Rust中如何传递非静态可变数据给线程并确保其在join前不销毁？

问题背景 #

我需要将一个Vec拆分为多个块分配给不同线程处理，在所有线程调用join()确认不再访问该Vec后返回它。但线程要求数据必须具有'static生命周期以防止访问时被销毁，请问有什么解决办法？

我尝试过以下两种方式，但都存在弊端：

• 使用Arc和Mutex：会限制原数据的可变性
• 让每个线程返回Vec再合并：对于3840*2160尺寸的Vec来说，效率极低

报错代码 #

{
    let mut data: Vec<u8> = vec![0; (render_config.resolution.0 * render_config.resolution.1 * 3) as usize];
    
    let chunks = data.chunks_mut(((render_config.resolution.0 * render_config.resolution.1 * 3) / 12) as usize);

    let mut ts = Vec::with_capacity(chunks.len());

    let mut offset = 0;
    for chunk in chunks {
        ts.push(thread::spawn(|| { Self::get_image_sub_threaded(self.clone(), render_config.clone(), is_screenshot, chunk, offset); }));
        offset += chunk.len();
    }

    for t in ts {
        t.join();
    }

    data
}

错误信息 #

error[E0597]: `data` does not live long enough
  --> src\rendering\camera.rs:26:22
   |
26 |         let chunks = data.chunks_mut(((render_config.resolution.0 * render_config.resolution.1 * 3) / 12) as usize);
   |                      ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ borrowed value does not live long enough
...
32 |             ts.push(thread::spawn(|| { Self::get_image_sub_threaded(self.clone(), render_config.clone(), is_screenshot, chunk, offset); }));
   |                     ---------------------------------------------------------------------------------------------------------------------- argument requires that `data` is borrowed for `'static`
...
41 |     }
   |     - `data` dropped here while still borrowed

解决方案：使用thread::scope #

thread::scope可以创建一个线程作用域，作用域内的线程可以安全引用作用域内的变量，无需满足'static生命周期约束，同时保证所有线程在作用域结束前完成执行，不会出现悬垂引用。以下是实现代码：

pub fn get_image_threaded(&self, render_config: &RenderConfig) -> Vec<u8> {
    let mut data: Vec<u8> = vec![0; (render_config.screenshot_resolution.0 * render_config.screenshot_resolution.1 * 3) as usize];
    
    let chunks = data.chunks_mut(((render_config.screenshot_resolution.0 * render_config.screenshot_resolution.1 * 3) as usize) / num_cpus::get());
    println!("CPUs available: {}", num_cpus::get());
    
    thread::scope(|scope| {
        let mut ts = Vec::with_capacity(chunks.len());
        let mut offset = 0;
        let mut i: u32 = 0;
        for chunk in chunks {
            let o = offset;
            let len = chunk.len() / 3;
            ts.push(scope.spawn(move || { Self::get_image_sub_threaded(self.clone(), render_config.clone(), true, chunk, o, i); }));
            offset += len;
            i += 1;
        }
        for t in ts {
            t.join().unwrap();
        }
    });

    data
}

内容的提问来源于stack exchange，提问作者Robert Lucas