Aggregating GitHub Statistics

Problem Statement

We wanted to count how many lines of code and content were written in 2025 across all repositories in an organization. GitHub's web UI shows contribution graphs but doesn't provide aggregate line counts. We needed to use the GitHub API to extract this data.

Key Discoveries

1. Three Different APIs for Different Purposes

GitHub provides multiple APIs, each suited for different queries:
APIBest ForLimitations
GraphQL Contributions APICommit counts, contribution totalsNo line-level stats
Search Commits APIFinding commits by date/authorMax 1000 results, no line stats
Contributor Stats APIWeekly line additions/deletionsReturns 202 if not cached, needs triggering

2. The Contributor Stats API is Quirky

The /repos/{owner}/{repo}/stats/contributors endpoint is the only way to get line-level statistics, but it has important behaviors:
  • Returns 202 on first request: GitHub computes stats asynchronously. The first request triggers computation and returns HTTP 202. You must wait and retry.
  • Data is cached: Once computed, stats are cached for about a week.
  • Weekly granularity: Data is returned in weekly buckets with Unix timestamps.

3. Unix Timestamps for Date Filtering

To filter for 2025 data, use Unix timestamp 1735689600 (January 1, 2025 00:00 UTC).

Step-by-Step Guide

Prerequisites

Install and authenticate the GitHub CLI:
bash 
# Install gh CLI (macOS)
brew install gh

# Authenticate
gh auth login

# Verify authentication
gh auth status

Step 1: Get High-Level Contribution Counts (GraphQL)

This gives you authoritative commit counts from GitHub's contribution graph:
bash 
gh api graphql -f query='
{
  viewer {
    contributionsCollection(from: "2025-01-01T00:00:00Z", to: "2025-12-31T23:59:59Z") {
      totalCommitContributions
      totalIssueContributions
      totalPullRequestContributions
      totalPullRequestReviewContributions
      contributionCalendar {
        totalContributions
      }
    }
  }
}'
Example output:
json 
{
  "totalCommitContributions": 2354,
  "totalIssueContributions": 222,
  "totalPullRequestContributions": 9,
  "contributionCalendar": { "totalContributions": 2595 }
}

Step 2: Find Which Repos Have Your Commits (Search API)

Identify repositories with activity in a specific time period:
bash 
# Count total commits in an org for 2025
gh api "search/commits?q=author:YOUR_USERNAME+org:YOUR_ORG+committer-date:2025-01-01..2025-12-31&per_page=100" \
  --jq '.total_count'

# Group by repository (limited to first 100 results)
gh api "search/commits?q=author:YOUR_USERNAME+org:YOUR_ORG+committer-date:2025-01-01..2025-12-31&per_page=100" \
  --jq '[.items | group_by(.repository.full_name) | .[] | {repo: .[0].repository.full_name, count: length}] | sort_by(-.count)'
Note: The Search API returns a maximum of 1000 results across all pages. Use it to identify active repositories, then query each repo's stats individually.

Step 3: Get Line Statistics for a Single Repository

This is the core query to get additions and deletions:
bash 
gh api "repos/OWNER/REPO/stats/contributors" \
  --jq '.[] | select(.author.login == "YOUR_USERNAME") |
    .weeks | map(select(.w >= 1735689600)) |
    {
      commits: (map(.c) | add),
      additions: (map(.a) | add),
      deletions: (map(.d) | add)
    }'
Example output:
json 
{
  "commits": 245,
  "additions": 3180613,
  "deletions": 3185840
}
Understanding the response:
  • .w = Unix timestamp for the week
  • .a = Lines added that week
  • .d = Lines deleted that week
  • .c = Commits that week

Step 4: Handle 202 Responses (Stats Not Ready)

If the API returns empty or 202, you need to trigger stats generation:
bash 
# Trigger stats computation (ignore output)
gh api "repos/OWNER/REPO/stats/contributors" >/dev/null 2>&1

# Wait a few seconds
sleep 3

# Retry the query
gh api "repos/OWNER/REPO/stats/contributors" \
  --jq '.[] | select(.author.login == "YOUR_USERNAME") | ...'

Step 5: Query Multiple Repositories

Loop through repositories and aggregate:
bash 
#!/bin/bash
USERNAME="your_username"
ORG="your_org"

# List all repos in org
repos=$(gh repo list "$ORG" --limit 100 --json name --jq '.[].name')

total_adds=0
total_dels=0
total_commits=0

for repo in $repos; do
    # Trigger stats (in case not cached)
    gh api "repos/$ORG/$repo/stats/contributors" >/dev/null 2>&1
done

# Wait for GitHub to compute
sleep 5

for repo in $repos; do
    result=$(gh api "repos/$ORG/$repo/stats/contributors" 2>/dev/null | \
      jq -r --arg user "$USERNAME" '.[] | select(.author.login == $user) |
        .weeks | map(select(.w >= 1735689600)) |
        "\(map(.c) | add // 0) \(map(.a) | add // 0) \(map(.d) | add // 0)"' 2>/dev/null)

    if [ -n "$result" ] && [ "$result" != "0 0 0" ]; then
        commits=$(echo "$result" | cut -d' ' -f1)
        adds=$(echo "$result" | cut -d' ' -f2)
        dels=$(echo "$result" | cut -d' ' -f3)

        if [ "$commits" -gt 0 ] 2>/dev/null; then
            printf "%-40s %5d commits  +%-10d  -%d\n" "$repo" "$commits" "$adds" "$dels"
            total_adds=$((total_adds + adds))
            total_dels=$((total_dels + dels))
            total_commits=$((total_commits + commits))
        fi
    fi
done

echo "----------------------------------------"
printf "TOTAL: %d commits, +%d / -%d lines\n" "$total_commits" "$total_adds" "$total_dels"

Step 6: Query a Single Repo (Quick Reference)

For quick one-off queries:
bash 
# Replace these values
OWNER="lossless-group"
REPO="lossless-content"
USERNAME="mpstaton"
YEAR_START=1735689600  # Jan 1, 2025

gh api "repos/$OWNER/$REPO/stats/contributors" \
  --jq ".[] | select(.author.login == \"$USERNAME\") |
    .weeks | map(select(.w >= $YEAR_START)) |
    {commits: (map(.c) | add), added: (map(.a) | add), deleted: (map(.d) | add)}"

Common Unix Timestamps for Filtering

DateUnix Timestamp
Jan 1, 20241704067200
Jan 1, 20251735689600
Jan 1, 20261767225600
Generate a timestamp for any date:
bash 
date -j -f "%Y-%m-%d" "2025-01-01" "+%s"

Caveats and Limitations

  1. Large line counts may include non-code: Submodule updates, vendored dependencies, and auto-generated files inflate line counts significantly.
  2. Stats API coverage is incomplete: Not all repos return stats immediately. Running queries multiple times over several minutes improves coverage.
  3. Rate limits apply: GitHub API has rate limits. For large organizations, space out requests or use authenticated requests for higher limits.
  4. Private repos require appropriate scopes: Ensure your gh token has repo scope for private repositories.

Example Results

Running these queries for mpstaton in the lossless-group org for 2025:
text 
From GitHub Contribution Graph (authoritative):
  Total Contributions:     2,595
  Total Commits:           2,354
  Issues Created:            222
  Pull Requests:               9

From Stats API (partial coverage):
  Commits Analyzed:          720
  Lines Added:        +4,262,627
  Lines Deleted:      -3,862,995
  Net Lines:            +399,632
bash 
gh api "repos/OWNER/REPO/stats/contributors" \
    --jq '.[] | select(.author.login == "mpstaton") |
      .weeks | map(select(.w >= 1735689600)) |
      {commits: (map(.c) | add), added: (map(.a) | add), deleted: (map(.d) | add)}'