from sage.modules.with_basis.indexed_element import IndexedFreeModuleElement
from mjo.eja.eja_operator import FiniteDimensionalEJAOperator
-from mjo.eja.eja_utils import _mat2vec
+from mjo.eja.eja_utils import _scale
+
class FiniteDimensionalEJAElement(IndexedFreeModuleElement):
"""
The definition of `x^2` is the unambiguous `x*x`::
- sage: set_random_seed()
sage: x = random_eja().random_element()
sage: x*x == (x^2)
True
A few examples of power-associativity::
- sage: set_random_seed()
sage: x = random_eja().random_element()
sage: x*(x*x)*(x*x) == x^5
True
We also know that powers operator-commute (Koecher, Chapter
III, Corollary 1)::
- sage: set_random_seed()
sage: x = random_eja().random_element()
sage: m = ZZ.random_element(0,10)
sage: n = ZZ.random_element(0,10)
We should always get back an element of the algebra::
- sage: set_random_seed()
sage: p = PolynomialRing(AA, 't').random_element()
sage: J = random_eja()
sage: x = J.random_element()
SETUP::
- sage: from mjo.eja.eja_algebra import HadamardEJA
+ sage: from mjo.eja.eja_algebra import (random_eja,
+ ....: HadamardEJA)
EXAMPLES:
The characteristic polynomial of an element should evaluate
to zero on that element::
- sage: set_random_seed()
- sage: x = HadamardEJA(3).random_element()
+ sage: x = random_eja().random_element()
sage: p = x.characteristic_polynomial()
- sage: x.apply_univariate_polynomial(p)
- 0
+ sage: x.apply_univariate_polynomial(p).is_zero()
+ True
The characteristic polynomials of the zero and unit elements
should be what we think they are in a subalgebra, too::
sage: J = HadamardEJA(3)
sage: p1 = J.one().characteristic_polynomial()
sage: q1 = J.zero().characteristic_polynomial()
- sage: e0,e1,e2 = J.gens()
- sage: A = (e0 + 2*e1 + 3*e2).subalgebra_generated_by() # dim 3
+ sage: b0,b1,b2 = J.gens()
+ sage: A = (b0 + 2*b1 + 3*b2).subalgebra_generated_by() # dim 3
sage: p2 = A.one().characteristic_polynomial()
sage: q2 = A.zero().characteristic_polynomial()
sage: p1 == p2
Ensure that we can always compute an inner product, and that
it gives us back a real number::
- sage: set_random_seed()
sage: J = random_eja()
sage: x,y = J.random_elements(2)
sage: x.inner_product(y) in RLF
The definition of a Jordan algebra says that any element
operator-commutes with its square::
- sage: set_random_seed()
sage: x = random_eja().random_element()
sage: x.operator_commutes_with(x^2)
True
Test Lemma 1 from Chapter III of Koecher::
- sage: set_random_seed()
sage: u,v = random_eja().random_elements(2)
sage: lhs = u.operator_commutes_with(u*v)
sage: rhs = v.operator_commutes_with(u^2)
Test the first polarization identity from my notes, Koecher
Chapter III, or from Baes (2.3)::
- sage: set_random_seed()
sage: x,y = random_eja().random_elements(2)
sage: Lx = x.operator()
sage: Ly = y.operator()
Test the second polarization identity from my notes or from
Baes (2.4)::
- sage: set_random_seed()
- sage: x,y,z = random_eja().random_elements(3)
- sage: Lx = x.operator()
- sage: Ly = y.operator()
- sage: Lz = z.operator()
- sage: Lzy = (z*y).operator()
- sage: Lxy = (x*y).operator()
- sage: Lxz = (x*z).operator()
- sage: bool(Lx*Lzy + Lz*Lxy + Ly*Lxz == Lzy*Lx + Lxy*Lz + Lxz*Ly)
+ sage: x,y,z = random_eja().random_elements(3) # long time
+ sage: Lx = x.operator() # long time
+ sage: Ly = y.operator() # long time
+ sage: Lz = z.operator() # long time
+ sage: Lzy = (z*y).operator() # long time
+ sage: Lxy = (x*y).operator() # long time
+ sage: Lxz = (x*z).operator() # long time
+ sage: lhs = Lx*Lzy + Lz*Lxy + Ly*Lxz # long time
+ sage: rhs = Lzy*Lx + Lxy*Lz + Lxz*Ly # long time
+ sage: bool(lhs == rhs) # long time
True
Test the third polarization identity from my notes or from
Baes (2.5)::
- sage: set_random_seed()
- sage: u,y,z = random_eja().random_elements(3)
- sage: Lu = u.operator()
- sage: Ly = y.operator()
- sage: Lz = z.operator()
- sage: Lzy = (z*y).operator()
- sage: Luy = (u*y).operator()
- sage: Luz = (u*z).operator()
- sage: Luyz = (u*(y*z)).operator()
- sage: lhs = Lu*Lzy + Lz*Luy + Ly*Luz
- sage: rhs = Luyz + Ly*Lu*Lz + Lz*Lu*Ly
- sage: bool(lhs == rhs)
+ sage: u,y,z = random_eja().random_elements(3) # long time
+ sage: Lu = u.operator() # long time
+ sage: Ly = y.operator() # long time
+ sage: Lz = z.operator() # long time
+ sage: Lzy = (z*y).operator() # long time
+ sage: Luy = (u*y).operator() # long time
+ sage: Luz = (u*z).operator() # long time
+ sage: Luyz = (u*(y*z)).operator() # long time
+ sage: lhs = Lu*Lzy + Lz*Luy + Ly*Luz # long time
+ sage: rhs = Luyz + Ly*Lu*Lz + Lz*Lu*Ly # long time
+ sage: bool(lhs == rhs) # long time
True
"""
EXAMPLES::
sage: J = JordanSpinEJA(2)
- sage: e0,e1 = J.gens()
sage: x = sum( J.gens() )
sage: x.det()
0
::
sage: J = JordanSpinEJA(3)
- sage: e0,e1,e2 = J.gens()
sage: x = sum( J.gens() )
sage: x.det()
-1
An element is invertible if and only if its determinant is
non-zero::
- sage: set_random_seed()
sage: x = random_eja().random_element()
sage: x.is_invertible() == (x.det() != 0)
True
Ensure that the determinant is multiplicative on an associative
subalgebra as in Faraut and Korányi's Proposition II.2.2::
- sage: set_random_seed()
- sage: J = random_eja().random_element().subalgebra_generated_by()
+ sage: x0 = random_eja().random_element()
+ sage: J = x0.subalgebra_generated_by(orthonormalize=False)
sage: x,y = J.random_elements(2)
sage: (x*y).det() == x.det()*y.det()
True
- The determinant in matrix algebras is just the usual determinant::
+ The determinant in real matrix algebras is the usual determinant::
- sage: set_random_seed()
sage: X = matrix.random(QQ,3)
sage: X = X + X.T
sage: J1 = RealSymmetricEJA(3)
sage: actual2 == expected
True
- ::
-
- sage: set_random_seed()
- sage: J1 = ComplexHermitianEJA(2)
- sage: J2 = ComplexHermitianEJA(2,field=QQ,orthonormalize=False)
- sage: X = matrix.random(GaussianIntegers(), 2)
- sage: X = X + X.H
- sage: expected = AA(X.det())
- sage: actual1 = J1(J1.real_embed(X)).det()
- sage: actual2 = J2(J2.real_embed(X)).det()
- sage: expected == actual1
- True
- sage: expected == actual2
- True
-
"""
P = self.parent()
r = P.rank()
The inverse in the spin factor algebra is given in Alizadeh's
Example 11.11::
- sage: set_random_seed()
sage: J = JordanSpinEJA.random_instance()
sage: x = J.random_element()
sage: while not x.is_invertible():
The identity element is its own inverse::
- sage: set_random_seed()
sage: J = random_eja()
sage: J.one().inverse() == J.one()
True
If an element has an inverse, it acts like one::
- sage: set_random_seed()
sage: J = random_eja()
sage: x = J.random_element()
sage: (not x.is_invertible()) or (x.inverse()*x == J.one())
The inverse of the inverse is what we started with::
- sage: set_random_seed()
sage: J = random_eja()
sage: x = J.random_element()
sage: (not x.is_invertible()) or (x.inverse().inverse() == x)
of an element is the inverse of its left-multiplication operator
applied to the algebra's identity, when that inverse exists::
- sage: set_random_seed()
- sage: J = random_eja()
- sage: x = J.random_element()
- sage: (not x.operator().is_invertible()) or (
- ....: x.operator().inverse()(J.one()) == x.inverse() )
+ sage: J = random_eja() # long time
+ sage: x = J.random_element() # long time
+ sage: (not x.operator().is_invertible()) or ( # long time
+ ....: x.operator().inverse()(J.one()) # long time
+ ....: == # long time
+ ....: x.inverse() ) # long time
True
Check that the fast (cached) and slow algorithms give the same
answer::
- sage: set_random_seed() # long time
sage: J = random_eja(field=QQ, orthonormalize=False) # long time
sage: x = J.random_element() # long time
sage: while not x.is_invertible(): # long time
True
"""
not_invertible_msg = "element is not invertible"
- if self.parent()._charpoly_coefficients.is_in_cache():
+
+ algebra = self.parent()
+ if algebra._charpoly_coefficients.is_in_cache():
# We can invert using our charpoly if it will be fast to
# compute. If the coefficients are cached, our rank had
# better be too!
if self.det().is_zero():
raise ZeroDivisionError(not_invertible_msg)
- r = self.parent().rank()
+ r = algebra.rank()
a = self.characteristic_polynomial().coefficients(sparse=False)
- return (-1)**(r+1)*sum(a[i+1]*self**i for i in range(r))/self.det()
+ return (-1)**(r+1)*algebra.sum(a[i+1]*self**i
+ for i in range(r))/self.det()
try:
inv = (~self.quadratic_representation())(self)
The identity element is always invertible::
- sage: set_random_seed()
sage: J = random_eja()
sage: J.one().is_invertible()
True
The zero element is never invertible in a non-trivial algebra::
- sage: set_random_seed()
sage: J = random_eja()
sage: (not J.is_trivial()) and J.zero().is_invertible()
False
Test that the fast (cached) and slow algorithms give the same
answer::
- sage: set_random_seed() # long time
sage: J = random_eja(field=QQ, orthonormalize=False) # long time
sage: x = J.random_element() # long time
sage: slow = x.is_invertible() # long time
element should always be in terms of minimal idempotents::
sage: J = JordanSpinEJA(4)
- sage: x = sum( i*J.gens()[i] for i in range(len(J.gens())) )
+ sage: x = sum( i*J.monomial(i) for i in range(len(J.gens())) )
sage: x.is_regular()
True
sage: [ c.is_primitive_idempotent()
The identity element is minimal only in an EJA of rank one::
- sage: set_random_seed()
sage: J = random_eja()
sage: J.rank() == 1 or not J.one().is_primitive_idempotent()
True
A non-idempotent cannot be a minimal idempotent::
- sage: set_random_seed()
sage: J = JordanSpinEJA(4)
sage: x = J.random_element()
sage: (not x.is_idempotent()) and x.is_primitive_idempotent()
idempotent if and only if it's idempotent with trace equal to
unity::
- sage: set_random_seed()
sage: J = JordanSpinEJA(4)
sage: x = J.random_element()
sage: expected = (x.is_idempotent() and x.trace() == 1)
Primitive idempotents must be non-zero::
- sage: set_random_seed()
sage: J = random_eja()
sage: J.zero().is_idempotent()
True
The identity element is never nilpotent, except in a trivial EJA::
- sage: set_random_seed()
sage: J = random_eja()
sage: J.one().is_nilpotent() and not J.is_trivial()
False
The additive identity is always nilpotent::
- sage: set_random_seed()
sage: random_eja().zero().is_nilpotent()
True
sage: J = JordanSpinEJA(5)
sage: J.one().is_regular()
False
- sage: e0, e1, e2, e3, e4 = J.gens() # e0 is the identity
+ sage: b0, b1, b2, b3, b4 = J.gens()
+ sage: b0 == J.one()
+ True
sage: for x in J.gens():
....: (J.one() + x).is_regular()
False
The zero element should never be regular, unless the parent
algebra has dimension less than or equal to one::
- sage: set_random_seed()
sage: J = random_eja()
sage: J.dimension() <= 1 or not J.zero().is_regular()
True
The unit element isn't regular unless the algebra happens to
consist of only its scalar multiples::
- sage: set_random_seed()
sage: J = random_eja()
sage: J.dimension() <= 1 or not J.one().is_regular()
True
sage: J = JordanSpinEJA(4)
sage: J.one().degree()
1
- sage: e0,e1,e2,e3 = J.gens()
- sage: (e0 - e1).degree()
+ sage: b0,b1,b2,b3 = J.gens()
+ sage: (b0 - b1).degree()
2
In the spin factor algebra (of rank two), all elements that
aren't multiples of the identity are regular::
- sage: set_random_seed()
sage: J = JordanSpinEJA.random_instance()
sage: n = J.dimension()
sage: x = J.random_element()
The zero and unit elements are both of degree one in nontrivial
algebras::
- sage: set_random_seed()
sage: J = random_eja()
sage: d = J.zero().degree()
sage: (J.is_trivial() and d == 0) or d == 1
Our implementation agrees with the definition::
- sage: set_random_seed()
sage: x = random_eja().random_element()
sage: x.degree() == x.minimal_polynomial().degree()
True
M = matrix([(self.parent().one()).to_vector()])
old_rank = 1
- # Specifying the row-reduction algorithm can e.g. help over
+ # Specifying the row-reduction algorithm can e.g. help over
# AA because it avoids the RecursionError that gets thrown
# when we have to look too hard for a root.
#
always the same, except in trivial algebras where the minimal
polynomial of the unit/zero element is ``1``::
- sage: set_random_seed()
sage: J = random_eja()
sage: mu = J.one().minimal_polynomial()
sage: t = mu.parent().gen()
The degree of an element is (by one definition) the degree
of its minimal polynomial::
- sage: set_random_seed()
sage: x = random_eja().random_element()
sage: x.degree() == x.minimal_polynomial().degree()
True
identity. We require the dimension of the algebra to be at least
two here so that said elements actually exist::
- sage: set_random_seed()
- sage: n_max = max(2, JordanSpinEJA._max_random_instance_size())
- sage: n = ZZ.random_element(2, n_max)
+ sage: d_max = JordanSpinEJA._max_random_instance_dimension()
+ sage: n = ZZ.random_element(2, max(2,d_max))
sage: J = JordanSpinEJA(n)
sage: y = J.random_element()
sage: while y == y.coefficient(0)*J.one():
The minimal polynomial should always kill its element::
- sage: set_random_seed()
- sage: x = random_eja().random_element()
- sage: p = x.minimal_polynomial()
- sage: x.apply_univariate_polynomial(p)
+ sage: x = random_eja().random_element() # long time
+ sage: p = x.minimal_polynomial() # long time
+ sage: x.apply_univariate_polynomial(p) # long time
0
The minimal polynomial is invariant under a change of basis,
and in particular, a re-scaling of the basis::
- sage: set_random_seed()
- sage: n_max = RealSymmetricEJA._max_random_instance_size()
- sage: n = ZZ.random_element(1, n_max)
+ sage: d_max = RealSymmetricEJA._max_random_instance_dimension()
+ sage: d = ZZ.random_element(1, d_max)
+ sage: n = RealSymmetricEJA._max_random_instance_size(d)
sage: J1 = RealSymmetricEJA(n)
sage: J2 = RealSymmetricEJA(n,orthonormalize=False)
sage: X = random_matrix(AA,n)
"""
if self.is_zero():
- # We would generate a zero-dimensional subalgebra
- # where the minimal polynomial would be constant.
- # That might be correct, but only if *this* algebra
- # is trivial too.
- if not self.parent().is_trivial():
- # Pretty sure we know what the minimal polynomial of
- # the zero operator is going to be. This ensures
- # consistency of e.g. the polynomial variable returned
- # in the "normal" case without us having to think about it.
- return self.operator().minimal_polynomial()
-
+ # Pretty sure we know what the minimal polynomial of
+ # the zero operator is going to be. This ensures
+ # consistency of e.g. the polynomial variable returned
+ # in the "normal" case without us having to think about it.
+ return self.operator().minimal_polynomial()
+
+ # If we don't orthonormalize the subalgebra's basis, then the
+ # first two monomials in the subalgebra will be self^0 and
+ # self^1... assuming that self^1 is not a scalar multiple of
+ # self^0 (the unit element). We special case these to avoid
+ # having to solve a system to coerce self into the subalgebra.
A = self.subalgebra_generated_by(orthonormalize=False)
- return A(self).operator().minimal_polynomial()
+
+ if A.dimension() == 1:
+ # Does a solve to find the scalar multiple alpha such that
+ # alpha*unit = self. We have to do this because the basis
+ # for the subalgebra will be [ self^0 ], and not [ self^1 ]!
+ unit = self.parent().one()
+ alpha = self.to_vector() / unit.to_vector()
+ return (unit.operator()*alpha).minimal_polynomial()
+ else:
+ # If the dimension of the subalgebra is >= 2, then we just
+ # use the second basis element.
+ return A.monomial(1).operator().minimal_polynomial()
SETUP::
sage: from mjo.eja.eja_algebra import (ComplexHermitianEJA,
- ....: QuaternionHermitianEJA)
+ ....: HadamardEJA,
+ ....: QuaternionHermitianEJA,
+ ....: RealSymmetricEJA)
EXAMPLES::
sage: J = ComplexHermitianEJA(3)
sage: J.one()
- e0 + e3 + e8
+ b0 + b3 + b8
sage: J.one().to_matrix()
- [1 0 0 0 0 0]
- [0 1 0 0 0 0]
- [0 0 1 0 0 0]
- [0 0 0 1 0 0]
- [0 0 0 0 1 0]
- [0 0 0 0 0 1]
+ +---+---+---+
+ | 1 | 0 | 0 |
+ +---+---+---+
+ | 0 | 1 | 0 |
+ +---+---+---+
+ | 0 | 0 | 1 |
+ +---+---+---+
::
sage: J = QuaternionHermitianEJA(2)
sage: J.one()
- e0 + e5
+ b0 + b5
sage: J.one().to_matrix()
- [1 0 0 0 0 0 0 0]
- [0 1 0 0 0 0 0 0]
- [0 0 1 0 0 0 0 0]
- [0 0 0 1 0 0 0 0]
- [0 0 0 0 1 0 0 0]
- [0 0 0 0 0 1 0 0]
- [0 0 0 0 0 0 1 0]
- [0 0 0 0 0 0 0 1]
+ +---+---+
+ | 1 | 0 |
+ +---+---+
+ | 0 | 1 |
+ +---+---+
This also works in Cartesian product algebras::
B = self.parent().matrix_basis()
W = self.parent().matrix_space()
- if self.parent()._matrix_basis_is_cartesian:
- # Aaaaand linear combinations don't work in Cartesian
- # product spaces, even though they provide a method
- # with that name.
- pairs = zip(B, self.to_vector())
- return sum( ( W(tuple(alpha*b_i for b_i in b))
- for (b,alpha) in pairs ),
- W.zero())
- else:
- # This is just a manual "from_vector()", but of course
- # matrix spaces aren't vector spaces in sage, so they
- # don't have a from_vector() method.
- return W.linear_combination( zip(B, self.to_vector()) )
+ # This is just a manual "from_vector()", but of course
+ # matrix spaces aren't vector spaces in sage, so they
+ # don't have a from_vector() method.
+ return W.linear_combination( zip(B, self.to_vector()) )
TESTS::
- sage: set_random_seed()
sage: J = random_eja()
sage: x,y = J.random_elements(2)
sage: x.operator()(y) == x*y
The explicit form in the spin factor algebra is given by
Alizadeh's Example 11.12::
- sage: set_random_seed()
sage: x = JordanSpinEJA.random_instance().random_element()
sage: x_vec = x.to_vector()
sage: Q = matrix.identity(x.base_ring(), 0)
Test all of the properties from Theorem 11.2 in Alizadeh::
- sage: set_random_seed()
sage: J = random_eja()
sage: x,y = J.random_elements(2)
sage: Lx = x.operator()
sage: J = RealSymmetricEJA(3)
sage: J.one()
- e0 + e2 + e5
+ b0 + b2 + b5
sage: J.one().spectral_decomposition()
- [(1, e0 + e2 + e5)]
+ [(1, b0 + b2 + b5)]
sage: J.zero().spectral_decomposition()
- [(0, e0 + e2 + e5)]
+ [(0, b0 + b2 + b5)]
TESTS::
The spectral decomposition should work in subalgebras, too::
sage: J = RealSymmetricEJA(4)
- sage: (e0, e1, e2, e3, e4, e5, e6, e7, e8, e9) = J.gens()
- sage: A = 2*e5 - 2*e8
+ sage: (b0, b1, b2, b3, b4, b5, b6, b7, b8, b9) = J.gens()
+ sage: A = 2*b5 - 2*b8
sage: (lambda1, c1) = A.spectral_decomposition()[1]
sage: (J0, J5, J1) = J.peirce_decomposition(c1)
sage: (f0, f1, f2) = J1.gens()
sage: f0.spectral_decomposition()
- [(0, f2), (1, f0)]
+ [(0, 1.000000000000000?*c2), (1, 1.000000000000000?*c0)]
"""
A = self.subalgebra_generated_by(orthonormalize=True)
SETUP::
- sage: from mjo.eja.eja_algebra import random_eja
+ sage: from mjo.eja.eja_algebra import (random_eja,
+ ....: HadamardEJA,
+ ....: RealSymmetricEJA)
+
+ EXAMPLES:
+
+ We can create subalgebras of Cartesian product EJAs that are not
+ themselves Cartesian product EJAs (they're just "regular" EJAs)::
+
+ sage: J1 = HadamardEJA(3)
+ sage: J2 = RealSymmetricEJA(2)
+ sage: J = cartesian_product([J1,J2])
+ sage: J.one().subalgebra_generated_by()
+ Euclidean Jordan algebra of dimension 1 over Algebraic Real Field
TESTS:
This subalgebra, being composed of only powers, is associative::
- sage: set_random_seed()
sage: x0 = random_eja().random_element()
- sage: A = x0.subalgebra_generated_by()
+ sage: A = x0.subalgebra_generated_by(orthonormalize=False)
sage: x,y,z = A.random_elements(3)
sage: (x*y)*z == x*(y*z)
True
Squaring in the subalgebra should work the same as in
the superalgebra::
- sage: set_random_seed()
sage: x = random_eja().random_element()
- sage: A = x.subalgebra_generated_by()
+ sage: A = x.subalgebra_generated_by(orthonormalize=False)
sage: A(x^2) == A(x)*A(x)
True
element... unless the original algebra was trivial, in which
case the subalgebra is trivial too::
- sage: set_random_seed()
sage: A = random_eja().zero().subalgebra_generated_by()
sage: (A.is_trivial() and A.dimension() == 0) or A.dimension() == 1
True
"""
powers = tuple( self**k for k in range(self.degree()) )
- A = self.parent().subalgebra(powers, associative=True, **kwargs)
+ A = self.parent().subalgebra(powers,
+ associative=True,
+ check_field=False,
+ check_axioms=False,
+ **kwargs)
A.one.set_cache(A(self.parent().one()))
return A
where there are non-nilpotent elements, or that we get the dumb
solution in the trivial algebra::
- sage: set_random_seed()
sage: J = random_eja()
sage: x = J.random_element()
sage: while x.is_nilpotent() and not J.is_trivial():
The trace of an element is a real number::
- sage: set_random_seed()
sage: J = random_eja()
sage: J.random_element().trace() in RLF
True
The trace is linear::
- sage: set_random_seed()
sage: J = random_eja()
sage: x,y = J.random_elements(2)
sage: alpha = J.base_ring().random_element()
sage: (alpha*x + y).trace() == alpha*x.trace() + y.trace()
True
+ The trace of a square is nonnegative::
+
+ sage: x = random_eja().random_element()
+ sage: (x*x).trace() >= 0
+ True
+
"""
P = self.parent()
r = P.rank()
The trace inner product is commutative, bilinear, and associative::
- sage: set_random_seed()
sage: J = random_eja()
sage: x,y,z = J.random_elements(3)
sage: # commutative
"""
return self.trace_inner_product(self).sqrt()
+
+
+class CartesianProductEJAElement(FiniteDimensionalEJAElement):
+ def det(self):
+ r"""
+ Compute the determinant of this product-element using the
+ determianants of its factors.
+
+ This result Follows from the spectral decomposition of (say)
+ the pair `(x,y)` in terms of the Jordan frame `\left\{ (c_1,
+ 0),(c_2, 0),...,(0,d_1),(0,d_2),... \right\}.
+ """
+ from sage.misc.misc_c import prod
+ return prod( f.det() for f in self.cartesian_factors() )
+
+ def to_matrix(self):
+ # An override is necessary to call our custom _scale().
+ B = self.parent().matrix_basis()
+ W = self.parent().matrix_space()
+
+ # Aaaaand linear combinations don't work in Cartesian
+ # product spaces, even though they provide a method with
+ # that name. This is hidden behind an "if" because the
+ # _scale() function is slow.
+ pairs = zip(B, self.to_vector())
+ return W.sum( _scale(b, alpha) for (b,alpha) in pairs )