Trait IdentFrag

Source

pub trait IdentFrag: Spanned {
    type BadIdent: Clone;

    // Required methods
    fn frag_to_tokens(
        &self,
        out: &mut TokenStream,
    ) -> Result<(), Self::BadIdent>;
    fn fragment(&self) -> String;

    // Provided method
    fn note_atoms(&self, out: &mut Vec<AtomForReport>) { ... }
}

Expand description

For use with ExpansionOutput.append_identfrag_toks etc.

Sort of like quote::IdentFragment. But:

Strings available directly, not via the inconvenient fmt,
identifier construction doesn’t involve format_ident! and panics.
paste::InputAtom passed through separately,

The main purpose of this trait is to allow deferral of errors from constructing bad identifiers. Some inputs (implementors of IdentFrag, typically those which are already tokens) can infallibly be appended as tokens.

But paste results aren’t converted to identifiers until the last moment: they can’tbed converted infallibly, and the error surfaces in convert_to_ident.

The framework methods append_*ident* propagate any error to the call site. Call sites which pass already-tokens can just use ? to convert the uninhabited error to syn::Error - or they can use IdentFragInfallible::unreachable.

Call sites which pass actually-fallible content (ie, paste results) end up with a syn::Error.