from sympy.core import Add, Mul, Pow, S from sympy.core.basic import Basic from sympy.core.expr import Expr from sympy.core.numbers import _sympifyit, oo, zoo from sympy.core.relational import is_le, is_lt, is_ge, is_gt from sympy.core.sympify import _sympify from sympy.functions.elementary.miscellaneous import Min, Max from sympy.logic.boolalg import And from sympy.multipledispatch import dispatch from sympy.series.order import Order from sympy.sets.sets import FiniteSet class AccumulationBounds(Expr): r"""An accumulation bounds. # Note AccumulationBounds has an alias: AccumBounds AccumulationBounds represent an interval `[a, b]`, which is always closed at the ends. Here `a` and `b` can be any value from extended real numbers. The intended meaning of AccummulationBounds is to give an approximate location of the accumulation points of a real function at a limit point. Let `a` and `b` be reals such that `a \le b`. `\left\langle a, b\right\rangle = \{x \in \mathbb{R} \mid a \le x \le b\}` `\left\langle -\infty, b\right\rangle = \{x \in \mathbb{R} \mid x \le b\} \cup \{-\infty, \infty\}` `\left\langle a, \infty \right\rangle = \{x \in \mathbb{R} \mid a \le x\} \cup \{-\infty, \infty\}` `\left\langle -\infty, \infty \right\rangle = \mathbb{R} \cup \{-\infty, \infty\}` ``oo`` and ``-oo`` are added to the second and third definition respectively, since if either ``-oo`` or ``oo`` is an argument, then the other one should be included (though not as an end point). This is forced, since we have, for example, ``1/AccumBounds(0, 1) = AccumBounds(1, oo)``, and the limit at `0` is not one-sided. As `x` tends to `0-`, then `1/x \rightarrow -\infty`, so `-\infty` should be interpreted as belonging to ``AccumBounds(1, oo)`` though it need not appear explicitly. In many cases it suffices to know that the limit set is bounded. However, in some other cases more exact information could be useful. For example, all accumulation values of `\cos(x) + 1` are non-negative. (``AccumBounds(-1, 1) + 1 = AccumBounds(0, 2)``) A AccumulationBounds object is defined to be real AccumulationBounds, if its end points are finite reals. Let `X`, `Y` be real AccumulationBounds, then their sum, difference, product are defined to be the following sets: `X + Y = \{ x+y \mid x \in X \cap y \in Y\}` `X - Y = \{ x-y \mid x \in X \cap y \in Y\}` `X \times Y = \{ x \times y \mid x \in X \cap y \in Y\}` When an AccumBounds is raised to a negative power, if 0 is contained between the bounds then an infinite range is returned, otherwise if an endpoint is 0 then a semi-infinite range with consistent sign will be returned. AccumBounds in expressions behave a lot like Intervals but the semantics are not necessarily the same. Division (or exponentiation to a negative integer power) could be handled with *intervals* by returning a union of the results obtained after splitting the bounds between negatives and positives, but that is not done with AccumBounds. In addition, bounds are assumed to be independent of each other; if the same bound is used in more than one place in an expression, the result may not be the supremum or infimum of the expression (see below). Finally, when a boundary is ``1``, exponentiation to the power of ``oo`` yields ``oo``, neither ``1`` nor ``nan``. Examples ======== >>> from sympy import AccumBounds, sin, exp, log, pi, E, S, oo >>> from sympy.abc import x >>> AccumBounds(0, 1) + AccumBounds(1, 2) AccumBounds(1, 3) >>> AccumBounds(0, 1) - AccumBounds(0, 2) AccumBounds(-2, 1) >>> AccumBounds(-2, 3)*AccumBounds(-1, 1) AccumBounds(-3, 3) >>> AccumBounds(1, 2)*AccumBounds(3, 5) AccumBounds(3, 10) The exponentiation of AccumulationBounds is defined as follows: If 0 does not belong to `X` or `n > 0` then `X^n = \{ x^n \mid x \in X\}` >>> AccumBounds(1, 4)**(S(1)/2) AccumBounds(1, 2) otherwise, an infinite or semi-infinite result is obtained: >>> 1/AccumBounds(-1, 1) AccumBounds(-oo, oo) >>> 1/AccumBounds(0, 2) AccumBounds(1/2, oo) >>> 1/AccumBounds(-oo, 0) AccumBounds(-oo, 0) A boundary of 1 will always generate all nonnegatives: >>> AccumBounds(1, 2)**oo AccumBounds(0, oo) >>> AccumBounds(0, 1)**oo AccumBounds(0, oo) If the exponent is itself an AccumulationBounds or is not an integer then unevaluated results will be returned unless the base values are positive: >>> AccumBounds(2, 3)**AccumBounds(-1, 2) AccumBounds(1/3, 9) >>> AccumBounds(-2, 3)**AccumBounds(-1, 2) AccumBounds(-2, 3)**AccumBounds(-1, 2) >>> AccumBounds(-2, -1)**(S(1)/2) sqrt(AccumBounds(-2, -1)) Note: `\left\langle a, b\right\rangle^2` is not same as `\left\langle a, b\right\rangle \times \left\langle a, b\right\rangle` >>> AccumBounds(-1, 1)**2 AccumBounds(0, 1) >>> AccumBounds(1, 3) < 4 True >>> AccumBounds(1, 3) < -1 False Some elementary functions can also take AccumulationBounds as input. A function `f` evaluated for some real AccumulationBounds `\left\langle a, b \right\rangle` is defined as `f(\left\langle a, b\right\rangle) = \{ f(x) \mid a \le x \le b \}` >>> sin(AccumBounds(pi/6, pi/3)) AccumBounds(1/2, sqrt(3)/2) >>> exp(AccumBounds(0, 1)) AccumBounds(1, E) >>> log(AccumBounds(1, E)) AccumBounds(0, 1) Some symbol in an expression can be substituted for a AccumulationBounds object. But it does not necessarily evaluate the AccumulationBounds for that expression. The same expression can be evaluated to different values depending upon the form it is used for substitution since each instance of an AccumulationBounds is considered independent. For example: >>> (x**2 + 2*x + 1).subs(x, AccumBounds(-1, 1)) AccumBounds(-1, 4) >>> ((x + 1)**2).subs(x, AccumBounds(-1, 1)) AccumBounds(0, 4) References ========== .. [1] https://en.wikipedia.org/wiki/Interval_arithmetic .. [2] https://fab.cba.mit.edu/classes/S62.12/docs/Hickey_interval.pdf Notes ===== Do not use ``AccumulationBounds`` for floating point interval arithmetic calculations, use ``mpmath.iv`` instead. """ is_extended_real = True is_number = False def __new__(cls, min, max): min = _sympify(min) max = _sympify(max) # Only allow real intervals (use symbols with 'is_extended_real=True'). if not min.is_extended_real or not max.is_extended_real: raise ValueError("Only real AccumulationBounds are supported") if max == min: return max # Make sure that the created AccumBounds object will be valid. if max.is_number and min.is_number: bad = max.is_comparable and min.is_comparable and max < min else: bad = (max - min).is_extended_negative if bad: raise ValueError( "Lower limit should be smaller than upper limit") return Basic.__new__(cls, min, max) # setting the operation priority _op_priority = 11.0 def _eval_is_real(self): if self.min.is_real and self.max.is_real: return True @property def min(self): """ Returns the minimum possible value attained by AccumulationBounds object. Examples ======== >>> from sympy import AccumBounds >>> AccumBounds(1, 3).min 1 """ return self.args[0] @property def max(self): """ Returns the maximum possible value attained by AccumulationBounds object. Examples ======== >>> from sympy import AccumBounds >>> AccumBounds(1, 3).max 3 """ return self.args[1] @property def delta(self): """ Returns the difference of maximum possible value attained by AccumulationBounds object and minimum possible value attained by AccumulationBounds object. Examples ======== >>> from sympy import AccumBounds >>> AccumBounds(1, 3).delta 2 """ return self.max - self.min @property def mid(self): """ Returns the mean of maximum possible value attained by AccumulationBounds object and minimum possible value attained by AccumulationBounds object. Examples ======== >>> from sympy import AccumBounds >>> AccumBounds(1, 3).mid 2 """ return (self.min + self.max) / 2 @_sympifyit('other', NotImplemented) def _eval_power(self, other): return self.__pow__(other) @_sympifyit('other', NotImplemented) def __add__(self, other): if isinstance(other, Expr): if isinstance(other, AccumBounds): return AccumBounds( Add(self.min, other.min), Add(self.max, other.max)) if other is S.Infinity and self.min is S.NegativeInfinity or \ other is S.NegativeInfinity and self.max is S.Infinity: return AccumBounds(-oo, oo) elif other.is_extended_real: if self.min is S.NegativeInfinity and self.max is S.Infinity: return AccumBounds(-oo, oo) elif self.min is S.NegativeInfinity: return AccumBounds(-oo, self.max + other) elif self.max is S.Infinity: return AccumBounds(self.min + other, oo) else: return AccumBounds(Add(self.min, other), Add(self.max, other)) return Add(self, other, evaluate=False) return NotImplemented __radd__ = __add__ def __neg__(self): return AccumBounds(-self.max, -self.min) @_sympifyit('other', NotImplemented) def __sub__(self, other): if isinstance(other, Expr): if isinstance(other, AccumBounds): return AccumBounds( Add(self.min, -other.max), Add(self.max, -other.min)) if other is S.NegativeInfinity and self.min is S.NegativeInfinity or \ other is S.Infinity and self.max is S.Infinity: return AccumBounds(-oo, oo) elif other.is_extended_real: if self.min is S.NegativeInfinity and self.max is S.Infinity: return AccumBounds(-oo, oo) elif self.min is S.NegativeInfinity: return AccumBounds(-oo, self.max - other) elif self.max is S.Infinity: return AccumBounds(self.min - other, oo) else: return AccumBounds( Add(self.min, -other), Add(self.max, -other)) return Add(self, -other, evaluate=False) return NotImplemented @_sympifyit('other', NotImplemented) def __rsub__(self, other): return self.__neg__() + other @_sympifyit('other', NotImplemented) def __mul__(self, other): if self.args == (-oo, oo): return self if isinstance(other, Expr): if isinstance(other, AccumBounds): if other.args == (-oo, oo): return other v = set() for a in self.args: vi = other*a v.update(vi.args or (vi,)) return AccumBounds(Min(*v), Max(*v)) if other is S.Infinity: if self.min.is_zero: return AccumBounds(0, oo) if self.max.is_zero: return AccumBounds(-oo, 0) if other is S.NegativeInfinity: if self.min.is_zero: return AccumBounds(-oo, 0) if self.max.is_zero: return AccumBounds(0, oo) if other.is_extended_real: if other.is_zero: if self.max is S.Infinity: return AccumBounds(0, oo) if self.min is S.NegativeInfinity: return AccumBounds(-oo, 0) return S.Zero if other.is_extended_positive: return AccumBounds( Mul(self.min, other), Mul(self.max, other)) elif other.is_extended_negative: return AccumBounds( Mul(self.max, other), Mul(self.min, other)) if isinstance(other, Order): return other return Mul(self, other, evaluate=False) return NotImplemented __rmul__ = __mul__ @_sympifyit('other', NotImplemented) def __truediv__(self, other): if isinstance(other, Expr): if isinstance(other, AccumBounds): if other.min.is_positive or other.max.is_negative: return self * AccumBounds(1/other.max, 1/other.min) if (self.min.is_extended_nonpositive and self.max.is_extended_nonnegative and other.min.is_extended_nonpositive and other.max.is_extended_nonnegative): if self.min.is_zero and other.min.is_zero: return AccumBounds(0, oo) if self.max.is_zero and other.min.is_zero: return AccumBounds(-oo, 0) return AccumBounds(-oo, oo) if self.max.is_extended_negative: if other.min.is_extended_negative: if other.max.is_zero: return AccumBounds(self.max / other.min, oo) if other.max.is_extended_positive: # if we were dealing with intervals we would return # Union(Interval(-oo, self.max/other.max), # Interval(self.max/other.min, oo)) return AccumBounds(-oo, oo) if other.min.is_zero and other.max.is_extended_positive: return AccumBounds(-oo, self.max / other.max) if self.min.is_extended_positive: if other.min.is_extended_negative: if other.max.is_zero: return AccumBounds(-oo, self.min / other.min) if other.max.is_extended_positive: # if we were dealing with intervals we would return # Union(Interval(-oo, self.min/other.min), # Interval(self.min/other.max, oo)) return AccumBounds(-oo, oo) if other.min.is_zero and other.max.is_extended_positive: return AccumBounds(self.min / other.max, oo) elif other.is_extended_real: if other in (S.Infinity, S.NegativeInfinity): if self == AccumBounds(-oo, oo): return AccumBounds(-oo, oo) if self.max is S.Infinity: return AccumBounds(Min(0, other), Max(0, other)) if self.min is S.NegativeInfinity: return AccumBounds(Min(0, -other), Max(0, -other)) if other.is_extended_positive: return AccumBounds(self.min / other, self.max / other) elif other.is_extended_negative: return AccumBounds(self.max / other, self.min / other) if (1 / other) is S.ComplexInfinity: return Mul(self, 1 / other, evaluate=False) else: return Mul(self, 1 / other) return NotImplemented @_sympifyit('other', NotImplemented) def __rtruediv__(self, other): if isinstance(other, Expr): if other.is_extended_real: if other.is_zero: return S.Zero if (self.min.is_extended_nonpositive and self.max.is_extended_nonnegative): if self.min.is_zero: if other.is_extended_positive: return AccumBounds(Mul(other, 1 / self.max), oo) if other.is_extended_negative: return AccumBounds(-oo, Mul(other, 1 / self.max)) if self.max.is_zero: if other.is_extended_positive: return AccumBounds(-oo, Mul(other, 1 / self.min)) if other.is_extended_negative: return AccumBounds(Mul(other, 1 / self.min), oo) return AccumBounds(-oo, oo) else: return AccumBounds(Min(other / self.min, other / self.max), Max(other / self.min, other / self.max)) return Mul(other, 1 / self, evaluate=False) else: return NotImplemented @_sympifyit('other', NotImplemented) def __pow__(self, other): if isinstance(other, Expr): if other is S.Infinity: if self.min.is_extended_nonnegative: if self.max < 1: return S.Zero if self.min > 1: return S.Infinity return AccumBounds(0, oo) elif self.max.is_extended_negative: if self.min > -1: return S.Zero if self.max < -1: return zoo return S.NaN else: if self.min > -1: if self.max < 1: return S.Zero return AccumBounds(0, oo) return AccumBounds(-oo, oo) if other is S.NegativeInfinity: return (1/self)**oo # generically true if (self.max - self.min).is_nonnegative: # well defined if self.min.is_nonnegative: # no 0 to worry about if other.is_nonnegative: # no infinity to worry about return self.func(self.min**other, self.max**other) if other.is_zero: return S.One # x**0 = 1 if other.is_Integer or other.is_integer: if self.min.is_extended_positive: return AccumBounds( Min(self.min**other, self.max**other), Max(self.min**other, self.max**other)) elif self.max.is_extended_negative: return AccumBounds( Min(self.max**other, self.min**other), Max(self.max**other, self.min**other)) if other % 2 == 0: if other.is_extended_negative: if self.min.is_zero: return AccumBounds(self.max**other, oo) if self.max.is_zero: return AccumBounds(self.min**other, oo) return (1/self)**(-other) return AccumBounds( S.Zero, Max(self.min**other, self.max**other)) elif other % 2 == 1: if other.is_extended_negative: if self.min.is_zero: return AccumBounds(self.max**other, oo) if self.max.is_zero: return AccumBounds(-oo, self.min**other) return (1/self)**(-other) return AccumBounds(self.min**other, self.max**other) # non-integer exponent # 0**neg or neg**frac yields complex if (other.is_number or other.is_rational) and ( self.min.is_extended_nonnegative or ( other.is_extended_nonnegative and self.min.is_extended_nonnegative)): num, den = other.as_numer_denom() if num is S.One: return AccumBounds(*[i**(1/den) for i in self.args]) elif den is not S.One: # e.g. if other is not Float return (self**num)**(1/den) # ok for non-negative base if isinstance(other, AccumBounds): if (self.min.is_extended_positive or self.min.is_extended_nonnegative and other.min.is_extended_nonnegative): p = [self**i for i in other.args] if not any(i.is_Pow for i in p): a = [j for i in p for j in i.args or (i,)] try: return self.func(min(a), max(a)) except TypeError: # can't sort pass return Pow(self, other, evaluate=False) return NotImplemented @_sympifyit('other', NotImplemented) def __rpow__(self, other): if other.is_real and other.is_extended_nonnegative and ( self.max - self.min).is_extended_positive: if other is S.One: return S.One if other.is_extended_positive: a, b = [other**i for i in self.args] if min(a, b) != a: a, b = b, a return self.func(a, b) if other.is_zero: if self.min.is_zero: return self.func(0, 1) if self.min.is_extended_positive: return S.Zero return Pow(other, self, evaluate=False) def __abs__(self): if self.max.is_extended_negative: return self.__neg__() elif self.min.is_extended_negative: return AccumBounds(S.Zero, Max(abs(self.min), self.max)) else: return self def __contains__(self, other): """ Returns ``True`` if other is contained in self, where other belongs to extended real numbers, ``False`` if not contained, otherwise TypeError is raised. Examples ======== >>> from sympy import AccumBounds, oo >>> 1 in AccumBounds(-1, 3) True -oo and oo go together as limits (in AccumulationBounds). >>> -oo in AccumBounds(1, oo) True >>> oo in AccumBounds(-oo, 0) True """ other = _sympify(other) if other in (S.Infinity, S.NegativeInfinity): if self.min is S.NegativeInfinity or self.max is S.Infinity: return True return False rv = And(self.min <= other, self.max >= other) if rv not in (True, False): raise TypeError("input failed to evaluate") return rv def intersection(self, other): """ Returns the intersection of 'self' and 'other'. Here other can be an instance of :py:class:`~.FiniteSet` or AccumulationBounds. Parameters ========== other : AccumulationBounds Another AccumulationBounds object with which the intersection has to be computed. Returns ======= AccumulationBounds Intersection of ``self`` and ``other``. Examples ======== >>> from sympy import AccumBounds, FiniteSet >>> AccumBounds(1, 3).intersection(AccumBounds(2, 4)) AccumBounds(2, 3) >>> AccumBounds(1, 3).intersection(AccumBounds(4, 6)) EmptySet >>> AccumBounds(1, 4).intersection(FiniteSet(1, 2, 5)) {1, 2} """ if not isinstance(other, (AccumBounds, FiniteSet)): raise TypeError( "Input must be AccumulationBounds or FiniteSet object") if isinstance(other, FiniteSet): fin_set = S.EmptySet for i in other: if i in self: fin_set = fin_set + FiniteSet(i) return fin_set if self.max < other.min or self.min > other.max: return S.EmptySet if self.min <= other.min: if self.max <= other.max: return AccumBounds(other.min, self.max) if self.max > other.max: return other if other.min <= self.min: if other.max < self.max: return AccumBounds(self.min, other.max) if other.max > self.max: return self def union(self, other): # TODO : Devise a better method for Union of AccumBounds # this method is not actually correct and # can be made better if not isinstance(other, AccumBounds): raise TypeError( "Input must be AccumulationBounds or FiniteSet object") if self.min <= other.min and self.max >= other.min: return AccumBounds(self.min, Max(self.max, other.max)) if other.min <= self.min and other.max >= self.min: return AccumBounds(other.min, Max(self.max, other.max)) @dispatch(AccumulationBounds, AccumulationBounds) # type: ignore # noqa:F811 def _eval_is_le(lhs, rhs): # noqa:F811 if is_le(lhs.max, rhs.min): return True if is_gt(lhs.min, rhs.max): return False @dispatch(AccumulationBounds, Basic) # type: ignore # noqa:F811 def _eval_is_le(lhs, rhs): # noqa: F811 """ Returns ``True `` if range of values attained by ``lhs`` AccumulationBounds object is greater than the range of values attained by ``rhs``, where ``rhs`` may be any value of type AccumulationBounds object or extended real number value, ``False`` if ``rhs`` satisfies the same property, else an unevaluated :py:class:`~.Relational`. Examples ======== >>> from sympy import AccumBounds, oo >>> AccumBounds(1, 3) > AccumBounds(4, oo) False >>> AccumBounds(1, 4) > AccumBounds(3, 4) AccumBounds(1, 4) > AccumBounds(3, 4) >>> AccumBounds(1, oo) > -1 True """ if not rhs.is_extended_real: raise TypeError( "Invalid comparison of %s %s" % (type(rhs), rhs)) elif rhs.is_comparable: if is_le(lhs.max, rhs): return True if is_gt(lhs.min, rhs): return False @dispatch(AccumulationBounds, AccumulationBounds) def _eval_is_ge(lhs, rhs): # noqa:F811 if is_ge(lhs.min, rhs.max): return True if is_lt(lhs.max, rhs.min): return False @dispatch(AccumulationBounds, Expr) # type:ignore def _eval_is_ge(lhs, rhs): # noqa: F811 """ Returns ``True`` if range of values attained by ``lhs`` AccumulationBounds object is less that the range of values attained by ``rhs``, where other may be any value of type AccumulationBounds object or extended real number value, ``False`` if ``rhs`` satisfies the same property, else an unevaluated :py:class:`~.Relational`. Examples ======== >>> from sympy import AccumBounds, oo >>> AccumBounds(1, 3) >= AccumBounds(4, oo) False >>> AccumBounds(1, 4) >= AccumBounds(3, 4) AccumBounds(1, 4) >= AccumBounds(3, 4) >>> AccumBounds(1, oo) >= 1 True """ if not rhs.is_extended_real: raise TypeError( "Invalid comparison of %s %s" % (type(rhs), rhs)) elif rhs.is_comparable: if is_ge(lhs.min, rhs): return True if is_lt(lhs.max, rhs): return False @dispatch(Expr, AccumulationBounds) # type:ignore def _eval_is_ge(lhs, rhs): # noqa:F811 if not lhs.is_extended_real: raise TypeError( "Invalid comparison of %s %s" % (type(lhs), lhs)) elif lhs.is_comparable: if is_le(rhs.max, lhs): return True if is_gt(rhs.min, lhs): return False @dispatch(AccumulationBounds, AccumulationBounds) # type:ignore def _eval_is_ge(lhs, rhs): # noqa:F811 if is_ge(lhs.min, rhs.max): return True if is_lt(lhs.max, rhs.min): return False # setting an alias for AccumulationBounds AccumBounds = AccumulationBounds