[vtkusers] Streamlines vs. Lagrangian drifters

[Prabhu Ramachandran]

hi,

>>>>> "moore" == christopher moore <aeolus at ocean.washington.edu> writes:

<snip>
    moore> Those of us who do oceanography or fluid dynamics might
    moore> benefit from a class like vtkStreamer that, instead of
    moore> rendering streamlines, animates the path of a drifting
    moore> particle.  Perhaps someone has done this already that might
    moore> help us out?  I was hoping for something with an accurate
    moore> time-stepping scheme - perhaps a Runge-Kutta (as opposed to
    moore> simple forward-euler)?  I'm a hack of a developer, but if
    moore> nobody else has done this I'll give it a try.  You'll have
    moore> to promise not to laugh at my code, though....

The Runge-Kutta InitialValueProblemSolver is already in the current
CVS source tree.  Both 2nd order and 4th order.

common/vtkRungeKutta2.h
common/vtkRungeKutta4.h

And you can set the type of integration desired in the vtkStreamer
using 

  // Description:
  // Set/get the integrator type to be used in the stream line
  // calculation. The object passed is not actually used but
  // is cloned with MakeObject by each thread/process in the
  // process of integration. The default is 2nd order Runge Kutta.
  vtkSetObjectMacro ( Integrator, vtkInitialValueProblemSolver );

However, as you say, a path line algorithm would be useful but it
would require a huge amount of data.  For each time step you would
require data for a field of vectors.  Which would be terribly
voluminous for any realistic data.  It should be possible to do
though.  Maybe this is another thing that would be ideal for
parallelization?

prabhu