"""Square-free decomposition algorithms and related tools. """ from sympy.polys.densearith import ( dup_neg, dmp_neg, dup_sub, dmp_sub, dup_mul, dmp_mul, dup_quo, dmp_quo, dup_mul_ground, dmp_mul_ground) from sympy.polys.densebasic import ( dup_strip, dup_LC, dmp_ground_LC, dmp_zero_p, dmp_ground, dup_degree, dmp_degree, dmp_degree_in, dmp_degree_list, dmp_raise, dmp_inject, dup_convert) from sympy.polys.densetools import ( dup_diff, dmp_diff, dmp_diff_in, dup_shift, dmp_shift, dup_monic, dmp_ground_monic, dup_primitive, dmp_ground_primitive) from sympy.polys.euclidtools import ( dup_inner_gcd, dmp_inner_gcd, dup_gcd, dmp_gcd, dmp_resultant, dmp_primitive) from sympy.polys.galoistools import ( gf_sqf_list, gf_sqf_part) from sympy.polys.polyerrors import ( MultivariatePolynomialError, DomainError) def _dup_check_degrees(f, result): """Sanity check the degrees of a computed factorization in K[x].""" deg = sum(k * dup_degree(fac) for (fac, k) in result) assert deg == dup_degree(f) def _dmp_check_degrees(f, u, result): """Sanity check the degrees of a computed factorization in K[X].""" degs = [0] * (u + 1) for fac, k in result: degs_fac = dmp_degree_list(fac, u) degs = [d1 + k * d2 for d1, d2 in zip(degs, degs_fac)] assert tuple(degs) == dmp_degree_list(f, u) def dup_sqf_p(f, K): """ Return ``True`` if ``f`` is a square-free polynomial in ``K[x]``. Examples ======== >>> from sympy.polys import ring, ZZ >>> R, x = ring("x", ZZ) >>> R.dup_sqf_p(x**2 - 2*x + 1) False >>> R.dup_sqf_p(x**2 - 1) True """ if not f: return True else: return not dup_degree(dup_gcd(f, dup_diff(f, 1, K), K)) def dmp_sqf_p(f, u, K): """ Return ``True`` if ``f`` is a square-free polynomial in ``K[X]``. Examples ======== >>> from sympy.polys import ring, ZZ >>> R, x,y = ring("x,y", ZZ) >>> R.dmp_sqf_p(x**2 + 2*x*y + y**2) False >>> R.dmp_sqf_p(x**2 + y**2) True """ if dmp_zero_p(f, u): return True for i in range(u+1): fp = dmp_diff_in(f, 1, i, u, K) if dmp_zero_p(fp, u): continue gcd = dmp_gcd(f, fp, u, K) if dmp_degree_in(gcd, i, u) != 0: return False return True def dup_sqf_norm(f, K): r""" Find a shift of `f` in `K[x]` that has square-free norm. The domain `K` must be an algebraic number field `k(a)` (see :ref:`QQ(a)`). Returns `(s,g,r)`, such that `g(x)=f(x-sa)`, `r(x)=\text{Norm}(g(x))` and `r` is a square-free polynomial over `k`. Examples ======== We first create the algebraic number field `K=k(a)=\mathbb{Q}(\sqrt{3})` and rings `K[x]` and `k[x]`: >>> from sympy.polys import ring, QQ >>> from sympy import sqrt >>> K = QQ.algebraic_field(sqrt(3)) >>> R, x = ring("x", K) >>> _, X = ring("x", QQ) We can now find a square free norm for a shift of `f`: >>> f = x**2 - 1 >>> s, g, r = R.dup_sqf_norm(f) The choice of shift `s` is arbitrary and the particular values returned for `g` and `r` are determined by `s`. >>> s == 1 True >>> g == x**2 - 2*sqrt(3)*x + 2 True >>> r == X**4 - 8*X**2 + 4 True The invariants are: >>> g == f.shift(-s*K.unit) True >>> g.norm() == r True >>> r.is_squarefree True Explanation =========== This is part of Trager's algorithm for factorizing polynomials over algebraic number fields. In particular this function is algorithm ``sqfr_norm`` from [Trager76]_. See Also ======== dmp_sqf_norm: Analogous function for multivariate polynomials over ``k(a)``. dmp_norm: Computes the norm of `f` directly without any shift. dup_ext_factor: Function implementing Trager's algorithm that uses this. sympy.polys.polytools.sqf_norm: High-level interface for using this function. """ if not K.is_Algebraic: raise DomainError("ground domain must be algebraic") s, g = 0, dmp_raise(K.mod.to_list(), 1, 0, K.dom) while True: h, _ = dmp_inject(f, 0, K, front=True) r = dmp_resultant(g, h, 1, K.dom) if dup_sqf_p(r, K.dom): break else: f, s = dup_shift(f, -K.unit, K), s + 1 return s, f, r def _dmp_sqf_norm_shifts(f, u, K): """Generate a sequence of candidate shifts for dmp_sqf_norm.""" # # We want to find a minimal shift if possible because shifting high degree # variables can be expensive e.g. x**10 -> (x + 1)**10. We try a few easy # cases first before the final infinite loop that is guaranteed to give # only finitely many bad shifts (see Trager76 for proof of this in the # univariate case). # # First the trivial shift [0, 0, ...] n = u + 1 s0 = [0] * n yield s0, f # Shift in multiples of the generator of the extension field K a = K.unit # Variables of degree > 0 ordered by increasing degree d = dmp_degree_list(f, u) var_indices = [i for di, i in sorted(zip(d, range(u+1))) if di > 0] # Now try [1, 0, 0, ...], [0, 1, 0, ...] for i in var_indices: s1 = s0.copy() s1[i] = 1 a1 = [-a*s1i for s1i in s1] f1 = dmp_shift(f, a1, u, K) yield s1, f1 # Now try [1, 1, 1, ...], [2, 2, 2, ...] j = 0 while True: j += 1 sj = [j] * n aj = [-a*j] * n fj = dmp_shift(f, aj, u, K) yield sj, fj def dmp_sqf_norm(f, u, K): r""" Find a shift of ``f`` in ``K[X]`` that has square-free norm. The domain `K` must be an algebraic number field `k(a)` (see :ref:`QQ(a)`). Returns `(s,g,r)`, such that `g(x_1,x_2,\cdots)=f(x_1-s_1 a, x_2 - s_2 a, \cdots)`, `r(x)=\text{Norm}(g(x))` and `r` is a square-free polynomial over `k`. Examples ======== We first create the algebraic number field `K=k(a)=\mathbb{Q}(i)` and rings `K[x,y]` and `k[x,y]`: >>> from sympy.polys import ring, QQ >>> from sympy import I >>> K = QQ.algebraic_field(I) >>> R, x, y = ring("x,y", K) >>> _, X, Y = ring("x,y", QQ) We can now find a square free norm for a shift of `f`: >>> f = x*y + y**2 >>> s, g, r = R.dmp_sqf_norm(f) The choice of shifts ``s`` is arbitrary and the particular values returned for ``g`` and ``r`` are determined by ``s``. >>> s [0, 1] >>> g == x*y - I*x + y**2 - 2*I*y - 1 True >>> r == X**2*Y**2 + X**2 + 2*X*Y**3 + 2*X*Y + Y**4 + 2*Y**2 + 1 True The required invariants are: >>> g == f.shift_list([-si*K.unit for si in s]) True >>> g.norm() == r True >>> r.is_squarefree True Explanation =========== This is part of Trager's algorithm for factorizing polynomials over algebraic number fields. In particular this function is a multivariate generalization of algorithm ``sqfr_norm`` from [Trager76]_. See Also ======== dup_sqf_norm: Analogous function for univariate polynomials over ``k(a)``. dmp_norm: Computes the norm of `f` directly without any shift. dmp_ext_factor: Function implementing Trager's algorithm that uses this. sympy.polys.polytools.sqf_norm: High-level interface for using this function. """ if not u: s, g, r = dup_sqf_norm(f, K) return [s], g, r if not K.is_Algebraic: raise DomainError("ground domain must be algebraic") g = dmp_raise(K.mod.to_list(), u + 1, 0, K.dom) for s, f in _dmp_sqf_norm_shifts(f, u, K): h, _ = dmp_inject(f, u, K, front=True) r = dmp_resultant(g, h, u + 1, K.dom) if dmp_sqf_p(r, u, K.dom): break return s, f, r def dmp_norm(f, u, K): r""" Norm of ``f`` in ``K[X]``, often not square-free. The domain `K` must be an algebraic number field `k(a)` (see :ref:`QQ(a)`). Examples ======== We first define the algebraic number field `K = k(a) = \mathbb{Q}(\sqrt{2})`: >>> from sympy import QQ, sqrt >>> from sympy.polys.sqfreetools import dmp_norm >>> k = QQ >>> K = k.algebraic_field(sqrt(2)) We can now compute the norm of a polynomial `p` in `K[x,y]`: >>> p = [[K(1)], [K(1),K.unit]] # x + y + sqrt(2) >>> N = [[k(1)], [k(2),k(0)], [k(1),k(0),k(-2)]] # x**2 + 2*x*y + y**2 - 2 >>> dmp_norm(p, 1, K) == N True In higher level functions that is: >>> from sympy import expand, roots, minpoly >>> from sympy.abc import x, y >>> from math import prod >>> a = sqrt(2) >>> e = (x + y + a) >>> e.as_poly([x, y], extension=a).norm() Poly(x**2 + 2*x*y + y**2 - 2, x, y, domain='QQ') This is equal to the product of the expressions `x + y + a_i` where the `a_i` are the conjugates of `a`: >>> pa = minpoly(a) >>> pa _x**2 - 2 >>> rs = roots(pa, multiple=True) >>> rs [sqrt(2), -sqrt(2)] >>> n = prod(e.subs(a, r) for r in rs) >>> n (x + y - sqrt(2))*(x + y + sqrt(2)) >>> expand(n) x**2 + 2*x*y + y**2 - 2 Explanation =========== Given an algebraic number field `K = k(a)` any element `b` of `K` can be represented as polynomial function `b=g(a)` where `g` is in `k[x]`. If the minimal polynomial of `a` over `k` is `p_a` then the roots `a_1`, `a_2`, `\cdots` of `p_a(x)` are the conjugates of `a`. The norm of `b` is the product `g(a1) \times g(a2) \times \cdots` and is an element of `k`. As in [Trager76]_ we extend this norm to multivariate polynomials over `K`. If `b(x)` is a polynomial in `k(a)[X]` then we can think of `b` as being alternately a function `g_X(a)` where `g_X` is an element of `k[X][y]` i.e. a polynomial function with coefficients that are elements of `k[X]`. Then the norm of `b` is the product `g_X(a1) \times g_X(a2) \times \cdots` and will be an element of `k[X]`. See Also ======== dmp_sqf_norm: Compute a shift of `f` so that the `\text{Norm}(f)` is square-free. sympy.polys.polytools.Poly.norm: Higher-level function that calls this. """ if not K.is_Algebraic: raise DomainError("ground domain must be algebraic") g = dmp_raise(K.mod.to_list(), u + 1, 0, K.dom) h, _ = dmp_inject(f, u, K, front=True) return dmp_resultant(g, h, u + 1, K.dom) def dup_gf_sqf_part(f, K): """Compute square-free part of ``f`` in ``GF(p)[x]``. """ f = dup_convert(f, K, K.dom) g = gf_sqf_part(f, K.mod, K.dom) return dup_convert(g, K.dom, K) def dmp_gf_sqf_part(f, u, K): """Compute square-free part of ``f`` in ``GF(p)[X]``. """ raise NotImplementedError('multivariate polynomials over finite fields') def dup_sqf_part(f, K): """ Returns square-free part of a polynomial in ``K[x]``. Examples ======== >>> from sympy.polys import ring, ZZ >>> R, x = ring("x", ZZ) >>> R.dup_sqf_part(x**3 - 3*x - 2) x**2 - x - 2 See Also ======== sympy.polys.polytools.Poly.sqf_part """ if K.is_FiniteField: return dup_gf_sqf_part(f, K) if not f: return f if K.is_negative(dup_LC(f, K)): f = dup_neg(f, K) gcd = dup_gcd(f, dup_diff(f, 1, K), K) sqf = dup_quo(f, gcd, K) if K.is_Field: return dup_monic(sqf, K) else: return dup_primitive(sqf, K)[1] def dmp_sqf_part(f, u, K): """ Returns square-free part of a polynomial in ``K[X]``. Examples ======== >>> from sympy.polys import ring, ZZ >>> R, x,y = ring("x,y", ZZ) >>> R.dmp_sqf_part(x**3 + 2*x**2*y + x*y**2) x**2 + x*y """ if not u: return dup_sqf_part(f, K) if K.is_FiniteField: return dmp_gf_sqf_part(f, u, K) if dmp_zero_p(f, u): return f if K.is_negative(dmp_ground_LC(f, u, K)): f = dmp_neg(f, u, K) gcd = f for i in range(u+1): gcd = dmp_gcd(gcd, dmp_diff_in(f, 1, i, u, K), u, K) sqf = dmp_quo(f, gcd, u, K) if K.is_Field: return dmp_ground_monic(sqf, u, K) else: return dmp_ground_primitive(sqf, u, K)[1] def dup_gf_sqf_list(f, K, all=False): """Compute square-free decomposition of ``f`` in ``GF(p)[x]``. """ f_orig = f f = dup_convert(f, K, K.dom) coeff, factors = gf_sqf_list(f, K.mod, K.dom, all=all) for i, (f, k) in enumerate(factors): factors[i] = (dup_convert(f, K.dom, K), k) _dup_check_degrees(f_orig, factors) return K.convert(coeff, K.dom), factors def dmp_gf_sqf_list(f, u, K, all=False): """Compute square-free decomposition of ``f`` in ``GF(p)[X]``. """ raise NotImplementedError('multivariate polynomials over finite fields') def dup_sqf_list(f, K, all=False): """ Return square-free decomposition of a polynomial in ``K[x]``. Uses Yun's algorithm from [Yun76]_. Examples ======== >>> from sympy.polys import ring, ZZ >>> R, x = ring("x", ZZ) >>> f = 2*x**5 + 16*x**4 + 50*x**3 + 76*x**2 + 56*x + 16 >>> R.dup_sqf_list(f) (2, [(x + 1, 2), (x + 2, 3)]) >>> R.dup_sqf_list(f, all=True) (2, [(1, 1), (x + 1, 2), (x + 2, 3)]) See Also ======== dmp_sqf_list: Corresponding function for multivariate polynomials. sympy.polys.polytools.sqf_list: High-level function for square-free factorization of expressions. sympy.polys.polytools.Poly.sqf_list: Analogous method on :class:`~.Poly`. References ========== [Yun76]_ """ if K.is_FiniteField: return dup_gf_sqf_list(f, K, all=all) f_orig = f if K.is_Field: coeff = dup_LC(f, K) f = dup_monic(f, K) else: coeff, f = dup_primitive(f, K) if K.is_negative(dup_LC(f, K)): f = dup_neg(f, K) coeff = -coeff if dup_degree(f) <= 0: return coeff, [] result, i = [], 1 h = dup_diff(f, 1, K) g, p, q = dup_inner_gcd(f, h, K) while True: d = dup_diff(p, 1, K) h = dup_sub(q, d, K) if not h: result.append((p, i)) break g, p, q = dup_inner_gcd(p, h, K) if all or dup_degree(g) > 0: result.append((g, i)) i += 1 _dup_check_degrees(f_orig, result) return coeff, result def dup_sqf_list_include(f, K, all=False): """ Return square-free decomposition of a polynomial in ``K[x]``. Examples ======== >>> from sympy.polys import ring, ZZ >>> R, x = ring("x", ZZ) >>> f = 2*x**5 + 16*x**4 + 50*x**3 + 76*x**2 + 56*x + 16 >>> R.dup_sqf_list_include(f) [(2, 1), (x + 1, 2), (x + 2, 3)] >>> R.dup_sqf_list_include(f, all=True) [(2, 1), (x + 1, 2), (x + 2, 3)] """ coeff, factors = dup_sqf_list(f, K, all=all) if factors and factors[0][1] == 1: g = dup_mul_ground(factors[0][0], coeff, K) return [(g, 1)] + factors[1:] else: g = dup_strip([coeff]) return [(g, 1)] + factors def dmp_sqf_list(f, u, K, all=False): """ Return square-free decomposition of a polynomial in `K[X]`. Examples ======== >>> from sympy.polys import ring, ZZ >>> R, x,y = ring("x,y", ZZ) >>> f = x**5 + 2*x**4*y + x**3*y**2 >>> R.dmp_sqf_list(f) (1, [(x + y, 2), (x, 3)]) >>> R.dmp_sqf_list(f, all=True) (1, [(1, 1), (x + y, 2), (x, 3)]) Explanation =========== Uses Yun's algorithm for univariate polynomials from [Yun76]_ recrusively. The multivariate polynomial is treated as a univariate polynomial in its leading variable. Then Yun's algorithm computes the square-free factorization of the primitive and the content is factored recursively. It would be better to use a dedicated algorithm for multivariate polynomials instead. See Also ======== dup_sqf_list: Corresponding function for univariate polynomials. sympy.polys.polytools.sqf_list: High-level function for square-free factorization of expressions. sympy.polys.polytools.Poly.sqf_list: Analogous method on :class:`~.Poly`. """ if not u: return dup_sqf_list(f, K, all=all) if K.is_FiniteField: return dmp_gf_sqf_list(f, u, K, all=all) f_orig = f if K.is_Field: coeff = dmp_ground_LC(f, u, K) f = dmp_ground_monic(f, u, K) else: coeff, f = dmp_ground_primitive(f, u, K) if K.is_negative(dmp_ground_LC(f, u, K)): f = dmp_neg(f, u, K) coeff = -coeff deg = dmp_degree(f, u) if deg < 0: return coeff, [] # Yun's algorithm requires the polynomial to be primitive as a univariate # polynomial in its main variable. content, f = dmp_primitive(f, u, K) result = {} if deg != 0: h = dmp_diff(f, 1, u, K) g, p, q = dmp_inner_gcd(f, h, u, K) i = 1 while True: d = dmp_diff(p, 1, u, K) h = dmp_sub(q, d, u, K) if dmp_zero_p(h, u): result[i] = p break g, p, q = dmp_inner_gcd(p, h, u, K) if all or dmp_degree(g, u) > 0: result[i] = g i += 1 coeff_content, result_content = dmp_sqf_list(content, u-1, K, all=all) coeff *= coeff_content # Combine factors of the content and primitive part that have the same # multiplicity to produce a list in ascending order of multiplicity. for fac, i in result_content: fac = [fac] if i in result: result[i] = dmp_mul(result[i], fac, u, K) else: result[i] = fac result = [(result[i], i) for i in sorted(result)] _dmp_check_degrees(f_orig, u, result) return coeff, result def dmp_sqf_list_include(f, u, K, all=False): """ Return square-free decomposition of a polynomial in ``K[x]``. Examples ======== >>> from sympy.polys import ring, ZZ >>> R, x,y = ring("x,y", ZZ) >>> f = x**5 + 2*x**4*y + x**3*y**2 >>> R.dmp_sqf_list_include(f) [(1, 1), (x + y, 2), (x, 3)] >>> R.dmp_sqf_list_include(f, all=True) [(1, 1), (x + y, 2), (x, 3)] """ if not u: return dup_sqf_list_include(f, K, all=all) coeff, factors = dmp_sqf_list(f, u, K, all=all) if factors and factors[0][1] == 1: g = dmp_mul_ground(factors[0][0], coeff, u, K) return [(g, 1)] + factors[1:] else: g = dmp_ground(coeff, u) return [(g, 1)] + factors def dup_gff_list(f, K): """ Compute greatest factorial factorization of ``f`` in ``K[x]``. Examples ======== >>> from sympy.polys import ring, ZZ >>> R, x = ring("x", ZZ) >>> R.dup_gff_list(x**5 + 2*x**4 - x**3 - 2*x**2) [(x, 1), (x + 2, 4)] """ if not f: raise ValueError("greatest factorial factorization doesn't exist for a zero polynomial") f = dup_monic(f, K) if not dup_degree(f): return [] else: g = dup_gcd(f, dup_shift(f, K.one, K), K) H = dup_gff_list(g, K) for i, (h, k) in enumerate(H): g = dup_mul(g, dup_shift(h, -K(k), K), K) H[i] = (h, k + 1) f = dup_quo(f, g, K) if not dup_degree(f): return H else: return [(f, 1)] + H def dmp_gff_list(f, u, K): """ Compute greatest factorial factorization of ``f`` in ``K[X]``. Examples ======== >>> from sympy.polys import ring, ZZ >>> R, x,y = ring("x,y", ZZ) """ if not u: return dup_gff_list(f, K) else: raise MultivariatePolynomialError(f)