[SOLVED] [c] shift bytes in an array

schneidz · 08-01-2019, 02:32 PM

hi, is there a smarter way to do this ?:

Code:

 c = fgetc(fstream);
 while(c != EOF)
 {
  carr[0] = carr[1]; carr[1] = carr[2]; carr[2] = carr[3]; carr[3] = carr[4]; carr[4] = carr[5]; carr[5] = carr[6]; carr[6] = carr[7]; carr[7] = carr[8]; carr[8] = c;
  printf("%c", carr[0]);
  if(strcmp(carr, "012345678") == 0)
   sprintf(carr, "hispanola");
}

rtmistler · 08-01-2019, 02:43 PM

Without libraries? yes and no. Yes, you can use pointers, but no it's all manually moved byte by byte, just with a loop or something.

With libraries? memmove(3) - Recommend you read the manual page thoroughly. It DOES allow for overlapping regions, but one has to be careful about src/dst and that you're not obliterating src as you go along.

Edit: And the manual page says that it does move bytes into a temporary array, so if you're moving left to right (array-wise) then it won't do the obliteration thing I feared. I'd test that though. (Me being shifty-eyed and not trusting all library calls.

)

wainamoinen · 08-01-2019, 04:42 PM

I'm not sure what you want to do exactly, but you can operate in groups of eight characters as an 64 bit integer, and do a right shift of 8 bits. NOTE that the right shift works on little endian architectures, on big endian architectures you would need to use a left shift.

So, on your example to shift a string of 9 characters plus '\0', the shift can be done as shown on the next code.

Code:

#include <stdio.h>
#include <stdint.h>

#define ORIGINAL  "abcdefghi"
#define CHARACTER 'x'

void Method1(void)
{
	char carr[10] = ORIGINAL;
	char c = CHARACTER;

	printf("\nMethod 1\n");
	printf("Original: %s\n", carr);
	carr[0] = carr[1]; carr[1] = carr[2]; carr[2] = carr[3]; carr[3] = carr[4]; carr[4] = carr[5]; carr[5] = carr[6]; carr[6] = carr[7]; carr[7] = carr[8]; carr[8] = c;
	printf("Shifted:  %s\n", carr);
}

void Method2(void)
{
	char carr[10] = ORIGINAL;
	char c = CHARACTER;

	printf("\nMethod 2\n");
	printf("Original: %s\n", carr);
	*(int64_t*)&carr >>= 8;
	carr[7] = carr[8];
	carr[8] = c;
	printf("Shifted:  %s\n", carr);
}

int main(void)
{
	Method1();
	Method2();

	return 0;
}

The result will be

Code:

gcc carr.c -Wall -Wextra -pedantic -o carr
./carr

Method 1
Original: abcdefghi
Shifted:  bcdefghix

Method 2
Original: abcdefghi
Shifted:  bcdefghix

If you want to use this method with larger strings, you will need to be sure that the buffer (carr) has a size of a multiple of 8 (plus the null terminator). And your code will be less clear and perhaps more difficult to maintain.

ntubski · 08-01-2019, 05:27 PM

Quote:

Originally Posted by schneidz

hi, is there a smarter way to do this ?

If by "this" you just mean "shift bytes in an array", then I second rtmistler's suggestion of memmove. If the question is more about recognizing a string in a stream, then I would structure it like this:

Code:

  const char* findme = "012345678";
  const char* findme_end = findme + strlen(findme);
  const char* found = findme;

  while(c != EOF)
  {
    c = fgetc(fstream);
    if (c == *found)
      found++;
    else
      found = findme;
    if (found == findme_end)
    {
      printf("hispanola");
      found = findme;
    }
  }

For more sophisticated stuff, you could look at using a lexer or similar.

Quote:

Originally Posted by rtmistler

Edit: And the manual page says that it does move bytes into a temporary array, so if you're moving left to right (array-wise) then it won't do the obliteration thing I feared. I'd test that though. (Me being shifty-eyed and not trusting all library calls.

)

memcpy is the one which does not work with overlapping src/dst, memmove should be fine.

rtmistler · 08-01-2019, 06:13 PM

Quote:

Originally Posted by schneidz

hi, is there a smarter way to do this ?:

Code:

 c = fgetc(fstream);
 while(c != EOF)
 {
  carr[0] = carr[1]; carr[1] = carr[2]; carr[2] = carr[3]; carr[3] = carr[4]; carr[4] = carr[5]; carr[5] = carr[6]; carr[6] = carr[7]; carr[7] = carr[8]; carr[8] = c;
  printf("%c", carr[0]);
  if(strcmp(carr, "012345678") == 0)
   sprintf(carr, "hispanola");
}

A second look by the way, because one thought to me originally was to use a pointer and move back one character, decide what to do with the new first character, and call it a day.

What's missing is a proper definition of the carr[] array, as well as a definition as to how this is tied, or not tied to the variable c.

If your intention is to rotate new characters into an array, we don't know.

rtmistler · 08-01-2019, 06:15 PM

I mean I'd just fill the array in from the end, or second to last location, with each character and adjust the pointer backwards as it goes, and provide the pointer once it's all saved.

And, you know, overflow avoidance to not go beneath the zero element and set the last location to NULL so as to make it a string, ... verify if each input character is printable.

schneidz · 08-01-2019, 07:23 PM

Quote:

Originally Posted by rtmistler

Without libraries? yes and no. Yes, you can use pointers, but no it's all manually moved byte by byte, just with a loop or something.

With libraries? memmove(3) - Recommend you read the manual page thoroughly. It DOES allow for overlapping regions, but one has to be careful about src/dst and that you're not obliterating src as you go along.

Edit: And the manual page says that it does move bytes into a temporary array, so if you're moving left to right (array-wise) then it won't do the obliteration thing I feared. I'd test that though. (Me being shifty-eyed and not trusting all library calls.

)

big up. my searches kept pointing me to shift() but i think thats for shifting bits (instead of whole bytes).

schneidz · 08-01-2019, 08:03 PM

Quote:

Originally Posted by rtmistler

A second look by the way, because one thought to me originally was to use a pointer and move back one character, decide what to do with the new first character, and call it a day.

What's missing is a proper definition of the carr[] array, as well as a definition as to how this is tied, or not tied to the variable c.

If your intention is to rotate new characters into an array, we don't know.

thanx for rechecking. for your curiosity, i wanted to copy an mvs-mainframe dataset substituting about a dozen org-id's into proper values (the dataset is mostly ascii but has random binaries and ascii-10 (lf)'s thrown in).
so sed, awk, cut would choke on it. weirdly, head was able to read the whole line. also fgets(), getline() wasnt able to read the whole line either.
so i decided to process the file one byte at a time instead.

ntubski · 08-02-2019, 05:03 AM

Quote:

Originally Posted by schneidz

bmy searches kept pointing me to shift() but i think thats for shifting bits (instead of whole bytes).

I think shift() is perl: https://perldoc.perl.org/functions/shift.html

C has some bit shifting operators, but no function called "shift", as far as I know.

Quote:

Originally Posted by schneidz

(the dataset is mostly ascii but has random binaries and ascii-10 (lf)'s thrown in).
so sed, awk, cut would choke on it.

Maybe they try and fail to decode as UTF-8? Setting LANG=C might help there.

schneidz · 08-02-2019, 06:11 AM

Quote:

Originally Posted by ntubski

...
C has some bit shifting operators, but no function called "shift", as far as I know.
...

good correction.

for completions sake:[untested]

Code:

c = fgetc(fstream);
 while(c != EOF)
 {
  memmove(carr, carr+1, 8); carr[8] = c;
  printf("%c", carr[0]);
  if(strcmp(carr, "012345678") == 0)
   sprintf(carr, "hispanola");
}