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

1061 lines
40 KiB

from sympy.core.random import randrange
from sympy.simplify.hyperexpand import (ShiftA, ShiftB, UnShiftA, UnShiftB,
MeijerShiftA, MeijerShiftB, MeijerShiftC, MeijerShiftD,
MeijerUnShiftA, MeijerUnShiftB, MeijerUnShiftC,
MeijerUnShiftD,
ReduceOrder, reduce_order, apply_operators,
devise_plan, make_derivative_operator, Formula,
hyperexpand, Hyper_Function, G_Function,
reduce_order_meijer,
build_hypergeometric_formula)
from sympy.concrete.summations import Sum
from sympy.core.containers import Tuple
from sympy.core.expr import Expr
from sympy.core.numbers import I
from sympy.core.singleton import S
from sympy.core.symbol import symbols
from sympy.functions.combinatorial.factorials import binomial
from sympy.functions.elementary.piecewise import Piecewise
from sympy.functions.special.hyper import (hyper, meijerg)
from sympy.abc import z, a, b, c
from sympy.testing.pytest import XFAIL, raises, slow, ON_CI, skip
from sympy.core.random import verify_numerically as tn
from sympy.core.numbers import (Rational, pi)
from sympy.functions.elementary.exponential import (exp, exp_polar, log)
from sympy.functions.elementary.hyperbolic import atanh
from sympy.functions.elementary.miscellaneous import sqrt
from sympy.functions.elementary.trigonometric import (asin, cos, sin)
from sympy.functions.special.bessel import besseli
from sympy.functions.special.error_functions import erf
from sympy.functions.special.gamma_functions import (gamma, lowergamma)
def test_branch_bug():
assert hyperexpand(hyper((Rational(-1, 3), S.Half), (Rational(2, 3), Rational(3, 2)), -z)) == \
-z**S('1/3')*lowergamma(exp_polar(I*pi)/3, z)/5 \
+ sqrt(pi)*erf(sqrt(z))/(5*sqrt(z))
assert hyperexpand(meijerg([Rational(7, 6), 1], [], [Rational(2, 3)], [Rational(1, 6), 0], z)) == \
2*z**S('2/3')*(2*sqrt(pi)*erf(sqrt(z))/sqrt(z) - 2*lowergamma(
Rational(2, 3), z)/z**S('2/3'))*gamma(Rational(2, 3))/gamma(Rational(5, 3))
def test_hyperexpand():
# Luke, Y. L. (1969), The Special Functions and Their Approximations,
# Volume 1, section 6.2
assert hyperexpand(hyper([], [], z)) == exp(z)
assert hyperexpand(hyper([1, 1], [2], -z)*z) == log(1 + z)
assert hyperexpand(hyper([], [S.Half], -z**2/4)) == cos(z)
assert hyperexpand(z*hyper([], [S('3/2')], -z**2/4)) == sin(z)
assert hyperexpand(hyper([S('1/2'), S('1/2')], [S('3/2')], z**2)*z) \
== asin(z)
assert isinstance(Sum(binomial(2, z)*z**2, (z, 0, a)).doit(), Expr)
def can_do(ap, bq, numerical=True, div=1, lowerplane=False):
r = hyperexpand(hyper(ap, bq, z))
if r.has(hyper):
return False
if not numerical:
return True
repl = {}
randsyms = r.free_symbols - {z}
while randsyms:
# Only randomly generated parameters are checked.
for n, ai in enumerate(randsyms):
repl[ai] = randcplx(n)/div
if not any(b.is_Integer and b <= 0 for b in Tuple(*bq).subs(repl)):
break
[a, b, c, d] = [2, -1, 3, 1]
if lowerplane:
[a, b, c, d] = [2, -2, 3, -1]
return tn(
hyper(ap, bq, z).subs(repl),
r.replace(exp_polar, exp).subs(repl),
z, a=a, b=b, c=c, d=d)
def test_roach():
# Kelly B. Roach. Meijer G Function Representations.
# Section "Gallery"
assert can_do([S.Half], [Rational(9, 2)])
assert can_do([], [1, Rational(5, 2), 4])
assert can_do([Rational(-1, 2), 1, 2], [3, 4])
assert can_do([Rational(1, 3)], [Rational(-2, 3), Rational(-1, 2), S.Half, 1])
assert can_do([Rational(-3, 2), Rational(-1, 2)], [Rational(-5, 2), 1])
assert can_do([Rational(-3, 2), ], [Rational(-1, 2), S.Half]) # shine-integral
assert can_do([Rational(-3, 2), Rational(-1, 2)], [2]) # elliptic integrals
@XFAIL
def test_roach_fail():
assert can_do([Rational(-1, 2), 1], [Rational(1, 4), S.Half, Rational(3, 4)]) # PFDD
assert can_do([Rational(3, 2)], [Rational(5, 2), 5]) # struve function
assert can_do([Rational(-1, 2), S.Half, 1], [Rational(3, 2), Rational(5, 2)]) # polylog, pfdd
assert can_do([1, 2, 3], [S.Half, 4]) # XXX ?
assert can_do([S.Half], [Rational(-1, 3), Rational(-1, 2), Rational(-2, 3)]) # PFDD ?
# For the long table tests, see end of file
def test_polynomial():
from sympy.core.numbers import oo
assert hyperexpand(hyper([], [-1], z)) is oo
assert hyperexpand(hyper([-2], [-1], z)) is oo
assert hyperexpand(hyper([0, 0], [-1], z)) == 1
assert can_do([-5, -2, randcplx(), randcplx()], [-10, randcplx()])
assert hyperexpand(hyper((-1, 1), (-2,), z)) == 1 + z/2
def test_hyperexpand_bases():
assert hyperexpand(hyper([2], [a], z)) == \
a + z**(-a + 1)*(-a**2 + 3*a + z*(a - 1) - 2)*exp(z)* \
lowergamma(a - 1, z) - 1
# TODO [a+1, aRational(-1, 2)], [2*a]
assert hyperexpand(hyper([1, 2], [3], z)) == -2/z - 2*log(-z + 1)/z**2
assert hyperexpand(hyper([S.Half, 2], [Rational(3, 2)], z)) == \
-1/(2*z - 2) + atanh(sqrt(z))/sqrt(z)/2
assert hyperexpand(hyper([S.Half, S.Half], [Rational(5, 2)], z)) == \
(-3*z + 3)/4/(z*sqrt(-z + 1)) \
+ (6*z - 3)*asin(sqrt(z))/(4*z**Rational(3, 2))
assert hyperexpand(hyper([1, 2], [Rational(3, 2)], z)) == -1/(2*z - 2) \
- asin(sqrt(z))/(sqrt(z)*(2*z - 2)*sqrt(-z + 1))
assert hyperexpand(hyper([Rational(-1, 2) - 1, 1, 2], [S.Half, 3], z)) == \
sqrt(z)*(z*Rational(6, 7) - Rational(6, 5))*atanh(sqrt(z)) \
+ (-30*z**2 + 32*z - 6)/35/z - 6*log(-z + 1)/(35*z**2)
assert hyperexpand(hyper([1 + S.Half, 1, 1], [2, 2], z)) == \
-4*log(sqrt(-z + 1)/2 + S.Half)/z
# TODO hyperexpand(hyper([a], [2*a + 1], z))
# TODO [S.Half, a], [Rational(3, 2), a+1]
assert hyperexpand(hyper([2], [b, 1], z)) == \
z**(-b/2 + S.Half)*besseli(b - 1, 2*sqrt(z))*gamma(b) \
+ z**(-b/2 + 1)*besseli(b, 2*sqrt(z))*gamma(b)
# TODO [a], [a - S.Half, 2*a]
def test_hyperexpand_parametric():
assert hyperexpand(hyper([a, S.Half + a], [S.Half], z)) \
== (1 + sqrt(z))**(-2*a)/2 + (1 - sqrt(z))**(-2*a)/2
assert hyperexpand(hyper([a, Rational(-1, 2) + a], [2*a], z)) \
== 2**(2*a - 1)*((-z + 1)**S.Half + 1)**(-2*a + 1)
def test_shifted_sum():
from sympy.simplify.simplify import simplify
assert simplify(hyperexpand(z**4*hyper([2], [3, S('3/2')], -z**2))) \
== z*sin(2*z) + (-z**2 + S.Half)*cos(2*z) - S.Half
def _randrat():
""" Steer clear of integers. """
return S(randrange(25) + 10)/50
def randcplx(offset=-1):
""" Polys is not good with real coefficients. """
return _randrat() + I*_randrat() + I*(1 + offset)
@slow
def test_formulae():
from sympy.simplify.hyperexpand import FormulaCollection
formulae = FormulaCollection().formulae
for formula in formulae:
h = formula.func(formula.z)
rep = {}
for n, sym in enumerate(formula.symbols):
rep[sym] = randcplx(n)
# NOTE hyperexpand returns truly branched functions. We know we are
# on the main sheet, but numerical evaluation can still go wrong
# (e.g. if exp_polar cannot be evalf'd).
# Just replace all exp_polar by exp, this usually works.
# first test if the closed-form is actually correct
h = h.subs(rep)
closed_form = formula.closed_form.subs(rep).rewrite('nonrepsmall')
z = formula.z
assert tn(h, closed_form.replace(exp_polar, exp), z)
# now test the computed matrix
cl = (formula.C * formula.B)[0].subs(rep).rewrite('nonrepsmall')
assert tn(closed_form.replace(
exp_polar, exp), cl.replace(exp_polar, exp), z)
deriv1 = z*formula.B.applyfunc(lambda t: t.rewrite(
'nonrepsmall')).diff(z)
deriv2 = formula.M * formula.B
for d1, d2 in zip(deriv1, deriv2):
assert tn(d1.subs(rep).replace(exp_polar, exp),
d2.subs(rep).rewrite('nonrepsmall').replace(exp_polar, exp), z)
def test_meijerg_formulae():
from sympy.simplify.hyperexpand import MeijerFormulaCollection
formulae = MeijerFormulaCollection().formulae
for sig in formulae:
for formula in formulae[sig]:
g = meijerg(formula.func.an, formula.func.ap,
formula.func.bm, formula.func.bq,
formula.z)
rep = {}
for sym in formula.symbols:
rep[sym] = randcplx()
# first test if the closed-form is actually correct
g = g.subs(rep)
closed_form = formula.closed_form.subs(rep)
z = formula.z
assert tn(g, closed_form, z)
# now test the computed matrix
cl = (formula.C * formula.B)[0].subs(rep)
assert tn(closed_form, cl, z)
deriv1 = z*formula.B.diff(z)
deriv2 = formula.M * formula.B
for d1, d2 in zip(deriv1, deriv2):
assert tn(d1.subs(rep), d2.subs(rep), z)
def op(f):
return z*f.diff(z)
def test_plan():
assert devise_plan(Hyper_Function([0], ()),
Hyper_Function([0], ()), z) == []
with raises(ValueError):
devise_plan(Hyper_Function([1], ()), Hyper_Function((), ()), z)
with raises(ValueError):
devise_plan(Hyper_Function([2], [1]), Hyper_Function([2], [2]), z)
with raises(ValueError):
devise_plan(Hyper_Function([2], []), Hyper_Function([S("1/2")], []), z)
# We cannot use pi/(10000 + n) because polys is insanely slow.
a1, a2, b1 = (randcplx(n) for n in range(3))
b1 += 2*I
h = hyper([a1, a2], [b1], z)
h2 = hyper((a1 + 1, a2), [b1], z)
assert tn(apply_operators(h,
devise_plan(Hyper_Function((a1 + 1, a2), [b1]),
Hyper_Function((a1, a2), [b1]), z), op),
h2, z)
h2 = hyper((a1 + 1, a2 - 1), [b1], z)
assert tn(apply_operators(h,
devise_plan(Hyper_Function((a1 + 1, a2 - 1), [b1]),
Hyper_Function((a1, a2), [b1]), z), op),
h2, z)
def test_plan_derivatives():
a1, a2, a3 = 1, 2, S('1/2')
b1, b2 = 3, S('5/2')
h = Hyper_Function((a1, a2, a3), (b1, b2))
h2 = Hyper_Function((a1 + 1, a2 + 1, a3 + 2), (b1 + 1, b2 + 1))
ops = devise_plan(h2, h, z)
f = Formula(h, z, h(z), [])
deriv = make_derivative_operator(f.M, z)
assert tn((apply_operators(f.C, ops, deriv)*f.B)[0], h2(z), z)
h2 = Hyper_Function((a1, a2 - 1, a3 - 2), (b1 - 1, b2 - 1))
ops = devise_plan(h2, h, z)
assert tn((apply_operators(f.C, ops, deriv)*f.B)[0], h2(z), z)
def test_reduction_operators():
a1, a2, b1 = (randcplx(n) for n in range(3))
h = hyper([a1], [b1], z)
assert ReduceOrder(2, 0) is None
assert ReduceOrder(2, -1) is None
assert ReduceOrder(1, S('1/2')) is None
h2 = hyper((a1, a2), (b1, a2), z)
assert tn(ReduceOrder(a2, a2).apply(h, op), h2, z)
h2 = hyper((a1, a2 + 1), (b1, a2), z)
assert tn(ReduceOrder(a2 + 1, a2).apply(h, op), h2, z)
h2 = hyper((a2 + 4, a1), (b1, a2), z)
assert tn(ReduceOrder(a2 + 4, a2).apply(h, op), h2, z)
# test several step order reduction
ap = (a2 + 4, a1, b1 + 1)
bq = (a2, b1, b1)
func, ops = reduce_order(Hyper_Function(ap, bq))
assert func.ap == (a1,)
assert func.bq == (b1,)
assert tn(apply_operators(h, ops, op), hyper(ap, bq, z), z)
def test_shift_operators():
a1, a2, b1, b2, b3 = (randcplx(n) for n in range(5))
h = hyper((a1, a2), (b1, b2, b3), z)
raises(ValueError, lambda: ShiftA(0))
raises(ValueError, lambda: ShiftB(1))
assert tn(ShiftA(a1).apply(h, op), hyper((a1 + 1, a2), (b1, b2, b3), z), z)
assert tn(ShiftA(a2).apply(h, op), hyper((a1, a2 + 1), (b1, b2, b3), z), z)
assert tn(ShiftB(b1).apply(h, op), hyper((a1, a2), (b1 - 1, b2, b3), z), z)
assert tn(ShiftB(b2).apply(h, op), hyper((a1, a2), (b1, b2 - 1, b3), z), z)
assert tn(ShiftB(b3).apply(h, op), hyper((a1, a2), (b1, b2, b3 - 1), z), z)
def test_ushift_operators():
a1, a2, b1, b2, b3 = (randcplx(n) for n in range(5))
h = hyper((a1, a2), (b1, b2, b3), z)
raises(ValueError, lambda: UnShiftA((1,), (), 0, z))
raises(ValueError, lambda: UnShiftB((), (-1,), 0, z))
raises(ValueError, lambda: UnShiftA((1,), (0, -1, 1), 0, z))
raises(ValueError, lambda: UnShiftB((0, 1), (1,), 0, z))
s = UnShiftA((a1, a2), (b1, b2, b3), 0, z)
assert tn(s.apply(h, op), hyper((a1 - 1, a2), (b1, b2, b3), z), z)
s = UnShiftA((a1, a2), (b1, b2, b3), 1, z)
assert tn(s.apply(h, op), hyper((a1, a2 - 1), (b1, b2, b3), z), z)
s = UnShiftB((a1, a2), (b1, b2, b3), 0, z)
assert tn(s.apply(h, op), hyper((a1, a2), (b1 + 1, b2, b3), z), z)
s = UnShiftB((a1, a2), (b1, b2, b3), 1, z)
assert tn(s.apply(h, op), hyper((a1, a2), (b1, b2 + 1, b3), z), z)
s = UnShiftB((a1, a2), (b1, b2, b3), 2, z)
assert tn(s.apply(h, op), hyper((a1, a2), (b1, b2, b3 + 1), z), z)
def can_do_meijer(a1, a2, b1, b2, numeric=True):
"""
This helper function tries to hyperexpand() the meijer g-function
corresponding to the parameters a1, a2, b1, b2.
It returns False if this expansion still contains g-functions.
If numeric is True, it also tests the so-obtained formula numerically
(at random values) and returns False if the test fails.
Else it returns True.
"""
from sympy.core.function import expand
from sympy.functions.elementary.complexes import unpolarify
r = hyperexpand(meijerg(a1, a2, b1, b2, z))
if r.has(meijerg):
return False
# NOTE hyperexpand() returns a truly branched function, whereas numerical
# evaluation only works on the main branch. Since we are evaluating on
# the main branch, this should not be a problem, but expressions like
# exp_polar(I*pi/2*x)**a are evaluated incorrectly. We thus have to get
# rid of them. The expand heuristically does this...
r = unpolarify(expand(r, force=True, power_base=True, power_exp=False,
mul=False, log=False, multinomial=False, basic=False))
if not numeric:
return True
repl = {}
for n, ai in enumerate(meijerg(a1, a2, b1, b2, z).free_symbols - {z}):
repl[ai] = randcplx(n)
return tn(meijerg(a1, a2, b1, b2, z).subs(repl), r.subs(repl), z)
@slow
def test_meijerg_expand():
from sympy.simplify.gammasimp import gammasimp
from sympy.simplify.simplify import simplify
# from mpmath docs
assert hyperexpand(meijerg([[], []], [[0], []], -z)) == exp(z)
assert hyperexpand(meijerg([[1, 1], []], [[1], [0]], z)) == \
log(z + 1)
assert hyperexpand(meijerg([[1, 1], []], [[1], [1]], z)) == \
z/(z + 1)
assert hyperexpand(meijerg([[], []], [[S.Half], [0]], (z/2)**2)) \
== sin(z)/sqrt(pi)
assert hyperexpand(meijerg([[], []], [[0], [S.Half]], (z/2)**2)) \
== cos(z)/sqrt(pi)
assert can_do_meijer([], [a], [a - 1, a - S.Half], [])
assert can_do_meijer([], [], [a/2], [-a/2], False) # branches...
assert can_do_meijer([a], [b], [a], [b, a - 1])
# wikipedia
assert hyperexpand(meijerg([1], [], [], [0], z)) == \
Piecewise((0, abs(z) < 1), (1, abs(1/z) < 1),
(meijerg([1], [], [], [0], z), True))
assert hyperexpand(meijerg([], [1], [0], [], z)) == \
Piecewise((1, abs(z) < 1), (0, abs(1/z) < 1),
(meijerg([], [1], [0], [], z), True))
# The Special Functions and their Approximations
assert can_do_meijer([], [], [a + b/2], [a, a - b/2, a + S.Half])
assert can_do_meijer(
[], [], [a], [b], False) # branches only agree for small z
assert can_do_meijer([], [S.Half], [a], [-a])
assert can_do_meijer([], [], [a, b], [])
assert can_do_meijer([], [], [a, b], [])
assert can_do_meijer([], [], [a, a + S.Half], [b, b + S.Half])
assert can_do_meijer([], [], [a, -a], [0, S.Half], False) # dito
assert can_do_meijer([], [], [a, a + S.Half, b, b + S.Half], [])
assert can_do_meijer([S.Half], [], [0], [a, -a])
assert can_do_meijer([S.Half], [], [a], [0, -a], False) # dito
assert can_do_meijer([], [a - S.Half], [a, b], [a - S.Half], False)
assert can_do_meijer([], [a + S.Half], [a + b, a - b, a], [], False)
assert can_do_meijer([a + S.Half], [], [b, 2*a - b, a], [], False)
# This for example is actually zero.
assert can_do_meijer([], [], [], [a, b])
# Testing a bug:
assert hyperexpand(meijerg([0, 2], [], [], [-1, 1], z)) == \
Piecewise((0, abs(z) < 1),
(z*(1 - 1/z**2)/2, abs(1/z) < 1),
(meijerg([0, 2], [], [], [-1, 1], z), True))
# Test that the simplest possible answer is returned:
assert gammasimp(simplify(hyperexpand(
meijerg([1], [1 - a], [-a/2, -a/2 + S.Half], [], 1/z)))) == \
-2*sqrt(pi)*(sqrt(z + 1) + 1)**a/a
# Test that hyper is returned
assert hyperexpand(meijerg([1], [], [a], [0, 0], z)) == hyper(
(a,), (a + 1, a + 1), z*exp_polar(I*pi))*z**a*gamma(a)/gamma(a + 1)**2
# Test place option
f = meijerg(((0, 1), ()), ((S.Half,), (0,)), z**2)
assert hyperexpand(f) == sqrt(pi)/sqrt(1 + z**(-2))
assert hyperexpand(f, place=0) == sqrt(pi)*z/sqrt(z**2 + 1)
def test_meijerg_lookup():
from sympy.functions.special.error_functions import (Ci, Si)
from sympy.functions.special.gamma_functions import uppergamma
assert hyperexpand(meijerg([a], [], [b, a], [], z)) == \
z**b*exp(z)*gamma(-a + b + 1)*uppergamma(a - b, z)
assert hyperexpand(meijerg([0], [], [0, 0], [], z)) == \
exp(z)*uppergamma(0, z)
assert can_do_meijer([a], [], [b, a + 1], [])
assert can_do_meijer([a], [], [b + 2, a], [])
assert can_do_meijer([a], [], [b - 2, a], [])
assert hyperexpand(meijerg([a], [], [a, a, a - S.Half], [], z)) == \
-sqrt(pi)*z**(a - S.Half)*(2*cos(2*sqrt(z))*(Si(2*sqrt(z)) - pi/2)
- 2*sin(2*sqrt(z))*Ci(2*sqrt(z))) == \
hyperexpand(meijerg([a], [], [a, a - S.Half, a], [], z)) == \
hyperexpand(meijerg([a], [], [a - S.Half, a, a], [], z))
assert can_do_meijer([a - 1], [], [a + 2, a - Rational(3, 2), a + 1], [])
@XFAIL
def test_meijerg_expand_fail():
# These basically test hyper([], [1/2 - a, 1/2 + 1, 1/2], z),
# which is *very* messy. But since the meijer g actually yields a
# sum of bessel functions, things can sometimes be simplified a lot and
# are then put into tables...
assert can_do_meijer([], [], [a + S.Half], [a, a - b/2, a + b/2])
assert can_do_meijer([], [], [0, S.Half], [a, -a])
assert can_do_meijer([], [], [3*a - S.Half, a, -a - S.Half], [a - S.Half])
assert can_do_meijer([], [], [0, a - S.Half, -a - S.Half], [S.Half])
assert can_do_meijer([], [], [a, b + S.Half, b], [2*b - a])
assert can_do_meijer([], [], [a, b + S.Half, b, 2*b - a])
assert can_do_meijer([S.Half], [], [-a, a], [0])
@slow
def test_meijerg():
# carefully set up the parameters.
# NOTE: this used to fail sometimes. I believe it is fixed, but if you
# hit an inexplicable test failure here, please let me know the seed.
a1, a2 = (randcplx(n) - 5*I - n*I for n in range(2))
b1, b2 = (randcplx(n) + 5*I + n*I for n in range(2))
b3, b4, b5, a3, a4, a5 = (randcplx() for n in range(6))
g = meijerg([a1], [a3, a4], [b1], [b3, b4], z)
assert ReduceOrder.meijer_minus(3, 4) is None
assert ReduceOrder.meijer_plus(4, 3) is None
g2 = meijerg([a1, a2], [a3, a4], [b1], [b3, b4, a2], z)
assert tn(ReduceOrder.meijer_plus(a2, a2).apply(g, op), g2, z)
g2 = meijerg([a1, a2], [a3, a4], [b1], [b3, b4, a2 + 1], z)
assert tn(ReduceOrder.meijer_plus(a2, a2 + 1).apply(g, op), g2, z)
g2 = meijerg([a1, a2 - 1], [a3, a4], [b1], [b3, b4, a2 + 2], z)
assert tn(ReduceOrder.meijer_plus(a2 - 1, a2 + 2).apply(g, op), g2, z)
g2 = meijerg([a1], [a3, a4, b2 - 1], [b1, b2 + 2], [b3, b4], z)
assert tn(ReduceOrder.meijer_minus(
b2 + 2, b2 - 1).apply(g, op), g2, z, tol=1e-6)
# test several-step reduction
an = [a1, a2]
bq = [b3, b4, a2 + 1]
ap = [a3, a4, b2 - 1]
bm = [b1, b2 + 1]
niq, ops = reduce_order_meijer(G_Function(an, ap, bm, bq))
assert niq.an == (a1,)
assert set(niq.ap) == {a3, a4}
assert niq.bm == (b1,)
assert set(niq.bq) == {b3, b4}
assert tn(apply_operators(g, ops, op), meijerg(an, ap, bm, bq, z), z)
def test_meijerg_shift_operators():
# carefully set up the parameters. XXX this still fails sometimes
a1, a2, a3, a4, a5, b1, b2, b3, b4, b5 = (randcplx(n) for n in range(10))
g = meijerg([a1], [a3, a4], [b1], [b3, b4], z)
assert tn(MeijerShiftA(b1).apply(g, op),
meijerg([a1], [a3, a4], [b1 + 1], [b3, b4], z), z)
assert tn(MeijerShiftB(a1).apply(g, op),
meijerg([a1 - 1], [a3, a4], [b1], [b3, b4], z), z)
assert tn(MeijerShiftC(b3).apply(g, op),
meijerg([a1], [a3, a4], [b1], [b3 + 1, b4], z), z)
assert tn(MeijerShiftD(a3).apply(g, op),
meijerg([a1], [a3 - 1, a4], [b1], [b3, b4], z), z)
s = MeijerUnShiftA([a1], [a3, a4], [b1], [b3, b4], 0, z)
assert tn(
s.apply(g, op), meijerg([a1], [a3, a4], [b1 - 1], [b3, b4], z), z)
s = MeijerUnShiftC([a1], [a3, a4], [b1], [b3, b4], 0, z)
assert tn(
s.apply(g, op), meijerg([a1], [a3, a4], [b1], [b3 - 1, b4], z), z)
s = MeijerUnShiftB([a1], [a3, a4], [b1], [b3, b4], 0, z)
assert tn(
s.apply(g, op), meijerg([a1 + 1], [a3, a4], [b1], [b3, b4], z), z)
s = MeijerUnShiftD([a1], [a3, a4], [b1], [b3, b4], 0, z)
assert tn(
s.apply(g, op), meijerg([a1], [a3 + 1, a4], [b1], [b3, b4], z), z)
@slow
def test_meijerg_confluence():
def t(m, a, b):
from sympy.core.sympify import sympify
a, b = sympify([a, b])
m_ = m
m = hyperexpand(m)
if not m == Piecewise((a, abs(z) < 1), (b, abs(1/z) < 1), (m_, True)):
return False
if not (m.args[0].args[0] == a and m.args[1].args[0] == b):
return False
z0 = randcplx()/10
if abs(m.subs(z, z0).n() - a.subs(z, z0).n()).n() > 1e-10:
return False
if abs(m.subs(z, 1/z0).n() - b.subs(z, 1/z0).n()).n() > 1e-10:
return False
return True
assert t(meijerg([], [1, 1], [0, 0], [], z), -log(z), 0)
assert t(meijerg(
[], [3, 1], [0, 0], [], z), -z**2/4 + z - log(z)/2 - Rational(3, 4), 0)
assert t(meijerg([], [3, 1], [-1, 0], [], z),
z**2/12 - z/2 + log(z)/2 + Rational(1, 4) + 1/(6*z), 0)
assert t(meijerg([], [1, 1, 1, 1], [0, 0, 0, 0], [], z), -log(z)**3/6, 0)
assert t(meijerg([1, 1], [], [], [0, 0], z), 0, -log(1/z))
assert t(meijerg([1, 1], [2, 2], [1, 1], [0, 0], z),
-z*log(z) + 2*z, -log(1/z) + 2)
assert t(meijerg([S.Half], [1, 1], [0, 0], [Rational(3, 2)], z), log(z)/2 - 1, 0)
def u(an, ap, bm, bq):
m = meijerg(an, ap, bm, bq, z)
m2 = hyperexpand(m, allow_hyper=True)
if m2.has(meijerg) and not (m2.is_Piecewise and len(m2.args) == 3):
return False
return tn(m, m2, z)
assert u([], [1], [0, 0], [])
assert u([1, 1], [], [], [0])
assert u([1, 1], [2, 2, 5], [1, 1, 6], [0, 0])
assert u([1, 1], [2, 2, 5], [1, 1, 6], [0])
def test_meijerg_with_Floats():
# see issue #10681
from sympy.polys.domains.realfield import RR
f = meijerg(((3.0, 1), ()), ((Rational(3, 2),), (0,)), z)
a = -2.3632718012073
g = a*z**Rational(3, 2)*hyper((-0.5, Rational(3, 2)), (Rational(5, 2),), z*exp_polar(I*pi))
assert RR.almosteq((hyperexpand(f)/g).n(), 1.0, 1e-12)
def test_lerchphi():
from sympy.functions.special.zeta_functions import (lerchphi, polylog)
from sympy.simplify.gammasimp import gammasimp
assert hyperexpand(hyper([1, a], [a + 1], z)/a) == lerchphi(z, 1, a)
assert hyperexpand(
hyper([1, a, a], [a + 1, a + 1], z)/a**2) == lerchphi(z, 2, a)
assert hyperexpand(hyper([1, a, a, a], [a + 1, a + 1, a + 1], z)/a**3) == \
lerchphi(z, 3, a)
assert hyperexpand(hyper([1] + [a]*10, [a + 1]*10, z)/a**10) == \
lerchphi(z, 10, a)
assert gammasimp(hyperexpand(meijerg([0, 1 - a], [], [0],
[-a], exp_polar(-I*pi)*z))) == lerchphi(z, 1, a)
assert gammasimp(hyperexpand(meijerg([0, 1 - a, 1 - a], [], [0],
[-a, -a], exp_polar(-I*pi)*z))) == lerchphi(z, 2, a)
assert gammasimp(hyperexpand(meijerg([0, 1 - a, 1 - a, 1 - a], [], [0],
[-a, -a, -a], exp_polar(-I*pi)*z))) == lerchphi(z, 3, a)
assert hyperexpand(z*hyper([1, 1], [2], z)) == -log(1 + -z)
assert hyperexpand(z*hyper([1, 1, 1], [2, 2], z)) == polylog(2, z)
assert hyperexpand(z*hyper([1, 1, 1, 1], [2, 2, 2], z)) == polylog(3, z)
assert hyperexpand(hyper([1, a, 1 + S.Half], [a + 1, S.Half], z)) == \
-2*a/(z - 1) + (-2*a**2 + a)*lerchphi(z, 1, a)
# Now numerical tests. These make sure reductions etc are carried out
# correctly
# a rational function (polylog at negative integer order)
assert can_do([2, 2, 2], [1, 1])
# NOTE these contain log(1-x) etc ... better make sure we have |z| < 1
# reduction of order for polylog
assert can_do([1, 1, 1, b + 5], [2, 2, b], div=10)
# reduction of order for lerchphi
# XXX lerchphi in mpmath is flaky
assert can_do(
[1, a, a, a, b + 5], [a + 1, a + 1, a + 1, b], numerical=False)
# test a bug
from sympy.functions.elementary.complexes import Abs
assert hyperexpand(hyper([S.Half, S.Half, S.Half, 1],
[Rational(3, 2), Rational(3, 2), Rational(3, 2)], Rational(1, 4))) == \
Abs(-polylog(3, exp_polar(I*pi)/2) + polylog(3, S.Half))
def test_partial_simp():
# First test that hypergeometric function formulae work.
a, b, c, d, e = (randcplx() for _ in range(5))
for func in [Hyper_Function([a, b, c], [d, e]),
Hyper_Function([], [a, b, c, d, e])]:
f = build_hypergeometric_formula(func)
z = f.z
assert f.closed_form == func(z)
deriv1 = f.B.diff(z)*z
deriv2 = f.M*f.B
for func1, func2 in zip(deriv1, deriv2):
assert tn(func1, func2, z)
# Now test that formulae are partially simplified.
a, b, z = symbols('a b z')
assert hyperexpand(hyper([3, a], [1, b], z)) == \
(-a*b/2 + a*z/2 + 2*a)*hyper([a + 1], [b], z) \
+ (a*b/2 - 2*a + 1)*hyper([a], [b], z)
assert tn(
hyperexpand(hyper([3, d], [1, e], z)), hyper([3, d], [1, e], z), z)
assert hyperexpand(hyper([3], [1, a, b], z)) == \
hyper((), (a, b), z) \
+ z*hyper((), (a + 1, b), z)/(2*a) \
- z*(b - 4)*hyper((), (a + 1, b + 1), z)/(2*a*b)
assert tn(
hyperexpand(hyper([3], [1, d, e], z)), hyper([3], [1, d, e], z), z)
def test_hyperexpand_special():
assert hyperexpand(hyper([a, b], [c], 1)) == \
gamma(c)*gamma(c - a - b)/gamma(c - a)/gamma(c - b)
assert hyperexpand(hyper([a, b], [1 + a - b], -1)) == \
gamma(1 + a/2)*gamma(1 + a - b)/gamma(1 + a)/gamma(1 + a/2 - b)
assert hyperexpand(hyper([a, b], [1 + b - a], -1)) == \
gamma(1 + b/2)*gamma(1 + b - a)/gamma(1 + b)/gamma(1 + b/2 - a)
assert hyperexpand(meijerg([1 - z - a/2], [1 - z + a/2], [b/2], [-b/2], 1)) == \
gamma(1 - 2*z)*gamma(z + a/2 + b/2)/gamma(1 - z + a/2 - b/2) \
/gamma(1 - z - a/2 + b/2)/gamma(1 - z + a/2 + b/2)
assert hyperexpand(hyper([a], [b], 0)) == 1
assert hyper([a], [b], 0) != 0
def test_Mod1_behavior():
from sympy.core.symbol import Symbol
from sympy.simplify.simplify import simplify
n = Symbol('n', integer=True)
# Note: this should not hang.
assert simplify(hyperexpand(meijerg([1], [], [n + 1], [0], z))) == \
lowergamma(n + 1, z)
@slow
def test_prudnikov_misc():
assert can_do([1, (3 + I)/2, (3 - I)/2], [Rational(3, 2), 2])
assert can_do([S.Half, a - 1], [Rational(3, 2), a + 1], lowerplane=True)
assert can_do([], [b + 1])
assert can_do([a], [a - 1, b + 1])
assert can_do([a], [a - S.Half, 2*a])
assert can_do([a], [a - S.Half, 2*a + 1])
assert can_do([a], [a - S.Half, 2*a - 1])
assert can_do([a], [a + S.Half, 2*a])
assert can_do([a], [a + S.Half, 2*a + 1])
assert can_do([a], [a + S.Half, 2*a - 1])
assert can_do([S.Half], [b, 2 - b])
assert can_do([S.Half], [b, 3 - b])
assert can_do([1], [2, b])
assert can_do([a, a + S.Half], [2*a, b, 2*a - b + 1])
assert can_do([a, a + S.Half], [S.Half, 2*a, 2*a + S.Half])
assert can_do([a], [a + 1], lowerplane=True) # lowergamma
def test_prudnikov_1():
# A. P. Prudnikov, Yu. A. Brychkov and O. I. Marichev (1990).
# Integrals and Series: More Special Functions, Vol. 3,.
# Gordon and Breach Science Publisher
# 7.3.1
assert can_do([a, -a], [S.Half])
assert can_do([a, 1 - a], [S.Half])
assert can_do([a, 1 - a], [Rational(3, 2)])
assert can_do([a, 2 - a], [S.Half])
assert can_do([a, 2 - a], [Rational(3, 2)])
assert can_do([a, 2 - a], [Rational(3, 2)])
assert can_do([a, a + S.Half], [2*a - 1])
assert can_do([a, a + S.Half], [2*a])
assert can_do([a, a + S.Half], [2*a + 1])
assert can_do([a, a + S.Half], [S.Half])
assert can_do([a, a + S.Half], [Rational(3, 2)])
assert can_do([a, a/2 + 1], [a/2])
assert can_do([1, b], [2])
assert can_do([1, b], [b + 1], numerical=False) # Lerch Phi
# NOTE: branches are complicated for |z| > 1
assert can_do([a], [2*a])
assert can_do([a], [2*a + 1])
assert can_do([a], [2*a - 1])
@slow
def test_prudnikov_2():
h = S.Half
assert can_do([-h, -h], [h])
assert can_do([-h, h], [3*h])
assert can_do([-h, h], [5*h])
assert can_do([-h, h], [7*h])
assert can_do([-h, 1], [h])
for p in [-h, h]:
for n in [-h, h, 1, 3*h, 2, 5*h, 3, 7*h, 4]:
for m in [-h, h, 3*h, 5*h, 7*h]:
assert can_do([p, n], [m])
for n in [1, 2, 3, 4]:
for m in [1, 2, 3, 4]:
assert can_do([p, n], [m])
@slow
def test_prudnikov_3():
if ON_CI:
# See https://github.com/sympy/sympy/pull/12795
skip("Too slow for CI.")
h = S.Half
assert can_do([Rational(1, 4), Rational(3, 4)], [h])
assert can_do([Rational(1, 4), Rational(3, 4)], [3*h])
assert can_do([Rational(1, 3), Rational(2, 3)], [3*h])
assert can_do([Rational(3, 4), Rational(5, 4)], [h])
assert can_do([Rational(3, 4), Rational(5, 4)], [3*h])
for p in [1, 2, 3, 4]:
for n in [-h, h, 1, 3*h, 2, 5*h, 3, 7*h, 4, 9*h]:
for m in [1, 3*h, 2, 5*h, 3, 7*h, 4]:
assert can_do([p, m], [n])
@slow
def test_prudnikov_4():
h = S.Half
for p in [3*h, 5*h, 7*h]:
for n in [-h, h, 3*h, 5*h, 7*h]:
for m in [3*h, 2, 5*h, 3, 7*h, 4]:
assert can_do([p, m], [n])
for n in [1, 2, 3, 4]:
for m in [2, 3, 4]:
assert can_do([p, m], [n])
@slow
def test_prudnikov_5():
h = S.Half
for p in [1, 2, 3]:
for q in range(p, 4):
for r in [1, 2, 3]:
for s in range(r, 4):
assert can_do([-h, p, q], [r, s])
for p in [h, 1, 3*h, 2, 5*h, 3]:
for q in [h, 3*h, 5*h]:
for r in [h, 3*h, 5*h]:
for s in [h, 3*h, 5*h]:
if s <= q and s <= r:
assert can_do([-h, p, q], [r, s])
for p in [h, 1, 3*h, 2, 5*h, 3]:
for q in [1, 2, 3]:
for r in [h, 3*h, 5*h]:
for s in [1, 2, 3]:
assert can_do([-h, p, q], [r, s])
@slow
def test_prudnikov_6():
h = S.Half
for m in [3*h, 5*h]:
for n in [1, 2, 3]:
for q in [h, 1, 2]:
for p in [1, 2, 3]:
assert can_do([h, q, p], [m, n])
for q in [1, 2, 3]:
for p in [3*h, 5*h]:
assert can_do([h, q, p], [m, n])
for q in [1, 2]:
for p in [1, 2, 3]:
for m in [1, 2, 3]:
for n in [1, 2, 3]:
assert can_do([h, q, p], [m, n])
assert can_do([h, h, 5*h], [3*h, 3*h])
assert can_do([h, 1, 5*h], [3*h, 3*h])
assert can_do([h, 2, 2], [1, 3])
# pages 435 to 457 contain more PFDD and stuff like this
@slow
def test_prudnikov_7():
assert can_do([3], [6])
h = S.Half
for n in [h, 3*h, 5*h, 7*h]:
assert can_do([-h], [n])
for m in [-h, h, 1, 3*h, 2, 5*h, 3, 7*h, 4]: # HERE
for n in [-h, h, 3*h, 5*h, 7*h, 1, 2, 3, 4]:
assert can_do([m], [n])
@slow
def test_prudnikov_8():
h = S.Half
# 7.12.2
for ai in [1, 2, 3]:
for bi in [1, 2, 3]:
for ci in range(1, ai + 1):
for di in [h, 1, 3*h, 2, 5*h, 3]:
assert can_do([ai, bi], [ci, di])
for bi in [3*h, 5*h]:
for ci in [h, 1, 3*h, 2, 5*h, 3]:
for di in [1, 2, 3]:
assert can_do([ai, bi], [ci, di])
for ai in [-h, h, 3*h, 5*h]:
for bi in [1, 2, 3]:
for ci in [h, 1, 3*h, 2, 5*h, 3]:
for di in [1, 2, 3]:
assert can_do([ai, bi], [ci, di])
for bi in [h, 3*h, 5*h]:
for ci in [h, 3*h, 5*h, 3]:
for di in [h, 1, 3*h, 2, 5*h, 3]:
if ci <= bi:
assert can_do([ai, bi], [ci, di])
def test_prudnikov_9():
# 7.13.1 [we have a general formula ... so this is a bit pointless]
for i in range(9):
assert can_do([], [(S(i) + 1)/2])
for i in range(5):
assert can_do([], [-(2*S(i) + 1)/2])
@slow
def test_prudnikov_10():
# 7.14.2
h = S.Half
for p in [-h, h, 1, 3*h, 2, 5*h, 3, 7*h, 4]:
for m in [1, 2, 3, 4]:
for n in range(m, 5):
assert can_do([p], [m, n])
for p in [1, 2, 3, 4]:
for n in [h, 3*h, 5*h, 7*h]:
for m in [1, 2, 3, 4]:
assert can_do([p], [n, m])
for p in [3*h, 5*h, 7*h]:
for m in [h, 1, 2, 5*h, 3, 7*h, 4]:
assert can_do([p], [h, m])
assert can_do([p], [3*h, m])
for m in [h, 1, 2, 5*h, 3, 7*h, 4]:
assert can_do([7*h], [5*h, m])
assert can_do([Rational(-1, 2)], [S.Half, S.Half]) # shine-integral shi
def test_prudnikov_11():
# 7.15
assert can_do([a, a + S.Half], [2*a, b, 2*a - b])
assert can_do([a, a + S.Half], [Rational(3, 2), 2*a, 2*a - S.Half])
assert can_do([Rational(1, 4), Rational(3, 4)], [S.Half, S.Half, 1])
assert can_do([Rational(5, 4), Rational(3, 4)], [Rational(3, 2), S.Half, 2])
assert can_do([Rational(5, 4), Rational(3, 4)], [Rational(3, 2), Rational(3, 2), 1])
assert can_do([Rational(5, 4), Rational(7, 4)], [Rational(3, 2), Rational(5, 2), 2])
assert can_do([1, 1], [Rational(3, 2), 2, 2]) # cosh-integral chi
def test_prudnikov_12():
# 7.16
assert can_do(
[], [a, a + S.Half, 2*a], False) # branches only agree for some z!
assert can_do([], [a, a + S.Half, 2*a + 1], False) # dito
assert can_do([], [S.Half, a, a + S.Half])
assert can_do([], [Rational(3, 2), a, a + S.Half])
assert can_do([], [Rational(1, 4), S.Half, Rational(3, 4)])
assert can_do([], [S.Half, S.Half, 1])
assert can_do([], [S.Half, Rational(3, 2), 1])
assert can_do([], [Rational(3, 4), Rational(3, 2), Rational(5, 4)])
assert can_do([], [1, 1, Rational(3, 2)])
assert can_do([], [1, 2, Rational(3, 2)])
assert can_do([], [1, Rational(3, 2), Rational(3, 2)])
assert can_do([], [Rational(5, 4), Rational(3, 2), Rational(7, 4)])
assert can_do([], [2, Rational(3, 2), Rational(3, 2)])
@slow
def test_prudnikov_2F1():
h = S.Half
# Elliptic integrals
for p in [-h, h]:
for m in [h, 3*h, 5*h, 7*h]:
for n in [1, 2, 3, 4]:
assert can_do([p, m], [n])
@XFAIL
def test_prudnikov_fail_2F1():
assert can_do([a, b], [b + 1]) # incomplete beta function
assert can_do([-1, b], [c]) # Poly. also -2, -3 etc
# TODO polys
# Legendre functions:
assert can_do([a, b], [a + b + S.Half])
assert can_do([a, b], [a + b - S.Half])
assert can_do([a, b], [a + b + Rational(3, 2)])
assert can_do([a, b], [(a + b + 1)/2])
assert can_do([a, b], [(a + b)/2 + 1])
assert can_do([a, b], [a - b + 1])
assert can_do([a, b], [a - b + 2])
assert can_do([a, b], [2*b])
assert can_do([a, b], [S.Half])
assert can_do([a, b], [Rational(3, 2)])
assert can_do([a, 1 - a], [c])
assert can_do([a, 2 - a], [c])
assert can_do([a, 3 - a], [c])
assert can_do([a, a + S.Half], [c])
assert can_do([1, b], [c])
assert can_do([1, b], [Rational(3, 2)])
assert can_do([Rational(1, 4), Rational(3, 4)], [1])
# PFDD
o = S.One
assert can_do([o/8, 1], [o/8*9])
assert can_do([o/6, 1], [o/6*7])
assert can_do([o/6, 1], [o/6*13])
assert can_do([o/5, 1], [o/5*6])
assert can_do([o/5, 1], [o/5*11])
assert can_do([o/4, 1], [o/4*5])
assert can_do([o/4, 1], [o/4*9])
assert can_do([o/3, 1], [o/3*4])
assert can_do([o/3, 1], [o/3*7])
assert can_do([o/8*3, 1], [o/8*11])
assert can_do([o/5*2, 1], [o/5*7])
assert can_do([o/5*2, 1], [o/5*12])
assert can_do([o/5*3, 1], [o/5*8])
assert can_do([o/5*3, 1], [o/5*13])
assert can_do([o/8*5, 1], [o/8*13])
assert can_do([o/4*3, 1], [o/4*7])
assert can_do([o/4*3, 1], [o/4*11])
assert can_do([o/3*2, 1], [o/3*5])
assert can_do([o/3*2, 1], [o/3*8])
assert can_do([o/5*4, 1], [o/5*9])
assert can_do([o/5*4, 1], [o/5*14])
assert can_do([o/6*5, 1], [o/6*11])
assert can_do([o/6*5, 1], [o/6*17])
assert can_do([o/8*7, 1], [o/8*15])
@XFAIL
def test_prudnikov_fail_3F2():
assert can_do([a, a + Rational(1, 3), a + Rational(2, 3)], [Rational(1, 3), Rational(2, 3)])
assert can_do([a, a + Rational(1, 3), a + Rational(2, 3)], [Rational(2, 3), Rational(4, 3)])
assert can_do([a, a + Rational(1, 3), a + Rational(2, 3)], [Rational(4, 3), Rational(5, 3)])
# page 421
assert can_do([a, a + Rational(1, 3), a + Rational(2, 3)], [a*Rational(3, 2), (3*a + 1)/2])
# pages 422 ...
assert can_do([Rational(-1, 2), S.Half, S.Half], [1, 1]) # elliptic integrals
assert can_do([Rational(-1, 2), S.Half, 1], [Rational(3, 2), Rational(3, 2)])
# TODO LOTS more
# PFDD
assert can_do([Rational(1, 8), Rational(3, 8), 1], [Rational(9, 8), Rational(11, 8)])
assert can_do([Rational(1, 8), Rational(5, 8), 1], [Rational(9, 8), Rational(13, 8)])
assert can_do([Rational(1, 8), Rational(7, 8), 1], [Rational(9, 8), Rational(15, 8)])
assert can_do([Rational(1, 6), Rational(1, 3), 1], [Rational(7, 6), Rational(4, 3)])
assert can_do([Rational(1, 6), Rational(2, 3), 1], [Rational(7, 6), Rational(5, 3)])
assert can_do([Rational(1, 6), Rational(2, 3), 1], [Rational(5, 3), Rational(13, 6)])
assert can_do([S.Half, 1, 1], [Rational(1, 4), Rational(3, 4)])
# LOTS more
@XFAIL
def test_prudnikov_fail_other():
# 7.11.2
# 7.12.1
assert can_do([1, a], [b, 1 - 2*a + b]) # ???
# 7.14.2
assert can_do([Rational(-1, 2)], [S.Half, 1]) # struve
assert can_do([1], [S.Half, S.Half]) # struve
assert can_do([Rational(1, 4)], [S.Half, Rational(5, 4)]) # PFDD
assert can_do([Rational(3, 4)], [Rational(3, 2), Rational(7, 4)]) # PFDD
assert can_do([1], [Rational(1, 4), Rational(3, 4)]) # PFDD
assert can_do([1], [Rational(3, 4), Rational(5, 4)]) # PFDD
assert can_do([1], [Rational(5, 4), Rational(7, 4)]) # PFDD
# TODO LOTS more
# 7.15.2
assert can_do([S.Half, 1], [Rational(3, 4), Rational(5, 4), Rational(3, 2)]) # PFDD
assert can_do([S.Half, 1], [Rational(7, 4), Rational(5, 4), Rational(3, 2)]) # PFDD
# 7.16.1
assert can_do([], [Rational(1, 3), S(2/3)]) # PFDD
assert can_do([], [Rational(2, 3), S(4/3)]) # PFDD
assert can_do([], [Rational(5, 3), S(4/3)]) # PFDD
# XXX this does not *evaluate* right??
assert can_do([], [a, a + S.Half, 2*a - 1])
def test_bug():
h = hyper([-1, 1], [z], -1)
assert hyperexpand(h) == (z + 1)/z
def test_omgissue_203():
h = hyper((-5, -3, -4), (-6, -6), 1)
assert hyperexpand(h) == Rational(1, 30)
h = hyper((-6, -7, -5), (-6, -6), 1)
assert hyperexpand(h) == Rational(-1, 6)