xdr-codec 0.4.4

XDR encode/decode runtime support. Pairs with xdrgen which generates code from specs.
Documentation

Rust XDR library

Build Status Crates.io Coverage Status

This crate provides a set of runtime routines to encode and decode basic XDR types, which can be used with xdrgen's automatically generated code, or with hand-written codecs.

This crate also implements XDR-RPC record marking in the form of the XdrRecordReader and XdrRecordWriter IO filters.

Usage

The easiest way to use this library is with xdrgen, which takes takes a specification in a .x file and generates all the necessary definitions for you.

However, you can manually implement the Pack and Unpack traits for your own types:

struct MyType {
    a: u32,
    b: Vec<u8>,
}

impl Pack<W> for MyType
    where W: Write
{
    fn pack(&self, out: &mut W) -> xdr_codec::Result<usize> {
        let mut sz = 0;

        sz += try!(self.a.pack(out));
        sz += try!(Opaque::borrowed(self.b).pack(out));

        Ok(sz)
    }
}

impl Unpack<R> for MyType
    where R: Read
{
    fn unpack(input: &mut In) -> Result<(Self, usize)> {
        let mut rsz = 0;
        let ret = MyType {
            a: { let (v, sz) = try!(Unpack::unpack(input)); rsz += sz; v },
            b: { let (v, sz) = try!(Opaque::unpack(input)); rsz += sz; v.into_owned() },
        };

        Ok((ret, rsz))
    }
}

or alternatively, put the following in src/mytype.x:

struct MyType {
    unsigned int a;
    opaque b<>;
}

then add a build.rs to your Cargo.toml:

extern crate xdrgen;

fn main() {
    xdrgen::compile("src/mytype.x").expect("xdrgen mytype.x failed");
}

then include the generated code in one of your modules:

extern crate xdr_codec;

// ...

include!(concat!(env!("OUT_DIR"), "/mytype_xdr.rs"));

Documentation

Complete documentation is here.

Changes in 0.4.2

Implement standard traits for char/unsigned char (i8/u8 in Rust).

Also handle short/unsigned short as an extension in .x files. They are still represented in memory as i32/u32.

Changes in 0.4

Version 0.4 added the bytecodec feature, which implements Pack and Unpack for byte types (i8 and u8). This is normally unwanted, since bytes suffer from massive padding on the wire when used individually, or in an array of bytes (opaque is the preferred way to transport compact byte arrays). However, some protocols are mis-specified to use padded byte arrays, so bytecodec is available for them.

Changes in 0.2

Versions starting with 0.2 introduced a number of breaking changes:

  • u8 no longer implements Pack/Unpack

    XDR doesn't directly support encoding individual bytes; if it did, it would require each one to be padded out to 4 bytes. xdr-codec 0.1 implemented Pack and Unpack for u8 primarily to allow direct use of a Vec<u8> as an XDR opaque<>. However, this also allowed direct use of u8::pack() which makes it too easy to accidentally generate a malformed XDR stream without proper padding.

    In 0.2, u8 no longer implements Pack and Unpack. Instead, xdr-codec has a Opaque<'a>(&'a [u8]) wrapper which does. This allows any [u8] slice to be packed and unpacked.

    It also has a set of helper functions for packing and unpacking both flexible and fixed-sized opaques, strings and general arrays. These make it straightforward to manage arrays in a way that is robust. This also allows xdrgen to generate code for fixed-sized arrays that's not completely unrolled unpack calls.

    (I'm not entirely happy with the proliferation of functions however, so I'm thinking about a trait-based approach that is more idiomatic Rust. That may have to be 0.3.)

  • Extensions to XDR record marking

    I added XdrRecordReaderIter which allows iteration over records. Previously all the records in the stream were flattened into a plain byte stream, which defeats the purpose of the records. XdrRecordReader still implements Read so that's still available, but it also implements IntoIterator so you can iterate records.

    The addition of more unit tests (see below) pointed out some poorly thought out corner cases, so now record generation and use of the EOR marker is more consistent.

  • More unit tests, including quickcheck generated ones

    I've increased the number of tests, and added quickcheck generated tests which cleared up a few corner cases.

License

Licensed under either of

at your option.

Contribution

Unless you explicitly state otherwise, any contribution intentionally submitted for inclusion in the work by you, as defined in the Apache-2.0 license, shall be dual licensed as above, without any additional terms or conditions.