Re: Gnumeric, vectorial drawings (and charts)

[Jean Brefort <jean brefort normalesup org>]

Le mercredi 10 août 2011 à 16:38 +0200, Emmanuel Pacaud a écrit :
> Hi,
>
> Le samedi 06 août 2011 à 11:36 +0200, Jean Brefort a écrit :
> > There are several possiblities to improve the situation:
> > * use librsvg,
> > * use lasem (we already optionally use it for maths equations in
> > charts),
> > * add appropriate code in goffice.
>
> > The second would avoid to depend on librsvg, but I don't know if
> > Emmanuel will actively support it in the future, even in the near
> > future. Actually we might also consider importing the mathml code in
> > goffice.
>
> Why would you want to do that ?

Just a possibility if lasem went unmaintained, which might happen as it
is a one man project.

> > The third and most demanding solution would be to import the svg paths
> > as canvas items. We already have most of what we need in the canvas, and
> > we should just write the parser.
>
> I think you highly underestimate  the difficulty of SVG rendering, shown
> by the fact that all the modern web browsers fail to have a decent
> coverage of the SVG 1.1 specification.

I don't underestimate anything. I know it would be a long term issue.

> >  This would also work for wmf files
> > (Valek wrote some demo code in goffice in mf-demo.c) and might be
> > extended to emf which we do not support at all for now (as far as I
> > know). I suppose that OpenDocument graphics (odg) would be a valuable
> > target too.
>
> My take would be to convert wmf or odg to svg and use the best svg
> library available (librsvg currently) for the rendering.

Not sure that converting would make things easier.

> > About charts, I'm wondering if it would be valuable to rewrite the
> > rendering part using the canvas items. It would help for user
> > interactions with the graph elements, but might result in some
> > performance loss (difficult to know without experimenting).
>
> I'm not sure the conversion of the GogObjects to canvas items would
> really help for user interaction. The canvas positioning logic (absolute
> move and size of objects) is different to what you want for the graph
> objects: relative position and size w.r.t. parent object, modification
> of object properties (axis bounds).

The canvas can support relative positioning. And we can any feature we
need. Anyway, not a project for now.

Thanks,
Jean