请教一个 Python 中线程共享数据的问题

由于数据较多，使用 dataclass 当 struct 用
线程中写，main 中只读，所有没有加锁实际使用发现，这个数据共享不是完全引用，变量地址（使用 id 查看两边地址）会有改变，导致 main 和 thread 中的变量不是一个东西了，监测失败。
tricky 的是，第一次创建的线程没有问题。跑完一次，第二次再来一次就大概率出问题，后续再尝试就一直会出问题了，偶尔会成功。请问如果要实现多线程共享数据的读写，有什么最佳实现和模板吗

Python

线程

共享

7 条回复 2024-11-13 18:32:53 +08:00

smdbh

2024-10-08 11:32:02 +08:00

补充下，我当前是将 dataclass 的结构当参数传入 threading.Thead 的参数中，这个操作是否有问题

djangovcps

2024-10-08 11:40:16 +08:00

threading.lock ?

F532uKfxnXPiFdnC

2024-10-08 12:01:58 +08:00

```python
import threading
import time
from concurrent.futures import ThreadPoolExecutor
import unittest
from dataclasses import dataclass, field
from threading import Lock
import multiprocessing

@dataclass
class SharedData:
value: int = 0
# Using a Lock to ensure thread-safety when accessing shared data
lock: Lock = field(default_factory=Lock, init=False, repr=False)

def increment(self):
with self.lock:
self.value += 1

def get_value(self):
with self.lock:
return self.value

def worker(data: SharedData, num_iterations: int):
local_sum = 0
for _ in range(num_iterations):
local_sum += 1
# Use a lock to safely update the shared data
with data.lock:
data.value += local_sum

class TestSharedDataThreadSafety(unittest.TestCase):
def test_concurrent_increments(self):
shared_data = SharedData()
# Use 2x CPU count for threads to test both CPU-bound and I/O-bound scenarios
num_threads = multiprocessing.cpu_count() * 2
num_iteratiOns= 1000000 // num_threads

with ThreadPoolExecutor(max_workers=num_threads) as executor:
futures = [executor.submit(worker, shared_data, num_iterations) for _ in range(num_threads)]
for future in futures:
future.result()

expected_value = num_threads * num_iterations
self.assertEqual(shared_data.get_value(), expected_value,
f"Expected {expected_value}, but got {shared_data.get_value()}")

def test_race_condition(self):
shared_data = SharedData()
race_detected = threading.Event()

def racer():
with shared_data.lock:
initial_value = shared_data.value
time.sleep(0.001) # Simulate some work
# Check if the value has changed, which would indicate a race condition
if initial_value == shared_data.value:
shared_data.value += 1
else:
race_detected.set()

threads = [threading.Thread(target=racer) for _ in range(100)]
for t in threads:
t.start()
for t in threads:
t.join()

self.assertFalse(race_detected.is_set(), "Race condition detected")

def test_stress_test(self):
shared_data = SharedData()
stop_flag = threading.Event()

def stress_worker():
local_sum = 0
while not stop_flag.is_set():
local_sum += 1
# Use a lock to safely update the shared data after intensive local computation
with shared_data.lock:
shared_data.value += local_sum

# Use CPU count for threads to maximize resource utilization
threads = [threading.Thread(target=stress_worker) for _ in range(multiprocessing.cpu_count())]
for t in threads:
t.start()

time.sleep(5) # Run for 5 seconds to simulate prolonged stress
stop_flag.set()

for t in threads:
t.join()

print(f"Stress test final value: {shared_data.get_value()}")

if __name__ == '__main__':
unittest.main()
```

ClericPy

2024-10-08 23:17:03 +08:00

show me your code?

milkpuff

2024-10-09 13:25:40 +08:00

修改 data.a, data 的 id 不会变。

UN2758

331 天前

@qianchengv 佬，你这大段的代码还不如贴一个 gist 链接呢，比这个方便阅读多了

UN2758

331 天前

多线程，要么加锁，要么用 queue 之类的库，至于你的问题二，地址会移动可能是因为对象大小改变导致的重新分配地址