I see. RankNTypes should be made available for TupleSections to work. The above error will occur when {-# LANGUAGE RankNTypes #-} is missing. Sorry for the stupid question.

How to avoid the following error on the test case Cannot use Control.Lens? I did not import anything that relates to Control.Lens.

uncaught exception: IOException of type UserError
user error (Could not parse solution correctly)

Which one is required now? 4 (2+avg 10 30) = 13.0 or (avg 4 2) + (avg 10 30) = 23.0? Thanks really.

After skimming the other discourses I found that this problem would very possibly not be tested ;)

Suppose

fn apply f x => f x
fn incr x => x + 1

Since function calls are right associative, then how can we parse apply incr 1 into apply incr 1 but not apply (incr 1)?

Do not modify the nested list in your implementation or you won't pass the test ;)

This kata may be more interesting if parameter passing is required. 2333

The standard of register names is not clarified in detail.
In Assembler interpreter (part I), they are described as re alphabetical (letters only). But actually in this kata, to pass the tests, you should allow register names to begin with _/letters and follow by _/letters/numbers, like most common identifiers.

Maybe Data.Map with zipped list?

Data.HashMap is not supported by the Haskell compiler. Could I use something instead?

Is it necessary to convert the number to a 3-base number (as a string) when implementing the O(nlogn) algorithm?

I feel sorry that I have used the wrong combinator (S but not I) in implementing function xor', thus modified the type of it.

Maybe you could use unsafeCoerce in Unsafe.Coerce.

I think the topic isomorphism means more things than constructing bijections. Finishing this kata only requires us to think about making one-to-one correspondence between two certains sets, yet the name isomorphism makes me think that there should be some closed operations on these two sets, which corresponds to each other.

for churchPow you could use (&) from Data.Function.