A measure of software complexity

In recounting the e6pietls-reas-ens---ereK-£-HSC-S f-Gfour tr-a-ditional problems in developing and maintaining computer software, many authorities have mentioned complexity (e.g., References 2, 8, 15). The complexity pointed to sometimes is the inherent complexity of the jobs the computer is to do; sometimes it is the complexity of the systems or programs that direct the computer to do the jobs. Yet thoughful observers have also long noticed that the inherent complexity of jobs may differ greatly from the apparent complexity of the software for those same jobs ~ e.g., References 26, 29). Anyone who has had the opportumty to study the diversity of approaches, designs, and codes resulting when different people independently produce software for the same job, keenly senses the difference between apparent software complexity and inherent job complexity (e.g., Reference 25). Intuition and common sense generally agree that software which appears simple is superior to software that appears complex, whatever the inherent complexity of the job (e.g., Reference 20). This position is in fact incorporated in the appraisal guidelines of structured design as "simplicity. " 23,24 But applying intuition and common sense is not really sufficient to obtain consistently simple software. What is needed is objective, quantitative, reliable, valid and convenient ways of measuring either the complexity or the simplicity in software. To that end, a number of proposals have been advanced. 1,7,11,12,17,19,21,22,26,32 This paper proposes an alternative measure of software complexity. The background of the measure is briefly given, and its computational procedure described. Then it is applied to a given software design of a small modular structured program. Mterward, the measure is compared with other alternative measures and with programmer ratings of the program. The paper closes with a discussion of the validity of the proposed measure of software complexity.