A quick test to see if mmap() or fopen() + fgetln() is faster for linear line-by-line file processing.

mmap_vs_fgetln.c

#include <stdio.h>
#include <stdlib.h>
#include <time.h>
#include <errno.h>
#include <string.h>
#include <sys/mman.h>
#include <sys/types.h>
#include <sys/stat.h>
#include <sys/fcntl.h>
#include <unistd.h>

double avg_strlen_mmap(char*);
double avg_strlen_normal(char*);

int main (int argc, char **argv) {
  time_t starttime, endtime;
  double avg;

  starttime = time(NULL);
  avg = avg_strlen_normal(argv[1]);
  endtime = time(NULL);

  printf("Normal file operation speed: %ld secs, average: %lf.\n",
      endtime - starttime, avg);

  starttime = time(NULL);
  avg = avg_strlen_mmap(argv[1]);
  endtime = time(NULL);

  printf("Memory-mapped file operation speed: %ld secs, average: %lf.\n",
      endtime - starttime, avg);

  return EXIT_SUCCESS;
}

double avg_strlen_mmap(char *filename) {
  double num = 0;
  double len = 0;

  int fd = open(filename, O_RDONLY);
  if (fd == -1) {
    printf("Can't open file.\n");
    exit(1);
  }

  // figure out file size
  struct stat statbuf;
  int result = fstat(fd, &statbuf);
  if (result == -1) {
    printf("fstat failed\n");
    exit(1);
  }
  size_t length = statbuf.st_size;

  // use mmap
  caddr_t base = mmap(NULL, length, PROT_READ, MAP_SHARED, fd, 0);

  if (base == (caddr_t) -1) {
    printf("Could not mmap\n");
    exit(1);
  }

  size_t bytes_read = 0;
  char *str;
  size_t current_line = 0;
  for (str = (char *)base; bytes_read < length; bytes_read++, str++) {
    if (str[0] == '\n') {
      num += current_line;
      len += 1.0;
      current_line = 0;
    } else {
      current_line++;
    }
  }
  num += current_line;
  //len += 1.0;

  munmap(base, length);
  close(fd);

  return num / len;
}

double avg_strlen_normal(char *filename) {
  double num = 0;
  double len = 0;

  FILE *f = fopen(filename, "r");

  char *line = NULL;
  char *buf = NULL;
  size_t length;

  do {
    line = fgetln(f, &length);
    if (!line) {
      clearerr(f);
      break;
    }

    if (line[length - 1] == '\n') {
      line[length - 1] = '\0';
    } else {
      if ((buf = (char *)malloc((length + 1) * sizeof(char))) == NULL) {
        printf("Malloc failed.\n");
        exit(1);
      }
      memcpy(buf, line, length);
      buf[length] = '\0';
      line = buf;
    }

    num += strlen(line);
    len += 1.0;

    if (buf != NULL) {
      free(buf);
      buf = NULL;
    }
  } while (line != NULL);

  fclose(f);

  return num / len;
}