Re: [guppi-list] Hacking continues unabated

[Jon Trowbridge <trow emccta com>]

On Wed, Aug 02, 2000 at 04:28:41PM +0100, Neil J Pilgrim wrote:
> Ah right...so a 'scatter plot' is not really that much of a composite
> object (implied by the config dialog) but just a particular combination
> of primary objects ?

If I understand you correctly, yes.

I've tried to keep the "quanta of plotting" relatively simple and
small.  So the plot that you see when you select "Plot->New Scatter
Plot" is built up out of:

  * One Text Item: the title
  * Two Axis Items
  * One Frame Item, which draws the gray rules on the plot.
  * One Scatter Item: which draws the points in the plot
  * One Slinreg Item: which does the regression calculation and
                        draws the line.

These six items are creates and arranged relative to one another by
a small piece of Scheme code that gets executed when the "Plot->New
Scatter Plot" menu item is chosen.

The scatter item does a plot based on one set of x-data and one set of
y-data only.  To get N superimposed scater plots, one would lay N
scatter items on top of one another.

What Guppi lacks, and what I am working on, is a way to organize these
"quanta" into higher-level chunks that higher-level operations can be
performed on.  For example, you could now, without that much
difficulty, create a scatter plot matrix of NxN scatter
items. However, there is no provision for using Guppi's UI to operate
on all N^2 objects simulatenously, because there is no way to reflect
the heirarchical structure of this situation: that the N^2 scatter
items really comprise a greater whole whose properties should be
manipulable.

How to do this in a nice way is the question at hand, and Guppi can't
really be a nice user-level tool until it is addressed.  I do have
some ideas about this, though, but my thinking about it is still in
a very preliminary state.

-JT

-- 
GNU/Linux: Free your mind and your OS will follow.