Synchronization and interfaces

[WarnerJan Veldhuis]

If my interface says:

public void readRecord();

does my class comply to the interface's method if I say:

public synchronized void readRecord() {
...
}

I've tried it, and it *does* compile, but does my class really implement the right interface-method?

[David Weitzman]

Interfaces cannot require or disallow use of the synchronized modifier in implementing methods. There are many ways to manage synchronization, and that should be (must be) an implementation detail.
The only place synchronization belongs in an interface is when your Javadocs describe the thread safety properties of a class or method, which is obviously not machine enforcable*.

* (Footnote) Not enforcable in Java and probably every other language with current technology, that is.

[Mr. C Lamont Gilbert]

Originally posted by David Weitzman:

The only place synchronization belongs in an interface is when your Javadocs describe the thread safety properties of a class or method, which is obviously not machine enforcable*.

* (Footnote) Not enforcable in Java and probably every other language with current technology, that is.

Java does not allow the synchronized modifier in an interface. And what do you mean not machine enforceable!? Java's behavior is defined by the JLS, not the machine it is running on.

[David Weitzman]

And what do you mean not machine enforceable!? Java's behavior is defined by the JLS, not the machine it is running on.
I'm not appealing to the specs here, but rather common sense. The bytecode verifier machine-enforces many rules related to interfaces (ensuring that return types are correct, all methods are there, etc.). Code that doesn't meet those specs won't be loaded and run. There is no reason why a language can't also have interfaces that specify, say, "This method takes a list of integers and returns a list of those integers sorted" in bytecode such that the bytecode verifiers for those languages won't load code that doesn't properly sort and terminate (if you're interested in stuff like that, look up Proof Carrying Code). Java is not such a language and surely never will be for very practical reasons, but not necessarily because of technical reasons.
But since Java obviously lacks those capabilities, it becomes clear that nothing one puts in an interface relating to synchronization can have the same useful and enforced meaning that "returnType methodName(params)" does. It would be like having a modifier "srt" which specifies that any data returned by the function must be sorted. Putting things like that in an interface which aren't meant to be enforced and will add no new capabilities to a language while requiring the programmer to type more are simply a waste of space.
This was the more general reasoning behind my statement above. In this case it would also be sufficient to show that in particular, having a "synchronized" modifier in interfaces would impose arbitrary restrictions that interfere with implementors instead of helping. That becomes apparent when you're familiar with the different levels of synchronization used in the wild, but I don't know how well I can explain it here. I'll just point to LinkedQueue from Doug Lea as an example of a class which would be harmed by method-level synchronization and call it a day.

[Mr. C Lamont Gilbert]

I agree. Don't know why I took maching to mean computer as opposed to JVM which is what you seem to have meant. Either way the neither investigates ones claims about his code.
[ March 09, 2004: Message edited by: CL Gilbert ]

[Avi Abrami]

Jan,
As far as I know, the synchronized keyword is not part of the method declaration, therefore the compiler considers the implementation method declaration (with the "synchronized" keyword) to match the interface declaration (without the "synchronized" keyword).
I understand from this, that it is not required to use "synchronized" in interface method declarations -- in fact, I would suggest that it is better not to use "synchronized" at all in any interface method declarations.
Good Luck,
Avi.

[WarnerJan Veldhuis]

Thanks all for your answers... Conclusion: having synchronized in the method declaration does not break the interface contract. That's what I wanted to know... Thanks

[Mr. C Lamont Gilbert]

yes that is correct. thread safety is not part of the contract. and synchronize does not tell you much at all about a method.
Also you can throw exceptions in an interface, but the method does not have to throw the exception if it does not really throw it.