[Paraview] block merging using Paraview

[Julien Bodart]

Thanks for your help and sorry for the mailing list. I will
investigate the custom filters to get rid of the visual artefact. I
still have your solution. I was just afraid to miss an obvious way to
do it. Thanks again for your time.
Julien

2010/1/28, Andy Bauer <andy.bauer at kitware.com>:
> Please keep the discussions on the mailing list so that others can
> contribute and benefit from the information.
>
> I think the best solution is to create a custom filter that stitches the two
> rectilinear grids together in the way you want.  Most other solutions will
> convert your rectilinear grids to unstructured grids and thus use a lot more
> memory.
>
> Usually when you split a grid into multiple grids you'll get visual
> artifacts.
>
> Andy
>
> On Thu, Jan 28, 2010 at 12:12 PM, Julien Bodart
> <julien.bodart at gmail.com>wrote:
>
>> I am more working on the first option.
>> Actually those two files share a single uniform grid, so the space between
>> the two blocks is exactly the cell size. (so in this simple case it is
>> quite
>> huge). The reverse question works as well:
>> what happens if I split one single data file (with a rectilinear grid) in
>> two files (with no connections), just two subsets and two subgrids, how to
>> load them and see a "single grid".
>>
>> In my case loading the two files gives you all the grid nodes and then all
>> the needed information, but it looks like you have two different blocks.
>> In a finite volume point of view, maybe it means "adding one cell between
>> the two blocks", but if the information is on the nodes like here, nothing
>> more is needed.
>>
>> Your first solution works for me. I mean what you get using these simple
>> files is exactly what I want.
>> The problem there is the memory consumption (for example on my pc I can
>> open two regular files using just 22% of the memory, the MergeBlock
>> filters
>> goes to 66% and then every action make paraview crash). It looks "the
>> brute
>> force" for something I was expecting to be just tricky. (no extra nodes
>> needed)
>>
>> I hope it is more descriptive.
>>
>>
>> 2010/1/28 Andy Bauer <andy.bauer at kitware.com>
>>
>>> I'm still not completely understanding what you want done.  Since the two
>>> data sets have significant space between them do you want to fill in that
>>> space with extra cells or do you want to translate one of the data sets
>>> to
>>> get rid of the space between?  If you want to add in cells then it might
>>> be
>>> easiest to merge the data sets (GroupDataSet filter and then MergeBlocks
>>> filter like below) and then use the vtkDelauney3D filter to create a new
>>> grid that keeps the point data.
>>>
>>> If you want to translate one of the grids to get rid of the space between
>>> them you'll need to use something like the CleanToGrid filter to get an
>>> unstructured grid from your rectilinear grid.  Then  you can use the
>>> Transform filter to move it to the proper position (in this case
>>> -0.10000001489999999 in the z direction for test.2.nc since the
>>> vtkCleantoGridFilter is quite finicky).  Then use GroupDataSets filter to
>>> merge the data sets into a multiblock (right click on the GroupDataSets
>>> filter to select one of the test.*.nc files and the the Transform
>>> filter).
>>> Finally, you'll need to use the vtkCleantoGrid filter again to get the
>>> proper connectivity.  You'll want to check what it does with the point
>>> data
>>> at the merged points though.
>>>
>>> If neither of these is what you want then you'll need to be very
>>> descriptive of what  you hope to accomplish in order for me to help you
>>> any
>>> more.
>>>
>>> Andy
>>>
>>>
>>> On Thu, Jan 28, 2010 at 5:51 AM, Julien Bodart
>>> <julien.bodart at gmail.com>wrote:
>>>
>>>> Andy,
>>>>
>>>> Thanks for your help.
>>>> I have attached two simple files I am trying to merge. But looking at
>>>> the
>>>> problem, I am afraid you will tell me that there is no solution.
>>>> As you will see there is no duplicate points,  thus Paraview cannot
>>>> guess
>>>> it.
>>>> What is the best way to make them paraview-compliant? Add duplicate
>>>> cells?
>>>> Thanks.
>>>>
>>>> Julien
>>>>
>>>> 2010/1/27 Andy Bauer <andy.bauer at kitware.com>
>>>>
>>>>> To be clear, I meant first use the group data set filter or merge block
>>>>> filter to put all of the cells in the same data set and then use the
>>>>> clean
>>>>> to grid filter to get rid of duplicate points.  Points won't get merged
>>>>> if
>>>>> they are not within a certain distance from each other.  So again the
>>>>> real
>>>>> problem isn't that all of the cells aren't in the same data set but the
>>>>> fact
>>>>> that the grid connectivity is not correct.  If you are doing this in
>>>>> parallel this can also have an affect on appearance due to not having
>>>>> the
>>>>> correct connectivity between cells on different partitions/processors.
>>>>>
>>>>> If you can give me a more detailed description of the problem and send
>>>>> the grid I can take a closer look at it.
>>>>>
>>>>> Andy
>>>>>
>>>>>
>>>>> On Wed, Jan 27, 2010 at 10:04 AM, Julien Bodart <
>>>>> julien.bodart at gmail.com> wrote:
>>>>>
>>>>>> Well, actually I think I tried all the filters, included the clean to
>>>>>> grid one.
>>>>>> In my case there is actually no duplicate point between files:
>>>>>>
>>>>>> for example:
>>>>>>
>>>>>> file-1, z-coord: 0.0,0.1,0.2
>>>>>> file-2, z-coord: 0.3,0.4,0.5
>>>>>>
>>>>>> And the grid is structured.
>>>>>>
>>>>>> Maybe in this case there is no way to make a single dataset?
>>>>>>
>>>>>>
>>>>>>
>>>>>> 2010/1/27 Andy Bauer <andy.bauer at kitware.com>
>>>>>>
>>>>>>> Try using the clean to grid filter.  Although all of the cells are in
>>>>>>> the same data set (probably an unstructured grid), the filter does
>>>>>>> not
>>>>>>> realize that there are duplicate points and because of this the grid
>>>>>>> connectivity is probably not what you're expecting.
>>>>>>>
>>>>>>> Andy
>>>>>>>
>>>>>>> On Wed, Jan 27, 2010 at 8:03 AM, Julien Bodart <
>>>>>>> julien.bodart at gmail.com> wrote:
>>>>>>>
>>>>>>>> Hi,
>>>>>>>>
>>>>>>>> I am new to Paraview, thanks to the new Netcdf Reader of 3.6
>>>>>>>> release.
>>>>>>>> (Thank you very much for that)
>>>>>>>> Therefore my problem is probably really trivial but I can't get it
>>>>>>>> resolved...
>>>>>>>> I am reading 2 or more netcdf file describing several part of the
>>>>>>>> grid.
>>>>>>>> The complete domain is a cube(rectilinear grid), sliced in a given
>>>>>>>> number of files. There is no ghost cells so there is no obvious
>>>>>>>> match
>>>>>>>> between files.
>>>>>>>> When I try to merge the different block (using the group dataset
>>>>>>>> filter or merge block), I end up with a split domain, leading to a
>>>>>>>> visual
>>>>>>>> gap between each sub-domain, whether I am plotting contour ,
>>>>>>>> vertical plane
>>>>>>>> or whatever.
>>>>>>>> Is there a way to "really" merge those subdomain to end up with a
>>>>>>>> single domain.
>>>>>>>>
>>>>>>>> Thanks in advance!
>>>>>>>>
>>>>>>>> Julien
>>>>>>>>
>>>>>>>>
>>>>>>>> _______________________________________________
>>>>>>>> Powered by 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
>>>>>>>>
>>>>>>>>
>>>>>>>
>>>>>>
>>>>>
>>>>
>>>
>>
>