Project

Clone detection; Software evolution; Meta-modeling; Software architecture

Lead
Software-Entwickler verbringen viel ihrer Zeit nicht mit dem Programmieren neuer Codes sondern mit dem Analysieren bestehender Quellcodes. Aktuelle Integrierte Entwicklungsumgebungen bieten weniger an als Textverarbeitungssysteme für Programme, welche nur eine begrenzte Unterstützung für die Entwickler zur Abfrage und Analyse von Softwaresystemen anbieten. In der Fortsetzung unseres laufenden SNF-Projektes (http://scg.unibe.ch/asa) suchen wir nach neuen Wegen, um den Entwicklern zu ermöglichen, effiziente Antworten über detaillierte Fragen der Softwaresysteme in der Entwicklung zu geben.
Lay summary
In diesem Projekt erforschen wir neue Wege, um Software Entwicklern in der Beantwortung von Fragen in Bezug auf die entwickelte Software zu unterstützen.  Das Projekt ist in vier parallele Forschungsteile für Doktoranden aufgebaut. Im ersten Teil, „Bewegliche Gewinnung/Methoden von Modellen“ erforschen wir wie Software Quellcodes und die dazugehörenden Informationen in Software Modellen umgesetzt, einfach abgefragt, analysiert und effizient gestaltet werden können. Die grösste Herausforderung ist es, solche Modelle ohne grossen Entwicklungsaufwand in einer komplexen Parsing-Technologie schnell zu programmieren und zu verfeinern.  Im zweiten Teil „Kontextabhängiger Werkzeugaufbau“ streben wir danach, Entwickler so zu unterstützen, dass neue Werkzeuge die gezielt für die Anwendungsdomäne funktionieren, möglichst einfach und schnell gebaut werden können. Der dritte Teil „Ökosystem mining“ stellt Entwicklern nützliche Informationen über andere zusammenhängende Projekte wie z.B. wiederverwendbare Code-Beispiele, Möglichkeiten zur Software Erneuerung, verbreitete Softwarefehler und Fehlerbehebung zur Verfügung.  Zum Schluss in Teil „Evolutionäre Überwachung“ verfolgen wir die Entwicklung des Softwaresystems  mit dem Ziel, die Entwickler über Schwachstellen und Verbesserungsmöglichkeiten laufend zu informieren.

Direct link to Lay Summary

Last update: 22.10.2015

A Systematic Literature Review of Software Visualization Evaluation

Author	Merino, Leonel
Persistent Identifier (PID)	http://scg.unibe.ch/research/softvis-eval
Repository	A Systematic Literature Review of Software Visualization Evaluation
Abstract

Through software visualization developers can augment their capabilities to analyze multiple aspects of complex software systems. However, not every software visualization is suitable to tackle a given development concern. To understand how effective a software visualization is, it must be evaluated on that particular development task. We present an overview of the evaluation of existing visualizations, and examine their characteristics. In particular, we review the complete literature body of papers published in the SOFTVIS/VISSOFT conferences. Among the 387 papers published to date, we included in our study 181 full papers from which we extracted evaluation strategies, data collection methods and other various aspects of the evaluations. We found that 62% of the proposed software visualization approaches do not include an evaluation. We believe that for software visualization approaches to be adopted by developers, visualizations not only must prove their effectiveness via evaluations, but also evaluations should include participants of the target audience and real-world software systems. Finally, we recommend researchers in the field to conduct surveys that can help them to identify what are the frequent and complex problems that affect developers.

Title	Year

Number	Title	Start	Funding scheme

Software developers actually spend much of their time not producing new code, but analysing the existing code base. Integrated Development Environments (IDEs) however are mostly glorified text editors, and offer only limited support for developers to query and analyse software systems. In this continuation of our running SNF project (http://scg.unibe.ch/asa), we proceed to explore new ways to enable developers to efficiently answer detailed questions about the software system under development.Like its predecessor, this project is designed as four related PhD tracks (including two end-of-thesis subprojects). We avoid any critical dependencies between the tracks, while offering many possibilities for fruitful collaboration.- Agile Model Extraction. Here we address the question: "How can we rapidly extract models from unknown source code?" The key insight is that we do not need precise parsers to extract useful models from software. By using approximate parsing techniques combined with heuristics to automatically recognise common programming language features, we expect to be able to iteratively construct analysable models of complex software systems in a fraction of the time it would take to build a conventional model importer.- Context-Aware Tooling. In this track we tackle the question: "How do we close the abstraction gap between application domains and IDEs?" We plan to extend and generalise our previous work on highly configurable tools, such as debuggers, browsers and inspectors, that can easily be adapted to the context of a given project, and enhance both the source code and the IDE itself with domain-specific and project-specific information. In particular, we will focus on domain-specific ways to present and query software systems.- Ecosystem Mining. Here we ask: "How can we mine the ecosystem to automatically discover intelligence relevant to a given project?" When evolving source code, developers commonly make use of dedicated social media (i.e., question and answer fora) as well as general search engines to search for answers to technical questions. In addition to these sources, the software ecosystem of related software systems can be a rich source of useful information, if properly analysed. We plan to mine these ecosystems to locate common, reusable code examples, opportunities for refactoring, common bugs, and bug fixes.- Evolutionary Monitoring. Finally we address the question: "How can we steer the evolution of a software system by monitoring stakeholder needs, technical debt, and ecosystem trends?" By monitoring the activities of the stakeholders (i.e., both users and developers), we hope to infer their needs and the chronic problems of the system. By monitoring "technical debt" (i.e., pain points where effort should be invested to avoid the future escalation of these problems), we expect to be able to better prioritise future development activities. By monitoring technical trends from the ecosystem, we expect to be able to detect new opportunities and also better estimate the cost of future change.In addition to disseminating the results of the research through academic publications, we will continue to release software artifacts (tools, IDE extensions etc.) as part of the Pharo development environment and the Moose analysis platform used by a wide community of academic and industrial partners.