Question

Haskell has the sum function

sum :: Num a => [a] -> a

Which can be nicely composed to sum a matrix by

sum . map sum :: Num a => [[a]] -> a

Going deeper, however, such as summing a cube, creates the restriction Num [a]

sum . map sum . map sum :: (Num a, Num [a]) => [[[a]]] -> a

Which, if you think about it, is natural. So with the former attempt to define the sumcube function blowing up in one’s face, we need to find a different path. One such attempt would be:

sum . map sum . map (map sum) :: Num a => [[[a]]] -> a

Which seems nowhere as natural as the summatrix function.

In my quest to posessing the mental tools for problem solving in Haskell, I am interested in knowing how to tackle this problem of summing a structure of any depth by, say, stacking map sums as in my third code example. Is this at all possible? And in that case, how would you do it?

Answer 1

You’ll have to work from the inside out. When you have a function f for summing a data structure, then sum . map f is the way to sum a list of those data structures.

                     sum  :: Num a =>   [a]   -> a
           sum . map sum  :: Num a =>  [[a]]  -> a
sum . map (sum . map sum) :: Num a => [[[a]]] -> a