Abstract
Despite its prominence as the most widely used programing language
in industry, there are only a small number of publications on
software metrics applied to COBOL. COBSTYLE, a one-pass, line-by
line style analyzer, written in PL/I, is used to analyze 638
COBOL programs. COBSTYLE differs from other style analyzers in
that it assesses penalty points for abuses in style, and considers
to a larger degree, overall program structure. Mean style scores for
23 style characteristics are obtained. The data produced by
COBSTYLE, together with programmer efficiency data, are stat-
istically analyzed, yielding results which empirically demonstrate
the benefits of good programming style. COBSTYLE, scores are shown
to have statistically significant correlations with the following:
(a) overall performance (as measured by students' final course
grade)- at the 0.005 significance level; (b) program correctness-
at the 0.01 significance level; and (c) total debugging time -
at the 0.05 significance level. An important aspect of this study
is the "after-the-fact" nature of the methodology; i.e., none of
the participants in this experiment were aware that the programs
to be submitted to a style analyzer for analysis.
