> I really hope that the Rust compiler has some really good error messages explaining what the problem is because I couldn't see what was wrong with this fragment until I read the explanation.
They're not always perfect, but in this case it's quite good.
err.rs:4:9: 4:10 error: cannot borrow `v` as mutable because it is also borrowed as immutable
err.rs:4 v.push(6); // oops, sorry x!
^
err.rs:3:17: 3:18 note: previous borrow of `v` occurs here; the immutable borrow prevents subsequent moves or mutable borrows of `v` until the borrow ends
err.rs:3 let x = v.get(2);
^
err.rs:5:6: 5:6 note: previous borrow ends here
err.rs:1 fn main() {
err.rs:2 let mut v = vec![1i32, 2, 3, 4, 5];
err.rs:3 let x = v.get(2);
err.rs:4 v.push(6); // oops, sorry x!
err.rs:5 }
^
If you want to try out code examples, the Rust playpen is nice: http://play.rust-lang.orgI'm quite surprised by Rust behaviour here: if x isn't used after 'v.push(6);' (doesn't seem too difficult to know by the compiler) then I would have expected the compiler to allow this code and to reject it only if x is used after the 'v.push(6)'..
Sure as he explained you can do
"let mut v = vec![1i32, 2, 3, 4, 5];
{
let x = v.get(2);
<work with x>
}
v.push(6);"
but this must be quite annoying..
Currently, borrows last for block scope, which is why this doesn't work. There is a feature in the making called SEME (single entry, multiple exit) which will resolve this and other annoying borrow hazards. It at least at one point was considered a priority for 1.0.
Saying that, after reading the explanation it became a bit more obvious, and I really like the way of describing the relationship to the pointer as a 'view' and 'loan', which as a c++ programmer (though clearly spending too much time with GC languages) I can understand.