best practice 🙏

I am always amazed at this genius, there are no words. Wish you success!

For a it doesn't matter, I just didn't think too much when coding the formula. But for r 0 should be excluded and 1 included.

This is tail recursive. However, Haskell doesn't need tail recursion to make its execution more memory-saving, due to its non-strict evaluation.

Since randomIO generates uniformly distributed numbers between [0,1], the (1 -) <$> part is redundant, IMO.

I'm actually more surprised that xenia managed to come up with the exact same solution I have, even naming of variables, and the error you have pointed out.

The drop 1 res part is incorrect. Since you append one element to res when '(' is encountered, you should remove it from the tail of res when ')' is encountered, namely init res.

uhm.. what is the point?

You're of course right an application should be applied; I think I needed this to have either an unambiguous parse, or a finitely defined one. I had some problems there at first. This is the BNF I ended up implementing; it's not necessarily a correct Lambda Calculus. It is correct for the tests, however.

My problem, and the suggestion part of what I wrote, is that some of your description is incorrect and/or confusing.

You can do with that, or not, what you want.

Sorry I didn't implement sepBy correctly. Now it works. Thank you :D

Your code should report error on invalid code. invoke(, b) is invalid, invoke(a, ) as well.

Are you asking for spoiling hints?

I've clicked "Run sample tests" hundreds of times and the results were all green. But "Attempt" just gives me the error message. Shouldn't the test case Should work when there're multiple parameters in section When there's no lambda be invoke(a, b) ? How can the error message requires me return Left "Hugh?" ?