snappy_framed

(Work in progress: Good enough for basic use, but may panic on corrupt input, and SnappyFramedEncoder assumes you normally write large buffers. APIs may change before 1.0.)

[API documentation][doc]

Rust support for [Snappy framed compression][framed], a streaming version of Google's high-speed [Snappy][snappy] compression scheme. We provide standard implementations of Read and Write for working with these streams.

Benchmarked at 807 MB/s with CRC32C checksums disabled:

test read::benches::decompress_speed ... bench:    972211 ns/iter (+/- 213429) = 807 MB/s

To use, add the following to the [dependencies] section of your Cargo.toml file:

snappy_framed = "*"

And import it as follows at the top-level of your crate:

extern crate snappy_framed;

You may also be interested in [snzip][snzip], a command-line tool for compressing and decompressing the [Snappy framed][framed] format.

Benchmarking notes

Using a nightly build of Rust, it's possible to profile and optimize the decompressor:

cargo bench --features unstable
valgrind --tool=callgrind target/release/snappy_framed-* \
  --bench decompress_speed
kcachegrind callgrind.out.*

Note that, by default, Rust and LLVM are extremely aggressive about inlining, which makes the profile hard to read. You may need to add #[inline(never)] to various functions to make them show up separately on the report.

//! [doc]: http://emk.github.io/snappy_framed-rs/snappy_framed/index.html //! [snappy]: http://code.google.com/p/snappy/ //! [framed]: http://code.google.com/p/snappy/source/browse/trunk/framing_format.txt //! [snzip]: https://github.com/kubo/snzip