pgit

pgit

A Git-like version control CLI backed by PostgreSQL with pg-xpatch delta compression.

Note: pgit is primarily a demo for pg-xpatch delta compression. It's not intended to replace git—but it is genuinely useful for importing a repo and running SQL analytics on your commit history.

Why pgit?

Import any git repo. Analyze it instantly.

pgit init
pgit import /path/to/your/repo --branch main
pgit analyze coupling

No scripts, no parsing git log output. Just answers.

file_a                    file_b                    commits_together
────────────────────────  ────────────────────────  ────────────────
src/parser.rs             src/lexer.rs              127
src/db/schema.go          src/db/migrations.go      84
README.md                 CHANGELOG.md              63

Need something custom? Everything is in PostgreSQL — write your own SQL:

-- Same analysis as above, as raw SQL
SELECT pa.path, pb.path, COUNT(*) as times_together
FROM pgit_file_refs a
JOIN pgit_paths pa ON pa.group_id = a.group_id
JOIN pgit_file_refs b ON a.commit_id = b.commit_id AND a.group_id < b.group_id
JOIN pgit_paths pb ON pb.group_id = b.group_id
GROUP BY pa.path, pb.path
ORDER BY times_together DESC;

Features

• Git-familiar commands: init, add, commit, log, diff, checkout, push, pull, clone
• Pre-built analyses: churn, coupling, hotspots, authors, activity, bus-factor — one command each
• SQL queryable: Run arbitrary queries on your entire repo history
• Delta compression: pg-xpatch achieves competitive compression with git's packfiles (benchmark results)
• Search across history: pgit search "pattern" searches all versions of all files
• PostgreSQL as remote: Connection URL is your "remote" - no separate auth system
• Local development: Uses Docker/Podman container for local database
• Import from Git: Migrate existing repositories with full history

Quick Start

# Start the local database
pgit local start

# Initialize a new repository
pgit init
pgit config user.name "Your Name"
pgit config user.email "you@example.com"

# Basic workflow
pgit add .
pgit commit -m "Initial commit"
pgit log

# Or import an existing git repo
pgit init
pgit import /path/to/git/repo --branch main

Analyze Your Repository

pgit includes pre-built analyses that are optimized for the underlying storage engine. No SQL knowledge needed:

pgit analyze churn                   # most frequently modified files
pgit analyze coupling                # files always changed together
pgit analyze hotspots --depth 2      # churn aggregated by directory
pgit analyze authors                 # commits per contributor
pgit analyze activity --period month # commit velocity over time
pgit analyze bus-factor              # files with fewest authors (knowledge silos)

All commands support --json, --raw (for piping), --limit, and --path (glob filter). Results are displayed in an interactive table with search, column expand/hide, and clipboard copy (y/Y).

Query Your Repository

For custom queries, pgit stores everything in PostgreSQL so you can query it directly:

# Built-in search across all history
pgit search "TODO" --path "*.rs"
pgit search --all "panic!" --ignore-case

# Raw SQL access
pgit sql "SELECT * FROM pgit_commits ORDER BY authored_at DESC LIMIT 10"

Example Queries

-- Most frequently changed files
SELECT p.path, COUNT(*) as versions
FROM pgit_file_refs r
JOIN pgit_paths p ON p.group_id = r.group_id
GROUP BY p.path
ORDER BY versions DESC
LIMIT 10;

-- Files by extension
SELECT
  COALESCE(NULLIF(SUBSTRING(path FROM '\.([^.]+)$'), ''), '(no ext)') as extension,
  COUNT(*) as file_count
FROM pgit_paths
GROUP BY extension
ORDER BY file_count DESC
LIMIT 15;

See pgit sql examples for more, or check docs/xpatch-query-patterns.md for query optimization tips.

Installation

Using Go (recommended)

go install github.com/imgajeed76/pgit/v3/cmd/pgit@latest

From GitHub Releases

Download pre-built binaries from Releases:

• Linux: pgit_*_linux_amd64.tar.gz or pgit_*_linux_arm64.tar.gz
• macOS: pgit_*_darwin_amd64.tar.gz or pgit_*_darwin_arm64.tar.gz
• Windows: pgit_*_windows_amd64.zip

Package Managers

# Debian/Ubuntu
sudo dpkg -i pgit_*_linux_amd64.deb

# RHEL/Fedora
sudo rpm -i pgit_*_linux_amd64.rpm

# Alpine
sudo apk add --allow-untrusted pgit_*_linux_amd64.apk

Requirements

• Docker or Podman - Required for the local database container
• PostgreSQL with pg-xpatch - Optional, only needed for remote operations (push/pull/clone)

Compression: pgit vs git

Benchmarked on 19 real repositories across 6 languages (193k total commits). Comparing git gc --aggressive packfile vs pgit actual data (excluding indexes for both):

Scorecard: pgit 9 wins, git 9 wins, 1 tie.

See the full benchmark results for all 19 repositories with detailed per-repo breakdowns, charts, and methodology.

pgit uses pg-xpatch delta compression with zstd. Results vary by repository — pgit tends to win on source-code-heavy repos with incremental changes, while git wins on repos with large vendored dependencies or binary assets.

# Build the benchmark tool
go build -o pgit-bench ./cmd/pgit-bench/

# Run against the curated repo list (19 repos, ~50 min at 3 parallel)
./pgit-bench --file bench_repos.txt --parallel 3 --report BENCHMARK.md

./pgit-bench https://github.com/tokio-rs/tokio

Commands

Remote Setup

To test push/pull/clone, you can spin up a pg-xpatch container as your remote:

# Start a pg-xpatch container (creates database 'myproject' automatically)
docker run -d --name pgit-remote \
  -e POSTGRES_USER=pgit \
  -e POSTGRES_PASSWORD=pgit \
  -e POSTGRES_DB=myproject \
  -p 5433:5432 \
  ghcr.io/imgajeed76/pg-xpatch:latest

# Add it as a remote and push
pgit remote add origin postgres://pgit:pgit@localhost:5433/myproject
pgit push origin

The database name can be anything you want - just make sure it matches in both the container and the connection URL. pgit will initialize the schema automatically on first push.

Shell Completions

# Bash
pgit completion bash > /etc/bash_completion.d/pgit

# Zsh
pgit completion zsh > "${fpath[1]}/_pgit"

# Fish
pgit completion fish > ~/.config/fish/completions/pgit.fish

Environment Variables

License

MIT

Built with ❤️ by Oliver Seifert