hexz-store 0.8.0

Storage backends for Hexz snapshots (local file, HTTP, S3)
Documentation 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30

//! Shared Tokio runtime for async storage backends.
//!
//! Lazily initialises a process-global multi-threaded runtime so that
//! `HttpBackend` and `S3Backend` share one thread pool rather than each
//! creating their own.

use std::io;
use std::sync::OnceLock;
use tokio::runtime::{Handle, Runtime};

static GLOBAL_RUNTIME: OnceLock<Runtime> = OnceLock::new();

/// Returns a handle to the shared Tokio runtime, creating it on first call.
///
/// # Errors
///
/// Returns an I/O error if the Tokio runtime cannot be created (e.g. the OS
/// cannot spawn worker threads).
pub fn global_handle() -> io::Result<Handle> {
    // OnceLock::get_or_init is infallible and doesn't propagate errors from
    // the init closure, so we try to create the runtime and store it only on
    // success.  On repeated calls we just hand back the existing handle.
    if let Some(rt) = GLOBAL_RUNTIME.get() {
        return Ok(rt.handle().clone());
    }
    let rt = Runtime::new()?;
    // Another thread may have raced us; that's fine — get_or_init returns
    // whichever value was stored first.
    Ok(GLOBAL_RUNTIME.get_or_init(|| rt).handle().clone())
}