You cannot select more than 25 topics Topics must start with a letter or number, can include dashes ('-') and can be up to 35 characters long.

512 lines
21 KiB
Python

6 years ago
# Created by Pearu Peterson, June 2003
from __future__ import division, print_function, absolute_import
import numpy as np
from numpy.testing import (assert_equal, assert_almost_equal, assert_array_equal,
assert_array_almost_equal, assert_allclose)
from scipy._lib._numpy_compat import suppress_warnings
from pytest import raises as assert_raises
from numpy import array, diff, linspace, meshgrid, ones, pi, shape
from scipy.interpolate.fitpack import bisplrep, bisplev
from scipy.interpolate.fitpack2 import (UnivariateSpline,
LSQUnivariateSpline, InterpolatedUnivariateSpline,
LSQBivariateSpline, SmoothBivariateSpline, RectBivariateSpline,
LSQSphereBivariateSpline, SmoothSphereBivariateSpline,
RectSphereBivariateSpline)
class TestUnivariateSpline(object):
def test_linear_constant(self):
x = [1,2,3]
y = [3,3,3]
lut = UnivariateSpline(x,y,k=1)
assert_array_almost_equal(lut.get_knots(),[1,3])
assert_array_almost_equal(lut.get_coeffs(),[3,3])
assert_almost_equal(lut.get_residual(),0.0)
assert_array_almost_equal(lut([1,1.5,2]),[3,3,3])
def test_preserve_shape(self):
x = [1, 2, 3]
y = [0, 2, 4]
lut = UnivariateSpline(x, y, k=1)
arg = 2
assert_equal(shape(arg), shape(lut(arg)))
assert_equal(shape(arg), shape(lut(arg, nu=1)))
arg = [1.5, 2, 2.5]
assert_equal(shape(arg), shape(lut(arg)))
assert_equal(shape(arg), shape(lut(arg, nu=1)))
def test_linear_1d(self):
x = [1,2,3]
y = [0,2,4]
lut = UnivariateSpline(x,y,k=1)
assert_array_almost_equal(lut.get_knots(),[1,3])
assert_array_almost_equal(lut.get_coeffs(),[0,4])
assert_almost_equal(lut.get_residual(),0.0)
assert_array_almost_equal(lut([1,1.5,2]),[0,1,2])
def test_subclassing(self):
# See #731
class ZeroSpline(UnivariateSpline):
def __call__(self, x):
return 0*array(x)
sp = ZeroSpline([1,2,3,4,5], [3,2,3,2,3], k=2)
assert_array_equal(sp([1.5, 2.5]), [0., 0.])
def test_empty_input(self):
# Test whether empty input returns an empty output. Ticket 1014
x = [1,3,5,7,9]
y = [0,4,9,12,21]
spl = UnivariateSpline(x, y, k=3)
assert_array_equal(spl([]), array([]))
def test_resize_regression(self):
"""Regression test for #1375."""
x = [-1., -0.65016502, -0.58856235, -0.26903553, -0.17370892,
-0.10011001, 0., 0.10011001, 0.17370892, 0.26903553, 0.58856235,
0.65016502, 1.]
y = [1.,0.62928599, 0.5797223, 0.39965815, 0.36322694, 0.3508061,
0.35214793, 0.3508061, 0.36322694, 0.39965815, 0.5797223,
0.62928599, 1.]
w = [1.00000000e+12, 6.88875973e+02, 4.89314737e+02, 4.26864807e+02,
6.07746770e+02, 4.51341444e+02, 3.17480210e+02, 4.51341444e+02,
6.07746770e+02, 4.26864807e+02, 4.89314737e+02, 6.88875973e+02,
1.00000000e+12]
spl = UnivariateSpline(x=x, y=y, w=w, s=None)
desired = array([0.35100374, 0.51715855, 0.87789547, 0.98719344])
assert_allclose(spl([0.1, 0.5, 0.9, 0.99]), desired, atol=5e-4)
def test_out_of_range_regression(self):
# Test different extrapolation modes. See ticket 3557
x = np.arange(5, dtype=float)
y = x**3
xp = linspace(-8, 13, 100)
xp_zeros = xp.copy()
xp_zeros[np.logical_or(xp_zeros < 0., xp_zeros > 4.)] = 0
xp_clip = xp.copy()
xp_clip[xp_clip < x[0]] = x[0]
xp_clip[xp_clip > x[-1]] = x[-1]
for cls in [UnivariateSpline, InterpolatedUnivariateSpline]:
spl = cls(x=x, y=y)
for ext in [0, 'extrapolate']:
assert_allclose(spl(xp, ext=ext), xp**3, atol=1e-16)
assert_allclose(cls(x, y, ext=ext)(xp), xp**3, atol=1e-16)
for ext in [1, 'zeros']:
assert_allclose(spl(xp, ext=ext), xp_zeros**3, atol=1e-16)
assert_allclose(cls(x, y, ext=ext)(xp), xp_zeros**3, atol=1e-16)
for ext in [2, 'raise']:
assert_raises(ValueError, spl, xp, **dict(ext=ext))
for ext in [3, 'const']:
assert_allclose(spl(xp, ext=ext), xp_clip**3, atol=1e-16)
assert_allclose(cls(x, y, ext=ext)(xp), xp_clip**3, atol=1e-16)
# also test LSQUnivariateSpline [which needs explicit knots]
t = spl.get_knots()[3:4] # interior knots w/ default k=3
spl = LSQUnivariateSpline(x, y, t)
assert_allclose(spl(xp, ext=0), xp**3, atol=1e-16)
assert_allclose(spl(xp, ext=1), xp_zeros**3, atol=1e-16)
assert_raises(ValueError, spl, xp, **dict(ext=2))
assert_allclose(spl(xp, ext=3), xp_clip**3, atol=1e-16)
# also make sure that unknown values for `ext` are caught early
for ext in [-1, 'unknown']:
spl = UnivariateSpline(x, y)
assert_raises(ValueError, spl, xp, **dict(ext=ext))
assert_raises(ValueError, UnivariateSpline,
**dict(x=x, y=y, ext=ext))
def test_lsq_fpchec(self):
xs = np.arange(100) * 1.
ys = np.arange(100) * 1.
knots = np.linspace(0, 99, 10)
bbox = (-1, 101)
assert_raises(ValueError, LSQUnivariateSpline, xs, ys, knots,
bbox=bbox)
def test_derivative_and_antiderivative(self):
# Thin wrappers to splder/splantider, so light smoke test only.
x = np.linspace(0, 1, 70)**3
y = np.cos(x)
spl = UnivariateSpline(x, y, s=0)
spl2 = spl.antiderivative(2).derivative(2)
assert_allclose(spl(0.3), spl2(0.3))
spl2 = spl.antiderivative(1)
assert_allclose(spl2(0.6) - spl2(0.2),
spl.integral(0.2, 0.6))
def test_nan(self):
# bail out early if the input data contains nans
x = np.arange(10, dtype=float)
y = x**3
w = np.ones_like(x)
# also test LSQUnivariateSpline [which needs explicit knots]
spl = UnivariateSpline(x, y, check_finite=True)
t = spl.get_knots()[3:4] # interior knots w/ default k=3
y_end = y[-1]
for z in [np.nan, np.inf, -np.inf]:
y[-1] = z
assert_raises(ValueError, UnivariateSpline,
**dict(x=x, y=y, check_finite=True))
assert_raises(ValueError, InterpolatedUnivariateSpline,
**dict(x=x, y=y, check_finite=True))
assert_raises(ValueError, LSQUnivariateSpline,
**dict(x=x, y=y, t=t, check_finite=True))
y[-1] = y_end # check valid y but invalid w
w[-1] = z
assert_raises(ValueError, UnivariateSpline,
**dict(x=x, y=y, w=w, check_finite=True))
assert_raises(ValueError, InterpolatedUnivariateSpline,
**dict(x=x, y=y, w=w, check_finite=True))
assert_raises(ValueError, LSQUnivariateSpline,
**dict(x=x, y=y, t=t, w=w, check_finite=True))
def test_increasing_x(self):
xx = np.arange(10, dtype=float)
yy = xx**3
x = np.arange(10, dtype=float)
x[1] = x[0]
y = x**3
w = np.ones_like(x)
# also test LSQUnivariateSpline [which needs explicit knots]
spl = UnivariateSpline(xx, yy, check_finite=True)
t = spl.get_knots()[3:4] # interior knots w/ default k=3
assert_raises(ValueError, UnivariateSpline,
**dict(x=x, y=y, check_finite=True))
assert_raises(ValueError, InterpolatedUnivariateSpline,
**dict(x=x, y=y, check_finite=True))
assert_raises(ValueError, LSQUnivariateSpline,
**dict(x=x, y=y, t=t, w=w, check_finite=True))
class TestLSQBivariateSpline(object):
# NOTE: The systems in this test class are rank-deficient
def test_linear_constant(self):
x = [1,1,1,2,2,2,3,3,3]
y = [1,2,3,1,2,3,1,2,3]
z = [3,3,3,3,3,3,3,3,3]
s = 0.1
tx = [1+s,3-s]
ty = [1+s,3-s]
with suppress_warnings() as sup:
r = sup.record(UserWarning, "\nThe coefficients of the spline")
lut = LSQBivariateSpline(x,y,z,tx,ty,kx=1,ky=1)
assert_equal(len(r), 1)
assert_almost_equal(lut(2,2), 3.)
def test_bilinearity(self):
x = [1,1,1,2,2,2,3,3,3]
y = [1,2,3,1,2,3,1,2,3]
z = [0,7,8,3,4,7,1,3,4]
s = 0.1
tx = [1+s,3-s]
ty = [1+s,3-s]
with suppress_warnings() as sup:
# This seems to fail (ier=1, see ticket 1642).
sup.filter(UserWarning, "\nThe coefficients of the spline")
lut = LSQBivariateSpline(x,y,z,tx,ty,kx=1,ky=1)
tx, ty = lut.get_knots()
for xa, xb in zip(tx[:-1], tx[1:]):
for ya, yb in zip(ty[:-1], ty[1:]):
for t in [0.1, 0.5, 0.9]:
for s in [0.3, 0.4, 0.7]:
xp = xa*(1-t) + xb*t
yp = ya*(1-s) + yb*s
zp = (+ lut(xa, ya)*(1-t)*(1-s)
+ lut(xb, ya)*t*(1-s)
+ lut(xa, yb)*(1-t)*s
+ lut(xb, yb)*t*s)
assert_almost_equal(lut(xp,yp), zp)
def test_integral(self):
x = [1,1,1,2,2,2,8,8,8]
y = [1,2,3,1,2,3,1,2,3]
z = array([0,7,8,3,4,7,1,3,4])
s = 0.1
tx = [1+s,3-s]
ty = [1+s,3-s]
with suppress_warnings() as sup:
r = sup.record(UserWarning, "\nThe coefficients of the spline")
lut = LSQBivariateSpline(x, y, z, tx, ty, kx=1, ky=1)
assert_equal(len(r), 1)
tx, ty = lut.get_knots()
tz = lut(tx, ty)
trpz = .25*(diff(tx)[:,None]*diff(ty)[None,:]
* (tz[:-1,:-1]+tz[1:,:-1]+tz[:-1,1:]+tz[1:,1:])).sum()
assert_almost_equal(lut.integral(tx[0], tx[-1], ty[0], ty[-1]),
trpz)
def test_empty_input(self):
# Test whether empty inputs returns an empty output. Ticket 1014
x = [1,1,1,2,2,2,3,3,3]
y = [1,2,3,1,2,3,1,2,3]
z = [3,3,3,3,3,3,3,3,3]
s = 0.1
tx = [1+s,3-s]
ty = [1+s,3-s]
with suppress_warnings() as sup:
r = sup.record(UserWarning, "\nThe coefficients of the spline")
lut = LSQBivariateSpline(x, y, z, tx, ty, kx=1, ky=1)
assert_equal(len(r), 1)
assert_array_equal(lut([], []), np.zeros((0,0)))
assert_array_equal(lut([], [], grid=False), np.zeros((0,)))
class TestSmoothBivariateSpline(object):
def test_linear_constant(self):
x = [1,1,1,2,2,2,3,3,3]
y = [1,2,3,1,2,3,1,2,3]
z = [3,3,3,3,3,3,3,3,3]
lut = SmoothBivariateSpline(x,y,z,kx=1,ky=1)
assert_array_almost_equal(lut.get_knots(),([1,1,3,3],[1,1,3,3]))
assert_array_almost_equal(lut.get_coeffs(),[3,3,3,3])
assert_almost_equal(lut.get_residual(),0.0)
assert_array_almost_equal(lut([1,1.5,2],[1,1.5]),[[3,3],[3,3],[3,3]])
def test_linear_1d(self):
x = [1,1,1,2,2,2,3,3,3]
y = [1,2,3,1,2,3,1,2,3]
z = [0,0,0,2,2,2,4,4,4]
lut = SmoothBivariateSpline(x,y,z,kx=1,ky=1)
assert_array_almost_equal(lut.get_knots(),([1,1,3,3],[1,1,3,3]))
assert_array_almost_equal(lut.get_coeffs(),[0,0,4,4])
assert_almost_equal(lut.get_residual(),0.0)
assert_array_almost_equal(lut([1,1.5,2],[1,1.5]),[[0,0],[1,1],[2,2]])
def test_integral(self):
x = [1,1,1,2,2,2,4,4,4]
y = [1,2,3,1,2,3,1,2,3]
z = array([0,7,8,3,4,7,1,3,4])
with suppress_warnings() as sup:
# This seems to fail (ier=1, see ticket 1642).
sup.filter(UserWarning, "\nThe required storage space")
lut = SmoothBivariateSpline(x, y, z, kx=1, ky=1, s=0)
tx = [1,2,4]
ty = [1,2,3]
tz = lut(tx, ty)
trpz = .25*(diff(tx)[:,None]*diff(ty)[None,:]
* (tz[:-1,:-1]+tz[1:,:-1]+tz[:-1,1:]+tz[1:,1:])).sum()
assert_almost_equal(lut.integral(tx[0], tx[-1], ty[0], ty[-1]), trpz)
lut2 = SmoothBivariateSpline(x, y, z, kx=2, ky=2, s=0)
assert_almost_equal(lut2.integral(tx[0], tx[-1], ty[0], ty[-1]), trpz,
decimal=0) # the quadratures give 23.75 and 23.85
tz = lut(tx[:-1], ty[:-1])
trpz = .25*(diff(tx[:-1])[:,None]*diff(ty[:-1])[None,:]
* (tz[:-1,:-1]+tz[1:,:-1]+tz[:-1,1:]+tz[1:,1:])).sum()
assert_almost_equal(lut.integral(tx[0], tx[-2], ty[0], ty[-2]), trpz)
def test_rerun_lwrk2_too_small(self):
# in this setting, lwrk2 is too small in the default run. Here we
# check for equality with the bisplrep/bisplev output because there,
# an automatic re-run of the spline representation is done if ier>10.
x = np.linspace(-2, 2, 80)
y = np.linspace(-2, 2, 80)
z = x + y
xi = np.linspace(-1, 1, 100)
yi = np.linspace(-2, 2, 100)
tck = bisplrep(x, y, z)
res1 = bisplev(xi, yi, tck)
interp_ = SmoothBivariateSpline(x, y, z)
res2 = interp_(xi, yi)
assert_almost_equal(res1, res2)
class TestLSQSphereBivariateSpline(object):
def setup_method(self):
# define the input data and coordinates
ntheta, nphi = 70, 90
theta = linspace(0.5/(ntheta - 1), 1 - 0.5/(ntheta - 1), ntheta) * pi
phi = linspace(0.5/(nphi - 1), 1 - 0.5/(nphi - 1), nphi) * 2. * pi
data = ones((theta.shape[0], phi.shape[0]))
# define knots and extract data values at the knots
knotst = theta[::5]
knotsp = phi[::5]
knotdata = data[::5, ::5]
# calculate spline coefficients
lats, lons = meshgrid(theta, phi)
lut_lsq = LSQSphereBivariateSpline(lats.ravel(), lons.ravel(),
data.T.ravel(), knotst, knotsp)
self.lut_lsq = lut_lsq
self.data = knotdata
self.new_lons, self.new_lats = knotsp, knotst
def test_linear_constant(self):
assert_almost_equal(self.lut_lsq.get_residual(), 0.0)
assert_array_almost_equal(self.lut_lsq(self.new_lats, self.new_lons),
self.data)
def test_empty_input(self):
assert_array_almost_equal(self.lut_lsq([], []), np.zeros((0,0)))
assert_array_almost_equal(self.lut_lsq([], [], grid=False), np.zeros((0,)))
class TestSmoothSphereBivariateSpline(object):
def setup_method(self):
theta = array([.25*pi, .25*pi, .25*pi, .5*pi, .5*pi, .5*pi, .75*pi,
.75*pi, .75*pi])
phi = array([.5 * pi, pi, 1.5 * pi, .5 * pi, pi, 1.5 * pi, .5 * pi, pi,
1.5 * pi])
r = array([3, 3, 3, 3, 3, 3, 3, 3, 3])
self.lut = SmoothSphereBivariateSpline(theta, phi, r, s=1E10)
def test_linear_constant(self):
assert_almost_equal(self.lut.get_residual(), 0.)
assert_array_almost_equal(self.lut([1, 1.5, 2],[1, 1.5]),
[[3, 3], [3, 3], [3, 3]])
def test_empty_input(self):
assert_array_almost_equal(self.lut([], []), np.zeros((0,0)))
assert_array_almost_equal(self.lut([], [], grid=False), np.zeros((0,)))
class TestRectBivariateSpline(object):
def test_defaults(self):
x = array([1,2,3,4,5])
y = array([1,2,3,4,5])
z = array([[1,2,1,2,1],[1,2,1,2,1],[1,2,3,2,1],[1,2,2,2,1],[1,2,1,2,1]])
lut = RectBivariateSpline(x,y,z)
assert_array_almost_equal(lut(x,y),z)
def test_evaluate(self):
x = array([1,2,3,4,5])
y = array([1,2,3,4,5])
z = array([[1,2,1,2,1],[1,2,1,2,1],[1,2,3,2,1],[1,2,2,2,1],[1,2,1,2,1]])
lut = RectBivariateSpline(x,y,z)
xi = [1, 2.3, 5.3, 0.5, 3.3, 1.2, 3]
yi = [1, 3.3, 1.2, 4.0, 5.0, 1.0, 3]
zi = lut.ev(xi, yi)
zi2 = array([lut(xp, yp)[0,0] for xp, yp in zip(xi, yi)])
assert_almost_equal(zi, zi2)
def test_derivatives_grid(self):
x = array([1,2,3,4,5])
y = array([1,2,3,4,5])
z = array([[1,2,1,2,1],[1,2,1,2,1],[1,2,3,2,1],[1,2,2,2,1],[1,2,1,2,1]])
dx = array([[0,0,-20,0,0],[0,0,13,0,0],[0,0,4,0,0],
[0,0,-11,0,0],[0,0,4,0,0]])/6.
dy = array([[4,-1,0,1,-4],[4,-1,0,1,-4],[0,1.5,0,-1.5,0],
[2,.25,0,-.25,-2],[4,-1,0,1,-4]])
dxdy = array([[40,-25,0,25,-40],[-26,16.25,0,-16.25,26],
[-8,5,0,-5,8],[22,-13.75,0,13.75,-22],[-8,5,0,-5,8]])/6.
lut = RectBivariateSpline(x,y,z)
assert_array_almost_equal(lut(x,y,dx=1),dx)
assert_array_almost_equal(lut(x,y,dy=1),dy)
assert_array_almost_equal(lut(x,y,dx=1,dy=1),dxdy)
def test_derivatives(self):
x = array([1,2,3,4,5])
y = array([1,2,3,4,5])
z = array([[1,2,1,2,1],[1,2,1,2,1],[1,2,3,2,1],[1,2,2,2,1],[1,2,1,2,1]])
dx = array([0,0,2./3,0,0])
dy = array([4,-1,0,-.25,-4])
dxdy = array([160,65,0,55,32])/24.
lut = RectBivariateSpline(x,y,z)
assert_array_almost_equal(lut(x,y,dx=1,grid=False),dx)
assert_array_almost_equal(lut(x,y,dy=1,grid=False),dy)
assert_array_almost_equal(lut(x,y,dx=1,dy=1,grid=False),dxdy)
def test_broadcast(self):
x = array([1,2,3,4,5])
y = array([1,2,3,4,5])
z = array([[1,2,1,2,1],[1,2,1,2,1],[1,2,3,2,1],[1,2,2,2,1],[1,2,1,2,1]])
lut = RectBivariateSpline(x,y,z)
assert_allclose(lut(x, y), lut(x[:,None], y[None,:], grid=False))
class TestRectSphereBivariateSpline(object):
def test_defaults(self):
y = linspace(0.01, 2*pi-0.01, 7)
x = linspace(0.01, pi-0.01, 7)
z = array([[1,2,1,2,1,2,1],[1,2,1,2,1,2,1],[1,2,3,2,1,2,1],
[1,2,2,2,1,2,1],[1,2,1,2,1,2,1],[1,2,2,2,1,2,1],
[1,2,1,2,1,2,1]])
lut = RectSphereBivariateSpline(x,y,z)
assert_array_almost_equal(lut(x,y),z)
def test_evaluate(self):
y = linspace(0.01, 2*pi-0.01, 7)
x = linspace(0.01, pi-0.01, 7)
z = array([[1,2,1,2,1,2,1],[1,2,1,2,1,2,1],[1,2,3,2,1,2,1],
[1,2,2,2,1,2,1],[1,2,1,2,1,2,1],[1,2,2,2,1,2,1],
[1,2,1,2,1,2,1]])
lut = RectSphereBivariateSpline(x,y,z)
yi = [0.2, 1, 2.3, 2.35, 3.0, 3.99, 5.25]
xi = [1.5, 0.4, 1.1, 0.45, 0.2345, 1., 0.0001]
zi = lut.ev(xi, yi)
zi2 = array([lut(xp, yp)[0,0] for xp, yp in zip(xi, yi)])
assert_almost_equal(zi, zi2)
def test_derivatives_grid(self):
y = linspace(0.01, 2*pi-0.01, 7)
x = linspace(0.01, pi-0.01, 7)
z = array([[1,2,1,2,1,2,1],[1,2,1,2,1,2,1],[1,2,3,2,1,2,1],
[1,2,2,2,1,2,1],[1,2,1,2,1,2,1],[1,2,2,2,1,2,1],
[1,2,1,2,1,2,1]])
lut = RectSphereBivariateSpline(x,y,z)
y = linspace(0.02, 2*pi-0.02, 7)
x = linspace(0.02, pi-0.02, 7)
assert_allclose(lut(x, y, dtheta=1), _numdiff_2d(lut, x, y, dx=1),
rtol=1e-4, atol=1e-4)
assert_allclose(lut(x, y, dphi=1), _numdiff_2d(lut, x, y, dy=1),
rtol=1e-4, atol=1e-4)
assert_allclose(lut(x, y, dtheta=1, dphi=1), _numdiff_2d(lut, x, y, dx=1, dy=1, eps=1e-6),
rtol=1e-3, atol=1e-3)
def test_derivatives(self):
y = linspace(0.01, 2*pi-0.01, 7)
x = linspace(0.01, pi-0.01, 7)
z = array([[1,2,1,2,1,2,1],[1,2,1,2,1,2,1],[1,2,3,2,1,2,1],
[1,2,2,2,1,2,1],[1,2,1,2,1,2,1],[1,2,2,2,1,2,1],
[1,2,1,2,1,2,1]])
lut = RectSphereBivariateSpline(x,y,z)
y = linspace(0.02, 2*pi-0.02, 7)
x = linspace(0.02, pi-0.02, 7)
assert_equal(lut(x, y, dtheta=1, grid=False).shape, x.shape)
assert_allclose(lut(x, y, dtheta=1, grid=False),
_numdiff_2d(lambda x,y: lut(x,y,grid=False), x, y, dx=1),
rtol=1e-4, atol=1e-4)
assert_allclose(lut(x, y, dphi=1, grid=False),
_numdiff_2d(lambda x,y: lut(x,y,grid=False), x, y, dy=1),
rtol=1e-4, atol=1e-4)
assert_allclose(lut(x, y, dtheta=1, dphi=1, grid=False),
_numdiff_2d(lambda x,y: lut(x,y,grid=False), x, y, dx=1, dy=1, eps=1e-6),
rtol=1e-3, atol=1e-3)
def _numdiff_2d(func, x, y, dx=0, dy=0, eps=1e-8):
if dx == 0 and dy == 0:
return func(x, y)
elif dx == 1 and dy == 0:
return (func(x + eps, y) - func(x - eps, y)) / (2*eps)
elif dx == 0 and dy == 1:
return (func(x, y + eps) - func(x, y - eps)) / (2*eps)
elif dx == 1 and dy == 1:
return (func(x + eps, y + eps) - func(x - eps, y + eps)
- func(x + eps, y - eps) + func(x - eps, y - eps)) / (2*eps)**2
else:
raise ValueError("invalid derivative order")