On 9/30/12 7:00 AM, Tillmann Rendel wrote:

Vasili I. Galchin wrote:

...

I would an examples of monads that are pure, i.e. no side-effects.

One view of programming in monadic style is: You call return and >>= all the time. (Either you call it directly, or do notation calls it for you). So if you want to understand whether a monad "has side-effects", you should look at the implementation of return and >>=. If the implementation of return and >>= is written in pure Haskell (without unsafePerformIO or calling C code etc.), the monad is pure.

I'm not sure return and bind will give you the information you seek, however. In order to obey the monad laws, return and bind must be (for all intents and purposes) pure. The place to look for "impurities" is the primitive operations of the monad; i.e., those which cannot be implemented using return and bind. These are the operations which take you out of a free monad and the operations which must be given some special semantic interpretation. -- Live well, ~wren

On Sep 30, 2012 10:56 AM, "Albert Y. C. Lai" wrote:

On 12-09-29 09:57 PM, Vasili I. Galchin wrote:

...

I would an examples of monads that are pure, i.e. no

side-effects.

What does "side effect" mean, to you? Definition?

When discussing monads, at least, a side effect is an effect that is triggered by merely evaluating an expression. A monad is an interface that decouples effects from evaluation.

Because some people say "State has no side effect", and some other people

say "State has side effects". The two groups use different definitions.

_______________________________________________ Haskell-Cafe mailing list Haskell-Cafe@haskell.org http://www.haskell.org/mailman/listinfo/haskell-cafe

On 12-09-30 06:33 PM, Jake McArthur wrote:

When discussing monads, at least, a side effect is an effect that is triggered by merely evaluating an expression. A monad is an interface that decouples effects from evaluation.

I don't understand that definition. Or maybe I do subconsciously. I have s :: State Int () s = do { x <- get; put (x+1) } Is there an effect triggered by merely evaluating s? I have m :: IO () m = if True then putStrLn "x" else putChar 'y' Is there an effect triggered by merely evaluating m? What counts as "evaluate"?

On 12-10-01 05:34 AM, Jon Fairbairn wrote:

"Albert Y. C. Lai" writes:

...

On 12-09-30 06:33 PM, Jake McArthur wrote:

...

When discussing monads, at least, a side effect is an effect that is triggered by merely evaluating an expression. A monad is an interface that decouples effects from evaluation.

I don't understand that definition. Or maybe I do subconsciously.

I have

s :: State Int () s = do { x <- get; put (x+1) }

Is there an effect triggered by merely evaluating s?

I have

m :: IO () m = if True then putStrLn "x" else putChar 'y'

Is there an effect triggered by merely evaluating m?

What counts as "evaluate"?

Evaluation! Consider m `seq` 42. m is evaluated, but to no effect.

Sure thing. So s has no side effect, and m has no side effect. Since they have no side effect, there is no effect to be decoupled from evaluation.