[Paraview] WG: Data handling in ParaView

[Stefan.Melber at dlr.de]

________________________________________
Von: Melber-Wilkending, Stefan
Gesendet: Donnerstag, 1. September 2016 16:52
An: Moreland, Kenneth
Betreff: AW: [Paraview] Data handling in ParaView

Hi Kenneth,

after some search i found a bug in the code: one variable which holds the offset from cell to cell was one to small - so this leads to the effect the cells are "connected". Fixing this and going from "unstructured grid" to "polygon data" i can now write in one file and read that in paraview.

By the way the pics i had attached are the version "each cell has its own file" - which looks good because the offset error did not work there. The pics with the offset error and one file are so ugly you cant find any cell there.

Regarding the colors i found it is not a good idea to take the cell number itself as color - it gives to much colors. Instead i use the modulo-function to limit the color to e.g. 100 instead 1e5.

For polygon data how to i "color" the triangle instead of the points? Do you have an example file?

Best regards,

     Stefan
________________________________________
Von: Moreland, Kenneth [kmorel at sandia.gov]
Gesendet: Donnerstag, 1. September 2016 15:54
An: Melber-Wilkending, Stefan; paraview
Betreff: RE: [Paraview] Data handling in ParaView

Stefan,

The idea of identifying each "cell" by a scalar value should work just fine and be scalable.

However, I do not understand what the issue is that the n-th cell's last point is connected to the n+1-th cell's first point. If there is no connection. I honestly do not see anything wrong with the pictures you posted except that in d4.png it looks like the lines of some "cells" are drawn on top of other lines. But that is exactly what I would expect if the cells are adjacent to each other, which I would also expect.

The only advice I can think of is to define the scalar values on the triangles (what ParaView calls cells) instead of the points. This will generally be more robust.

-Ken

-----Original Message-----
From: ParaView [mailto:paraview-bounces at paraview.org] On Behalf Of Stefan Melber
Sent: Thursday, September 1, 2016 4:01 AM
To: paraview <paraview at paraview.org>
Subject: [EXTERNAL] [Paraview] Data handling in ParaView

Hi,


i have (maybe) a simple problem: For the visualization of the so called dual-cells of an unstructured grid i use currently the block-structured vtu-format. The data thereby consists of many complex cells which are itself consists of a number of triangles - see the picture d1 attached.
Therefore i have to write out EVERY cell in another vtu-file and group them with am vtm-file together. This way works well with a small number of cells (~1000 - as for the picture) however the aim is to go up to
100e6 cells. Depending on the numbers i got from 1000 cells this process will take ~70 days!

The bottleneck is clearly the huge number of files i have to write (and read in ParaView). So are there ideas to have another representation in only one file for such kind of data?

I did some experiments writing all cells in one vtk(vtu)-file and give them scalar values depending on the cell number. But then, the n-th cell (its last point) is connected to the n+1 (its first point) and so on.
Looks ugly if the n-th cell is not a direct neighbor to the n+1 cell - and that can be the case in my data. And - cutting then the resulting stuff in ParaView i did not get the cutting pictures (like d4) i like to have ...

Best regards and hoping for some nice ideas,


         Stefan