I was just wondering how getchar() is implemented? Is it something like the following? It is quite inefficient to read this way, single byte. Does it use some buffering?
Pseudo code:
int getchar() {
char buf[1];
int n = read(0, buf, 1);
if(n < 1)
printf("Read failed");
return buf[0];
}
Considering the number of implementations for the standard C library, it's not possible to provide a definite answer, but the most common ones seem to follow the same general guidelines.
By definition, getchar() uses the stream infrastructure of the standard C library i.e. FILE and its friendly functions. In most modern C library implementations, the file streams are buffered to a degree, with the buffer size and behavior usually being tunable via setvbuf().
I am aware of at least one case (glibc) where files may optionally - via an extra option to fopen() - be accessed via memory mapping (i.e. mmap()) rather than read()/write(). To avoid issues when mixed with calls to higher level functions, such as scanf(), getchar() is forced to use the same buffering structures.
Without information from a profiler, I'd worry more about the structural complexity of any code that uses getchar() than about any performance issues caused by its use.
Here is a very simple implementation.
int mygetchar(void)
{
static char buf[BUFSIZ];
static char *bufp = buf;
static int i = 0;
if (i == 0)
{
i = read(0, buf, 1);
bufp = buf;
}
if ( --i >= 0 )
{
return *bufp++;
}
return EOF;
}
