Refactoring is the process of changing the internal structure of software to improve its quality without modifying its external behavior. Empirical studies have repeatedly shown that refactoring has a positive impact on the understandability and maintainability of software systems. However, before carrying out refactoring activities, developers need to identify refactoring opportunities. Currently, refactoring opportunity identification heavily relies on developers’ expertise and intuition. In this paper, we investigate the effectiveness of machine learning algorithms in predicting software refactorings. More specifically, we train six different machine learning algorithms (i.e., Logistic Regression, Naive Bayes, Support Vector Machine, Decision Trees, Random Forest, and Neural Network) with a dataset comprising over two million refactorings from 11,149 real-world projects from the Apache, F-Droid, and GitHub ecosystems. The resulting models predict 20 different refactorings at class, method, and variable-levels with an accuracy often higher than 90 percent. Our results show that (i) Random Forests are the best models for predicting software refactoring, (ii) process and ownership metrics seem to play a crucial role in the creation of better models, and (iii) models generalize well in different contexts.

Abstract Software developers, including those of the Android mobile platform, constantly seek to improve their applications’ maintainability and evolvability. Code smells are commonly used for this purpose, as they indicate symptoms of design problems. However, although the literature presents a variety of code smells, such as God Class and Long Method, characteristics that are specific to the underlying technologies are not taken into account. The presentation layer of an Android app, for example, implements specific architectural decisions from the Android platform itself (such as the use of Activities, Fragments, and Listeners) as well as deal with and integrate different types of resources (such as layouts and images). Through a three-step study involving 246 Android developers, we investigated code smells that developers perceive for this part of Android apps. We devised 20 specific code smells and collected the developers’ perceptions of their frequency and importance. We also implemented a tool that identifies the proposed code smells and studied their prevalence in 619 open-source Android apps. Our findings suggest that: 1) developers perceive smells specific to the presentation layer of Android apps; 2) developers consider these smells to be of high importance and frequency; and 3) the proposed smells occur in real-world Android apps. Our domain-specific smells can be leveraged by developers, researchers, and tool developers for searching potentially problematic pieces of code.

Software testing involves probing into the behavior of software systems to uncover faults. Most testing activities are complex and costly, so a practical strategy that has been adopted to circumvent these issues is to automate software testing. There has been a growing interest in applying machine learning (ML) to automate various software engineering activities, including testing-related ones. In this paper, we set out to review the state-of-the art of how ML has been explored to automate and streamline software testing and provide an overview of the research at the intersection of these two fields by conducting a systematic mapping study. We selected 48 primary studies. These selected studies were then categorized according to study type, testing activity, and ML algorithm employed to automate the testing activity. The results highlight the most widely used ML algorithms and identify several avenues for future research. We found that ML algorithms have been used mainly for test-case generation, refinement, and evaluation. Also, ML has been used to evaluate test oracle construction and to predict the cost of testing-related activities. The results of this paper outline the ML algorithms that are most commonly used to automate software-testing activities, helping researchers to understand the current state of research concerning ML applied to software testing. We also found that there is a need for better empirical studies examining how ML algorithms have been used to automate software-testing activities.

Architecture-Driven Modernization (ADM) intends to standardize software reengineering by relying on a family of standard metamodels. Knowledge-Discovery Metamodel (KDM) is the main ADM ISO metamodel aiming at representing all aspects of existing legacy systems. One of the internal KDM metamodels is called Structure, responsible for representing architectural abstractions (Layers, Components and Subsystems) and their relationships. Planned Architecture (PA) is an artifact that involves not only the architectural abstractions of the system but also the access rules that must exist between them and be maintained over time. Although PAs are frequently used in Architecture-Conformance Checking processes, up to this moment, there is no contribution showing how to specify and serialize PAs in ADM-based modernization projects. Therefore, in this paper we present an approach that i) involves a DSL (Domain-Specific Language) for the specification of PAs using the Structure metamodel concepts, and ii) proposes a strategy for the serialization of PAs as a Structure metamodel instance without modifying it. We have conducted a comparison between DCL-KDM and other techniques for specifying and generating PAs. The results showed that DCL-KDM is an efficient alternative to to generate instances of the Structure metamodel as a PA and to serialize it.