Giuliano Antoniol

Recovering traceability links between code and documentation

Software system documentation is almost always expressed informally in natural language and free text. Examples include requirement specifications, design documents, manual pages, system development journals, error logs, and related maintenance reports. We propose a method based on information retrieval to recover traceability links between source code and free text documents. A premise of our work is that programmers use meaningful names for program items, such as functions, variables, types, classes, and methods. We believe that the application-domain knowledge that programmers process when writing the code is often captured by the mnemonics for identifiers; therefore, the analysis of these mnemonics can help to associate high-level concepts with program concepts and vice-versa. We apply both a probabilistic and a vector space information retrieval model in two case studies to trace C++ source code onto manual pages and Java code to functional requirements. We compare the results of applying the two models, discuss the benefits and limitations, and describe directions for improvements.

Comparison and Evaluation of Clone Detection Tools

Many techniques for detecting duplicated source code (software clones) have been proposed in the past. However, it is not yet clear how these techniques compare in terms of recall and precision as well as space and time requirements. This paper presents an experiment that evaluates six clone detectors based on eight large C and Java programs (altogether almost 850 KLOC). Their clone candidates were evaluated by one of the authors as an independent third party. The selected techniques cover the whole spectrum of the state-of-the-art in clone detection. The techniques work on text, lexical and syntactic information, software metrics, and program dependency graphs.

Feature Location Using Probabilistic Ranking of Methods Based on Execution Scenarios and Information Retrieval

This paper recasts the problem of feature location in source code as a decision-making problem in the presence of uncertainty. The solution to the problem is formulated as a combination of the opinions of different experts. The experts in this work are two existing techniques for feature location: a scenario-based probabilistic ranking of events and an information-retrieval-based technique that uses latent semantic indexing. The combination of these two experts is empirically evaluated through several case studies, which use the source code of the Mozilla Web browser and the Eclipse integrated development environment. The results show that the combination of experts significantly improves the effectiveness of feature location as compared to each of the experts used independently

EEG data compression techniques

Electroencephalograph (EEG) and Holter EEG data compression techniques which allow perfect reconstruction of the recorded waveform from the compressed one are presented and discussed. Data compression permits one to achieve significant reduction in the space required to store signals and in transmission time. The Huffman coding technique in conjunction with derivative computation reaches high compression ratios (on average 49% on Holter and 58% on EEG signals) with low computational complexity. By exploiting this result a simple and fast encoder/decoder scheme capable of real-time performance on a PC was implemented. This simple technique is compared with other predictive transformations, vector quantization, discrete cosine transform (DCT), and repetition count compression methods. Finally, it is shown that the adoption of a collapsed Huffman tree for the encoding/decoding operations allows one to choose the maximum codeword length without significantly affecting the compression ratio. Therefore, low cost commercial microcontrollers and storage devices can be effectively used to store long Holter EEGs in a compressed format.

An exploratory study of the impact of antipatterns on class change- and fault-proneness

Antipatterns are poor design choices that are conjectured to make object-oriented systems harder to maintain. We investigate the impact of antipatterns on classes in object-oriented systems by studying the relation between the presence of antipatterns and the change- and fault-proneness of the classes. We detect 13 antipatterns in 54 releases of ArgoUML, Eclipse, Mylyn, and Rhino, and analyse (1) to what extent classes participating in antipatterns have higher odds to change or to be subject to fault-fixing than other classes, (2) to what extent these odds (if higher) are due to the sizes of the classes or to the presence of antipatterns, and (3) what kinds of changes affect classes participating in antipatterns. We show that, in almost all releases of the four systems, classes participating in antipatterns are more change-and fault-prone than others. We also show that size alone cannot explain the higher odds of classes with antipatterns to underwent a (fault-fixing) change than other classes. Finally, we show that structural changes affect more classes with antipatterns than others. We provide qualitative explanations of the increase of change- and fault-proneness in classes participating in antipatterns using release notes and bug reports. The obtained results justify a posteriori previous work on the specification and detection of antipatterns and could help to better focus quality assurance and testing activities.

Design pattern recovery in object-oriented software

An approach to recover object oriented design patterns from design and code is presented. The pattern recovery process is based on a multi-stage filtering strategy to avoid combinatorial explosion on large software systems. To maintain independence from the language and the case tools adopted in developing software, both design and code are mapped into an intermediate representation. The multi-stage searching strategy allows to safely determine pattern candidates. To assess the effectiveness of the pattern recovery process a portable environment written in Java has been developed. Based on this environment, experimental results on public domain and industrial software were obtained and are discussed in the paper. Evidence is shown that, by exploiting information about method calls as a further constraint beyond the structural ones, the number of false positives is reduced.

CERBERUS: Tracing Requirements to Source Code Using Information Retrieval, Dynamic Analysis, and Program Analysis

The concern location problem is to identify the source code within a program related to the features, requirements, or other concerns of the program. This problem is central to program development and maintenance. We present a new technique called prune dependency analysis that can be combined with existing techniques to dramatically improve the accuracy of concern location. We developed CERBERUS, a potent hybrid technique for concern location that combines information retrieval, execution tracing, and prune dependency analysis. We used CERBERUS to trace the 360 requirements of RHINO, a 32,134 line Java program that implements the ECMAScript international standard. In our experiment, prune dependency analysis boosted the recall of information retrieval by 155% and execution tracing by 104%. Moreover, we show that our combined technique outperformed the other techniques when run individually or in pairs.

DeMIMA: A Multilayered Approach for Design Pattern Identification

Design patterns are important in object-oriented programming because they offer design motifs, elegant solutions to recurrent design problems, which improve the quality of software systems. Design motifs facilitate system maintenance by helping to understand design and implementation. However, after implementation, design motifs are spread throughout the source code and are thus not directly available to maintainers. We present DeMIMA, an approach to identify semi-automatically micro-architectures that are similar to design motifs in source code and to ensure the traceability of these micro-architectures between implementation and design. DeMIMA consists of three layers: two layers to recover an abstract model of the source code, including binary class relationships, and a third layer to identify design patterns in the abstract model. We apply DeMIMA to five open-source systems and, on average, we observe 34% precision for the considered 12 design motifs. Through the use of explanation-based constraint programming, DeMIMA ensures 100% recall on the five systems. We also apply DeMIMA on 33 industrial components.

Analyzing cloning evolution in the Linux kernel

Abstract Identifying code duplication in large multi-platform software systems is a challenging problem. This is due to a variety of reasons including the presence of high-level programming languages and structures interleaved with hardware-dependent low-level resources and assembler code, the use of GUI-based configuration scripts generating commands to compile the system, and the extremely high number of possible different configurations. This paper studies the extent and the evolution of code duplications in the Linux kernel. Linux is a large, multi-platform software system; it is based on the Open Source concept, and so there are no obstacles in discussing its implementation. In addition, it is decidedly too large to be examined manually: the current Linux kernel release (2.4.18) is about three million LOCs. Nineteen releases, from 2.4.0 to 2.4.18, were processed and analyzed, identifying code duplication among Linux subsystems by means of a metric-based approach. The obtained results support the hypothesis that the Linux system does not contain a relevant fraction of code duplication. Furthermore, code duplication tends to remain stable across releases, thus suggesting a fairly stable structure, evolving smoothly without any evidence of degradation.

Combining Probabilistic Ranking and Latent Semantic Indexing for Feature Identification

The paper recasts the problem of feature location in source code as a decision-making problem in the presence of uncertainty. The main contribution consists in the combination of two existing techniques for feature location in source code. Both techniques provide a set of ranked facts from the software, as result to the feature identification problem. One of the techniques is based on a scenario based probabilistic ranking of events observed while executing a program under given scenarios. The other technique is defined as an information retrieval task, based on the latent semantic indexing of the source code. We show the viability and effectiveness of the combined technique with two case studies. A first case study is a replication of feature identification in Mozilla, which allows us to directly compare the results with previously published data. The other case study is a bug location problem in Mozilla. The results show that the combined technique improves feature identification significantly with respect to each technique used independently