Re: [Haskell-cafe] Fighting the monad stack, MonadIO

Thanks a lot for all explanations! It looks like 'ioAction' is the key to the solution and if the Browser module did not provide/expose this function, no IO actions could be run inside the BrowserAction monad? If yes, is this a general concept/pattern how to hide functionality of a underlying monad, in this case hide IO entirely? Adam On Apr 10, 2008, at 11:02 AM, Judah Jacobson wrote:

On Thu, Apr 10, 2008 at 7:50 AM, Adam Smyczek wrote:

...

For a small webapi binding I try to implement a session like monad by building a stack including BrowserAction from Network.Browser module as following:

newtype RBAction a = RBAction { exec :: ErrorT String (StateT RBState BrowserAction) a } deriving (Functor, Monad, MonadState RBState)

I would like the RBAction to implement MonadIO as well, but fight with the liftIO function for hours now, without success. Any idea how the implementation of liftIO could look like?

In order to make a stack of monads an instance of MonadIO, you need to be able to run IO actions in the innermost monad (BrowserAction). Ideally, that monad itself would be an instance of MonadIO; maybe the authors of HTTP didn't do so because it would add a dependency on the mtl package (which defines the MonadIO class).

Luckily, though, Network.Browser provides ioAction :: IO a -> BrowserAction a which is exactly what you need to write the MonadIO instances yourself.

-- Solution #1 --

...

instance MonadIO BrowserAction where liftIO = ioAction

Then you can add MonadIO to the deriving clause for RBAction.

-- Solution #2 --

...

instance MonadIO RBAction where liftIO = RBAction . lift . lift . ioAction

This pulls IO actions manually through the stack of transformers. It might be better because if the HTTP package ever provided its own instance of MonadIO BrowserAction, that would conflict with #1.

Hope that helps, -Judah