Designite and its Sibling Tools for Software Engineering Research

This presentation provides an overview of Designite and its sibling tools (such as DesigniteJava, CodeSplit, and QScored) and their features. The focus of the presentation is to elaborate on the ways these tools can be used in software engineering research.

Transcript

1. Designite and its Sibling Tools for Software Engineering Research Tushar Sharma https://tusharma.in CSER Spring 2022 meeting
2. Overview • What are Designite, DesigniteJava, QScored … • How these tools can be used in SE research • How I used these tools in my research
3. Dr. Tushar Sharma Assistant professor at Dalhousie University, Canada PhD • Athens University of Economics and Business, Greece Industry experience • Siemens Research (7 + 2) Books • Refactoring for software design smells Tools/platforms • Designite • QScored
5. Smells: Types http://www.tusharma.in/smells/
6. A survey on software smells https://www.sciencedirect.com/science/article/abs/pii/S0164121217303114
7. Demo http://www.designite-tools.com What is Designite?
8. Designite is a software design quality assessment tool for C# and Java. Free academic licenses.
9. DesigniteJava Designite (C#)
11. Console application – C# version Relevant for large scale empirical studies
12. DesigniteJava Tool also generate CSV files for large- scale analysis
13. DesigniteJava plugin for IntelliJ
14. CodeSplit • CodeSplitJava splits Java projects by putting classes and methods into separate files. • Useful in ML4SE applications java -jar CodeSplitJava.jar -i <path of the input source folder> -o <path of the output folder> -m <mode>
15. An open platform Allows users to upload their projects’ analysis reports Language agnostic comprehensive visualization Detailed visualization for many code quality aspects Quality ranking Computes quality score and relative quality ranking for all projects in its corpus REST APIs APIs to upload projects’ quality reports and search QScored corpus based on quality parameters Quality badges Allows users to show their projects’ ranking and score to the world https://qscored.com/
16. QScored dataset The dataset contains • Detected code smells (7 kinds of architecture, 20 design, and 11 implementation smells) • 10 project-level, 2 package-level, 12 class-level, and 3 method-level code quality metrics • Historical trend of relative ranking of the repositories.
17. Using Designite toolset in research projects
18. Explored characteristics - Frequency - Smell density and project size - Smell correlation - Smell colocation - Causality
19. Explored characteristics - Frequency - Smell density and project size - Smell correlation - Smell colocation - Causality Research questions C# C# C# Analyze smells
20. Explored characteristics - Smell detection using deep learning without feature engineering - Smell detection using transfer learning
22. 10-90 20-80 30-70 40-60 50-50 60-40 Java 10.0 45.7 80.0 94.4 97.3 100.0 C# 20.8 65.9 90.3 96.7 97.8 100.0 0.0 20.0 40.0 60.0 80.0 100.0 Projects (in percent) Pareto categories
23. Potential research applications Empirical research • Smells (implementation, design, or architecture) as one of the dimensions • Code quality metrics Smell detection extensions • IDE integrations (e.g., impact of showing smells in the IDEs on code quality) ML4SE • Identifying smells using ML techniques • Active learning
24. Potential research applications Repository selection • QScored supports repository selection based on the code quality of the open-source repositories (among its own corpus) • APIs to perform parameterized search • Get top 100 Java repositories where lines of code is more than 50,000
25. Impacting research community by tools • Software quality assessment tools for C# and Java • Free academic license • Used by 135+ universities worldwide (full list) • Extensively used by SE researchers • 120+ stars on GitHub • An open platform for code quality ranking and visualization. • Hosts quality data of 100+ K GitHub repositories (> 1 billion LOC) • Available for research community as dataset https://qscored.com https://designite-tools.com
26. Relevant links Designite (website) - https://www.designite-tools.com/ DesigniteJava (GitHub) - https://github.com/tushartushar/DesigniteJava Smell catalog - http://www.tusharma.in/smells/ Designite plugin for IntelliJ - https://plugins.jetbrains.com/plugin/13380-designitejava QScored (website) – https://qscored.com/ CodeSplit for Java - https://github.com/tushartushar/CodeSplitJava
27. Thank you!! tushar@dal.ca Software engineering Machine learning Code smell Technical debt Refactoring Code quality Binary symbol reconstruction SE4ML ML4SE SMART lab Dalhousie University https://web.cs.dal.ca/~tushar/smart/