String.intern() 线程安全有保证吗？

Question

Is there any documented guarantee that String.intern() is thread safe? The javadoc alludes to it but does not address it directly:

Returns a canonical representation for the string object. A pool of strings, initially empty, is maintained privately by the class String.

When the intern method is invoked, if the pool already contains a string equal to this String object as determined by the equals(Object) method, then the string from the pool is returned. Otherwise, this String object is added to the pool and a reference to this String object is returned.

It follows that for any two strings s and t, s.intern() == t.intern() is true if and only if s.equals(t) is true.

All literal strings and string-valued constant expressions are interned. String literals are defined in section 3.10.5 of the The Java™ Language Specification.

Notably, the javadoc says that a String from the pool is guaranteed to be returned, but not that the pool itself is thread safe (so as written it appears to leave the door open for a pool entry to be replaced in the event of competing threads, although I think this interpretation is unlikely).

And the JDK source of String shows intern() to be a native method which does not shed any light on its thread safety:

public native String intern();

My concern is specifically related to whether the following would be fully thread safe, guaranteeing only one MyObject is created (not merely stored in the cache) for each unique string value in the face of concurrent requests:

public void myMethod(String myString)
{
    // Get object from cache, the cache itself is thread safe
    MyObject object = myCache.get(myString);
    if (object == null)
    {
        synchronized(myString.intern())
        {
            // Retry cache to avoid race condition
            object = myCache.get(myString);
            if (object == null)
            {
                object = new MyObject(myString);
                // ... do some startup / config of the object ...
                myCache.put(object);
            }
        }
    }
    // do something useful with the object
}

I wish to avoid synchronizing on either the method or the cache itself since the creation of the object can take some time (entails network access). There are workarounds such as maintaining a local thread safe cache / pool of strings but it is not worth doing unless necessary. The memory implications of String.intern() (inability to remove interned strings from the cache) are not relevant for this particular use case (small number of strings being used).

I believe that String.intern() is thread safe and that the code above is fine, but the lack of direct confirmation from a reputable source leaves me slightly concerned.

This question has been asked here multiple times before, but no concrete answer with references has been provided:

• Is String.Intern() thread safe
• How do intern'd strings behave between different threads and classloaders?

Answer 1

That actually sounds like you might want a Guava Striped<Lock> , which maps objects to locks in a hashed way. Synchronizing on interned strings seems like a potentially dangerous hack that could have weird side effects if any other code you're using had the same idea.