[vtk-developers] A heads up about Temporal classes

[John Biddiscombe]

You may safely ignore this email, but I just thought I'd send a quick 
heads up about a few changes I made. Below is the CVS log for 
TemporalInterpolator

And I also committed some fixes to ParticleTracer last week. Things are 
working quite smoothly for me now, so if anyone did want to try out the 
particle tracer. In principle, it should work. Particles no longer 
disappear etc and most bugs are fixed (NB most, not al)

The interesting part about the CVS commit I just made to interpolator is 
that I tried changing the algorithm type to vtkDataObjectAlgorithm as 
discussed (off list) around IEEE paper writing time and it worked ok. We 
may therefore consider that vtkTemporalDataSetAlgorithm is actually only 
necessary for special cases where we specifically want to export a 
vtkTemporalDataSet. The Cache/Interpolator/other classes could be simple 
DataObjectAlgorithms and all is well. An Advanced reader which can 'do 
hyperslabs' in time might benefit from being temporal. As discussed, 
using DataObjectAlgorithm is preferable to TemporalAlgorithm because 
then filters downstream can do type checking directly on the data 
(SetInputPortInformation actually) rather than having to know what is 
'inside'  a collection of temporal objects at execution time.

JB

CVS log
BUG: There was a nasty bug in the interpolator caused by using 
CopyStructure, and then setting the points with a new interpolated data 
array - this caused the points in the original data to be overwritten 
since the vtkPoints object was shallow copied from the first input. This 
is now fixed
ENH: Add a new Discrete mode of operation so that one can request 
interpolated data at particular steps. This is so that we can simulate a 
discrete time provider using interpolated values from some other 
provider. Some filters will not want inputs with continuous time, so we 
canmuse this discrete mode.
ENH:I've added a whole bunch of commented out code temporarily as I've 
been experimenting with changing the algorithm type from 
vtkTemporalDataSetAlgorithm to just vtkDataObjectAlgorithm and allowing 
the executive to do all the time/temporal management. It works as 
expected, but I'm going to leave it alone for a couple of weeks until 
I've investigated a couple of other temporal issues.


-- 
John Biddiscombe,                            email:biddisco @ cscs.ch
http://www.cscs.ch/about/BJohn.php
CSCS, Swiss National Supercomputing Centre  | Tel:  +41 (91) 610.82.07
Via Cantonale, 6928 Manno, Switzerland      | Fax:  +41 (91) 610.82.82