C file size discrepency

问题

I am trying to learn C, and am currently working on a toy script. Right now, it simply opens a text file, reads it char by char, and spits it out onto the command line.

I looked up how to see the size of a file (using fseek() and then ftell()), but the result it returns doesn't match up with the number I get from counting the characters in a while loop as I iterate through the file.

I'm wondering if the discrepency is due to windows using \r\n and not just \n, since the discrepency seems to be #newlines+1.

Below is the script I am working on:

#include <stdio.h>
#include <stdlib.h>

int main()
{
        FILE * fp = fopen("test.txt", "r");

        fseek(fp, 0, SEEK_END);
        char * stringOfFile = malloc(ftell(fp));
        printf("allocated %d characters for file\n", ftell(fp));
        fseek(fp,0,SEEK_SET);//reset pointer

        char tmp = getc(fp); //current letter in file
        int i=0;
        while (tmp != EOF) //End-Of-File (defined in stdio.h)
        {
                *(stringOfFile+i) = tmp;
                tmp = getc(fp);
                i++;
        }
        fclose(fp);
        printf("Turns out we had %d characters to store.\nThe file was as follows:\n", i);
        printf("%s", stringOfFile);
}

And the output I get (with a simple test file you can see from the output) is:

allocated 67 characters for file
Turns out we had 60 characters to store.
The file was as follows:
line1
line2
line3
line4
line5
(last)line6

lmnopqrstuvw▬$YL Æ

where the tail bits of the printing seem to be garbage from allocating too much memory to the string.

Thanks in advance for any help/answer you can provide!

回答1:

If you're running windows:

FILE * fp = fopen("test.txt", "r");

opens the file in text mode which implies \r\n conversion to \n

So if your file has 7 lines, the conversion removes 7 chars (that is, if the file was using Windows-style line termination)

The fix is to open it in binary mode

FILE * fp = fopen("test.txt", "rb");

so ftell and reading chars one by one should match.

Of course, that's wasting space & not very convenient to have \r chars in your text, so you could allocate like you're doing, and in the end perform a realloc to shrink down the allocated memory with the actual number of chars (since it's smaller, it's ok)

stringOfFile = realloc(stringOfFile,i+1);

Note that since I've taken the need to add the nul-terminator into account, I've added 1 to the number of chars, so if there aren't any \r chars in the file, the realloc could increase the size of the block by 1.

So, as I was hinting at, don't forget to nul-terminate your string or printf doesn't stop properly:

stringOfFile[i] = '\0';

(unless you don't care about creating a C-string, since storing the string size + display char-by-char is also correct)

We've see that the ftell method is tricky, and in some cases, when the stream is for instance the output of a command (popen returns a FILE * but you cannot fseek it) or a socket, whatever, this principle cannot be applied since we don't know the size of the data in advance.

In the general case, it would be better to:

allocate a small buffer
read char by char and store
if buffer is full, call realloc to increase the size by some step (not at every char, performance would be bad)
in the end, call realloc again to adjust the size more precisely

(that solves the binary/text issue transparently as well)

Note that if you're working with large files (>4GB) you have to use 64-bit unsigned integers for positions and fopen64 flavours of I/O functions (and all offset variables like i should be unsigned / conform to return type of ftell or you'll start having problems at 2GB). Well, I suppose it doesn't matter much when processing moderately small text files.

Also, check David answer. With text files, putting the result of getc in a char should work, but not in the general case with binary files.

回答2:

    char tmp = getc(fp); //current letter in file
    int i=0;
    while (tmp != EOF) //End-Of-File (defined in stdio.h)

You need to check the value returned by getc for EOF. Instead, you convert it to a char and then check whether that's equal to EOF converted to a char. But what if the value of char that converts to EOF is actually in the file? Check the docs, getc returns an int.

You have other mistakes as well.

来源：https://stackoverflow.com/questions/48157920/c-file-size-discrepency

标签

filesize

file-pointer