Root cause of the failing CI run was that examples/tests referencing feature-gated items were auto-detected (no required-features), so `--all-targets` compiled them under feature combinations where those modules didn't exist. Fixes: - mytheclipse Cargo.toml: declare the `high_level` example and `race_stress` integration test with required-features = ["full"]; `cargo build --all-targets` now skips them when full is off. This clears the whole test-matrix (workspace default, all-features, and every single-feature config) which all failed on E0432/E0433. - lib.rs: auto_metrics_service depends on service_builder, so regate it behind all(observability, resiliency) instead of observability alone (observability-only build compiled the module without resiliency). - mytheclipse-tracing: gate `pub mod fmt` behind any tracing-subscriber-providing feature so --no-default-features compiles. - mytheclipse-queue: gate `pub mod worker` behind in-memory (worker.rs requires tokio, only provided by in-memory). - clippy -D warnings fixes: deprecated base64 0.22 free fns -> Engine (paseto), unused key field, needless mut (service_builder), unused import/dead var/missing is_empty (bg_join), dead is_expired (dlock), while-let-iterator->for (parallel_map), type_complexity (shutdown_guard), MutexGuard held across await (middleware, now clones Arc'd layers), if-let-Err->is_err (queue), unused CliBuilder fields now wired into clap. - rustdoc -D warnings: resolve retry/MetricsBridge/CircuitBreaker/KeyRing intra-doc links and fix the unparseable lifecycle.rs code fence. - cargo fmt --all to satisfy the Rustfmt gate.
mytheclipse-queue
A unified job queue abstraction so your background work isn't locked to one
transport. Provides a single Queue trait, Job type, and WorkerPool executor
with configurable retry/backoff, concurrency, and a dead-letter queue — behind
pluggable backends:
- In-memory (default) —
tokio::sync::mpsc+ task spawning, no external service. - Redis (
redis) — LIST-based queue with atomic moves. - NATS JetStream (
nats) — durable consumer with ACK/NACK. - PostgreSQL (
postgres) —SKIP LOCKEDpolling.
All backends share the same WorkerPool driver; swapping is a one-line change
at construction time.
Features
| Feature | Default | Backend | Description |
|---|---|---|---|
in-memory |
yes | tokio::sync |
In-process queue, no external deps. |
redis |
no | redis crate (fred) |
Redis/Valkey list-based queue. |
nats |
no | async-nats |
NATS JetStream durable consumer. |
postgres |
no | tokio-postgres |
PostgreSQL SKIP LOCKED queue. |
Usage
use mytheclipse_queue::{InMemoryQueue, WorkerPool, Job, JobHandler, JobFuture};
fn print_handler() -> impl JobHandler {
struct PrintHandler;
impl JobHandler for PrintHandler {
fn handle(&self, job: Job) -> JobFuture {
Box::pin(async move {
println!("payload: {:?}", job.payload);
Ok(())
})
}
}
PrintHandler
}
#[tokio::main]
async fn main() -> Result<(), Box<dyn std::error::Error>> {
let queue = InMemoryQueue::new();
queue.enqueue("email", b"hello".to_vec()).await?;
let pool = WorkerPool::new(queue, 4);
pool.start("email", print_handler());
Ok(())
}
Swap InMemoryQueue::new() for RedisQueue::connect("redis://127.0.0.1") (with
the redis feature) or NatsQueue::connect("nats://127.0.0.1") (with the nats
feature) to move to a distributed broker without touching handler code.