Zig● experimentalv0.2.1MITsince 2025 // ★ 318 // ⑂ 22
fragmentize
deterministic file chunker
a content-defined chunker (rolling hash with gear-table) that produces deterministic, dedupe-friendly chunks from arbitrary file streams. mostly an excuse to write zig. partly an excuse to think about borg, casync, and the long-tail of dedupe storage.
01 / install
$ zig build -Doptimize=ReleaseFast
$ ./zig-out/bin/fragmentize --help
02 / usage
$ fragmentize chunk huge_disk.img --avg=64K --min=16K --max=256K
$ fragmentize stats huge_disk.frags
# chunks: 14322 avg: 64.1K dedupe: 38.4%
03 / features
- rastrigin-gear rolling hash
- tunable min/avg/max chunk sizes
- 32 MB/s on a thinkpad x1 (no avx tricks)
- deterministic across architectures
04 / changelog
v0.2.12026.04fix endianness bug on aarch64
v0.2.02026.02deterministic across arches
v0.1.02025.12first checkin. zig is good actually.