CppSATD: A Reusable Self-Admitted Technical Debt Dataset in C++

CppSATD is the first dataset of Self-Admitted Technical Debt (SATD) comments that focuses on C++ projects, one of the most widely-used programming languages. In addition to multi-class SATD annotations, the dataset also captures source code context around each comment, offering valuable insight into where and how SATD occurs in practice.

CppSATD is currently the largest multi-type SATD dataset with contextual source code, consisting of over 531,000 annotated comments, including more than 13,000 manually classified SATD instances.

Contents

File
Description

data_and_scripts.zip
Contains the full CppSATD dataset (cppsatd.csv), the manually annotated subset (manual_annotations.csv), and the Python extraction scripts with pattern files (cppsatd.data.extraction.zip).

repos.zip
Source code of the 5 C++ repositories from which SATD comments were extracted.

Annotation_Reference_Document.pdf
Guideline document used to ensure consistent annotation of SATD types among annotators.

SATD Types Covered

Design/Code Debt

Requirement Debt

Defect Debt

Test Debt

Documentation Debt

Each comment is labeled as one of the above SATD types or marked as NON-SATD.

How to Use

Explore Annotations: Use cppsatd.csv to analyze comment distributions, frequencies, and types.

Use Manual Labels: Rely on manual_annotations.csv for high-quality ground truth data in training or evaluation.

Reproduce Extraction: Refer to cppsatd.data.extraction.zip to understand or replicate the data collection and pattern matching process.

Inspect Repos: Use repos.zip if you want to see the actual C++ projects and their corresponding comments.

Checksums

A SIGNATURES.MD file is included to ensure the integrity of all files in this replication package. You can verify the data_and_scripts.zip using its corresponding MD5, SHA1, or SHA256 digest.

Contact

For questions, suggestions, or collaboration inquiries, please contact the dataset authors.