Development plan refined on 1st June, 2024

[yungyuc]

On the completion of "add NASA plot3d mesh format loader" #333 and "euler1d gui and plotting area user experience enhancement" #308, we have the basic functionalities for the non-linear solver of the Euler equation. We are ready for integrating all features to the main GUI, and remove the proof-of-concept (experimental) code from the "App". After the integration, we will have a main GUI equipping the features of 1D plotting (like matplotlib), 3D viewing (Qt3D/OpenGL), basic parameter controlling widgets, Python console, and an information pane. The basic GUI layout could be like:

The "App" structure was used to house the proof-of-concept code, and it has finished the job. Having said that, we still need a house for the experimental code. The "App" structure itself should be revamped to provide a cleaner plugin system that has access to the main GUI features.

Underneath the GUI, there are three groups of development: (i) the CESE method, (ii) unstructured mesh generation, and (iii) the computing system. They should be reorganized and improved.

The project modmesh was originally derived from the master project solvcon. Once the GUI gets cleaned up, we will start to restructure the code in modmesh and assemble it back with solvcon.

The CESE method

For the CESE method, there are two sub-groups. The first is to improve the core solver and the boundary condition treatments of the 1D solver. The second is to port the two- and three-dimensional solver from the old solvcon code base to modmesh.

Mesh generation

There is not any existing code for generating the unstructured mesh generation. We need a plan for describing the geometry, implementing Delaunay triangulation, and improving the mesh.

Computing system

The computing system involves two major enhancements for the simple array: (i) tabular data structure for fast C++ and flexible Python and (ii) profiling.

Deployment

Reorganizing the GUI and developing the associated sub-systems will make solvcon a complete platform for solving conservation laws. But it is already complex and difficult to build. We need a streamlined flow for deployment and it may start with the most challenging platform: Windows. The startup work involves 3 things:

• Identify the essential binaries
  • Qt library: windeployqt.exe copy required qt libraries to the release folder.  See https://doc.qt.io/qt-6/windows-deployment.html.
  • Python: Python Windows embeddable package provides a compact python environment. See https://www.python.org/downloads/windows/.
• Use Cmake to automate the windows binary build process.
• Use Github Actions to make the nightly build.

Summary

I summarize all points in what follows:

• Redesign GUI and code structure
  • Reorganize the features
    • 1D plotting (like matplotlib)
      • Boundary-aware plotting (issue "Adjust the boundary to imporve the plotting" Adjust the boundary to improve the plotting #335)
    • 3D viewing (Qt3D/OpenGL)
    • Basic parameter controlling widgets
    • Python console
    • An information pane
  • Revamp the "App" to a cleaner plugin system
• The CESE method
  • One-dimensional CESE solver
    • Compare the differences between the NRBC treatment available in the literature (issue "Add more NRBC treatment in the 1D Euler solver" Add more NRBC treatment in the 1D Euler solver #330)
  • Port the two- and three-dimensional solver from the old solvcon code base
• Unstructured mesh generation
  • Prototype for geometry operation, Delaunay triangulation, and mesh improvement.
• Computing system
  • Tabular data structure, e.g., data frame (issue "Add Dataframe" Add Dataframe #332)
  • Profiler
    • Profiler serialization
      • Wrap the call profiler (pr "Quentin/wrap call profiler" Add Python wrap to CallProfiler #315)
      • Prototype a serializer of the call profiler (pr "Serialize/Deserialize the call profiler" Prototype JSON serializer for CallProfiler #323)
    • Profile data analyzer
• Deployment
  • Nightly build of modmesh on windows
    • Identify the essential binaries
      • Qt library: windeployqt.exe copy required qt libraries to the release folder.  See https://doc.qt.io/qt-6/windows-deployment.html.
      • Python: Python Windows embeddable package provides a compact python environment. See https://www.python.org/downloads/windows/.
    • Use Cmake to automate the windows binary build process.
    • Use Github Actions to make the nightly build.

[yungyuc]

I created the following projects to follow up the plan:

• viewer and GUI revamping (solvcon project 8)
• The CESE method
  • one-dimensional solver (solvcon project 11)
  • port 2/3D Euler solver to modmesh (solvcon project 9)
• tabular data processing (solvcon project 6), which is the model problem leading to code development for profiling
• modmesh nightly build (solvcon project 7)

Mesh generation is too pre-mature. A project should be created after we can make a clearer picture.