PyAsync

Thread and Process-based parallelism for Python. Simple, explicit API for running tasks in parallel.

• Threads for I/O-bound tasks (HTTP requests, file operations)
• Processes for CPU-bound tasks (computations, data processing)

No magic, just straightforward parallel execution.

Installation

pip install python-async

Quick Start

I/O-Bound Tasks (Threads)

import pyasync
import requests

def fetch(url):
    return requests.get(url).json()

# Run 3 requests in parallel - takes ~1 second, not ~3 seconds!
results = pyasync.parallel(
    lambda: fetch("https://api.example.com/users/1"),
    lambda: fetch("https://api.example.com/users/2"),
    lambda: fetch("https://api.example.com/users/3")
)

CPU-Bound Tasks (Processes)

import pyasync
from functools import partial

def heavy_compute(n):
    return sum(i * i for i in range(n))

# Run computations in parallel processes (bypasses GIL)
results = pyasync.cpu_parallel(
    partial(heavy_compute, 10_000_000),
    partial(heavy_compute, 20_000_000),
    partial(heavy_compute, 30_000_000),
    timeout=10.0  # Optional timeout
)

API

Thread-Based (I/O-Bound)

parallel(*callables)

Run multiple functions in parallel threads. Returns results in order.

results = pyasync.parallel(
    lambda: requests.get("https://api1.com"),
    lambda: requests.get("https://api2.com"),
    lambda: requests.get("https://api3.com")
)
# All 3 run simultaneously!

background(callable)

Start a function in the background. Returns a Task.

task = pyasync.background(lambda: slow_operation())

# Do other work while it runs...
print("Working...")

# Get result when ready
result = task.result()

run(callable)

Run a single function in the thread pool.

result = pyasync.run(lambda: requests.get("https://api.com"))

---

Process-Based (CPU-Bound)

Note: Functions must be picklable. Use functools.partial instead of lambdas.

cpu_parallel(*callables, timeout=None, max_workers=None)

Run multiple functions in parallel processes with true parallelism.

from functools import partial

def compute(n):
    return sum(i * i for i in range(n))

results = pyasync.cpu_parallel(
    partial(compute, 1_000_000),
    partial(compute, 2_000_000),
    partial(compute, 3_000_000),
    timeout=10.0,    # TimeoutError if exceeded
    max_workers=4    # Limit processes
)

cpu_background(callable)

Start a function in a background process. Returns a CpuTask with fine-grained control.

task = pyasync.cpu_background(partial(heavy_compute, 100_000_000))

# Monitor status
print(f"Running: {task.running}")
print(f"Done: {task.done}")

# Wait with timeout
try:
    result = task.result(timeout=30.0)
except TimeoutError:
    print("Task took too long!")
    task.cancel()

cpu_run(callable, timeout=None)

Run a single function in a separate process and wait for result.

try:
    result = pyasync.cpu_run(partial(compute, 100_000_000), timeout=10.0)
except TimeoutError:
    print("Computation took too long!")

---

CpuTask Class

CpuTask provides fine-grained control over CPU-bound tasks:

Property/Method	Description
.done	True if task completed
.running	True if task is currently running
.cancelled	True if task was cancelled
.result(timeout=None)	Wait and get result (raises TimeoutError)
.exception(timeout=None)	Get exception if task failed
.cancel()	Attempt to cancel (only works if not started)
.add_done_callback(fn)	Add completion callback

---

CpuExecutor Context Manager

For advanced control over process pools:

with pyasync.CpuExecutor(max_workers=4, timeout=30.0) as executor:
    # Submit individual tasks
    task1 = executor.submit(compute, 1_000_000)
    task2 = executor.submit(compute, 2_000_000)
    
    # Get results
    print(task1.result())
    print(task2.result())
    
    # Or wait for all
    results = executor.wait_all()
    
    # Batch processing with map
    results = list(executor.map(compute, [1_000_000, 2_000_000, 3_000_000]))

Examples

Parallel Tasks (Threads)

$ python examples/simple_parallel.py

=== Simple Parallel Tasks ===

[Task A] Starting...
[Task B] Starting...
[Task C] Starting...
[Task A] Done!
[Task C] Done!
[Task B] Done!

Results: ['Task A completed', 'Task B completed', 'Task C completed']
Total time: 2.01s (longest task was 2s)

CPU-Bound Parallel Tasks (Processes)

$ python examples/cpu_parallel_tasks.py
============================================================
CPU-Bound Parallel Tasks Example
============================================================

1. Parallel Prime Counting
----------------------------------------
Ranges: [(1, 100000), (100000, 300000), (300000, 500000), (500000, 1000000)]
Primes found: [9592, 16405, 15541, 36960]
Total primes: 78498

Sequential time: 1.11s
Parallel time:   0.76s
Speedup:         1.5x

2. Background Task with Monitoring
----------------------------------------
Task started...
  done: False
  cancelled: False

Task completed!
  done: True
  result: 9592 primes found

Parallel API Calls

$ python examples/parallel_api_calls.py

=== Parallel API Calls ===

Fetching 3 users in parallel...
  - Leanne Graham (Sincere@april.biz)
  - Ervin Howell (Shanna@melissa.tv)
  - Clementine Bauch (Nathan@yesenia.net)

When to Use

Use Case	Function	Why
Multiple HTTP requests	parallel()	I/O-bound, threads work great
File operations	parallel()	I/O-bound
Data processing	cpu_parallel()	CPU-bound, needs true parallelism
Image/video processing	cpu_parallel()	CPU-bound
Long computation with timeout	cpu_run(fn, timeout=10)	Fine-grained control
Background computation	cpu_background()	Monitor and cancel if needed

Testing

python -m unittest discover -s tests -v

License

MIT