An empirical study on the relation between identifiers and fault proneness

Poorly-chosen identifiers have been reportedin the literature as misleading and increasing the programcomprehension effort. Identifiers are composed of terms,which can be dictionary words, acronyms, contractions, orsimple strings. We conjecture that the use of identical termsin different contexts may increase the risk of faults. Weinvestigate our conjecture using a measure combining termentropy and term context-coverage to study whether certainterms increase the odds ratios of methods to be fault-prone.Entropy measures the physical dispersion of termsin a program: the higher the entropy, the more scatteredacross the program the terms. Context coverage measuresthe conceptual dispersion of terms: the higher their contextcoverage, the more unrelated the methods using them.We compute term entropy and context-coverage of termsextracted from identifiers in Rhino 1.4R3 and ArgoUML0.16. We show statistically that methods containing termswith high entropy and context-coverage are more fault-pronethan others.

Uncontrolled Keywords

Source code identifiers, fault models, program comprehension