def test_vector_evalf():
a, b = symbols('a b')
v = pi * A.x
assert v.evalf(2) == Float('3.1416', 2) * A.x
v = pi * A.x + 5 * a * A.y - b * A.z
assert v.evalf(3) == Float('3.1416', 3) * A.x + Float('5', 3) * a * A.y - b * A.z
assert v.evalf(5, subs={a: 1.234, b:5.8973}) == Float('3.1415926536', 5) * A.x + Float('6.17', 5) * A.y - Float('5.8973', 5) * A.z
def test_systematic_basic():
def s(sympy_object, numpy_array):
_ = [sympy_object + numpy_array,
numpy_array + sympy_object,
sympy_object - numpy_array,
numpy_array - sympy_object,
sympy_object * numpy_array,
numpy_array * sympy_object,
sympy_object / numpy_array,
numpy_array / sympy_object,
sympy_object ** numpy_array,
numpy_array ** sympy_object]
x = Symbol("x")
y = Symbol("y")
sympy_objs = [
Rational(2, 3),
Float("1.3"),
x,
y,
pow(x, y)*y,
Integer(5),
Float(5.5),
]
numpy_objs = [
array([1]),
array([3, 8, -1]),
array([x, x**2, Rational(5)]),
array([x/y*sin(y), 5, Rational(5)]),
]
for x in sympy_objs:
for y in numpy_objs:
s(x, y)
def test_issue_3728():
assert (Rational(1, 2)*array([2*x, 0]) == array([x, 0])).all()
assert (Rational(1, 2) + array(
[2*x, 0]) == array([2*x + Rational(1, 2), Rational(1, 2)])).all()
assert (Float("0.5")*array([2*x, 0]) == array([Float("1.0")*x, 0])).all()
assert (Float("0.5") + array(
[2*x, 0]) == array([2*x + Float("0.5"), Float("0.5")])).all()
def test_dyadic_evalf():
N = ReferenceFrame('N')
a = pi * (N.x | N.x)
assert a.evalf(3) == Float('3.1416', 3) * (N.x | N.x)
s = symbols('s')
a = 5 * s * pi* (N.x | N.x)
assert a.evalf(2) == Float('5', 2) * Float('3.1416', 2) * s * (N.x | N.x)
assert a.evalf(9, subs={s: 5.124}) == Float('80.48760378', 9) * (N.x | N.x)
def test_RootSum_evalf():
rs = RootSum(x**2 + 1, exp)
assert rs.evalf(n=20, chop=True).epsilon_eq(Float("1.0806046117362794348"))
assert rs.evalf(n=15, chop=True).epsilon_eq(Float("1.08060461173628"))
rs = RootSum(x**2 + a, exp, x)
assert rs.evalf() == rs
def test_Number():
assert precedence(Integer(0)) == PRECEDENCE["Atom"]
assert precedence(Integer(1)) == PRECEDENCE["Atom"]
assert precedence(Integer(-1)) == PRECEDENCE["Add"]
assert precedence(Integer(10)) == PRECEDENCE["Atom"]
assert precedence(Rational(5, 2)) == PRECEDENCE["Mul"]
assert precedence(Rational(-5, 2)) == PRECEDENCE["Add"]
assert precedence(Float(5)) == PRECEDENCE["Atom"]
assert precedence(Float(-5)) == PRECEDENCE["Add"]
def test_Type__cast_check__complex_floating_point():
val9_11 = 123.456789049 + 0.123456789049j
raises(ValueError, lambda: c64.cast_check(.12345678949 + .12345678949j))
assert abs(val9_11 - c64.cast_check(val9_11) - 4.9e-8) < 1e-8
dcm21 = Float('0.123456789012345670499') + 1e-20j # 21 decimals
assert abs(dcm21 - c128.cast_check(dcm21) - 4.99e-19) < 1e-19
v19 = Float('0.1234567890123456749') + 1j * Float('0.1234567890123456749')
raises(ValueError, lambda: c128.cast_check(v19))
def test_purestr():
assert purestr(Symbol('x')) == "Symbol('x')"
assert purestr(Basic(S(1), S(2))) == "Basic(Integer(1), Integer(2))"
assert purestr(Float(2)) == "Float('2.0', precision=53)"
assert purestr(Symbol('x'), with_args=True) == ("Symbol('x')", ())
assert purestr(Basic(S(1), S(2)), with_args=True) == \
('Basic(Integer(1), Integer(2))', ('Integer(1)', 'Integer(2)'))
assert purestr(Float(2), with_args=True) == \
("Float('2.0', precision=53)", ())
def test_RR_Float():
f1 = Float("1.01")
f2 = Float("1.0000000000000000000001")
assert f1._prec == 53
assert f2._prec == 80
assert RR(f1)-1 > 1e-50
assert RR(f2)-1 < 1e-50 # RR's precision is lower than f2's
RR2 = RealField(prec=f2._prec)
assert RR2(f1)-1 > 1e-50
assert RR2(f2)-1 > 1e-50 # RR's precision is equal to f2's
def test_function_comparable():
assert sin(x).is_comparable is False
assert cos(x).is_comparable is False
assert sin(Float('0.1')).is_comparable is True
assert cos(Float('0.1')).is_comparable is True
assert sin(E).is_comparable is True
assert cos(E).is_comparable is True
assert sin(Rational(1, 3)).is_comparable is True
assert cos(Rational(1, 3)).is_comparable is True
def test_rf_ff_eval_hiprec():
maple = Float('6.9109401292234329956525265438452')
us = ff(18, Rational(2, 3)).evalf(32)
assert abs(us - maple) / us < 1e-31
maple = Float('6.8261540131125511557924466355367')
us = rf(18, Rational(2, 3)).evalf(32)
assert abs(us - maple) / us < 1e-31
maple = Float('34.007346127440197150854651814225')
us = rf(Float('4.4', 32), Float('2.2', 32))
assert abs(us - maple) / us < 1e-31
def test_issue_16759():
d = sympify({.5: 1})
assert S.Half not in d
assert Float(.5) in d
assert d[.5] is S.One
d = sympify(OrderedDict({.5: 1}))
assert S.Half not in d
assert Float(.5) in d
assert d[.5] is S.One
d = sympify(defaultdict(int, {.5: 1}))
assert S.Half not in d
assert Float(.5) in d
assert d[.5] is S.One
def _mpc_(self):
"""
Convert self to an mpmath mpc if possible
"""
from sympy.core.numbers import I, Float
re_part, rest = self.as_coeff_Add()
im_part, imag_unit = rest.as_coeff_Mul()
if not imag_unit == I:
# ValueError may seem more reasonable but since it's a @property,
# we need to use AttributeError to keep from confusing things like
# hasattr.
raise AttributeError("Cannot convert Add to mpc. Must be of the form Number + Number*I")
return (Float(re_part)._mpf_, Float(im_part)._mpf_)
def test_scalar_numpy():
if not np:
skip("numpy not installed.")
assert represent(Integer(1), format='numpy') == 1
assert represent(Float(1.0), format='numpy') == 1.0
assert represent(1.0 + I, format='numpy') == 1.0 + 1.0j
def Ar(element: Any) -> Float:
"""Retrieve the relative atomic mass of the element.
:param element: Any object that provides a __str__ or __repr__ method
that gives a string of the symbol of an element.
:return: The relative atomic mass of the element (in amu).
:raise LookupError: Raised when the element symbol can not be found in data.
>>> Ar('H')
1.01
>>> Ar(sympy.Symbol('H'))
1.01
>>> class Foo:
... def __repr__(self):
... return 'H'
>>> Ar(Foo())
1.01
>>> Ar('Ha') # not an element
Traceback (most recent call last):
...
LookupError: unknown element symbol 'Ha'
"""
symbol = str(element)
if symbol in ATOMIC_MASSES:
return Float(ATOMIC_MASSES[symbol], 3)
raise LookupError(f"unknown element symbol '{symbol}'")
def test_rf_lambdify_mpmath():
from sympy.utilities.lambdify import lambdify
x, y = symbols('x,y')
f = lambdify((x, y), rf(x, y), 'mpmath')
maple = Float('34.007346127440197')
us = f(4.4, 2.2)
assert abs(us - maple) / us < 1e-15
def test_sympy_lambda():
mpmath.mp.dps = 50
sin02 = mpmath.mpf("0.19866933079506121545941262711838975037020672954020")
f = lambdify(x, sin(x), "sympy")
assert f(x) == sin(x)
prec = 1e-15
assert -prec < f(Rational(1, 5)).evalf() - Float(str(sin02)) < prec
def test__sympify():
x = Symbol('x')
f = Function('f')
# positive _sympify
assert _sympify(x) is x
assert _sympify(1) == Integer(1)
assert _sympify(0.5) == Float("0.5")
assert _sympify(1 + 1j) == 1.0 + I * 1.0
# Function f is not Basic and can't sympify to Basic. We allow it to pass
# with sympify but not with _sympify.
# https://github.com/sympy/sympy/issues/20124
assert sympify(f) is f
raises(SympifyError, lambda: _sympify(f))
class A:
def _sympy_(self):
return Integer(5)
a = A()
assert _sympify(a) == Integer(5)
# negative _sympify
raises(SympifyError, lambda: _sympify('1'))
raises(SympifyError, lambda: _sympify([1, 2, 3]))
def dotest(s):
x = Symbol("x")
y = Symbol("y")
l = [Rational(2), Float("1.3"), x, y, pow(x, y) * y, 5, 5.5]
for x in l:
for y in l:
s(x, y)
return True
def test_hbar():
assert hbar.is_commutative is True
assert hbar.is_real is True
assert hbar.is_positive is True
assert hbar.is_negative is False
assert hbar.is_irrational is True
assert hbar.evalf() == Float(1.05457162e-34)
def test_precision():
f1 = Float("1.01")
f2 = Float("1.0000000000000000000001")
for u in [
1, 1e-2, 1e-6, 1e-13, 1e-14, 1e-16, 1e-20, 1e-100, 1e-300, f1, f2
]:
result = construct_domain([u])
v = float(result[1][0])
assert abs(u - v) / u < 1e-14 # Test relative accuracy
result = construct_domain([f1])
y = result[1][0]
assert y - 1 > 1e-50
result = construct_domain([f2])
y = result[1][0]
assert y - 1 > 1e-50
def _pi_coeff(arg, cycles=1):
"""
When arg is a Number times pi (e.g. 3*pi/2) then return the Number
normalized to be in the range [0, 2], else None.
When an even multiple of pi is encountered, if it is multiplying
something with known parity then the multiple is returned as 0 otherwise
as 2.
Examples:
>>> from sympy.functions.elementary.trigonometric import _pi_coeff as coeff
>>> from sympy import pi
>>> from sympy.abc import x, y
>>> coeff(3*x*pi)
3*x
>>> coeff(11*pi/7)
11/7
>>> coeff(-11*pi/7)
3/7
>>> coeff(4*pi)
0
>>> coeff(5*pi)
1
>>> coeff(5.0*pi)
1
>>> coeff(5.5*pi)
3/2
>>> coeff(2 + pi)
"""
arg = sympify(arg)
if arg is S.Pi:
return S.One
elif not arg:
return S.Zero
elif arg.is_Mul:
cx = arg.coeff(S.Pi)
if cx:
c, x = cx.as_coeff_Mul() # pi is not included as coeff
if c.is_Float:
# recast exact binary fractions to Rationals
m = int(c * 2)
if Float(float(m) / 2) == c:
c = Rational(m, 2)
if x is not S.One or not (c.is_Rational and c.q != 1):
if x.is_integer:
c2 = c % 2
if c2 == 1:
return x
elif not c2:
if x.is_even is not None: # known parity
return S.Zero
return 2 * x
else:
return c2 * x
return cx
else:
return Rational(c.p % (2 * c.q), c.q)
def test_sympify_numpy():
if not numpy:
skip('numpy not installed. Abort numpy tests.')
np = numpy
def equal(x, y):
return x == y and type(x) == type(y)
assert sympify(np.bool_(1)) is S(True)
try:
assert equal(sympify(np.int_(1234567891234567891)),
S(1234567891234567891))
assert equal(sympify(np.intp(1234567891234567891)),
S(1234567891234567891))
except OverflowError:
# May fail on 32-bit systems: Python int too large to convert to C long
pass
assert equal(sympify(np.intc(1234567891)), S(1234567891))
assert equal(sympify(np.int8(-123)), S(-123))
assert equal(sympify(np.int16(-12345)), S(-12345))
assert equal(sympify(np.int32(-1234567891)), S(-1234567891))
assert equal(sympify(np.int64(-1234567891234567891)),
S(-1234567891234567891))
assert equal(sympify(np.uint8(123)), S(123))
assert equal(sympify(np.uint16(12345)), S(12345))
assert equal(sympify(np.uint32(1234567891)), S(1234567891))
assert equal(sympify(np.uint64(1234567891234567891)),
S(1234567891234567891))
assert equal(sympify(np.float32(1.123456)), Float(1.123456, precision=24))
assert equal(sympify(np.float64(1.1234567891234)),
Float(1.1234567891234, precision=53))
assert equal(sympify(np.longdouble(1.123456789)),
Float(1.123456789, precision=80))
assert equal(sympify(np.complex64(1 + 2j)), S(1.0 + 2.0 * I))
assert equal(sympify(np.complex128(1 + 2j)), S(1.0 + 2.0 * I))
assert equal(sympify(np.longcomplex(1 + 2j)), S(1.0 + 2.0 * I))
#float96 does not exist on all platforms
if hasattr(np, 'float96'):
assert equal(sympify(np.float96(1.123456789)),
Float(1.123456789, precision=80))
#float128 does not exist on all platforms
if hasattr(np, 'float128'):
assert equal(sympify(np.float128(1.123456789123)),
Float(1.123456789123, precision=80))
def test_issue_8245():
a = S("6506833320952669167898688709329/5070602400912917605986812821504")
assert rel_check(a, a.n(10))
assert rel_check(a, a.n(20))
assert rel_check(a, a.n())
# prec of 30 is enough to fully capture a as mpf
assert Float(a, 30) == Float(str(a.p), '') / Float(str(a.q), '')
for i in range(31):
r = Rational(Float(a, i))
f = Float(r)
assert (f < a) == (Rational(f) < a)
# test sign handling
assert (-f < -a) == (Rational(-f) < -a)
# test equivalence handling
isa = Float(a.p, '') / Float(a.q, '')
assert isa <= a
assert not isa < a
assert isa >= a
assert not isa > a
assert isa > 0
a = sqrt(2)
r = Rational(str(a.n(30)))
assert rel_check(a, r)
a = sqrt(2)
r = Rational(str(a.n(29)))
assert rel_check(a, r)
assert Eq(log(cos(2)**2 + sin(2)**2), 0) is S.true
def test_issue_2877():
f = Float(2.0)
assert (x + f).subs({f: 2}) == x + 2
def r(a, b, c):
return factor(a * x**2 + b * x + c)
e = r(5.0 / 6, 10, 5)
assert nsimplify(e) == 5 * x**2 / 6 + 10 * x + 5
def test_Type__cast_check__integers():
# Rounding
raises(ValueError, lambda: integer.cast_check(3.5))
assert integer.cast_check('3') == 3
assert integer.cast_check(Float('3.0000000000000000000')) == 3
assert integer.cast_check(
Float('3.0000000000000000001')) == 3 # unintuitive maybe?
# Range
assert int8.cast_check(127.0) == 127
raises(ValueError, lambda: int8.cast_check(128))
assert int8.cast_check(-128) == -128
raises(ValueError, lambda: int8.cast_check(-129))
assert uint8.cast_check(0) == 0
assert uint8.cast_check(128) == 128
raises(ValueError, lambda: uint8.cast_check(256.0))
raises(ValueError, lambda: uint8.cast_check(-1))
def fit_traj_segment(p_start,
p_end,
v_start,
v_end,
p_max,
v_max,
a_max,
j_max,
independent_variable=Symbol('t')):
'''
This function selects a motion profile for a general trajectory segment with a given start/end velocities/positions
considering the start and end accelerations/jerks are zeros
'''
# Step_1. calculate jerk_sign_and_duration
segment_jerks_and_durations = calculate_jerk_sign_and_duration(
p_start,
p_end,
v_start,
v_end,
p_max,
v_max,
a_max,
j_max,
independent_variable=Symbol('t'))
# Step_2: generate pos, vel, acc, jrk using the calculated "segment_jerks_and_durations"
p0 = p_start
v0 = v_start
a0 = 0.0
times = [0.0]
jerk_functions = []
acceleration_functions = []
velocity_functions = []
position_functions = []
# Integrate jerk starting from the start of the trajectory and going all the way through the end.
for j0, T in segment_jerks_and_durations:
times.append(times[-1] + T)
j = Float(j0)
a = integrate(j, independent_variable) + a0
v = integrate(a, independent_variable) + v0
p = integrate(v, independent_variable) + p0
jerk_functions.append(j)
acceleration_functions.append(a)
velocity_functions.append(v)
position_functions.append(p)
a0 = a.subs({independent_variable: T})
v0 = v.subs({independent_variable: T})
p0 = p.subs({independent_variable: T})
position = PiecewiseFunction(times, position_functions,
independent_variable)
velocity = PiecewiseFunction(times, velocity_functions,
independent_variable)
acceleration = PiecewiseFunction(times, acceleration_functions,
independent_variable)
jerk = PiecewiseFunction(times, jerk_functions, independent_variable)
return position, velocity, acceleration, jerk
def test_scalar_scipy_sparse():
if not np:
skip("numpy not installed.")
if not scipy:
skip("scipy not installed.")
assert represent(Integer(1), format='scipy.sparse') == 1
assert represent(Float(1.0), format='scipy.sparse') == 1.0
assert represent(1.0 + I, format='scipy.sparse') == 1.0 + 1.0j
def test_lambertw():
k = Symbol('k')
assert LambertW(x, 0) == LambertW(x)
assert LambertW(x, 0, evaluate=False) != LambertW(x)
assert LambertW(0) == 0
assert LambertW(E) == 1
assert LambertW(-1 / E) == -1
assert LambertW(-log(2) / 2) == -log(2)
assert LambertW(oo) is oo
assert LambertW(0, 1) is -oo
assert LambertW(0, 42) is -oo
assert LambertW(-pi / 2, -1) == -I * pi / 2
assert LambertW(-1 / E, -1) == -1
assert LambertW(-2 * exp(-2), -1) == -2
assert LambertW(2 * log(2)) == log(2)
assert LambertW(-pi / 2) == I * pi / 2
assert LambertW(exp(1 + E)) == E
assert LambertW(
x**2).diff(x) == 2 * LambertW(x**2) / x / (1 + LambertW(x**2))
assert LambertW(x, k).diff(x) == LambertW(x, k) / x / (1 + LambertW(x, k))
assert LambertW(sqrt(2)).evalf(30).epsilon_eq(
Float("0.701338383413663009202120278965", 30), 1e-29)
assert re(LambertW(2, -1)).evalf().epsilon_eq(Float("-0.834310366631110"))
assert LambertW(-1).is_real is False # issue 5215
assert LambertW(2, evaluate=False).is_real
p = Symbol('p', positive=True)
assert LambertW(p, evaluate=False).is_real
assert LambertW(p - 1, evaluate=False).is_real is None
assert LambertW(-p - 2 / S.Exp1, evaluate=False).is_real is False
assert LambertW(S.Half, -1, evaluate=False).is_real is False
assert LambertW(Rational(-1, 10), -1, evaluate=False).is_real
assert LambertW(-10, -1, evaluate=False).is_real is False
assert LambertW(-2, 2, evaluate=False).is_real is False
assert LambertW(0, evaluate=False).is_algebraic
na = Symbol('na', nonzero=True, algebraic=True)
assert LambertW(na).is_algebraic is False
assert LambertW(p).is_zero is False
n = Symbol('n', negative=True)
assert LambertW(n).is_zero is False
def test_core_numbers():
for c in (Catalan, Catalan(), ComplexInfinity, ComplexInfinity(),
EulerGamma, EulerGamma(), Exp1, Exp1(),
GoldenRatio, GoldenRatio(), Half, Half(), ImaginaryUnit,
ImaginaryUnit(), Infinity, Infinity(), Integer, Integer(2), NaN,
NaN(), NegativeInfinity, NegativeInfinity(), NegativeOne,
NegativeOne(), Number, Number(15), NumberSymbol, NumberSymbol(),
One, One(), Pi, Pi(), Rational, Rational(1, 2), Float,
Float("1.2"), Zero, Zero()):
check(c)
def evalf(self, n=15, subs=None, maxn=100, chop=False, strict=False, quad=None, verbose=False):
"""
Evaluate the given formula to an accuracy of n digits.
Optional keyword arguments:
subs=<dict>
Substitute numerical values for symbols, e.g.
subs={x:3, y:1+pi}. The substitutions must be given as a
dictionary.
maxn=<integer>
Allow a maximum temporary working precision of maxn digits
(default=100)
chop=<bool>
Replace tiny real or imaginary parts in subresults
by exact zeros (default=False)
strict=<bool>
Raise PrecisionExhausted if any subresult fails to evaluate
to full accuracy, given the available maxprec
(default=False)
quad=<str>
Choose algorithm for numerical quadrature. By default,
tanh-sinh quadrature is used. For oscillatory
integrals on an infinite interval, try quad='osc'.
verbose=<bool>
Print debug information (default=False)
"""
from sympy import Float, Number
n = n if n is not None else 15
if subs and is_sequence(subs):
raise TypeError('subs must be given as a dictionary')
# for sake of sage that doesn't like evalf(1)
if n == 1 and isinstance(self, Number):
from sympy.core.expr import _mag
rv = self.evalf(2, subs, maxn, chop, strict, quad, verbose)
m = _mag(rv)
rv = rv.round(1 - m)
return rv
if not evalf_table:
_create_evalf_table()
prec = dps_to_prec(n)
options = {'maxprec': max(prec, int(maxn*LG10)), 'chop': chop,
'strict': strict, 'verbose': verbose}
if subs is not None:
options['subs'] = subs
if quad is not None:
options['quad'] = quad
try:
result = evalf(self, prec + 4, options)
except NotImplementedError:
# Fall back to the ordinary evalf
v = self._eval_evalf(prec)
if v is None:
return self
try:
# If the result is numerical, normalize it
result = evalf(v, prec, options)
except NotImplementedError:
# Probably contains symbols or unknown functions
return v
re, im, re_acc, im_acc = result
if re:
p = max(min(prec, re_acc), 1)
re = Float._new(re, p)
else:
re = S.Zero
if im:
p = max(min(prec, im_acc), 1)
im = Float._new(im, p)
return re + im*S.ImaginaryUnit
else:
return re
def add_terms(terms, prec, target_prec):
"""
Helper for evalf_add. Adds a list of (mpfval, accuracy) terms.
Returns
-------
- None, None if there are no non-zero terms;
- terms[0] if there is only 1 term;
- scaled_zero if the sum of the terms produces a zero by cancellation
e.g. mpfs representing 1 and -1 would produce a scaled zero which need
special handling since they are not actually zero and they are purposely
malformed to ensure that they can't be used in anything but accuracy
calculations;
- a tuple that is scaled to target_prec that corresponds to the
sum of the terms.
The returned mpf tuple will be normalized to target_prec; the input
prec is used to define the working precision.
XXX explain why this is needed and why one can't just loop using mpf_add
"""
terms = [t for t in terms if not iszero(t)]
if not terms:
return None, None
elif len(terms) == 1:
return terms[0]
# see if any argument is NaN or oo and thus warrants a special return
special = []
from sympy.core.numbers import Float
for t in terms:
arg = Float._new(t[0], 1)
if arg is S.NaN or arg.is_infinite:
special.append(arg)
if special:
from sympy.core.add import Add
rv = evalf(Add(*special), prec + 4, {})
return rv[0], rv[2]
working_prec = 2*prec
sum_man, sum_exp, absolute_error = 0, 0, MINUS_INF
for x, accuracy in terms:
sign, man, exp, bc = x
if sign:
man = -man
absolute_error = max(absolute_error, bc + exp - accuracy)
delta = exp - sum_exp
if exp >= sum_exp:
# x much larger than existing sum?
# first: quick test
if ((delta > working_prec) and
((not sum_man) or
delta - bitcount(abs(sum_man)) > working_prec)):
sum_man = man
sum_exp = exp
else:
sum_man += (man << delta)
else:
delta = -delta
# x much smaller than existing sum?
if delta - bc > working_prec:
if not sum_man:
sum_man, sum_exp = man, exp
else:
sum_man = (sum_man << delta) + man
sum_exp = exp
if not sum_man:
return scaled_zero(absolute_error)
if sum_man < 0:
sum_sign = 1
sum_man = -sum_man
else:
sum_sign = 0
sum_bc = bitcount(sum_man)
sum_accuracy = sum_exp + sum_bc - absolute_error
r = normalize(sum_sign, sum_man, sum_exp, sum_bc, target_prec,
rnd), sum_accuracy
return r
def evalf_mul(v, prec, options):
res = pure_complex(v)
if res:
# the only pure complex that is a mul is h*I
_, h = res
im, _, im_acc, _ = evalf(h, prec, options)
return None, im, None, im_acc
args = list(v.args)
# see if any argument is NaN or oo and thus warrants a special return
special = []
from sympy.core.numbers import Float
for arg in args:
arg = evalf(arg, prec, options)
if arg[0] is None:
continue
arg = Float._new(arg[0], 1)
if arg is S.NaN or arg.is_infinite:
special.append(arg)
if special:
from sympy.core.mul import Mul
special = Mul(*special)
return evalf(special, prec + 4, {})
# With guard digits, multiplication in the real case does not destroy
# accuracy. This is also true in the complex case when considering the
# total accuracy; however accuracy for the real or imaginary parts
# separately may be lower.
acc = prec
# XXX: big overestimate
working_prec = prec + len(args) + 5
# Empty product is 1
start = man, exp, bc = MPZ(1), 0, 1
# First, we multiply all pure real or pure imaginary numbers.
# direction tells us that the result should be multiplied by
# I**direction; all other numbers get put into complex_factors
# to be multiplied out after the first phase.
last = len(args)
direction = 0
args.append(S.One)
complex_factors = []
for i, arg in enumerate(args):
if i != last and pure_complex(arg):
args[-1] = (args[-1]*arg).expand()
continue
elif i == last and arg is S.One:
continue
re, im, re_acc, im_acc = evalf(arg, working_prec, options)
if re and im:
complex_factors.append((re, im, re_acc, im_acc))
continue
elif re:
(s, m, e, b), w_acc = re, re_acc
elif im:
(s, m, e, b), w_acc = im, im_acc
direction += 1
else:
return None, None, None, None
direction += 2*s
man *= m
exp += e
bc += b
if bc > 3*working_prec:
man >>= working_prec
exp += working_prec
acc = min(acc, w_acc)
sign = (direction & 2) >> 1
if not complex_factors:
v = normalize(sign, man, exp, bitcount(man), prec, rnd)
# multiply by i
if direction & 1:
return None, v, None, acc
else:
return v, None, acc, None
else:
# initialize with the first term
if (man, exp, bc) != start:
# there was a real part; give it an imaginary part
re, im = (sign, man, exp, bitcount(man)), (0, MPZ(0), 0, 0)
i0 = 0
else:
# there is no real part to start (other than the starting 1)
wre, wim, wre_acc, wim_acc = complex_factors[0]
acc = min(acc,
complex_accuracy((wre, wim, wre_acc, wim_acc)))
re = wre
im = wim
i0 = 1
for wre, wim, wre_acc, wim_acc in complex_factors[i0:]:
# acc is the overall accuracy of the product; we aren't
# computing exact accuracies of the product.
acc = min(acc,
complex_accuracy((wre, wim, wre_acc, wim_acc)))
use_prec = working_prec
A = mpf_mul(re, wre, use_prec)
B = mpf_mul(mpf_neg(im), wim, use_prec)
C = mpf_mul(re, wim, use_prec)
D = mpf_mul(im, wre, use_prec)
re = mpf_add(A, B, use_prec)
im = mpf_add(C, D, use_prec)
if options.get('verbose'):
print("MUL: wanted", prec, "accurate bits, got", acc)
# multiply by I
if direction & 1:
re, im = mpf_neg(im), re
return re, im, acc, acc