format-pack Git Command Guide

The git format-pack format stores Git objects in compressed delta-encoded archives to optimize storage space, network transfer, and repository performance. Pack files are the primary storage mechanism for Git repositories.

Pack Format Specification

Pack File Header:

Field	Size	Description
Signature	4 bytes	’P’, ‘A’, ‘C’, ‘K’
Version	4 bytes	Version 2 or 3 (network byte order)
Object Count	4 bytes	Number of objects in pack (network byte order)

Object Entry Format:

Undeltified Objects:

• type and length (3-bit type, (n-1)*7+4-bit length)
• compressed object data

Delta Objects:

• type and length (3-bit type, (n-1)*7+4-bit length)
• base object name (SHA-1 for OBJ_REF_DELTA)
• compressed delta data

Delta Types:

Type Value	Name	Base Reference
6 (0b110)	OBJ_REF_DELTA	SHA-1 hash of base object
7 (0b111)	OBJ_OFS_DELTA	Negative offset to base object in pack

Object Types:

Type Value	Name	Description
1	OBJ_COMMIT	Commit object
2	OBJ_TREE	Tree object
3	OBJ_BLOB	Blob (file content) object
4	OBJ_TAG	Tag object

Caution

Visibility of packed data depends on creator of the pack. Some objects may be unreachable if pack prerequisites aren’t satisfied.

Examples of git pack operations:

Inspect pack files

ls -la .git/objects/pack/

List all pack files in the repository.

Verify pack integrity

git verify-pack .git/objects/pack/pack-*.idx

Check pack file integrity and show object statistics.

Create custom pack

git pack-objects custom-pack < object-list.txt

Create a custom pack file from specified objects.

Repack repository

git repack -a -d

Repack all objects into a single pack, removing old packs.

Unpack objects

find .git/objects/pack -name "*.pack" -exec git unpack-objects {} +

Unpack objects from pack files to loose objects.

Verify all packs

git for-each-ref --format='%(objectname)' | git pack-objects verify-pack

Verify packs contain expected objects.

Create thin pack

git pack-objects --thin thin-pack < object-list.txt

Create thin pack that references objects not in the pack.

Tip

Use git gc to automatically manage pack optimization. Manual pack operations are rarely needed in normal workflows.

How does Git decide when to create pack files?

Git creates pack files during git gc, git repack, or when the number of loose objects exceeds gc.autoPackLimit (default 50).

What types of objects can be packed?

All Git objects (commits, trees, blobs, tags) can be packed using delta compression to save space.

What’s the difference between thin and normal packs?

Thin packs reference base objects not included in the pack, allowing smaller transfers at the cost of requiring access to referenced objects.

How do I inspect what’s in a pack file?

Use git verify-pack -v .git/objects/pack/pack-*.idx to see all objects, their sizes, and delta chains.

How do I recover from a corrupted pack file?

For corrupted packs, try git fsck --full first. If recovery fails, use backups or attempt unpacking with git unpack-objects < packfile.

What affects pack compression size?

Compression depends on similarity between objects, Git version pack compression algorithm settings, and whether deltas are enabled.

Why do some objects remain loose?

Recent objects stay loose for performance. Only gc/periodic maintenance consolidates them into packs.

Applications of the pack format

1. Storage Efficiency: Compresses thousands of objects into single files with delta compression
2. Memory Mapping: Allows efficient memory mapping of object data during operations
3. Network Efficiency: Enables fast push/pull operations with thin pack optimization
4. Clone Performance: Speeds up initial cloning by bundling repository data
5. Backup Strategy: Facilitates repository backups and archival
6. Scalability: Supports repositories with millions of objects through multi-pack-index