Having written a bunch of CGIs myself, I'd recommend that you REALLY DON'T use C -- rather C++ or (PREFERABLY) PERL.
Here is a simple example in C++ that does what you want:
#include <iostream>
#include <fstream>
#include <string>
#include <algorithm>
#define kNumScores (10)
using namespace std;
struct score_t
{
string name;
unsigned long score;
bool operator<(const score_t &inVal) const
{ return score > inVal.score; } // this line is confusing
};
int main()
{
ifstream inFile("my_scores.dat");
score_t scores[kNumScores + 1];
// Read the scores
for(unsigned i = 0; i < kNumScores; ++i)
{
string score;
getline(inFile, scores[i].name);
getline(inFile, score);
scores[i].score = atoi(score.c_str());
}
inFile.close();
// Add the new score
scores[kNumScores].name = "New Name";
scores[kNumScores].score = 150;
// Sort the scores
sort(scores, scores + kNumScores + 1);
ofstream outFile("my_scores.dat");
for(unsigned i = 0; i < kNumScores; ++i)
{
outFile << scores[i].name << endl;
outFile << scores[i].score << endl;
}
outFile.close();
return 0;
}
There are some fun bugs to try and find in the above code that have to do with concurrency (and multiple people entering data at the EXACT same time). This can be fixed with some configuration changes within Apache (causing it to run only a single instantiation of a CGI at a time).
Now, having given you all this, you really don't want to use C/C++, but rather PERL. The reason has to do with the fact that C/C++ can have exploited buffer overflows quite easily, and PERL is much friendlier at ignoring exploit attempts.
(If you choose not to user PERL, use C++ and syntax like the above. [The code presented is actually reasonably good at ignoring buffer overflow problems, etc.])
Good luck with whatever your solution.