fixed length arrays have their size encoded

[oohal]

When a encoding a structure that contains a fixed length array the length of the array is also encoded. For example, this code:

extern crate rustc_serialize;
extern crate bincode;

use bincode::SizeLimit;
use bincode::rustc_serialize::encode;

#[derive(RustcEncodable)]
struct TestStruct {
    a : u16,
    b : [u8; 6],
}

fn main() {
    let test = TestStruct {a: 0xffff, b: [1, 2, 3, 4, 5, 6]};
    let encoded: Vec<u8> = encode(&test, SizeLimit::Infinite).unwrap();

    println!("header: {:?}", encoded);
}

Produces this output:

header: [255, 255, 0, 0, 0, 0, 0, 0, 0, 6, 1, 2, 3, 4, 5, 6]

This behaviour makes it hard to use bincode for unpacking externally defined binary formats (e.g ELF headers). I suppose this library isn't really designed for that, but it works well enough for BE formats so it would be nice to see this fixed in the next major version.

[TyOverby]

I don't think that there is a way to do this with rustc-serialize. At least, there isn't a trait method for statically sized arrays.

[Matvey-Kuk]

+1 for this idea

[TyOverby]

This feature is only possible if we drop RustcSerialize support. I'm planning on doing that eventually, but it's on the backlist until serde hits 1.0

[Matvey-Kuk]

@TyOverby thank you! Just wanted to mention there is somebody else who need this =)

[aldanor]

Very much in support of this as well. Now that serde 0.9 is out, is it a good time to revisit this? (also now that custom derive will get stabilised in a few weeks)

[darfink]

I'd like to chime in on this as well. Working with external data and this issue hit me, ended up having to use a fork.

[mikedilger]

I also ran into this. What if we change the code to skip these lengths if the rustc_serialize feature is disabled?

(I noticed https://github.com/serde-rs/bench says it is like bincode w/o these lengths)

[TyOverby]

Since there's so much interest in this, I think I'll begin work on removing rustc-serialize support from bincode altogether. I've wanted to do this for a long time, but since serde is hitting 1.0 and macros are landing, I think it'll be pretty reasonable to move on to.

[TyOverby]

rustc-serialize removal happening here: #95

[aldanor]

That's great, thanks! Given that this will be a (very-) breaking change release, does it make sense to break things a little further and add endianness support?