Re: bug in functions 'roundup' and 'ceil'

[Jody Goldberg <jgoldberg home com>]

On Mon, Aug 13, 2001 at 09:35:41PM +0100, Nick Lamb wrote:
> Perhaps someone who has Excel installed can reverse engineer the "fix"
> used in Excel and we can try to match it in Gnumeric? My understanding
> is that Gnumeric is trying first to be compatible with Excel, and
> second to be a good spreadsheet in its own right; so since there's no
I disagree with this.  Gnumeric's primary goal is to be an excellent
spreadsheet.   XL compatibility is important, but is only a means to
an end.  In situations where XL is demonstrably inferior we need not
hang ourselves beside them.  Our users may need the same noose, but
it should be optional, and possibly not the default.

> fix to this problem except teaching binary floating point math in
> primary schools we should try to be broken in the same way as Excel :)
This is not the sort of thing that can be easily reverse engineered.
There are several layers of abstraction between the input and the
display (formatting, internal storage formats etc).  It would be
interesting to see what OpenCalc has done in this situation, but
second guessing XL for this will be a waste of time.