Deriving Typeable for data families

[chak]

See Simon's proposal in the attachment.

[chak]

Spec

[mitar]

Ping.

This would be really great to have. Currently (7.0 from repository) it fails with:

    Can't make a derived instance of
      `Typeable (...)':
      `...' is a type family

But without Typeable I cannot make instance of Data for data types I have associated with my type class.

[simonpj]

OK, I've done this

Mon Nov 15 23:11:46 GMT Standard Time 2010  simonpj@microsoft.com
  * Deriving Typeable changes
  
  * Fix a bug that led to a crash with
      data family T a
      deriving Functor T
  
  * Allow deriving Typeable for data families
      data family T a
      deriving Typeable1 T
  
  * Some refactoring and tidying

    M ./compiler/typecheck/TcDeriv.lhs -47 +58

Simon

[mitar]

Will this be included in 7.0.2?

[simonpj]

We don't add features in patch releases, just fix bugs. There is always pressure to put more features in as the deadline approaches, and we have to stop somewhere.

Still I suppose you could argue that this patch just fixes a bug. Adding it would break no programs. So I don't have a strong opinion. Ian, Simon, I defer to you.

[mitar]

Without it my  library cannot compile. This is why I am pushing for this.

[heatsink]

Replying to mitar:

As a workaround, you can derive from Typeable without this feature by writing your own instance.

data family T a

instance Typeable a => Typeable (T a) where
  typeOf x = mkTyConApp (mkTyCon "T") [typeOf (unT x)]
    where
      unT :: T a -> a
      unT = undefined

[mitar]

This is really equal to what would deriving do?

[mitar]

Wouldn't it be enough to define?

instance Typeable1 T where
  typeOf1 _ = mkTyConApp (mkTyCon "T") []