git-2025-features（git 2025 功能）

📌 NOTE: For detailed Git 2.49+ features (git-backfill, path-walk API, zlib-ng), see git-2-49-features.md skill.

🚨 CRITICAL GUIDELINES

Windows File Path Requirements

MANDATORY: Always Use Backslashes on Windows for File Paths

When using Edit or Write tools on Windows, you MUST use backslashes (\) in file paths, NOT forward slashes (/).

Examples:

• ❌ WRONG: D:/repos/project/file.tsx
• ✅ CORRECT: D:\repos\project\file.tsx

This applies to:

• Edit tool file_path parameter
• Write tool file_path parameter
• All file operations on Windows systems

Documentation Guidelines

NEVER create new documentation files unless explicitly requested by the user.

• Priority: Update existing README.md files rather than creating new documentation
• Repository cleanliness: Keep repository root clean - only README.md unless user requests otherwise
• Style: Documentation should be concise, direct, and professional - avoid AI-generated tone
• User preference: Only create additional.md files when user specifically asks for documentation

Git 2025 Features - Advanced Capabilities

Git 2.49 (March 2025) - Latest

Major additions: git-backfill, path-walk API, zlib-ng performance, improved delta compression.

See git-2-49-features.md for complete coverage.

Git 2.48-2.49 Features

Reftables Migration (Completed in 2.48)

What: New reference storage format replacing loose ref files and packed-refs.

Benefits:

• Faster ref operations (50-80% improvement)
• Atomic ref updates
• Better scalability for repositories with many refs
• Reflogs fully migratable (completed in 2.48)

Migration:

# Check current ref storage format
git config core.refStorage

# Migrate to reftables
git refs migrate --ref-storage=reftables

# Verify migration
git fsck --full
git log --oneline -5

# Roll back if needed (before critical operations)
git refs migrate --ref-storage=files

When to use:

• Repositories with 10,000+ refs
• High-frequency branch operations
• CI/CD systems creating many temporary refs
• Monorepos with extensive branching

Performance Milestones (2.48-2.49)

Git 2.48:

• Memory leak free status achieved
• Stable memory usage in long-running operations

Git 2.49:

• zlib-ng integration: 20-30% faster compression
• Path-walk API: 50-70% better delta compression
• New name-hashing algorithm for optimal packfiles

Benefits automatically in:

• Large repository clones
• Extended rebase sessions
• Bulk operations (filter-repo, GC, repack)

Sparse-Checkout (Enhanced in 2.48)

What: Check out only a subset of files from repository.

Use cases:

• Monorepos (work on one service)
• Large repositories (reduce disk usage)
• Build systems (fetch only needed files)

Cone Mode (Default - Recommended):

# Clone with sparse-checkout
git clone --filter=blob:none --sparse <repo-url>
cd <repo>

# Initialize sparse-checkout in cone mode
git sparse-checkout init --cone

# Add directories to checkout
git sparse-checkout set src/api src/shared docs

# Add more directories
git sparse-checkout add tests/integration

# View current patterns
git sparse-checkout list

# Check what would be matched
git sparse-checkout check-rules src/api/users.ts

# Disable sparse-checkout
git sparse-checkout disable

Advanced Patterns (Non-Cone Mode):

# Enable pattern mode
git sparse-checkout init --no-cone

# Add patterns (one per line)
git sparse-checkout set \
  "*.md" \
  "src/api/*" \
  "!src/api/legacy/*"

# Read patterns from file
git sparse-checkout set --stdin < patterns.txt

Reapply Rules:

# After merge/rebase that materialized unwanted files
git sparse-checkout reapply

Partial Clone

What: Clone repository without downloading all objects initially.

Filters:

1. blob:none - Defer all blobs (fastest, smallest)
2. tree:0 - Defer all trees and blobs
3. blob:limit=1m - Defer blobs larger than 1MB

Usage:

# Clone without blobs (fetch on demand)
git clone --filter=blob:none <repo-url>

# Clone without large files
git clone --filter=blob:limit=10m <repo-url>

# Combine with sparse-checkout
git clone --filter=blob:none --sparse <repo-url>
cd <repo>
git sparse-checkout set src/api

# Convert existing repository to partial clone
git config extensions.partialClone origin
git config remote.origin.promisor true
git fetch --filter=blob:none

# Prefetch all missing objects
git fetch --unshallow

Combine Partial Clone + Sparse-Checkout:

# Ultimate efficiency: Only objects for specific directories
git clone --filter=blob:none --sparse <repo-url>
cd <repo>
git sparse-checkout set --cone src/api
git checkout main

# Result: Only have objects for src/api

Check promisor objects:

# Verify partial clone status
git config extensions.partialClone

# See promisor packfiles
ls -lah .git/objects/pack/*.promisor

# Force fetch specific object
git rev-list --objects --missing=print HEAD | grep "^?"

Git Worktrees

What: Multiple working directories from one repository.

Benefits:

• Work on multiple branches simultaneously
• No need to stash/commit before switching
• Parallel work (review PR while coding)
• Shared.git (one fetch updates all)

Basic Operations:

# List worktrees
git worktree list

# Create worktree for existing branch
git worktree add ../project-feature feature-branch

# Create worktree with new branch
git worktree add -b new-feature ../project-new-feature

# Create worktree from remote branch
git worktree add ../project-fix origin/fix-bug

# Remove worktree
git worktree remove ../project-feature

# Clean up stale worktree references
git worktree prune

Advanced Patterns:

# Worktree for PR review while coding
git worktree add ../myproject-pr-123 origin/pull/123/head
cd ../myproject-pr-123
# Review PR in separate directory
cd -
# Continue coding in main worktree

# Worktree for hotfix
git worktree add --detach ../myproject-hotfix v1.2.3
cd ../myproject-hotfix
# Make hotfix
git switch -c hotfix/security-patch
git commit -am "fix: patch vulnerability"
git push -u origin hotfix/security-patch

# Worktree organization
mkdir -p ~/worktrees/myproject
git worktree add ~/worktrees/myproject/feature-a -b feature-a
git worktree add ~/worktrees/myproject/feature-b -b feature-b
git worktree add ~/worktrees/myproject/pr-review origin/pull/42/head

Best Practices:

1. Organize directory structure:

~/projects/
  myproject/           # Main worktree
  myproject-feature/   # Feature worktree
  myproject-review/    # Review worktree

1. Clean up regularly:

# Remove merged worktrees
git worktree list | grep feature | while read wt branch commit; do
  if git branch --merged | grep -q "$branch"; then
    git worktree remove "$wt"
  fi
done

1. Shared configuration:

• .git/config applies to all worktrees
• .git/info/exclude applies to all worktrees
• Each worktree has own index and HEAD

Scalar (Large Repository Tool)

What: Tool for optimizing very large repositories (Microsoft-developed).

# Install scalar (comes with Git 2.47+)
scalar register <path>

# Clone with scalar optimizations
scalar clone --branch main <repo-url>

# Enables automatically:
# - Sparse-checkout (cone mode)
# - Partial clone (blob:none)
# - Multi-pack-index
# - Commit-graph
# - Background maintenance

# Unregister
scalar unregister <path>

# Delete repository
scalar delete <path>

Git Backfill (Experimental)

What: Background process to fetch missing objects in partial clone.

# Fetch missing blobs in background
git backfill

# Configure batch size
git backfill --min-batch-size=1000

# Respect sparse-checkout patterns
git backfill --sparse

Performance Comparison

Traditional Clone:

git clone large-repo
# Size: 5GB, Time: 10 minutes

Sparse-Checkout:

git clone --sparse large-repo
git sparse-checkout set src/api
# Size: 500MB, Time: 3 minutes

Partial Clone:

git clone --filter=blob:none large-repo
# Size: 100MB, Time: 1 minute

Partial Clone + Sparse-Checkout:

git clone --filter=blob:none --sparse large-repo
git sparse-checkout set src/api
# Size: 50MB, Time: 30 seconds

When to Use Each Feature

Sparse-Checkout:

• ✓ Monorepos
• ✓ Working on specific services/modules
• ✓ Limited disk space
• ✗ Need entire codebase often

Partial Clone:

• ✓ CI/CD pipelines
• ✓ Large repositories
• ✓ Good network connectivity
• ✗ Offline work frequently

Worktrees:

• ✓ Parallel development
• ✓ PR reviews during work
• ✓ Multiple branch testing
• ✗ Low disk space

Combine All:

• ✓ Massive monorepos (Google scale)
• ✓ Multiple simultaneous tasks
• ✓ Minimal local storage
• ✓ Fast network connection

Troubleshooting

Sparse-checkout not working:

# Verify configuration
git config core.sparseCheckout
git config core.sparseCheckoutCone

# Re-apply patterns
git sparse-checkout reapply

# Check patterns
git sparse-checkout list

Missing objects in partial clone:

# Fetch specific object
git fetch origin <commit>

# Fetch all missing
git fetch --unshallow

# Verify promisor config
git config extensions.partialClone

Worktree issues:

# Locked worktree
git worktree unlock <path>

# Corrupted worktree
git worktree remove --force <path>
git worktree prune

# Branch already checked out
git checkout --ignore-other-worktrees <branch>

Migration Guide

From traditional to optimized workflow:

# 1. Current large clone
cd large-project
du -sh .git  # 5GB

# 2. Create optimized new clone
cd ..
git clone --filter=blob:none --sparse large-project-new
cd large-project-new
git sparse-checkout set src/api src/shared

# 3. Verify size
du -sh .git  # 50MB

# 4. Switch workflow
cd ../large-project-new

# 5. Delete old clone when comfortable
rm -rf ../large-project

Resources

• Git Partial Clone Documentation
• Git Sparse-Checkout Guide
• Git Worktree Best Practices
• Scalar Documentation