Skip to main content

Activity Analysis

Git history is a powerful place to look for anyone seeking to better understand code-related activities in an engineering organization. Commit history is a high-fidelity view into the actual behaviors and patterns of teams, as it's directly tied to an output of engineering organizations (code).

MergeStat can be used to analyze git activity for various purposes.

  • How frequently are we merging into main? (As a proxy for how effective we are at shipping code)
  • Who are the "experts" in certain repos or parts of a codebase? (So we know who to target for questions)
  • What parts of a project are seeing a lot of recent change (churn)? (Does this reflect a current feature initiative or do we have a flaky area in our codebase?)
  • What are trends in our coding activities over time?
  • How much (relative) effort is spent on what projects or parts of a codebase?
  • How much (relative) effort is spent on refactoring old code vs writing new code?
  • etc..

If it's a question with an answer encoded in git history, MergeStat should be able to answer it!

Git Activity​

By looking only at git commits (and associated file changes), we can answer some interesting questions. MergeStat makes this raw data accessible through SQL for ad-hoc querying. In addition, tools such as dbt can be used to define models on top of this raw data for easier use in downstream reporting and analytics.

We maintain this dbt project to showcase how raw git commit activity data can be turned into rollups and metrics for an analysis on the git activity of an engineering org.

Relatative repo activity

More on the Way!

We're putting together more examples of how git activity can be used to extract valuable insights about an engineering organization. Stay tuned for more!

Active Contributors​

It can be important to know how many active contributors a project has:

  • Many SaaS products charge based on the number of "active developers" (number of people who've committed code within a time window).
  • An evaluation of the health of an open-source or internal codebase should take into account the number of recent contributors. This can be a proxy for how maintained a codebase is.
  • Finding who the most recent contributors are to a project is a useful way to identify the right people to contact with questions.

Show the monthly count of unique authors (by email)​

SELECT
    date_trunc('month', author_when),
    count(distinct author_email)
FROM git_commits
JOIN repos ON repos.id = git_commits.repo_id
-- WHERE repo LIKE '...' -- uncomment to filter by repo
GROUP BY 1
ORDER BY 1 DESC

Show a list of authors in the last 90 days​

SELECT
    DISTINCT(author_email),
    author_name
FROM git_commits
JOIN repos ON repos.id = git_commits.repo_id
WHERE author_when > now() - '90 days'::interval