Brian Beckman: The Zen of Stateless State - The State Monad - Part 2

Download this episode

Download Video


Concurrency is a problem that faces all developers as we move to the age of ManyCore processor architectures. Managing state is an important aspect of programming generally and for parallel programming especially. The great Brian Beckman demonstrates three ways of labeling a binary tree with unique integer node numbers: (1) by hand, (2) non-monadically, but functionally, by threading an updating counter state variable through function arguments, and (3) monadically, by using a partially generalized state-monad implementation to handle the threading via composition. Of course during this lesson from one of the masters of mathematical programming, we wind through various conversational contexts, but always stay true to the default topic in a stateful monadic way (watch/listen to this piece to understand what this actually means Smiley)

This is another great conversation with astrophysicist and programming master Brian Beckman. Brian is one of the true human treasures of Microsoft. If you don't get mondas, this is a great primer. Even if you don't care about monadic data types, this is worth your time, especially if you write code for a living. This is part 2 of a 2 part series.

See part 1 here. 

Below, you will find several exercises for generalizing the constructions further. Here are the source files you need for playing with these algorithms in visual studio or your favorite Haskell environment. Brian will monitor this thread so start your coding engines!!

Exercise 1: generalize over the type of the state, from int
to <S>, say, so that the SM type can handle any kind of
state object. Start with Scp<T> --> Scp<S, T>, from
"label-content pair" to "state-content pair".

Exercise 2: go from labeling a tree to doing a constrained
container computation, as in WPF. Give everything a
bounding box, and size subtrees to fit inside their
parents, recursively.

Exercise 3: promote @return and @bind into an abstract
class "M" and make "SM" a subclass of that.

Exercise 4 (HARD): go from binary tree to n-ary tree.

Exercise 5: Abstract from n-ary tree to IEnumerable; do
everything in LINQ! (Hint: SelectMany).

Exercise 6: Go look up monadic parser combinators and
implement an elegant parser library on top of your new
state monad in LINQ.

Exercise 7: Verify the Monad laws, either abstractly
(pencil and paper), or mechnically, via a program, for the
state monad.

Exercise 8: Design an interface for the operators @return
and @bind and rewrite the state monad so that it implements
this interface. See if you can enforce the monad laws
(associativity of @bind, left identity of @return, right
identity of @return) in the interface implementation.

Exercise 9: Look up the List Monad and implement it so that it implements the same interface.

Exercise 10: deconstruct this entire example by using
destructive updates (assignment) in a discipline way that
treats the entire CLR and heap memory as an "ambient
monad." Identify the @return and @bind operators in this
monad, implement them explicitly both as virtual methods
and as interface methods.



Available formats for this video:

Actual format may change based on video formats available and browser capability.

    The Discussion

    • User profile image
      Can anyone (Brian?) suggest a good book(s) as a introduction to category theory? I'm really looking for just enough to underpin monad theory.
    • User profile image
      perhaps somewhat off-topic from what .tom wanted, but the wikipedia pages on monadology:
      and monism
      was helpful to me to get my head around what the heck a monad "is" (it turns out its pretty hard to define Smiley ) and are pretty good too Smiley
    • User profile image


      // F# code

      open System
      open System.Text

      // State Monad

      //  label a binary tree to demonstrate state-monad
      //  implement non-monadically and monadically

      type Tree<'a> =
      | Leaf of string*'a
      | Branch of Tree<'a>*Tree<'a>  

      // prints binary tree
      // val printTree : Tree<'a> -> unit
      let printTree(a)=
          let rec print(a,level)=
                  let emptyString =new String(' ',level*2)
                  printfn "%s" emptyString
                  match a with
                  |Leaf (sym,e)-> Console.Write(emptyString)
                                  Console.Write("Leaf: "+sym+" ")
                  |Branch (left,right) -> Console.Write(emptyString)
      //non-monad version
      let rec labelTreeNM(t,s) =
               match t with
               |Leaf(sym,_)-> let l=Leaf(sym,s)
               |Branch(left,right)-> let(sL,nLeft)=labelTreeNM(left,s)

      let demoTree=Branch(Leaf("A",0),Branch(Leaf("B",0),Branch(Leaf("C",0),Leaf("D",0))))

      let (_,demoTreeNM)=labelTreeNM(demoTree,0)

      // monad version
      type State<'s,'a> = State of ('s ->'s*'a)

      ////type StateMonad =
      //  class
      //    new : unit -> StateMonad
      //    static member
      //      Bind : sm:State<'a,'b> * f:('b -> State<'a,'c>) -> State<'a,'c>
      //    static member Return : a:'a -> State<'b,'a>
      //  end

      type StateMonad() =
              static member Return(a) = State (fun s -> s, a)
              static member Bind(sm,f) =
                 State (fun s0 ->let (s1,a1)= match sm with
                                              | State h -> h s0
                                 let (s2,a2)= match f a1 with
                                              | State h->h s1

      // succinct functors for state monad

      //val ( >>= ) : State<'a,'b> -> ('b -> State<'a,'c>) -> State<'a,'c>
      let (>>=)m  f = StateMonad.Bind( m, f)
      //val Return : 'a -> State<'b,'a>
      let Return =StateMonad.Return

      // Tree<'a> -> State<int,Tree<int>>
      let rec mkMonad(t)
                =match t with
                 |Leaf(sym,_) -> State(fun s->((s+1),Leaf(sym,s)))
                 |Branch(oldL,oldR)-> mkMonad(oldL)>>=
                                        (fun newL->mkMonad(oldR) >>=
                                                    fun newR->Return(Branch(newL,newR)))

      // monad version
      let monadLabel(t,s)= let(nS,nT)=  match mkMonad(t) with
                                        | State f-> f(s)

      let mTree=monadLabel(demoTree,0)



    • User profile image

      Look on YouTube.  Category theory is used in Linear Algebra and most importantly quantum mechanics.  You I suggest the Stanford University Open Course ware on Quantum Mechanics, there is also probability some good Linear Algebra open course wear at Stanford as well.  Quantum Mechanics II Lecture series starts with a good opener on linear algebra, as it pertains to QED.  All linear algebra operations are function operations of observable events.  It does get much into category theory and complex math, but it is the 'real' stuff that is observable and that is what the course shows how to design mathematical proofs of.  Here is the link to the courses:


      Lecture 1 | Modern Physics: Quantum Mechanics (Stanford) 



      General Link:


      Linerar Algerbra:

      Lecture 1 | Introduction to Linear Dynamical Systems 




    Comments closed

    Comments have been closed since this content was published more than 30 days ago, but if you'd like to continue the conversation, please create a new thread in our Forums, or Contact Us and let us know.