Git history and code archaeology

When the input is a Git repository, RepoDNA reads its history to answer questions that the files alone cannot: how the project grew, who worked on what, what changes most, and how active it is now.

repodna history
repodna hotspots
repodna show --section history,contributors

History view of RepoDNA's own analysis

How history is read#

RepoDNA runs the git executable, read-only, through a hardened runner: it streams git log with raw changes and line counts in one pass, and reads branches, tags, and remotes separately. It does not link a Git library, so it supports every repository format your Git supports. Without Git, the history sections are marked unavailable and the rest of the analysis still runs.

Which commits: those reachable from HEAD, newest first, up to analysis.max_commits (50,000 by default; --max-commits for one run). If the limit is reached, git.historyTruncated is set and reports say so.
Shallow clones (for example CI checkouts with fetch-depth: 1) are detected and flagged with git.shallow; history-based results then describe only the fetched commits. For a full picture, analyze a full clone.
Merge commits are counted but carry no file changes of their own.
Renames are followed, so a file's history continues across moves.
Identities are grouped by e-mail address after applying the repository's .mailmap.
The artifact stores up to analysis.artifact_commits (1,000) commits with hash, date, author, and subject; statistics use every analyzed commit.

Everything that should not depend on when you run the analysis (recent churn, "recent" windows) is measured backwards from the latest commit, so the same revision always gives the same numbers. Only the activity level, which answers "is it active now?", is measured against the analysis time, and the artifact records that time.

What it reports#

Result	Meaning
Timeline	Commits, authors, and lines added and removed per month, and commits per day. Histories shorter than three months are charted per day; reports chart long histories per quarter or year.
Contributors	Per identity: commits, lines added and removed, first and last commit, active days, and main areas.
Ownership	Share of commits by the top contributor, and how many contributors authored half of all commits.
Activity level	`very-active` (at least 30 commits in the last 30 days), `active` (at least 5), `moderate` (at least one in 90 days), `low` (at least one in 365 days), or `dormant`.
Releases	Tags that look like versions (`v1.2.3`, `2.0`, `v1.0.0-rc.1`), with their dates and the commits between them.
Dormant periods	Gaps without commits of at least `dormant_days` (90).
File history	Per file: commits, authors, lines added and removed, when it first appeared and last changed, and earlier paths.
Hotspots	Files that change often and are complex; see below.
Recent changes	What changed in the `recent_days` (90) before the latest commit: commits, files, directories, and dependency changes.

Hotspots#

A hotspot is a file where change and complexity meet: code that is modified often and is hard to modify safely. Files changed in at least hotspot_min_commits (3) commits are ranked by a weighted sum of percentile ranks:

Signal	Weight
Commits touching the file	30%
Churn (lines added and removed)	20%
Commits in the recent window	15%
Highest function complexity	15%
Code lines	10%
Distinct authors	5%
Files that import it	5%

Each hotspot lists the signals that put it there, such as "Changed in 42 commits" or "Contains a function with complexity 31". See also complexity.

People#

History describes how a repository was worked on, not the quality of anyone's work. RepoDNA reports identities and their commits, never judgments about people. Squash merges, pair programming, bots, and people using several e-mail addresses all distort attribution.

--anonymize (or privacy.anonymize_contributors) replaces names with "Contributor 1", "Contributor 2", and so on in the artifact.
--no-commit-messages (or privacy.include_commit_messages = false) keeps commit subjects out of the artifact.
--privacy public replaces names with pseudonyms in reports and exports.

See privacy.

Findings#

activity.hotspot, activity.high-churn-area, contributors.concentration, evolution.dormant-period, evolution.abandoned-area, and evolution.new-module; see the rules reference. For how the repository changed over time, see the Time Machine.