statics in inner classes

[Roedy Green]

I think I almost have it understanding inner, nested, top-level,
public etc classes and the equivalent enums.

What I don't understand is the restriction on no static variables in
inner nested non-static instance classes. Why?

What puzzles me is it seems ok to derive inner classes from a class
that has some static variables. It only seems upset about new statics.

I finally figured out why you can only access final enclosing local
variables in your anonymous inner classes.

I am gradually putting it together and will post in the next day or
two at http://mindprod.com/jgloss/innerclasses.html

The other puzzle is why enums constructors cannot access static
fields.



--
Bush crime family lost/embezzled $3 trillion from Pentagon.
Complicit Bush-friendly media keeps mum. Rumsfeld confesses on video.
http://www.infowars.com/articles/us/mckinney_grills_rumsfeld.htm

Canadian Mind Products, Roedy Green.
See http://mindprod.com/iraq.html photos of Bush's war crimes

 

[Chris Smith]

What I don't understand is the restriction on no static variables in
inner nested non-static instance classes. Why?

It appears to be just an arbitrary rule. There is certainly an obvious
behavior that could have been assigned to it, but Sun decided against
it, for whatever reason. I wouldn't hunt for clues to anything
profound.

What puzzles me is it seems ok to derive inner classes from a class
that has some static variables. It only seems upset about new statics.

Since static members are not inherited, that's really not a problem.

The other puzzle is why enums constructors cannot access static
fields.

I'm unaware of such a restriction. At least, the following code
compiles with Eclipse 3.1RC3.

public enum Test
{
A(42), B(7)

;

static int a = 5;
public int aa;

Test(int val)
{
aa = val * a;
}
}

--
www.designacourse.com
The Easiest Way To Train Anyone... Anywhere.

Chris Smith - Lead Software Developer/Technical Trainer
MindIQ Corporation

 

[Lasse Reichstein Nielsen]

I'm unaware of such a restriction. At least, the following code
compiles with Eclipse 3.1RC3.

But did you test it? Both Test.A.aa and Test.B.aa has the value 0.

public enum Test
{
A(42), B(7)

This corresponds to the declarations:

public static final A = new Test(42);
public static final B = new Test(42);

however ...

static int a = 5;

the static variable "a" only gets its value later, so at the time of
creation of A and B, it has the default value.

To confuze it even more, had you made "a" final, it would have been a
constant, and would be inlined without a problem.


I'm not sure what Roedy Green's problem is, excatly. Enum constructors
can access static fields just fine, they just have to consider when
they are doing it ... which is before static initialiation reaches
the assignments to static fields.

/L

 

[Chris Smith]

But did you test it? Both Test.A.aa and Test.B.aa has the value 0.

Ah. I missed that; another ugly implementation artifact. Thanks for
pointing it out!

Unfortunately, these kinds of things are building up, making Java into a
language that has to be learned from version 1.0 forward, instead of
making sense statically in its current form.

--
www.designacourse.com
The Easiest Way To Train Anyone... Anywhere.

Chris Smith - Lead Software Developer/Technical Trainer
MindIQ Corporation

 

[Roedy Green]

It appears to be just an arbitrary rule. There is certainly an obvious
behavior that could have been assigned to it, but Sun decided against
it, for whatever reason. I wouldn't hunt for clues to anything
profound.

I've been chewing on it. One thing that hit me is it may speed garbage
collection. Classes I think are harder to scavenge that object. You
don't want millions of little throwaway class objects littering the
pool. By refusing static members, you can dispense with a separate
class object for each anonymous class.

--
Bush crime family lost/embezzled $3 trillion from Pentagon.
Complicit Bush-friendly media keeps mum. Rumsfeld confesses on video.
http://www.infowars.com/articles/us/mckinney_grills_rumsfeld.htm

Canadian Mind Products, Roedy Green.
See http://mindprod.com/iraq.html photos of Bush's war crimes

 

[Chris Smith]

I've been chewing on it. One thing that hit me is it may speed garbage
collection. Classes I think are harder to scavenge that object. You
don't want millions of little throwaway class objects littering the
pool. By refusing static members, you can dispense with a separate
class object for each anonymous class.

No, you definitely need a separate Class object (and all the other
underlying class stuff) for each anonymous inner class. Anonymous
classes are treated in exactly the same way as any other class, and the
JVM generally doesn't even pay attention to the difference.

I don't think it would matter anyway. Class GC is done per class
loader, so unless you added additional class loaders, there isn't any
additional work to be done in garbage collection. An anonymous class is
ALWAYS loaded by the same classloader as its containing class (or
conceivably by its ancestor, but that requires strange contortions).

--
www.designacourse.com
The Easiest Way To Train Anyone... Anywhere.

Chris Smith - Lead Software Developer/Technical Trainer
MindIQ Corporation

 

[Raymond DeCampo]

It appears to be just an arbitrary rule. There is certainly an obvious
behavior that could have been assigned to it, but Sun decided against
it, for whatever reason. I wouldn't hunt for clues to anything
profound.

According to the specification, inner classes may have static fields
that are compile time constants. So there's one of those clues you
suggested we not hunt for.

Ray

 

[Thomas G. Marshall]

It appears to be just an arbitrary rule. There is certainly an obvious
behavior that could have been assigned to it, but Sun decided against
it, for whatever reason. I wouldn't hunt for clues to anything
profound.

Me neither, but I would offer a possibility. Well actually, more
accurately, they are stray thoughts forming in me head.

Throughout the documentation, sun seems to regard inner classes as members
of enclosing instances. Basically, even though it looks like a simple class
namespace issue, it actually regards the inner class as a member, with rules
similar to other members. I of course am very likely to have this wrong,
but it seems this way, perhaps I'll go dig up what I'm referring to today.

Take for example, one of the goofiest things in java:

X.Y xy = new X().new Y();

For when Y is a non-static class it requires the context of the X /object/
to instantiate it. Where as when Y is a static inner:

X.Y xy = new X.Y();

Works just fine. So going back to:

new X().new Y();

The inner class Y is (perhaps?) not able to have a static member because
there is no static class for it to belong to? That is, X, the outer class
is (perhaps?) by definnition an implied static. The inner however, is not,
and therefore there is no class per se to actually contain the statics.

This is far from a connecting of the dots, and is likely wrong, but it's the
best I can do right now.

 

[Roedy Green]

The inner however, is not,
and therefore there is no class per se to actually contain the statics.

If there were statics for inner classes what are they -- one class per
set of objects tied to mother? or to me more logical one class per set
of such objects ever created anywhere.


What happens when your inner class extends a class that already has
some statics? Does it stop you?

--
Bush crime family lost/embezzled $3 trillion from Pentagon.
Complicit Bush-friendly media keeps mum. Rumsfeld confesses on video.
http://www.infowars.com/articles/us/mckinney_grills_rumsfeld.htm

Canadian Mind Products, Roedy Green.
See http://mindprod.com/iraq.html photos of Bush's war crimes

 

[Lasse Reichstein Nielsen]

If there were statics for inner classes what are they -- one class per
set of objects tied to mother?

That would be consistent. A member class belongs to the enclosing
instance, so there really is one class per enclosing instance. It's
just not implemented like that under the hood, but that is the idea.

What happens when your inner class extends a class that already has
some statics? Does it stop you?

Extending a class that contains static members will allow you to
access the members through the extending type as well. But that's just
being nice. They are still resolved, at compile time, to the static
members of the class declaring them, and the compiled bytecode need
not refer to the extending class at all.

/L

 

[Thomas G. Marshall]

Roedy Green coughed up:

On Sat, 09 Jul 2005 16:02:48 GMT, "Thomas G. Marshall"



If there were statics for inner classes what are they -- one class per
set of objects tied to mother?

I'm sorry, I don't understand this statement.

or to me more logical one class per set
of such objects ever created anywhere.

Nor this one. Can you rephrase?

What happens when your inner class extends a class that already has
some statics? Does it stop you?

From what, adding further statics? You cannot. From accessing the existing
ones in the super class? You can, but only because java allows you to
access superclass statics from extended classes (and I'm not particularly
glad about that, since it seems a break in what statics should be (belonging
to a class, not a class hierarchy), but thems the facts, and it doesn't hurt
the discussion any).

If you have these classes:

A
X
X.Y

And this in A:

public static int num;

and this relationship

A <|-- X.Y

Then accessing X.Y.num is always possible. I don't like it much, but it's
true of any subclass of a class holding a static. For yucks, this would be
the test app:

class A
{
static int num = 5;
}

public class StaticTest
{
class B extends A
{
static int cant = 6; // compile error
}

public static void main(String[] args)
{
System.out.println("Can access B.num---> " + B.num);
}
}

 

[Roedy Green]

I'm sorry, I don't understand this statement.

Does conceptually there a different class object allocated per
instance of the mother object or is there one shared by all.

Perhaps the restriction on no statics was to avoid deciding what it
means.


In other would statics on an inner class be for communication between
inner class members of a gives object only, or between inner class
members of all inner class objects no matter what object they are
attached to.


By not allowing statics they avoid the question.

--
Bush crime family lost/embezzled $3 trillion from Pentagon.
Complicit Bush-friendly media keeps mum. Rumsfeld confesses on video.
http://www.infowars.com/articles/us/mckinney_grills_rumsfeld.htm

Canadian Mind Products, Roedy Green.
See http://mindprod.com/iraq.html photos of Bush's war crimes

 

[Thomas Schodt]

I think I almost have it understanding inner, nested, top-level,
public etc classes and the equivalent enums.

I am gradually putting it together and will post in the next day or
two at http://mindprod.com/jgloss/innerclasses.html

As anonymous classes do not have constructors
you might want to mention something about
instance initializer blocks here.

http://mindprod.com/jgloss/anonymousclasses.html

 

[Thomas G. Marshall]

Does conceptually there a different class object allocated per
instance of the mother object or is there one shared by all.

Perhaps the restriction on no statics was to avoid deciding what it
means.


In other would statics on an inner class be for communication between
inner class members of a gives object only, or between inner class
members of all inner class objects no matter what object they are
attached to.


By not allowing statics they avoid the question.

Not to avoid the question per se, but yes, you're on the right track. Since
they are part of enclosing *instances*, having a static belong to it is a
no-no.

(Please, no one here who is junior confuse this with accessing a static from
an object).

 

[Thomas G. Marshall]

Roedy Green coughed up:

On Mon, 11 Jul 2005 01:21:00 GMT, "Thomas G. Marshall"


Does conceptually there a different class object allocated per
instance of the mother object or is there one shared by all.

Perhaps the restriction on no statics was to avoid deciding what it
means.


In other would statics on an inner class be for communication between
inner class members of a gives object only, or between inner class
members of all inner class objects no matter what object they are
attached to.


By not allowing statics they avoid the question.

Roedy, it's of course up to you, but I would think that you might want an
example at least *similar* to my

X.Y xy = new X().new Y(); // non-static inner Y

vs.

X.Y xy = new X.Y(); // static inner Y

to eventually make it to your jgloss. I believe that the top formalism for
non-static inner classes is flat-out unknown to a vast majority of java
engineers.

It helps to explain why the following is not allowed (and it confuses nearly
every beginner):

public class Static3
{
// NON-STATIC inner class
class Inner
{
}

public static void main(String[] args)
{
// following disalllowed:
// "non-static variable this cannot be
// referenced from a static context"
Inner inner = new Inner();
}
}

$ com Static3.java
javac 1.5.0-beta2
Static3.java:17: non-static variable this cannot be referenced
from a static context
Inner inner = new Inner();
^
1 error
$

It is complaining about the variable "this". This is implied:

Inner inner = this.new Inner();

....because Inner is not static. This is the same error message when any
member variable is accessed from a static context:

public class Static4
{
int num = 5;

public static void main(String[] args)
{
int x = num;
}
}


$ com Static4.java
javac 1.5.0-beta2
Static4.java:11: non-static variable num cannot be referenced
from a static context
int x = num;
^
1 error
$


--
Having a dog that is a purebred does not qualify it for breeding. Dogs
need to have several generations of clearances for various illnesses
before being bred. If you are breeding dogs without taking care as to
the genetic quality of the dog (again, being purebred is *not* enough),
you are what is known as a "backyard breeder" and are part of the
problem. Most of the congenital problems of present day dogs are
traceable directly to backyard breeding. Spay or neuter your pet
responsibly, and don't just think that you're somehow the exception and
can breed a dog without taking the care described.