fgets read a whole file

alaios · 08-21-2005, 10:48 AM

Hi is it possible read a whole value and keep it in one string variable? If not how i can tell the fget to read till he finds a \n character

Is it possible instead of using so big number (see below) to put a constant that means either infinity or either a very big number
fgets(tempstr,10000000,fp);

Hko · 08-21-2005, 12:33 PM

While possible with fgets(), I think the low-level read() syscall is more suitable and effecient for reading entire files.

Here's how to do that. The program checks how big the file is, and uses only the minimal amount of memory needed to store the string.

Code:

#include <stdio.h>
#include <stdlib.h>
#include <unistd.h>
#include <sys/types.h>
#include <sys/stat.h>
#include <fcntl.h>

#define FILENAME "/your/file.txt"

int main()
{
    int fd;
    struct stat st;
    char *string;

    /* Open the file */
    fd = open(FILENAME, O_RDONLY);
    if (fd < 0) {  /* Check if succesfully opened. */
        perror("Error opening file");
        return 1;
    }

    /* Get its size in bytes */
    if (fstat(fd, &st) < 0) {   /* Check if succesful. */
        perror("Error getting file size");
        return 1;
    }

    /* Get that much memory for the string.
     * But get 1 byte more, to store the '\0' to terminate the string.
     */
    string = malloc(st.st_size + 1);   /* 1 byte more */
    if (string == NULL) {  /* Check we really got it. */
        fprintf(stderr, "Error: out of memory.\n");
        return 1;
    }

    /* Read the entire file into the string. No more, no less. */
    if (read(fd, string, st.st_size) < 0) { /* Check for error */
        perror("Error reading the file");
        return 1;
    }

    /* Make sure the string is terminated. */
    string[st.st_size] = '\0';

    /*** Once here, the entire file is read into 'string' without errors. ***/

    /* Now just output the string. */
    puts(string);

    /* Done, exit succesfully (i.e. return exit code 0) */
    return 0;
}

alaios · 08-21-2005, 01:19 PM

Thx a lot really..... Very nice code well documented ... nice cool try
Where have u learned all these? The most tutorials i have read never talk abou those topics....
Where have u learned about #include <fcntl.h> library
Thx again

Hko · 08-21-2005, 04:37 PM

Quote:

Originally posted by alaios
Where have u learned all these?

In books on C-programming for UNIX/Linux.[/quote]

Those read(2) and open(2) functions are the actual functions that talk to the kernel (low-level). fgets(3), fprintf(3) and friends use those low-level functions.

When reading line-by-line, read(2) is harder to use than fgets(). fgets(3) and other stdio-functions try to optimize things through buffering by line.

But when you want to read an entire file at once, you can get to do the buffering yourself by using read(2) directly. So you can buffer-by-file, which is optimal.

You can create a buffer dynamically, exactly as big as you need for the file. But you'll need to know the file size, which you can do with fstat(2).

Quote:

Where have u learned about #include <fcntl.h> library

I read that in: man 2 open

Quote:

Thx again

You're welcome.

paulsm4 · 08-21-2005, 06:57 PM

Hi -

1. HKo's solution is entirely excellent - great job!

2. I thought I'd mention that "fgets()" is most appropriate when
a) You're doing line-by-line input from a text file
... AND ...
b) You're quite certain that no line will ever exceed your buffer size (yes,
fgets() will prevent buffer overflows - but you'll still have to detect and
deal with the partial input).

3. I thought I'd *also* mention that if you wanted to deal with "the whole
file", then mmap might be an interesting option for you to consider: use the
Linux VM manager to simply treat the file as a contiguous area of memory:

http://www.ecst.csuchico.edu/~beej/guide/ipc/mmap.html

'Hope that helps .. PSM