Hi Ethan,
> I've been playing around with VisAD a bit, but I seriously think that I'm
> not getting anywhere by myself. I really need help. So if any of these sound
> too easy for you, please be patient. Thanks.
>
> I'm basically just trying to do a 3D plot, but don't really understand how
> VisAD works, particularly the data model. I've read the VisAD guide, but to
> no avail. Too much information which I don't understand, probably. I've also
> gone through some of the progams in the examples directory, but i can't find
> any documentation describing what each test program is trying to do, so not
> too much help there either.
Sorry for your problems. The root of the problem is that I
am better at writing software than at writing documentation
(although the length of the Guide proves that I try).
> >From what I know, there are basically 3 steps to creating a plot.
> 1) creating the data model
> 2) creating the mappings with ScalarMap()
> 3) creating the display itself
Correct.
> I'm really stuck at the data model part. So I hope to solve just the data
> model in this mail. Hopefully I'll have less problems...
>
> Basically I'm just displaying a 3D plot, and my data is a time series of
> various financial instruments. Below is some description of my data model.
> - I believe it would be (RealType.Time -> (value1, value2,value3))
> - valueX has no units.
> - valueX are all float/double values
> - valueX are sampled at equal time intervals
>
> I believe this is about as simple a model as it can get, but alas, how do I
> do it???
>
> >From the little understanding I have, I think creating the model should
> be...
>
> RealType time = {RealType.Time};
>
> RealType value1 = new RealType("value1", Null, Null);
> RealType value2 = new RealType("value2", Null, Null);
> RealType value3 = new RealType("value3", Null, Null);
> RealType[] values = {value1, value2, value3};
>
> RealTupleType time_type = new RealTupleType(time);
> RealTupleType valuestuple = new RealTupleType(values);
>
> FunctionType time_values = new FunctionType(time_type, valuestuple);
Good up to here.
> FlatField data = FlatField.makeField(time_values); // is this right ???
FlatField.makeField is only intended as a helper for our
examples/TestNN.java programs. Please don't use it. You'll
need to construct a FlatField explicitly, first constructing
a Set of time samples:
> Linear1DSet ??? // how & where do I create the Linear1DSet?
This will be a set of times. This is where you use DateTime.
For example, in examples/Test03.java we use:
// different time resolution test
// 2 May 99, 15:51:00
double start = new DateTime(1999, 122, 57060).getValue();
Set time_set
new Linear1DSet(time_type, start, start + 3000.0, ntimes);
You'd do something like this, where ntimes is the number of
time samples in your data. If your times are not evenly spaced,
then use a Gridded1DSet, also as in examples/Test03.java.
Once you have a Set, then construct your FlatField by:
FlatField data = new FlatField(time_values, time_set);
Then you will need to fill in the values for value1,
value2 and value3, as follows:
> And populating the model with data should be...
>
> double val1 = ...
> double val2 = ...
> double val3 = ...
> double[] valss = {val1, val2, val3};
>
> // add record to new_data
> RealTuple record = new RealTuple(valuestuple, values);
>
> index = ??? // can i use the DateTime class to convert a string of time
> to some index?
> new_data.setSample(index, record);
Instead, do this:
double[][] values = new double[3][ntimes];
for (int i=0; i<ntimes; i++) {
values[0][i] = value1 at time step i
values[0][i] = value2 at time step i
values[0][i] = value3 at time step i
}
data.setSamples(values);
> I'm sure there's something wrong somewhere, or things that I may have
> overlooked. Would really appreciate if anyone can help me out here. Thanks a
> million in advance.
Good luck, and please let the list know if you have further
problems (I am leaving for 3 weeks travel this Friday, so
hopefully others on the list can help while I am away).
> Also, below are some of the other things which I want to be able to do.
> Would these affect the data model in any way?
> - the FlatField will be written out to a .nc file at later time
The visad.data.netcdf.Plain.save() method should create a
netCDF file from your FlatField with no problems.
> - for scalar mappings, all the 4 variables can be assigned to any display
> types
> e.g. animation of v1 vs v2 vs v3, or time vs v2 vs v3 with v1 varying
> RGB
This should be no problem. You'll use statement like:
DisplayImplJ3D display = new DisplayImplJ3D("finance");
display.addMap(new ScalarMap(value1, Display.XAxis));
. . .
And don't forget to create a JFrame and add() display.getComponent()
to the JFrame or a sub-JPanel.
> - remote visualization of the data from another machine.
This isn't too difficult. See Test14.java and Test15.java
in visad/examples. You construct a DataReferenceImpl to your
data, and a RemoteDataReference from that, then link it to a
URL using a RemoteServerImpl. On the remote machine you
construct a DisplayImplJ3D, then a RemoteDisplayImpl from that,
retrieve the RemoteDataReference from the URL via the
RemoteServer, and link them using:
RemoteDisplayImpl.addReference(RemoteDataReference);
Cheers,
Bill
----------------------------------------------------------
Bill Hibbard, SSEC, 1225 W. Dayton St., Madison, WI 53706
hibbard@xxxxxxxxxxxxxxxxx 608-263-4427 fax: 608-263-6738
http://www.ssec.wisc.edu/~billh/vis.html
