Distinguishing Identifiers Using Code Inspection

Sujata S. Deshmukh, S. Pratap Singh

Abstract: Software quality is not defined in terms of quality attributes but instead must be inferred from characteristics that correlate to quality attributes and defect attributes. Readability of source code in one such attribute. Software characteristics have been identified by empirical research, which correlate well to source code. One of such software characteristic is choice of identifier name. Identifier names are primary means by which source code authors communicate their concepts to their readers. They are key mechanisms by which readers of source code such as maintenance programmers access and understands source code. There exists a relationship between flawed identifier names and software maintenance cost. Studying such a relationship is useful to assess whether naming conventions have impact on maintenance effort and to gain a deeper and finer-grained understanding of which program comprehension issues lead to code quality problems.We present in this paper the indentifier analysis method that can identify flaws in Java identifier names. We found that percentage of using compound word identifier names is significantly more than simple word identifier names. Also percentage of compound word identifier names with no dictionary meaning is significant at variable, class and method level identifiers. It unnecessarily increase the splitting overhead of compound word identifier names before finding their dictionary meaning without any clear benefit in readability of source code.

Keywords: Code Inspection, identifier analysis