Nerd Sniping: What does the code do at the end of the page? | Hacker News Reader

Nerd Sniping: What does the code do at the end of the page?

neil.fraser.name

It returns false (also in Ruby) --- the "finally" (or "ensure") clause runs after the function body, but before the function epilogue.

It's not really a "paradox" so much as it is very bad style to "return" from a "finally".

Edit

My partner Dave has a much better trick question:

    def foo
        begin
            return false
        ensure
            next
        end
    end

"You win", I told him. "Nah, Google won. But nobody really wins if you write code like that."

The last trick question doesn't work in Ruby, at least not Ruby 1.8.7. Gives me a "LocalJumpError: unexpected next".

I realize that's not the point, but I figured I'd mention it anyway.

Here's a good article that explains what happens: http://blog.leshill.org/blog/2009/11/17/ensure-with-explicit...

Huh, interesting. The code supplied is:

try { return true; } finally { return false; }

Which returns false. I guess this is because of the guarantee that the finally block will be run, no matter what. The JS interpreter must be catching the return and holding it pending the execution of the finally block.

Checking with

try { return true; } finally { alert('hi'); }

Shows this to be true. (It both returns true AND alerts 'hi'.)

I see it that the "finally" is run right before the return. So, the statement "true" would be evaluated, then control would go to the "finally" block, where the method would return false.

Its a tricky question, but not something I would use to find a "keeper". From the Java tutorial - "The finally block _always_ executes when the try block exits." After the interpreter executes the "return true;" statement, it exits the try block. And then, if you know that "The finally block _always_ executes when the try block exits.", you know that the code block will return false;

It's not the answer that makes this a "keeper" screen, but the candidate's progression of facial expressions.

in .NET its illegal - "Control cannot leave the body of a finally clause"

It appears MS thought this through. The contrived example is but ONE way a mediocre programmer could really munge this stuff up if the language is not specific about protecting exit points from a finall{} block.

What about throwing exceptions in the finally block? Isn't it an exit point too?

When you call non-trivial 3rd party code from your finally block, how do you ensure it doesn't throw? Or what do you do if it does? Indefinitely nested try blocks?

Indeed. I can't think of a reason as to why any language would even allow it.

I know it's horrible in practice, and I really can't see a valid use for this particular case, but forbidding things like this (or other things like multiple inheritance) just because it can be abused has never seemed like a good idea to me.

It's not just that it can be abused, though - every use of it is an abuse. Finally{} is not for decisions and manipulating values and program logic, it's for cleaning up resources.

I wonder about the merits of this sort of thing as an interview screen, though. There's not necessarily a strong correlation between knowledge of language esoterica and other more important qualities like perseverance and raw intelligence.

The point is not to assess language knowledge, but disposition. Even if the candidate is unamused, they should still instantly get what the interviewer found punny about it.

The test is not whether you know the correct answer, but whether your brow furrows when you read the code. If you don't find the snippet interesting you fail.

The ultimate screen would be to follow the candidate home and see if they Google or test the answer.

it's not so mysterious if you know what the intent of a 'finally' block is.

I agree; the trickiest thing about it is the confusing relationship between the syntax for exception handling and the syntax for cleanup blocks, which are only vaguely related concepts.

That code uses one of several gotchas in java: http://www.cs.arizona.edu/projects/sumatra/hallofshame

Error: Norman coordinate.

It's also worth noting that this won't work in Python before 2.5 (or maybe 2.4).

In 2.3 it does. Why wouldn't it?

I swear I read that this behavior changed somehow in Python 2.5. But I'm not sure how. But it does at least appear to work in python 2.4.

In 2.5 they made try/except/finally possible.

You mean after 2.5 right?

try: return True finally: return False

-- There's an error in your program: * 'return' outside function ...

Just sayin'

It works if you put it in a function (2.5.1):

  >>> def test():
  ...  try: return True
  ...  finally: return False
  ... 
  >>> test()
  False

2.6.2 as well. I liked this little snippet, though I grumbled a little on the inside and would be appalled if I ever found something like that in serious code...

A couple things can happen I guess, it depends on the interpreting language.

It's just a dumbed down version of Duff's device :-)

How is this related to Duff's device?

It isn't even a very little bit related to Duff's device.

Simple. It too snips nerds, but only the weaker ones.

Reply on news.ycombinator.com