Transducers: map(prop(p)) != pluck(p) ?

[monfera]

I must be overlooking something obvious:
With full application:

R.map(R.prop('a'))([{a: 1}, {a: 2}, {a: 3}, {a: 4}])
[1, 2, 3, 4]
R.pluck('a')([{a: 1}, {a: 2}, {a: 3}, {a: 4}])
[1, 2, 3, 4]

So far, so good. Now with transducers:

var numbers = [{a: 1}, {a: 2}, {a: 3}, {a: 4}];
var transducer = R.compose(R.map(R.prop('a')), R.map(R.add(1)), R.take(2));
R.transduce(transducer, R.flip(R.append), [], numbers); 
[2, 3]

It is still good, however this won't work:

var numbers = [{a: 1}, {a: 2}, {a: 3}, {a: 4}];
var transducer = R.compose(R.pluck('a'), R.map(R.add(1)), R.take(2));
R.transduce(transducer, R.flip(R.append), [], numbers); 
Uncaught TypeError: xf.step is not a function
    at _arrayReduce (http://localhost:63342/d6/ramda.js:4201:26)
    at _reduce (http://localhost:63342/d6/ramda.js:4230:24)
    at Object.<anonymous> (http://localhost:63342/d6/ramda.js:5927:16)
    at Object.<anonymous> (http://localhost:63342/d6/ramda.js:1358:27)
    at Object.transduce (http://localhost:63342/d6/ramda.js:1041:31)
    at <anonymous>:5:3
    at Object.InjectedScript._evaluateOn (<anonymous>:895:140)
    at Object.InjectedScript._evaluateAndWrap (<anonymous>:828:34)
    at Object.InjectedScript.evaluate (<anonymous>:694:21)

[monfera]

Doing a quick and dirty, and replacing R.pluck with tpluck in the above snippet, I also get the expected [2, 3].

function tpluck(prop) {
    return R.map(function(x) {return x[prop];});
}

[CrossEye]

The problem, of course, is that R.pluck is not properly transduced.

If I'm not mistaken, and I can't tell for sure until I get in front of a computer, it's because pluck is built on an internal _map and not on the tranducerified public map. This, I suppose is more ammunition in my case for reducing the number of such internal functions we use. I'd rather be dogfooding our stuff as much as we can. It's not clear to me how easy an overall change this would be. But it should be simple to change pluck.

[monfera]

Probably it's not the underlying cause of the issue, but I noticed that ramda transducers don't follow the old or the new transducers protocol (https://github.com/cognitect-labs/transducers-js#the-transducer-protocol) which I ran into the hard way. Using Kefir.js for 'FRP', I failed to switch from transducers.js / transducers-js to ramda in a test. Is it due to some fundamental difference in the implementation approach?

[buzzdecafe]

@monfera HEAD version should be compliant with new protocol. see: #1006

[monfera]

Interesting, I was already up to date and Kefir still balked at a point which I could solve by switching a R.map inside a R.compose to a t.map (transducers.js).

[buzzdecafe]

please keep me posted -- i am interested to know what places we missed and need to bring into compliance. thanks