I must say, the code got a whole lot more complicated. If you think it became too unreadable, I'd be happy to discuss how we can improve it. But I felt like these checks were important to catch user errors and give meaningful feedback.

The problem is that the inputs are ambiguous, and it is not possible to tell what the userintends to do. For now this is solved by just looking at the first entry.

Concerning the pull request: Do you prefer if all commits are squashed into one, or will you take care of this when merging?

Not sure how the failed checks are related to my changes. Are those just flaky checks?

Sorry for the late reply.

Do you prefer if all commits are squashed into one, or will you take care of this when merging?

You are welcome to squash. If not, we'll do it on merge.

Not sure how the failed checks are related to my changes. Are those just flaky checks?

The CI runs have been deleted meanwhile. I assume the failures were unrelated. We had a couple of failing tests recently. Please try to rebase, that should likely fix it.

Thanks for your feedback! I added your remarks, squashed the commits and rebased onmain.
mypy failing appears to be unrelated.

Thanks! Had to useassert_array_almost_equal for all checks – I assume there are some slight rounding errors caused by floating-point arithmetic. Similar tests also use theassert_array_almost_equal which was already imported.

The last commit had some old unused parameter in one test case, causing pytest to complain about a missing fixture.

Sorry for the confusion. I amended that commit already but forgot to push 😅

"power cycled" to get CI to re-run against current main (which should fix the mypy failures).

Thanks@3j14 and congratulations on your first contribution to Matplotlib! 🎉