Keep symmetric behaviour when composing or using transducers with arrays and objects

[xgbuils]

We have this code:

var mapSquare = R.map(x => x * x);
var mapInc = R.map(R.inc);
var intoArray = R.into([]);
var intoObject = R.into({});

var fn = R.compose(mapSquare, mapInc);
var xFn = R.compose(mapInc, mapSquare);

var arr = [1, 2, 3];
var obj = {a: 1, b: 2, c: 3};

fn(arr) // [4, 9, 16]
intoArray(xFn, arr) // [4, 9, 16]

// However ...
fn(obj) // {a: 4, b: 9, c: 16}
intoObject(xFn, obj) // throwing: reduce: list must be array or iterable
// but I also expect {a: 4, b: 9, c: 16}

I guess there are similar issues using filter. I could do a PR to fix it.

Best regards!

[CrossEye]

I'd have to think carefully about what it means to support transducing on objects. I think about take(2) for instance, We think in Ramda as {a: 1, b: 2, c: 3} as equivalent to {b: 2, c: 3, a: 1}. The thought that take(2) would yield different results for them makes me pretty suspicious.

But if you have ideas, please go ahead and share them.

[CrossEye]

Sorry, hadn't seen the PR when I responded. I'll check there.