access to `&self` in `display()`

[colin-kiegel]

compile error: self is not available in a static method. Maybe a self argument is missing? [E0424]

I would like to reuse the error description in the Display::fmt function on some occasions. It would be nice to do this by delegation to self.description(). There might be addional usecases for &self access.

quick_error! {
    #[derive(Debug)]
    pub enum LoaderError {
        ArrayTemplateNotFound(name: String) {
            description("Template not found")
            display("{}: {:?}", self.description(), name)
        }
    }
}

[tailhook]

Well, I'm okay with this proposal. But I'm not sure it's possible. You will need to define name "self" in this code too because of macro hygiene. Like display2(self, "{}: {:?}", self.description(), name).

[colin-kiegel]

You are right - this is macro hygiene. It should work by passing a closure like display_fn(|self_| { format!("{}", self_.description()), or possibly better display_fn(|self_, f| { write!(f, "{}", self_.description()) - since using write! should save the allocation done by format!.

The question then is how to react if someone defines both display and display_fn (or whatever the name will be).

[colin-kiegel]

Ok, you were right again. Passing a closure didn't work. With PR #8 the macro uses closures internally now, but the signature of display_fn is like you suggested. What do you think? Should it be called display_ref instead?

I was also thinking about trying to overload the definition of display("...") and display(X, "..."), but I am not sure if that could lead to problems in some corner cases, where "..." might be behind an identifier X. I didn't test that.

Regarding the question of duplicate display vs. display_fn implementations - the current design will silently only take the first implementation in order of appearance. This is the current behaviour with multiple display implementations.

[tailhook]

Well, I think the syntax should be:

display(self) -> ("...", a, b)

Because there is a similar from(x) -> (..) construction. Otherwise it's fine.

[colin-kiegel]

Hm, I can't get it to work - rust either complains

src/lib.rs:616:25: 616:43 error: expected ident, found "I/O error: {err}"
src/lib.rs:616                 display("I/O error: {err}", err=err)
                                       ^~~~~~~~~~~~~~~~~~

or

src/lib.rs:626:25: 626:30 error: format argument must be a string literal.
src/lib.rs:626                 display(self_) -> ("{} {}: {}", self_.description(), place, err)
                                       ^~~~~

Depending on which matcher comes first:

    (FIND_DISPLAY_IMPL $name:ident $item:ident
        { display($self_:ident) -> ($($exprs:tt)*) $($tail:tt)* }
    ) => {
        |$self_: &$name, f: &mut ::std::fmt::Formatter| { write!(f, $($exprs)*) }
    };
    (FIND_DISPLAY_IMPL $name:ident $item:ident
        { display($($exprs:tt)*) $($tail:tt)* }
    ) => {
        |_, f: &mut ::std::fmt::Formatter| { write!(f, $($exprs)*) }
    };

[tailhook]

@colin-kiegel, I think I've figured out a way to do that in #10. Could you take a look if I'm missing something? I will probably squash the commits then.