[Haskell-cafe] Sequencing Operations in a Monad

[Ronald Guida]

SevenThunders wrote:
 > OK so check out what really happens with liftM2.  Suppose I have an IO
 > containing an involved matrix computation called s.  For simplicity 
we might
 > assume that
 >
 > s :: IO (Int)  
 >
 > and the Int is an index into an array containing a bunch of matrices in C
 > land.  Assume that s is determined by a succession of many IO operations
 > that have lots of side effects and are fairly computationally intensive.
 > Also assume that s is unevaluated.
 >
 > Now do an operation like
 >
 > q = liftM2 MultMatrix s s
 >
 > What happens is that s is 'evaluated' twice when q is evaluated
 >
 > e.g.
 > do
 >     qint <- q
 >
 >
 > That becomes evident when we look at liftM2's definition
 > liftM2 f  =  \a b -> do { a' <- a; b' <- b; return (f a' b') }
 >
 > the statements
 > a' <- a   and b' <- b will cause s to be evaluated twice.
 >
 > Therein lies my problem.

Here's your solution:

  do
     -- Compare this to liftM2 and your definition of q
     s' <- s   -- this evaluates s once and for all
     qint <- return $ MultMatrix s' s'

-- Ron