unwrap tag (tagged)

[etaty]

I was trying to play with tagged type
https://gitter.im/milessabin/shapeless?at=56ab8938dc33b33c7548a5a2

[clhodapp]

Can you add a test similar to https://github.com/etaty/shapeless/blob/features/tagged-unwrap/core/src/test/scala/shapeless/unwrapped.scala#L109 ? I'm concerned about whether it will pass.

[etaty]

Yep good catch.
test.illTyped("unwrapped: String @@ TestTag") don't work
unless I write
val unwrapped : String = wrapped.unwrap

[etaty]

that's weird, I am curious for the reason why it keeps the tagged type.

[clhodapp]

It's because e.g. String @@ TestTag is the most-specific T that can be extracted from T @@ TestTag, so that's what the compiler chooses to infer (as strange as it is, (String @@ TestTag) @@ TestTag is actually the same as String @@ TestTag and the compiler is able to see that this is the case and reason from it). As far as I currently know, you actually need to be working in the context of a macro to actually strip tags off in the general case, such as what's needed here.

[clhodapp]

Also, the version of this in your PR actually leaves off the maybe-important feature of being able to see through multiple wrappers over the same type. This is the purpose of all of those chain parameters in the other instance providers. Also, rather than constructing a new instance, you can take advantage of a micro-optimization I make in some of my other instances: Taking an implicit chain and simply casting it. This works because Unwrapped instances for e.g. tagged types don't have to actually change the object, they just need to make the types line up.

[milessabin]

What's the status of this? I'm aiming for an RC1 before the end of the week and it'd be nice if we could get this in by then.

[etaty]

Hi Miles,
I know nothing about macro.
And I did not realise that in my case, I did not need the unwrap. (I am also not sure it is a needed feature).
So I won't probably finish the pull-request by the end of the week.
(you can close/reject)

[milessabin]

No worries ... if you'd like to work on this I'll punt it to 2.3.1?

[milessabin]

Moved to 2.3.1 ...

[clhodapp]

Just got around to checking how easy it is to define an untagger with Miles's higher order unification patch (scala/scala#5102). It's pretty easy, I think: https://gist.github.com/clhodapp/a2f1b6fc011df9b60b107bd3034aef6d

[clhodapp]

Actually, that technique seems to work even without higher order unification!

[clhodapp]

https://gist.github.com/clhodapp/3862b54ddda3dc45a581b1733342a7bc

[milessabin]

Do either or both of you have time to try and get this over the finish line for 2.3.1?

[clhodapp]

I'll make a PR this evening or next.

[milessabin]

Oh, blimey! Missed this ... sorry :-(

LGTM. Many thanks ... merging :-)