Re: GODataScalar, GODataVector and GODataMatrix removal

[Jean Bréfort <jean brefort normalesup org>]

Le samedi 18 avril 2009 à 21:25 +0200, Emmanuel Pacaud a écrit :
> Le samedi 18 avril 2009 à 13:06 -0600, Andreas J. Guelzow a écrit :
> > On Sat, 2009-04-18 at 18:25 +0200, Emmanuel Pacaud wrote:
> > > Hi,
> > > 
> > > Here's the patches that removes GODataScalar, GODataVector and
> > > GODataMatrix and moves their functionality into GOData.
> > > 
> > > It could be possible to go further and also merge the GODataFooVal,
> > > GODataFooStr and GnmGOFooData classes into GODataVal, GODataStr and
> > > GnmGOData.
> > > 
> > > Comments ?
> > 
> > While I can see that there is a point in moving the GODataScalar and
> > friends into GOData, I do not like the deletion of those classes, I
> > think they and their API should be retained (and possibly marked as
> > depreciated) so not to break applications unneccessarily.
> 
> Right. I'm going to work on a patch that leaves the API as it is.
> 
> > I would also prefer that the changes to GODATA would then allow
> > arbitrary dimensional objects, not such be limited to dimension 2.
> 
> It would be nice, yes, but I'm not sure how the API would look like.
> 
> > What is the point of negative n_columns and n_rows? If there is no
> > meaning to that then why are they changed to signed ints?
> 
> GODataMatrixSize defines the meaning of negative values. I'm not sure
> there's a use for them though.

I'm not sure either. 0 size for rows or columns, or both, is enough to
tell that there are no valid data in the matrix. I do not remember why I
made that choice (as usual not enough comments...).