Reading a text file backwards in C

Question

What's the best way to read a file backwards in C? I know at first you may be thinking that this is no use whatsoever, but most logs etc. append the most recent data at the end of the file. I want to read in text from the file backwards, buffering it into lines - that is

abc
def
ghi

should read ghi, def, abc in lines.

So far I have tried:

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

    void read_file(FILE *fileptr)
    {
        char currentchar = '\0';
        int size = 0;

        while( currentchar != '\n' )
        {
            currentchar = fgetc(fileptr); printf("%c\n", currentchar);
            fseek(fileptr, -2, SEEK_CUR);
            if( currentchar == '\n') { fseek(fileptr, -2, SEEK_CUR); break; }
            else size++;

        }
        char buffer[size]; fread(buffer, 1, size, fileptr);
        printf("Length: %d chars\n", size);
        printf("Buffer: %s\n", buffer);


    }


    int main(int argc, char *argv[])
    {
        if( argc < 2) { printf("Usage: backwards [filename]\n"); return 1; }

        FILE *fileptr = fopen(argv[1], "rb");
        if( fileptr == NULL ) { perror("Error:"); return 1; }

        fseek(fileptr, -1, SEEK_END); /* Seek to END of the file just before EOF */
        read_file(fileptr);


        return 0;


    }

In an attempt to simply read one line and buffer it. Sorry that my code is terrible, I am getting so very confused. I know that you would normally allocate memory for the whole file and then read in the data, but for large files that constantly change I thought it would be better to read directly (especially if I want to search for text in a file).

Thanks in advance

* Sorry forgot to mention this will be used on Linux, so newlines are just NL without CR. *

Answer 1

I recommend a more portable (hopefully) way of file size determination since fseek(binaryStream, offset, SEEK_END) is not guaranteed to work. See the code below.

I believe that files should be at least minimally buffered at the kernel level (e.g. buffering at least one block per file by default), so seeks should not incur significant amount of extra I/O and should only advance the file position internally. If the default buffering is not satisfactory, you may try to use setvbuf() to speed up the I/O.

#include <limits.h>
#include <string.h>
#include <stdio.h>

/* File must be open with 'b' in the mode parameter to fopen() */
long fsize(FILE* binaryStream)
{
  long ofs, ofs2;
  int result;

  if (fseek(binaryStream, 0, SEEK_SET) != 0 ||
      fgetc(binaryStream) == EOF)
    return 0;

  ofs = 1;

  while ((result = fseek(binaryStream, ofs, SEEK_SET)) == 0 &&
         (result = (fgetc(binaryStream) == EOF)) == 0 &&
         ofs <= LONG_MAX / 4 + 1)
    ofs *= 2;

  /* If the last seek failed, back up to the last successfully seekable offset */
  if (result != 0)
    ofs /= 2;

  for (ofs2 = ofs / 2; ofs2 != 0; ofs2 /= 2)
    if (fseek(binaryStream, ofs + ofs2, SEEK_SET) == 0 &&
        fgetc(binaryStream) != EOF)
      ofs += ofs2;

  /* Return -1 for files longer than LONG_MAX */
  if (ofs == LONG_MAX)
    return -1;

  return ofs + 1;
}

/* File must be open with 'b' in the mode parameter to fopen() */
/* Set file position to size of file before reading last line of file */
char* fgetsr(char* buf, int n, FILE* binaryStream)
{
  long fpos;
  int cpos;
  int first = 1;

  if (n <= 1 || (fpos = ftell(binaryStream)) == -1 || fpos == 0)
    return NULL;

  cpos = n - 1;
  buf[cpos] = '\0';

  for (;;)
  {
    int c;

    if (fseek(binaryStream, --fpos, SEEK_SET) != 0 ||
        (c = fgetc(binaryStream)) == EOF)
      return NULL;

    if (c == '\n' && first == 0) /* accept at most one '\n' */
      break;
    first = 0;

    if (c != '\r') /* ignore DOS/Windows '\r' */
    {
      unsigned char ch = c;
      if (cpos == 0)
      {
        memmove(buf + 1, buf, n - 2);
        ++cpos;
      }
      memcpy(buf + --cpos, &ch, 1);
    }

    if (fpos == 0)
    {
      fseek(binaryStream, 0, SEEK_SET);
      break;
    }
  }

  memmove(buf, buf + cpos, n - cpos);

  return buf;
}

int main(int argc, char* argv[])
{
  FILE* f;
  long sz;

  if (argc < 2)
  {
    printf("filename parameter required\n");
    return -1;
  }

  if ((f = fopen(argv[1], "rb")) == NULL)
  {
    printf("failed to open file \'%s\'\n", argv[1]);
    return -1;
  }

  sz = fsize(f);
//  printf("file size: %ld\n", sz);

  if (sz > 0)
  {
    char buf[256];
    fseek(f, sz, SEEK_SET);
    while (fgetsr(buf, sizeof(buf), f) != NULL)
      printf("%s", buf);
  }

  fclose(f);
  return 0;
}

I've only tested this on windows with 2 different compilers.

Answer 2

You could just pipe the input through the program tac, which is like cat but backwards!

http://linux.die.net/man/1/tac