Git Sparse Index and Partial Clones for Large Monorepos

Home » Git » Git Sparse Index and Partial Clones for Large Monorepos
April 30, 2025 Git
Git Sparse Index and Partial Clones for Large Monorepos

When a Git repository grows to hundreds of thousands of files, even simple operations like git status become slow because Git must read the full index. Two relatively recent features — partial clones (fetch only the objects you need) and the sparse index (compress index entries for paths outside your sparse-checkout cone) — together cut checkout time and memory usage by an order of magnitude.

Problem: Checking out a large Git monorepo — one containing WordPress themes, plugins, and build artifacts — takes minutes because Git downloads the full history and all files even when only a specific package is being worked on.

Solution: Use a partial clone with git clone --filter=blob:none to skip downloading file blobs at clone time (fetched on demand), and a sparse index with git sparse-checkout init --cone to limit the working tree to specific directories. Both features work together to dramatically reduce checkout time for large repos.


The commands below set up a partial clone with blobless filtering, enable a cone-mode sparse-checkout scoped to one package, activate the sparse index, and verify the speedup with git status timing.


# 1. Partial clone — skip downloading file blobs until they are needed
git clone --filter=blob:none --no-checkout \
    https://github.com/org/giant-monorepo.git
cd giant-monorepo

# 2. Enable cone-mode sparse-checkout (faster path matching than pattern mode)
git sparse-checkout init --cone

# 3. Check out only the package you care about
git sparse-checkout set packages/my-package

# 4. Activate the sparse index (Git 2.35+)
#    Compresses index entries for paths outside the cone into a single entry
git config core.sparseIndex true
git checkout main

# 5. Verify — should be very fast even in a huge repo
time git status

# 6. Add another package to the working tree later
git sparse-checkout add packages/shared-utils

# 7. Check which paths are in the sparse-checkout cone
git sparse-checkout list

# 8. Pre-fetch blobs for your cone so offline work is fully supported
git fetch --filter=blob:none origin main
git checkout -b feature/my-work origin/main


# Inspect index size difference
# Without sparse index:
git config core.sparseIndex false && git status > /dev/null
wc -c .git/index

# With sparse index:
git config core.sparseIndex true
git sparse-checkout reapply   # rebuilds the index in sparse form
wc -c .git/index              # typically 10–50× smaller in large repos


NOTE: The sparse index is incompatible with some Git operations that need the full tree (e.g., git merge across cone boundaries); Git will automatically densify the index for those operations and re-sparsify afterwards, so it is safe to enable globally on a developer machine.

Related posts