Different rules for conditional function call (e.g. fn? x )

[tad-lispy]

I was very much surprised today to learn that this code doesn't throw:

fn = 5
fn? 10

I was expecting it to throw TypeError: fn is not a function. Turns out it compiles to:

if (typeof fn === "function") {
  fn(x);
}

IMO the language is too lenient here for it's user's own sake. The above example is quite certainly an application developers mistake and an error would be very helpful. Current implementation can lead to a very subtle bugs.

I suggest it compiles to this instead:

if (typeof fn !== "undefined" && fn !== null) {
  fn(x);
}

Real life example

I had a perfectly good piece of code like this:

if data.value? then do_something_with data.value
else default_value

Then I decided that it will read better if I put the conditional logic on new lines:

if data.value? th # <-- but by mistake I left th out of then, a stupid typo!
  do_something_with data.value
else
 default_value

Now my code was compiling and running without errors, but the left branch of the if statement was never being executed. You can imagine that in a large codebase it was quite hard to figure out where does the wrong value come from.

[carlmathisen]

This is a feature, not a bug, and IMO a very important feature of CoffeeScript.

An example usage of the existential operator from the CS docs:

zip = lottery.drawWinner?().address?.zipcode

Or fn? 10 if you will.

Whenever I write code in JavaScript instead of CoffeeScript, this is the number one feature I miss. :)

[tad-lispy]

Thanks for your comment @carlmathisen. I fully agree with you that conditional function call is a very good feature.

I'm not saying it should be removed, but rather that it's implementation should be different. Example you have provided works just as well in both current and proposed implementation, unless you expect that drawWinner can be defined and not to be a function at the same time and your intention is not to call it then. I doubt this is the case. My guess is that you want to call it if it exists. Correct me if I'm wrong :)

Consider how existential family of operators in CoffeeScript is consistent, except for the function invocation:

operator	meaning
a?	does a exist?
a ? b	if a exists then a else b
a ?= b	set a to b if a does not exist
a?.b	take b of a if a exists
a? b	call a if a is a function

What I suggest is to change the meaning of a? b to mean call a if a exists, which IMO is more consistent (and thus intuitive) and also prevents kind of bugs I described above.

[tad-lispy]

Ok, I've just realised that original title for this issue wasn't the best I could have come up with :)

[connec]

This has come up before (e.g. #2315, and I'm sure there are more recent ones). There's been no consensus so far.

Personally I agree that the semantics should be the same for function calls as they are for other existence checks (e.g != null), both for consistency and to catch hard-to-debug cases such as these.

[tad-lispy]

Thanks @connec. Closing in favour of #2315. I did my research, but missed this one.