Re: [guppi-list] New Guppi Snapshot

[Neil J Pilgrim <eennjp electeng leeds ac uk>]

Jon Trowbridge wrote:
> On Tue, Aug 29, 2000 at 01:05:23PM +0100, Neil J Pilgrim wrote:
> > Jon Trowbridge wrote:

[gnumeric configure bumpf]
> Yeah, that happens... Jody still hasn't checked in the guppi-gnumeric
> stuff, so there are some gaps.  It shouldn't cause any problems,
> though.

It didn't appear to, just mentioning it case :)

> It will create the Makefile, but won't build anything.  At least in
> theory...

I didn't notice - guppi takes ages to build over many files, so I wasn't
paying attention; haven't read the python makefile which results...

> > Pie-chart rotation is fine; repeated dragging-out behaves very
> >  oddly?
> 
> In what way?

Hmmm. It doesn't appear to be doing it much now, but before it was
randomly jumping the the zoom level back to normal?

More specifically, if I pull out (and in) the 3% wedge repeatedly then
the pie becomes smaller and smaller until it is just a small dot with
percentages around. Every now and then I manage to get it back to a full
size pie. Basically I guess the re-scaling of the rest of the pie when
you're dragging in and out the wedges is not working? Incidentally, if
you drag all of the wedges out (with right-button) then you can only
pull them further out with the left button, whereas I would think that
you should be able to also push them in?

> > Re multiple views, will multiple views only be possible on multiple
> > windows (unlinked views)? I'm trying to see a reason for linked views...
> > You could have something like in grace, where there is a 'stack' of
> > views which you can push new ones onto, plus cycle through. Ideally it
> > would be like an 'undo' list, showing all the changes that you've made,
> > perhaps?
> 
> Yeah, this isn't very flexible.  One problem has been that guppi's
> "engine" hasn't been good about handling on-the-fly changes to plots,
> (i.e. adding new elements, etc.) which means things have been pretty
> static in terms of the layout.  This is why there has always been a
> linear regression element there by default... it couldn't really be
> added later.

Ah. Hard-coded into the 'classes' ? 
[tangent]
I have to confess to wondering why C++ was neglected before: GTK+ uses
quasi-OO techniques so what difference would it make to people needing
to learn the code structuring, etc.?
[/tangent]

> The good news is that I've recently fixed this, and plots can be build
> up or torn down while being displayed.  So this sort of thing should
> be getting better in the future.

...including the elusive configuration system ?

> > Should 'home' be a 'mouse tool'?
> 
> I agree that it is a bit strange, but it can be nice to "home" only
> part of your graph.  i.e. you can "home" the x-axis but leave the
> y-axis zoomed in.  Of course, the easiest thing is to just press the
> "Home" key, but I guess the theory is to put everything on the toolbar
> for the benefit of new users...

Aha! I hadn't noticed the ability to zoom only on one axis (or home
too). It certainly is useful then!

> > You cannot start a reframe outside of the plot, but can end it
> > outside.
> 
> I original confined the mouse to the window in question during tool
> ops, but this became annoying for things like dragging.  I figured
> that there was no harm in this.

Well no, it enables extension of the range in plots, for example. My
point was really that you should be able to do the drag in reverse - ie.
start 'outside' and work inwards. Not that important, just slightly
asymmetric ;)

> > Print preview of the 'homed' plot is fine, but if zoomed in the boxplots
> > (and data) goes outside the plot frame.
> 
> Go outside the frame in the print preview?  There is a bug in
> gnome-print 0.20 that causes clipping to not work properly when
> "printing" to a gdk-pixbuf.  That might be the problem.  (It is fixed
> in CVS; the rumor is that gnome-print 0.21 will be released real soon
> now.)

That sounds like a reasonable explanation for what I'm seeing.

> > In fact, large amounts of:
> > ---
> > Gtk-WARNING **: invalid cast from `(unknown)' to `GuppiPlotCanvas'
> 
> Yikes... that isn't supposed to happen.  Can you tell me the specific
> sequence actions that trigger those warnings?

So far I've not managed to trigger those again, I just noticed them when
I went back to the xterm I loaded from...I'll let you know if I find
them again. I think it happened when I hit the home on the graph centre,
but I'm not sure specifically what I did before that.

Further point: in selection of size data, you cannot now select 'none'
once something has been chosen other than that?

-- 
Neil is awaiting the all-new configuration system ;)