VKVassilios Karakoidas
All publications

Conference · 2015

Generating the Blueprints of the Java Ecosystem

Vassilios Karakoidas, Dimitris Mitropoulos, Panos Louridas, Georgios Gousios, Diomidis Spinellis

A dataset of static-analysis metrics for 22,730 JAR files from Maven Central, covering object-oriented design, program size and package design, published so that others can ask questions of the Java ecosystem without rebuilding the corpus.

Published in
MSR '15: Proceedings of the 2015 International Working Conference on Mining Software Repositories, 2015
Citations
17 on Google Scholar, read 5 September 2026 — 9 of 30 by count
Cite as
KMGSL15
DOI (external link, opens in a new tab)the dataset

The idea

Maven Central is the de-facto library repository for the Java ecosystem, which makes it the closest thing available to a census of real Java code. Measuring it once, thoroughly, and publishing the measurements is worth more to the research community than any single result drawn from them.

How it was built

  • A local snapshot of the repository, filtered to Java projects that ship both a binary and a source JAR — the source JAR is what proves the project is Java — keeping only the latest version of each.
  • Three analysis tools, each covering a different aspect: ckjm (object-oriented design, including the Chidamber and Kemerer metrics), JDepend (package design) and CLMT, a cross-language metric tool (program size).
  • A task queue and a pool of Python workers that downloaded each JAR, ran the three tools on it and wrote the results into a MySQL database.

What is in it

11,365 projects, 449,213 files, 59,436 packages, 74.5 million lines of code (40.9 million of them source lines), 370,518 classes, 66,352 interfaces — and 32,844,836 individual measurements.

The example use the paper gives is measuring domain-specific language usage across the ecosystem, which is the question the XRDS column takes up. The dataset has its own page: Java Ecosystem Blueprints.

Written from the authors' copy of the paper (mirrored at Columbia's Network Security Lab); the published version is behind the DOI above and is not hosted here.