Assessing sofware quality by micro patterns detection

One of the goals of Software Engineering is to reduce, or at least to try to control, the defectiveness of software systems during the development phase. Software engineers need to have empirical evidence that software metrics are related to software quality. Unfortunately, software quality is quite an elusive concept, software being an immaterial entity that cannot be physically measured in traditional ways. In general, software quality means many things. In software, the narowest sense of product quality is commonly recognized as absence or low incidence of bugs in the product. It is also the most basic meaning of confermance to requirements, because if the software contains too many functional defects, the basic requirement of providing the desired function is not met. To increase overall customer satisfaction as well as satisfaction toward various quality attributes, the quality attributes must be taken into account in the planning and design of software. To improve quality during development, we need models of the develompment process, and within the process we need to select and deploy specific methods and approaches, and employ proper tools and technologies. It is necessary to know measures of the characteristics and quality parameters of the development process and its stages, as well as metrics and models to help ensure that the development process is under control to meet the product’s quality objectives. Software quality metrics tend to measure whether software is well structured, not too simple and not too complex, with cohesive modules that minimize their coupling. Many quality metrics have been proposed for software, depending also on the paradigm and languages used there are metrics for structured programming, object-oriented programming, aspect-oriented programming, and so on. The use of traditional metrics as quality indicators is very dicult. The Lines of Code (LOC) metric (very related to faults), is dicult to use, you can not say to a team of developers to develop classes by imposing a predefined number of lines of code. The use of the micro patterns (introduced by Gil and Maman) metrics, that capture concepts of good or bad programming (like anti patterns) can help developers to focus on those classes that belong to categories of micro patterns prone to fault. The relationship between traditional metrics and micro patterns is useful for enabling these new metrics to evaluate software quality. Micro patterns are similar to design patterns, but their characteristic is that they can be identified automatically, and are at a lower level of abstraction with respect to design patterns. This thesis tackles the problem of measuring software quality in Object Oriented (OO) systems by using such novel approaches based on micro patterns that can be a useful metrics in order to measure the quality of software by showing that certain categories of micro patterns are more fault prone than others, and that the classes that do not correspond to any category of micro patterns are more likely to be faulty. Many empirical studies were performed to validate empirically CK suite under these two aspects, showing an acceptable correlation between CK metrics values and software fault-proneness and diculty of maintenance. In OO, micro patterns can help to identify the portions of code that should be improved (for example those where encapsulation is not respected), and highlight other portions that make up good design practices. The design patterns, defined in the early nineties were an important breakthrough at analysis and design level, but are dicult to be automatically supported at the coding level. There are tools claiming to help finding the usage of design patterns in code, but in practice they are used in a very limited way. On the contrary, micro patterns are defined at coding level, and it is relatively easy to recognize them automatically, thus being able to implement formal conditions on the structure of the class. Thesis overview The thesis is organized according to this scheme: • Chapter 2 provides an overview of the concept of software metrics; • Chapter 3 presents an overview of the design patterns catalogs; • Chapter 4 discusses the micro patterns catalog using the definitions made by Gil and Maman; • Chapter 5 discusses the interpretation of Micro Patterns given by Arcelli and Maggioni; • Chapter 6 present the study of the evolution of five particular micro patterns (anti patterns) in different releases of the Eclipse and NetBeans systems, and the correlations between anti patterns and faults. The analysis confirms previous findings regarding the high coverage of micro patterns onto the system classes, and show that anti patterns not only represent bad Object Oriented programming practices, but may also be associated to the production of worse quality software, since they present a significantly enhanced fault proneness. • Chapter 7 present a study that aims to show, through empirical studies of open source software systems, which categories of micro patterns are more correlated to faults. Gil and Maman demonstrated, and subsequent studies confirmed, that 75% of the classes of a software system are covered by micro patterns. In this chapter is also analyzed the relationship between faults and the remaining 25% of classes that do not match with any micro pattern. We found that these classes are more likely to be faultprone than the others. We also studied the correlation among all the micro patterns of the catalog, in order to verify the existence of relationships between them. • Chapter 8 present a study on micro patterns in different releases of two software systems developed with Object Oriented technologies and Agile process. In this chapter we present some empirical results on two case studies of systems devel- oped with Agile methodologies, and compare them to previous results obtained for non Agile systems. In particular we have verified that the distribution of micro patterns in a software system developed using Agile methodologies does not differ from the distribution studied in other systems, and that the micro patterns fault-proneness is about the same. We also analyzed how the distribution of micro patterns changes in different releases of the same software system. We demonstrate that there is a relationship between the number of faults and the classes that do not match with any micro patterns. We found that these classes are more likely to be fault-prone than the others even in software developed with Agile methodologies • Chapter 9 present the Java tool used in order to extract from the source code the informations about micro patterns distributions. • Chapter 10 discusses the related works in the field.