Why does returning a C-string from a function result in random characters?

Question

I've had to stop coding so many projects because of this weird quirk that I'm fed up enough to ask and risk looking like an idiot, so here goes...

I wrote a function like this:

const char* readFileToString(const char* filename) {
    const char* result;
    std::ifstream t(filename);
    std::stringstream buffer;
    buffer << t.rdbuf();
    result = buffer.str().c_str();
    return result;
}

I would expect that, if file.txt contains hello, that readFileToString("file.txt") should return hello. Instead, it returns garbled text, something along the lines of H�rv�0. However, if I put a std::cout << result; just before the return, it'll print hello.

Is this some weird, impossible quirk with C++? How do I fix it?

Answer 1

It's neither weird nor impossible; you returned a pointer to a buffer that went out of scope. The const char* doesn't "own" the string data, it only refers to it. Or, it used to! Once returned, that pointer is now invalid. You shall not dereference it.

I suggest you stick with std::string instead of venturing into advanced pointer techniques.

std::string readFileToString(const char* filename)
{
    std::ifstream t(filename);
    std::stringstream buffer;
    buffer << t.rdbuf();
    return buffer.str();
}

Unfortunately, I am not aware of any way to avoid a copy here.

If you don't mind shuffling your design about a little, and if you have some way to avoid a stream-to-string copy down the line, you could do this instead:

void readFileToStream(const char* filename, std::ostream& os)
{
    std::ifstream t(filename);
    os << t.rdbuf();
}

You may wish to return bool signifying the stream's state, but again you can do that at the callsite anyway.

Answer 2

Please see the commentary below:

const char* readFileToString(const char* filename) {
    const char* result;
    std::ifstream t(filename);
    std::stringstream buffer; // Behind the scenes some memory will/is allocated
    buffer << t.rdbuf();      // Memory is getting filled 
    result = buffer.str().c_str(); // Getting the address of that memory
    return result;
    // Buffer getting destroyed along with the allocated memory (what result points to)
}

.. Here results points to an invalid memory location

So hence it being corrupted