@rae: Impredicative types and dynamic types -- useful in parsing JSON

I demonstrate how using impredicative types and dynamic types (as they work in Haskell) can be useful for parsing a number in JSON (stored as a Scientific).

[raehik]

Thanks Richard! Now I understand I came across impredicative types the other day in fumieval's mason library for bytestring serializing. I apparently had to instantiate (?) the funky builder type to expose in a typeclass, or I'd get weird type errors, so I lost the ability to conveniently define serializers for any bytestring-like (strict, lazy, short). Wonder if it's an interaction with impredicative types & typeclasses or if I needed to tinker more.

[Dregntael]

Nice video! I wonder whether you could also get this to work with the original version of floatingOrInteger by using existential types AnyFloat and AnyInteger that encapsulate the universally quantified value?

[stevemao]

I'm curious if impredicative types can be used as the backend of dependent types? I think most people use singleton to implement dependent types so just wondering.

[BjrnSH]

Hmm, to get rid of the fromInteger, I'm wondering if you couldn't simply move the last "Right" to before the "case"?

[danieldiaz]

10:37
Instead of

cabal repl --build-depends=scientific

we could also do

cabal install --lib --package-env . scientific
ghci

(scientific will only be visible in the local folder)

[eddiemundo]

The evaluation plugin probably didn't work because the file didn't compile.