -- Defined in `GHC.Real'
> :info Floating
...
instance Floating Float -- Defined in `GHC.Float'
instance Floating Double -- Defined in `GHC.Float'
To see why these classes are mutually exclusive, let's have a look at some of the methods in both classes (this is going to be a bit handwavy). fromInteger in Integral converts an Integral number to an Integer, without loss of precision. In a way, Integral captures the essence of being (a subset of) the mathematical integers.
On the other hand, Floating contains methods such as pi and exp, which have a pronounced 'real number' flavour.
If there were a type that was both Floating and Integral, you could write toInteger pi and have a integer that was equal to 3.14159... - and that's not possible :-)
That said, you should change all your type signatures to use Double instead of Int; after all, not all triangles have integer sides, or angles that are an integral number of degrees!
If you absolutely don't want that for whatever reason, you also need to convert the sides (the a and b arguments) in lawOfCosines to Double. That's possible via
lawOfCosines aInt bInt gamma = sqrt $ a*a + b*b - 2*a*b*(cos (toRadians gamma)) where
a = fromInteger aInt
b = fromInteger bInt
The type signature for toRadians says it takes an Int but returns a Double. In some programming languages, the conversion from one to the other (but not back) happens automatically. Haskell is not such a language; you must manually request conversion, using fromIntegral.
The errors you are seeing are all coming from various operations which don't work on Int, or from trying to add Int to Double, or similar. (E.g., / doesn't work for Int, pi doesn't work for Int, sqrt doesn't work for Int...)