Conversation

deech
@deech
At least in Haskell in places I've worked the over use of TC's is terrible but that's more because of abysmal records than any category theory concerns.
1
6
deech
@deech
And *de*abstraction is nearly impossible, once you fix the dictionary of a TC and realize down the road an entry doesn't belong you're out of luck. eg. peek/poke should never have been bundled with Storable but here we are ...
3
1
Anthony Cowley
@a_cowley
Apologies for the tangent, but can you expand on your thinking about Storable? It’s probably my most-used type class after Show, and I can’t recall having significant negative feelings about it.
1
3
deech
@deech
Several times I've wanted to bind to C structs that I wanted to read but never write but I'm forced to provide a poke, poke{Elem,Byte}Off. Nowadays I just have c2hs gen up the specific thing I need.
1
3
Anthony Cowley
@a_cowley
I put an error there if I’m not supposed to write. That example does get to the problem of it being near impossible to slice up interfaces to a universally right granularity.
1
2
deech
@deech
It's difficult to get right, but another big problem is Haskell culture heavily incentivizes programming against interfaces and once you've built up an entire ecosystem around them the mistakes become millstones.
1
5
Philippa Cowderoy
@flippac
It doesn't help that the tools for relating typeclass signatures/interfaces are, er, kinda shitty for that job. Strictly nominal, no way to say "this is all I actually use of it"
1
2