THE SUBPROGRAM PROBLEM FOR SOFTWARE METRIC DESIGN

In the design of software metrics, increasing attention has been given to the complexity arising from the pattern of coupling among subprogr ams. We call it the ''subprogram problem'' - to discover the most effe ctive means whereby one might judiciously assign a measure to the subp rogram calls appearing in the context of tightly-coupled modules. We p hrase our solution in the language of matrix analysis, giving rise to both an exact and an iterative computational approach. The resulting m ethodology is uniform over a wide range of software metrics, integrati ng an accounting for the coupling complexity into the extended meaning of a particular metric. A running example is provided, fully illustra ting the technique.