Per-request timeouts

[joshtriplett]

I'd love to set a timeout on a per-request basis that overrides the timeout in the config. That way, individual requests can have a default timeout, while known long-running requests can disable the timeout.

[yoshuawuyts]

This is not a perfect answer, but using something like futures-time (blog) could perhaps do what you want it to:

use futures_time::prelude::*;
use futures_time::time::Duration;

let res = surf::get("api.example.com")
    .timeout(Duration::from_secs(2))    // provided by futures-time
    .await??;                           // the first `?` is surf, the second `?` is futures-time

It only allows defining timeouts shorter than the default; if they're longer the default will run first. So it may not fully suit your use case. But it might be a good way to get things working in the shorter-term without needing to modify surf directly.

There is probably a bigger question here too about what the right ways of structuring timeout hierarchies are, and for example a non-async version of surf should reason about timeouts (if at all). Buy maybe it's simply enough to say: "if you have an arbitrary operation you want to be able to time-out, you should call the timeout method on the future". I'm not sure heh.

Anyway, I hope this is helpful!