Re: Object graphics axis

[davidf(at)dfanning.com (David Fanning)]

Karri Kaksonen (karri.kaksonen@picker.fi) writes:

> I just came back from Davids course about object graphics and it
> occured to me that a reason why positioning the axis objects is
> so difficult is that the range depends on the location of the
> axis.
> 
> The manual says that the range of the axis is set by a vector:
> 
> [-Xmin/(Xmax-Xmin), 1/(Xmax-Xmin)]
> 
> This may work if the length of the axis is 1.0 in normalized
> coordinates. In the course I chose the length to be 2.0 and in
> order to get it right I just tried out different values until
> I found it to be closer to:
> 
> [-1-Xmin*2/(Xmax-Xmin), 2/(Xmax-Xmin)]
> 
> I thought about this on my flight home last night and what I am
> afraid of is that the -1 in the first element may actually depend
> on where the axis is drawn on the screen. My location of the axis
> was at [-1.0, -1.0]. If this is the case then RSI should do
> something about it before version 5.3 comes out. Otherwise you have
> to update the range vector every time you reposition the axis.
> 
> I run some tests at home and found out that the Location-keyword
> sets two coordinates out of three. And the coordconv sets the
> third coordinate, range and the scale.
> 
> range = [-50 150]
> normalized_len = 1.6
> 
> Location = [not_valid y_pos z_pos]
> scale = normalized_len/(range[1]-range[0])
> Xcoordconv = [x_pos - range[0]*scale, scale]
> 
> This does not sound logical to me. I would rather see:
> 
> Location = [x_pos, y_pos, z_pos]
> Scale = normalized_len/(range[1]-range0])
> Offset = -range[0]*scale
> 
> Perhaps RSI could add "scale" and "offset" to make the objects behave
> like the user expects them to work.

I agree completely that this does not work like the
user expects it to work. You, Ben, me...everyone I know...
has found it incredibly difficult to understand. And the
IDL documentation is decidedly unhelpful. I found it
frustrating too that the NORM_COORD routine supplied
in the object/examples directory, and referred to in
the documentation, also only works with data that
starts at zero. In other words, with data created
with the ever popular DIST function. :-)

After spending literally hours trying to make *real*
data show up in object graphics plots I threw up my
hands and threw myself on the mercy of the IDL
programmers who wrote the code. They supplied me
with an algorithm, which with a few slight modifications
to make it actually work, reliably creates the
translation and scaling factors necessary to place
a real data range into a particular view.

The relevant piece of code is this:

   scale = [((position[0]*range[1])-(position[1]*range[0])) / $
    (range[1]-range[0]), (position[1]-position[0])/(range[1]-range[0])]

where "range" is a two-element vector with the minimum and
maximum data range, and "position" is a two-element vector
with the minimum and maximum location in the arbitrary 
coordinate space of the view. This is the code that
can be found in NORMALIZE:

   http://www.dfanning.com/programs/normalize.pro

Why this code hasn't found its way back to RSI is
beyond me, but I know for a fact that I can't 
write object programs without it. I've tried.

Cheers,

David

P.S. This is not to say you *can't* write object
programs without NORMALIZE. I've seen plenty of
object programs from RSI which seem to be working
just fine. I'm only suggesting that the people who
wrote those programs seem to be a hell of a lot 
smarter than I am. :-)

-- 
David Fanning, Ph.D.
Fanning Software Consulting
Phone: 970-221-0438 E-Mail: davidf@dfanning.com
Coyote's Guide to IDL Programming: http://www.dfanning.com/
Toll-Free IDL Book Orders: 1-888-461-0155