ghci's :info and :browse break encapsulation

[Feuerbach]

In ghci, :info shows representations of abstract data types. I consider this undesirable, since it breaks abstraction and confuses users.

Example:

Prelude> :m +Data.Unique 
Prelude Data.Unique> :i Unique 
newtype Unique = Data.Unique.Unique Integer
  	-- Defined in `Data.Unique'

I would expect it to behave in the same way as Haddock does, i.e. just show "data Unique".

Same about the :browse command.

[igloo]

Hmm, so I guess you'd get:

Prelude> :i IO
data IO a
        -- Defined in ‛GHC.Types’

Prelude> :i GHC.IO.IO
newtype IO a
  = GHC.Types.IO (GHC.Prim.State# GHC.Prim.RealWorld
                  -> (# GHC.Prim.State# GHC.Prim.RealWorld, a #))
        -- Defined in ‛GHC.Types’

and presumably you would still get:

$ ghci Unique.hs
*Data.Unique> :i Unique
newtype Unique = Unique Integer         -- Defined at Unique.hs:42:9

I'm rather ambivalent about this.

[Feuerbach]

Oh. Apparently scoping rules in ghci are more complicated than I imagined (re GHC.IO.IO). Where can I read more about them?

[igloo]

Note that GHC.IO exports the internals of the IO type:

module GHC.IO (
    IO(..), ...
...

Scope rules are defined in  http://www.haskell.org/onlinereport/haskell2010/ (probably the Modules chapter in particular).

[Feuerbach]

Sorry, I wasn't clear. I am puzzled by the fact that in a bare ghci session, without importing any additional modules except Prelude,

1) The type IO is available under the name GHC.IO.IO 2) The data constructor GHC.IO.IO is in scope

In ordinary Haskell, you'd need to import GHC.IO for that to be the case. Hence my question.

[igloo]

Ah, I see. That's documented in the user's guide here:

http://www.haskell.org/ghc/docs/latest/html/users_guide/interactive-evaluation.html#ghci-import-qualified

http://www.haskell.org/ghc/docs/latest/html/users_guide/ghci-invocation.html#id486684