[Haskell-cafe] State monad strictness - how?

Udo Stenzel u.stenzel at web.de
Thu Jan 11 07:44:21 EST 2007


Yitzchak Gale wrote:
> You're right, it is not in the docs. I don't think anyone would
> have planned it that way. StateT is strict only because there
> happens to be a line in a do-expression that looks like:
> 		(a, s') <- runStateT m s
> The tuple pattern-match causes the strictness.
> That appears accidental, so it seems to be just an honest bug.

I agree that this is an accident, but the bug is in lazy State, for
three reasons:

- Being strict in the (result,state) pair does not for the evaluation of
  either result or state.  Not being strict could only ever be useful
  for a following action that made no use of either state or result, and
  I have a hard time imagining why you'd ever want to write such a
  beast, let alone in monadic style.  In fact, an unboxed tuple would be
  even better.

- Assuming that the State monad is lazy in the tuple, and you need to be
  strict in the state component, you are hosed.  No amount of 'seq' will
  help you.  On the other hand, were it strict and you needed it to be
  lazy, you could achieve that by manually boxing the data involved.

- (>>=) should also be head strict in the state component.  Again, if
  this is wrong, you can repair it.  If laziness turns out wrong, you
  can't.  Moreover, for most data types that you want to build lazily,
  especially lists, head strictness doesn't make a difference, as long
  as the tail is lazily evaluated.  For data where you need strictness,
  such as integers or tuples of them, having strictness available make
  all the difference.

I'd be fine with laziness being configurable, of course, but if it
isn't, I want strict state.  Come to think of it, it's probably just a
bad idea that _|_ and (_|_,_|_) are different things.


-Udo
-- 
The Seventh Commandments for Technicians:
	Work thou not on energized equipment, for if thou dost, thy
	fellow workers will surely buy beers for thy widow and console
	her in other ways.
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 189 bytes
Desc: Digital signature
Url : http://www.haskell.org/pipermail/haskell-cafe/attachments/20070111/d58a1bd0/attachment-0001.bin


More information about the Haskell-Cafe mailing list