aehobak

Aehobak transcodes binary patches from bsdiff. The goal is a byte-oriented format, compact and optimised for patch application speed. As compression efficiency is content-dependent, one should verify with a suitable corpus. The following results are for LZ4-compressed bsdiff patches of build artifacts that are no larger than 3% of the target object size. The bench example can report the same metrics for provided files.

LZ4-compressed aehobak patches yield a median reduction of 74.9%.

Uncompressed aehobak patches yield a median reduction of:

• 48.1% over LZ4-compressed bsdiff patches
• 99.1% over uncompressed bsdiff patches

Direct application of aehobak patches can achieve 75% of memcpy speed and is panic-free.

Usage

let old = vec![1, 2, 3, 4, 5];
let new = vec![1, 2, 4, 6];
let mut patch = Vec::new();
let mut encoded = Vec::new();

bsdiff::diff(&old, &new, &mut patch).unwrap();
aehobak::encode(&patch, &mut encoded).unwrap();

let mut decoded = Vec::with_capacity(patch.len());
let mut patched = Vec::with_capacity(new.len());
aehobak::decode(&mut encoded.as_slice(), &mut decoded).unwrap();
bsdiff::patch(&old, &mut decoded.as_slice(), &mut patched).unwrap();
assert_eq!(patched, new);

Diffing Files

fn diff_files(orig_file: &str, file: &str, patch_file: &str) -> std::io::Result<()> {
    let old = std::fs::read(orig_file)?;
    let new = std::fs::read(file)?;
    let mut patch = Vec::new();
    let mut encoded = Vec::new();

    bsdiff::diff(&old, &new, &mut patch)?;
    aehobak::encode(&patch, &mut encoded)?;
    std::fs::write(patch_file, &encoded)
}

Patching Files

fn patch_file(orig_file: &str, patch_file: &str, file: &str) -> std::io::Result<()> {
    let old = std::fs::read(orig_file)?;
    let patch = std::fs::read(patch_file)?;
    let mut new = Vec::with_capacity(10_000_000);

    aehobak::patch(&old, &patch, &mut new)?;
    std::fs::write(file, &new)
}