Code Review: Monad and run time error

[cd155]

In the following code, it will have run time error because isBalancedWithHeight' could return a Nothing, I did not tell the program what to do if isBalancedWithHeight' l or isBalancedWithHeight' r is Nothing. (the comment part)

isBalancedWithHeight' :: BiTree a -> Maybe Int
isBalancedWithHeight' Empty' = Just 0
isBalancedWithHeight' (Node' _ (l,r))
    -- | isBalancedWithHeight l == Nothing || isBalancedWithHeight r == Nothing = 
    --     Nothing
    | abs(lh - rh) <= 1 = Just (max lh rh + 1)
    | otherwise = Nothing
    where Just lh = isBalancedWithHeight' l
          Just rh = isBalancedWithHeight' r

I was puzzled why the following code wouldn't have a run time error as the previous code does

isBalancedWithHeight :: BiTree a -> Maybe Int
isBalancedWithHeight Empty' = Just 0
isBalancedWithHeight (Node' _ (l,r)) = do
  lh <- isBalancedWithHeight l
  rh <- isBalancedWithHeight r
  if abs (lh-rh) <= 1
    then Just $ max lh rh + 1
    else Nothing

After enabling Haskell Language Server in vscode, lh and rh be automatically inferred as Int rather than Mabye Int. Can anyone tell me why and how this code handle when isBalancedWithHeight l or isBalancedWithHeight r return Nothing?

[cd155]

The following is responded by ChatGPT

The reason why lh and rh are inferred as Int rather than Maybe Int is because they are extracted using the <- operator inside a do block, which automatically unwraps the Just value inside the Maybe monad.

Regarding the handling of Nothing values, in the do block, if isBalancedWithHeight l or isBalancedWithHeight r returns Nothing, the entire computation will short-circuit and return Nothing, indicating that the tree is not balanced. This is because in a do block, if any step in the sequence returns Nothing, the entire computation returns Nothing. Therefore, you do not need to handle the case explicitly like in the commented out section of the first code snippet.

[smiths]

I'm very far from an expert in Haskell, but this looks like a reasonable response from ChatGPT. @cd155 you gave it the above code snippets and asked your question? Maybe we can replace some TAs with AI in our courses. 😄 Of course I'd rather increase TA hours, but having another option for students to learn is a good thing.

[cd155]

I am satisfied with the answer that ChatGPT gave, but I feel something is still missing. For example, the chat didn't say why the <- can automatically unwrap the value in the 'do' block. Other than that, the response is quite impressive.

[cd155]

@smiths I also hear some people use ChatGPT to do their homework. I know most institutions use plagiarism software. It seems a challenge that artifacts work is quite similar to human creation. It also raises a question of how we distinguish artifacts and human creation.

[smiths]

Yes, I think ChatGPT is becoming an issue for homework. I'm lucky to be teaching the capstone project right now where it isn't a worry. However, in my second year design course in the past many of my assignment questions could have been answered by ChatGPT. If I was working on a course like that now, I would have to spend time checking my questions with ChatGPT. I know the Faculty is investigating options for detecting academic dishonesty using ChatGPT.

[JacquesCarette]

ChatGPT has a correct answer here, even though it's phrased in a way that you'd expect a beginner to use and not an experienced Haskell programmer.

The <- isn't "unwrapping" the Int, it is using the definition of what happens in the Maybe monad, specifically the meaning of bind (aka >>=). The second paragraph alludes to that, in a roundabout way.

I was speaking with a prof yesterday who caught multiple (Master's !!!) students in his course using ChatGPT for handing in answers to some essay questions. It's a real problem, and it will get worse before it gets better.