should the `<-` syntax be removed?

[BennoLossin]

Instead : always asks for an initializer and since T implements Init<T>, this should be fine.

[BennoLossin]

The benefit of no <- syntax is that the code can be rustfmt'ed, which is something we very much want to have. However missing the <- could lead to inference issues and it's less clear when a value is in-place initialized vs moved.

[antonio-hickey]

I will say as a new contributor and someone who is not as well versed in the project, seeing this (<-) did confuse me at first requiring some digging around to what was going on.

However it does seem like a good idea to make in-place init's explicit.

What about an attribute marker or a wrapper marker?
It becomes more verbose, but we can get rustfmt'ed and still have explicit in-place init's.

// Attribute Marker
init!(Self {
    a: Box::new(42),
    #[in_place]
    bar: init!(Bar { b: 42 }),
}? AllocError)

// Wrapper Marker
init!(Self {
    a: Box::new(42),
    bar: in_place(init!(Bar { b: 42 })),
}? AllocError)

[BennoLossin]

The attribute suggestion is great, it was also noted by @nbdd0121 on zulip somewhere (haven't dug up the discussion). I just forgot to jot it down here as well.

this is also somewhat related to the rust-lang effort of in-place init, if we choose to not have <- there, then I'm more inclined to remove it here as well.

Another alternative would be to have a wrapper script for rustfmt that replaced field <- value, with #[init] field: value, then ran rustfmt, and then reverted that change again, but that might be too fragile.