>So the problem is that at the end of a while loop the comp might still calc and the while already goes further?? was that the problem??
Hm, no, I don't think so. The problem is that floating point calculations might be a bit off due to the internal representation even when the calculation is logically / mathematically correct.
So something like
floor(tStart,1/3,1/4) + 1/3 <= floor(tEnd,1/3,1/4)
might return false even (with your numbers for tStart and tEnd entered), even when it should return true, because floating point arithmetic calculations were not 100% correct, just a little off (for example, 1/3 cannot be represented correctly in the used base system).
[edit]: Even this should be more stable than my first version, though in a mathmetical sense, both should be equal:
...
while floor(tStart+ (iterno()-1)/3 ,1/3,1/4) <= floor(tEnd,1/3,1/4);
>While rounding I would get one part more.... any idea?
So your requirement is that the sum of your rounded numbers equals ShotsAll, right? So you need an unbiased tie-breaking when rounding.
I don't think there is the one and only one correct answer to this, maybe you can try something like
TEST:
LOAD recno() as ID,
TEST,
floor(TEST) as FLOOR,
TEST - floor(TEST) as DIFF
INLINE [
TEST
20
20.5
19.5
18.5
21.5
18.6
19.4
];
JOIN LOAD
round(sum(TEST)-SUM(FLOOR)) as TOSHARE
Resident TEST;
LOAD ID,
RANGESUM(FLOOR, if(recno()<=TOSHARE,1,0)) as RESULT
Resident TEST order by DIFF desc, FLOOR desc;
