Reading binary data into struct with ifstream

Question

I'm trying to read binary data from a file using ifstream.

Specifically, I'm trying to populate this "Header" struct with data read from a file:

struct Header {
    char id[16];
    int length;
    int count;
};

Now, if I read the file in this way, the result is exactly what I want:
```
input.read((char*)&hdr, sizeof(hdr));
```

But if I instead read each variable of the struct manually, the results are gibberish:

input.read((char*)&hdr.id,     sizeof(hdr.id));
input.read((char*)&hdr.length, sizeof(hdr.length));
input.read((char*)&hdr.count,  sizeof(hdr.count));

My question is, what is happening here that makes these two methods return different results?

You read into length and count, how have you defined them? Or, were they meant to be hdr.length and hdr.count? — Chris McGrath
– Chris McGrath, Commented Mar 5, 2014 at 8:37
If you think it relevant how you wrote the data to that file (or someone else did), you'd be right; it is so. Short answer: sizeof(Header) is required to be "at least" as big as the sum of the size of its members, and can be bigger when including implementation member padding. The members aren't guaranteed to be buttressed up against each other when writing the struct as a whole, yet you're reading it expecting they are. I.e. If you're "writing" a orange, trying to "read" a bag a tangerines isn't necessarily going to work. — WhozCraig
– WhozCraig, Commented Mar 5, 2014 at 11:38
@ChrisMcGrath Sorry, that was a copy-paste error, fixed it now — Dan
– Dan, Commented Mar 5, 2014 at 17:35
@Dan You "only know how its supposed to be read." And how is that ? The first method assumes whoever wrote this did it with the same layout and padding that you're using. If there is no padding issue, then the second will work (and is nearly the preferred way to do this to begin with, both for reading and writing). Given the layout and size of the data members in that struct, it would be odd if there were padding between elements, after the last element, or both, but I've seen odder things. sizeof(hdr) compared to sizeof() each member will tell much. — WhozCraig
– WhozCraig, Commented Mar 5, 2014 at 18:10
@Dan Don't count on it always being that way. You're fortunate neither padding nor endianess were an issue, because they usually are. — WhozCraig
– WhozCraig, Commented Mar 6, 2014 at 9:53

Marcel Zebrowski · Accepted Answer · 2018-08-25 17:54:01Z

13

It is also possible to read the struct in one step.

i.e. fh.read((char*)&h, sizeof(Header));

answered Aug 25, 2018 at 17:54

Marcel Zebrowski

1,01711 silver badges17 bronze badges

Sign up to request clarification or add additional context in comments.

3 Comments

Jean-Christophe Over a year ago

With this method, care must be taken to avoid memory alignment padding. See Structure padding and packing

starriet 차주녕 Over a year ago

@Jean-Christophe Thanks, it seems quite important. Could you give us a simple example of when the above code fails? Rather, it seems the code in the accepted answer has the problem you mentioned(structure padding): If the structure was padded, reading from it one element by one element could be corrupted because of the padded memory, no? (unless the fstream::read moves the position including the padded memory)

Jean-Christophe Over a year ago

@starriet This is not portable. The above code will lead to corrupted read if the file is initially serialized on a different instruction set. A typical case would be messaging from a 32-bit machine to a 64-bit machine. An other case of failure would come form the reading of an unpacked structure/scalar on packed data.

Blaz Bratanic · Accepted Answer · 2014-03-06 08:16:48Z

9

As the comment above states, you are probably missing hdr.length and hdr.count. I tried it with gcc 4.8 and clang 3.5 and it works correctly.

#include <iostream>
#include <fstream>

#pragma pack(push, r1, 1)
struct Header {
    char id[15];
    int length;
    int count;
};
#pragma pack(pop, r1)

int main() {
  Header h = {"alalalala", 5, 10};

  std::fstream fh;
  fh.open("test.txt", std::fstream::out | std::fstream::binary);
  fh.write((char*)&h, sizeof(Header));
  fh.close();

  fh.open("test.txt", std::fstream::in | std::fstream::binary);

  fh.read((char*)&h.id, sizeof(h.id));
  fh.read((char*)&h.length, sizeof(h.length));
  fh.read((char*)&h.count, sizeof(h.count));

  fh.close();

  std::cout << h.id << " " << h.length << " " << h.count << std::endl;
}

edited Mar 6, 2014 at 8:16

answered Mar 5, 2014 at 8:48

Blaz Bratanic

2,27914 silver badges19 bronze badges

3 Comments

WhozCraig Over a year ago

Change the id field to char id[15] and run that again, just for kicks.

Blaz Bratanic Over a year ago

You are right. Added pragma directives to fix alignment.

Jamie Over a year ago

This would be much improved by handling byte order. If this is written on a big endian system and read on a small endian system, things would go poorly.

Collectives™ on Stack Overflow

Reading binary data into struct with ifstream

2 Answers 2

3 Comments

3 Comments

Your Answer

Linked

Hot Network Questions

Collectives™ on Stack Overflow

2 Answers 2

3 Comments

3 Comments

Your Answer

Sign up or log in

Post as a guest

Linked

Related