This PR handles date aspects on x-axis-related quantities. The stats refer to the y-axis and their unit does not matter in this PR. For simplicity and clarity, let's not use dates for them.

@timhoffm

return [{
       'coords': datetimes,
       'vals': [0.1, 0.5, 0.2],
       'mean': datetimes[1],
       'median': datetimes[1],
       'min': datetimes[0],
       'max': datetimes[-1],
       'quantiles': datetimes

The quantities are the y-axis values- that, and the mean, median, min and max values for the quantities spread is what you are asking me to change? Just wanted to confirm

Something like this:

   'coords': datetimes,
       'vals': [0.1, 0.5, 0.2],
       'mean': 0.5,
       'median': 0.5,
       'min': 0.1,
       'max': 0.2,
       'quantiles': [0.1, 0.5, 0.2]

Yes, but please do not use obviously nonsensical data (like max < mean).

This test creates a figure but doesn't close it. While matplotlib typically handles this, it's better practice to either use a pytest fixture that manages figure lifecycle, or explicitly close the figure to avoid resource leaks in the test suite. Consider adding plt.close(fig) after the violin call or using a context manager.

The error message "np.datetime64 'position' values, require np.timedelta64 'widths'" has the same grammatical issue as the previous error message. The comma after "values" creates a comma splice. Consider removing the comma or restructuring the message for better clarity and grammar.

Accessing positions[0] without checking if positions is non-empty could raise an IndexError. While the code validates that len(positions) == N on line 9052, if N (which equals len(vpstats)) is 0, positions could be an empty list. Consider adding a check for empty positions/vpstats or documenting that vpstats must be non-empty.

The variable width0 is assigned from widths when np.isscalar(widths) is True, but at this point in the code, widths has already been converted to a list on line 9057 when it was a scalar. This means width0 will always be assigned from widths[0], making the conditional logic incorrect. The check should use the original widths parameter before conversion.

The match string for pytest.raises is split across two lines without proper line continuation formatting. This could lead to issues if formatting changes or the line is refactored. Consider using proper line continuation with parentheses or joining into a single line for better maintainability and consistency with pytest best practices.

The error message "datetime/date 'position' values, require timedelta 'widths'" has incorrect grammar. The comma after "values" creates a comma splice. The message should either be: "datetime/date 'position' values require timedelta 'widths'" (remove comma) or "For datetime/date 'position' values, timedelta 'widths' are required" (restructure). Additionally, the message would be more helpful if it included an example as mentioned in the PR description.

While these tests verify that incorrect combinations of datetime positions with float widths raise errors, there's no test for the successful case where datetime positions are paired with timedelta widths (or np.datetime64 with np.timedelta64). Add positive test cases to ensure that the correct usage actually works and doesn't raise an error, similar to test_numeric_positions_with_float_widths_ok.

The test function name test_mixed_positions_datetime_and_numeric_behaves is unclear. The suffix "behaves" doesn't indicate what behavior is being tested. Consider renaming to test_mixed_positions_datetime_and_numeric_raises to be consistent with the other test names in this group and clearly indicate that this test expects an error to be raised.

The match string for pytest.raises is split across two lines without proper line continuation formatting. This could lead to issues if formatting changes or the line is refactored. Consider using proper line continuation with parentheses or joining into a single line for better maintainability and consistency with pytest best practices.

The match string for pytest.raises is split across two lines without proper line continuation formatting. This could lead to issues if formatting changes or the line is refactored. Consider using proper line continuation with parentheses or joining into a single line for better maintainability and consistency with pytest best practices.