racjin 1.0.1

Compress and decompress Racjin game archives (CFC.DIG / CDDATA.DIG)

racjin

Python implementation of the Racjin compression algorithm used in PS2/PSP/Wii game archives.

Handles CFC.DIG and CDDATA.DIG archive files found in games by Racjin and other Japanese studios.

If this helped you, consider buying me a coffee ☕

Supported Games

• Fullmetal Alchemist and the Broken Angel (PS2)
• Fullmetal Alchemist 2: Curse of the Crimson Elixir (PS2)
• Fullmetal Alchemist 3: Kami o Tsugu Shoujo (PS2)
• Naruto: Uzumaki Chronicles 1 & 2 (PS2)
• Bleach: Soul Carnival 2 (PSP)
• Naruto Shippuden: Legends: Akatsuki Rising (PSP)
• Naruto Shippuden: Ultimate Ninja Impact (PSP)
• Other titles using the same archive format

Installation

uv add racjin

Or with pip:

pip install racjin

For development:

uv sync

CLI Usage

Extract an entire archive

# Auto-detects format and endianness
racjin extract CDDATA.DIG output/

# Specify format explicitly
racjin extract CFC.DIG output/ --format cfc

# For Wii/GameCube (big-endian)
racjin extract CDDATA.DIG output/ --big-endian

Compress / decompress individual files

# Decompress (requires known decompressed size)
racjin decompress compressed.bin decompressed.bin --size 577152

# Compress
racjin compress input.bin compressed.bin

Python API

from racjin import compress, decompress

# Decompress
with open("compressed.bin", "rb") as f:
    compressed_data = f.read()

decompressed = decompress(compressed_data, expected_size)

# Compress
with open("input.bin", "rb") as f:
    original_data = f.read()

compressed = compress(original_data)

# Round-trip
assert decompress(compress(data), len(data)) == data

Algorithm

The Racjin compression uses 9-bit tokens packed into a bitstream:

• Bit 8 = 1: Literal byte (lower 8 bits stored directly)
• Bit 8 = 0: Back-reference using a context-sensitive sliding window
  • Bits 3-7: 5-bit frequency index
  • Bits 0-2: 3-bit length (actual length = value + 1, max 8 bytes)
  • Window is a 8192-entry table indexed by (freq_index + last_byte * 32)

Tokens are folded using a rotating bit shift (0-7), packing 8 tokens (72 bits) into 9 bytes.

Archive Formats

Three structure variants are supported:

Format	Field Layout	Notes
cfc	offset, comp_size, section_count|flags, decomp_size	Most common (post-2005)
cddata	offset, decomp_size, section_count, comp_size	All sizes in sectors (2004-era)
cddata-old	offset, size, section_count	No compression (pre-2004)

All offsets are in 2048-byte disc sectors.

Credits

Algorithm reverse-engineered by Raw-man (C++ implementation). This is a Python port.

License

GPL-3.0 (same as the reference implementation)