identify R.__ by special property rather than by identity

[davidchambers]

Closes #1052

The '@@functional/placeholder' property name was proposed by @paldepind in #1052 (comment).

---

If we decide to merge this pull request, I wonder how likely it is that the Underscore contributors would agree to add this property to the _ object.

[buzzdecafe]

🐄
let's do it, and adjust to reactions of others later

[megawac]

Heads up, there used to be a deopt associated with property names that cannot be accessed by dot notation (v8)

[CrossEye]

I think this is a good pass at this. Later, we may want to consider a Symbol instead, for those environments supporting them, but that's not urgent.

@megawac: although I doubt that would stop us (as this style of property name for such interfaces if getting to be a defacto standard, ISTM), I am curious to hear more about that de-opt. Do you have any references for it?

[paldepind]

All this testing looks superfluous to me. curry is based on curryN and curryN is tested for the exact same behavior

[davidchambers]

I don't follow. We need these tests to ensure we're not comparing to R.__ by identity, don't we?

[paldepind]

My point is that these tests should be for curryN. You've made no change to curry – yet you've changed the tests for curry but not for curryN.

[CrossEye]

Well, since curry is public and its behavior has changed, its tests should probably reflect that. The fact that it is implemented in terms of curryN is not a public guarantee, so our tests shouldn't assume this.

OTOH, we probably also need similar tests for curryN, as that one is also public.

[paldepind]

That is not an unreasonable way to see it. I agree.

[davidchambers]

I've added tests for R.curryN. Thanks for pointing out this omission, @paldepind.

[paldepind]

Hehe. It was not really my initial intent. But I'm glad it lead to something useful.

[CrossEye]

🌿

[jdalton]

If we decide to merge this pull request, I wonder how likely it is that the Underscore contributors would agree to add this property to the _ object.

Not likely from my end. Lodash/Underscore support customizable placholders in a different way, based on the .placeholder property on methods with placeholder support, e.g. _.partial.placeholder.