I have noticed a difference in report style between the
following error messages:
interface A {}
interface A_ {}
class B implements A, A_ {}
class C implements A, A_ {}
public class Main
{ static void f( final java.lang.String s ){}
public static void main( java.lang.String[] args )
{ f( Math.random() < 0.5 ? new B() : new C() ); }}
Main.java:49: f(java.lang.String) in Main cannot be applied to (java.lang.Object&A&A_)
{ f( Math.random() < 0.5 ? new B() : new C() ); }}
^
Here, one can notice that "java.lang.Object" is being
mentioned. If "A_" is removed, one might expect
"java.lang.Object&A" as type, but gets only "A":
interface A {}
class B implements A {}
class C implements A {}
public class Main
{ static void f( final java.lang.String s ){}
public static void main( java.lang.String[] args )
{ f( Math.random() < 0.5 ? new B() : new C() ); }}
Main.java:45: f(java.lang.String) in Main cannot be applied to (A)
{ f( Math.random() < 0.5 ? new B() : new C() ); }}
^
Why is "java.lang.Object" dropped now? Or, why was it
mentioned in the first case?
And why can I not declare variables of such "&"-types?
interface A {}
interface B {}
public class Main
{ static void f( final java.lang.String s ){}
public static void main( java.lang.String[] args )
{ A&B a; }}
How are these static types (in parentheses and with "&")
called, which are mentioned in error message, but can not be
used in the source code?
OK, if this was needed, I could write
interface C implements A, B {} (...) C a;
(The compiler used was from a recent mustang (1.6) build.)
Thomas Hawtin - 28 Nov 2005 13:44 GMT
> I have noticed a difference in report style between the
> following error messages:
[quoted text clipped - 15 lines]
> mentioned. If "A_" is removed, one might expect
> "java.lang.Object&A" as type, but gets only "A":
Presumably so as not to give prominence to one interface over another.
The first type mentioned in an intersection type is the only one that
can be a class and is used for erasure.
> interface A {}
> class B implements A {}
[quoted text clipped - 12 lines]
>
> And why can I not declare variables of such "&"-types?
You can declare a generic parameter using &. I guess there wasn't
sufficient reason to allow it for variable declarations. I don't know
off hand what kind of further mess to the grammar it would make.
> interface A {}
> interface B {}
[quoted text clipped - 11 lines]
>
> interface C implements A, B {} (...) C a;
You can write:
<T extends A&B> void fn(T t) {
...
}
(Which will produce a method with erasure void fn(A).)
Tom Hawtin

Signature
Unemployed English Java programmer
http://jroller.com/page/tackline/
Roedy Green - 28 Nov 2005 17:08 GMT
> { A&B a; }}
The JLS
http://java.sun.com/docs/books/jls/second_edition/html/interfaces.doc.html#35470
suggests that interfaces can extend multiple interfaces, just as
classes can implement multiple interfaces.
So you should be able to define an interface C, that extends both A
and B and gives the effect you want without new syntax.

Signature
Canadian Mind Products, Roedy Green.
http://mindprod.com Java custom programming, consulting and coaching.
Ingo R. Homann - 29 Nov 2005 08:10 GMT
Hi,
>>{ A&B a; }}
>
[quoted text clipped - 5 lines]
> So you should be able to define an interface C, that extends both A
> and B and gives the effect you want without new syntax.
But that does not do what Stefan asked for - the following will not work:
interface A {}
interface B {}
class X implements A, B {}
interface C extends A, B {}
C c=new X(); // error
Of course I'm assuming here, that I cannot change the code in the first
three lines (e.g. 'Comparable', 'Serializable', 'String' or something
like that).
Ciao,
Ingo
Roedy Green - 29 Nov 2005 11:15 GMT
On Tue, 29 Nov 2005 09:10:02 +0100, "Ingo R. Homann"
<ihomann_spam@web.de> wrote, quoted or indirectly quoted someone who
said :
>class X implements A, B {}
>interface C extends A, B {}
>
>C c=new X(); // error
so he would have to write
class X implements C
or
class X implements A,B,C

Signature
Canadian Mind Products, Roedy Green.
http://mindprod.com Java custom programming, consulting and coaching.
Ingo R. Homann - 29 Nov 2005 14:05 GMT
Hi Roedy,
As I said before, consider:
A=Comparable
B=Serializable
X=String
> so he would have to write
> class X implements C
> or
> class X implements A,B,C
Unfortunately, you can not do so.
Ciao,
Ingo
P.Hill - 29 Nov 2005 04:26 GMT
If you'd kindly tell us what you are trying to implement
in Java or discover about the language we might better be
able to help you better.
[...]
> Here, one can notice that "java.lang.Object" is being
> mentioned. If "A_" is removed, one might expect
> "java.lang.Object&A" as type, but gets only "A":
In one case A is a common base type which the compiler can use to
identify the type of object returned by your cryptic expression.
The Object&A&_A is a bit of odd notation that the compiler
used to identify the type of the expression in order to say it
didn't have any such method that took such a type, it means
AN object which also implements A and _A. It could have said
A&_A and it might in the next release, I doubt such cryptic notation is
part of any standard.
> Why is "java.lang.Object" dropped now? Or, why was it
> mentioned in the first case?
Dropped? It doesn't need to mention it. All types derive from
java.lang.Object.
> And why can I not declare variables of such "&"-types?
Why would you want to? (see note at top and bottom).
> How are these static types (in parentheses and with "&")
> called, which are mentioned in error message, but can not be
> used in the source code?
> interface C implements A, B {} (...) C a;
Yes, if you want to invent a type you ought to give it a name.
Why don't you want to give it a name?
There is use of & in the generic notation (as mentioned in another
reply), but without further information about YOUR goal, I can't
guess what you are trying to accomplish, so I can't help accomplish it,
but I'd bet you are trying to program in another language when you
should try to write in Java.
-Paul
Stefan Ram - 29 Nov 2005 10:14 GMT
>If you'd kindly tell us what you are trying to implement
>in Java or discover about the language we might better be
>able to help you better.
What I actually would like to do is to write an expression
that contains a cast of a subexpression to a base class of its
class so that this cast is needed. (I.e., compilation should
fail without the cast, or at least the value of the expression
should change without the cast) This expression should not
require additional declarations (of methods or classes).
An example in Java 1.4 would be
Math.random() < 0.5 ?( java.lang.Object )System.in : System.out
In Java 1.4, the expression will produce a compile-time error
if the cast operator "( java.lang.Object )" is removed.
But in Java 1.5 the cast is not needed anymore.
The observation you responded to was a side-effect in this
search.