Launchpad has imported 3 comments from the remote bug at https://www.libreoffice.org/bugzilla/show_bug.cgi?id=73410.
If you reply to an imported comment from within Launchpad, your comment will be sent to the remote bug automatically. Read more about Launchpad's inter-bugtracker facilities at https://help.launchpad.net/InterBugTracking. ------------------------------------------------------------------------ On 2014-01-08T21:02:45+00:00 Bjbraams-2 wrote: Created attachment 91706 Calc spreadsheet to demonstrate rounding errors for large argument values This report concerns LibreOffice Calc; functions roundup, rounddown, trunc, ceiling, floor, int, even and odd. A minimal set of instructions to reproduce the core of the problem follows. Open a new blank spreadsheet. In the A1 cell enter "=power(2,50)+1" (without the quotes) and in the B1 cell enter "=rounddown(a1,0)-a1" (without the quotes). Expected result: B1=0. Actual result: B1=5. More generally, let y=power(2,50)+1. I expect that rounding y to integer will return y exactly, for any of the functions roundup, rounddown, trunc, ceiling, floor and int. I also expect that even(y)=y+1 and odd(y)=y. Actual results: functions roundup(y,0), rounddown(y,0), trunc(y,0), ceiling(y,1), floor(y,1) and int(y) all return y+5. Moreover, even(y)=y and odd(y)=y+6 according to Calc. (Note that the second argument to roundup, rounddown and trunc indicates number of places behind the decimal symbol and the second argument to ceiling and floor indicates the unit.) There are similar problems with these rounding functions when rounding to some number (>0) of places behind the decimal symbol. The attached spreadsheet illustrates the issue. I have highlighted in yellow the cells where the result is different from what is expected. Row 5 shows the precise results described here; other rows show similar calculations for values y=power(2,x)+1 for other values of x. Reply at: https://bugs.launchpad.net/ubuntu/+source/libreoffice/+bug/1261048/comments/7 ------------------------------------------------------------------------ On 2014-01-10T11:32:55+00:00 Markus Mohrhard wrote: Please read about floating point arithmetics. http://en.wikipedia.org/wiki/Floating_point#Accuracy_problems Reply at: https://bugs.launchpad.net/ubuntu/+source/libreoffice/+bug/1261048/comments/10 ------------------------------------------------------------------------ On 2014-01-10T11:51:26+00:00 Christopher M. Penalver wrote: Markus Mohrhard, thank you for your comment. While the prior work done in this area is appreciated, in that it replicates the output of Excel (yes, I already know LibreOffice isn't Excel), I think closing this as a Won't Fix is premature, especially predicated on a post from Wikipedia. Despite this, venerable Gnumeric ( https://projects.gnome.org/gnumeric/ ) does in fact provide a more accurate result than LibreOffice or Excel in this instance: lsb_release -rd Description: Ubuntu Trusty Tahr (development branch) Release: 14.04 apt-cache policy gnumeric gnumeric: Installed: 1.12.9-1 Candidate: 1.12.9-1 Version table: *** 1.12.9-1 0 500 http://us.archive.ubuntu.com/ubuntu/ trusty/universe amd64 Packages 100 /var/lib/dpkg/status I think this should be discussed much further technically before making such a judgement. Thank you for your understanding. Reply at: https://bugs.launchpad.net/ubuntu/+source/libreoffice/+bug/1261048/comments/11 -- You received this bug notification because you are a member of Desktop Packages, which is subscribed to libreoffice in Ubuntu. https://bugs.launchpad.net/bugs/1261048 Title: [Upstream] Wrong results from rounding functions for large argument Status in LibreOffice Productivity Suite: Confirmed Status in “libreoffice” package in Ubuntu: Won't Fix Bug description: 1) lsb_release -rd Description: Ubuntu Trusty Tahr (development branch) Release: 14.04 2) apt-cache policy libreoffice-calc libreoffice-calc: Installed: 1:4.1.3-0ubuntu3 Candidate: 1:4.1.3-0ubuntu3 Version table: *** 1:4.1.3-0ubuntu3 0 500 http://us.archive.ubuntu.com/ubuntu/ trusty/main amd64 Packages 100 /var/lib/dpkg/status I am using Ubuntu 12.04.3 LTS and within that LibreOffice 3.5.7.2, Build ID: 350m1(Build:2) but this is also reproducible in LO Trunk 4.3.0.0.alpha0+ on Windows Vista: What is expected to happen at a terminal: cd ~/Desktop && wget https://bugs.launchpad.net/ubuntu/+source/libreoffice/+bug/1261048/+attachment/3942054/+files/LibreOfficeRoundingIssues.ods && localc --nologo LibreOfficeRoundingIssues Is that for cell D5 is it 0. What happens instead is that it is 5. This would be an issue with Calc numerical precision, as the actual outcome of 5 is also the same with Excel. WORKAROUND: Use gnumeric: apt-cache policy gnumeric gnumeric: Installed: 1.12.9-1 Candidate: 1.12.9-1 Version table: *** 1.12.9-1 0 500 http://us.archive.ubuntu.com/ubuntu/ trusty/universe amd64 Packages 100 /var/lib/dpkg/status --- ApportVersion: 2.0.1-0ubuntu17.6 Architecture: i386 DistroRelease: Ubuntu 12.04 InstallationMedia: Ubuntu 12.04.3 LTS "Precise Pangolin" - Release i386 (20130820.1) MarkForUpload: True Package: libreoffice 1:3.5.7-0ubuntu5 PackageArchitecture: i386 ProcEnviron: TERM=xterm PATH=(custom, no user) LANG=en_US.UTF-8 SHELL=/bin/bash ProcVersionSignature: Ubuntu 3.8.0-35.50~precise1-generic 3.8.13.13 Tags: precise running-unity Uname: Linux 3.8.0-35-generic i686 UpgradeStatus: No upgrade log present (probably fresh install) UserGroups: adm cdrom dip lpadmin plugdev sambashare sudo To manage notifications about this bug go to: https://bugs.launchpad.net/df-libreoffice/+bug/1261048/+subscriptions -- Mailing list: https://launchpad.net/~desktop-packages Post to : desktop-packages@lists.launchpad.net Unsubscribe : https://launchpad.net/~desktop-packages More help : https://help.launchpad.net/ListHelp
