putting laziness to work
DESCRIPTION
Putting Laziness to Work. Why use laziness. Laziness has lost of interesting uses Build cyclic structures. Finite representations of infinite data. Do less work, compute only those values demanded by the final result. - PowerPoint PPT PresentationTRANSCRIPT
![Page 1: Putting Laziness to Work](https://reader036.vdocument.in/reader036/viewer/2022062410/568164d3550346895dd70338/html5/thumbnails/1.jpg)
Putting Laziness to Work
![Page 2: Putting Laziness to Work](https://reader036.vdocument.in/reader036/viewer/2022062410/568164d3550346895dd70338/html5/thumbnails/2.jpg)
Why use laziness
• Laziness has lots of interesting uses– Build cyclic structures. Finite representations of infinite
data.– Do less work, compute only those values demanded by
the final result.– Build infinite intermediate data structures and actually
materialize only those parts of the structure of interest.• Search based solutions using enumerate then test .
– Memoize or remember past results so that they don’t need to be recomputed
![Page 3: Putting Laziness to Work](https://reader036.vdocument.in/reader036/viewer/2022062410/568164d3550346895dd70338/html5/thumbnails/3.jpg)
Cyclic structures
• cycles:: [Int]• cycles = 1 : 2 : 3 : cycles
1 2 3
![Page 4: Putting Laziness to Work](https://reader036.vdocument.in/reader036/viewer/2022062410/568164d3550346895dd70338/html5/thumbnails/4.jpg)
Cyclic Trees
• data Tree a = Tip a | Fork (Tree a) (Tree a)
• t2 = Fork (Fork (Tip 3) (Tip 4)) (Fork (Tip 9) t2)
Fork
Fork Fork
Tip 3 Tip 4 Tip 9
![Page 5: Putting Laziness to Work](https://reader036.vdocument.in/reader036/viewer/2022062410/568164d3550346895dd70338/html5/thumbnails/5.jpg)
Mutually Cyclic(t3,t4) = ( Fork (Fork (Tip 11) t3) t4 , Fork (Tip 21) (Fork (Tip 33) t3) )
Fork
Tip 11
Tip 21Fork
Fork
Fork
Tip 33
![Page 6: Putting Laziness to Work](https://reader036.vdocument.in/reader036/viewer/2022062410/568164d3550346895dd70338/html5/thumbnails/6.jpg)
Prime numbers and infinite listsprimes :: [Integer]primes = sieve [2..] where sieve (p:xs) = p : sieve [x | x<-xs , x `mod` p /= 0]
![Page 7: Putting Laziness to Work](https://reader036.vdocument.in/reader036/viewer/2022062410/568164d3550346895dd70338/html5/thumbnails/7.jpg)
Dynamic Programming
• Consider the function
fib :: Integer -> Integerfib 0 = 1fib 1 = 1fib n = fib (n-1) + fib (n-2)
LazyDemos> :set +sLazyDemos> fib 301346269(48072847 reductions, 78644372 cells, 1 garbage
collection)
• takes about 9 seconds on my machine!
![Page 8: Putting Laziness to Work](https://reader036.vdocument.in/reader036/viewer/2022062410/568164d3550346895dd70338/html5/thumbnails/8.jpg)
Why does it take so long?
• Consider (fib 6)fib 6
fib 5 fib 4
fib 2
fib 1 fib 0
fib 3
fib 2
fib 1 fib 0
fib 1
fib 3
fib 2
fib 1 fib 0
fib 1
fib 4
fib 2
fib 1 fib 0
fib 3
fib 2
fib 1 fib 0
fib 1
![Page 9: Putting Laziness to Work](https://reader036.vdocument.in/reader036/viewer/2022062410/568164d3550346895dd70338/html5/thumbnails/9.jpg)
What if we could remember past results?
• Strategy– Create a data structure– Store the result for every (fib n) only if (fib n) is
demanded.– If it is ever demanded again return the result in the
data structure rather than re-compute it• Laziness is crucial• Constant time access is also crucial– Use of functional arrays
![Page 10: Putting Laziness to Work](https://reader036.vdocument.in/reader036/viewer/2022062410/568164d3550346895dd70338/html5/thumbnails/10.jpg)
Lazy Arraysimport Data.Arraytable = array (1,5) [(1,'a'),(2,'b'),(3,'c'),(5,'e'),(4,'d')]
• The array is created once• Any size array can be created• Slots cannot be over written• Slots are initialized by the list• Constant access time to value stored in every slot
![Page 11: Putting Laziness to Work](https://reader036.vdocument.in/reader036/viewer/2022062410/568164d3550346895dd70338/html5/thumbnails/11.jpg)
Taming the duplication
fib2 :: Integer -> Integerfib2 z = f z where table = array (0,z) [ (i, f i) | i <- range (0,z) ] f 0 = 1 f 1 = 1 f n = (table ! (n-1)) + (table ! (n-2))
LazyDemos> fib2 301346269(4055 reductions, 5602 cells)
Result is instantaneous on my machine
![Page 12: Putting Laziness to Work](https://reader036.vdocument.in/reader036/viewer/2022062410/568164d3550346895dd70338/html5/thumbnails/12.jpg)
Can we abstract over this pattern?
• Can we write a memo function that memoizes another function.
• Allocates an array• Initializes the array with calls to the function• But, We need a way to intercept recursive calls
![Page 13: Putting Laziness to Work](https://reader036.vdocument.in/reader036/viewer/2022062410/568164d3550346895dd70338/html5/thumbnails/13.jpg)
A fixpoint operator does the trick
• fix f = f (fix f)
• g fib 0 = 1• g fib 1 = 1• g fib n = fib (n-1) + fib (n-2)
• fib1 = fix g
![Page 14: Putting Laziness to Work](https://reader036.vdocument.in/reader036/viewer/2022062410/568164d3550346895dd70338/html5/thumbnails/14.jpg)
Generalizing
memo :: Ix a => (a,a) -> ((a -> b) -> a -> b) -> a -> bmemo bounds g = f where arrayF = array bounds [ (n, g f n) | n <- range bounds ] f x = arrayF ! x
fib3 n = memo (0,n) g n
fact = memo (0,100) g where g fact n = if n==0 then 1 else n * fact (n-1)
![Page 15: Putting Laziness to Work](https://reader036.vdocument.in/reader036/viewer/2022062410/568164d3550346895dd70338/html5/thumbnails/15.jpg)
Representing Graphs
import STimport qualified Data.Array as Atype Vertex = Int
-- Representing graphs:
type Table a = A.Array Vertex atype Graph = Table [Vertex] -- Array Int [Int]
Index for each node
Edges (out of) that index
1
2 3
4 5
78
910
6
1 [2,3]2 [7,4]3 [5]4 [6,9,7]5 [8]6 [9]7 [9]8 [10]9 [10]10
[]
![Page 16: Putting Laziness to Work](https://reader036.vdocument.in/reader036/viewer/2022062410/568164d3550346895dd70338/html5/thumbnails/16.jpg)
Functions on graphs
type Vertex = Inttype Edge = (Vertex,Vertex)
vertices :: Graph -> [Vertex]
indices :: Graph -> [Int]
edges :: Graph -> [Edge]
![Page 17: Putting Laziness to Work](https://reader036.vdocument.in/reader036/viewer/2022062410/568164d3550346895dd70338/html5/thumbnails/17.jpg)
Building Graphs
buildG :: Bounds -> [Edge] -> Graph
graph = buildG (1,10) [ (1, 2), (1, 6), (2, 3), (2, 5), (3, 1), (3, 4), (5, 4), (7, 8), (7, 10), (8, 6), (8, 9), (8, 10) ]
1
2
5
6
47
8
9
10
3
![Page 18: Putting Laziness to Work](https://reader036.vdocument.in/reader036/viewer/2022062410/568164d3550346895dd70338/html5/thumbnails/18.jpg)
DFS and Forests
data Tree a = Node a (Forest a) type Forest a = [Tree a]
nodesTree (Node a f) ans = nodesForest f (a:ans)
nodesForest [] ans = ansnodesForest (t : f) ans = nodesTree t (nodesForest f ans)
• Note how any tree can be spanned • by a Forest. The Forest is not always• unique.
1
2
5
6
47
8
9
10
3
![Page 19: Putting Laziness to Work](https://reader036.vdocument.in/reader036/viewer/2022062410/568164d3550346895dd70338/html5/thumbnails/19.jpg)
DFS
• The DFS algorithm finds a spanning forest for a graph, from a set of roots.
dfs :: Graph -> [Vertex] -> Forest Vertex
dfs :: Graph -> [Vertex] -> Forest Vertexdfs g vs = prune (A.bounds g) (map (generate g) vs)
generate :: Graph -> Vertex -> Tree Vertexgenerate g v = Node v (map (generate g) (g `aat` v))
Array indexing
An infinite cyclic tree
![Page 20: Putting Laziness to Work](https://reader036.vdocument.in/reader036/viewer/2022062410/568164d3550346895dd70338/html5/thumbnails/20.jpg)
Sets of nodes already visited
import qualified Data.Array.ST as Btype Set s = B.STArray s Vertex Bool
mkEmpty :: Bounds -> ST s (Set s)mkEmpty bnds = newSTArray bnds False
contains :: Set s -> Vertex -> ST s Boolcontains m v = readSTArray m v
include :: Set s -> Vertex -> ST s ()include m v = writeSTArray m v True
Mutable array
![Page 21: Putting Laziness to Work](https://reader036.vdocument.in/reader036/viewer/2022062410/568164d3550346895dd70338/html5/thumbnails/21.jpg)
Pruning already visited paths
prune :: Bounds -> Forest Vertex -> Forest Vertexprune bnds ts = runST (do { m <- mkEmpty bnds; chop m ts })
chop :: Set s -> Forest Vertex -> ST s (Forest Vertex)chop m [] = return []chop m (Node v ts : us) do { visited <- contains m v ; if visited then chop m us else do { include m v ; as <- chop m ts ; bs <- chop m us ; return(Node v as : bs) } }
![Page 22: Putting Laziness to Work](https://reader036.vdocument.in/reader036/viewer/2022062410/568164d3550346895dd70338/html5/thumbnails/22.jpg)
Topological Sort
postorder :: Tree a -> [a]postorder (Node a ts) = postorderF ts ++ [a]
postorderF :: Forest a -> [a]postorderF ts = concat (map postorder ts)
postOrd :: Graph -> [Vertex]postOrd = postorderF . Dff
dff :: Graph -> Forest Vertexdff g = dfs g (vertices g)
![Page 23: Putting Laziness to Work](https://reader036.vdocument.in/reader036/viewer/2022062410/568164d3550346895dd70338/html5/thumbnails/23.jpg)
A
B C
D E
G
H
I
J
F
A
B C
D E
G
H
I
J
F
(0,””)
(1,A)(5,A)
(2,CA)
(3,ECA)
(4,HECA)
(6,BA)
(7,DBA)
(8,GDBA)
(9,DBA)
A. Control Flow Grapha :: Graph Char a
B. DFS Labeled Graphb :: Graph Char (Int,[Char])dfsnum :: v->Intdfsnum v = fst(apply b v)dfspath:: v->[v]dfspath v = snd(apply b v)
A
B C
D E
G
H
I
J
F
(A)
(A)(A)
(C)
(C)
(A)
(B)
(B)
(D)
(D)
C. Semi-Dominator Labeled Graphc :: Graph Char Charsemi :: v->vsemi = apply c
A
B C
D G E
J
H
F I
D. Semi-Dominator Graphd :: Graph Char a
A
B C
D G E
J
H
F
I
E. Dominator Graphe :: Graph Char a
dfslabel f map g induce