Help specifying a specific size type in C++

**indigo0086** · 06-11-2007

I am reading the header of a file and I know to use an unsigned char of 8 bits but what about an 8 byte portion of the header? Do I use an unsigned char[8] or unsigned unsigned char* for data types. I would think if I know the specific size I use an array and if it changes during runtime to dynamically allocate it, but generally what should I use.

Also I see that the fread function from the C standard library reads into a (void *) data type while the fstream class reads it into a char*. Is it safe to cast an unsigned char* into a char* or should I start off using a char* from the beginning even if it might affect the reading of the byte.

**Mortissus** · 06-11-2007

You can use a fixed size array to read the file. However, if your intention is to create an struct to interpret the content of a file you should be aware that the compiler might change the size of the members of the struct to optimize it, for example:

You define a struct:

Code:

struct MyStruct {
    char user_id[5];
    char valid;     
}

Then, you read a file into the struct like this:

Code:

MyStruct ms;
read(fd, &ms, sizeof(MyStruct));

The compiler could change the size of, for example, user_id, to a power of 2, in this case it would be 8. You would read 9 bytes, instead of 6. In gcc, to solve this problem you use #pragma

Code:

#pragma pack(1)
//define your struct here
#pragma pack

It is safe to cast from unsigned for reading, for example. You must take care to properly use the data inside the array, if casted.

**indigo0086** · 06-11-2007

well in my case there are several indicators that I ignore so rather then read to the sizeof(MyStruct), I would be reading to the size of the individual members of the struct specifying the size in bytes. The problem I'm running into is if I have an integer value that I need to read from the file.

Code:

struct a
{
   char data1[2];
   int data2;
}

read(file, (char*)data2, sizeof(data2));

I get something called a "segmentation fault" during runtime while debugging.

**UMR_Student** · 06-11-2007

Code:

read(file, (char*)data2, sizeof(data2));

//Should be this
read(file, static_cast<char*>(&data2), sizeof data2);

**brewbuck** · 06-11-2007

Originally Posted by Mortissus

The compiler could change the size of, for example, user_id, to a power of 2, in this case it would be 8. You would read 9 bytes, instead of 6. In gcc, to solve this problem you use #pragma

Code:

#pragma pack(1)
//define your struct here
#pragma pack

Ack. I'd never recommend changing the packing of a structure just for the "convenience" of being able to read it in a single call. Now you've limited yourself to running on systems where the compiler can pack structures. Just take the hit and read/write the data members individually.

And if you INSIST on packing the structure, and you're using GCC, don't use a pragma, this is deprecated. Instead use __attribute__((__packed__))

**Salem** · 06-11-2007

Or realise that you can never solve the endian problem, even if you manage to fix the alignment problem with various "pack" approaches.

At which point, read the file as a byte stream and manually add each byte to the appropriate member of the struct.

**indigo0086** · 06-11-2007

OK now I'm having a weird problem. When I try to use file.read(data1, 8) it gives me the string and some other weird data when I try to output it. When I use file.get(data1, 9) it gives me the proper input (at least as far as char goes). Should I not be using file.read()? What is the difference between that and file.get() besides the .get() size indicates n-1.

**robwhit** · 06-11-2007

I hope you resized data1 accordingly.

if you read() in data1, and then output it as a string, it won't have the NULL terminator.

**swoopy** · 06-11-2007

>What is the difference between that and file.get() besides the .get() size indicates n-1
file.get(data1, 9) will read text of up to 8 characters, and add a string terminator('\0') to the end. file.read(data1, 8) will simply read 8 bytes, but no string terminator is added. All C-style strings require a string terminator, which can be added manually if necessary.

**indigo0086** · 06-11-2007

Then why am I getting these extra characters. For example this is what I get♠

Code:

typedef struct
{
    char title[8];
    char type[8];
    int xpts;
    int ypts;
} TerrainHeader;

int main()
{
    TerrainHeader ter;
    std::fstream inputfile("scape.ter", ios::in);
    inputfile.read(ter.title, 8);
    cout << ter.title;
    return 0;
}

The output is "TERRAGEN π┬┬w¶╕├wx$="

the 8 bytes should be the string "TERRAGEN" but when I output it it doesn't display properly

**robwhit** · 06-11-2007

http://www.cprogramming.com/tutorial/lesson9.html

**Daved** · 06-11-2007

C style strings indicate their length with a null character that terminates the string. You are reading in 8 bytes for the eight characters, but you never add a null character to indicate the end of the string. In addition, your arrays only have room for eight characters, so if you did add the null character you would be accessing data passed the end of the array.

This only matters if you are using functions that assume a C style string. For example, operator<< assumes a null-terminated string, so it continues to output until it finds a null. If you aren't going to use the char array with functions expecting a C style string, you can leave it as is.

When you read in 9 characters, it apparently is reading in a null byte and making your code work?

Maybe I missed something, but I don't necessarily see the need to use read() and read in binary mode here.

**dwks** · 06-11-2007

It was just explained to you! Since you're using read(), you need to add the NULL terminator yourself.

Code:

int main()
{
    TerrainHeader ter;
    std::fstream inputfile("scape.ter", ios::in);
    inputfile.read(ter.title, 7);
    ter.title[7] = 0;
    cout << ter.title;
    return 0;
}

"TERRAGEN" cannot fit into a char array with 8 characters. You need 9 with the NULL. So:

Code:

typedef struct
{
    char title[9];
    char type[8];
    int xpts;
    int ypts;
} TerrainHeader;

int main()
{
    TerrainHeader ter;
    std::fstream inputfile("scape.ter", ios::in);
    inputfile.read(ter.title, 8);
    ter.title[8] = 0;
    cout << ter.title;
    return 0;
}

Or you could simplify things and just use get().

Code:

typedef struct
{
    char title[9];
    char type[8];
    int xpts;
    int ypts;
} TerrainHeader;

int main()
{
    TerrainHeader ter;
    std::fstream inputfile("scape.ter", ios::in);
    inputfile.get(ter.title, 9);
    cout << ter.title;
    return 0;
}

[edit] Three replies! Wow. I was too busy adding colour to my post . . . [/edit]
[edit=2] Make that four replies.

[/edit]

**swoopy** · 06-11-2007

> char title[8];
This should be:

Code:

    char title[9];

To make room for the string terminator.
> inputfile.read(ter.title, 8);
Now add the string terminator to make it a string:

Code:

    inputfile.read(ter.title, 8);
    ter.title[inputfile.gcount()] = '\0';

**indigo0086** · 06-11-2007

ah, how dumb of me. But it's perfectly ok to use read for non string data?

Thread: Help specifying a specific size type in C++

Thread Tools

Search Thread

Display

Help specifying a specific size type in C++

Similar Threads

char Handling, probably typical newbie stuff

failure to import external C libraries in C++ project

Please Help - Problem with Compilers

ras.h errors

Learning OpenGL