问题
I have some static initializer code:
someMethodThatTakesAHashMap(new HashMap<K, V>() {
{
put("a","value-a");
put("c","value-c");}
});
For some reason I am receiving a warning from Eclipse: The serializable class does not declare a static final serialVersionUID.
Is this complaining about the anonymous class? What can I do about that, or should I just suppress it.
回答1:
The syntax you're using is called double-brace initialization - which is actually an "instance initialization block that is part of an anonymous inner class" (certainly not a hack). So, when using this notation, you are actually defining a new class(!).
The "problem" in your case is that HashMap implements Serializable. This interface doesn't have any methods and serves only to identify the semantics of being serializable. In other words, it's a marker interface and you concretely don't have to implement anything. But, during deserialization, Java uses a version number called a serialVersionUID to verify that the serialized version is compatible with the target. If you don't provide this serialVersionUID, it will be calculated. And, as documented in the javadoc of Serializable, the calculated value is extremely sensitive and it is thus recommended be explicitly declare it to avoid any deserialization problems. And this is what Eclipse is "complaining" about (note that this is just a warning).
So, to avoid this warning, you could add a serialVersionUID to your annonymous inner class:
someMethodThatTakesAHashMap(new HashMap<String, String>() {
private static final long serialVersionUID = -1113582265865921787L;
{
put("a", "value-a");
put("c", "value-c");
}
});
But you loose the conciseness of the syntax (and you may not even need it).
Another option would thus be to ignore the warning by adding a @SuppressWarnings("serial") to the method where you are calling someMethodThatTakesAHashMap(Map). This seems more appropriate in your case.
That all being said, while this syntax is concise, it has some drawbacks. First, if you hold a reference on the object initialized using a double-brace initialization, you implicitly hold a reference to the outer object which won't be eligible for garbage collection. So be careful. Second (this sounds like micro optimization though), double-brace initialization has a very a little bit of overhead. Third, this technique actually uses anonymous inner classes as we saw and thus eats a bit of permgen space (but I doubt that this is really a problem unless you really abuse them). Finally - and this is maybe the most important point - I am not sure it makes the code more readable (it's not a well known syntax).
So, while I like to use it in tests (for the conciseness), I tend to avoid using it in "regular" code.
回答2:
Yes, you could suppress the warning, but I'd rewrite it like this:
HashMap<String, String> map = new HashMap<String, String>();
map.put("a","value-a");
map.put("c","value-c");
someMethodThatTakesAHashMap(map);
No suppressing needed, and much better to read, IMO.
回答3:
I generally agree with Bart K., but for informational purposes:
The warning can also be eliminated by adding the field, which can be automatically generated by hitting ctrl+1.
The warning can also be suppressed by adding the @SuppressWarnings("serial") annotation before the definition.
The anonymous class implements Serializeable, and Serializeable requires this static field so that versions can be distinguished when serializing and de-serializing. More information here:
http://www.javablogging.com/what-is-serialversionuid/
回答4:
The ImmutableMap class from the Google Collections library is useful for this situation. e.g.
someMethodThatTakesAHashMap(ImmutableMap.<K, V>builder().put("a","value-a").put("c","value-c").build());
or
someMethodThatTakesAHashMap(ImmutableMap.of("a","value-a","c","value-c"));
回答5:
To address the other half of your question, "should I suppress it?" --
Yes. In my opinion, this is a terrible warning. serialVersionUID should by default not be used, not the other way around.
If you don't add serialVersionUID, the worst thing that happens is that two versions of an object that are actually serialization-compatible are deemed incompatible. serialVersionUID is a way to declare that the serialization compatibility has not changed, overriding Java's default assessment.
Using serialVersionUID, the worst thing that happens is that you inadvertently fail to update the ID when the class's serialized form changes in an incompatible way. At best, you also get a runtime error. At worst, something worse happens. And imagine how easy it is to fail to update it.
回答6:
Your intent was to initialize an anonymous instance of HashMap. The warning is clue that your code is doing more than you intended.
What we're looking for is a way to initialize an anonymous HashMap instance. What we have above creates an anonymous subclass of HashMap then creates an anonymous instance of that anonymous class.
Because the code does more than was intended, I'd call it a hack.
What we really want is something like this:
foo(new HashMap<String, String>({"a", "value-a"}, {"c", "value-c"}));
But alas this isn't valid Java. There isn't a way to do anything this in a type-safe way using an array of key/value pairs. Java simple doesn't have the expressive power.
The Google Collection's ImmutableMap.of static methods are close but it means creating a version of the factory method for various numbers of key/value pairs. (See finnw's answer.)
So keep things simple. Go with Bart K's solution unless your code is littered with this initialization. If so use ImmutableMap. Or roll your own HashMap subclass with the "of" style factory methods. Or create these "of" style factory methods in a utility class. Here's one for two key/value pairs:
public final MapUtil {
public static <K,V> Map<K,V> makeMap(K k1, V v1, K k2, V v2) {
Map<K,V> m = new HashMap<K,V>();
m.put(k1, v1);
m.put(k2, v2);
return m;
}
}
Embrace the verbosity and take solace in the knowledge your corporate co-workers are wearing the same shackles as you.
来源：https://stackoverflow.com/questions/1514881/understanding-this-warning-the-serializable-class-does-not-declare-a-static-fin