merces / entropy Goto Github PK

Because I read the whole file at once, currently entropy does not support large files. I believe the solution would be reading the file in chunks. Here's what I tried:

vector<char> buff(1024, 0);
streamsize total = 0;

// Count occurrence of each possible byte, from zero to 255.
unsigned int counted_bytes[256] = { 0 };

while (!f.eof()) {
    f.read(buff.data(), buff.size());

    for (int j = 0; j < buff.size(); j++) {
        unsigned char c = buff[j];
        counted_bytes[c]++;
    }
    total += f.gcount();
}

f.close();

cout << calculate_entropy(counted_bytes, total) << " " << argv[i] << "\n";

However, the final entropy is not correctly calculated. Any ideas why?

This is critical. If you try to calculate the entropy for more than one file, the first file entropy is correct, but the others are not. See:

./entropy /bin/ls
2.71 /bin/ls

./entropy /bin/pwd
1.12 /bin/pwd

./entropy /bin/ls /bin/pwd
2.71 /bin/ls
2.09 /bin/pwd

do not forget -lstdc++fs in the one line compilation directive:

g++ -std=c++2a -o entropy entropy.cpp -lstdc++fs

Also, using std::setprecision requires:
#include // std::setprecision

If you provide entropy with more than one path where at least one of them is not a regular file, entropy will correctly fail for this file, but it will report the wrong file in the error message:

./entropy /bin/ls /bin
2.71 /bin/ls
"/bin/ls" isn't a regular file, skipping.