Software graphs and programmer awareness

TL;DR

This paper investigates the structure of dependency graphs in object-oriented software, revealing distinct degree distributions and proposing a model that attributes these patterns to programmer awareness of type relationships.

Contribution

It introduces a simple generative model explaining the differing in-degree and out-degree distributions based on programmer awareness, validated across multiple Java applications.

Findings

In-degree distribution resembles a power-law.

Out-degree distribution resembles an exponential.

Model aligns well with observed data across diverse applications.

Abstract

Dependencies between types in object-oriented software can be viewed as directed graphs, with types as nodes and dependencies as edges. The in-degree and out-degree distributions of such graphs have quite different forms, with the former resembling a power-law distribution and the latter an exponential distribution. This effect appears to be independent of application or type relationship. A simple generative model is proposed to explore the proposition that the difference arises because the programmer is aware of the out-degree of a type but not of its in-degree. The model reproduces the two distributions, and compares reasonably well to those observed in 14 different type relationships across 12 different Java applications.