WebAssembly support

[ibaryshnikov]

Is it a goal or non-goal?
It worked out of the box before 0.1, but not supported by inventory yet.

[Enet4]

Thank you for reporting. WebAssembly support is indeed planned, but it did not occur to me that the flexible transfer syntax registry would break compatibility.

If inventory really can't work in a wasm environment, we'll have to look for an alternate implementation.

[ibaryshnikov]

Thanks for a quick reply!

There's a related issue amethyst/amethyst#1893, but linkme support for wasm is not ready yet dtolnay/linkme#6

From the other side, it's possible to have mutable statics

lazy_static! {
    static ref REGISTRY: Mutex<TransferSyntaxRegistry> = {
        Mutex::new(TransferSyntaxRegistry { m: initialize_codecs() })
    };
}

the drawback is that it can affect performance.

One more option is to use decoder builder, keeping a local copy of transfer syntax registry per decoder

let decoder = DecoderBuilder::default()
    .with_transfer_syntax(extend it)
    .with_transfer_syntax(extend again)
    .build()?;
let object = decoder.from_reader(input)?;

What do you think? May be there is a better alternative which I missed.

[Enet4]

That would be a similar approach to the data dictionary, which is a type parameter in many places of the implementation. So that may indeed be feasible for a scenario where we can't have an inventory of transfer syntaxes, while still keeping an intuitive default of using said inventory.

I've been busy at a conference this week, but I may be able to continue pushing some code and roadmap text soon.

[Enet4]

Considering that #17 has been merged, we can consider this resolved. We should strive to keep WebAssembly as a working target, although there is much to be done regarding testing. Please create a new issue if another problem emerges when using DICOM-rs in WebAssembly.