shove 0.11.3

Async tasks via pubsub on steroids. Comes with built-in support for complex queue configurations, audit logs, autoscaling consumer groups and more.
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
31
32
33
34
35
36
37

//! Internal `PublisherImpl` trait. Backend-specific publisher structs
//! implement this; users call the public `Publisher<B>` wrapper that
//! delegates here.

use std::collections::HashMap;

use crate::error::Result;
use crate::topic::Topic;

// Methods are anchored by the InMemory port's `_anchor_*` helpers in
// `backend::mod` under the `inmemory` feature. Under
// `--no-default-features` no backend is compiled, so the trait methods
// genuinely have no call site; `dead_code` is expected there and the
// per-trait allow avoids polluting the default build with warnings
// until Phase 5+ adds the generic wrappers.
#[allow(dead_code)]
pub(crate) trait PublisherImpl: Send + Sync {
    fn publish<T: Topic>(&self, msg: &T::Message) -> impl Future<Output = Result<()>> + Send;

    fn publish_with_headers<T: Topic>(
        &self,
        msg: &T::Message,
        headers: HashMap<String, String>,
    ) -> impl Future<Output = Result<()>> + Send;

    /// Publish a batch, returning the number of messages the backend
    /// confirmed as accepted alongside the overall result. On `Ok(())` the
    /// caller can assume `succeeded == msgs.len()`; on `Err(_)` the count
    /// reflects how many were accepted before the failure surfaced (e.g.
    /// SNS chunks, Kafka per-partition acks, RabbitMQ confirms), so the
    /// `messages_published_total` counter doesn't have to choose between
    /// overcounting failures or undercounting successes.
    fn publish_batch<T: Topic>(
        &self,
        msgs: &[T::Message],
    ) -> impl Future<Output = (u64, Result<()>)> + Send;
}