Re: [Getfem-commits] please merge branch devel-tetsuo-xml

[Konstantinos Poulios]

The FOOTER_WRITTEN enumeration entry is not used anywhere, so it has to be removed.

Is it intentional that you did not support exporting slices in the vtu format? What prevents us from exporting slices?

For now, you can continue working on your current branch, and when the code is ready for merging we can move the final result to a new clean branch to merge to master.

Best regards

Kostas

On Sun, May 10, 2020 at 10:34 PM Konstantinos Poulios <address@hidden> wrote:

the other thing is the big overlapping between the vtk_export class and the new vtu_export class. Duplicating code is not a very good strategy. I think you should just keep the old vtk_export class and add an optional argument for the user to choose if exporting to xml format or not.

Best regards

Kostas

On Sun, May 10, 2020 at 10:23 PM Konstantinos Poulios <address@hidden> wrote:

by the way, the "remove_dof_unused" function you should make it work in-place on the passed by reference V vector, as was the code that you replaced with this function. The line

gmm::copy(remove_dof_unused(V, pmf_dof_used, Q), W)

and the temporary W vector are completely unnecessary, and just add redundant copy operations.

Best regards

Kostas

On Sun, May 10, 2020 at 10:16 PM Konstantinos Poulios <address@hidden> wrote:

Dear Tetsuo,

Thanks for your answer. Your code looks quite nice actually. I have one question about the lines

      std::vector<scalar_type> W(Q*pmf_dof_used.card());
      gmm::copy(remove_dof_unused(V, pmf_dof_used, Q), W);
      write_dataset_(V, name, qdim);

Since you do not do anything with vector W, what is the meaning of having it? Should the last line be:

      write_dataset_(W, name, qdim);

instead?

Apart from that, I think we need a bit cleaner workflow without too many unnecessary commits. I remember that I had advised you in the past against too large commits, but the ideal is somewhere in the middle. The commits must in general be organized in logical units from the perspective of someone looking at the git history. The work you have done here, I would put it into one or two commits. All the forth and back during the development, it doesn't need to be part of the repository history.

You can just make a new branch and put the outcome of your development in one or two commits and then we can merge that branch. Sorry for being picky, but establishing some good development habits will make our life easier in the future.

Best regards

Kostas

On Thu, May 7, 2020 at 3:23 PM Tetsuo Koyama <address@hidden> wrote:

P.S.
I had a typo

>That is a good point. It maybe a good idea of using library, but we
have to use cmake to link vtk librayr.
That is a good point. It maybe a good idea of using library, but we
have to use cmake to link vtk library.

2020年5月7日(木) 22:17 Tetsuo Koyama <address@hidden>:
>
> Dear Kostas
>
> Thank you very much for taking the time to review.
>
> > I think it is an important contribution to add vtu support, especially if it is binary/compressed, just ascii is not very useful.
> Thanks. I agree that binary/compressed is important. After this change
> is confirmed, I would like to add that option.
>
> > However we might need to discuss a bit on how to do it. As far as I can see you have used boost for xml writing. I think we
> > had dropped our dependency on boost and I am not very keen on reintroducing a dependency on boost.
> I agree with the policy that projects don't use boost. In the end, I
> made changes to eliminate the dependence on boost in the end. If there
> is any remaining dependence, please point out . I am sorry that the
> commit is complicated. The current vtu object does not require
> dependency to boost even if when extending to binaries.
>
> >Before we merge this, I would like to hear some arguments for one solution or another. The first thing to check is what others do.
> > How is vtu export implemented in other software like e.g. fenics? What is the more future-proof way of implementing vtu support?
> I didn't search fenics but meshio package (This is a major package
> which is used to convert mesh format. You can install by `apt install
> python3-meshio`) and mayavi2.
> Both are built by full scratches of writing text and binary like
> getfem project is doing. They reffer vtk file format pdf.
> https://vtk.org/wp-content/uploads/2015/04/file-formats.pdf
>
> > What is a solution with least dependencies? If we have to depend on an external library it might be better to depend
> > on vtk directly https://www.paraview.org/Wiki/VTK/Examples/Cxx/IO/WriteVTU
> That is a good point. It maybe a good idea of using library, but we
> have to use cmake to link vtk librayr. I think it is difficult to use
> with getfem using automake. (Is there any plan to use cmake in
> getfem?)
>
> This is hello world of vtk library.
> https://lorensen.github.io/VTKExamples/site/Cxx/GeometricObjects/CylinderExample/
>
> > Have you done some research regarding these questions?
> That is all. If I need I search of fenics I will do it !
>
> > There is also another thing that I would like to ask you about. Could you please don't use markup in your git commit description? It might look nice in your git client but it looks ugly and difficult to read on other's systems.
> Thank you for pointing it out. I used emoji prefix which is popular in
> my local. It is not good to use it in a global community. Sorry.
>
> > just to add that for compressed vtu files I use the attached conversion script based on binary vtk files exported from getfem.
> Thanks. I'll use it.
>
> Thank you for reading.
>
> BR Tetsuo
>
> 2020年5月7日(木) 19:21 Konstantinos Poulios <address@hidden>:
> >
> > just to add that for compressed vtu files I use the attached conversion script based on binary vtk files exported from getfem.
> >
> > On Thu, May 7, 2020 at 11:59 AM Konstantinos Poulios <address@hidden> wrote:
> >>
> >> Dear Tetsuo
> >>
> >> I think it is an important contribution to add vtu support, especially if it is binary/compressed, just ascii is not very useful. However we might need to discuss a bit on how to do it. As far as I can see you have used boost for xml writing. I think we had dropped our dependency on boost and I am not very keen on reintroducing a dependency on boost.
> >>
> >> Before we merge this, I would like to hear some arguments for one solution or another. The first thing to check is what others do. How is vtu export implemented in other software like e.g. fenics? What is the more future-proof way of implementing vtu support? What is a solution with least dependencies? If we have to depend on an external library it might be better to depend on vtk directly https://www.paraview.org/Wiki/VTK/Examples/Cxx/IO/WriteVTU
> >>
> >> Have you done some research regarding these questions?
> >>
> >> There is also another thing that I would like to ask you about. Could you please don't use markup in your git commit description? It might look nice in your git client but it looks ugly and difficult to read on other's systems.
> >>
> >> Best regards
> >> Kostas
> >>
> >>
> >>
> >>
> >> On Thu, May 7, 2020 at 2:07 AM Tetsuo Koyama <address@hidden> wrote:
> >>>
> >>> Dear getfem project
> >>>
> >>> Could you merge devel-tetsuo-xml?
> >>> This branch is addition of vtu_export class.
> >>> By using this class we can export xml unstructured grid format vtk
> >>> (only ascii format and write_point_data).
> >>> I tested it by using meshio package (https://github.com/nschloe/meshio).
> >>> In the future, the binary format and write_cell_data method may be extended.
> >>>
> >>> Thank you for reading.
> >>>
> >>> BR Tetsuo
> >>>