The changes
function has type Frameworks t => Behavior t a -> Moment t (Event t (Future a))
. Future
is abstract and there is only one function that consumes it (reactimate'
).
However, I can easily write the following function:
changes' :: Frameworks t => Behavior t a -> Moment t (Event t a)
changes' b = fmap (fmap const b <@>) (changes b)
to get a normal (non-Future
) event.
Is something wrong with that function? If not, why does the original changes
function have a more restrictive type?
The function
changes
returns different values than the functionchanges'
that you describe. The key point is the following:Consider a Behavior defined by
stepper
(oraccumB
), which happens to change at time t0. What value does the Behavior have at this moment in time? The answer is that the Behavior takes on the new value for all times that are strictly larger than the time of change, t > t0, and that it still has its old value at time t0. In other words, thechanges'
function returns an event whose values are the old values of the Behavior at the time of change. In contrast, thechanges
function returns the new ("future") values. For various reasons that have to do with recursion, the new values are wrapped in aFuture
type, so that they cannot be accessed until thereactimate'
phase.EDIT: Tobias has drawn a picture for illustration: