Signature of return value of generic function?

[eitsupi]

Wouldn't it be potentially useful in documentation and testing if we could set a class for the return value of a generic function?

For example, Rust defines a trait with a method as like:

pub trait Summary {
    fn summarize(&self) -> String;
}

In the absence of such a thing, I don't think we can rule out the possibility of completely different return values for each method.

For example, in the past, there has been some discussion about what the dplyr::pull() methods in the arrow package should return (apache/arrow#32705).
I imagine that if the return class of the dplyr::pull() generic function were defined, this confusion could be avoided.

[hadley]

How would such a constraint be enforced?

[eitsupi]

How would such a constraint be enforced?

I think it is the same issue as #450