Lifetime syntax suggestion

1

Lifetimes in Rust are difficult to learn how to use them properly, and I guess many people lose interest in the language because of that. I wonder if it has ever been considered adopting a syntax that allows representing this:

struct Foo<'a, 'b, 'c> {
    field1: &'a String,
    field2: &'b String,
    field3: &'c String,
}

fn bar<'r>(s: Foo<'_, 'r, '_>, t: Foo<'_, '_, 'r>) -> &'r String

Like this:

struct Foo {
    field1: &String,
    field2: &String,
    field3: &String,
}

fn bar(s: Foo, t: Foo) -> &String
where 'return: 's.field2 | 't.field3

This is not a full proposal, I'd just like to know if something similar has already been discussed. If not, do you guys think this might be worth trying?

2

It probably has come up before, maybe even several times. I don't think such big changes for syntax are or will ever realistically be in scope.

3

fields are private by default.
I think it is understanding variance that priority but not sure of a syntax that would help.

4

One place this came up in the past is the alternatives section of RFC 2115 and associated discussion.

1 Like
5

Thank you for the links! :+1:

6

Whilst we are here, could some kind soul explain what Milack27's example actually means?

Every time I start to look into those life time ticks I get severe headache and my brain freezes over.

I'm sure there is a simple concept in there waiting to come out for me but this kind of thing:

fn bar<'r>(s: Foo<'_, 'r, '_>, t: Foo<'_, '_, 'r>

remains nothing but line noise at the moment.

8

Exactly my point. In the bar function, I only need to reference one lifetime for each parameter. But because Foo itself has three generic lifetime parameters, I'm forced to reference all of them in both s and t. The ones with the '_ placeholder are irrelevant to this context, but they have to be there anyway.

9

This topic was automatically closed 90 days after the last reply. New replies are no longer allowed.