Colauntangle: 300 Synthesized Tangled Git Commits from Java and C# Projects

Colauntangle Post-cutoff dataset contains 300 synthesized tangled commits collected from open-source Java and C# projects. The benchmark is constructed by combining pairs of atomic commits using git cherry-pick, following the methodology of the original Flexeme benchmark. Each data point contains a tangled commit and its corresponding atomic commits, which serve as ground-truth decomposition answers.

Use Cases

• Benchmarking code untangling algorithms based on synthesized entangled commits.
• Training machine learning models to separate concerns in composite code changes.
• Evaluating the performance of commit decomposition tools against a known ground truth.
• Studying the characteristics of tangled changes in Java and C# ecosystems.

Strengths

• Contains 300 data points, each with a synthesized tangled commit and its corresponding atomic commits.
• Uses a defined methodology (git cherry-pick) consistent with an established benchmark (Flexeme).
• Sources commits from 14 named, representative open-source projects (9 Java, 5 C#).

Limitations

• Column-level documentation is absent; field semantics must be inferred after download.
• Row count is unknown, which may limit suitability assessment.
• Data is synthesized, which may not fully reflect the distribution of real-world tangled commits.

Provenance

Source

figshare, author colauntangle colauntangle

Collection Method

Commits were synthesized by combining pairs of atomic commits from selected open-source repositories using git cherry-pick.

Freshness

Last updated 2026-05-14 18:05:23