Software Heritage Graph Dataset

Software Heritage is the largest existing public archive of software source code and development history. The dataset is a fully deduplicated Merkle DAG representation linking file content, directories, commits, and repository states from major forges, distributions, and package managers. Author and committer information is anonymized.

Use Cases

• Analyze software evolution patterns based on the linked Version Control System commits.
• Study code provenance and deduplication based on the Merkle DAG representation.
• Research source code distribution and origins based on data from major forges and package managers.
• Investigate digital preservation techniques based on the archive's crawling and timestamp information.

Strengths

• Described as the largest existing public archive of software source code.
• Links together file content, directories, commits, and repository states into a fully deduplicated graph.
• Sources include major development forges, FOSS distributions, and language-specific package managers.

Limitations

• Column-level documentation is absent; field semantics must be inferred after download.
• Row count and dataset size are unknown, which may limit suitability assessment.
• Author and committer information is anonymized, limiting attribution studies.

Provenance

Source

Software Heritage organization.

Collection Method

Periodic crawls of major development forges, FOSS distributions, and package managers.

Time Range

Timestamps about when artifacts were observed are included, but specific coverage is not stated.

Freshness

Last update date is unknown; freshness unverified.

Geography

Spatial coverage is not specified.