Default way of executing code in Haskell - performance

In the following generalized code:
nat = [1..xmax]
xmax = *insert arbitrary Integral value here*
setA = [2*x | x <- nat]
setB = [3*x | x <- nat]
setC = [4*x | x <- nat]
setD = [5*x | x <- nat]
setOne = setA `f` setB
setTwo = setC `f` setD
setAll = setOne ++ setTwo
setAllSorted = quicksort setAll
(please note that 'f' stands for a function of type
f :: Integral a => [a] -> [a] -> [a]
that is not simply ++)
how does Haskell handle attempting to print setAllSorted?
does it get the values for setA and setB, compute setOne and then only keep the values for setOne in memory (before computing everything else)?
Or does Haskell keep everything in memory until having gotten the value for setAllSorted?
If the latter is the case then how would I specify (using main, do functions and all that other IO stuff) that it do the former instead?
Can I tell the program in which order to compute and garbage collect? If so, how would I do that?

The head of setAllSorted is necessarily less-than-or-equal to every element in the tail. Therefore, in order to determine the head, all of setOne and setTwo must be computed. Furthermore, since all of the sets are constant applicative forms, I believe they will not be garbage collected after being computed. The numbers themselves will likely be shared between the sets, but the cons nodes that glue them together will likely not be (your luck with some will depend upon the definition of f).

Due to laziness, Haskell evaluates things on-demand. You can think of the printing done at the end as "pulling" elements from the list setAllSorted, and that might pull other things with it.
That is, running this code it goes something like this:
Printing first evaluates the first element of setAllSorted.
Since this comes from a sorting procedure, it will require all the elements of setAll to be evaluated. (Since the smallest element could be the last one).
Evaluating the first element of setAll requires evaluating the first element of setOne.
Evaluating the first element of setOne depends on how f is implemented. It might require all or none of setA and setB to be evaluated.
After we're done printing the first element of setAllSorted, setAll will have been fully evaluated. There are no more references to setOne, setTwo and the smaller sets, so all of these are by now eligible for garbage collection. The first element of setAllSorted can also be reclaimed.
So in theory, this code will keep setAll in memory most of the time, while setAllSorted, setOne and setTwo will likely only occupy a constant amount of space at any time. Depending on the implementation of f, the same may be true for the smaller sets.

Related

SML Syntax Breakdown

I am trying to study SML (for full transparency this is in preparation for an exam (exam has not started)) and one area that I have been struggling with is higher level functions such as map and foldl/r. I understand that they are used in situations where you would use a for loop in oop languages (I think). What I am struggling with though is what each part in a fold or map function is doing. Here are some examples that if someone could break them down I would be very appreciative
fun cubiclist L = map (fn x=> x*x*x) L;
fun min (x::xs) = foldr (fn (a,b) => if (a < b) then a else b) x xs;
So if I could break down the parts I see and high light the parts I'm struggling with I believe that would be helpful.
Obviously right off the bat you have the name of the functions and the parameters that are being passed in but one question I have on that part is why are we just passing in a variable to cubiclist but for min we pass in (x::xs)? Is it because the map function is automatically applying the function to each part in the map? Also along with that will the fold functions typically take the x::xs parameters while map will just take a variable?
Then we have the higher order function along with the anonymous functions with the logic/operations that we want to apply to each element in the list. But the parameters being passed in for the foldr anonymous function I'm not quite sure about. I understand we are trying to capture the lowest element in the list and the then a else b is returning either a or b to be compared with the other elements in the list. I'm pretty sure that they are rutnred and treated as a in future comparisons but where do we get the following b's from? Where do we say b is the next element in the list?
Then the part that I really don't understand and have no clue is the L; and x xs; at the end of the respective functions. Why are they there? What are they doing? what is their purpose? is it just syntax or is there actually a purpose for them being there, not saying that syntax isn't a purpose or a valid reason, but does they actually do something? Are those variables that can be changed out with something else that would provide a different answer?
Any help/explanation is much appreciated.
In addition to what #molbdnilo has already stated, it can be helpful to a newcomer to functional programming to think about what we're actually doing when we crate a loop: we're specifying a piece of code to run repeatedly. We need an initial state, a condition for the loop to terminate, and an update between each iteration.
Let's look at simple implementation of map.
fun map f [] = []
| map f (x :: xs) = f x :: map f xs
The initial state of the contents of the list.
The termination condition is the list is empty.
The update is that we tack f x onto the front of the result of mapping f to the rest of the list.
The usefulness of map is that we abstract away f. It can be anything, and we don't have to worry about writing the loop boilerplate.
Fold functions are both more complex and more instructive when comparing to loops in procedural languages.
A simple implementation of fold.
fun foldl f init [] = init
| foldl f init (x :: xs) = foldl f (f init x) xs
We explicitly provide an initial value, and a list to operate on.
The termination condition is the list being empty. If it is, we return the initial value provided.
The update is to call the function again. This time the initial value is updated, and the list is the tail of the original.
Consider summing a list of integers.
foldl op+ 0 [1,2,3,4]
foldl op+ 1 [2,3,4]
foldl op+ 3 [3,4]
foldl op+ 6 [4]
foldl op+ 10 []
10
Folds are important to understand because so many fundamental functions can be implemented in terms of foldl or foldr. Think of folding as a means of reducing (many programming languages refer to these functions as "reduce") a list to another value of some type.
map takes a function and a list and produces a new list.
In map (fn x=> x*x*x) L, the function is fn x=> x*x*x, and L is the list.
This list is the same list as cubiclist's parameter.
foldr takes a function, an initial value, and a list and produces some kind of value.
In foldr (fn (a,b) => if (a < b) then a else b) x xs, the function is fn (a,b) => if (a < b) then a else b, the initial value is x, and the list is xs.
x and xs are given to the function by pattern-matching; x is the argument's head and xs is its tail.
(It follows from this that min will fail if it is given an empty list.)

Is there a performance difference between head$filter and head$dropWhile with Haskell Strings?

I'm working on lists of "People" objects in Haskell, and I was wondering if there was any difference in performance between head$dropWhile and head$filter to find the first person with a given name. The two options and a snip of the datatype would be:
datatype Person = Person { name :: String
, otherStuff :: StuffTypesAboutPerson }
findPerson :: String -> [Person] -> Person
findPerson n = head $ dropWhile (\p -> name p /= n)
findPerson n = head $ filter (\p -> name p == n)
My thought was, filter would have to compare the full length of n to the full length of every name until it finds the first one. I would think dropWhile would only need to compare the strings until the first non-matching Char. However, I know there is a ton of magic in Haskell, especially GHC. I would prefer to use the filter version, because I think it's more straight-forward to read. However, I was wondering if there actually is any performance difference? Even if it's negligible, I'm also interested from a curiosity standpoint at this point.
Edit: I know I also need to protect from errors with Maybe, etc, but I left that out to simplify the code example.
There are several approaches to the problem
findPerson n = head $ dropWhile (\p -> name p /= n)
findPerson n = head $ filter (\p -> name p == n)
findPerson n = fromJust $ find (\p -> name p == n)
The question also points out two facts:
when x,y are equal strings, == needs to compare all the characters
when x,y are different strings, /= only needs to compare until the first different character
This is correct, but does not consider the other cases
when x,y are equal strings, /= needs to compare all the characters
when x,y are different strings, == only needs to compare until the first different character
So, between == and /= there is no performance winner. We can expect that, at most, one of them will perform an additional not w.r.t. the other one.
Also, all the three implementations of findPerson mentioned above, essentially perform the same steps. Given xs :: [Person], they will all scan xs until a matching name is found, and no more. On all the persons before the match, the name will be compared against n, and this comparison will stop at the first different character (no matter what comparison we use above). The matching person will have their name compared completely with n (again, in all cases).
Hence, the approaches are expected to run in the same time. There might be a very small difference between them, but it could be so small that it would be hard to detect. You can try to experiment with criterion and see what happens, if you wish.

transferring an imperative for-loop into idiomatic haskell

I have some difficulties to transfer imperative algorithms into a functional style. The main concept that I cannot wrap my head around is how to fill sequences with values according to their position in the sequence. How would an idiomatic solution for the following algorithm look in Haskell?
A = unsigned char[256]
idx <- 1
for(i = 0 to 255)
if (some_condition(i))
A[i] <- idx
idx++
else
A[i] = 0;
The algorithm basically creates a lookup table for the mapping function of a histogram.
Do you know any resources which would help me to understand this kind of problem better?
One of the core ideas in functional programming is to express algorithms as data transformations. In a lazy language like Haskell, we can even go a step further and think of lazy data structures as reified computations. In a very real sense, Haskell's lists are more like loops than normal linked lists: they can be calculated incrementally and don't have to exist in memory all at once. At the same time, we still get many of the advantages of having a data type like that ability to pass it around and inspect it with pattern matching.
With this in mind, the "trick" for expressing a for-loop with an index is to create a list of all the values it can take. Your example is probably the simplest case: i takes all the values from 0 to 255, so we can use Haskell's built-in notation for ranges:
[0..255]
At a high level, this is Haskell's equivalent of for (i = 0 to 255); we can then execute the actual logic in the loop by traversing this list either by a recursive function or a higher-order function from the standard library. (The second option is highly preferred.)
This particular logic is a good fit for a fold. A fold lets us take in a list item by item and build up a result of some sort. At each step, we get a list item and the value of our built-up result so far. In this particular case, we want to process the list from left to right while incrementing an index, so we can use foldl; the one tricky part is that it will produce the list backwards.
Here's the type of foldl:
foldl :: (b -> a -> b) -> b -> [a] -> b
So our function takes in our intermediate value and a list element and produces an updated intermediate value. Since we're constructing a list and keeping track of an index, our intermediate value will be a pair that contains both. Then, once we have the final result, we can ignore the idx value and reverse the final list we get:
a = let (result, _) = foldl step ([], 1) [0..255] in reverse result
where step (a, idx) i
| someCondition i = (idx:a, idx + 1)
| otherwise = (0:a, idx)
In fact, the pattern of transforming a list while keeping track of some intermediate state (idx in this case) is common enough so that it has a function of its own in terms of the State type. The core abstraction is a bit more involved (read through ["You Could Have Invented Monads"][you] for a great introduction), but the resulting code is actually quite pleasant to read (except for the imports, I guess :P):
import Control.Applicative
import Control.Monad
import Control.Monad.State
a = evalState (mapM step [0..255]) 1
where step i
| someCondition i = get <* modify (+ 1)
| otherwise = return 0
The idea is that we map over [0..255] while keeping track of some state (the value of idx) in the background. evalState is how we put all the plumbing together and just get our final result. The step function is applied to each input list element and can also access or modify the state.
The first case of the step function is interesting. The <* operator tells it to do the thing on the left first, the thing on the right second but return the value on the left. This lets us get the current state, increment it but still return the value we got before it was incremented. The fact that our notion of state is a first-class entity and we can have library functions like <* is very powerful—I've found this particular idiom really useful for traversing trees, and other similar idioms have been quite useful for other code.
There are several ways to approach this problem depending on what data structure you want to use. The simplest one would probably be with lists and the basic functions available in Prelude:
a = go 1 [] [0..255]
where
go idx out [] = out
go idx out (i:is) =
if condition i
then go (idx + 1) (out ++ [idx]) is
else go idx (out ++ [0]) is
This uses the worker pattern with two accumulators, idx and out, and it traverses down the last parameter until no more elements are left, then returns out. This could certainly be converted into a fold of some sort, but in any case it won't be very efficient, appending items to a list with ++ is very inefficient. You could make it better by using idx : out and 0 : out, then using reverse on the output of go, but it still isn't an ideal solution.
Another solution might be to use the State monad:
a = flip runState 1 $ forM [0..255] $ \i -> do
idx <- get
if condition i
then do
put $ idx + 1 -- idx++
return idx -- A[i] = idx
else return 0
Which certainly looks a lot more imperative. The 1 in flip runState 1 is indicating that your initial state is idx = 1, then you use forM (which looks like a for loop but really isn't) over [0..255], the loop variable is i, and then it's just a matter of implementing the rest of the logic.
If you want to go a lot more advanced you could use the StateT and ST monads to have an actual mutable array with a state at the same time. The explanation of how this works is far beyond the scope of this answer, though:
import Control.Monad.State
import Control.Monad.ST
import qualified Data.Vector as V
import qualified Data.Vector.Mutable as MV
a :: V.Vector Int
a = runST $ (V.freeze =<<) $ flip evalStateT (1 :: Int) $ do
a' <- lift $ MV.new 256
lift $ MV.set a' 0
forM_ [0..255] $ \i -> do
when (condition i) $ do
idx <- get
lift $ MV.write a' i idx
put $ idx + 1
return a'
I simplified it a bit so that each element is set to 0 from the start, we begin with an initial state of idx = 1, loop over [0..255], if the current index i meets the condition then get the current idx, write it to the current index, then increment idx. Run this as a stateful operation, then freeze the vector, and finally run the ST monad side of things. This allows for an actual mutable vector hidden safely within the ST monad so that the outside world doesn't know that to calculate a you have to do some rather strange things.
Explicit recursion:
a = go 0 1
where go 256 _ = []
go i idx | someCondition i = idx : go (i+1) (idx+1)
| otherwise = 0 : go (i+1) idx
Unfolding: (variant of the explicit recursion above)
a = unfoldr f (0,1)
where f (256,_) = Nothing
f (i,idx) | someCondition i = Just (idx,(i+1,idx+1))
| otherwise = Just (0 ,(i+1,idx ))
Loops can usually be expressed using different fold functions. Here is a solution which uses foldl(you can switch to foldl' if you run into a stackoverflow error):
f :: (Num a) => (b -> Bool) -> a -> [b] -> [a]
f pred startVal = reverse . fst . foldl step ([], startVal)
where
step (xs, curVal) x
| pred x = (curVal:xs, curVal + 1)
| otherwise = (0:xs, curVal)
How to use it? This function takes a predicate (someCondition in your code), the initial value of an index and a list of element to iterate over. That is, you can call f someCondition 1 [0..255] to obtain the result for the example from your question.

FP homework. Is it possible to define a function using nested pattern matching instead of auxiliary function?

I am solving the Programming assinment for Harvard CS 51 programming course in ocaml.
The problem is to define a function that can compress a list of chars to list of pairs where each pair contains a number of consequent occurencies of the character in the list and the character itself, i.e. after applying this function to the list ['a';'a';'a';'a';'a';'b';'b';'b';'c';'d';'d';'d';'d'] we should get the list of [(5,'a');(3,'b');(1,'c');(4,'d')].
I came up with the function that uses auxiliary function go to solve this problem:
let to_run_length (lst : char list) : (int*char) list =
let rec go i s lst1 =
match lst1 with
| [] -> [(i,s)]
| (x::xs) when s <> x -> (i,s) :: go 0 x lst1
| (x::xs) -> go (i + 1) s xs
in match lst with
| x :: xs -> go 0 x lst
| [] -> []
My question is: Is it possible to define recursive function to_run_length with nested pattern matching without defining an auxiliary function go. How in this case we can store a state of counter of already passed elements?
The way you have implemented to_run_length is correct, readable and efficient. It is a good solution. (only nitpick: the indentation after in is wrong)
If you want to avoid the intermediary function, you must use the information present in the return from the recursive call instead. This can be described in a slightly more abstract way:
the run length encoding of the empty list is the empty list
the run length encoding of the list x::xs is,
if the run length encoding of xs start with x, then ...
if it doesn't, then (x,1) ::run length encoding of xs
(I intentionally do not provide source code to let you work the detail out, but unfortunately there is not much to hide with such relatively simple functions.)
Food for thought: You usually encounter this kind of techniques when considering tail-recursive and non-tail-recursive functions (what I've done resembles turning a tail-rec function in non-tail-rec form). In this particular case, your original function was not tail recursive. A function is tail-recursive when the flows of arguments/results only goes "down" the recursive calls (you return them, rather than reusing them to build a larger result). In my function, the flow of arguments/results only goes "up" the recursive calls (the calls have the least information possible, and all the code logic is done by inspecting the results). In your implementation, flows goes both "down" (the integer counter) and "up" (the encoded result).
Edit: upon request of the original poster, here is my solution:
let rec run_length = function
| [] -> []
| x::xs ->
match run_length xs with
| (n,y)::ys when x = y -> (n+1,x)::ys
| res -> (1,x)::res
I don't think it is a good idea to write this function. Current solution is OK.
But if you still want to do it you can use one of two approaches.
1) Without changing arguments of your function. You can define some toplevel mutable values which will contain accumulators which are used in your auxilary function now.
2) You can add argument to your function to store some data. You can find some examples when googling for continuation-passing style.
Happy hacking!
P.S. I still want to underline that your current solution is OK and you don't need to improve it!

Can I always convert mutable-only algorithms to single-assignment and still be efficient?

The Context
The context of this question is that I want to play around with Gene Expression Programming (GEP), a form of evolutionary algorithm, using Erlang. GEP makes use of a string based DSL called 'Karva notation'. Karva notation is easily translated into expression parse trees, but the translation algorithm assumes an implementation having mutable objects: incomplete sub-expressions are created early-on the translation process and their own sub-expressions are filled-in later-on with values that were not known at the time they were created.
The purpose of Karva notation is that it guarantees syntactically correct expressions are created without any expensive encoding techniques or corrections of genetic code. The problem is that with a single-assignment programming language like Erlang, I have to recreate the expression tree continually as each sub expression gets filled in. This takes an inexpensive - O(n)? - update operation and converts it into one that would complete in exponential time (unless I'm mistaken). If I can't find an efficient functional algorithm to convert K-expressions into expression trees, then one of the compelling features of GEP is lost.
The Question
I appreciate that the K-expression translation problem is pretty obscure, so what I want is advice on how to convert an inherently-non-functional algorithm (alg that exploits mutable data structures) into one that does not. How do pure functional programming languages adapt many of the algorithms and data structures that were produced in the early days of computer science that depend on mutability to get the performance characteristics they need?
Carefully designed immutability avoids unecessary updating
Immutable data structures are only an efficiency problem if they're constantly changing, or you build them up the wrong way. For example, continually appending more to the end of a growing list is quadratic, whereas concatenating a list of lists is linear. If you think carefully, you can usually build up your structure in a sensible way, and lazy evaluation is your friend - hand out a promise to work it out and stop worrying.
Blindly trying to replicate an imperative algorithm can be ineffecient, but you're mistaken in your assertion that functional programming has to be asymptotically bad here.
Case study: pure functional GEP: Karva notation in linear time
I'll stick with your case study of parsing Karva notation for GEP. (
I've played with this solution more fully in this answer.)
Here's a fairly clean pure functional solution to the problem. I'll take the opportunity to name drop some good general recursion schemes along the way.
Code
(Importing Data.Tree supplies data Tree a = Node {rootLabel :: a, subForest :: Forest a} where type Forest a = [Tree a].)
import Data.Tree
import Data.Tree.Pretty -- from the pretty-tree package for visualising trees
arity :: Char -> Int
arity c
| c `elem` "+*-/" = 2
| c `elem` "Q" = 1
| otherwise = 0
A hylomorphism is the composition of an anamorphism (build up, unfoldr) and a catamorphism (combine, foldr).
These terms are introduced to the FP community in the seminal paper Functional Programming with Bananas, Lenses and Barbed wire.
We're going to pull the levels out (ana/unfold) and combine them back together (cata/fold).
hylomorphism :: b -> (a -> b -> b) -> (c -> (a, c)) -> (c -> Bool) -> c -> b
hylomorphism base combine pullout stop seed = hylo seed where
hylo s | stop s = base
| otherwise = combine new (hylo s')
where (new,s') = pullout s
To pull out a level, we use the total arity from the previous level to find where to split off this new level, and pass on the total arity for this one ready for next time:
pullLevel :: (Int,String) -> (String,(Int,String))
pullLevel (n,cs) = (level,(total, cs')) where
(level, cs') = splitAt n cs
total = sum $ map arity level
To combine a level (as a String) with the level below (that's already a Forest), we just pull off the number of trees that each character needs.
combineLevel :: String -> Forest Char -> Forest Char
combineLevel "" [] = []
combineLevel (c:cs) levelBelow = Node c subforest : combineLevel cs theRest
where (subforest,theRest) = splitAt (arity c) levelBelow
Now we can parse the Karva using a hylomorphism. Note that we seed it with a total arity from outside the string of 1, since there's only one node at the root level. Correspondingly we apply head to the result to get this singleton back out after the hylomorphism.
karvaToTree :: String -> Tree Char
karvaToTree cs = let
zero (n,_) = n == 0
in head $ hylomorphism [] combineLevel pullLevel zero (1,cs)
Linear Time
There's no exponential blowup, nor repeated O(log(n)) lookups or expensive modifications, so we shouldn't be in too much trouble.
arity is O(1)
splitAt part is O(part)
pullLevel (part,cs) is O(part) for grab using splitAt to get level, plus O(part) for the map arity level, so O(part)
combineLevel (c:cs) is O(arity c) for the splitAt, and O(sum $ map arity cs) for the recursive call
hylomorphism [] combineLevel pullLevel zero (1,cs)
makes a pullLevel call for each level, so the total pullLevel cost is O(sum parts) = O(n)
makes a combineLevel call for each level, so the total combineLevel cost is O(sum $ map arity levels) = O(n), since the total arity of the entire input is bound by n for valid strings.
makes O(#levels) calls to zero (which is O(1)), and #levels is bound by n, so that's below O(n) too
Hence karvaToTree is linear in the length of the input.
I think that puts to rest the assertion that you needed to use mutability to get a linear algorithm here.
Demo
Let's have a draw of the results (because Tree is so full of syntax it's hard to read the output!). You have to cabal install pretty-tree to get Data.Tree.Pretty.
see :: Tree Char -> IO ()
see = putStrLn.drawVerticalTree.fmap (:"")
ghci> karvaToTree "Q/a*+b-cbabaccbac"
Node {rootLabel = 'Q', subForest = [Node {rootLabel = '/', subForest = [Node {rootLabel = 'a', subForest = []},Node {rootLabel = '*', subForest = [Node {rootLabel = '+', subForest = [Node {rootLabel = '-', subForest = [Node {rootLabel = 'b', subForest = []},Node {rootLabel = 'a', subForest = []}]},Node {rootLabel = 'c', subForest = []}]},Node {rootLabel = 'b', subForest = []}]}]}]}
ghci> see $ karvaToTree "Q/a*+b-cbabaccbac"
Q
|
/
|
------
/ \
a *
|
-----
/ \
+ b
|
----
/ \
- c
|
--
/ \
b a
which matches the output expected from this tutorial where I found the example:
There isn't a single way to do this, it really has to be attempted case-by-case. I typically try to break them down into simpler operations using fold and unfold and then optimize from there. Karva decoding case is a breadth-first tree unfold as others have noted, so I started with treeUnfoldM_BF. Perhaps there are similar functions in Erlang.
If the decoding operation is unreasonably expensive, you could memoize the decoding and share/reuse subtrees... though it probably wouldn't fit into a generic tree unfolder and you'd need to write specialized function to do so. If the fitness function is slow enough, it may be fine to use a naive decoder like the one I have listed below. It will fully rebuild the tree each invocation.
import Control.Monad.State.Lazy
import Data.Tree
type MaxArity = Int
type NodeType = Char
treeify :: MaxArity -> [Char] -> Tree NodeType
treeify maxArity (x:xs) = evalState (unfoldTreeM_BF (step maxArity) x) xs
treeify _ [] = fail "empty list"
step :: MaxArity -> NodeType -> State [Char] (NodeType, [NodeType])
step maxArity node = do
xs <- get
-- figure out the actual child node count and use it instead of maxArity
let (children, ys) = splitAt maxArity xs
put ys
return (node, children)
main :: IO ()
main = do
let x = treeify 3 "0138513580135135135"
putStr $ drawTree . fmap (:[]) $ x
return ()
There are a couple of solutions when mutable state in functional programming is required.
Use a different algorithm that solves the same problem. E.g. quicksort is generally regarded as mutable and may therefore be less useful in a functional setting, but mergesort is generally better suited for a functional setting. I can't tell if this option is possible or makes sense in your case.
Even functional programming languages usually provide some way to mutate state. (This blog post seems to show how to do it in Erlang.) For some algorithms and data structures this is indeed the only available option (there's active research on the topic, I think); for example hash tables in functional programming languages are generally implemented with mutable state.
In your case, I'm not so sure immutability really leads to a performance bottleneck. You are right, the (sub)tree will be recreated on update, but the Erlang implementation will probably reuse all the subtrees that haven't changed, leading to O(log n) complexity per update instead of O(1) with mutable state. Also, the nodes of the trees won't be copied but instead the references to the nodes, which should be relatively efficient. You can read about tree updates in a functional setting in e.g. the thesis from Okasaki or in his book "Purely Functional Data Structures" based on the thesis. I'd try implementing the algorithm with an immutable data structure and switch to a mutable one if you have a performance problem.
Also see some relevant SO questions here and here.
I think I figured out how to solve your particular problem with the K trees, (the general problem is too hard :P). My solution is presented in some horrible sort of hybrid Python-like psudocode (I am very slow on my FP today) but it doesn't change a node after you create one (the trick is building the tree bottom-up)
First, we need to find which nodes belong to which level:
levels currsize nodes =
this_level , rest = take currsize from nodes, whats left
next_size = sum of the arities of the nodes
return [this_level | levels next_size rest]
(initial currsize is 1)
So in the +/*abcd, example, this should give you [+, /*, abcd]. Now you can convert this into a tree bottom up:
curr_trees = last level
for level in reverse(levels except the last)
next_trees = []
for root in level:
n = arity of root
trees, curr_trees = take n from curr_trees, whats left
next_trees.append( Node(root, trees) )
curr_trees = next_trees
curr_trees should be a list with the single root node now.
I am pretty sure we can convert this into single assignment Erlang/Haskell very easily now.

Resources