Add Box::into_inner.

[crlf0710]

This adds a Box::into_inner method to the Box type. I actually suggest deprecating the compiler magic of *b if this gets stablized in the future.

r? @m-ou-se

[KamilaBorowska]

Personally I would rather prefer to have DerefMove to allow other types to be special like Box, but that's just my opinion.

[fmease]

Should the documentation mention deref'ing the box as an alternative? This might prevent confusion.

[m-ou-se]

I actually suggest deprecating the compiler magic of *b if this gets stablized in the future.

That's a rather large change; one that should be discussed with the lang team first. Has there been any discussion on this? How about alternatives like a moving-out-of-deref trait as suggested above?

Without that language change, is there a good use case for this new function?

[crlf0710]

@m-ou-se Yes, there is still a use case (which i think is good).
if i have a type S which is Copy and i have a Box<S> value, and i want to take the S value out.

#[derive(Copy, Clone, Debug)]
struct S;

    let boxed_val = Box::new(S);  
    let val = *boxed_val;
    // this copies the value out (with Deref), but the memory is not deallocated.
    println!("{:?}", boxed_val); //it's still alive now, needs another `drop` to deallocate.
    // do something more.

This seems a footgun because the behavior of that line depends on whether S is Copy. With this function i can easily do what i want. A moving-out-of-deref trait won't be able to fix this.

(Deprecating the compiler magic of *b)
That's a rather large change; one that should be discussed with the lang team first.

Indeed! But i think whether or not that change will happen, this convenience function serves its purpose well. And there's already Pin::into_inner(and many others) in the standard library so this is not an inconsistent addition. And i'm only purposing adding it as an unstable method.

[JohnCSimon]

merge conflict
@JohnCSimon label: +S-waiting-on-author -S-waiting-on-author

[crlf0710]

Rebased and squashed.

[pickfire]

Isn't this the same as *c? Why is the deref bad and we want to remove it?

[m-ou-se]

This was already discussed. See the conversation above.

[m-ou-se]

@bors r+ rollup

[bors]

📌 Commit ce7de07 has been approved by m-ou-se