https://bugs.freedesktop.org/show_bug.cgi?id=86005

--- Comment #5 from Jindrich Svorc <[email protected]> ---
(In reply to Robinson Tryon (qubit) from comment #4)
> (In reply to Jindrich Svorc from comment #3)

> Sure, but you aren't representing 0.45 or 0.6 as floating-point numbers :-)

Right, ok, that explains it. Thank you. 
I thought the floating number should give you exact result. No additional
rounding needed. But apparently it is not valid in all cases.
It might make sense to round it for N-1 bits. 

Just one additional comment, It was not visible anywhere that the number in the
cell is actually 0.999999999980 and not 1. I am not sure whether it is
deliberately but it was difficult to find the root-cause. Even the sum in the
status bar showed 1.

Maybe if the INT function round the number for 14 decimal places before it is
applied it can help. But I am not sure whether I can foresee all the
consequences.

Thanks for explanation. 

> 
> > Interestingly, if you change the equation to get result = 2 it calculates it
> > correctly.
> 
> Indeed! Look at this:
> (8.45-8)/0.6 -> 0.749 999 999 999 999 000 00
> 
>  0.45/0.6    -> 0.750 000 000 000 000 000 00
> (7.45-7)/0.6 -> 0.750 000 000 000 000 000 00
> 
> (3.45-3)/0.6 -> 0.749 999 999 999 999 000 00
> (9.45-9)/0.6 -> 0.749 999 999 999 999 000 00
> 
> > I would not care too much since I usually don't need the 15th place but the
> > problem is that INT function gives me 0 instead of 1 so it suddenly creates
> > a big difference.
> 
> Perhaps you could use 10 or 15 places of precision? That should hopefully
> round the number up or down as expected...

-- 
You are receiving this mail because:
You are the assignee for the bug.
_______________________________________________
Libreoffice-bugs mailing list
[email protected]
http://lists.freedesktop.org/mailman/listinfo/libreoffice-bugs

Reply via email to