Re: xmgrace

[John Gill <jng earthenware-services org>]

On Tue, 2004-07-13 at 12:18, Jody Goldberg wrote:
> On Tue, Jul 06, 2004 at 10:20:45PM -0400, Allin Cottrell wrote:
> > 
> > I'll put in my twopence-worth on this.  I'm a gnumeric fan, but I have 
> > to agree with Andrzej's point that gnumeric's graphing capacities are 
> > quite primitive.  And the response, that if we just wait a little 
> > these capacities will be terrific, without any need for third-party 
> > plugins, has worn thin over the several years I've heard it made.
>
> It's sad but true that the charting engine lost on the order of 2
> years of development, probably closer to 3 trying to go through
> bonobo to guppi.  It was the right choice at the time, but failed
> miserably for a variety of social and technical reasons.  Believe me
> when I say that writing everything from scratch was not my first
> choice.  However, the price has been paid, and we've got a shiny new
> engine now.  It's fairly easy to add new capabilities to bring
> things up to par with other applications.
 
Seems to me that work on the scripting interface is probably the best route to supporting other plotting packages.

With a good scripting interface creating functions to extract data and send it to your favourite plotting package should be fairly easy.  It ought to be possible using this approach to provide plugins to support other plot packages + if down the line a particular package proves to be a. stable and b. a clear winner then maybe switching to that as the default plotting would make sense.

Last time using an external package came up I'd suggested using the python chaco package -- hindsight says this would have been a big mistake + today I'd be suggesting the python matplotlib package.  

This one, together with python scripting would probably work well in gnumeric -- as per gretl for gnuplot, matplotlib is happy to spit out png's that you could then attach as images on your spreadsheet.

(to be fair gnuplot has been around an awful lot longer and going that direction would be a lot less risky than chaco/matplotlib + probably xmgrace).

John