How do Software Practitioners Write Programs? An Empirical Study

Abstract

Findings of a research investigation into problem decomposition strategies employed by software practitioners during program design are introduced. A cognitive model of designer behavior advances the view that program design is a problem solving task involving repeated application of decomposition and elaboration. The former activity is viewed as goal generation, while the latter activity is considered to consist of the allocation of known ‘clusters’ or plans to the existing decomposition structure. The paper argues that problem decomposition is not performed in an idealized top-down manner, but is influenced by availability effects caused by problem specification characteristics and subject familiarity with component parts of the problem. In addition, it argues that the elaboration of a decomposition involves either the retrieval of known process action clusters or the transformation and re-categorization of information into new clusters. These characteristics of the model provide specific experimental hypotheses. Results from an observational experiment evaluating these hypotheses provide empirical support for the model. The results obtained are discussed in relation to other empirical investigations into program design and directions for further experimentation are presented.