JPlag: Looking Back, Moving Forward

[tsaglam]

For the past five years, I have been project lead for JPlag. This time is now coming to an end. When I took over maintenance together with @sebinside (who has recently moved on as well), JPlag was in a different place: maintained, but with development largely dormant. In this post, I will briefly reflect on the past years and discuss what comes next for JPlag.

JPlag Legacy

As you may know, JPlag was initially developed in 1996 by Guido Malpohl at the group of Prof. Walter Tichy at KIT. At the time, Java 1.0 was just released. Initially, it was used internally, but later it was made accessible via a web service, allowing users to submit programs. In 2015, the web service was retired, and the project was moved to GitHub. Over the years, JPlag became one of the most widely used tools for detecting source code plagiarism. However, development was dormant, and in 2020, with the retirement of Prof. Tichy, the maintenance was transferred to us at SDQ (also a group at KIT).

In 2020, JPlag still used its legacy HTML-based UI, which was generated from hardcoded string literals across several thousand lines of Java code (Report, AllMatches, Program, and more...). By then, several features were already broken, since the implementation relied on browser features that had been blocked for security reasons. The codebase still contained a discontinued web service and was largely legacy code. Due to its age, the codebase contained its own implementation of HashTable and ArrayList (both were not available in Java 1.0). JPlag itself worked reliably, but lacked support for modern programming languages and an API for 3rd-party integration. If you want to explore JPlag as it was, we preserved the 2020 version in the legacy branch¹.

The Last Five Years

To address these issues, we began the complete modernization of JPlag, basically reviving the project. Looking back, we have released many significant features. Here are the most important ones:

• A redesigned, modern web-based UI.
• Proper clustering analysis to detect group plagiarism.
• Support for 20 programming languages and multi-language submissions.
• A unified module system for adding support for new languages.
• Defense mechanisms against obfuscation attempts².
• Bundling of the viewer into the application.
• A web-based demo for JPlag.
• Parallelization of the entire JPlag pipeline for improved performance.
• An API for 3rd-party integration (JPlag is used in education tools such as Artemis or CodeGrade).
• Distribution via Maven Central.

Here is also a comparison of the legacy UI and the modern report viewer:

Overview (legacy):	Overview (now):

Comparison View (legacy):	Comparison View (now):

However, there were also many internal improvements:

• We completely overhauled and modernized the codebase.
• Comprehensive unit tests, integration tests, E2E tests, and language tests (core coverage 84%).
• Modern CI/CD pipeline (automated build, test, lint, documentation consistency, and release workflows).
• Extensive documentation-as-code (Wiki deployed from the repository).
• Enforce project-wide code style and conventions (via Spotless/Checkstyle/ESLint/Prettier).

Beyond development and maintenance, my academic research focused on research around source code plagiarism and (automated) obfuscation attacks³. Looking back, I especially enjoyed the discussions at ICSE and the opportunity to deliver a keynote at the MODELS Educators Symposium.

Project Growth in Numbers

In the past five years, JPlag has evolved into a highly active open-source project⁴, and I am proud of what our team has accomplished. In the following, I want to look at some stats to highlight how JPlag has developed over the last five years:

Metric	5 Years Ago	Today
⭐ Stars	320	1,650
📦 Commits5	111	5,500
🔀 Pull Requests	19	2,200
🐞 Issues	81	3526
👥 Contributors	13	47
💬 Supported Languages	7	20
🍴 Forks	112	351

Click to see how the project popularity grew over time…

(in 2024 someone posted JPlag to Hacker News, where it reached the front page, thus leading to strong growth)

Moving forward

With recent releases, JPlag has reached a stable state with fewer bugs and fewer breaking changes. In the near future, we want to focus on more user-facing features. Stay tuned for future releases. As always, if you have requests or ideas, please open an issue.

As my time as project lead comes to an end, I will pass the torch to @robinmaisch, who has already been an active maintainer. I have no doubt JPlag is in good hands. While I will remain part of the maintainer team, I will step back from day-to-day development and mostly watch from the sidelines.

I want to take this opportunity to thank everyone who joined this journey, especially our maintainer team and student developers (with a big shout-out to @sebinside, @dfuchss, @robinmaisch, @uuqjz, @Kr0nox, and @TwoOfTwelve), as well as all external contributors.

Footnotes

1. see the https://github.com/jplag/JPlag/tree/legacy branch and the archived website at https://web.archive.org/web/20020221210020/http://www.jplag.de/. ↩

2. we introduced subsequence match merging (--match-merging) and token-sequence normalization (--normalize, only Java/C++). ↩

3. see https://jplag.de if you want to read up on recent research around JPlag. If you are really interested, check out my doctoral dissertation. ↩

4. see https://github.com/jplag/JPlag/graphs/contributors for an activity graph. ↩

5. only counts commits on GitHub (the project was moved to GitHub in Spring 2015). ↩

6. 52 of which are currently open. ↩