An efficient high-quality hierarchical clustering algorithm for automatic inference of software architecture from the source code of a software system

TL;DR

This paper presents a high-quality hierarchical clustering algorithm that automatically infers software architecture from source code, enabling better comprehension and maintenance of large, complex object-oriented systems.

Contribution

The paper introduces a novel efficient hierarchical clustering algorithm that reconstructs software architecture at the subsystem level from source code, aiding comprehension and maintenance.

Findings

Reconstructs architectural diagrams from source code automatically

Reduces software maintenance costs significantly

Enables understanding of complex software systems at subsystem level

Abstract

It is a high-quality algorithm for hierarchical clustering of large software source code. This effectively allows to break the complexity of tens of millions lines of source code, so that a human software engineer can comprehend a software system at high level by means of looking at its architectural diagram that is reconstructed automatically from the source code of the software system. The architectural diagram shows a tree of subsystems having OOP classes in its leaves (in the other words, a nested software decomposition). The tool reconstructs the missing (inconsistent/incomplete/inexistent) architectural documentation for a software system from its source code. This facilitates software maintenance: change requests can be performed substantially faster. Simply speaking, this unique tool allows to lift the comprehensible grain of object-oriented software systems from OOP class-level to subsystem-level. It is estimated that a commercial tool, developed on the basis of this work, will reduce software maintenance expenses 10 times on the current needs, and will allow to implement next-generation software systems which are currently too complex to be within the range of human comprehension, therefore can't yet be designed or implemented. Implemented prototype in Open Source: http://sourceforge.net/p/insoar/code-0/1/tree/