valentinaschueller / richards-swe-coupling Goto Github PK

Issue was related to my DUNE setup.

two places where dt is used at the moment:

• Groundwater.dt.value in the weak form
• Groundwater.scheme.model.dt

If one should be changed, both should be changed (use @property for this?). But: is Groundwater.scheme.model.dt actually used anywhere? It seems like an unused value, then we could get rid of it

see

• https://black.readthedocs.io/en/stable/guides/using_black_with_other_tools.html
• https://flake8.pycqa.org/en/latest/user/configuration.html
• https://pycqa.github.io/isort/docs/configuration/profiles.html

We should introduce a relaxation step in the code, either before sending height data or after receiving it. I don't think preCICE itself supports it so it should be part of river.py/ groundwater.py

I am unhappy with the way I am currently setting time step sizes, time windows, and simulation duration. For our analysis it should be possible:

• to easily switch those settings in the same place (although some of it is related to precice-config.xml and some of it to the solvers
• to enforce that both solvers use the same time step size
• to enforce that preCICE doesn't make my solvers use a different time step size than the one I set in the solvers

see https://precice.org/couple-your-code-implicit-coupling.html as a reference

basically a homogeneous Neumann BC¹ but should be labelled as such

At the end of the while loop in both Python scripts, I only call end_time_step() if precice.action_read_iteration_checkpoint() is required. This is in line with the preCICE tutorial. ¹
With this simple check, subcycling/multirate doesn't work properly. end_time_step() is called in two different cases:

• in each iteration for "intermediate" time steps
• only in the final iteration for the final time step of a time window

Ideally, I would like to save model output only in the final iteration for every time step. Maybe check https://precice.org/couple-your-code-time-step-sizes.html for this?

Originally posted by @valentinaschueller in #14 (comment)

• Follow https://precice.org/couple-your-code-preparing-your-solver.html for this.
• River should receive flux, groundwater should receive water height.
• use consistent naming for everything!

implicit coupling, same time step size

• save output data to CSV files
• save figures directly instead of showing them

... with minimum recompilation overhead

• is this possible without recompilation? (probably not)
• allow user to set value of these boundary conditions

Mostly to enable easier testing and scripting

The absolute value of analysis.get_S() and process_results.compute_convergence_rate() should be (almost) equal, especially for homogeneous Dirichlet boundary conditions.

Instead of storing the numpy arrays for the axis, height, and potential, store the data directly in a DataArray (not just when saving the output in the end). This should also allow us to easily deal with subcycling issues (#15), since the data is clearly linked to a (potentially already existing) time step

use argparse to have a proper command line interface.

I want simulations to be set up in a cleaner way to prevent inconsistencies between the Python executables and the preCICE configuration. This is related to #3, #4, #5, #6, #13.

Idea:

• Use a jinja2 template for the preCICE configuration files (see Benjamin's newer oscillator example implementation)
• The place where one defines the values for the jinja2 template will be the place where all other simulation settings can be selected
• Start out with a Python executable for this. Later, think about moving these settings into a YAML file?