Sorry for taking so long to get around to this!

I agree that using NFData1 is the way to go here. But since you're in town—there were also similar datatypes in Data.Functor.Sum and Data.Functor.Product that were added to base at the same time that Compose was. Can you add NFData(1) instances for them as well to complete the trio?

Thanks!

No problem, I have updated the branch with the additional instances.

Looks good to me.

@hvr, should we merge this now, or wait until after the imminent release of the next deepseq that will be bundled with GHC 8.2?

@RyanGlScott if we're confident that these instances are correct, I see no reason to delay (there's a few other things I want to cleanup before I feel good enough about tagging 1.4.3) ... while GHC 8.2.1 RC1 won't have them, it's a given we're gonna have a GHC 8.2.1 RC2... and that's gonna have the final deepseq-1.4.3