Skip to content

Improve path to Clone::clone call when deriving Clone #3371

Open
@CohenArthur

Description

In the meantime we could generate something like <to_clone_type as #[lang = "clone"]>::clone(to_clone) but this is extra work. It's not super difficult, so I'll mark it as a good-first-pr

Originally posted by @CohenArthur in #3343 (comment)

When derivine Clone, we have to perform "clone calls" to sub items. e.g. to clone a struct Foo { a: T, b: U } we need to create the following Clone implementation:

impl<T: #[lang = "clone"], U: #[lang = "clone"]> #[lang = "clone"] for Foo {
    fn clone(&self) -> Self {
        Foo { a: Clone::clone(self.a), b: Clone::clone(self.b)
    }
}

The issue is that just calling Clone::clone is not valid, especially if the module structure is complex. So we can run into name collision issues with a user-defined Clone trait, which is not good. Similarly, we don't want to call a.clone() as that may collide with a user defined trait.

We can fix this by using the field's type and casting it as the known, official Clone trait and calling that trait's method. To do this, we need to generate the following code:

impl<T: #[lang = "clone"], U: #[lang = "clone"]> #[lang = "clone"] for Foo {
    fn clone(&self) -> Self {
        Foo { a: <T as #[lang = "clone"]>::clone(self.a), b: <U as #[lang = "clone"]>::clone(self.b)
    }
}

Note that I think we could also do something like #[lang = "clone"]::clone(self.a), but I think this requires yet another rework of lang items in the AST... and I'm not sure it's worth it

Metadata

Assignees

No one assigned

    Type

    No type

    Projects

    No projects

    Relationships

    None yet

    Development

    No branches or pull requests

    Issue actions