Overcoming the Prevalent Decomposition of Legacy Code

Abstract

The potential benefits of advanced separation of concerns (ASOC) techniques are well known and many programmers find the idea of using them appealing. For new software engineering projects these modularization mechanisms offer guidelines of how to structure the system modules. But how can legacy systems profit from them? Code related to concerns not represented in the current modularization has to be carefully identified and extracted while preserving system integrity. This paper presents a refactoring tool that aids in the extraction of concerns that are ill-represented in the prevalent OOP decomposition 1. Mining for Concerns While contemporary modularization techniques such as OOP have proven to be successful, their approach of modularizing software systems according to a single concern is inherently insufficient and might not provide enough structure for developing complex systems 6, 7, 8. Concerns not represented in the current system decomposition can decrease the code quality, as they have to be pressed into the primary decomposition. We call such concerns hidden concerns (HCs). Code related to these concerns can show two symptoms of poor modularity: it can be scattered over the whole project or it can be tangled with other code. Code tangling is a state where lines related to different concerns are interwoven. ASOC techniques promise to overcome these problems by providing constructs to represent otherwise hidden concerns. However, regardless of which ASOC technique is used, software developers face the same problems when applying these paradigms to legacy systems: How to identify and extract the code related to a hidden concern? Due to the scattered nature of hidden concerns, searching for them in existing code is a non-trivial task.