[SOLVED] How to convert negative integer to byte array?

Basel · 10-26-2010, 05:14 PM

I need to convert an integer to a byte array of size 2 and vice versa. The code shown below works well for positive values but not for negative values. Also, using an array of size four makes the conversion works. However, I am limited to an array of size 2.

Code:

#include <stdio.h>

main() {
	int input = -1;
	unsigned char buff[2];
	buff[0] = input & 0xFF;
	buff[1] = (input & 0xFF00) >> 8;
//	buff[2] = (input & 0xFF0000) >> 16;
//	buff[3] = (input & 0xFF000000) >> 24;
	int value = 0;
	value  = buff[0];
	value += buff[1] << 8;
//	value += buff[2] << 16;
//	value += buff[3] << 24;
	printf("value[%d]\n", value);
}

Code:

// output with array of size 2
value[65535]
// output with array of size 4
value[-1]

dugan · 10-26-2010, 05:17 PM

Uhm, point out to your teacher that you're not using 16-bit machines?

On a modern desktop PC, an byte is 8 bits and an int is 32 bits. That means you cannot convert an int to two bytes without throwing away half the data in it. For a small positive number, you're throwing away leading zeros, so you get the same value. Negative numbers, being represented in twos complement, use all 32 bits. That means you cannot convert them to two-byte values without changing the number.

On a 16-bit machine (read: a 286 or 386SX), an int is 16-bits, and therefore can be represented with two bytes.

BTW, I'm currently reading Code, by Charles Petzold (http://www.amazon.ca/Code-Language-C...dp/0735611319/), which is a very good overview of issues like this.

Basel · 10-26-2010, 05:36 PM

Thanks dugan for your insightful reply.
EDIT: thanks for recommending the book

As a work around for the problem, using "short int" would do the job. However, a "short int" may be larger than 2 bytes.

ntubski · 10-26-2010, 10:22 PM

You could sign-extend the number when converting to int:

Code:

	int value = 0;
	value  = buff[0];
	value += buff[1] << 8;

        if (buff[1] & 0x80) {
            // for bonus points use a value that doesn't assume int size.
            value |= 0xFFFF0000;
        }

Values that really don't fit into 2 bytes will still be changed, but it works for negative numbers greater than -2^16 + 1.

dugan · 10-27-2010, 11:09 AM

You could also convert the negative number to positive before converting it to a byte array, then convert the contents of the byte array to its twos complement representation. That would work for as many negative numbers as positive ones.

theNbomr · 10-27-2010, 12:35 PM

Does this help?

Code:

#include <stdio.h>
#include <stdlib.h>

union intToChar {
    int   integer;
    char  bytes[4];
};

int	main( int argc, char * argv[] ){

union intToChar i2c;

    i2c.integer = 0x12345678;
    printf( "%02X %02X %02X %02X\n", (unsigned int)i2c.bytes[0],
    				     (unsigned int)i2c.bytes[1],
				     (unsigned int)i2c.bytes[2],
				     (unsigned int)i2c.bytes[3] );
    exit(0);
}

--- rod.