Needs unit tests

[sherm1]

In the remote past sd/fast was embedded in larger applications where there were plenty of unit tests. Now there should be some standalone unit tests to make sure future changes don't break anything.

At a minimum there should be tests that we can generate code in the supported languages and then compile and link the resulting code. After that, we should have tests that run the generated code and check the answers against some oracular source.

[DonyDominic]

Hey, I’d like to help with setting up the tests.

I had a couple of quick questions:

1. Do you have a preference for the framework — CTest or something else?

2. What language would you suggest for the test harness — Bash, C, or Python?

Open to your thoughts on what would fit best.

[sherm1]

Sounds great, thanks Dony. I am most familiar with Google's gtest framework so I'd prefer to use that. The test cases would then be in C++ but could call the generated C functions using extern "C". I'm also thinking about adding C++ directly as one of the sdfast output languages -- currently its old-style C and Fortran only.

[esba1ley]

While I'm not doing anything with gtest, I have been going through the examples, and I found something odd about example/c/sphere* . When you set up the initial condition for the inner sphere to drop vertically from the middle of the outer sphere, and only have rotation the inner sphere about the x axis, the inner ball bounces back and forth in the Y/Z plane as expected:

sphere_animation_x_rot_only.mp4

When there is rotation about the y axis, the ball continues to spin at the same rate, bouncing vertically:

sphere_animation_y_rot_only.mp4

When there is rotation about the z axis, however, on first impact, all rotation ceases and the inner sphere just bounces vertically.

sphere_animation_z_rot_only.mp4

I'd have expected to see the ball bounce in the X/Y plane for the z-spin case. Am I missing something?

Here is my sphere testing directory. You can build the sim in the build dir by typing

$ make

and you can run the sim and analysis Python script to plot and make the animation by:

$ make analysis

the Python package requirements are in requirements.txt, and you'll also need the ffmpeg tool to create the .mp4 animation file. I saved the test cases that generated the animations above in the directory test_cases

sphere.zip

[sherm1]

Love the animations! I would also expect the about-X and about-Z cases to be symmetric. Would be interesting to figure out why they aren't. It's been several decades since that example was written so it is probably as much of a mystery to me as to anyone else!

It's hard to tell from the animations whether the rotations are purely about the specified axes. I suppose that would be a first thing to check.