10d 14h 13m 33s
Make it easier to read than something like
Under ANSI C, you have the various functions under time.h and the “official” way to calculate elapsed time is to call difftime() (as you portably can't assume that time_t is in sections, or even a integral type). That returns, as a double, the number of seconds between the two time values.
The problem is, you can't really take that result, stick it into a struct tm and call mktime() to renormalize it, as the value could overflow an int (which is what each field is defined as in struct tm). An int is defined as a minimum of 16 bits, which isn't even enough to record the number of seconds in a day. Sure, on modern systems ints are probably 32 bits in size, but that only leaves you some 68 years before the seconds overflow (it should be fun in 2038).
So it's more portable to do the math directly.
So, I thought I'd be cute and try to do the minimal amount of math possible, and in looking over math.h I saw modf(), which splits a double into its integral portion and fractional portion. So, I tried:
diff = difftime(end,start); diff /= 60.0; tmp = modf(diff,&diff); sec = tmp * 60.0; diff /= 60.0; tmp = modf(diff,&diff); min = tmp * 60.0; diff /= 24.0; tmp = modf(diff,&diff); hour = tmp * 24.0; diff /= 365.0; tmp = modf(diff,&diff); day = tmp * 365.0; year = diff;
only to have it fail miserably. Even several variations on that didn't work. So I bit the bullet and did it the old fasion way:
#define SECSMIN (60.0) #define SECSHOUR (60.0 * 60.0) #define SECSDAY (60.0 * 60.0 * 24.0) #define SECSYEAR (60.0 * 60.0 * 24.0 * 365.2422) diff = difftime(end,start); year = (int)(diff / SECSYEAR); diff -= ((double)year) * SECSYEAR; day = (int)(diff / SECSDAY); diff -= ((double)day) * SECSDAY; hour = (int)(diff / SECSHOUR); diff -= ((double)hour) * SECSHOUR; min = (int)(diff / SECSMIN); diff -= ((double)min) * SECSMIN; sec = (int)(diff);