PCR (Python Concurrency Runtime)

日本語版はこちら (Japanese Version)

PCR is a high-performance, task-based concurrency runtime for Python that combines flexibility with a powerful execution engine. It automatically builds a DAG (Directed Acyclic Graph) of complex task dependencies and executes them using the most suitable executor based on task characteristics like CPU load, I/O wait, or distributed environment.

Key Features

• Smart Scheduling: Automatically detects task types (CPU-bound vs I/O-bound) and dispatches them to Thread, Process, or Remote Workers.
• Resilience: Built-in support for retries, timeouts, and automatic worker isolation.
• Advanced Visualization: Automatically generates execution timelines (Gantt charts) and DAG structures (Mermaid charts).
• Async Integration: Transparently handles async def functions, enabling high-efficiency I/O parallelism on a single thread.
• Distributed Execution: Scalable distributed processing via a lightweight FastAPI-based worker server.

Quick Start

from pcr import task, run, parallel
import time

@task
def fetch_data(id):
    time.sleep(1)
    return f"Data {id}"

@task(cpu_bound=True)
def process_data(data):
    # Heavy computation executed in a ProcessPool
    return data.upper()

# Build the DAG
t1 = fetch_data(1)
t2 = fetch_data(2)
combined = process_data(parallel([t1, t2]))

# Execute with visual monitoring
result = run(combined, monitor=True, export_timeline="timeline.txt")
print(result)

Installation

pip install .

Testing

Keep your runtime healthy with pytest:

pytest

Project Structure

• pcr/: Core source code.
• tests/: Comprehensive test suite using pytest.
• examples/: Sample code for various use cases (Distributed, Async, etc.).

License

MIT License - feel free to use and contribute!