Program Design II (PK2)

SML Assignment 4 of Spring 2006

The Power of Problem Generalisation

• Improving the time or space complexity of an algorithm.

• Dealing with problems for which a recursively defined algorithm cannot readily be found.

In any recurrence for a divide-and-conquer algorithm, first make explicit the costs of every part of the divide, conquer (recurse), and combine steps, and relate these costs to the relevant parts of the algorithm. Clearly state any assumptions you make. Then maximally simplify the resulting expressions before continuing.

To derive tight asymptotic bounds Θ(...), use the Master Theorem (MT) where possible. If the MT is applicable, then show in detail which case you apply, why it is applicable, and how you apply it. You can assume that the regularity condition holds, if need be. If the MT is not applicable, then first explain why that is so and then use any other suitable theorem or method seen in the course, again giving the full details of your reasoning.

The file binTrees.sml has the beginning of a realisation of a polymorphic abstract datatype (ADT) 'a bTree for binary trees, specified on slides 6.19 to 6.24. The file trees.sml has sample binary trees for your tests and the examples below. Add the following functions to that ADT, but declare it as a non-abstract datatype, just to ease the testing:

1. The function preorder returns the preorder walk, as a list, of a binary tree.
   Example: preorder smallIntTree = [1,2,5,7,3,6,10,13]
   Requirements: Do not use any help functions.
   Can you make your function tail-recursive? Why / Why not?

2. The function preorder' has the same specification as preorder.
   Requirements: Use exactly one (local) help function, to be called preorderAcc, which must be obtained by introducing an accumulator. This technique is also known as performing a descending generalisation on the specification of preorder. Do not use any standard or library functions for lists, not even @.
   Can you make your help function tail-recursive? Why / Why not?

3. The hidden function preorders takes a list of binary trees as an argument and returns the concatenation of the preorder walks of the trees in that list. This specification was obtained by performing a tupling generalisation on the specification of preorder. This technique aims at reducing the number of recursive calls in each clause.
   Example: preorders [smallIntTree,anotherSmallIntTree] = [1,2,5,7,3,6,10,13,2,1,7,5,13,3,6,10]
   Requirements: Do not use any help functions. Do not use any standard or library functions for lists, not even @. Each clause must have at most one recursive call.
   Can you make your function tail-recursive? Why / Why not?

4. The function preorder'' has the same specification as preorder.
   Requirement: Use preorders as a (local) help function.

5. Discuss your answers to questions 1 to 4. In particular, if no tail recursion in the strict sense has been achieved, would things change with an inorder walk or a postorder walk? Would a combination of the two generalisation techniques help to achieve tail recursion in the strict sense with one of the three kinds of walks?

6. Give a recurrence for the running time T of your preorder function on a balanced binary tree. Derive a tight asymptotic bound for T.

7. Give a recurrence for the running time T' of your preorderAcc function on a balanced binary tree. Derive a tight asymptotic bound for T'.

8. Give a recurrence for the running time T'' of your preorders function on a balanced binary tree. Derive a tight asymptotic bound for T''.

9. Discuss your answers to questions 6 to 8.

Submission

• SML functions for Questions 1 to 4 above, in a .sml file, executable under Moscow ML version 2.01, and documented, in English, according to the coding convention of the course.

• Replies, in English, to Questions 5 to 9 above, in a .txt or .pdf or .html file.