Optional @?x fails to parse

[STRd6]

@?trigger "changed"

Currently fails with a parse error but should parse as

this?.trigger("changed")

In strict mode this can be null or undefined.

[edemaine]

Oddly, CoffeeScript treats @?trigger "changed" like @? trigger "changed", i.e., as two function calls. This is pretty weird though.

This is related to a question I asked a while back on Discord: should x?y be shorthand for x?.y, just like how x?(y) is shorthand for x?.(y) and x?[y] is shorthand for x?.[y]? I think making @?x shorthand for this?.x is less controversial, though, because @x is already shorthand for this.x which includes the dot.

[vendethiel]

Oddly, CoffeeScript treats @?trigger "changed" like @? trigger "changed", i.e., as two function calls. This is pretty weird though.

It's self-consistent - the parens are invisible, not the dot.
@?foo is this && this(foo) in CoffeeScript, not this && this.foo() (just like a?b is a && a(b), not a && a.b). Coffee considers @ to be callable (which, well, is fair enough.)

[STRd6]

Hmm, we may want to go for consistency with CoffeeScript then, or at least note it and have a compat flag.

[vendethiel]

I'd say CoffeeScript is not consistent with itself here: the only call introducers are supposed to be whitespace and (.

Coco and LiveScript did change this (a?b is a?.b), though I'm not 100% sure it was meant to change the meaning of a?b then (Coco 0.4.2: "Enabled ADI on ?.", but the commit adds a test without removing or changing another one)

[edemaine]

TIL CoffeeScript treats f?g like f? g also. So I agree it's somewhat consistent on this point, but it doesn't match my own intuitive definition of implicit calls which would have matched @vendethiel's:

the only call introducers are supposed to be whitespace and (.

However, CoffeeScript's actual definition is vague:

You don’t need to use parentheses to invoke a function if you’re passing arguments. The implicit call wraps forward to the end of the line or block expression.

I don't see any CoffeeScript tests about this, nor could I find an issue.

But when we define a meaning for @?foo and/or foo?bar, I agree we should probably add a coffeeCompat flag, even though I couldn't find any real-world usage.