Eremey Valetov
ff06506bbc
Test corpus (empty, text, binary, compressible, incompressible) with reference archives created by original UC2 v2.3 in DOSBox. Two CTest tests: test_identify (magic detection) and test_extract (full extraction pipeline verified byte-for-byte against corpus).
31 lines
1.7 KiB
Plaintext
31 lines
1.7 KiB
Plaintext
UltraCompressor II was a DOS-era archiver created by Nico de Vries between
|
|
1992 and 1996. It was notable for its advanced deduplication system called
|
|
"master blocks", file versioning within archives, and competitive compression
|
|
ratios on the hardware of its day.
|
|
|
|
The archiver used an LZ77 sliding-window compressor with Huffman entropy
|
|
coding. The algorithm operates on a 64KB circular buffer with hash-chain
|
|
match finding. Matches of 3 to 32760 bytes are supported, with lazy
|
|
evaluation to find better matches at adjacent positions.
|
|
|
|
Huffman trees are serialized using delta coding against the previous block's
|
|
tree, with a nested Huffman code for the delta symbols. This is remarkably
|
|
efficient for typical data where consecutive blocks have similar symbol
|
|
distributions.
|
|
|
|
The deduplication system works by identifying common data blocks across files
|
|
and storing them only once as "master blocks". When a file's compressed data
|
|
matches an existing master, only a reference is stored. This was ahead of its
|
|
time -- modern tools like borg and restic use similar content-defined chunking.
|
|
|
|
UC2 v3.0.0 is a cross-platform revival of this archiver, built on Jan
|
|
Bobrowski's clean-room portable decompressor (libunuc2). The project brings
|
|
UC2 back as a modern, portable C99 tool that runs on Linux, macOS, Windows,
|
|
and even DOS via DJGPP cross-compilation.
|
|
|
|
This text file serves as part of the test corpus for verifying the extraction
|
|
pipeline. It contains enough English prose to exercise the typical symbol
|
|
distribution paths in the decompressor, including the Huffman tree generation
|
|
and the LZ77 back-reference matching for repeated phrases like "master blocks"
|
|
and "compression" which appear multiple times.
|