Question48 of Valentin Crettaz's mock exam

[Claire Yang]

Code:

public class Question48{
public static void main(String[] args) {
int i = 4*6-3/2<<2*5>>>1%2-4^3;
System.out.println(i);
}
}

Exam explanation: the expression should be ((((4*6)-(3/2))<<(2*5))>>>((1%2-4))^3.
My question is that since 1%2-4=-3, can shift operator take negative integer as right operand? I can't find any comment/example like the code in many books.

I tried the following 2 code:

public static void main(String[] args) {
    int j=3;
    for (int n=-1; n>-4; n--) {
       j=j>>>n;
       System.out.print(j);
    }
}

I got output 0 0 0.

public static void main(String[] args) {
    int j=-3;
        for (int n=-1; n>-4; n--) {
            j=j>>>n;
            System.out.print(j);
        }
}

I got output 1 0 0.

Could somebody tell me why?
Thanks.

[Barry Gaunt]

You have j=j>>>n. Do you really mean that?
Don't worry about the sign of the righthand operand.
Just take the low-order 5 (int lefthand operand) or 6 (long lefthand operand) bits of the righthand operand and shift by that number of bits (taken as a positive number).
For example:
-1 >>> -1
The lefthand operator is an int so we take the
low-order 5 bits of the -1 on the righthand side. We get 11111, that's 31 as an unsigned number. Shift the lefthand side to the right by 31 bits, filling on left with 0s, gives
just 1 bit remaining in the lowest order position.
So -1 >>> -1 is 1.
All other cases are similar.
-Barry
[ September 23, 2002: Message edited by: Barry Gaunt ]
[ September 23, 2002: Message edited by: Barry Gaunt ]

[Shishio San]

Hi
The same question was posted before. I kept the answer to it as a reference. It might help:

Using % to reduce the right-hand side (RHS) operand is fine for positive operands. However, when you have negative RHS operands, you should go by the low-order bits (5 for int and 6 for long). That is,
x >> -4; // reduced to x >> 28 when x is int
x >> -4; // reduced to x >> 60 when x is long
What Joshua Bloch (author of Effective Java and the Collections API) pointed out to me a while back was that reduction is done by masking the RHS operand with 0x1F for int shifts and 0x3F for long shifts, not by using %.
For int:
-4 = 1111 11.. 1111 1100 (32 bits)
& 0x1F = & 0000 00.. 0001 1111
-----------------------------------------
28 = 0000 00.. 0001 1100
For long:
-4 = 1111 11.. 1111 1100 (64 bits)
& 0x3F = & 0000 00.. 0011 1111
-----------------------------------------
60 = 0000 00.. 0011 1100
shifting operands must be a primitive integer type, which of course includes long and excludes float and double.

And the answer to your original post is 3 by the way.
I hope this helps.

[Claire Yang]

I understand this question now, thanks greatly for your replies!

[Valentin Crettaz]

chia,
Welcome to Javaranch
We'd like you to read the Javaranch Naming Policy and change your publicly displayed name (change it here) to comply with our unique rule. Thank you.
PS: Quote from the naming policy:

For your publicly displayed name, use a first name, a space, and a last name. Obviously fictitious names or improperly formatted names may be locked out.