Conversation

The more I read Charity Major's threads the less confident I am that types as we use them today are a net win.

Hillel

@hillelogram

Mar 31, 2019

I'd be interested in hearing more about this. Which threads make you feel that way, and why?

deech

@deech

Mar 31, 2019

Grab me for a beer for a better take but basically runtime monitoring in langs like Haskell is light years behind eg. Smalltalk.

Ranjit Jhala

@RanjitJhala

Mar 31, 2019

The issue is not with types at the source but with GHC’s typed _intermediate representations_ which make it difficult to write transforms that make such monitoring possible in other languages.

Replying to

and

Could you expand on why they make it difficult? Is it the fact they are typed? Or the fact that the IRs are in there in the first place? Or something else? Asking because I’m rather interested in doing type preserving compilation:

5:04 AM · Apr 1, 2019Twitter for iPhone

Replying to

and

It is tricky to ensure that every transform preserves typed-ness, e.g. difficult to write a transform that just "traces" the inputs and outputs for each function call (a trivial thing with a python "decorator"), as the inputs must have "Show" instances etc.