Re: COUPNCD COUPPCD Aarrrghhhh!

[Neil Booth <neil daikokuya demon co uk>]

Andreas J. Guelzow wrote:-

> I would agree with Jody wholeheartedly, especially since Excel 2000 
> seems to calculate more correct values. (I could have saved myself lots 
> of trouble if I didn't have to chase a moving target.)

But doing this _is_ a moving target.  The latest is XP, so if this is
policy there's little point updating for 2000.  We need to find
someone with XP.

I'd prefer we DTRT, then let MS converge towards us 8-) But if others
prefer to match XP, that's fine with me.

> Could somebody please update Jp's regression files to refer to the Excel 
> 2000 results and place them somewhere were we can get to them. I would 
> be very willing to change the code (again) to have just one version!

We can remove the XL compatibility from COUPNCD and COUPPCD, since
2000 appears to do that perfectly.  I'm assuming, of course, that XP
doesn't re-break them 8-)

But we probably to keep the function coup_cd_xl, since other functions
like COUPDAYS are still broken in 2000 (and therefore maybe XP).  I
have figured out what the XL algorithm for COUPDAYS with Act/Act is (I
can get a perfect score on the regression sheet), and it's easier to
implement with this function still around.

There's little point continuing until we have the latest regression
files.

Cheers,

Neil.