Collaborative Identification of Code Smells: A Multi-Case Study

To Collaborate or Not to Collaborate?: Improving the Identification of Code Smells

“…Figure 1.1 illustrates how developers working collaboratively, whom we refer to as collaborators, may benefit from the exchange of knowledge on the identification of code smells. The figure provides an example of knowledge exchange extracted from a real development scenario reported in our recent study (63). In the context of this thesis, the developer's knowledge may include information about the purpose of the software system, details on the code structure, or definitions of code smell types.…”

“…In fact, the aforementioned example suggests that the collaborators actually benefit of discussions when identifying a code smell. Thus, the collaborative smell identification emerges as a possible way to improve the e ectiveness of developers otherwise working in isolation, which has been shown to be limited for several reasons, such as the inherent di culty to confirm of refute a code smell suspect (27,44,55,63).…”

“…Although the collaborative smell identification potentially improves the developers' e ectiveness, it is rarely applied by organizations due to the following reasons (43,60,63). First, there is limited knowledge of how to conduct collaborative smell identification (63). Second, due to the limited knowledge, organizations assume that collaborative smell identification does not worth when compare to individual smell identification (43).…”

“…Code elements represent basic units of the software system decomposition, such as classes or methods (30,37). Regardless of the code element a ected by a code smell, its identification is more di cult than it is usually assumed or advertised (27,63). In fact, the identification of code smells in practice consists of three basic steps, as follows (63,65).…”

“…In addition, other studies observe that the identification of code smells may Chapter 1. Introduction 15 become di cult or error-prone when performed by single developers (17,63). Overall, there is empirical evidence that the identification of code smells by single developers has shortcomings that should be addressed.…”