Compare simulation data across all integrators

[Mikasa0503]

Summary

Add tests/unit/test_integrators.py and compare the complete sim.data pytree for every supported integrator against a separately stepped default-integrator Sim after two simulation steps. Require matching tree structures, compare floating-point arrays with rtol=0.02 and atol=1e-4, and use strict comparisons for array shapes/dtypes and non-floating arrays. Compare PRNG key data exactly.

Rebased onto main at 62a3146a408c5fd5d0c2451de22e29bd74a71b18, which includes the merged #134 fix. The review diff contains only the new test file.

Validation

On macOS arm64 / CPU:

• pixi run --locked -e tests python -m pytest -q tests/unit/test_integrators.py: 3 passed.
• MPLBACKEND=Agg pixi run --locked -e tests tests: 746 passed, 50 skipped, 12 deselected, 6 warnings. Rendering exclusions follow the project configuration.
• pixi run --locked -e tests ruff check: passed.
• pixi run --locked -e tests ruff format --check --diff: 153 files already formatted.
• git diff --check: passed.

Addresses #106.

[Copilot]

Copilot review overview

🟡 Changes recommended

Extend finiteness checks to cover the omitted state tensors.

Get a fresh assessment by requesting another Copilot review.

Review effort: Lite
Findings: 1

Open (1)

• Smoke test omits SimState fields from non-finite checks · New

What changed in this PR

Adds parameterized smoke tests covering all supported simulation integrators.

Changes:

• Tests two simulation steps for each integrator.
• Verifies step advancement and finite state tensors.

File	Summary
tests/​unit/​test_sim.py	Adds integrator smoke tests; force, torque, and rotor_vel are not included in finiteness checks.

---

💡 Add a code-review agent skill or configure MCP servers for context-aware, tailored reviews. Learn more in the docs.

[Copilot]

Copilot review overview

🟢 Approval recommended

No unresolved blocking issues were identified.

Review effort: Lite
Findings: None

Resolved since last review (1)

• Smoke test omits SimState fields from non-finite checks

[amacati]

Hi, sorry for the long wait, and thanks for contributing! test_sim.py is already extremely overloaded, and I think this is a great opportunity to split things up. Could you add this test in a new test_integrators.py file? Also, it would be great if the test always tested that the values in sim.data etc are approximately equal to the default integrator. Over two timesteps, there should not be a lot of relative deviation. See

crazyflow/tests/unit/test_sim.py

Lines 151 to 158 in d28ec70

	data = jax.tree.flatten_with_path(sim.data)[0]
	default_data = jax.tree.flatten(sim.default_data)[0]
	for i, (path, value) in enumerate(data):
	default_value = default_data[i]
	if isinstance(value, jnp.ndarray):
	array_compare_assert(value, default_value, name=path)
	else:
	assert value == default_value, f"{path} value mismatch"

for an example pattern

[amacati]

See my feedback in the comment above

[amacati]

Hi Hugo, #134 has been merged, so you should be able to pull, rebase off of main, and then we can merge this. By the way, what is your use-case for testing the integrators? Are you working on drone research and want to use the semi-implicit Euler integrator?

[amacati]

(Btw, this was closed automatically because the integration fix was merged. Just rebase and reopen)

[Mikasa0503]

Yes, I’m doing drone research and wanted to get more familiar with crazyflow by contributing tests.