On Wed, Sep 21, 2005 at 12:19:24PM +0200, Eike Rathke wrote:
> Hi dev,
>
> On Tue, Sep 20, 2005 at 23:33:36 +0200, Eike Rathke wrote:
>
> > > http://software.newsforge.com/article.pl?sid=05/09/09/192250&tid=93
> >
> > Thanks for the pointer.
>
> And for those who are interested, here is what came to my mind yesterday:
>
> http://jroller.com/page/erAck?entry=openformula_specification_draft_for_opendocument
What is the goal of an open-formula specification ? I'd assume it is
to improve interoperability between versions and implementors of the
standard. If that is the case then there is already a standard, MS
Excel. Given the complexity and prevalence of it's evaluation
engine it seems wasteful for us to produce a standard that does
anything that is not a superset of Excel's conventions.
Note, I'm not saying that OOo needs to use Sheet!A1:A2 style
references, or to dumb down the analytics to conform to known
errors. However, handling the intricacies of
- iterative expressions
- array iteration
- implicit intersection
- natural language references
- the calling conventions for all of MS Excel's std functions
is a huge task. Any time the proposed standard and MS Excel's
conventions differ our workload increases substantially. We end up
having to provide a mapping to and from the convention in the xls
importer/exporter.
Things like locale conventions for decimal points, thousands
separators, argument and array separators are handle-able. That is
just display logic. Even address conventions can be handled
somewhat (I have a patch to add R1C1 style refs to OOo) although
that can start to break functions like INDIRECT or ADDRESS.
However, getting different calling or evaluation conventions for
functions is a slow torture.
---------------------------------------------------------------------
To unsubscribe, e-mail: [EMAIL PROTECTED]
For additional commands, e-mail: [EMAIL PROTECTED]
