Subj : Re: Reverse words in a string (Another Interview question)
To   : comp.programming
From : Randy
Date : Tue Oct 04 2005 02:54 pm

Roger Willcocks wrote:
> "Randy" <joe@burgershack.com> wrote in message 
> news:dhsgsb$dot$1@joe.rice.edu...
....
> 
> That seems like a lot of work... it's hard to beat the canonical solution:
> 
> #include <stdio.h>
> 
> void revtok(char* str, char sep)
> {
>     char *tmp, *end;
>     do {
>         while (*str && *str == sep) str++;
>         end = str;
>         while (*end && *end != sep) end++;
>         tmp = str; str = end;
>         while (--end > tmp) { char t = *tmp; *tmp++ = *end; *end = t; }
>     } while (*str);
> }
> 
> void main(int argc, char* argv[])
> {
>     char str[] = "when in the course of human events";
>     revtok(str, '\0');
>     revtok(str, ' ');
>     printf("%s\n", str);
> }
> 
> --
> Roger

That's a nice algorithm -- reverse the entire string then reverse each
word back to right-way-round.  Very clean.

But it does introduce an interesting issue.  Your code's two phase
reversal of tokens -- the entire string then each word itself -- is
counterintuitive.  For the next person to understand what your code is
doing, comments would be essential.

Short of using a language that has builtins to do this, I definitely
prefer your approach to manipulating each word.  Very palindromic.

    Randy

.