Domain model-driven software engineering: A method for discovery of dependency links

Abstract

Context

Dependency management often suffers from labor intensity and complexity in creating and maintaining the dependency relations in practice. This is even more critical in a distributed development, in which developers are geographically distributed and a wide variety of tools is used. In those settings, different interpretations of software requirements or usage of different terminologies make it challenging to predict the change impact.

Objective

is (a) to describe a method facilitating change management in geographically distributed software engineering by effective discovery and establishment of dependency links using domain models; (b) to evaluate the effectiveness of the proposed method.

Method

A domain model, providing a common reference point, is used to manage development objects and to automatically support dependency discovery. We propose to associate (annotate) development objects with the concepts from the model. These associations are used to compute dependency among development objects, and are stepwise refined to direct dependency links (i.e. enabling product traceability). To evaluate the method, we conducted a laboratory-based randomized experiment on two real cases. Six participants were using an implemented prototype and two comparable tools to perform simulated tasks.

Results

In the paper we elaborate on the proposed method discussing its functional steps. Results from the experiment show that the method can be effectively used to assist in discovery of dependency links. Users have discovered on average fourteen percent more dependency links than by using the comparable tools.

Conclusions

The proposed method advocates the use of domain models throughout the whole development life-cycle and is apt to facilitate multi-site software engineering. The experimental study and results suggest that the method is effective in the discovery of dependencies among development objects.

Introduction

Organizations encounter a large number of stakeholders, and have broad geographical distribution alongside a wide variety of tools; cooperative and multi-site software engineering is the norm rather than the exception [11]. The output from a system’s development process ranges from information analysis and architecture to design and code. Thus, during software development, a collection of heterogeneous software artifacts, or so called ‘product fragments’ (i.e., requirements specification, design, code, test scenarios and documentation, etc.) is made.

Distributed development projects have special settings and needs. When groups of developers reside at different sites, different interpretations of product fragment specifications, software requirements or the use of different terminology can lead to incompatible product fragments [55]. Another problem here is the management of product fragment diversity, the discovery and maintenance of dependency links among the product fragments. Possibly, heterogeneous development environments make it difficult to integrate tools and establish traceability links [12]. Dependency management often suffers from an extensive effort and complexity of creating and maintaining the traces [14]. This complexity is a result of [15]: syntactic and semantic differences, informal/semi-formal languages (e.g., requirements, design), different stakeholders being in charge of different software artifacts, increasingly rapid pace of change, etc.

To solve the problem, we propose using domain models not only to guide the design of a software application, but also to actually access and manage the information produced during the software development. The domain model here is a shared conceptualization of a domain and can also be referred to as a domain model. A study by Cataldo et al. [7] found that “traditional software dependencies, such as syntactic relationships, tend to capture a relatively narrow view of product dependencies.” Therefore, we use a domain model as a conceptual and logical representation of a problem domain in software development [33], [47] to establish dependency links. In our method, associations of product fragments using concepts from the domain model constitute the base to calculate a semantic distance among the product fragments. The computed semantic distance is further used to assess the impact of change and establish dependency links among the product fragments.

The contribution of this paper is twofold. First, the method presented here was refined based on method acceptance tests reported in [48] and was applied in a prototype implementation. Second, effectiveness of the proposed method is evaluated and the results from laboratory-based randomized experiment [42] are analyzed.

The paper is structured as follows. In the next section, a problem boundary and the overall method are presented. Then the settings of the randomized experiment are discussed, followed by an analysis of its results. Later the proposed method is compared to related work. Finally, conclusions are summarized and future work is discussed.