Archive Structure

This document describes the high-level physical layout of a 7z archive file.

Physical Layout

A 7z archive consists of three main regions arranged sequentially:

┌─────────────────────────────┐  Offset 0x00
│     Signature Header        │  32 bytes (fixed)
│     (Start Header)          │
├─────────────────────────────┤  Offset 0x20
│                             │
│     Pack Data               │  Variable size
│     (Compressed Streams)    │  (may be empty)
│                             │
├─────────────────────────────┤  Offset 0x20 + NextHeaderOffset
│                             │
│     Next Header             │  NextHeaderSize bytes
│     (Header Database)       │
│                             │
└─────────────────────────────┘  EOF

Region Details

Signature Header

Offset: 0x00 Size: 32 bytes (fixed)

The signature header is always present and always 32 bytes. It contains:

• Archive signature (magic bytes)
• Format version
• CRC of header fields
• Pointer to the next header (offset and size)
• CRC of next header data

See 03-SIGNATURE-HEADER for the complete format.

Pack Data

Offset: 0x20 (immediately after signature header) Size: Variable (may be zero)

The pack data region contains all compressed data streams concatenated together:

┌─────────────────────────────┐
│     Pack Stream 0           │  PackSize[0] bytes
├─────────────────────────────┤
│     Pack Stream 1           │  PackSize[1] bytes
├─────────────────────────────┤
│          ...                │
├─────────────────────────────┤
│     Pack Stream N-1         │  PackSize[N-1] bytes
└─────────────────────────────┘

Key properties:

• Pack streams are stored contiguously with no gaps
• The number and sizes of streams are defined in 06-PACK-INFO
• Total pack data size = sum of all PackSize values
• Region may be empty if archive contains only empty files/directories

Next Header

Offset: 0x20 + NextHeaderOffset Size: NextHeaderSize bytes

The next header contains the archive metadata. It may be:

1. Plain Header - Uncompressed metadata starting with property ID 0x01
2. Encoded Header - Compressed/encrypted metadata starting with property ID 0x17

See 05-HEADER-STRUCTURE for header organization.

Data Flow

Reading an Archive

1. Read and validate signature header (32 bytes)
2. Calculate next header position: 32 + NextHeaderOffset
3. Seek to next header position
4. Read NextHeaderSize bytes
5. Verify NextHeaderCRC matches
6. Parse header (decompress if encoded)
7. Extract pack info, unpack info, substreams info, files info
8. For each file to extract:
   • Locate the folder containing the file
   • Decompress the folder's pack streams through coder chain
   • Extract the file's substream from the unpack stream

Writing an Archive

1. Collect all files to archive
2. Organize files into folders (one per file, or solid)
3. For each folder:
   • Concatenate file data
   • Apply coder chain (filters, compression, encryption)
   • Write resulting pack streams
4. Build header structures:
   • PackInfo with stream positions and sizes
   • UnpackInfo with folder and coder definitions
   • SubStreamsInfo with per-file sizes
   • FilesInfo with metadata
5. Optionally encode (compress) the header
6. Calculate CRCs
7. Write signature header with correct offsets

Empty Archive

An archive with no files has:

NextHeaderOffset = 0
NextHeaderSize = size of minimal header

The minimal header contains:

• 0x01 (Header property ID)
• 0x00 (End property ID)

Total empty archive size: 32 bytes (signature) + 2 bytes (minimal header) = 34 bytes.

Size Constraints

Maximum Sizes

Field	Maximum Value	Notes
NextHeaderOffset	2^64 - 1	Theoretical maximum
NextHeaderSize	2^64 - 1	Theoretical maximum
Total archive size	2^64 - 1	~18 EB theoretical
Practical archive size	Implementation-defined	Often 2^63 for signed offsets

Minimum Sizes

Archive Type	Minimum Size	Breakdown
Empty archive	34 bytes	32 (signature) + 2 (header: 01 00)
Single empty file	Variable	32 + header with FilesInfo
Single small file	Variable	32 + pack data + header

Empty archive calculation:

• Signature header: 32 bytes (fixed)
• Minimal next header: 2 bytes (0x01 Header + 0x00 End)
• Total: exactly 34 bytes

Offset Calculations

All offsets in the header are relative to specific reference points:

Offset Field	Reference Point
NextHeaderOffset	End of signature header (byte 32)
PackPos	End of signature header (byte 32)

Example Offset Calculation

For an archive with:

• 1000 bytes of pack data
• 200 bytes of header

Signature Header:  bytes 0-31     (32 bytes)
Pack Data:         bytes 32-1031  (1000 bytes)
Next Header:       bytes 1032-1231 (200 bytes)

NextHeaderOffset = 1000 (offset from byte 32 to byte 1032)
NextHeaderSize = 200

Validation Requirements

Implementations MUST verify:

1. Signature valid: First 6 bytes match 37 7A BC AF 27 1C
2. Version acceptable: Major version is 0
3. Offset in bounds: 32 + NextHeaderOffset + NextHeaderSize <= file_size
4. No overlap: Pack data region does not overlap header region
5. CRCs match: StartHeaderCRC and NextHeaderCRC are correct

Self-Extracting Archives

SFX archives prepend executable code before the signature header:

┌─────────────────────────────┐  Offset 0x00
│     Executable Stub         │  Variable size
│     (PE, ELF, or Mach-O)    │
├─────────────────────────────┤  SFX Offset
│     Signature Header        │  32 bytes
├─────────────────────────────┤
│     Pack Data               │
├─────────────────────────────┤
│     Next Header             │
└─────────────────────────────┘

To read an SFX archive:

1. Search for signature bytes (up to 1 MiB)
2. Validate version bytes after signature
3. Use found offset as new base for all calculations

See 15-SFX-ARCHIVES for complete SFX handling.

See Also

• Signature Header - Signature header format
• Header Structure - Header organization
• Pack Info - Pack stream information
• SFX Archives - Self-extracting archives