Handling errors is flattened structs validation

[Techassi]

I would like to ask, if there is any recommended way of handling the following situation?

Let's assume we have a struct, which defines common attributes which are always present / supported independent of the type of item (field vs variant):

#[derive(Debug, FromMeta)]
#[darling(and_then = ItemAttributes::validate)]
pub struct ItemAttributes {
    pub my_common_attr: String,
}

impl ItemAttributes {
    fn validate(self) -> Result<Self, Error> {
		return Err(Error::custom("validation failed").with_span(ident));
        // What to put here, we have no access to ident yet     ^^^^^
    }
}

Next, we have the field and variant specific structs, which flatten the common struct into it:

#[derive(Debug, FromField)]
#[darling(
    attributes(versioned),
    forward_attrs(allow, doc, cfg, serde),
    and_then = FieldAttributes::validate
)]
pub(crate) struct FieldAttributes {
    #[darling(flatten)]
    pub(crate) common: ItemAttributes,

    // The ident (automatically extracted by darling) cannot be moved into the
    // shared item attributes because for struct fields, the type is
    // `Option<Ident>`, while for enum variants, the type is `Ident`.
    pub(crate) ident: Option<Ident>,
}

impl FieldAttributes {
    fn validate(self) -> Result<Self, Error> {
        return Err(Error::custom("validation failed").with_span(&self.ident));
        // This is totally fine and results in great DX         ^^^^^^^^^^^
    }
}

Now to my question: Is there a way to get access to the ident of the field / variant to ultimately get access to its span?. This could allow handling common validation with the same amount of DX as handling item specific validation.

[TedDriggs]

Magic fields like ident don’t work with flatten, so even if the types were identical you wouldn’t be able to do what’s depicted here.

In my opinion, a better solution would be to change ItemAttributes::my_common_attr to be a SpannedValue<String> or a syn::LitStr so you have the span of that specific field, and put the error there rather than being on the variant or field ident.

[Techassi]

Okay, thanks for the input!

I would've hoped there is still a way to access the field or variant ident, because that provides the best DX in the situation this is used in my code. For now, I will follow your recommendation, but I would love to see this being possible in the future.