Lucid and Efficient Case Analysis

This paper describes a new scheme for building static search trees, using multiway radix search trees. We present this method for code generation of switch statements in imperative languages. We show that, for sparse case sets, the method produces faster code on average than existing methods, requiring O(1) time with a small constant for the average search. We then apply this method to the problem of code generation for generic functions in object-oriented languages, and nd that its use improves clarity as well as eeciency. 1. Introduction Switch statements in C, like case statements in Pascal and Ada, are useful conditional control constructs. These statements represent multiway tree control structures , whereas if statements represent binary tree control structures. In this paper we present a new code generation method for switch statements, which on the average generates faster code for sparse sets of cases than existing methods, and apply it to generic function dispatch in object-oriented languages. The method can be thought of as generating a multiway radix search tree on the case labels, and is referred to as the MRST method throughout this paper. Although the MRST method is appropriate for most searches where a static search tree can be generated, only its application to switch statements is considered here. There has been considerable work in the past ((8], 5], 2], 3] and 6]) regarding the Pascal case statement and code generation. Generating code for switch statements is discussed in 4] and 9]. A scheme similar to MRST, but restricted to binary radix search trees, can be found in 7]. We begin, in section 2, by describing the MRST method of code generation for switch statements. In section 3, we compare it with four other code generation methods. We analyze the methods in terms of storage requirements and time, providing the worst case and expected case scenarios for each method. We then perform an empirical study of the MRST method with a few representative examples. Finally, in section 5, we present an implementation mechanism for generic functions in object-oriented languages which exploits the properties of our MRST method.