Parse default field values as Type::Verbatim

[DJMcNab]

See #1774 (noting that a true fix will require 3.x and #1851)

• RFC: [RFC] Default field values rust-lang/rfcs#3681
• Tracking issue: Tracking issue for RFC 3681: Default field values rust-lang/rust#132162
• Feature gate: #![feature(default_field_values)]

#[derive(Default)]
struct Pet {
    name: Option<String>,
    age: i128 = 42,
    //        ^^^^
}

Adapted from #1851. cc @estebank

This PR is an alternative to #1851 but avoiding the breaking-change. Instead, if a named field has a default value, the field's entire type and the default value is parsed as a Type::Verbatim. Our users can then detect and support this case, by re-parsing the verbatim type into the type and default value expression.

Semantically of course, the default value is not part of the type, but syntactically, the default value is in type position. Our handling of negative inherent impls has a similar shape.

The interaction with unnamed_fields was a bit unclear to me, so I took my cue from #1851 (of allowing both features to be combined). Note that since rust-lang/rust#131045, unnamed_fields is not supported by upstream Rust.

I chose not to implement the solution you proposed in #1851 (comment), both because the idiomatic implementation strategy was unclear to me, and because if I understand correctly, the proposed solution makes supporting default_field_values in dependent macros unfeasible.

Thanks for all your great work!

[Veetaha]

This seems like a bearable hack. Although I do forsee that existing proc macros will still break if they encounter the default field value syntax, this makes it easier to hackily get the default expression from the type for macros that want to start supporting this syntax (and I would be interested in that too with bon).

An alternative way to approach this could be to add a new syn::Item variant with versioning applied e.g. syn::Item::StructV2(StructV2) where StructV2 is the same as the current Struct but with the additional default field. Although, this would also necessitate the FieldV2 struct and bloating the crate with lots of similar/but-slightly-different types. This idea seems crazy, but IDK, just wanted to state it here, and maybe someone could be inspired with it

[jplatte]

Gentle ping, @dtolnay