Jaguarpaw Blog - How to read Pipes

by Tom Ellis on 29th March 2013

How to read Pipes

Initially I found it rather hard to understand how the Proxy types from the Pipes library work. This is what I discovered. I’m going to use the ProxyCorrect type for the examples.

The data definitions are

data ProxyCorrect a' a b' b m r =
     Proxy { unProxy :: m (ProxyF a' a b' b r (ProxyCorrect a' a b' b m r)) }

-- | The base functor for the 'ProxyCorrect' type
data ProxyF a' a b' b r x
    = Request a' (a  -> x)
    | Respond b  (b' -> x)
    | Pure    r

Then

• a and a' are channels for communication with upstream
• b and b' are channels for communication with downstream
• the primed type variables a' and b' are requests (i.e. the data travels upstream)
  • a' our request to upstream
  • b' is downstream’s request to us
• the unprimed type variables a and b are responses (i.e. the data travels downstream)
  • a is upstream’s response to us
  • b is our response to downstream

thus

• a' and b are what we emit (upstream and downstream respectively)
• a and b' are what we receive (from upstream and downstream respectively)

m is the monad that a' and b are generated in. r is the type of the final return value of the proxy when it terminates.

A crucial point is that

ProxyCorrect a' a b' b m r

is just about to emit a request or response though we don’t know which. If you want a proxy that is ready to receive a request you must use

\b' -> ProxyCorrect a' a b' b m r

and if you want a proxy that is ready to receive a response you must use

\a -> ProxyCorrect a' a b' b m r