Re: [was ghc-devs] Reasoning backwards with type families

Happy to go with the bogusness if it works better than injective types, feel free to submit a patch. :) On Thu, 14 Dec 2017 at 9:08 pm, Anthony Clayden < anthony_clayden@clear.net.nz> wrote:

On Thu, 14 Dec 2017 at 4:13 PM, Clinton Mead wrote:

...

I've panicked GHC enough whilst developing Freelude so whilst I'm not sure exactly what you're saying I'd be hestiant about exploiting anything bogus (8.2 btw seems far more stable than 8.0 btw).

;-) Fair enough.

"bogus" is SPJ's way of saying: it works, but it isn't supported by deep type theory. 'C'est brutal mais ca marche.'

And that particular exploit has been stable since 2004 at least: the HList library totally relied on it until Closed Type Families arrived.

AntC