python-trio/trio
Trio
Trio is an asynchronous I/O runtime and concurrency library for Python. It provides a system for executing non-blocking network and disk operations through a centralized event loop and task scheduler.
The library is built on a structured concurrency model, which ensures that asynchronous tasks are bound to a specific lifetime and cannot outlive the scope that started them. It utilizes a nursery-based task manager to track task lifecycles in a parent-child tree, preventing orphaned concurrent operations by requiring child tasks to be joined before their parent scope exits.
The framework covers asynchronous I/O programming and high concurrency networking, allowing for the management of multiple parallel operations and network connections. It also includes capabilities for concurrent task coordination and async resource management to ensure clean shutdowns of tasks and system resources.
Features
- Concurrency Libraries - Provides a complete runtime and library for managing asynchronous I/O and parallel tasks in Python.
- Asynchronous Event Loops - Provides a single-threaded event loop that manages a queue of pending I/O events and callbacks.
- Resource Lifecycle Managers - Manages the lifecycle of network sockets and processes to ensure clean resource disposal and prevent leaks.
- Scope-Based Resource Cleanup - Provides predictable teardown of active operations by automatically releasing resources when tasks go out of scope.
- High-Concurrency Networking - Enables the creation of Python servers and clients capable of maintaining thousands of active connections efficiently.
- Asynchronous I/O Libraries - Provides a non-blocking interface and event loop for handling simultaneous network connections and file operations in Python.
- Async I/O Runtimes - Implements a centralized event loop and task scheduler for executing non-blocking network and disk operations.
- Structured Concurrency Managers - Implements a structured concurrency model where asynchronous tasks are strictly bound to the lifetime of their creating scope.
- Structured Concurrency Hierarchies - Uses a nursery-based task hierarchy to ensure all child concurrent operations are joined before the parent scope exits.
- Cancellation Propagation - Propagates cancellation signals down the task tree using scoped exceptions to ensure predictable resource cleanup.
- Nursery Managers - Implements nursery-based task management to ensure child tasks are always joined, preventing orphaned concurrent operations.
- Cooperative Schedulers - Schedules multiple concurrent tasks on a single thread by requiring functions to voluntarily yield control.
- Event-Driven I/O - Handles high-concurrency network connections and subprocess monitoring using an event-driven asynchronous I/O framework.
- Asynchronous Shutdown Coordination - Coordinates the asynchronous shutdown of tasks by blocking process termination until all registered operations complete.
- I/O Polling - Implements low-level system selector interfaces to efficiently wake asynchronous tasks when I/O data is ready.
- Process Signal Management - Processes OS-level interrupt signals to ensure concurrent tasks shut down cleanly without leaving orphaned processes.
- Concurrent Task Runners - Manages and monitors multiple concurrent asynchronous tasks to ensure a predictable and organized order of execution.
- Alternative Runtimes - Pythonic library for asynchronous I/O.
- Asynchronous Programming - Simplifies async I/O with a structured concurrency model.
- Networking - Library for asynchronous concurrency and I/O.
