I manage a lot of Git repositories -- a few hundred at last count, spread across several directory trees. Checking each one by hand to see what's dirty, what needs pushing, or what's fallen behind got old fast.

git-all scans a directory tree, finds every Git repo (including submodules and worktrees), and runs a Git command in each one -- reporting only the repos that need attention. Clean repos stay silent, so you only see what matters.

Installation

Nix

nix run github:jwiegley/git-all -- status ~/projects

Cabal

cabal install

Make sure ~/.cabal/bin is on your PATH.

Usage

git-all                        # Status of repos under .
git-all status ~/projects      # Status under given directories
git-all -j8 status .           # 8 parallel workers
git-all -p status .            # Show "NEED PULL" sections
git-all -U status .            # Include untracked files
git-all fetch ~/projects       # Fetch all repos
git-all log ~/projects         # Run any git command

The output uses == markers for normal results and ## for errors. Only the first ten lines per repo are shown, with a count of any remaining.

Building from source

nix develop                    # Enter dev shell with all tools
cabal build                    # Build
cabal test                     # Run tests
nix flake check                # Run all checks (build, lint, format)

Build targets

nix run .#format -- Main.hs    # Format code with fourmolu
nix run .#lint -- Main.hs      # Lint with hlint

How it works

A producer thread walks the directory tree, writing (gitDir, workTree) pairs to a channel as it discovers .git directories. The main thread consumes the channel lazily and runs the requested Git command in each repository. Top-level directories are scanned in parallel (controlled by -j); subdirectories are scanned sequentially to avoid thread explosion.

License

BSD-3-Clause -- see LICENSE.md.