[Paraview] Particle trajectories with Paraview - How to to use an existing dataset (xml / hdf5) - error messages with "Particle Pathlines"

Thu Oct 21 07:01:29 EDT 2010

All you need do is remove the Collection "unset" Grid from around the actual grid. You've created a collection of 1 temporal grid, hence the multiblock structure.

JB

From: Oliver Fochler [mailto:fochler at th.physik.uni-frankfurt.de]
Sent: 21 October 2010 12:52
To: Biddiscombe, John A.
Cc: paraview at paraview.org
Subject: Re: [Paraview] Particle trajectories with Paraview - How to to use an existing dataset (xml / hdf5) - error messages with "Particle Pathlines"

Dear John, Dear Paraview users,

after having been abroad for a couple of days and thus busy with other stuff, I returned to my visualization issue.
I've given one of your solutions a try and discarded unused blocks from my data set, sticking just with one block. Unfortunately this very naive approach doesn't work. I would guess I'm still having a multi-block with incorrect time stamps - just that the multi-block now consists of only one block.

I think I'll be able to modify the routines that create the xdmf data sets. But I would need some hints on how the data structure should actually look like in order to work as expected.
Looking at the xmf-File (see attachments for full example) my data is now organized like this:
<Grid Name="euclidean position" GridType="Collection" CollectionType="Unset">
      <Grid Name="Event 1" GridType="Collection" CollectionType="Temporal">
        <Grid Name="Frame 1" GridType="Uniform">
          <Topology Name="Topology" TopologyType="Polyvertex" Dimensions="629"/>
          <Geometry Name="Geometry" GeometryType="XYZ">
            <DataItem NumberType="Float" Precision="4" Dimensions="629 3" Format="HDF">trajectoryTest2.h5:/Case/Event 1/Frame 1/euclidean position</DataItem>
          </Geometry>
          <Time TimeType="Single" Value="0.2"/>
          <Attribute Name="azimuthal angle" AttributeType="Scalar" Center="Node">
            <DataItem NumberType="Float" Precision="4" Dimensions="629" Format="HDF">trajectoryTest2.h5:/Case/Event 1/Frame 1/azimuthal angle</DataItem>
          </Attribute>
..............
So, to me it looks as though I've got a temporal collection and inside of this there are grids called "Frame x" with each of these representing one time step. The "Frame"-grid is stamped with a time as above (<Time TimeType="Single" Value="0.2"/>) and contains several points XYZ with some assigned attributes.
How should the structure look like in order to work as expected (and discussed below)? Should there be a time stamp associated with the temporal collection grid one level above?
Or is there a deeper problem with the information stored in the h5 file?

Thank you guys so much for helping me on this one. I really appreciate it!

Cheers,
Oliver

On 07/10/10 17:17, Oliver Fochler wrote:

Thank you so much for your very detailed answer!

I think, I'll go with solution 2 and try modifying our converter such

that it outputs the blocks to separate files. Or rather, in a first

attempt, I will simply discard blocks 2 through 5 completely, since I

don't really need this information right now anyway.

Best regards,

Oliver

On 07/10/10 08:52, Biddiscombe, John A. wrote:

Oliver,

I received your files (off list) and gave them a try. I see now what the problem is.

You've got a temporal collection, and inside this, is a multi-block set of particles. What is happening is that the pathline filter is being called once per block by the executive, but the DATA_TIME_STEPS key is stamped on the multi-block wrapper, and each time the pathline filter is executed on a sub-block, the sub-block is extracted from the collection and passed in - without - the DATA_TIME_STEPS key stamped on it.

Now this is fixable in two ways.

1) Have the executive stamp the sub-blocks with the wrapper's TIME key as they are extracted and passed in.

2) Have the Xdmf reader stamp the sub-blocks as they are generated (easier to do IMHO).

BUT.

Even if we make one of these fixes, I suspect the pathline filter will fail. This is because the pathline filter caches particle positions (w.r.t time) each time it is executed and builds a trail for each particle ID. If it is looped over a multiblock dataset - the executive reuses the same copy of the pathline filter for each block - so the pathline filter will get highly confused as it receives 5 executions for each time step - and the ID's present will be also occur 5 times (assuming you have Ids from 0-N for each block, if you have 0-5N and they are not duplicated, then it might just be OK).

Anyway. This is a serious problem as the pathline filter won't operate on the multiblock data. I'll add this to my list of 'one day I must fix this' - actually, I will soon need to use the pathline filter on multiblock data myself, so this is a good heads-up that I need to do something.

Here are ideas -

Solution 1: Doesn't work. If you use Extract-Block filter in ParaView, it just creates a new multiblock dataset with the unwanted blocks removed - it doesn't solve the problem of the sub-block stamping, you'd still need to fix 2) above in the reader. I have my own vtkFlattenDataset Filter (similar to extract blocks but removes the multiblock hierarchy) which I tried and it does work (for one block, but I didn't check for multiple to see if Ids are reused across blocks as my filter appends the datasets if >1 are taken). Unfortunately, this filter is not very well tested so not safe for public consumption.

Solution 2: If you can you split your blocks into separate Xdmf files so that you load 5 files instead of 1 (sorry, this is painful), then all will be ok. _This will at least work_

Solution 3: I could add multiblock support for the pathline filter, but this would take a spare afternoon/day which I don't have right now.

If I think of a solution 4, I'll post again.

JB

-----Original Message-----

From: paraview-bounces at paraview.org<mailto:paraview-bounces at paraview.org> [mailto:paraview-bounces at paraview.org] On Behalf Of Biddiscombe, John A.

Sent: 06 October 2010 15:30

To: Oliver Fochler; paraview at paraview.org<mailto:paraview at paraview.org>

Subject: Re: [Paraview] Particle trajectories with Paraview - How to to use an existing dataset (xml / hdf5) - error messages with "Particle Pathlines"

Oliver

The Xdmf file (or reader) must be not handling the time properly, can you send (or ftp if big) me xdmf and hdf files and I'll have a look at it. I use the pathline filter extensively and know it works ok.

JB

-----Original Message-----

From: paraview-bounces at paraview.org<mailto:paraview-bounces at paraview.org> [mailto:paraview-bounces at paraview.org] On Behalf Of Oliver Fochler

Sent: 06 October 2010 14:36

To: paraview at paraview.org<mailto:paraview at paraview.org>

Subject: [Paraview] Particle trajectories with Paraview - How to to use an existing dataset (xml / hdf5) - error messages with "Particle Pathlines"

Dear Paraview users,

I know that this topic has probably been discussed a couple of times.

But so far I haven't been able to get find any real solutions from the

discussion threads to my specific problem. I'm afraid I will need a

detailed dummy / newbie explanation..

What I would like to do:

----------------------------------------

Plot the trajectory of a moving particle, i.e. trace its path with some

sort of line. The position of the particle is given as x,y,z coordinates

(euclidean position) at certain distinct time steps.

The ability to fade out the track after n time steps would be nice. And

of course I need more than one particle, but that doesn't really matter.

The data format I have:

----------------------------------------

I'm using a converter written by someone else to convert the output of

my simulations to a format that is understood by paraview. In principle

I do have the source code of this converter, but I haven't really

bothered to look into it yet (and don't really have the time to do so now).

The output I get from this converter is hdf5, i.e. a xml file and a .h5

file. The data is organized in frames, one frame per time step. See

below for an excerpt from my xml file.

In principle this dataset just holds the positions and momenta of point

like particles (plus some more scalar values) at given time steps.

What I've done so far / Error messages:

--------------------------------------

Loading this dataset (the xml file) into paraview works fine. It

displays all the points and animating is no problem. Each frame is

correctly associated with the actual time (as denoted in the xml file)

and I can loop through the time steps to see the particles evolve,

inspect scalar properties with glyphs and so on.

But how can I get the trajectory of a particle? Or rather the polyline

connecting the coordinate points the particle has passed at the given

time steps.

I've tried using the "Particle Pathlines" filter but it gives me error

messages like

 "vtkCompositeDataPipeline (0x5c93630): Algorithm

vtkTemporalPathLineFilter(0x54ff590) returned failure for request:

vtkInformation (0x55b9160)"

for Paraview 3.6.2 or

"vtkTemporalPathLineFilter (0x4b7a680): The input dataset did not have a

valid DATA_TIME_STEPS information key"

and

"vtkCompositeDataPipeline (0x4b7c0c0): Algorithm

vtkTemporalPathLineFilter(0x4b7a680) returned failure for request:

vtkInformation"

for Paraview 3.8.0

I would be very glad if you had some suggestions!

Cheers,

Oliver

Here's the beginning of my xml file:

<?xml version="1.0" ?>

<!DOCTYPE Xdmf SYSTEM "Xdmf.dtd" []>

<Xdmf xmlns:xi="http://www.w3.org/2003/XInclude"<http://www.w3.org/2003/XInclude> Version="2.2"

Name="Case: BAMPS_test">

  <Domain Name="myDomain">

    <Grid Name="euclidean position" GridType="Collection"

CollectionType="Unset">

      <Grid Name="Event 1" GridType="Collection" CollectionType="Temporal">

        <Grid Name="Frame 1" GridType="Uniform">

          <Topology Name="Topology" TopologyType="Polyvertex"

Dimensions="1019"/>

          <Geometry Name="Geometry" GeometryType="XYZ">

            <DataItem NumberType="Float" Precision="4" Dimensions="1019

3" Format="HDF">BAMPS_test.h5:/Case/Event 1/Frame 1/euclidean

position</DataItem>

          </Geometry>

          <Time TimeType="Single" Value="0.2"/>

          <Attribute Name="azimuthal angle" AttributeType="Scalar"

Center="Node">

            <DataItem NumberType="Float" Precision="4" Dimensions="1019"

Format="HDF">BAMPS_test.h5:/Case/Event 1/Frame 1/azimuthal angle</DataItem>

          </Attribute>

          <Attribute Name="electric charge" AttributeType="Scalar"

Center="Node">

            <DataItem NumberType="Float" Precision="4" Dimensions="1019"

Format="HDF">BAMPS_test.h5:/Case/Event 1/Frame 1/electric charge</DataItem>

          </Attribute>

          <Attribute Name="energy" AttributeType="Scalar" Center="Node">

            <DataItem NumberType="Float" Precision="4" Dimensions="1019"

Format="HDF">BAMPS_test.h5:/Case/Event 1/Frame 1/energy</DataItem>

          </Attribute>

_______________________________________________

Powered by www.kitware.com<http://www.kitware.com>

Visit other Kitware open-source projects at http://www.kitware.com/opensource/opensource.html

Please keep messages on-topic and check the ParaView Wiki at: http://paraview.org/Wiki/ParaView

Follow this link to subscribe/unsubscribe:

http://www.paraview.org/mailman/listinfo/paraview

_______________________________________________

Powered by www.kitware.com<http://www.kitware.com>

Visit other Kitware open-source projects at http://www.kitware.com/opensource/opensource.html

Please keep messages on-topic and check the ParaView Wiki at: http://paraview.org/Wiki/ParaView

Follow this link to subscribe/unsubscribe:

http://www.paraview.org/mailman/listinfo/paraview

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://www.paraview.org/pipermail/paraview/attachments/20101021/4bbb79e1/attachment-0001.htm>