Question

我想用hyper下载大文件（500mb），并且如果下载失败则能够恢复。

有没有办法使用hyper来为每个收到的数据块运行某些功能？ send()方法返回Result<Response>，但我找不到Response上任何通过块返回迭代器的方法。理想情况下，我可以做类似的事情：

client.get(&url.to_string())
    .send()
    .map(|mut res| {
        let mut chunk = String::new();
        // write this chunk to disk
    });

这是可能的，或者只有在Hyper下载整个文件后才会调用map吗？

Answer 1

对于每个接收到的数据块，有没有办法使用hyper运行某些功能？

超级Response实施Read。这意味着Response是一个流，您可以像通常使用流一样从中读取任意数据块。

对于它的价值，这是我用来从ICECat下载大文件的一段代码。我正在使用Read界面以显示终端中的下载进度。

此处的变量response是Hyper的Response的实例。

{
    let mut file = try_s!(fs::File::create(&tmp_path));
    let mut deflate = try_s!(GzDecoder::new(response));

    let mut buf = [0; 128 * 1024];
    let mut written = 0;
    loop {
        status_line! ("icecat_fetch] " (url) ": " (written / 1024 / 1024) " MiB.");
        let len = match deflate.read(&mut buf) {
            Ok(0) => break,  // EOF.
            Ok(len) => len,
            Err(ref err) if err.kind() == io::ErrorKind::Interrupted => continue,
            Err(err) => return ERR!("{}: Download failed: {}", url, err),
        };
        try_s!(file.write_all(&buf[..len]));
        written += len;
    }
}

try_s!(fs::rename(tmp_path, target_path));
status_line_clear();

我想用hyper下载大文件（500mb），如果下载失败则可以恢复。

这通常使用HTTP“Range”标头实现（参见RFC 7233）。

并非每个服务器都支持“Range”标头。我见过很多带有自定义HTTP堆栈的服务器，没有正确的“Range”支持，或者由于某种原因禁用了“Range”标头。因此，跳过Hyper的Response块可能是必要的后备。

但是如果你想加快速度并节省流量，那么恢复停止下载的主要方法应该是使用“Range”标题。

P.S。使用Hyper 0.12时，Hyper返回的响应体为Stream，并且对于收到的每个数据块运行一些函数，我们可以使用for_each流组合器：

extern crate futures;
extern crate futures_cpupool;
extern crate hyper; // 0.12
extern crate hyper_rustls;

use futures::Future;
use futures_cpupool::CpuPool;
use hyper::rt::Stream;
use hyper::{Body, Client, Request};
use hyper_rustls::HttpsConnector;
use std::thread;
use std::time::Duration;

fn main() {
    let url = "https://steemitimages.com/DQmYWcEumaw1ajSge5PcGpgPpXydTkTcqe1daF4Ro3sRLDi/IMG_20130103_103123.jpg";

    // In real life we'd want an asynchronous reactor, such as the tokio_core, but for a short example the `CpuPool` should do.
    let pool = CpuPool::new(1);
    let https = HttpsConnector::new(1);
    let client = Client::builder().executor(pool.clone()).build(https);

    // `unwrap` is used because there are different ways (and/or libraries) to handle the errors and you should pick one yourself.
    // Also to keep this example simple.
    let req = Request::builder().uri(url).body(Body::empty()).unwrap();
    let fut = client.request(req);

    // Rebinding (shadowing) the `fut` variable allows us (in smart IDEs) to more easily examine the gradual weaving of the types.
    let fut = fut.then(move |res| {
        let res = res.unwrap();
        println!("Status: {:?}.", res.status());
        let body = res.into_body();
        // `for_each` returns a `Future` that we must embed into our chain of futures in order to execute it.
        body.for_each(move |chunk| {println!("Got a chunk of {} bytes.", chunk.len()); Ok(())})
    });

    // Handle the errors: we need error-free futures for `spawn`.
    let fut = fut.then(move |r| -> Result<(), ()> {r.unwrap(); Ok(())});

    // Spawning the future onto a runtime starts executing it in background.
    // If not spawned onto a runtime the future will be executed in `wait`.
    // 
    // Note that we should keep the future around.
    // To save resources most implementations would *cancel* the dropped futures.
    let _fut = pool.spawn(fut);

    thread::sleep (Duration::from_secs (1));  // or `_fut.wait()`.
}

如何下载带有超级和恢复错误的大文件？

1 个答案: