Information Retrieval (IR) approaches are used to leverage textual or unstructured data generated during the software development process to support various software engineering (SE) tasks (e.g., concept location, traceability link recovery, change impact analysis, etc.). Two of the most important steps for applying IR techniques to support SE tasks are preprocessing the corpus and configuring the IR technique, and these steps can significantly influence the outcome and the amount of effort developers have to spend for these maintenance tasks. We present the use of Genetic Algorithms (GAs) to automatically configure and assemble an IR process to support SE tasks. The approach named IR-GA determines the (near) optimal solution to be used for each step of the IR process without requiring any training. We applied IR-GA on three different SE tasks and the results of the study indicate that IR-GA outperforms approaches previously used in the literature, and that it does not significantly differ from an ideal upper bound that could be achieved by a supervised approach and a combinatorial approach.

Information Retrieval (IR) approaches are nowadays used to support various software engineering tasks, such as feature location, traceability link recovery, clone detection, or refactoring. However, previous studies showed that inadequate instantiation of an IR technique and underlying process could significantly affect the performance of such approaches in terms of precision and recall. This paper proposes the use of Genetic Algorithms (GAs) to automatically configure and assemble an IR process for software engineering tasks. The approach (named GA-IR) determines the (near) optimal solution to be used for each stage of the IR process, i.e., term extraction, stop word removal, stemming, indexing and an IR algebraic method calibration. We applied GA-IR on two different software engineering tasks, namely traceability link recovery and identification of duplicate bug reports. The results of the study indicate that GA-IR outperforms approaches previously published in the literature, and that it does not significantly differ from an ideal upper bound that could be achieved by a supervised and combinatorial approach.

Locating features in large software systems is a fundamental maintenance task for developers when fixing bugs and extending software. We introduce In Situ Impact Insight (I3), a novel user interface to support feature location. In addition to a list of search results, I3 provides support for developers during browsing and inspecting the retrieved code entities. In situ visualizations augment results and source code with additional information relevant for further exploration. Developers are able to retrieve details on the textual similarity of a source code entity to the search query and to other entities, as well as the information on co-changed entities from a project's history. Execution traces recorded during program runs can be used as filters to further refine the search results. We implemented I3 as an Eclipse plug-in and tested it in a user study involving 18 students and professional developers that were asked to perform three feature location tasks chosen from the issue tracking system of jEdit. The results of our study suggest that I3's user interface is intuitive and unobtrusively supports developers with the required information when and where they need it.

Coupling is a fundamental property of software systems, and numerous coupling measures have been proposed to support various development and maintenance activities. However, little is known about how developers actually perceive coupling, what mechanisms constitute coupling, and if existing measures align with this perception. In this paper we bridge this gap, by empirically investigating how class coupling - as captured by structural, dynamic, semantic, and logical coupling measures - aligns with developers' perception of coupling. The study has been conducted on three Java open-source systems - namely ArgoUML, JHotDraw and jEdit - and involved 64 students, academics, and industrial practitioners from around the world, as well as 12 active developers of these three systems. We asked participants to assess the coupling between the given pairs of classes and provide their ratings and some rationale. The results indicate that the peculiarity of the semantic coupling measure allows it to better estimate the mental model of developers than the other coupling measures. This is because, in several cases, the interactions between classes are encapsulated in the source code vocabulary, and cannot be easily derived by only looking at structural relationships, such as method calls.

Information Retrieval (IR) methods, and in particular topic models, have recently been used to support essential software engineering (SE) tasks, by enabling software textual retrieval and analysis. In all these approaches, topic models have been used on software artifacts in a similar manner as they were used on natural language documents (e.g., using the same settings and parameters) because the underlying assumption was that source code and natural language documents are similar. However, applying topic models on software data using the same settings as for natural language text did not always produce the expected results. Recent research investigated this assumption and showed that source code is much more repetitive and predictable as compared to the natural language text. Our paper builds on this new fundamental finding and proposes a novel solution to adapt, configure and effectively use a topic modeling technique, namely Latent Dirichlet Allocation (LDA), to achieve better (acceptable) performance across various SE tasks. Our paper introduces a novel solution called LDA-GA, which uses Genetic Algorithms (GA) to determine a near-optimal configuration for LDA in the context of three different SE tasks: (1) traceability link recovery, (2) feature location, and (3) software artifact labeling. The results of our empirical studies demonstrate that LDA-GA is able to identify robust LDA configurations, which lead to a higher accuracy on all the datasets for these SE tasks as compared to previously published results, heuristics, and the results of a combinatorial search.

Software trace ability is the ability to describe and follow the life of a requirement in both a forward and backward direction by defining relationships to related development artifacts. A plethora of different trace ability recovery approaches use information retrieval techniques, which depend on the quality of the textual information in requirements and software artifacts. Not only is it important that stakeholders use meaningful names in these artifacts, but also it is crucial that the same names are used to specify the same concepts in different artifacts. Unfortunately, the latter is difficult to enforce and as a result, software trace ability approaches are not as efficient and effective as they could be - to the point where it is questionable whether the anticipated economic and quality benefits were indeed achieved. We propose a novel and automatic approach for expanding corpora with relevant documentation that is obtained using external function call documentation and sets of relevant words, which we implemented in Trace Lab. We experimented with three Java applications and we show that using our approach the precision of recovering trace ability links was increased by up to 31% in the best case and by approximately 9% on average.

Research studies in software maintenance are notoriously hard to reproduce due to lack of datasets, tools, implementation details (e.g., parameter values, environmental settings) and other factors. The progress in the field is hindered by the challenge of comparing new techniques against existing ones, as researchers have to devote a lot of their resources to the tedious and error-prone process of reproducing previously introduced approaches. In this paper, we address the problem of experiment reproducibility in software maintenance and provide a long term solution towards ensuring that future experiments will be reproducible and extensible. We conducted a mapping study of a number of representative maintenance techniques and approaches and implemented them as a library of experiments and components that we make publicly available with TraceLab, called the Component Library. The goal of these experiments and components is to create a body of actionable knowledge that would (i) facilitate future research and would (ii) allow the research community to contribute to it as well. In addition, to illustrate the process of using and adapting these techniques, we present an example of creating new techniques based on existing ones, which produce improved results.

Feature location is a fundamental step in software evolution tasks such as debugging, understanding, and reuse. Numerous automated and semi-automated feature location techniques (FLTs) have been proposed, but the question remains: How do we objectively determine which FLT is most effective? Existing evaluations frequently use bug fix data, which includes the location of the fix, but not what other code needs to be understood to make the fix. Existing evaluation measures such as precision, recall, effectiveness, mean average precision (MAP), and mean reciprocal rank (MRR) will not differentiate between a FLT that ranks higher these related elements over completely irrelevant ones. We propose an alternative measure of relevance based on the likelihood of a developer finding the bug fix locations from a ranked list of results. Our initial evaluation shows that by modeling user behavior, our proposed evaluation methodology can compare and evaluate FLTs fairly.

Approaches that support software maintenance need to be evaluated and compared against existing ones, in order to demonstrate their usefulness in practice. However, oftentimes the lack of well-established sets of benchmarks leads to situations where these approaches are evaluated using different datasets, which results in biased comparisons. In this data paper we describe and make publicly available a set of benchmarks from six Java applications, which can be used in the evaluation of various software engineering (SE) tasks, such as feature location and impact analysis. These datasets consist of textual description of change requests, the locations in the source code where they were implemented, and execution traces. Four of the benchmarks were already used in several SE research papers, and two of them are new. In addition, we describe in detail the methodology used for generating these benchmarks and provide a suite of tools in order to encourage other researchers to validate our datasets and generate new benchmarks for other subject software systems. Our online appendix: http://www.cs.wm.edu/semeru/data/msr13/.

Question & answer (Q&A) websites, such as Stack Overflow (SO), are widely used by developers to find and provide answers to technical issues and concerns in software development. Mobile development is not an exception to the rule. In the latest SO dump, more than 400K questions were labeled with tags related to mobile technologies. Although, previous works have analyzed the main topics and trends in SO threads, there are no studies devoted specifically to mobile development. In this paper we used topic modeling techniques to extract hot-topics from mobile-development related questions. Our findings suggest that most of the questions include topics related to general questions and compatibility issues, and the most specific topics, such as crash reports and database connection, are present in a reduced set of questions.

SUMMARY Feature location is the activity of identifying an initial location in the source code that implements functionality in a software system. Many feature location techniques have been introduced that automate some or all of this process, and a comprehensive overview of this large body of work would be beneficial to researchers and practitioners. This paper presents a systematic literature survey of feature location techniques. Eighty‐nine articles from 25 venues have been reviewed and classified within the taxonomy in order to organize and structure existing work in the field of feature location. The paper also discusses open issues and defines future directions in the field of feature location. Copyright © 2011 John Wiley & Sons, Ltd.

The paper presents an adaptive approach to perform impact analysis from a given change request to source code. Given a textual change request (e.g., a bug report), a single snapshot (release) of source code, indexed using Latent Semantic Indexing, is used to estimate the impact set. Should additional contextual information be available, the approach configures the best-fit combination to produce an improved impact set. Contextual information includes the execution trace and an initial source code entity verified for change. Combinations of information retrieval, dynamic analysis, and data mining of past source code commits are considered. The research hypothesis is that these combinations help counter the precision or recall deficit of individual techniques and improve the overall accuracy. The tandem operation of the three techniques sets it apart from other related solutions. Automation along with the effective utilization of two key sources of developer knowledge, which are often overlooked in impact analysis at the change request level, is achieved. To validate our approach, we conducted an empirical evaluation on four open source software systems. A benchmark consisting of a number of maintenance issues, such as feature requests and bug fixes, and their associated source code changes was established by manual examination of these systems and their change history. Our results indicate that there are combinations formed from the augmented developer contextual information that show statistically significant improvement over standalone approaches.

Similarly to other fields in software engineering, the results of case studies involving feature location techniques (FLTs) are hard to reproduce, compare, and generalize, due to factors such as, incompatibility of different datasets, lack of publicly available implementation or implementation details, or the use of different metrics for evaluating FLTs. To address these issues, we propose a solution for creating, conducting, and sharing experiments in feature location based on TraceLab, a framework for conducting research. We argue that this solution would allow rapid advancements in feature location research because it will enable researchers to create new FLTs in the form of TraceLab templates or components, and compare them with existing ones using the same datasets and the same metrics. In addition, it will also allow sharing these FLTs and experiments within the research community. Our proposed solution provides (i) templates and components for creating new FLTs and instantiating existing ones, (ii) datasets that can be used as inputs for these FLTs, and (iii) metrics for comparing these FLTs. The proposed solution can be easily extended with new FLTs (in the form of easily configurable templates and components), datasets, and metrics.

The paper presents an exploratory study of two feature location techniques utilizing three strategies for splitting identifiers: Camel Case, Samurai and manual splitting of identifiers. The main research question that we ask in this study is if we had a perfect technique for splitting identifiers, would it still help improve accuracy of feature location techniques applied in different scenarios and settings? In order to answer this research question we investigate two feature location techniques, one based on Information Retrieval and the other one based on the combination of Information Retrieval and dynamic analysis, for locating bugs and features using various configurations of preprocessing strategies on two open-source systems, Rhino and jEdit. The results of an extensive empirical evaluation reveal that feature location techniques using Information Retrieval can benefit from better preprocessing algorithms in some cases, and that their improvement in effectiveness while using manual splitting over state-of-the-art approaches is statistically significant in those cases. However, the results for feature location technique using the combination of Information Retrieval and dynamic analysis do not show any improvement while using manual splitting, indicating that any preprocessing technique will suffice if execution data is available. Overall, our findings outline potential benefits of putting additional research efforts into defining more sophisticated source code preprocessing techniques as they can still be useful in situations where execution information cannot be easily collected.

The paper presents an adaptive approach to perform impact analysis from a given change request (e.g., a bug report) to source code. Given a textual change request, a single snapshot (release) of source code, indexed using Latent Semantic Indexing, is used to estimate the impact set. Additionally, the approach configures the best-fit combination of information retrieval, dynamic analysis, and data mining of past source code commits to produce an improved impact set. The tandem operation of the three techniques sets it apart from other related solutions.

Acquiring general understanding of large software systems and components from which they are built can be a time consuming task, but having such an understanding is an important prerequisite to adding features or fixing bugs. In this paper we propose the tool, namely TopicXP, to support developers during such software maintenance tasks by extracting and analyzing unstructured information in source code identifier names and comments using Latent Dirichlet Allocation. TopicXPenables developers to gain an overview of a software system under analysis by extracting and visualizing natural language topics, which generally correspond to concepts or features implemented in software classes. TopicXPis implemented as an open-source Eclipse plug-in, which proposes interactive visualization of topics along with structural dependencies between underlying classes implementing these topics. The paper also presents the results of a preliminary user study aimed at evaluating TopicXP.

Data fusion is the process of integrating multiple sources of information such that their combination yields better results than if the data sources are used individually. This paper applies the idea of data fusion to feature location, the process of identifying the source code that implements specific functionality in software. A data fusion model for feature location is presented which defines new feature location techniques based on combining information from textual, dynamic, and web mining analyses applied to software. A novel contribution of the proposed model is the use of advanced web mining algorithms to analyze execution information during feature location. The results of an extensive evaluation indicate that the new feature location techniques based on web mining improve the effectiveness of existing approaches by as much as 62%.