CSCI 2720 Lists II Eileen Kraemer University of Georgia September 13, 2005 Stack : Linked Memory • Push X0 • Push X1 • Push X2 X2 X1 X0 Stack: Linked Memory • function MakeEmptyStack():ptr return • function IsEmptyStack(ptr L):bool return L = • Function Top(ptr L):info if IsEmptyStack(L) then error else return Info(L) Stack: Linked Memory • Function Pop(locative L):info if IsEmptyStack(L) then error else X <- Top(L) L <= Next(L) return x Stack:Linked Memory • Procedure Push(info x, locative L): P <- NewCell(Node) Info(P) <- x Next(P) <- L L <= P What is a locative? • A “new data type that makes the coding of algorithms smoother” • Elegant to write • Difficult to implement What is a locative?? • Like a pointer variable (most of the time) • Key(P), Next(P) .. • In assignments, keeps track of value assigned, and the place in memory that it came from … • P <= Q … updates value of P, and what used to point to P now points to Q … • In practice, we do this by using trailing pointers Queue: Linked Memory • Function MakeEmptyQueue():ptr L <- NewCell(Queue) Front(L) <- Back(L) <- return L • Function IsEmptyQueue(ptr L):boolean return Front(L) = Queue: Linked Memory • Function Enqueue(info x, ptr L): P <- NewCell(Node) Info(P) <- x Next(P) <- If IsEmptyQueue(L) then Front(L) <- P else Next(Back(L)) <- P Back(L) <- P Queue: Linked Memory • Function Dequeue(ptr L):info If IsEmptyQueue(L) then error else X <- Info(Front(L)) Front(L) <- Next(Front(L)) If Front(L) = then Back (L) <= return x Queue: Linked Memory • Function Front(ptr L): info if IsEmptyQueue(L) then error else return Info(Front(L)) Stacks and Recursion • Recursion • Pro: expressive power • Cost: overhead = time + memory • Stacks used in implementing recursion • Works because subprogram invocations end in the opposite order from their beginning (LIFO property) Classical Example of Recursive Algorithm: The Towers of Hanoi • The Legend. In an ancient city in India, so the legend goes, monks in a temple have to move a pile of 64 sacred disks from one location to another. The disks are fragile; only one can be carried at a time. A disk may not be placed on top of a smaller, less valuable disk. And, there is only one other location in the temple (besides the original and destination locations) sacred enough that a pile of disks can be placed there. .jedi Source: http://www.math.toronto.edu/mathnet/games/towers.html The Towers of Hanoi Source: http://www.mathematik.uni-muenchen.de/~hinz/tower.jpg The Towers of Hanoi • How should the monks proceed? • Will they make it? The Towers of Hanoi • Recursive solution! 1. 2. For N = 0 do nothing Move the top N-1 disks from Src to Aux (using Dst as an intermediary peg) Move the bottom disks from Src to Dst Move N-1 disks from Aux to Dst (using Src as an intermediary peg) 3. 4. • The first call: Solve(3, 1, 2, 3) The Towers of Hanoi 1. 2. 3. 4. 5. 6. 7. Move Move Move Move Move Move Move from Src to Dst from Src to Aux from Dst to Aux from Src to Dst from Aux to Src from Aux to Dst from Src to Dst Towers of Hanoi applet • http://www.mazeworks.com/hanoi / The Towers of Hanoi • How much time will monks spend? ¹ • For one disk, only one move is necessary • For two disks, we need three moves • For n disks??? The Towers of Hanoi • Let Tn denote the number of moves needed to move n disks. • T1 = 1 • T2 = 3 • Tn = 2*Tn-1 + 1 The Towers of Hanoi • Let Tn denote the number of moves needed to move n disks. • T1 = 1 • T2 = 3 • Tn = 2*Tn-1 + 1 • Tn = 2n - 1 • How to prove it? Induction! • Base case: T1 = 1 = 21 - 1. OK! • Inductive hypothesis: Tn = 2n -1 • Inductive step: we show that: Tn+1=2n+1-1. Tn+1=2*Tn+1=2*(2n-1)+1= =2n+1-2 + 1= 2n+1-1. QED! Towers of Hanoi • Suppose it takes one minute for a monk to move a disk. The whole task hence would take 264-1 minutes = (210)6*24-1 minutes ≈ (103)6*15 minutes ≈ 25*1016 hours ≈ 1016 days = 1000000000000000 days ≈ the age of universe Recursion • Sierpiński triangle • Wacław Sierpiński – Polish mathematician 1882-1969 Sierpiński Triangle • Draw a black triangle • Draw a white triangle in the middle of the triangle. • Call the procedure for three left black triangles Tail recursion • A special form of recursion in which the last operation of a function is a recursive call. • The recursion may be optimized away by executing the call in the current stack frame and returning its result rather than creating a new stack frame. Recursive form int max_list(list l, int max_so_far) { if (null == l) return max_so_far; if (max_so_far < head(l)) return max_list(tail(l), head(l)); else return max_list(tail(l), max_so_far); } Note …. • The return value of the current invocation is just the return value of the recursive call. • A compiler could optimize it so it doesn't allocate new space for l and max_so_far on each invocation or tear down the stack on the returns. Iterative form int max_list(list l, int max_so_far) { for (;;) { if (null == l) return max_so_far; if (max_so_far < head(l)) { max_so_far = head(l); l = tail(l); } else { max_so_far = max_so_far; l = tail(l); } }} Tail recursion • Now no need to allocate new memory for the parameters or get rid of it during the returns, so this will run faster. • This example simple enough to do by hand ---much harder for complex recursive data structures, such as trees. • If compiler is good enough to find and rewrite tail recursion, it will also • collapse the loop test • eliminate the assignment of max_so_far to itself, • hoist the assignment of l after the test Final form … int max_list(list l, int max_so_far) { while (null != l){ if (max_so_far < head(l)) { max_so_far = head(l); } l = tail(l); } return max_so_far; } SLL -- Singly Linked Lists • Pro: • Insertion: (1) • Access(L,I) -- can’t do it in O(1) … • What is the running time of access??? (more later) • … but given an item in the list, can get to the next item in O(1) … which can give us a traversal (visit each item once) in O(n) Simple List traversal Procedure Traverse(ptr P): // visit nodes of SLL, starting at P While P != { Visit(Key(P)) P <- Next(P) } // if list is already in lexicographic order, // will give in-order traversal “Trickier Traversals” Example (zig-zag scan) <canary, cat, chickadee, coelacanth, collie, corn, cup> K = crabapple • Want to find, given word w, the last word in L that alphabetically precedes w and ends with same letter as w One approach … • Keep both forward and back pointers: canary <=>cat <=>chickadee <=> coelacanth <=>collie <=>corn <=>cup • Pro: (1) from any element to predecessor • forward to cup, backward to collie • Con: 2x the pointer memory, all the time Another approach • Use trailing pointers Function FindLast(ptr L, key w): key // find last word in list L ending w/same letter as w // return f there is no such word P <- L Q <- While P != and Key(P) < w do if Key(P) ends with same letter as w then Q <- P P <- Next(P) If Q = then return else return Key(Q) Back ptrs v. Trailing ptrs • For our example, trailing ptrs saves memory • One extra pointer, only while searching • With other search conditions, can be too complex, code too specialized … Stack based … • Start at beginning of list • Stack pointers to all cells during forward traversal • Pop pointers from stack to do backward traversal Link inversion traversal • Idea: actually place stack into the list itself • “turn around” the next pointers • Temporarily destroys linked list Link Inversion Traversal StartTraversal(L): P Q L Forward(P,Q): P Q Q Next(Q) Next(Q) P Back(P,Q): P Next(P) Next(P) Q Q P Doubly Linked Lists • Coming soon ….