Re: [office-formula] LOOKUP function

[Kohei Yoshida <kyoshida@novell.com>]

On Thu, 2008-03-06 at 17:30 +0100, Eike Rathke wrote: 
> Hi Kohei,
> 
> On Monday, 2008-02-25 11:30:17 -0500, Kohei Yoshida wrote:
> 
> > * 3-parameter case ([lookup value], [lookup matrix], [result
> > matrix(vector)])
> 
> Um, you wrote the syntax with every parameter being optional, that's
> probably not intended?

Ah, no.  You're right; that was not intended.  Actually I didn't mean
that to be a syntax definition in the ODFF spec & never realized that
the angular brackets were used to indicate an optional parameter. ;-)

> 
> > 
> >   2) The directions of the lookup matrix and the result matrix can 
> >     differ.  [NOTE: the ODFF spec does not mention this.] 
> 
> Just to make sure that I find the right wording while editing the spec:
> 
> If the lookup matrix is a row vector or square or taller than wide, and
> the result vector is a column vector, then the value corresponding
> offset-wise is returned. Same if the lookup matrix is wider than tall
> and the result vector is a row vector.
> 
> Or could someone come up with a less confusing wording?
> 
> >   3) The lengths of the search vector and the result vector do not need 
> >     to be identical; however, when the match position falls outside 
> >     the length of the result vector, an error is raised if the result 
> >     vector is given as an array object.  If it's a cell range, it gets 
> >     automatically extended to the length of the searched vector.
> 
> But in the direction of the result vector, as I just found out.
> And: if just a single cell reference is passed, a column vector is
> generated.

Ah, good discovery.

> 
> >     If the cell range cannot be extended due to size limit, then '0'
> >     is displayed.
> 
> How odd..

Yeah.  So, this is a corner case specific to one particular application.
I'm not really sure if it's worthwhile to spec this, to be honest.

> 
>   Eike
> 
-- 
Kohei Yoshida - OpenOffice.org Engineer - Novell, Inc.
<kyoshida@novell.com>