COUNTIFS, AVERAGEIFS and ECMA spec vs. Excel behavior

[Eike Rathke <erack@sun.com>]

Hi,

Thoughts addressed especially to Doug ;-)

With the latest revision of the spec uploaded this morning I added
drafts for COUNTIFS, AVERAGEIF and AVERAGEIFS.

As already mentioned earlier, the ECMA spec for COUNTIFS has a first
parameter 'count-range' that is not implemented by Excel2007. Looking at
the definition for AVERAGEIFS revealed that the syntax and arguments
description of COUNTIFS seems had been copied from AVERAGEIFS, including
its parameter 'average-range' that then was renamed. The examples given
for COUNTIFS do match the Excel2007 implementation, but not the ECMA
definition.

Furthermore, AVERAGEIFS 'average-range' argument is decribed as "The
actual cells that are averaged are determined by using the top, left
cell in average-range as the beginning cell, and then including cells
that correspond in size and shape to cell-range." This seems to had been
copied from the description of AVERAGEIF, but is not implemented by
Excel2007. Instead, all ranges have to be of the same dimension in shape
and size, else an error is returned. In Excel2007 the described behavior
is implemented only for the third optional AVERAGEIF parameter
'average-range'.

Given these discrepancies I followed with my definition what Excel2007
actually implements.

Doug, can you confirm that the Excel2007 implementation is indeed the
desired behavior and will not change to what the ECMA spec says instead?

Thanks
  Eike

-- 
Automatic string conversions considered dangerous. They are the GOTO statements
of spreadsheets.  --Robert Weir on the OpenDocument formula subcommittee's list.

PGP signature