Puzzle from comp.lang.c++

[NMTop40]

Thought I might bring this here as a bit of fun:

Hi

I recently had to write a small code in a competition ,but my code was
rejected cause it failed in 1 of test cases.


The problm was .....we are given vector of strings....each string
consists of either 1 or 2("12122" 0r "2121" so on..)...i had to find
the that string where percentage of '1' is minimum.Now the problem and
solution both are trivial but i was told that comparing double with <
or > sign doesn't ensure a correct solution always ie it fails for
number differing by a very small value.


Also if two string have indentical small percentage of '1 then the
smallest indexed string is to be returned(ie string of index 0 in
vector is identical with index 4 then 0 is to be returned.)


The code goes something like this.....

#include<iostream> 
#include<string> 
#include<sstream> 
#include<vector> 

using namespace std; 


class Elections
{ 
public: 
        int visit(vector<string> like); 
}; 

int Elections::visit(vector<string>like) 
{ 
        double min=1000,temp=0; 
        int index=0; 

        for(int i=0;i<like.size();i++) 
        { 
                temp=0; 
                for(int j=0;j<like.size();j++) 
                        if(like[i][j]=='1') 
                                temp++; 

                        temp=temp/like[i].size(); 

                        if(min>temp) 
                        { 
                                min=temp; 
                                index=i; 
                        } 
        } 
        return index; 
}

The test case for which my solution fails is
{"11122211222111111112121", "11122211222111111112121",
"11122211222111111112121", "11122211222111111112121",
"11122211222111111112121", "11122211222111111112121",
"11122211222111111112121", "11122211222111111112121",
"11122211222111111112121", "11122211222111111112121"}

The above case has all equal string yet it returned 9(index instead of
0) according to the system test they carried out.
I tried running the above code in a different compiler and the answer i
got was correct.(ie 0)


I am not sure , besides the floating point comparison ambiguity what
other reason could have led to different results in two different
compilers.


thanks in advance


Somesh