Refactor Request/Resonse's TryFrom impl by generating match stmt

[DevinR528]

No description provided.

[DevinR528]

I wonder if this is worth doing until try is finally (:smile:) part of the language? I agree it cleans up repetitive parts of the code, I worry that it's at the expense of readability of the macro. This is by far the most complicated part of ruma that consumers sometimes use. These explicit matches are a good clue to what's going on in the generated code.

Just a thought, I'll still work on resolving the issues.

[jplatte]

So you think the existing code was actually more readable? I have to say I'm not entirely convinced about the changes so far either.. :/

[DevinR528]

Yeah, I guess after tinkering with it I would say it was more readable before. My preference (for readability) would be putting as big a code block as possible inside the quote! call, the less broken up the to-be-generated code the better.

[jplatte]

Okay. Do you think the macro_rules! solution would overall improve readability, though?

[DevinR528]

Removing the inner quote! would help

quote! {
    let request_body: RequestBody =
    ruma_api::try_deserialize!(
        ruma_api::exports::serde_json::from_slice(request.body().as_slice()),
        ::ruma_api::error::RequestDeserializationError,
    );
}

Or try_request_deserialize! and try_response_deserialize! macros instead of passing the error?

[jplatte]

I think try_deserialize!(request, ...) and try_deserialize(response, ...) would be the shortest solution that would actually work (I think request or response has to be part of the macro input so it can be referred to properly – if you don't get it working ping me on Matrix, this could be tricky).

[jplatte]

Nice, you made it work! :)

I think the way I've seen this written usually is like this: @request, so without the extra space. Where did you get the idea with the @?

[DevinR528]

I remember reading somewhere that the @ was meant to signal internal details of the macro, if you meant with spaces I probably added that as I can't remember exactly where I saw that.

I was surprised that doubling the token $kind, $kind actually worked. Did you have a specific technique in mind?

[jplatte]

I remember reading somewhere that the @ was meant to signal internal details of the macro

Okay :)
Yeah that's what I use too and would basically be my reasoning as well. So no, I didn't have anything else in mind, just the same thing without the extra space, like you've implemented now.

I was surprised that doubling the token $kind, $kind actually worked.

Why not? ^^

[jplatte]

CI failed because of formatting, please fix.