Console Progress Bar

[Bit Hunter]

Today i was not feeling like working with anything particular, so i decided to work on a little emulation of progress bar in ANSI C for a console app.

Was wondering if there is anything that can be optimized in the code below?

#include <stdio.h>
#ifdef __MINGW32__
#include <windows.h>
#else
#define Sleep(t)	usleep((t) * 1000)
#endif

int main(int argc, char **argv) {
	if (argc < 3) return 1;

	int len = atoi(argv[1]);
	int step = atoi(argv[2]);
	int i;
	char buf[len+1];
	char c = '-';

	float fstep = (float)len / (float)step;

	for (i = 0; i < len; i++) {
		buf[i] = '-';
		buf[i+1] = '\0';
	}

	printf("[%s] %c %02.0f\%\r", buf, c, 0.0);
	fflush(stdout);

	for (i = 1; i <= step; i++) {
		Sleep(3);

		int j, l;
		for (j = 0; j < len; j++) {
			if (j < i*(fstep)) {
				buf[j] = '#';
				l = j;
			} else {
				buf[j] = '-';
			}
			buf[j+1] = '\0';
		}

		if (i != step) buf[l] = '>';

		int k = i%4;
		if (k == 0) c = '-';
		if (k == 1) c = '\\';
		if (k == 2) c = '|';
		if (k == 3) c = '/';

		float per = i/(float)step*100;
		printf ("[%-*s] %c %02.0f%% \r", len, buf, c, per);
		fflush(stdout);
	}

	printf("\r\n");
	return 0;
}

Should be able to compile with GCC as well as mingw32.

Attached are the screenshots of the application running on lenny and xp.

Edited April 6, 2011 by Bit Hunter

[Jason Cooper]

Just looking at it and this loop jumps out as being inefficient.

for (i = 0; i < len; i++) {
    buf[i] = '-';
    buf[i+1] = '\0';
}

Every time round the loop you put write to two locations in memory when you could do

for (i = 0; i < len; i++) {
    buf[i] = '-';
}
buf[i] = '\0';

[Bit Hunter]

Thanks, must have forgotten.

I think switch is faster then if:

if (k == 0) c = '-';
if (k == 1) c = '\\';
if (k == 2) c = '|';
if (k == 3) c = '/';

switch (k) {
default:
case 0:
        c = '-';
        break;
case 1:
        c = '/';
        break;
case 2:
        c = '|';
        break;
case 3:
        c = '\\';
        break;
}

Edited April 6, 2011 by Bit Hunter