MantisBT - ParaView
View Issue Details
0015967ParaView(No Category)public2016-02-01 16:232016-08-12 10:00
Alan Scott 
Kitware Robot 
normalminorhave not tried
closedmoved 
 
5.1 
Sandia
incorrect functionality
0015967: Exodus reader fails for bad time_whole steps
The Exodus reader fails reading in time data if the 'time_whole' entry is bad. EnSight does read these files correctly, in the sense that you can actually see the time steps (even if they have incorrect times.). This is confusing and needs to be fixed.

Attaching a file that shows the problem. Open the dataset, and you will see that it looks like there is only one timestep in the dataset.

Linux, 4.4.0, local server. There are 100 timesteps in this dataset.
No tags attached.
? cube_e_1_CDL.e (123,760) 2016-02-01 16:23
https://www.vtk.org/Bug/file/10017/cube_e_1_CDL.e
Issue History
2016-02-01 16:23Alan ScottNew Issue
2016-02-01 16:23Alan ScottFile Added: cube_e_1_CDL.e
2016-02-01 16:24Alan ScottDescription Updatedbug_revision_view_page.php?rev_id=1497#r1497
2016-02-01 16:41Ken MorelandNote Added: 0035693
2016-02-01 16:53Alan ScottNote Added: 0035694
2016-02-01 16:53Alan ScottTarget Version => 5.1
2016-02-12 13:34Ken MorelandNote Added: 0035732
2016-08-12 10:00Kitware RobotNote Added: 0038989
2016-08-12 10:00Kitware RobotStatusbacklog => closed
2016-08-12 10:00Kitware RobotResolutionopen => moved
2016-08-12 10:00Kitware RobotAssigned To => Kitware Robot

Notes
(0035693)
Ken Moreland   
2016-02-01 16:41   
In the process of fixing this issue, I urge you to be careful to not break existing functionality. The problem is that ParaView is doing what should be the right thing, but it is given bad data.

I believe the issue with this file (and similar ones) is that all time steps are reported as being at time 0. (Note that this is technically a malformed Exodus file. According to the documentation, "The only restriction on the time values is that they must monotonically increase," which these are not.) ParaView sees that they are all at time zero and combines them. This is actually desired behavior because there are situations where we need to throw out overlapping data (particularly during simulation restarts).

Probably the Exodus reader will need to be changed to have an added mode where the time values are ignored and time steps are used in their place. The easiest implementation would be an "ignore time values" checkbox in the properties panel to turn on this behavior (off by default). A more user-friendly version would have a condition in the reader where if there was more than one time step and all time values are the same (or maybe all zero), then ignore the time values.
(0035694)
Alan Scott   
2016-02-01 16:53   
Absolutely amazing - after writing this bug up with one customer, a second just called me asking for this feature/bug fix. The reason is that when some Sierra codes don't converge, they dump their current state (that is the wording the customer used) to the Exodus dataset, using the last timestep. This means they write the last timestep twice - once as a good timestep, the second as the timestep of interest - the failed one.

I like Ken's idea about explicitly adding a "ignore time values" checkbox. As Ken stated, if the Exodus file breaks with the spec, lets make the user explicitly turn off time values.
(0035732)
Ken Moreland   
2016-02-12 13:34   
Recently on the mailing list there was a user that ran into a similar problem except with a sequence of "restart" files (http://paraview.markmail.org/thread/7bcj5facb7v7sddk [^]). Basically he had a sequence of Exodus files to be interpreted as a time series, but ParaView was merging them all together because they all had the same time value (because the time value was ignored when writing the files because it was a file sequence).

Anyway, the point is that the Exodus file series reader should also support this option.
(0038989)
Kitware Robot   
2016-08-12 10:00   
Resolving issue as `moved`.

This issue tracker is no longer used. Further discussion of this issue may take place in the current ParaView Issues page linked in the banner at the top of this page.