[Paraview] Live data?

[Pat LeSmithe]

Thanks to both Kent and Berk for your replies.  I'll look into the
transient data support, although the code is probably a bit beyond my
understanding at this point.  Is it in a particular directory/function
sub-tree?

The primary use for me now is high-frequency "monitoring" of
"generalized sensors," e.g., processing and visualizing live data
streaming from experiments, network connections, and perhaps financial
markets.  I think a general description is multiple scalar data defined
on a dynamic 3D mesh.  In the simplest case, a fixed 2D sub-grid is
known in advance, and only "z" values of individual points get updated
(i.e., no dis/appearing acts).  Other scenarios may entail less
structure: connecting certain "dots" on the fly, moving surfaces, or
just scattering points.  I'm not sure that vector, tensor, or volume
plots will be relevant for me, although I imagine the latter will be
especially so for forthcoming applications of real-time medical imaging(?).

Since the information usually arrives from disparate sources at
irregular intervals over an indefinite period, the time steps are
neither well-defined nor countable in advance.  If necessary, I can
multiplex the signals into mesh-wide updates delivered periodically to
the visualization application, but low latency is important.  That is,
frame dropping is OK for the sake of soft real-time-ness.  I'm just
getting started with VTK, so I'm not sure about how it (and its
descendants) handles direct partial updates to a data set.  Although the
overall data rate in my case is not very high, it might rise
significantly in the future.

At this stage, I'm not sure about a proper format for visualizing
multi-dimensional time series with ParaView, VisIt, Mayavi, a custom VTK
wrapper, 3D graphics engine such as OGRE, etc.  I'm hoping to "serve"
the aggregated signal to the visualization pipeline either via a network
connection or from a single growing file.  Is HDF5 a possible way to go?
 Of course, I can start with a simple "streamable" human-readable or VTK
format.

By the way, is there an option in ParaView to automatically "Apply" all
changes to the pipeline?

Also, having found Octaviz (http://octaviz.sourceforge.net/), I'm
curious about the possibility of integrating popular scientific,
numerical, and statistical computing suites (e.g., Octave, R, SciPy,
NumPy) and high-level visualization environments (e.g., ParaView,
VisIT).  (Or at least greatly facilitating it, if licensing issues
arise.)  As I understand it, one issue is two-way communication of data
between the computing (in the sense of SciPy, etc.) and visualization
engines.  Perhaps the existing parallelization of the latter can even
help to parallelize the former.  Of course, another issue is writing
simple commands like plot3(), but already, Octaviz appears to go quite
far.  How about making the Python shell dockable, perhaps as a view?  If
this is old hat, I apologize.

One more: Is grid-enabled visualization on the horizon?

Thanks again.


Berk Geveci wrote:
>> Does ParaView now have the "animating live data" features discussed at
>>
>> http://paraview.org/Wiki/Animating_Live_Data  ?
> 
> We are working on something along these lines. Can you describe your
> use cases? What is the data source?
> 
> -berk