C++ binary file - Writing ints - strange behaviour

Question

I have a simple vector of ints, and I want to write it into a binary file. For example:

#include <fstream>
#include <vector>

int main () {
    std::vector<uint32_t> myVector{5, 10, 15, 20 };
    // write vector to bin file
    std::ofstream outfile("./binary_ints.data",  std::ios_base::binary|std::ios::trunc);
    std::copy(myVector.begin(), myVector.end(), std::ostreambuf_iterator<char>(outfile));
    outfile.close(); 
    return 0;
}

Then, if I inspect the file "binary_ints.data" in hex mode, I have this:

00000000: 050a 0f14 0a

Thats ok!.

However, If myVector has this data:

std::vector<uint32_t> myVector{3231748228};

Then, the hex stored is weird:

00000000: 840a

84 in Hex doesn't match with the Int 3231748228.

What's happening here? Thanks.

std::copy doesn't transform one element into multiple new elements, nor does std::ostreambuf_iterator. — chris
You could have done some experimentation to see what happens if the int value is > 255. — PaulMcKenzie
I'm curious where the extra newline is coming from (other than the one that is actually part of the vector data). The program should not be writing one after the vector data. — Michael Burr
@MichaelBurr, it is actually an interesting question. I did not see any newline in my tests during answer preparation. But as long as it is not a part of the question and I do not know how to reproduce it, I just did not do anything to resolve this "issue". — Innokentiy Alaytsev

Innokentiy Alaytsev Innokentiy Alaytsev · Accepted Answer · 2017-12-07T18:56:33

The problem is that each value in your std::vector<uint32_t> is interpreted as char during your std::copy () invocation. 3231748228 is represented in hex as ‭C0A09084. std::copy () takes uint32_t value, truncates it to single byte which is 0x84 on Little-endian processor. After writing byte 0x84 in file byte 0x0a is added which corresponds to new line character.

A possible solution is to use ofstream::write() instead of std::copy ():

#include <fstream>
#include <vector>

int main () {
    std::vector<uint32_t> myVector{3231748228 };
    // write vector to bin file
    std::ofstream outfile("./binary_ints.data",  std::ios_base::binary|std::ios::trunc);

    outfile.write (
        (char*)(myVector.data ()), 
        myVector.size () * sizeof (decltype (myVector)::value_type));

    outfile.close();
    return 0;
}

Note the use of decltype (). The same effect may be achivied by just writing sizeof (uint32_t), but with decltype () you may be sure that the code remains correct even if you change myVector value type.

C++ binary file - Writing ints - strange behaviour

1 Answers