I looked over Test46/47.java which you said would have examples of code to
create glyph plots. But is there a way to create the glyph plot by just
creating mappings in the mapping dialog box from the visad GUI? I'm trying
to see what an input data file with a lot of data points looks like when it
is displayed in a glyph plot. Alternatively, how did you see the output of
Michelle Kam (408) 742-2881
Lockheed Martin Space Systems Co. SSM/ATC/MSIS
1111 Lockheed Martin Way, Sunnyvale, CA 94089
From: Bill Hibbard [mailto:billh@xxxxxxxxxxxxx]
Sent: Tuesday, July 08, 2003 3:51 AM
To: Michelle Kam
Subject: RE: .csv file
> The visad spreadsheet was able to load in my csv file that had 7 input
> when I modified the header to what you suggested. However, I am trying to
> produce a 3-D glyph plot in which 3 fields are mapped to 3 parameters such
> the x, y, and z axis, and 3 other fields mapped to 3 additional parameters
> local to that (x, y, z) point.
> However, when I load in my file with 7 input fields and create the
> for a glyph plot, I only see about 3 small points on different faces of
> cube rather than many points scattered throughout. Do you think the scarce
> number of plots could be from the values I gave each input field in my
> file (some possibly too small) or could it be from the function I set up
> the header?:
> (index) -> (GPA, Distance, Phone, Shoesize, Year, Xoffice, Yoffice,
> Firstname(Text), Lastname(Text)).
It is perfectly OK to have dependent variables mapped to
x, y, z, etc (anything but Animation or SelectValue, which
must be mapped from independent variables in 1-D domains).
If you want glyph plots, you'd do that with ScalarMaps to
Shape, which requires some code to build VisADGeometryArrays
and pass them to the setShapes() method of the ShapeControl
(see Test46.java and Test47.java in visad/examples). I don't
know why your points are all on the faces of your box - it may
just be the nature of your data - but its not because of
using dependent variables.
> I'm not exactly sure why these input fields like GPA and distance should
> dependent variables and if they are going to be dependent variables, are
> arbitrarily assigning the index to it?
Think of it this way, using a MathType of:
((GPA, Distance, Phone, Shoesize, Year, Xoffice, Yoffice) ->
says that name is a unique function of (GPA, Distance, Phone,
Shoesize, Year, Xoffice, Yoffice), which isn't necessarily so.
If everyone has a private office, then perhaps all the other
variables are unique functions of (Xoffice, Yoffice), in which
case you'd need an Irregular2DSet domain sampling for office
locations. But unless you're trying to visualize how shoesize
and GPA can be smoothly interpolated between offices (does GPA
drop off and shoesize increase in the executive suites?), you
don't need these to be independent variables.
Reminds me of a t-shirt I saw. It had a big C++ on it, and under
that in small letters it said "My boss's grade point average".