Add Option::unwrap_unchecked

[joshlf]

Now that NonNull has been added and Option<NonNull<T>> is preferred over *mut T, if you have code that used to know that a *mut T was non-null and dereferenced it, you now know that your Option<NonNull<T>> is Some and can call .unwrap() on it. However, .unwrap() can incur runtime overhead that dereferencing *mut T never would. Since a lot of performance-critical code is written using raw pointers (and now, NonNull), this runtime overhead may be worth worrying about.

I propose, for these cases, to introduce an unsafe unwrap_unchecked method to Option that returns a T by simply assuming that the Option is currently Some(T).

[kennytm]

There is already an external package providing this function. https://crates.io/crates/unsafe_unwrap

[joshlf]

Ah good to know. I still think it could be useful in the stdlib, but it's definitely less of a concern with that crate.

[aleksmelnikov]

My compare of default unwrap() and unsafe_unwrap():

$ git clone https://github.com/nvzqz/unsafe-unwrap-rs.git
$ cd unsafe-unwrap-rs
$ cargo build
$ cargo bench

The bench output:

test benches::bench_normal_unwrap_1000 ... bench:       2,112 ns/iter (+/- 709)
test benches::bench_unsafe_unwrap_1000 ... bench:         814 ns/iter (+/- 89)

Code compare:

// default unwrap()
impl<T> Option<T> {
    pub fn unwrap(self) -> T {
        match self {
            Some(val) => val,
            None => panic!("called `Option::unwrap()` on a `None` value"),
        }
  
    }
}

// unsafe_unwrap()
impl<T> UnsafeUnwrap<T> for Option<T> {
    #[inline]
    unsafe fn unsafe_unwrap(self) -> T {
     if let Some(x) = self { x } else { unreachable() }
    }
}

I tried to change code of unsafe_unwrap():

impl<T> UnsafeUnwrap<T> for Option<T> {
    #[inline]
    unsafe fn unsafe_unwrap(self) -> T {
     //if let Some(x) = self { x } else { unreachable() }
     if let Some(x) = self { x } else { panic!("PANIC") }
    }
}

New bench output:

test benches::bench_normal_unwrap_1000 ... bench:       2,111 ns/iter (+/- 56)
test benches::bench_unsafe_unwrap_1000 ... bench:       2,110 ns/iter (+/- 13)

So, panic! macro slows down default unwrap() in ~2.5 times!

[ojeda]

Available in nightly now. For Result too.