pygoroutine 0.1.9

🚀 Go-like concurrency in Python with channels and true parallelism.

pygoroutine 🚀

Go-like Concurrency in Python.

pygoroutine brings the simplicity and power of Go's concurrency model—goroutines and channels—to Python. It provides a dead-simple API to make concurrent programming feel effortless and intuitive, whether you're dealing with I/O-bound or CPU-bound tasks.

Key Features

• Dead-Simple Concurrency: Fire-and-forget tasks with a single go() call.
• Go-style Channels: Elegant communication using ch << value to send and for item in ch: to receive.
• Powerful Concurrency Patterns: Go-like select, WaitGroup, Context, and Once primitives for sophisticated coordination.
• True Parallelism: Bypass the GIL for CPU-bound tasks with process=True.
• Unified API: Handles async and regular functions automatically.
• Robust Lifecycle Management: An optional GoroutineManager provides fine-grained control for libraries and complex applications.

Installation

pip install pygoroutine

Quick Start: The Go-like Way

This example demonstrates the core features: starting a concurrent task with go() and communicating with it over a channel.

import time
from gopy import go, nc

def producer(ch):
    """A producer "goroutine" that sends numbers over a channel."""
    print("Producer starting...")
    for i in range(5):
        message = f"Message #{i+1}"
        print(f"-> Sending: '{message}'")
        ch << message  # Send a value into the channel
        time.sleep(0.5)
    
    ch.close()
    print("Producer finished.")

def main():
    ch = nc()
    go(producer, ch)

    # The main thread becomes the consumer.
    print("Consumer waiting for messages...")
    for received_message in ch:
        print(f"<- Received: '{received_message}'")
    
    print("Consumer finished. All tasks complete.")

if __name__ == "__main__":
    main()

Core Concepts

1. The go() Function The go() function is the heart of the library. It runs any function or coroutine concurrently without blocking and returns a Future object.

from gopy import go
import time

def my_sync_task(name):
    time.sleep(1)
    return f"Sync task '{name}' finished."

future = go(my_sync_task, "A")
print("Main thread is not blocked.")

You can optionally wait for the result
result = future.result()
print(result)

2. Channels for Communication Channels provide a safe and elegant way for your concurrent tasks to communicate. Send: channel << value Receive (Loop): for item in channel: Receive (Single): item = channel.get() Close: channel.close()

3. True Parallelism for CPU-Bound Tasks Bypass Python's GIL by running CPU-bound tasks in a separate process with the process=True flag.

from gopy import go

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

This runs in another process, utilizing another CPU core.

future = go(sum_squares, 10_000_000, process=True)
result = future.result()
print(f"Result from process: {result}")

Advanced Go-like Patterns

pygoroutine also includes implementations of Go's most powerful concurrency primitives.

select: Waiting on Multiple Channels

The select statement waits for several channel operations to be ready and executes the first one that is.

from gopy import go, nc, select, Case, GET
import time

ch1 = nc()
ch2 = nc()

def worker(ch, delay, msg):
    time.sleep(delay)
    ch << msg

go(worker, ch1, 0.2, "from ch1")
go(worker, ch2, 0.1, "from ch2")

# select blocks until one of the cases is ready
ready_case = select([
    Case(ch1, GET),
    Case(ch2, GET),
])

# The result is attached to the case object

print(f"Received '{ready_case.value}' from the first ready channel.")

Output: Received 'from ch2' from the first ready channel.

WaitGroup: Waiting for a Group of Tasks

A WaitGroup blocks until a collection of goroutines has finished.

from gopy import go, WaitGroup, defer
import time

wg = WaitGroup()

def worker(id):
    with defer(wg.done): # Ensures wg.done() is called on exit
        print(f"Worker {id} starting...")
        time.sleep(0.5)
        print(f"Worker {id} finished.")

wg.add(3) # Set the counter
for i in range(3):
    go(worker, i)

print("Main thread waiting...")
wg.wait() # Blocks until the counter is zero
print("All workers are done.")

Context: Cancellation and Timeouts

A Context provides a standardized way to signal cancellation or deadlines across multiple goroutines.

from gopy import go, new_context_with_timeout, TimeoutError
import time

def slow_worker(ctx):
    print("Worker starting, has 3 seconds to complete.")
    for i in range(3):
        if ctx.is_done():
            print(f"Worker cancelled: {ctx.err()}")
            return
        time.sleep(1)
        print(f"Worker heartbeat {i+1}...")
    print("Worker finished successfully.")

# Create a context that times out after 1.5 seconds
ctx = new_context_with_timeout(1.5)
future = go(slow_worker, ctx=ctx)

try:
    future.result()
except TimeoutError as e:
    print(f"Main thread caught error: {e}")

Once: Do Something Exactly Once

A Once object ensures that a given function is executed only one time, no matter how many concurrent tasks try to call it. It's perfect for thread-safe lazy initialization.

from gopy import go, Once, WaitGroup

initializer = Once()

def setup_resource():
    print("--- Initializing shared resource ONCE ---")

def worker(id, wg):
    defer(wg.done)
    print(f"Worker {id} requesting resource.")
    initializer.do(setup_resource)
    print(f"Worker {id} has resource.")

wg = WaitGroup()
wg.add(3)
for i in range(3):
    go(worker, i, wg)
wg.wait()

The "Initializing shared resource" message will only print once.

Advanced Usage: The GoroutineManager

For libraries or applications needing explicit setup and teardown, use the GoroutineManager. It provides a context manager for clean, predictable lifecycle management.

from gopy import GoroutineManager
import time

def worker(ch):
    time.sleep(0.1)
    ch << "done"

with GoroutineManager() as app:
    ch = app.nc()
    app.go(worker, ch)
    result = ch.get()
    print(f"Received '{result}' from worker.")

print("Manager has been shut down.")

License

This project is licensed under the MIT License - see the LICENSE file for details.