[Paraview] Fwd: paraview making up numbers

[Berk Geveci]

---------- Forwarded message ----------
From: Berk Geveci <berk.geveci at kitware.com>
Date: Tue, Jul 8, 2008 at 10:51 AM
Subject: Re: [Paraview] paraview making up numbers
To: John Doe <ufnoise at gmail.com>


Let me try to answer your questions.

> 3. Different elements in the same region (lines, triangles), and
> having completely different data arrays per element type.

There is no file format that supports this because VTK/ParaView does
not support it. The best way of achieving it is to write a different
block per cell type that has different arrays.

> I tried this and it created a vtm file, a directory, and a file in the
> directory.  I could see places where there are a log of "nans" and
> 1e307's.  Can paraview read nan's and what does it do with them.

No. ParaView cannot read or handle nans. Nor does it support a concept
of invalid value. All values of an arrays have to be defined. The only
exception to this is the concept of "blanking" for structured grids
(which blanks a whole point/cell not mark a value invalid).

>  Is there anyway to create a flat file format for a vtm file?

No.

> Is the ensight format any easier, or just superior?  Any
> recommendations for a format appreciated.

Writing Ensight is definitely easier. Whether it is superior or not
depends on your use case.

> Requirements:
> 1. Different regions (or pieces) have different data arrays with
> completely unrelated names.
>
> Optional (But would be really, really nice):
> 1. Documented
> 2. Single file format

You should look into the following formats:

Ensight. See http://www.ensight.com/downloads/cat_view-82.html for
documentation.
* It is well documented
* Supports multiple blocks/parts
* Easy to read/write
* It is not a single file format

Exodus. See http://endo.sandia.gov/SEACAS/Documentation/exodusII.pdf
* It is well documented
* Supports multiple blocks/parts
* Requires a library to read/write
(http://sourceforge.net/projects/exodusii). It uses netcdf under the
covers so you can directly the netcdf library too
* It is a single file format

Xdmf: See xdmf.org
* It is somewhat well documented
* Supports multiple blocks/parts
* Requiers library (HDF5) to read/write
* Each dataset is usually 2 files (one xml, one hdf5)

-berk

> On Thu, Jul 3, 2008 at 1:38 PM, John Doe <ufnoise at gmail.com> wrote:
>> While I am learning about this new format.  Would it be reasonable for
>> me to file a request that data arrays be initialized to zero in
>> regions they are not specified?  If not possible, would it be more
>> reasonable to file a request that the software warns people that
>> paraview will not initialize data arrays?
>>
>> Thanks,
>>
>> Juan
>>
>> On Thu, Jul 3, 2008 at 1:23 PM, Berk Geveci <berk.geveci at kitware.com> wrote:
>>> No, I am afraid it did not make into the documentation yet. It is easy
>>> to create an example using ParaView. Load 2 unstructured grids, group
>>> them together using the Group filter and then write the output as a
>>> vtm file.
>>>
>>> -berk
>>>
>>> On Thu, Jul 3, 2008 at 12:35 PM, John Doe <ufnoise at gmail.com> wrote:
>>>> Is there any documentation for the multi-block format?  It is not part
>>>> of the file formats document I found online.
>>>>
>>>> Thanks,
>>>>
>>>> Juan
>>>>
>>>> On Wed, Jul 2, 2008 at 2:43 PM, Berk Geveci <berk.geveci at kitware.com> wrote:
>>>>> It is required that all pieces in an XML unstructured file have the
>>>>> exact same set of arrays. It is by coincidence that you are
>>>>> experiencing crashes. You need to use multi-block dataset to group
>>>>> together pieces with different fields. You may want to look at using
>>>>> the XML-based multi-block file format (vtm extension) or EnSight
>>>>> format for this. The former is not well documented yet, I am afraid.
>>>>>
>>>>> -berk
>>>>>
>>>>> On Tue, Jul 1, 2008 at 4:46 PM, John Doe <ufnoise at gmail.com> wrote:
>>>>>> I have devices composed of pieces using an unstructured mesh in the
>>>>>> xml file format.  Some of the pieces don't have data in them.  It
>>>>>> appears that paraview makes up numbers for these regions.  Sometimes
>>>>>> the numbers are unphysical, e.g. 1e+307.  Sometimes the numbers seem
>>>>>> to be based on data from adjacent pieces.  Is there any way to tell
>>>>>> paraview to use a reasonable values in regions for which data is not
>>>>>> presented.
>>>>>>
>>>>>> This becomes important as each region (piece) of my device can have
>>>>>> completely different physics.
>>>>>>
>>>>>> Thanks,
>>>>>>
>>>>>> Juan
>>>>>> _______________________________________________
>>>>>> ParaView mailing list
>>>>>> ParaView at paraview.org
>>>>>> http://www.paraview.org/mailman/listinfo/paraview
>>>>>>
>>>>>
>>>>
>>>
>>
>