为什么GHC不能针对常数进行优化？

Question

import Data.List
a = foldl' (+) 0 [1..99999999]
main = putStrLn $ show $ a

This program takes a while to run. But a does not depend on anything and thus is constant. It could be perfectly calculated at compile time. Why is not GHC optimizing for this? Is there a flag for it to do so, or should I just replace that kind of constant calculation by the values themselves?

Answer 1

This is not a perfect solution, but as kqr already remarked you can of course achieve your goal with Template Haskell:

{-# LANGUAGE TemplateHaskell #-}

import Language.Haskell.TH
import Data.List

a :: Integer
a = $( return . LitE . IntegerL $ foldl' (+) 0 [1..99999999] )

main = print a

This generates the integer literal 4999999950000000 from the fold expression, before actually starting to compile the program.

Answer 2

This is the same as the discussion covered by this reddit thread. Basically, what you want is but a simple case which is uncommon in practice. Optimizing with constant folding quickly turns into problems regarding Gödel's theorem and the Halting problem.