5 Golden Rules of widening, boxing & varargs

[Anand Shrivastava]

Dear Friends,

Here I found a lot of confusion with regard to autoboxing, varargs and widening. Therefore, with the help of all the ranchers and moderators I propose to compile the rules in this regard which we shall call the Golden Rules. My perception is as under, on which I invite you all for an exhaustive discussion and to incorporate any modifications, if required, so that it may help all the SCJP aspirants.

Rules :

1. Primitive Widening > Boxing > Varargs.
2. Widening and Boxing (WB) not allowed.
3. Boxing and Widening (BW) allowed.
4. While overloading, Widening + vararg and Boxing + vararg can only be used in a mutually exclusive manner i.e. not together.
5. Widening between wrapper classes not allowed

Now Let me illustrate each of them with instances :

    
Overloaded methods               Invoked by saying     Called method
doX(Integer i) & doX(long l)       doX(5)              long (by Rule 1)
doX(int...i) & doX(Integer i)      doX(5)              Integer (by Rule 1)
doX(Long l) & doX(int...i)         doX(5)              int...i (Rule 2 & 1)
doX(Long l) & doX(Integer...i)     doX(5)              Integer...i(R. 2&1)
doX(Object o) & doX(Long l)        doX(5)              Object o (Rule 2&3)
doX(Object o) & doX(int...i)       doX(5)              Object o (Rule 3&1)
doX(Object o) & doX(long l)        doX(5)              long l (Rule 3&1)
doX(long...l) & doX(Integer...i)   doX(5)              ambiguous (Rule 4)
doX(long...l) & doX(Integer i)     doX(5)              Integer (Rule 1)
doX(Long l)                        Integer i;          error (Rule 5)
                                   doX(i)
doX(Long l) & doX(long...l)        Integer i;          long...l(Rule 5 & 1)
                                   doX(i)

Dear friends, I am also a learner. So if i have made a mistake somewhere, please let me know. Moderators kindly help in this regard to consolidate the rules. Thanks in anticipation

[ July 28, 2008: Message edited by: Anand Shrivastava ]

[Milan Sutaria]

great work Anand .. Keep it up
just want to add little to your rule#4... widening+varArgs & Boxing+varargs are individually allowed (but not allowed in overloaded version of method)

[Anand Shrivastava]

Thanks Milan, I acknowledge that rule 4 should be happily worded as under:

Rule 4 - While overloading Widening + vararg and boxing + vararg are mutually exclusive of each other. Editing it accordingly.

Thanks.
[ July 28, 2008: Message edited by: Anand Shrivastava ]

[Jean-Luc Thirion]

Does that imply that the combination "boxing + Object Widening" is ok also for assignement ?

For instance, I think this will compile (I am quite sure):

 Object o = 5; 
Number n = 40; 

Sorry, I know I can test that but I am not at home right now.
[ October 22, 2008: Message edited by: Jean-Luc Thirion ]

[Bert Bates]

I didn't verify the technical details, but such a chart is a GREAT idea!

Also Anand, you can edit away those multiple edit notes on the bottom of your first post.

Bert

[Ankit Garg]

Great Anand

[Rajshekhar Paul]

Great post, Anand!

[Punit Singh]

Nice research Anand

One more example for (Boxing+Widening) is preferred over (Boxing+Varargs).

doX(Integer... i) doX(Object o) doX(5) (Object o) based of rule 1, 3.
[ December 18, 2008: Message edited by: punit singh ]

[Fabian Angy]

Yeah thanks Anand !

[Jayati Das]

This is surely a helpful note -Thanks !

[geeta vemula]

doX(Long l) & doX(long...l) Integer i; long...l(Rule 5 & 1) doX(i)

In this is it following the series as unboxing-widening-varargs?

Thanks,

Geeta Vemula

[Punit Singh]

yeah, it is following unboxing+widening+varargs.

[aparna shinde]

these rules are really helpful !!! Good work anand

[Abhi vijay]

  public class Devk{
          public static void tMeth(Integer... i){
               System.out.print("A");
          }
          public static void tMeth(int... i){
               System.out.print("B");
          }
          public static void main(String args[]){
               tMeth(Integer.valueOf("7"));
          }
     }  

Here I am getting compiler error???

[Ruben Soto]

Good catch. It seems there is an ambiguity in this case:

 
Overloaded methods                 Invoked by saying     Called method
doX(Integer... i) & doX(int...i)   doX(new Integer(1))   Ambiguous call

The call doX(1) is ambiguous as well. This looks like Rule 4 needs to be expanded, and a new rule similar to 4 added.
4. While overloading, Widening (or the identity conversion) + vararg and Boxing + vararg can only be used in a mutually exclusive manner i.e. not together.
4' While overloading, vararg and unboxing + vararg can only be used in a mutually exclusive manner i.e. not together.

[ December 20, 2008: Message edited by: Ruben Soto ]
[ December 20, 2008: Message edited by: Ruben Soto ]

[akash azal]

can anyone clear about line no 9 in anand's post

[Punit Singh]

akash azal wrote:can anyone clear about line no 9 in anand's post

doX(Object o) & doX(long l)        doX(5)              long l (Rule 31)

Here doX(5) will call doX(long l) that is rule 1, means int----> long that is primitive widening.

for doX(Object o), int ----> Integer boxing and then
Integer---->Object that is Widening. //rule 3

So here rule 1 preferred over rule 3.

[akash azal]

how it is processed for linr no8

[Punit Singh]

akash azal wrote:how it is processed for linr no8

line no 8 is Boxing + widening//from rule 3
preffered over
varargs //from rule 1

Var-args should be your last preference always.

You can do one thing, just remember rule 1 as
1. widening>boxing>varargs

so you can see boxing and widening comes before var-args so doX(Object) in line 8 called.
same way for line 9, widening comes before boxing, that's why widening is preferred over (widening+boxing), so doX(long) is called.

[Jose Francisco Ruiz Massieu]

well, i think that rule 4 it is not completly true due if we invoke line 10 with

byte b = 5;
doX(b);

it works even with overloading, (run doX(long... l) by rules 1 and 2), so in this case it is perfectly legal to have both Widening + vararg and Boxing + vararg together

what do you think???

maybe:

4) Widening + varargs , Boxing + vararg together
4.1) Without overloading Is allowed
4.2) With overloading
4.2.1) Without widening Not allowed

doX(int...i)
doX(Integer...i)

doX(5)              //ambiguous (Rule 4)

4.2.2) With widening Is allowed

doX(int...i)
doX(Integer...i)

byte b = 5;
doX(b);                  // ok, (run doX(int... i) by rules 1 and 2)
doX(b, b);              // ok, (run doX(int... i) by rules 1 and 2)
doX(b, 5);              // ok, (run doX(int... i) by rules 1 and 2)

[Ruben Soto]

Jose Francisco,

The reason why the cases that you show work is because you can't box a byte to an Integer, so the method that uses the boxing + varargs is not even considered. Let me know if that makes sense.

[Jose Francisco Ruiz Massieu]

yes I know, thats why it says (run doX(int... i) by rules 1 and 2), but what it's important (I think) is that, in this case both Widening + varargs , Boxing + vararg can be together, in the same class. So

While overloading, Widening + vararg and Boxing + vararg can only be used in a mutually exclusive manner i.e. not together

doesn't apply at all, do it???, what do you think???

[Punit Singh]

Jose Francisco Ruiz Massieu wrote:yes I know, thats why it says (run doX(int... i) by rules 1 and 2), but what it's important (I think) is that, in this case both Widening + varargs , Boxing + vararg can be together, in the same class. So

While overloading, Widening + vararg and Boxing + vararg can only be used in a mutually exclusive manner i.e. not together

doesn't apply at all, do it???, what do you think???

byte---->Integer... is not boxing+varargs.
it is
byte---->int //widening then int --->Integer //boxing then Integer... //var-args
so
byte--->Integer... is widening+boxing+var-args

So above rule is right, you are misjudging it.

[Natalie Ap]

Hi all,

Can you please explain to me why the following is gives a compile time 'ambiguous method' error and which rule covers the following possibility:

   
public class A2 {

	public static void main(String[] args) {
		A2 a = new A2();
		a.getNum(4);
	}
	
	public void getNum(int... n) {
		System.out.println("A: getNum(int... n)");
	}
	
	public void getNum(Integer... n) {
		System.out.println("D: getNum(Integer... n)");
	}
}

[Punit Singh]

a.getNum(4);

here 4 is int

so for getNum(int... n)
it is only varargs.
int ---> int...

for getNum(Integer... n)
it is boxing + varargs.
int--->Integer--->varargs
Here rule no 4 applies, but we need to make rule no 4 more clear to cover this behavior also.

[Jia Tan]

doX(int... i) & doX(long... l), call doX(5) get ambiguous too

[Jia Tan]

I guess the rules is: you cannot have more than one primitive widening + var-args (WV) overloaded methods, and you cannot have var-args(V) co-exists with other primitive widening + var-args (WV) overloaded methods

the following are all resulted in ambiguous by calling doX(5):

doX(int... i) & doX(long... l)
doX(int... i) & doX(float...f)
doX(long... l) & doX(double... d)

[Ruben Soto]

Punit Singh wrote:a.getNum(4);

here 4 is int

so for getNum(int... n)
it is only varargs.
int ---> int...

for getNum(Integer... n)
it is boxing + varargs.
int--->Integer--->varargs
Here rule no 4 applies, but we need to make rule no 4 more clear to cover this behavior also.

Read my post from 12/20/08 in this thread:

The call doX(1) is ambiguous as well. This looks like Rule 4 needs to be expanded, and a new rule similar to 4 added.
4. While overloading, Widening (or the identity conversion) + vararg and Boxing + vararg can only be used in a mutually exclusive manner i.e. not together.
4' While overloading, vararg and unboxing + vararg can only be used in a mutually exclusive manner i.e. not together.

[Jia Tan]

I would suggest the 4th rule being changed to:

4. Combination of the following overloaded methods will result in compiler error:
a. Var-args (V) of primitive type same as argument primitive type
b. Widening + Var-args (WV)
c. Boxing/Unboxing + Var-args (BV)

Example:

doX(int… i) & doX(Integer… i)	doX(5)	4a & 4c
doX(int… i) & doX(Integer… i)	doX(new integer(1))	4a & 4c
doX(int… i) & doX(long… i)	doX(5)	4a & 4b
doX(long… l) & doX(float… f)	doX(5)	4b & 4b
doX(long… l) & doX(Integer… i)	doX(5)	4b & 4c

[Aakash Goel]

what a amazing post!

just wanted to clarify one thing though:

doX(Long l) & doX(long...l)        Integer i;          long...l(Rule 5 & 1)  
                                   doX(i)  

The conversion choosing long... here is 'Un'boxing and then Widening. Right?

[Punit Singh]

Aakash Goel wrote:what a amazing post!

just wanted to clarify one thing though:

doX(Long l) & doX(long...l)        Integer i;          long...l(Rule 5 & 1)  
                                   doX(i)  

The conversion choosing long... here is 'Un'boxing and then Widening. Right?

Yes you are right.

[Sourabhkapoor Kapoor]

public class xt{

public static void main(String argv[]){
int i=2;
doNumber(i);
}

public static void doNumber(int... dn){ //line 4
System.out.print("c");
}


public static void doNumber(Integer... dn){ //line 5
System.out.print("b");
}

public static void doNumber(Object dn){ line 6
System.out.print("e");
}

}


Answer: e
Modified rule 4 needs to change i think..Please correct me If i am wrong

line 4:rule 4(a)
line 5 :rule boxing var-args 4(c)
Still can compile no ambiguity cuz line 6 will help the compiler to resolve ambiguity.

[Ankit Garg]

Sourabhkapoorscjp Kapoor please check your private messages for some administrative matter...

[zhong chen]

Some addition to this:

public class WBV {

    public static void main(String[] args) {
        new WBV().getNum(5);
    }
/*
    public void getNum(int... n) {
        System.out.println("int...");
    }
*/
/*
    public void getNum(int[] n) {  // won't compile if coexist with int... or Integer[],
                                   // but Integer... is ok and preceeds int[]
        System.out.print("int[]");
    }
*/
/*
    public void getNum(Integer... n) {
        System.out.println("Integer...");
    }
*/

    public void getNum(Integer[] n) {  // won't compile if coexist with Integer... or int[]
                                       // but int... is ok and preceeds Integer[]
        System.out.println("Integer[]");
    }


    public void getNum(Object n) { // this version beats int..., int[], Integer... and Integer[]
        System.out.println("Object");
    }

}

[zhong chen]

I found this in Examlab mockup 1.

Why this code compiles without error? Aren't these static methods overloaded?

public class Gened{

     public static void main(String argv[]){
          doNumber(2);
          doNumber(2L);
          doNumber(new int[]{1,2,3,4});
          doNumber(new Integer(2));
          doNumber(new Integer[2]);
     }

     public static void doNumber(int dn){
          System.out.print("a");
     }

     public static void doNumber(int... dn){
          System.out.print("b");
     }

     public static void doNumber(Integer dn){
          System.out.print("c");
     }

     public static void doNumber(Integer... dn){
          System.out.print("d");
     }

     public static void doNumber(Object dn){
          System.out.print("e");
     }

}

[Atwal Usha]

Yes, they are overloaded but why do you think the code should result in compile time error.

[Ankur kothari]

doX(Object o) & doX(int...i)

doX(5);

why is doX(Object o)preferred over the other one? one is boxing and widening and other is varargs......

[Rajiv Chopra]

Anand Shrivastava wrote:Dear Friends,

Here I found a lot of confusion with regard to autoboxing, varargs and widening. Therefore, with the help of all the ranchers and moderators I propose to compile the rules in this regard which we shall call the Golden Rules. My perception is as under, on which I invite you all for an exhaustive discussion and to incorporate any modifications, if required, so that it may help all the SCJP aspirants.

Rules :

1. Primitive Widening > Boxing > Varargs.
2. Widening and Boxing (WB) not allowed.
3. Boxing and Widening (BW) allowed.
4. While overloading, Widening + vararg and Boxing + vararg can only be used in a mutually exclusive manner i.e. not together.
5. Widening between wrapper classes not allowed

Now Let me illustrate each of them with instances :

    
Overloaded methods               Invoked by saying     Called method
doX(Integer i) & doX(long l)       doX(5)              long (by Rule 1)
doX(int...i) & doX(Integer i)      doX(5)              Integer (by Rule 1)
doX(Long l) & doX(int...i)         doX(5)              int...i (Rule 2 & 1)
doX(Long l) & doX(Integer...i)     doX(5)              Integer...i(R. 2&1)
doX(Object o) & doX(Long l)        doX(5)              Object o (Rule 2&3)
doX(Object o) & doX(int...i)       doX(5)              Object o (Rule 3&1)
doX(Object o) & doX(long l)        doX(5)              long l (Rule 3&1)
doX(long...l) & doX(Integer...i)   doX(5)              ambiguous (Rule 4)
doX(long...l) & doX(Integer i)     doX(5)              Integer (Rule 1)
doX(Long l)                        Integer i;          error (Rule 5)
                                   doX(i)
doX(Long l) & doX(long...l)        Integer i;          long...l(Rule 5 & 1)
                                   doX(i)

Dear friends, I am also a learner. So if i have made a mistake somewhere, please let me know. Moderators kindly help in this regard to consolidate the rules. Thanks in anticipation

[ July 28, 2008: Message edited by: Anand Shrivastava ]

Great Work done!!

[Ben Power]

Thanks, as many others said, this has been really helpful as I attempt to organize my thoughts for the SCJP (tomorrow!).

Ben

[swaraj gupta]

hi anand!
Better include one more rule to this post...
" A var-arg method will only be chosen if and only if no other non-var-arg method is possible. "