什么时候 ConcurrentDictionary TryMove 返回 false

如果字典不包含给定键的值，它只会返回 false 吗? 或者由于线程竞争条件(如另一个线程添加/更新某些内容) ，它也会返回 false 吗？

密码问题:

ConcurrentDictionary<int, string> cd = new ConcurrentDictionary<int, string>();


// This might fail if another thread is adding with key value of 1.
cd.TryAdd(1, "one");


// Will this ever fail if no other thread ever removes with the key value of 1?
cd.TryRemove(1);

编辑: 我认为它只有在不包含给定键值的情况下才会返回 false，但是需要绝对确定。

20457 次浏览
小开

The ConcurrentDictionary does not suffer from race conditions. That's why you use it.

Return Value

true if an object was removed successfully; otherwise, false.

小开
最佳答案

While Mitch is right that a ConcurrentDictionary is not vulnerable to race conditions, I think the answer to the question you are asking is that yes, if the key is present, TryRemove will work and will return true.

In the code you posted, there's no way that TryRemove would return false since cd is a local variable not accessed anywhere else. But if some code elsewhere were given a reference to this ConcurrentDictionary object and were removing keys on a separate thread, then it's possible that TryRemove could return false, even here -- but only because the key was already removed, not because some other action is being performed on the dictionary and the key is somehow "stuck" there.

小开

One other point to make:

// This might fail if another thread is adding with key value of 1.
cd.TryAdd(1, "one");

This comment is incorrect and possibly suffers from the same misconception about what it means to 'try'. It's not about a concurrent attempt to add, it's whether a value has already been added with key 1.

Consider a standard Dictionary<TKey,TValue>. The equivalent code would be:

if (!d.Contains(1))
d.Add(1, "one");

This requires two operations. There's no way to design such an API to be threadsafe, as cd might have a value with key 1 added between the call to Contains and Add, which would then result in Add throwing.

The concurrent collections have APIs that logically bundle these test-and-do pairs into single atomic operations, behind a single API.


提交答案