.zdelta Files

The zdelta cargo feature adds file-level helpers on top of the diffing engine: zstd compression and tar containers for shipping diffs.

[dependencies]
zbsdiff = { version = "1.6", features = ["zdelta"] }

A .zdelta file is a zstd frame holding one of two shapes:

• Single-target shape: the decompressed payload is a raw bsdiff patch (magic BSDIFF40) transforming exactly one source into one target.
• Container shape: the decompressed payload is a tar archive holding a meta.json descriptor plus one or more patch or payload files, used when several files change together (for example pooled package archives). Containers written by write_container carry the format identifier zoi.zdelta.v1 in meta.json.

Generating a container

use serde_json::json;
use zbsdiff::zdelta;

let patch = zdelta::diff_bytes(old_bytes, new_bytes)?;
let meta = json!({
    "format": zdelta::ZDELTA_FORMAT_ID,
    "entries": ["data.bsdiff"],
});
zdelta::write_container(
    &meta,
    &[("data.bsdiff".to_string(), patch)],
    "update.zdelta".as_ref(),
)?;

Use write_container_bytes when the compressed bytes should stay in memory, and zstd_compress for compressing a raw single patch.

Applying a .zdelta file

use zbsdiff::zdelta::{self, ZDelta};

let rebuilt = match zdelta::load("update.zdelta".as_ref())? {
    ZDelta::Container(_, files) => {
        let patch = files.get("data.bsdiff").expect("patch entry");
        zdelta::apply_bytes(old_bytes, patch)?
    }
    ZDelta::Single(patch) => zdelta::apply_bytes(old_bytes, &patch)?,
};

load (and load_bytes) detects either shape automatically. Decompression is capped at 64 GiB of output per file. See the custom formats guide for decoupling the diffing engine from storage entirely.