[Paraview] Starting a batch animation in the middle -- fail

[Bill Sherman]

Hello,

I have been successfully using pvbatch to generate animated image
sequences of state files created with the ParaView GUI.  However,
I am not able to begin an animation in the middle -- which I
desperately need to do in order to parallelize my renderings.

I mentioned this before in July (and again in person at SC'14),
but haven't yet reached a solution.

The documentation for the WriteAnimation() method in "simple.py"
indicates that this is possible, but fails in two ways.

1) The WriteAnimation() method doesn't actually implement the two
most important parameters needed: SetStartFileCount() and
SetPlaybackTimeWindow().

2) Even doing the operation manually, SetPlaybackTimeWindow()
affects the time of time-varying data, but does not affect the
time of animation cues such as the cameras!  or translation
values, etc used to move objects around.

I've also experimented a lot with setting various values of the
GetAnimationScene() -- such as: StartTime, Duration (which seems
to have no affect whatsoever), and AnimationTime.


I created a tar file with an example state file and batch rendering
script:
	http://www.freevr.org/Downloads/pvanimationtest.tar.gz

The scene has three time-varying elements:
	* Annotated Time -- works
	* Time-varying data (meshes of digits 0-9) -- works
	* An arrow moving through a sphere -- fails

NOTE: camera moves also fail, basically in the same way as the
arrow.

So how the arrow fails is that even when I begin the animation
process in the middle, the arrow (and any camera moves) always
start as though it were the first frame.  By the time it gets
to animation-time 5.0, the arrow should be through the sphere,
but instead it appears to the left.


To run a test, just do:
	% pvbatch 3Drender.py

This will print out some information, and may print warnings about
bad interpolation -- these are fine because it's a result of the
interpolating between polygonal objects (the digits) that aren't
intended to be morphed.  The effect is sufficient that it works
to tell me that data-time is working.

Newly rendered frames go into the "Frames" directory.  There are
other directories of the form "Frames-exN", which are various
experiments I've run to figure out the problem.

The state file that I am using is "arrowspheredatai.pvsm" (which is
loaded in the Python script).


This is a most desperate hour.

Please help me O PV wizards, you are my only hope.

	Thanks in abundance,
	Bill

--
Bill Sherman
Sr. Technology Advisor
Advanced Visualization Lab
Pervasive Technology Inst
Indiana University
shermanw at indiana.edu