Cannot move out of dereference for struct with vector.

[JP-Ellis]

I have encountered an issue which seems very unusual to me and I believe it might be a bug.

Below is a small example which illustrates the issue.

struct CountDown {
    count: uint,
    // Including the vector causes `Some(self.current())` to fail later on.
    vector: Vec<uint>
}

impl CountDown {
    fn new() -> CountDown {
        CountDown {
            count: 10,
            vector: vec!(3u, 4, 5)
        }
    }

    fn current(self) -> uint {
        self.count
    }
}

impl Iterator<uint> for CountDown {
    fn next(&mut self) -> Option<uint> {
        self.count -= 1;
        if self.count == 0 {
            None
        } else {
            // Whether the vector is in the struct or not, this work:
            //Some(self.count)

            // When the struct contains a vector, the following causes the error:
            //   cannot move out of dereference of `&mut`-pointer
            // If the vector is not in the struct, this work fine.
            Some(self.current())
        }
    }
}

fn main() {
    for n in CountDown::new() {
        println!("n = {}", n);
    }
}

It appears to me quite strange that the presence of a Vector within the struct (even if it isn't used) causes the issue. It also seems strange to me that returning a struct field works, but returning a method which itself returns that struct field fails.

Admittedly, I don't really understand what cannot move out of dereference of '&mut'-pointer really means; a Google search doesn't yield very much, and the documentation doesn't seem to contain much regarding the error.

[huonw]

This is a little subtle and comes down to moves vs. copies, due to current taking self by value.

    fn current(self) -> uint {

If you change that to fn current(&self) -> uint, i.e. take self by reference it will work.

The reason it works when there is no Vec is CountDown becomes Copy, i.e. a by-value use doesn't have to consume the source, essentially it just duplicates the CountDown instance out of the &mut self in next giving current its own independent copy. As soon as you have a Vec this is illegal (see my linked stack overflow answer above), and so the compiler complains.

Thanks for filing, but this isn't a bug, closing. :)

[JP-Ellis]

Thank you for clarifying this for me. Sorry for bringing up a non-issue, though hopefully others with the same issue might stumble upon this and find their answer.