rargz

rargz is a fast parallel tar + zstd archiver. It walks a directory, emits a deterministic tar stream, and compresses chunks concurrently with independent zstd workers.

671,352 files → 27 seconds:

Features

• Parallel tar stream generation with independent zstd workers
• Optional progress indicator (auto-disables when stderr is not a TTY)
• Produces standard .tar.zst by default (compatible with tar --zstd or zstd -d)
• Optional .rargz chunked format (designed for future parallel decompression)
• Streaming extraction (stdin to filesystem)

Installation

cargo install rargz

From source:

cargo install --path .

Archiving

Create a standard .tar.zst archive and write it to stdout:

rargz path/to/input > archive.tar.zst

Enable the chunked .rargz format:

rargz --format rargz path/to/input > archive.rargz

Tune chunk size (bytes) and thread count:

rargz --chunk-size 4194304 --jobs 8 path/to/input > archive.tar.zst

Raise the compression level (zstd 1–22, default 3) and silence the progress spinner:

rargz --level 9 --no-progress path/to/input > archive.tar.zst

Extraction

Extraction reads from stdin and writes the output to the directory specified with -o/--output:

rargz --extract -o /path/to/output < archive.tar.zst

or

cat archive.tar.zst | rargz --extract -o /path/to/output

.tar.zst streams are decompressed sequentially for compatibility.
.rargz streams are chunked and designed for parallel decompression (WIP until format is finalized).

You can also extract directly from a file without an explicit pipe:

rargz --extract -o /path/to/output archive.tar.zst

Format details

• .tar.zst mode produces a pure tar stream compressed with zstd. Fully compatible with standard tools:

  • tar --zstd -xf archive.tar.zst
  • zstd -d archive.tar.zst | tar -xf -

• .rargz mode adds a small header (RARGZ\0, version, chunk size) and length-prefixes each compressed chunk. This enables random access and parallel decompression.

Inspection

Read-only operations work on both .tar.zst and .rargz streams supplied via a file or stdin.

• --list prints every entry path:

  rargz --list archive.tar.zst
  # or
  rargz --list < archive.tar.zst
  

• --count reports how many entries are present:

  rargz --count archive.tar.zst
  

Both modes consume the entire archive to validate it; they exit non-zero on structural errors.

Performance notes

• --jobs controls the compression worker pool (defaults to the number of logical CPUs).
• --chunk-size specifies the amount of tar data each compressor sees at once; larger chunks improve ratio, smaller chunks reduce memory per worker.
• --format rargz emits independently compressed chunks with a lightweight header for future parallel extraction support.
• --no-progress disables the progress indicator (also auto-disables when stderr is not a TTY).

Safety

• Parent directory components (..) are removed from archive paths to prevent extraction from writing outside the target tree.
• Any error in the pipeline cancels all worker threads immediately.

License

MIT. See LICENSE for details.