Add support for generic functions to #[debug_handler]

[jplatte]

I've had an idea for a while that would make generic handlers pretty much fully checkable, that is:

#[debug_handler(with(T = String, T = u64; U = i8, U = i16; N = 0))]
fn handler<T, U, const N: usize>(extract_t: Path<T>, extract_u: Json<U>, extract_n: Foo<N>)
where
    T: Deserialize,
    U: Deserialize,
{
    // ...
}

could make debug_handler add FromRequest assertions for Path<String>, Path<u64>, Json<i8>, Json<i16> and Foo<0> while also adding Send assertions for the futures returned by

• handler::<String, i8, 0>
• handler::<String, i16, 0>
• handler::<u64, i8, 0>
• handler::<u64, i16, 0>

If we were to do that, we could just require users to add this with() annotation before any parameters are checked on a generic function (even ones that don't mention the generic types) as it should be easy enough.

Originally posted in #1251 (comment)

[jplatte]

For anybody who wants to work on this, here's some tips:

• If you aren't familiar with the debug_handler macro yet, go read its documentation to get an understanding of what it does
• We also have tests for it here, the most interesting ones are probably the ones in fail, where every .rs test case has a corresponding .stderr file showing the error generated by the macro
• You can find the macro implementation here
• If you have never worked with procedural macro code before and find the code hard to read, maybe my blog series on proc-macros over at blog.turbo.fish can help :)
• The important parts for this this feature are the functions check_inputs_impl_from_request for the FromRequest implementation assertions and check_future_send for the Send assertions
• I'm pretty sure there's nothing to be done for lifetime generics, those should just be an error
• It's probably worthwhile to start with type generic support and add const generic support later, if at all
• Use syn::visit_mut to substitute the concrete types for uses of a generic parameter, this ensures you find them all even in very complex cases like <<T as Trait>::Foo as some_crate::OtherTrait<U>>::Bar

[slessans]

@jplatte I am interested in taking a crack at this. thanks for the extensive write-up!

[slessans]

@jplatte here is a proof-of-concept that implements this for generic handler args for early feedback if you have some time.

#1265

will move onto return types next

[RoKu1]

Is it still open for taking. I am new here, was looking for some simple issue to work on.

[jplatte]

No, this is mostly done in #1265.

[RoKu1]

Hi, Is there any simple task I can take up.

[0xdeepmehta]

any updates here?

[davidpdrsn]

@0xdeepmehta No. I still have to review the PR.

[jplatte]

As per latest comments in the PR, I'll close this since it's probably not worth the maintenance burden to support.