smartasync 0.4.0

Unified sync/async API decorator with automatic context detection

SmartAsync

Sync or async: this WAS the question

Unified sync/async API decorator with automatic context detection

SmartAsync allows you to write async methods once and call them in both sync and async contexts without modification. It automatically detects the execution context and adapts accordingly.

Features

• ✅ Automatic context detection: Detects sync vs async execution context at runtime
• ✅ Zero configuration: Just apply the @smartasync decorator
• ✅ Asymmetric caching: Smart caching strategy for optimal performance
• ✅ Compatible with __slots__: Works with memory-optimized classes
• ✅ Pure Python: No dependencies beyond standard library

Installation

pip install smartasync

Quick Start

from smartasync import smartasync
import asyncio

class DataManager:
    @smartasync
    async def fetch_data(self, url: str):
        """Fetch data - works in both sync and async contexts!"""
        async with httpx.AsyncClient() as client:
            response = await client.get(url)
            return response.json()

# Sync context - no await needed
manager = DataManager()
data = manager.fetch_data("https://api.example.com/data")

# Async context - use await
async def main():
    manager = DataManager()
    data = await manager.fetch_data("https://api.example.com/data")

asyncio.run(main())

How It Works

SmartAsync uses asyncio.get_running_loop() to detect the execution context:

• Sync context (no event loop): Executes with asyncio.run()
• Async context (event loop running): Returns coroutine to be awaited

Asymmetric Caching

SmartAsync uses an intelligent caching strategy:

• ✅ Async context detected: Cached forever (can't transition from async to sync)
• ⚠️ Sync context: Always rechecked (can transition from sync to async)

This ensures correct behavior while optimizing for the most common case (async contexts in web frameworks).

Use Cases

1. CLI + HTTP API

from smartasync import smartasync
from smpub import PublishedClass, ApiSwitcher

class DataHandler(PublishedClass):
    api = ApiSwitcher()

    @api
    @smartasync
    async def process_data(self, input_file: str):
        """Process data file."""
        async with aiofiles.open(input_file) as f:
            data = await f.read()
        return process(data)

# CLI usage (sync)
handler = DataHandler()
result = handler.process_data("data.csv")

# HTTP usage (async via FastAPI)
# Automatically works without modification!

2. Testing

@smartasync
async def database_query(query: str):
    async with database.connect() as conn:
        return await conn.execute(query)

# Sync tests
def test_query():
    result = database_query("SELECT * FROM users")
    assert len(result) > 0

# Async tests
async def test_query_async():
    result = await database_query("SELECT * FROM users")
    assert len(result) > 0

3. Mixed Codebases

Perfect for gradually migrating sync code to async without breaking existing callers.

Performance

• Decoration time: ~3-4 microseconds (one-time cost)
• Sync context: ~102 microseconds (dominated by asyncio.run() overhead)
• Async context (first call): ~2.3 microseconds
• Async context (cached): ~1.3 microseconds

For typical CLI tools and web APIs, this overhead is negligible compared to network latency (10-200ms).

Advanced Usage

With __slots__

SmartAsync works seamlessly with __slots__ classes:

from smartasync import smartasync

class OptimizedManager:
    __slots__ = ('data',)

    def __init__(self):
        self.data = []

    @smartasync
    async def add_item(self, item):
        await asyncio.sleep(0.01)  # Simulate I/O
        self.data.append(item)

Cache Reset for Testing

@smartasync
async def my_method():
    pass

# Reset cache between tests
my_method._smartasync_reset_cache()

Limitations

• ⚠️ Cannot transition from async to sync: Once in async context, cannot move back to sync (this is correct behavior)
• ⚠️ Sync overhead: Always rechecks context in sync mode (~2 microseconds per call)

Thread Safety

SmartAsync is safe for all common use cases:

✅ Safe scenarios (covers 99% of real-world usage):

• Single-threaded applications (CLI tools, scripts)
• Async event loops (inherently single-threaded)
• Web servers with request isolation (new instance per request)
• Thread pools with instance-per-thread pattern

⚠️ Theoretical concern (anti-pattern, not recommended):

• Sharing a single instance across multiple threads in a thread pool

Why this isn't a real issue: The anti-pattern scenario defeats SmartAsync's purpose. If you're using thread pools with shared instances, you should use async workers instead for better performance and natural concurrency.

Recommendation: Create instances per thread/request, or better yet, use async patterns natively.

Related Projects

SmartAsync is part of the Genro-Libs toolkit:

• smartswitch - Rule-based function dispatch
• smpub - CLI/API framework (uses SmartAsync for async handlers)
• gtext - Text transformation and templates

Contributing

Contributions welcome! Please see CONTRIBUTING.md for guidelines.

License

MIT License - see LICENSE file for details.

Credits

Author: Giovanni Porcari (Genropy Team) Part of: Genro-Libs developer toolkit