testing

[tvdbogert]

Let's think a little about testing before we start moving the testing code into the repo.

Do we want to organize that differently?

There are tests that we use during development, to see if everything is working correctly. Then there are also tests that are tools for the user, for instance, to inspect the maximum isometric moments, if they had altered any muscle properties.

For development, do we want unit tests, and an automated process that checks everything?

@dblana it would be good to hear your thoughts first, since your experience is more recent than mine

[dblana]

We have the following tests:

• Visualization
• Speed
• Moment arms
• Equilibrium
• Derivatives
• Isometric moment-angle curves
• Isokinetic moment-angular velocity curves
• Simulation with Matlab solver
• Simulation with real-time solver

For a user who is not making any changes to the model, the tests that might be useful are the visualization, speed and real-time simulation. The isometric/isokinetic curves and output from the simulation with the Matlab solver could be provided for the standard model, as they should look the same for all users.

If a user makes any changes to model parameters, they will need to run the Equilibrium test (so perhaps this is not really a "test") to generate a new passive equilibrium before they can run any simulations. They may also want to look at moment arms, isometric/isokinetic curves and simulation outputs.

I think that the only test that is just for model developers is the Derivatives test. Unit tests would be great, I'm not sure what a "unit" would be in our case. I'll need to look into this a bit 😬