What is the reason why “synchronized” is not allowed in Java 8 interface methods?

java java-8 synchronized default-method jsr335

In Java 8, I can easily write:

interface Interface1 {
    default void method1() {
        synchronized (this) {
            // Something
        }
    }

    static void method2() {
        synchronized (Interface1.class) {
            // Something
        }
    }
}

I will get the full synchronisation semantics that I can use also in classes. I cannot, however, use the synchronized modifier on method declarations:

interface Interface2 {
    default synchronized void method1() {
        //  ^^^^^^^^^^^^ Modifier 'synchronized' not allowed here
    }

    static synchronized void method2() {
        // ^^^^^^^^^^^^ Modifier 'synchronized' not allowed here
    }
}

Now, one can argue that the two interfaces behave the same way except that Interface2 establishes a contract on method1() and on method2(), which is a bit stronger than what Interface1 does. Of course, we might also argue that default implementations should not make any assumptions about concrete implementation state, or that such a keyword simply wouldn't pull its weight.

Question:

What is the reason why the JSR-335 expert group decided not to support synchronized on interface methods?

Synchronized is an implementation behavior and it changes final byte code result made by compiler so it can be used beside a code. It has no sense in method declaration. It should be confusing what has compiler produce if synchronized is on abstraction layer.

@MartinStrejc: That might be an explanation for omitting default synchronized, yet not necessarily for static synchronized, although I would accept that the latter might've been omitted for consistency reasons.

I'm not sure if this question adds any value as the synchronized modifier may be overriden in subclasses, hence it would only matter if there was something as final default methods. (Your other question)

@skiwi: The overriding argument is not sufficient. Subclasses may override methods that are declared synchronized in super classes, effectively removing synchronization. I wouldn't be surprised that not supporting synchronized and not supporting final is related, though, maybe because of multiple inheritance (e.g. inheriting void x() and synchronized void x(), etc.). But that's speculation. I'm curious about an authoritative reason, if there is one.

>>"Subclasses may override methods that are declared synchronized in super classes, effectively removing synchronization"... only if they don't call super which requires a full re-implementation and possible access to private members. Btw, there is a reason those methods are called "defenders" - they are present to allow easier adding new methods.

Brian Goetz

While at first it might seem obvious that one would want to support the synchronized modifier on default methods, it turns out that doing so would be dangerous, and so was prohibited.

Synchronized methods are a shorthand for a method which behaves as if the entire body is enclosed in a synchronized block whose lock object is the receiver. It might seem sensible to extend this semantics to default methods as well; after all, they are instance methods with a receiver too. (Note that synchronized methods are entirely a syntactic optimization; they're not needed, they're just more compact than the corresponding synchronized block. There's a reasonable argument to be made that this was a premature syntactic optimization in the first place, and that synchronized methods cause more problems than they solve, but that ship sailed a long time ago.)

So, why are they dangerous? Synchronization is about locking. Locking is about coordinating shared access to mutable state. Each object should have a synchronization policy that determines which locks guard which state variables. (See Java Concurrency in Practice, section 2.4.)

Many objects use as their synchronization policy the Java Monitor Pattern (JCiP 4.1), in which an object's state is guarded by its intrinsic lock. There is nothing magic or special about this pattern, but it is convenient, and the use of the synchronized keyword on methods implicitly assumes this pattern.

It is the class that owns the state that gets to determine that object's synchronization policy. But interfaces do not own the state of the objects into which they are mixed in. So using a synchronized method in an interface assumes a particular synchronization policy, but one which you have no reasonable basis for assuming, so it might well be the case that the use of synchronization provides no additional thread safety whatsoever (you might be synchronizing on the wrong lock). This would give you the false sense of confidence that you have done something about thread safety, and no error message tells you that you're assuming the wrong synchronization policy.

It is already hard enough to consistently maintain a synchronization policy for a single source file; it is even harder to ensure that a subclass correctly adhere to the synchronization policy defined by its superclass. Trying to do so between such loosely coupled classes (an interface and the possibly many classes which implement it) would be nearly impossible and highly error-prone.

Given all those arguments against, what would be the argument for? It seems they're mostly about making interfaces behave more like traits. While this is an understandable desire, the design center for default methods is interface evolution, not "Traits--". Where the two could be consistently achieved, we strove to do so, but where one is in conflict with the other, we had to choose in favor of the primary design goal.

Note also that in JDK 1.1, the synchronized method modifier appeared in the javadoc output, misleading people into thinking that it was part of the specification. This was fixed in JDK 1.2. Even if it appears on a public method, the synchronized modifier is part of the implementation, not the contract. (Similar reasoning and treatment occurred for the native modifier.)

A common mistake in early Java programs was to sprinkle enough synchronized and thread safe components around and you had an almost thread safe program. The problem was this usually worked ok but it broken in surprising and brittle ways. I agree that understanding how your locking works is a key to robust applications.

@BrianGoetz Very good reason. But why is synchronized(this) {...} allowed in a default method? (As shown in Lukas's question.) Doesn't that allow the default method to own the state of the implementation class too? Don't we want to prevent that too? Will we need a FindBugs rule to find the cases for which uninformed developers do that?

@Geoffrey: No, there's no reason to restrict this (though it should always be used with care.) The sync block requires the author to explicitly select a lock object; this allows them to participate in the synchronization policy of some other object, if they know what that policy is. The dangerous part is assuming that synchronizing on 'this' (which is what sync methods do) is actually meaningful; this needs to be a more explicit decision. That said, I expect sync blocks in interface methods to be pretty rare.

@GeoffreyDeSmet: For the same reason you can do e.g. synchronized(vector). It you want to be safe, you should never use a public object (such as this itself) for locking.

zhenke zhu

public class ParentSync {

public synchronized void parentStart() {
    System.out.println("I am " + this.getClass() + " . parentStarting. now:" + nowStr());
    try {
        Thread.sleep(30000);
    } catch (InterruptedException e) {
        e.printStackTrace();
    }
    System.out.println("I am " + this.getClass() + " . parentFinished. now" + nowStr());
}

private String nowStr() {
    return new SimpleDateFormat("yyyy-MM-dd HH:mm:ss").format(new Date());
}
}


public class SonSync1 extends ParentSync {
public void sonStart() {
    System.out.println("I am " + this.getClass() + ". sonStarting,calling parent now ... ");
    super.parentStart();
    System.out.println("I am " + this.getClass() + ". sonFinished");
}
}



public class SonSync2 extends ParentSync {

public void sonStart() {
    System.out.println("I am " + this.getClass() + ". sonStarting,calling parent now ... ");
    super.parentStart();
    System.out.println("I am " + this.getClass() + ". sonFinished");
}
}



public class SyncTest {
public static void main(String[] args) throws Exception {

    new Thread(() -> {
        new SonSync1().sonStart();
    }).start();

    new Thread(() -> {
        new SonSync2().sonStart();
    }).start();

    System.in.read();
}
}

result:

I am class com.common.interface18_design.whynotsync_onmethod.SonSync1. sonStarting,calling parent now ... 
I am class com.common.interface18_design.whynotsync_onmethod.SonSync2. sonStarting,calling parent now ... 
I am class com.common.interface18_design.whynotsync_onmethod.SonSync2 . parentStarting. now:2019-04-18 09:50:08
I am class com.common.interface18_design.whynotsync_onmethod.SonSync1 . parentStarting. now:2019-04-18 09:50:08
I am class com.common.interface18_design.whynotsync_onmethod.SonSync1 . parentFinished. now2019-04-18 09:50:38
I am class com.common.interface18_design.whynotsync_onmethod.SonSync1. sonFinished
I am class com.common.interface18_design.whynotsync_onmethod.SonSync2 . parentFinished. now2019-04-18 09:50:38
I am class com.common.interface18_design.whynotsync_onmethod.SonSync2. sonFinished

(sorry for using parent class as example)

from the result , we could know that the parent class lock is owned by every sub class, SonSync1 and SonSync2 object have different object lock . every lock is independency. so in this case , i think it is not dangerous using a synchronized in a parent class or a common interface . could anyone explain more on this ?

What is the reason why “synchronized” is not allowed in Java 8 interface methods?

Follow WeChat

Want to stay one step ahead of the latest teleworks?

相似问题

Platform

Support

Contact US