The coding period officially started on May 27th, 2019. I have already discussed my project in detail in the previous posts and would just summarize it here. The goal of the project is to preserve the named metadata in
.geo files through to the tagged regions (MeshValueCollection) in dolfinx. This would help the end user to easily model with a large number of tagged regions.
A typical mesh workflow with FEniCS is as follows:
- Creation of geometric model: This task is carried out in a popular meshing software gmsh.
Conversion of geometric model: Then we convert the .msh file created by gmsh into XDMF file that could be read by DOLFIN using meshio.
- Assignment and Application: Next, we need to assign the material properties and apply loads and boundary condition. This is done with the help of “Mesh Function” and “Mesh Value Collection” in FEniCS.
The problem with the current implementation of the mesh workflow is that if a user has many different tagged regions in the mesh that s/he wishes to utilize in the simulation then there is no direct method of doing so. What was required by my GSoC project was to create a new mesh workflow that would seamlessly allow users to integrate different boundaries into their simulation. The agreed-upon method was to use the
<Information> tag of XDMF to carry the information about various tagged regions from
dolfin via meshio. So, in essence, we could keep the current process of working with a mesh and add methods to meshio and dolfinx.
The changes to be made to different parts of the mesh workflow is as follows:
- Creation of geometric model: There is no need for change in this part.
Conversion of geometric model: We need to update meshio to write the map of string tag to number tag in
<Information>element of the XDMF file. The required changes to meshio are done and here is the CircleCI build corresponding to the change.
- Assignment and Application:
The last part of the project was to add the capabilities to dolfin to read
<Information>tag data. A new function named read_tags() was added to the XDMFFile.cpp and the functionality of it was tested both in C++ as well as python version of the code. Here is the CircleCI build for the same.
A basic skeleton of the new mesh workflow has been prepared in the first phase. In the next phase of programming, I would try to make these methods much more robust and add further functionality.