Ownership

Rust is notorious for its borrow checker, the mechanism that enforces memory safety at compile time. The core rule is that every value has a single owner, and once that ownership is transferred, the original variable can no longer be used. This is called a move. A move occurs whenever a variable is passed to a function or assigned to another variable. After the move, the original binding is invalid. Consider this example from The Book:

src/main.rs

fn main() {
    let s1 = String::from("hello");
    let s2 = s1;
    println!("{s1}, world!");
}

Running this code will cause the following error:

error[E0382]: borrow of moved value: `s1`
 --> src/main.rs:4:16
  |
2 |     let s1 = String::from("hello");
  |         -- move occurs because `s1` has type `String`, which does not implement the `Copy` trait
3 |     let s2 = s1;
  |              -- value moved here
4 |     println!("{s1}, world!");
  |                ^^ value borrowed here after move
  |
  = note: this error originates in the macro `$crate::format_args_nl` which comes from the expansion of the macro `println` (in Nightly builds, run with -Z macro-backtrace for more info)
help: consider cloning the value if the performance cost is acceptable
  |
3 |     let s2 = s1.clone();
  |                ++++++++


For more information about this error, try `rustc --explain E0382`.

There is a lot of output here, but let us focus on the source of the error for now. First, we see that let s2 = s1; moves the value in s1 to s2, and then we are told that the println! macro mistakenly seeks to borrow the value that has been moved. Since that value is no longer owned by s1, it causes an error.

Cloning

In the example above, we are also given helpful advice about how to fix the error using .clone(), which copies the value rather than moving it. Cloning is probably the simplest way to reuse the original value, as almost everything in Rust can be cloned.

src/main.rs

fn main() {
    let s1 = String::from("hello");
    let s2 = s1.clone();
    println!("{s1}, world!");
}

hello, world!

Functions

The above example also highlights that the println macro seeks to borrow the value in s1. In fact, any time you pass a variable to a function, its value is moved, meaning the originally defined variable can no longer be used to access the value outside of the function. For example, let’s define a mean() function:

fn mean(x: Vec<f64>) -> f64 {
    let mut total = 0.0;
    let n = x.len();
    
    for xi in x {
        total += xi;
    }
    
    total / (n as f64)
}

Calling this function twice on the same vector will fail because the first call moves x into the function:

let x = vec![0.0, 3.14, 10.1, 44.8];

let avg1 = mean(x); // ⬅️ x moved here!
let avg2 = mean(x); // ❌ compiler error

As we just saw, one way of handling this is to clone the data before passing it to the function.

let x  = vec![0.0, 3.14, 10.1, 44.8];

let avg1 = mean(x.clone()); // ⬅️ x cloned here!
let avg2 = mean(x);         // ✅ compiler happy

While cloning in this way is safe and correct, it has the unfortunate downside of allocating a new copy of the data each time — not an issue for small vectors like x, but as they grow, finding a better approach will become more urgent.

Borrowing

Fortunately, there is a more efficient approach known as borrowing, which involves passing a reference to a value rather than moving it. In Rust, we signal a reference by placing an ampersand & in front of a variable or type name, for example, &x or &Vec<f64>. These tell the compiler that we are borrowing a value rather than taking ownership. To use the mean()function with a reference, notice that we have to update its signature to make the reference type explicit:

fn mean(x: &Vec<f64>) -> f64 {
    let mut total = 0.0;
    let n = x.len();
    
    for xi in x {
        total += *xi;
    }
    
    total / (n as f64)
}

fn main() {
    let x = vec![1.0, 2.0, 3.0];
    let avg = mean(&x); // 👈 borrowing `x`
    println!("x is still usable: {:?}", x);
}

x is still usable: [1.0, 2.0, 3.0]

An important restriction on referencing is that borrowed values cannot be moved or mutated — only read. You can borrow a value as many times as you like, you just cannot change it.

Slices

Rather than reference the entire array or vector, we can reference a slice, or a contiguous sequence of elements within it. These are denoted in Rust using &[T] where T stands for a generic type. For example, &[f64] means a reference slice of 64-bit floating points. If we want to demonstrate the use of reference slices with our mean() function, we will as before have to update its signature to make that type explicit:

src/main.rs

fn mean(x: &[f64]) -> f64 {
    let mut total = 0.0;
    let n = x.len();
    
    for xi in x {
        total += *xi;
    }
    
    total / (n as f64)
}

fn main() {
    let x = [0.0, 20.0, 742.3];  // array
    let y = vec![1.0, 2.0, 3.0]; // vector

    println!("the mean of x is: {}", mean(&x));
    println!("the mean of y is: {}", mean(&y));
}

the mean of x is: 254.1
the mean of y is: 2

There are two things to note about this example. The first is very subtle. We updated the signature for mean to take x: &[f64], a slice. However, we passed &x and &y to the function, which are references to arrays and vectors. What is going on here? The answer is an implicit conversion. Both arrays and vectors have methods for converting their generic references to slices, and those are invoked when passing them into a function with this signature.

Following from that, the second thing to note is that a function written to accept &Vec<f64> works only on vectors. That is because there is no method for converting array references to vector references. But as we just saw, both can be converted to slices. Rewriting a function to accept &[f64], thus, makes it more general at no cost. So, in practice, we should prefer slices over vector references whenever the function only needs to read the data — it is more flexible and equally efficient.