dabeaz/curioArchived
Curio
Curio est une bibliothèque Python dédiée à la concurrence structurée et à la programmation réseau asynchrone. Elle sert de framework pour gérer des milliers de connexions TCP simultanées grâce à des sockets non bloquants et des entrées/sorties à haut débit.
La bibliothèque organise les opérations asynchrones dans une hiérarchie de coroutines parentes et enfants pour garantir une terminaison et un nettoyage fiables des tâches dépendantes. Elle fournit une boîte à outils pour coordonner ces opérations concurrentes via des groupes de tâches et des délais d'attente (timeouts).
Le framework couvre un large éventail de capacités de gestion de la concurrence, incluant la communication inter-tâches via des files d'attente et la synchronisation de ressources partagées à l'aide de verrous (locks) et de sémaphores. Il inclut également la prise en charge de l'exécution de processus externes et le déchargement des tâches bloquantes vers des threads séparés pour maintenir la réactivité de la boucle d'événements.
Features
- Asynchronous Network Frameworks - Provides a foundational framework for building high-throughput, event-driven network servers and clients using asynchronous sockets.
- Structured Concurrency Hierarchies - Implements parent-child task relationships to ensure automatic lifecycle management and cancellation propagation.
- Task Coordination Strategies - Implements coordination strategies such as task groups and timeouts to ensure system stability.
- Asynchronous Network Clients - Provides non-blocking socket operations for high-concurrency network communication.
- Network I/O Multiplexing - Implements a central event loop for monitoring and managing thousands of concurrent network sockets.
- Python Coroutine Implementations - Provides a runtime implementation of coroutines specifically tailored for Python's asynchronous execution model.
- Structured Concurrency - Ensures child tasks are automatically terminated when their parent task completes or fails.
- Structured Concurrency Managers - Provides a framework for managing task lifecycles and cancellation through hierarchical scopes.
- Concurrent Connection Handling - Handles high-volume simultaneous TCP connections with minimal resource usage through asynchronous sockets.
- Asynchronous Task Orchestrators - Implements an orchestrator for coordinating concurrent operations via non-blocking event loops and task groups.
- Non-Blocking Event Loops - Uses a non-blocking event loop to manage multiple concurrent operations without blocking the main thread.
- Hierarchical Task Structures - Organizes asynchronous operations into nested logic structures to ensure reliable task cleanup.
- Asynchronous Network Programming - Provides a comprehensive framework for building servers and clients that handle concurrent connections via event loops.
- Blocking Work Offloaders - Offloads blocking I/O and CPU-heavy external process work to a separate thread pool.
- Thread Pool Offloading - Executes synchronous or CPU-intensive work in separate thread pools to prevent event loop stalling.
- Inter-Task Message Queues - Provides synchronized asynchronous queues to decouple producer and consumer coroutines.
- Inter-Coroutine Communication - Provides low-level mechanisms for passing values and signals between individual coroutines.
- Concurrent Resource Synchronization - Ships a set of locks, semaphores, and queues to synchronize shared state between concurrent coroutines.
- Coroutine Communication Channels - Facilitates the exchange of signals and information between concurrent coroutines using queues.
- Shared Resource Synchronization - Prevents race conditions using synchronization primitives like locks and semaphores.
- Time-Based Event Scheduling - Manages a priority queue of timed events to trigger task resumes at precise timestamps.
- Alternative Runtimes - Library for coroutine-based concurrency.
