reexport path while using derive

[kuqmua]

imagine you creating new crate what depends on serde and reexport it

pub use serde::*;

In your crate you creating a new proc_macro. In proc macro you creating something like

let pub_struct_something = quote::quote!{
    #[derive(your_new_crate::Serialize)]
    pub struct Something {
        my_field: bool
    }
};

quote::quote!{
    #pub_struct_something
    #some_other_logic
}.into()

The problem is - generated code is still trying to import modules from serde, not from your_new_crate.
I suggest wrap "expand_derive_serialize"(or something) in

pub fn derive_serialize(input: TokenStream) -> TokenStream {
    let mut input = parse_macro_input!(input as DeriveInput);
    ser::expand_derive_serialize(&mut input)
        .unwrap_or_else(syn::Error::into_compile_error)
        .into()
}

into new public function, something like

"expand_derive_serialize_WITH_PATH_TO_CRATE"

with additional(maybe few) parameters what change all serde occcurences to parameter value.
Of course defalut would be just serde.

For example new implementation would be

pub fn derive_serialize(input: TokenStream) -> TokenStream {
    let mut input = parse_macro_input!(input as DeriveInput);
    ser::expand_derive_serialize_WITH_PATH_TO_CRATE(&mut input, "serde")
        .unwrap_or_else(syn::Error::into_compile_error)
        .into()
}

and you can write something like

let generated_serialize_token_stream_for_something_struct = serde::expand_derive_serialize_WITH_PATH_TO_CRATE(&mut possible_something_struct_token_stream, "your_crate");

Well i can write implemenation mannualy in my macro but in time i think it will be outdated i just wont know about some vulnerability or new optimization.

[dtolnay]

I would prefer not to work around this in the way you are asking.

rust-lang/rust#54363 (comment) has a writeup of my preferred approach.