[Dune-devel] Updates on psurface convert

[Markus Blatt]

Hi Xiaoxue,

I must admit, I am totally confused now. The last time you reported
about your changes to psurface, you did that as advised by
cloning psurface, creating a branch, and comitting your changes. Oliver
answered to your blog entry on this list and commented and requested
some changes, see
http://lists.dune-project.org/pipermail/dune-devel/2013-July/000383.html
. Unfortunately, there was no response to his email (at
least until your new report).

On Thu, Jul 25, 2013 at 04:13:06PM -0400, Xiaoxue Gong wrote:
> I have write a new psurface-convert code to convert the psurface into vtu
> file format. The code (in new project dune-psurface-convert) is save is
> git://gitorious.org/dune-psurface-convert/dune-psurface-convert.git
> This is I have made use of  the vtu writer in /dune/grid/io/file/vtk/vtk
> writer.hh, instead of  the third party code. It is written in a more
> C++-like style.

It seems like you did address the comments of Oliver in this new
project. What I do not understand is why you did not apply these
changes to the repository that you already created like advised (xiaoxueiriss-psurface).

Instead, you create a brand new repository (dune-psurface-convert), which seems to include the
code of psurface, but is not a clone of it. This means it does not
have the commit history of psurface available. On my system the
complete history looks like this:

mblatt at legolas:~/src/dune/gsoc/iris/dune-psurface-convert$ git log
commit 67b3982d6ee7d0f2074be5749cc5ef73da6dc7ca
Author: irisxiaoxue <xiaoxue198912 at gmail.com>
Date:   Thu Jul 25 15:42:01 2013 -0400

    xi

(Nice commit comment by the way ;) I think you should definitely read
http://tbaggery.com/2008/04/19/a-note-about-git-commit-messages.html )

Now this poses a serious problem for our collaboration as nobody will
be able to easily apply your changes to the upstream psurface
repository. The cause of this is that git has no way of knowing what
part of your initial commit is made by you and what part is from the
upstream psurface repository. Additionally there are many
automatically created files in that commit, that should not be there
(e.g the symbolic link am pointing to
/home/xiaoxue/GCoS/0722/dune-common-2.2.1/am, autom4te.cache, etc.)

If this is the way it was supposed to be, then please explain your
reasoning for doing it this way. Otherwise you will have to transfer
your changes manually to your original xiaoxueiriss-psurface
repository. Please make sure that for a fresh checkout of the
repository  the only additional files (compared to the upstream
psurface repository) are really needed and not some automatically
created ones. Please do this fast as we otherwise have no means to
evaluate your changes.

Additionally please comment on what of the points of the original work
plan of Oliver
http://lists.dune-project.org/pipermail/dune-devel/2013-June/000356.html
you adressed until now.

> The code could deal with the cube.par.am and sphere.par.am. 

What does could mean?

> The out put is
> shown in the attachment. I have trouble reading the trice.par.am, with an
> error message of  "You're trying to read a file in the old-style AmiraMesh
> format.
> But the format is not supported anymore :-(".

I am afraid you have to deal with this alone, as I am no AmiraMesh
expert and Oliver is not available until August 11. Maybe a simple
websearch for "amira User's Guide" might help? If it does not we have
to postpone this issue until Oliver is back
> 
> The amiramesh have the following information
>  1) 'Patches'
>   2) 'BaseGridVertexCoords
>   3) 'BaseGridTriangles'
>   4) 'NumNodesAndParameterEdgesPerTriangle'
>   5) 'Nodes'
>    For each node, five floats are stored containing first the local
> position in
>    barycentric coordinates on the respective triangle and x, y, and
> z-coordinates of   the image position.
>   6) 'ParameterEdges'
>      All edges of the graphs on all triangles.  Each entry is a tuple of
>      two integers storing the indices of the two endnodes of the edge.
>      The counting starts anew for each triangle.  The three invented
>      corner nodes always have the indices 0, 1, and 2, respectively.
>   7) 'EdgePoints
> 
> I stored the vertices (2) and nodes (5) (node positions) into point
> coordinates. The Triangles (3) and Edges (6) are stored as cell in vtu.
> Patches(1) is stored as cell data. The imagine position is stored as the
> point data on nodes and vertices. I have not figure out how to store Edge
> Points yet.
> Do you think it is the right way to store the psurface information into vtu
> file?

Please add some more elaborate documentation on the format you used
for storing psurface information into vtk files to the
repository under directory doc. Additionally add documentation on how to
replicate your results. Once that and merge of the two repositories as
detailed above is done, I will evaluate it. Before that this is not
possible. Please also note that this is mandatory for the approaching
midterm evaluations. 

BTW you need to notify us once this is done.


Best regards,

Markus

-- 
Do you need more support with DUNE or HPC in general? 

Dr. Markus Blatt - HPC-Simulation-Software & Services http://www.dr-blatt.de
Hans-Bunte-Str. 8-10, 69123 Heidelberg, Germany
Tel.: +49 (0) 160 97590858  Fax: +49 (0)322 1108991658