This seems to break the test added for #509

Yeah, #509 breaks.

$foo: (
    (key1): (value-1-0),
    key2: value-2-0,
    (key6): (value-6-0),
    key-3-0 key-3-1 key-3-2: value-3-0 value-3-1 value-3-2,
    key4: (value-4-0, value-4-1, value-4-2),
    key5: (key-5-0: value-5-1),
    (key-7-0: key-7-1): (value-7-0: value-7-1),
    (key-8-0, key-8-1, key-8-2): (value-8-0, value-8-1, value-8-2),
);

div {
    foo: map-get((foo: 1, bar: 2), foo);
    foo: map-get((foo: 1, bar: 2), bar);
    foo: map-get((foo: 1, bar: 2), baz);
    foo: map-get((), foo);
    foo: map-get($foo, (key-5-0: value-5-1));
    foo: map-get($foo, (key2));
    foo: map-get($foo, (key-3-0 key-3-1 key-3-2));
}

Results in an ~/dev/sass/spec/spec/libsass-closed-issues/issue_509/input.scss:7: unclosed parenthesis

Tested against s3.4 and it works just fine.

Lovely! Apparently, this breaks now on master. This wasn't an issue with this particular pull.

Update:_ ugh. I'm a brilliant professional who is not at all an idiot. I misread the Travis output... I thought "current" meant master. Nope. It's this pull request failing. 3.1 then!

I'm not sure what was up with this all seems fine now. I think master was failing for a short while.

Nevermind, there was a regression here. I've fixed it.

Coverage increased (+0.01%) when pulling b17f435 on xzyfer:fix/map-key-as-numbers into f660ede on sass:master.