remove Realized

[MaximilianAlgehed]

@abailly-iohk how do I make the formatter ruin my code? Didn't we used to have a script for that somewhere?

Checklist:

• Check source-code formatting is consistent

[ghost]

Love removing code! Minor suggestion to deprecate instead of remove Realized

[ghost]

how do I make the formatter ruin my code? Didn't we used to have a script for that somewhere?

No, we just use fourmolu with the provided fourmolu.yaml file. what's the problem?

[ghost]

Ah it seems the CI is failing on cabal formatting 🤷

[jorisdral]

Since Realized would be removed, is there a suggested alternative for running tests in a monad other than IO?

[MaximilianAlgehed]

Since Realized would be removed, is there a suggested alternative for running tests in a monad other than IO?

Yes, if you're using IOSim the method is to index the state type by the monad, so if before you had ThreadId in the state type and you were running in IOSim s you'd now have ThreadId m in the state type instead.

[jorisdral]

Since Realized would be removed, is there a suggested alternative for running tests in a monad other than IO?

Yes, if you're using IOSim the method is to index the state type by the monad, so if before you had ThreadId in the state type and you were running in IOSim s you'd now have ThreadId m in the state type instead.

👍 I figured this was probably going to be the suggested method. Thanks!

[MaximilianAlgehed]

I'll remember to put something in the release notes about this!

[MaximilianAlgehed]

@abailly-iohk can we agree - in the light of @jorisdral's comment about -werror - that perhaps deprecating isn't necessary and we should just rip the band-aide off?

[ghost]

Ah yes, I don't mind although I could argue that an error triggered by -Werror does not qualify as a breaking change and PVP is just wrong, but I won't :) Please revert my change and go ahead removing it

[MaximilianAlgehed]

@abailly-iohk any clue what's up with CI?

[ghost]

@abailly-iohk any clue what's up with CI?

Some nix f**ery 🤷

[jorisdral]

Since Realized would be removed, is there a suggested alternative for running tests in a monad other than IO?

Yes, if you're using IOSim the method is to index the state type by the monad, so if before you had ThreadId in the state type and you were running in IOSim s you'd now have ThreadId m in the state type instead.

👍 I figured this was probably going to be the suggested method. Thanks!

So one thing this does not address is how to handle actions where the result type mentions the monad m. quickcheck-dynamic requires the result types of actions to be Typeable, but for IOSim s to be typeable, s has to be typeable, which is not the case for ST-style monad runners (unless you use something like stToIO). The upside of the Realized type family was that the action type and realized result could be different, and then we didn't have to have this problem. Any other ideas on how to mitigate this, except for using something like stToIO?