Integrating simulation code with Jupyter

[uturuncoglu]

I am trying to prototype an idea and i just need your suggestion. I would like to integrate the numerical simulation (written in Fortran) with Jupyter notebook and try to explore the data structures (core data types such as arrays etc. and user derived ones that could be very complex) and simulation output while it is running. In this case, simulation is running in parallel using MPI directives. I use the in situ visualization approach before to analyze the simulation data but this is more generic. According to my current understanding, the Fortran code need to be wrapped by the C++ API and the new C++ API will interact with the Python and eventually Jupyter notebook. There might be also direct interaction between Fortran and Python but I’m not sure F2PY can provide that capability. If so, an example implementation would be great to look at it and understand. If not, the C++ interface will be responsible to translate Fortran data structures to Python ones and if we have user defined data types in the Fortran side, those also need to be defined in the Python side (may be using special Python module). So, it seems that the actual implementation will have different software layers and i am just trying to see the overall picture of the implementation. Anyway, if you have any suggestion or paper/books that you could point to me, that would be great.