Guidelines on the aesthetic quality of UML class diagrams

Abstract

In the past, formatting guidelines have proved to be a successful method to improve the readability of source code. With the increasing success of visual specification languages such as UML for model-driven software engineering visual guidelines are needed to standardize the presentation and the exchange of modeling diagrams with respect to human communication, understandability and readability. In this article, we introduce a new and encompassing taxonomy of visual guidelines capturing the aesthetic quality of UML class diagrams. We propose these guidelines as a framework to improve the aesthetic quality and thus the understandability of UML class diagrams. To validate this claim, we describe in detail a controlled experiment carried out as a pilot study to gather preliminary insights on the effects of some of the guideline rules on the understandability of UML class diagrams.

Introduction

In 1995, Booch, Rumbaugh and Jacobson integrated their software modeling approaches into a common approach, today known as the Unified Modeling Language (UML). The commonly agreed notation is still one of the main success factors of UML. However, even after several iterations, the de-facto standard modeling language still comprises inconsistencies and weaknesses, e.g. in formal semantics. One additional drawback of UML regarding human communication is the large number of possibilities for designing and displaying models, e.g. presentation options. In this article, we discuss the aesthetic quality of UML class diagrams, i.e. the quality of the presentation in terms of readability and layout. In fact, UML class diagrams represent the most stable diagram type of UML in the last years. Based on a web survey on the usage of UML, class diagrams appear to be the most used diagram type in the UML [7]. More than 170 responses showed that class diagrams are more often used than use case diagrams, sequence diagrams and even use case narratives.

After twelve years of UML and four years of UML version 2, one would assume that current tools should be able to perfectly support the UML. In our recent study on more than 60 UML tools [14] we found that most tools are still far away from a significant compliance level. The model interchange standard is often disregarded and the diagram interchange standard is not adopted at all. Thus, collaborative work is still restricted to the same tool (family). Further, the tools do not provide the necessary functionality for assessing or enforcing visual guidelines.

One problem in interchanging and communicating diagrams is the lack of commonly agreed publicly available guidelines. Such guidelines describe how to model high-quality, readable and aesthetically appealing UML (class) diagrams. The OMG avoided discussing this problem and removed almost all hints on style and layout from UML 2. In the meantime, several proposals on visual guidelines for UML class diagrams have been published.

Early guidelines focused on graph drawing aspects, e.g. the number of edge crossings. In recent years, however, the focus shifted more towards guidelines also incorporating perceptional, architectural or semantic information. In this article, we therefore discuss current approaches based on criteria derived from several related disciplines such as human diagram perception, software engineering, software visualization and graph drawing and propose a taxonomy of guidelines for UML class diagrams.

Apart from achieving aims such as improving communication, a commonly agreed visual guideline for high-quality UML class diagrams would also influence the quality of related tools. Standardized aesthetic quality assessment features may be offered similar to code quality tools today. A standard on aesthetic quality would also foster the development of specialized layout algorithms to visualize the (intermediary) results in model-driven development in a standardized way supporting exchange and understandability among software engineers. As an example, we realized related tools for our layout guideline, e.g. layout algorithms implementing our guideline, layout metrics and a regression test mechanism based on our guideline [13].

One of the most serious problems encountered on the way to developing a set of common guidelines is empirical validation, as there is no formal theory that can predict whether a specific visual guideline is able to improve the understandability of a diagram. In order to achieve credibility for an approach it is often regarded as crucial to provide empirical evidence of its usefulness [30]. This is often called “evidence-based software engineering” [8]. This is required as only empirical evidence can ensure the real-world usefulness of theories of software engineering [3].

Experimentation already has some history in visual guidelines. The major result of these studies is that the reduction of edge bends is more important than optimizing a layout with respect to the number of crossings. Some recent evaluations also consider advanced issues such as architectural importance or perceptual issues. To contribute to a common visual guideline for UML class diagrams, we discuss existing guidelines in terms of a new taxonomy on guideline rules and present an empirical validation of some of these rules.

This article is structured as follows: In the next section we introduce some basic terminology. In Section 3 we list related work on guidelines, tools, algorithms and empirical studies. Section 4 introduces a new taxonomy for visual guidelines for UML class diagrams. In Section 5, we describe an empirical validation for some guidelines. Finally, in Section 6 we summarize our results, draw conclusions and discuss work to be done in this field in the future.