Bug 72826

Summary:	rounding error in combin()
Product:	LibreOffice	Reporter:	Toralf Förster <toralf.foerster>
Component:	Calc	Assignee:	Not Assigned <libreoffice-bugs>
Status:	RESOLVED NOTABUG
Severity:	normal	CC:	gerard.fargeot, jmadero.dev, rb.henschel
Priority:	medium
Version:	4.1.3.2 release
Hardware:	x86 (IA32)
OS:	Linux (All)
Whiteboard:
Crash report or crash signature:		Regression By:
Attachments:	calculate the odds of a german lottery separated factors of the hypgeomdist function - the diff of 20 is still here

Description Toralf Förster 2013-12-18 09:59:25 UTC

Created attachment 90913 [details]
calculate the odds of a german lottery

The attached file contains in sheet "analysis", N6 the value
92,520,104,463,809,100 (origin in [1])

However there are people who claim in [2] a number of 92,520,104,463,809,120
(diff is 20).

Rounding errors in column D or E might be reason - but that would mean, that MS Excel under a different system would suffer exactly from the same issue ...

Any chance to clarify it ?


[1] https://github.com/toralf/bingo/blob/master/bingo.ods
[2] http://www.mathematik.uni-kl.de/~ruckdesc/pubs/bingoPaper-KroisandtRuckdeschel.pdf

Comment 1 Toralf Förster 2013-12-18 13:19:40 UTC

Created attachment 90923 [details]
separated factors of the hypgeomdist function - the diff of 20 is still here

in this modified sheet any division of the involved numbers is avoided - but the value in cell "analysis":M8 still differs (marginally) from the official number (which is assumed to be correct)

Comment 2 Regina Henschel 2013-12-18 16:50:57 UTC

The attached document is not suitable.
- Please reduce it, so that it contains only the function call, which you think has a wrong result, and only the needed input values.
- Say where to find this function call.
- Say what value you get and what value do you expect.
Please keep in mind, that we are volunteers, working here in our spare time.

Comment 3 Toralf Förster 2013-12-19 10:56:21 UTC

sure - in the mean while I think, it is not a bug.

BTW where can I find the precision of LO for multiplying 2 big integers ?

Comment 4 GerardF 2013-12-19 16:02:19 UTC

There is no error in the result of COMBIN function.
You are just trying to calculate with high numbers (>15 digits).

See http://en.wikipedia.org/wiki/Floating_point

Comment 5 Joel Madero 2013-12-21 02:16:20 UTC

This has been discussed to the point of exhaustion on mailing lists and other bugs - not a bug