The Data -- The State of Public Code

One CSV with one row per repository carrying at least one contribution authored or committed from a public sector e-mail address as found in the Software Heritage archive snapshot of March 2026.

Contributor identities are pseudonymised, only the domain is preserved, so per-domain and per-country analyses remain possible.

File: world-gov-un_deduplicated_anonymised.csv.zst
Contents: 302,615 rows × 48 columns, zstd-compressed CSV
Size: 52 MB compressed
License: CC BY 4.0

Download the data

Decompress with zstd -d; pandas, DuckDB, Polars and R read .csv.zst directly. Full method in the report.

Schema

Timestamps are Unix epoch seconds unless noted. Empty cells mean the information was not available for that repository. “Contributed” always means: with a government author or committer.

Column	Description
Identification
`origin_id`	Internal Software Heritage identifier of the origin.
`origin_url`	URL of the repository as archived by Software Heritage.
Contribution counts
`num_contributed_revs`, `num_contributed_rels`, `num_contributed_revrels`	Contributed revisions (commits), releases, and their sum.
`num_contributed_revs_in_main_branch`	Contributed revisions reachable from the main branch.
`total_revs`, `total_rels`, `total_revrels`, `total_revs_in_main_branch`	Same counts, all contributors.
`num_revrels`	Alias of `num_contributed_revrels` (pipeline compatibility).
First / last government contribution
`first_contributed_revrel`, `last_contributed_revrel`	SWHID of the earliest / latest contributed revision or release.
`first_contributed_revrel_ts`, `last_contributed_revrel_ts`	Their timestamps.
`first_contributed_revrel_author`, `first_contributed_revrel_committer`, `last_contributed_revrel_author`, `last_contributed_revrel_committer`	Their author and committer (pseudonymised).
`first_revrel_ts`, `last_revrel_ts`	First / last revision or release overall, any author.
Citation metadata
`has_citation.cff`, `has_codemeta.json`	Whether the repository ships a `CITATION.cff` / `codemeta.json`.
`citation_cff`, `codemeta_json`	Reference to the file content when present.
Graph impact metrics (experimental)
`forward_paths`, `forward_paths_to_roots`, `forward_descendants`, `histhost_forward_paths`, `histhost_forward_paths_to_roots`, `histhost_forward_descendants`	Propagation of the repository's contents through the Software Heritage graph (`histhost_`: restricted to hosted history). Floating-point estimates; can overflow to `inf` for widely reused code.
Archive status
`last_visit_ts`, `visits_enabled`	Last Software Heritage crawl (ISO datetime); whether the origin is still crawled.
Hosting-platform metadata (where a forge API provided it)
`num_forks`, `num_likes`, `num_followers`, `num_stars`	Forks, stars/likes, watchers; `num_stars` comes from an alternate source -- prefer `num_likes`.
`creation_date`, `modification_date`, `hosting_repo_age_days`	Repository creation / last-modification dates and age on its platform.
`licenses`, `keywords`, `description`	Declared licenses (SPDX URLs), topics, description.
`programming_languages`, `primary_programming_language`	Detected languages and the dominant one.
Fork provenance
`forked_from`, `is_fork_with_newer_contributions`	Origin forked from, when known; whether the fork received government contributions after the fork point.