Question

How do you reverse a string in C or C++ without requiring a separate buffer to hold the reversed string?

Answer 1

The standard algorithm is to use pointers to the start / end, and walk them inward until they meet or cross in the middle. Swap as you go.

---

Reverse ASCII string, i.e. a 0-terminated array where every character fits in 1 char. (Or other non-multibyte character sets).

void strrev(char *head) {   if (!head) return;   char *tail = head;   while(*tail) ++tail;    // find the 0 terminator, like head+strlen   --tail;               // tail points to the last real char                         // head still points to the first   for( ; head < tail; ++head, --tail) {       // walk pointers inwards until they meet or cross in the middle       char h = *head, t = *tail;       *head = t;           // swapping as we go       *tail = h;   } } 

// test program that reverses its args #include <stdio.h>  int main(int argc, char **argv) {   do {     printf('%s ',  argv[argc-1]);     strrev(argv[argc-1]);     printf('%s\n', argv[argc-1]);   } while(--argc);    return 0; } 

The same algorithm works for integer arrays with known length, just use tail = start + length - 1 instead of the end-finding loop.

(Editor’s note: this answer originally used XOR-swap for this simple version, too. Fixed for the benefit of future readers of this popular question. XOR-swap is highly not recommended; hard to read and making your code compile less efficiently. You can see on the Godbolt compiler explorer how much more complicated the asm loop body is when xor-swap is compiled for x86-64 with gcc -O3.)

---

Ok, fine, let’s fix the UTF-8 chars…

(This is XOR-swap thing. Take care to note that you must avoid swapping with self, because if *p and *q are the same location you’ll zero it with a^a==0. XOR-swap depends on having two distinct locations, using them each as temporary storage.)

Editor’s note: you can replace SWP with a safe inline function using a tmp variable.

#include <bits/types.h> #include <stdio.h>  #define SWP(x,y) (x^=y, y^=x, x^=y)  void strrev(char *p) {   char *q = p;   while(q && *q) ++q; /* find eos */   for(--q; p < q; ++p, --q) SWP(*p, *q); }  void strrev_utf8(char *p) {   char *q = p;   strrev(p); /* call base case */    /* Ok, now fix bass-ackwards UTF chars. */   while(q && *q) ++q; /* find eos */   while(p < --q)     switch( (*q & 0xF0) >> 4 ) {     case 0xF: /* U+010000-U+10FFFF: four bytes. */       SWP(*(q-0), *(q-3));       SWP(*(q-1), *(q-2));       q -= 3;       break;     case 0xE: /* U+000800-U+00FFFF: three bytes. */       SWP(*(q-0), *(q-2));       q -= 2;       break;     case 0xC: /* fall-through */     case 0xD: /* U+000080-U+0007FF: two bytes. */       SWP(*(q-0), *(q-1));       q--;       break;     } }  int main(int argc, char **argv) {   do {     printf('%s ',  argv[argc-1]);     strrev_utf8(argv[argc-1]);     printf('%s\n', argv[argc-1]);   } while(--argc);    return 0; } 

• Why, yes, if the input is borked, this will cheerfully swap outside the place.
• Useful link when vandalising in the UNICODE: http://www.macchiato.com/unicode/chart/
• Also, UTF-8 over 0x10000 is untested (as I don’t seem to have any font for it, nor the patience to use a hexeditor)

Examples:

$ ./strrev Räksmörgås ░▒▓○◔◑◕●  ░▒▓○◔◑◕● ●◕◑◔○▓▒░  Räksmörgås sågrömskäR  ./strrev verrts/.