Object mode fallback deprecation in numba.jit

[bcbnz]

In the future, numba.jit will no longer fall back to object mode: https://numba.readthedocs.io/en/latest/reference/deprecation.html#deprecation-of-object-mode-fall-back-behaviour-when-using-jit Note that prominent deprecation warnings will be emitted in 0.57.0 (not yet released; I see them because I am currently using the latest git version of numba) and fallback support will be removed in 0.59.0.

The only place where @jit is used is in src/quaternion/calculus.py for fd_indefinite_integral:

quaternion/src/quaternion/calculus.py

Lines 309 to 316 in a34a20d

	@jit
	def fd_indefinite_integral(f, t):
	Sfdt = np.empty_like(f)
	Sfdt[0] = 0.0
	for i in xrange(1, len(t)):
	for j in xrange(f.shape[1]):
	Sfdt[i, j] = Sfdt[i - 1, j] + (f[i, j] + f[i - 1, j]) * ((t[i] - t[i - 1]) / 2.0)
	return Sfdt

All other JIT-compiled functions use @njit which disables object mode.

There are no tests which hit this function -- it is only used if SciPy is not available:

quaternion/src/quaternion/calculus.py

Lines 461 to 481 in a34a20d

	try:
	from scipy.interpolate import InterpolatedUnivariateSpline
	spline = spline_evaluation
	derivative = spline_derivative
	antiderivative = spline_indefinite_integral
	indefinite_integral = spline_indefinite_integral
	definite_integral = spline_definite_integral
	except ImportError:
	# import warnings
	# warning_text = (
	# "\n\n" + "!" * 57 + "\n" +
	# "Could not import from scipy, which means that derivatives\n" +
	# "and integrals will use less accurate finite-differencing\n" +
	# "techniques. You may want to install scipy." +
	# "\n" + "!" * 57 + "\n"
	# )
	# warnings.warn(warning_text)
	derivative = fd_derivative
	antiderivative = fd_indefinite_integral
	indefinite_integral = fd_indefinite_integral
	definite_integral = fd_definite_integral

I tried reusing test_antiderivative since fd_indefinite_integral is used to provide antiderivative if SciPy is not available, but it didn't accept the 1D input (and looking closer, test_antiderivative is really testing quaternion.calculus.spline not quaternion.calculus.antiderivative).

Unfortunately I am not familiar enough with quaternion integrals to write a suitable test to prove it, but I would guess that @jit could be replaced with @njit to future-proof the function.