Re: [Haskell-cafe] ANNOUNCE: pipes-core 0.0.1

On 12-03-11 01:36 PM, Chris Smith wrote:

On Sun, Mar 11, 2012 at 11:22 AM, Mario Blažević wrote:

...

No, idP does terminate once it consumes its input. Your idP>> p first reproduces the complete input, and then runs p with empty input.

This is just not true. idP consumes input forever, and (idP>> p) = idP, for all pipes p.

If it is composed with another pipe that terminates, then yes, the *composite* pipe can terminate, so for example ((q>+> idP)>> p) may actually do something with p. But to get that effect, you need to compose before the monadic bind... so for example (q>+> (idP>> p)) = (q>+> idP) = q. Yes, q can be exhausted, but when it is, idP will await input, which will immediately terminate the (idP>> p) pipe, producing the result from q, and ignoring p entirely.

Sorry. I was describing the way it's done in SCC, and I assumed that pipes and pipes-core behaved the same. But GHCi says you're right:

:{ | runPipe ((fromList [1, 2, 3] >> return []) | >+> (idP >> fromList [4, 5] >> return []) | >+> consume) | :} [1,2,3]

May I enquire what was the reason for the non-termination of idP? Why was it not defined as 'forP yield' instead? The following command runs the way I expected.

:{ | runPipe ((fromList [1, 2, 3] >> return []) | >+> (forP yield >> fromList [4, 5] >> return []) | >+> consume) | :} [1,2,3,4,5]