handle WdlPair - to fix #1703

[delocalizer]

JsArray is the simplest... but do you think JsObject with keys "left" and "right", or "0" and "1" would be better?
"left" and "right" would be consistent with the wdl4s code, but not extend very well to larger tuple types if you ever decide to go there.

[delocalizer]

This was to fix #1703

[cjllanwarne]

Sorry that I missed this over the Thanksgiving break. This is awesome.

Personally I would favour a JsObject containing left and right members. I agree that this doesn't extend well to 3,4,5-tuples, etc, but I suspect for anything above 2 we'd want some sort of generic Tuple type which could have it's own rules for objectifying. In that case I could imagine a Tuple which happened to be length 2 might objectify differently from a Pair.... And IMO that'd be fine.

But I'm not the only one with an opinion, I'm sure! @kcibul, @geoffjentry?

[cjllanwarne]

Two other comments

• Could you add a ticket to allow the same style of JSON to be used for Pair inputs (whatever we decide here re JsArrays vs JsObjects
• Could you add a test case?

[geoffjentry]

@cjllanwarne @delocalizer I don't have a strong opinion either way although I think left/right looks nicer.

I'm not concerned about expandability into tuples. Chris what you said is pretty much exactly what we laid out in the "but what if we want tuples?" discussion when talking about Pairs. My take on that is if that happens then the pair simply becomes semantic sugar over a tuple2 at which point we can map left/right to whatever representation a tuple2 uses.

[delocalizer]

@cjllanwarne I added a test - is that the kind of thing you were after?

[Horneth]

FWIW I like the JsObject version better as well. Thanks for this !

[cjllanwarne]

Looks good. 👍

[geoffjentry]

👍

[kshakir]

Reminder, while some of the tests should work, the entire current Travis test suite will never pass due to #1717.

[geoffjentry]

Thanks @delocalizer !