Okapi emits warnings when using a closure with no arguments

[ThouCheese]

When writing a closure with no arguments, Okapi emits a warning on compilation.

#[openapi]
#[get("/?<id>")]
fn get_by_id(id: Option<i32>) -> String {
    let default = || 1;
    let id = id.unwrap_or_else(default);
    format!("got {}", id)
}

This gives the following result:

warning: unnecessary parentheses around type
  --> src/main.rs:16:1
   |
16 | #[get("/?<id>")]
   | ^^^^^^^^^^^^^^^^ help: remove these parentheses
   |
   = note: `#[warn(unused_parens)]` on by default

warning: 1 warning emitted

I've looked at the emitted code using cargo expand, but i couldn't figure out the cause of this warning.

[ralpha]

I was already in the code so took a look.
Apparently it is complaints about:

fn get_by_id(id: Option<(i32)>) -> String { ... }
                        ^^^^^ These

They are added by

okapi/rocket-okapi-codegen/src/lib.rs

Line 47 in 9807408

fn preserve_span_information(input: TokenStream) -> TokenStream {

The function:

fn preserve_span_information(input: TokenStream) -> TokenStream {
    // Outputting the input unmodified would cause its span information to be
    // lost when being consumed by other macros. But parsing it in then quoting
    // it back out causes span information to be preserved.
    // See https://github.com/GREsau/okapi/issues/12
    // and https://github.com/rust-lang/rust/issues/43081
    let parsed_input: syn::Item = syn::parse(input).unwrap();

    // Nested generics cause span bugs - we can work around this by wrapping all
    // generic type parameters in parentheses.
    // https://github.com/rust-lang/rust/pull/48258
    let parsed_input = GenericTypeVisitor.fold_item(parsed_input);

    quote!(#parsed_input).into()
}

Looking at: #12 there are similar issues there (that you reported 😛 ).

I'm not familiar with that issue so will leave this to someone else (@GREsau 😉 ).

When commenting out this line I don't have the error anymore 😄

okapi/rocket-okapi-codegen/src/lib.rs

Line 31 in 9807408

input = preserve_span_information(input);

maybe adding a strategic #[allow(unused_parens)] solves this. 🤫

(btw thanks for mentioning cargo expand did not know about it, could be useful in the future)

[GREsau]

Thanks for looking into this - yep, I think #[allow(unused_parens)] is going to be the simplest fix for this

[GREsau]

Actually now that I've taken a closer look, I think the best fix is to just remove the fixes that I added for #12 (which I'm very happy about, since that was a horrible hack!)

When I remove preserve_span_information, it not only fixes this issue, but I also can't reproduce the original issues reported in #12 - I can only assume that they've since been fixed in rustc somehow?

[GREsau]

This should be now fixed in v0.5.0 🙂