[[]] = f()

[satyr]

$ coffee -bpe '[[]] = f()'
false;

[satyr]

Related (?):

$ coffee -bpe 'for k, {} of o then'
var k;
for (k in o) {
  false;
}

[OnesimusUnbound]

Also this one

[] = 1
{} = [1, 3, 5]
[] = {}

I think is has to do with attempting destructuring assignment.

[satyr]

[] = 1 ...

Those are expected.

$ js -e 'print([] = 1)'
1

$ coffee -e 'console.log [] = 1'
1

[OnesimusUnbound]

The expression [] = 1 will generate the JavaScript false;

> coffee -bpe "[] = 1"
false;

Using your first example, when the expression [[]] = f() is a parameter of console.log

> coffee -bpe "console.log [[]] = f()"
var _ref;
console.log((_ref = f(), _ref));

Again, the false; statement seems to appear in {} = expr or [] = expr. one - liner.

[satyr]

The expression [] = 1 will generate the JavaScript false;

Oh right. Was assuming those were unaffected from the regression (at 4ce374b); not the case apparently.

[michaelficarra]

Looks like all the cases in this ticket were fixed by 230dea1 as well. Closing.

[satyr]

230dea1

This effectively reverts #870.
$ bin/coffee -bpe '[a, [], b] = c'
var a, b;
a = c[0], c[1], b = c[2];

[michaelficarra]

@satyr: Yeah, that optimization is now gone. But it was breaking other things anyway, so it was a bad optimization. I will open #870 so that someone can make a good optimization or just implement the [first, ..., last] syntax.