Events: parallel

parallel removes event-level serialization for a bus, so multiple events can be in-flight simultaneously. Companion runnable example:

Lifecycle impact

Events still enqueue and are tracked in history.
The bus does not gate execution with an event semaphore.
Handler-level concurrency rules still apply within each event.
Ordering guarantees become weaker under load because events can overlap.

Execution order example

Python
TypeScript
Go
Rust

import asyncio
from abxbus import BaseEvent, EventBus

class ParallelEvent(BaseEvent):
    order: int

bus = EventBus('ParallelEventBus', event_concurrency='parallel', event_handler_concurrency='parallel')

in_flight = 0
max_in_flight = 0
release = asyncio.Event()

async def handler(_: ParallelEvent) -> None:
    global in_flight, max_in_flight
    in_flight += 1
    max_in_flight = max(max_in_flight, in_flight)
    await release.wait()
    await asyncio.sleep(0.01)
    in_flight -= 1

bus.on(ParallelEvent, handler)

bus.emit(ParallelEvent(order=0))
bus.emit(ParallelEvent(order=1))

await asyncio.sleep(0)
release.set()
await bus.wait_until_idle()

assert max_in_flight >= 2

import { BaseEvent, EventBus } from 'abxbus'
import { z } from 'zod'

const ParallelEvent = BaseEvent.extend('ParallelEvent', { order: z.number() })

const bus = new EventBus('ParallelEventBus', {
  event_concurrency: 'parallel',
  event_handler_concurrency: 'parallel',
})

let inFlight = 0
let maxInFlight = 0
let release!: () => void
const gate = new Promise<void>((resolve) => {
  release = resolve
})

bus.on(ParallelEvent, async () => {
  inFlight += 1
  maxInFlight = Math.max(maxInFlight, inFlight)
  await gate
  await new Promise((resolve) => setTimeout(resolve, 10))
  inFlight -= 1
})

bus.emit(ParallelEvent({ order: 0 }))
bus.emit(ParallelEvent({ order: 1 }))

await new Promise((resolve) => setTimeout(resolve, 0))
release()
await bus.waitUntilIdle()

if (maxInFlight < 2) throw new Error('expected overlapping events')

package main

import (
	"sync"
	"time"

	abxbus "github.com/ArchiveBox/abxbus/abxbus-go"
)

func main() {
	bus := abxbus.NewEventBus("ParallelEventBus", &abxbus.EventBusOptions{
		EventConcurrency:        abxbus.EventConcurrencyParallel,
		EventHandlerConcurrency: abxbus.EventHandlerConcurrencyParallel,
	})

	var mu sync.Mutex
	inFlight := 0
	maxInFlight := 0
	release := make(chan struct{})

	bus.On("ParallelEvent", "handler", func(event *abxbus.BaseEvent) (any, error) {
		mu.Lock()
		inFlight++
		if inFlight > maxInFlight {
			maxInFlight = inFlight
		}
		mu.Unlock()

		<-release
		time.Sleep(10 * time.Millisecond)

		mu.Lock()
		inFlight--
		mu.Unlock()
		return nil, nil
	}, nil)

	bus.Emit(abxbus.NewBaseEvent("ParallelEvent", map[string]any{"order": 0}))
	bus.Emit(abxbus.NewBaseEvent("ParallelEvent", map[string]any{"order": 1}))

	time.Sleep(10 * time.Millisecond)
	close(release)
	bus.WaitUntilIdle(nil)

	if maxInFlight < 2 {
		panic("expected overlapping events")
	}
}

use abxbus::{
    event,
    event_bus::{EventBus, EventBusOptions},
    BaseEvent,
    types::{EventConcurrencyMode, EventHandlerConcurrencyMode},
};
use futures::executor::block_on;

event! {
    struct ParallelEvent {
        event_result_type: (),
    }
}

let bus = EventBus::new_with_options(
    Some("ParallelEventBus".to_string()),
    EventBusOptions {
        event_concurrency: EventConcurrencyMode::Parallel,
        event_handler_concurrency: EventHandlerConcurrencyMode::Parallel,
        ..EventBusOptions::default()
    },
);

bus.on(ParallelEvent, |_event: ParallelEvent| async move { Ok(()) });
bus.emit(ParallelEvent { ..Default::default() });
bus.emit(ParallelEvent { ..Default::default() });
block_on(bus.wait_until_idle(None));

Notes

Use when throughput matters more than deterministic event ordering.
Combine with idempotent handlers and explicit external coordination when needed.

​Lifecycle impact

​Execution order example

​Notes

Lifecycle impact

Execution order example

Notes