学ぶ Async Locks and asyncio.Lock()

メニューを表示するにはスワイプしてください

The asyncio event loop runs in a single thread, so two coroutines never execute at the exact same moment. But they can still interleave – one coroutine can be interrupted at any await point, and another can modify shared state before the first resumes. asyncio.Lock() prevents this.

The Problem: Interleaved State Mutation

Without a lock, shared state can become inconsistent when multiple coroutines modify it concurrently:


              12345678910111213141516171819202122232425
            
import asyncio
import nest_asyncio

nest_asyncio.apply()

shared_counter = 0

# Incrementing a shared counter without a lock
async def increment(label):
    global shared_counter

    current_value = shared_counter
    await asyncio.sleep(0)  # Yielding control – another coroutine can run here
    shared_counter = current_value + 1
    print(f"{label}: counter = {shared_counter}")

async def main():
    await asyncio.gather(
        increment("task_A"),
        increment("task_B"),
        increment("task_C"),
    )
    print(f"Final counter: {shared_counter}")  # Expected 3, may get 1

asyncio.run(main())

await asyncio.sleep(0) yields control without actually sleeping. At that point, another coroutine reads the same stale value and overwrites it.

Using `asyncio.Lock()`

A lock ensures only one coroutine can execute the protected block at a time:


              123456789101112131415161718192021222324252627
            
import asyncio
import nest_asyncio

nest_asyncio.apply()

shared_counter = 0
lock = asyncio.Lock()

# Incrementing a shared counter safely with a lock
async def increment(label):
    global shared_counter

    async with lock:  # Acquiring the lock
        current_value = shared_counter
        await asyncio.sleep(0)  # Safe – no other coroutine can enter the lock
        shared_counter = current_value + 1
        print(f"{label}: counter = {shared_counter}")

async def main():
    await asyncio.gather(
        increment("task_A"),
        increment("task_B"),
        increment("task_C"),
    )
    print(f"Final counter: {shared_counter}")  # Always 3

asyncio.run(main())

The async with lock block is entered by only one coroutine at a time. Any other coroutine that tries to acquire the lock waits until it is released.

A Realistic Example: Shared Cache

Locks are commonly used to protect shared data structures like caches:


              1234567891011121314151617181920212223242526272829303132333435363738
            
import asyncio
import httpx
import nest_asyncio

nest_asyncio.apply()

cache = {}
cache_lock = asyncio.Lock()

# Fetching a post with a shared cache protected by a lock
async def fetch_post_cached(client, post_id):
    async with cache_lock:
        if post_id in cache:
            print(f"Cache hit for post {post_id}")
            return cache[post_id]

    response = await client.get(
        f"https://jsonplaceholder.typicode.com/posts/{post_id}"
    )
    title = response.json()["title"]

    async with cache_lock:
        cache[post_id] = title

    return title

async def main():
    async with httpx.AsyncClient() as client:
        results = await asyncio.gather(
            fetch_post_cached(client, 1),
            fetch_post_cached(client, 1),  # Second request hits cache
            fetch_post_cached(client, 2),
        )

    for result in results:
        print(result)

asyncio.run(main())

When to Use Locks

Locks are needed when:

Multiple coroutines read and then write the same variable;
A check-then-act sequence must be atomic (e.g., cache lookup then insert);
A shared data structure (list, dict) is modified by more than one coroutine.

Locks are not needed for read-only access to shared data – concurrent reads are always safe.

すべて明確でしたか？

フィードバックありがとうございます！

セクション 3. 章 2

AIに質問する

何でも質問するか、提案された質問の1つを試してチャットを始めてください

セクション 3. 章 2

Async Locks and asyncio.Lock()

The Problem: Interleaved State Mutation

Using asyncio.Lock()

A Realistic Example: Shared Cache

When to Use Locks

Using `asyncio.Lock()`