Re: [office-formula] BITAND, etc. - what about bit limits/sign bit?Gnumeric folks, willing to change what they do?

Andreas J. Guelzow wrote:
>
> I think we should figure out a _good_ specification with less concern
> about what current applications implement.
A good ideal, but we also need to worry very much about importing/exporting
current documents.
>  If gnumeric depends on its
> current implementation, gnumeric would be free to add it as a gnumeric
> specific function.
>   
That's true.  But if handling 32-bit signed values specially really IS 
important,
then we need to know that.  Which is why I raise this as a discussion item,
to help find out what's important.

> So more importantly, how do you envision a BITAND interface? Assuming
> that we only allow integer arguments or truncate to integers what do we
> do with negative numbers? For example if we require the specification of
> a bit length as a third argument, Gnumeric could use its 2-argument
> function as if it has an optional bit length argument defaulting to 32.
>   
I have no trouble requiring non-negative integers as inputs.  That would 
make sense.

The problem with having a third argument is that BITAND, BITOR, etc.
could very rationally be extended to support a list of NumberSequences,
and the two uses conflict.

> Let's write good definitions.
>   

I wholeheartedly agree there!  And having discussions like this helps.

--- David A. Wheeler