An investigation into the effects of the counting method used on software science measurements

Abstract

Professor Maurice Halstead of Purdue University first defined a set of properties of algorithms in 1972. The properties are defined in terms of the number of unique operators, unique operands, total operators, and total operands used to express the algorithm. Since 1972, independent experiments have measured various sets of algorithms and have supported Halstead's theories concerning these properties. Also, new properties have been defined and experiments performed to study them.This paper reports a study in which different methods of counting operators and operands are applied to a fixed set of 34 algorithms written in PL/I. Some properties of the algorithms vary significantly depending on the counting method chosen; other properties remain stable. Although no one counting method can be shown to be best, the results do indicate the importance of the counting method to the overall measurement of an algorithm. Moreover, the results provide a reminder of how sensitive some of the measurements are and of how careful researchers must be when drawing conclusions from software science measurements.

DRCE Maintainability Model for Component Based Systems using Soft Computing Techniques

“…Using fuzzy, one can make use of English words as input and output instead of numeric data. These words are also called crisp input or linguistic variable (Elshoff, 1978). Figure 2 demonstrates the input and output of the proposed maintainability model.…”

Rigor in software complexity measurement experimentation

“…Lister [39] To reinforce the danger of unquestioned reliance of correlation coefficients, Lister provides the following table (Table 2) showing relationships between predicted and actual software lengths (N^ and N under Halstead's software science [27]) for a sample of thirty-one PL/1 programs [21], for various definitions of predicted length. Although the correlations are comparable, the errors associated with the predictions vary significantly.…”

APL compared with other languages according to Halstead's theory

“…Elshoff (8) has shown in 1978 that the procedure of counting functions and arguments may have a relatively large influence on the results of Halsteaa's theory. This applies particularly to APL since the theory does not take into account in this language the distinction between primitive functions and operators.…”