https://bugs.freedesktop.org/show_bug.cgi?id=63242
Mike Kaganski <[email protected]> changed: What |Removed |Added ---------------------------------------------------------------------------- Status|RESOLVED |REOPENED Resolution|NOTABUG |--- Severity|major |enhancement Priority|high |medium Ever confirmed|0 |1 --- Comment #7 from Mike Kaganski <[email protected]> --- Although the problem is clearly the hardware limitation (one can perform the steps that OP had described, and set all cells to show 20 decimal digits to see the real representation of the values and understand the underlying cause of this behaviour), here is what I wanted to draw attention to. The INT function (and maybe some others) is used in areas other than those where one deals with microscopic values. I.e., it is difficult to think of a use case where one performed high-accuracy arithmetics with numbers in a range of 10^-5, and then would truncate the result using INT. Thus, in most (if not all) usecases, taking into account these tiny deviations of the hardware representation of actual number is not only awkward, but is a real BUG. Who would care that the result of (4.01-4.0)*100 is represented as 0,99999999999997900000 when INT'ing it? Therefore, I disagree that this is NOTABUG. I set it to REOPENED and Enhancement request (but it may still be classified as a BUG by those more competent than I), and propose to add a threshold of accuracy to such functions (probably customizable somewhere in advanced options in UI, up to current behavior, but set to a reasonable value, say, like 10^(-4), by default). So that INT(0.999) would give 0, but INT(0.9999) would give 1. Of course, this setting need to be kept with the spreadsheet to give consistent results when opening it in another location. This could possibly require changes to ODF standard, and thus be difficult to do. -- 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
