]> gitweb.michael.orlitzky.com - sage.d.git/commitdiff
Rename _rho to _restrict_to_space and make it work on subspaces instead.
authorMichael Orlitzky <michael@orlitzky.com>
Sun, 16 Aug 2015 15:51:26 +0000 (11:51 -0400)
committerMichael Orlitzky <michael@orlitzky.com>
Sun, 16 Aug 2015 15:51:26 +0000 (11:51 -0400)
mjo/cone/cone.py

index 821b95958e81f1e9a93dcf176f774a4c82e9317f..3c582e681a8ea0c6d985cc89d5492a5481578801 100644 (file)
@@ -91,34 +91,41 @@ def _basically_the_same(K1, K2):
 
 
 
-def _rho(K, K2=None):
+def _restrict_to_space(K, W):
     r"""
-    Restrict ``K`` into its own span, or the span of another cone.
+    Restrict this cone a subspace of its ambient space.
 
     INPUT:
 
-    - ``K2`` -- another cone whose lattice has the same rank as this
-                cone.
+    - ``W`` -- The subspace into which this cone will be restricted.
 
     OUTPUT:
 
-    A new cone in a sublattice.
+    A new cone in a sublattice corresponding to ``W``.
 
-    EXAMPLES::
+    EXAMPLES:
+
+    When this cone is solid, restricting it into its own span should do
+    nothing::
 
         sage: K = Cone([(1,)])
-        sage: _rho(K) == K
+        sage: _restrict_to_space(K, K.span()) == K
         True
 
+    A single ray restricted into its own span gives the same output
+    regardless of the ambient space::
+
         sage: K2 = Cone([(1,0)])
-        sage: _rho(K2).rays()
+        sage: K2_S = _restrict_to_space(K2, K2.span()).rays()
+        sage: K2_S
         N(1)
         in 1-d lattice N
         sage: K3 = Cone([(1,0,0)])
-        sage: _rho(K3).rays()
+        sage: K3_S = _restrict_to_space(K3, K3.span()).rays()
+        sage: K3_S
         N(1)
         in 1-d lattice N
-        sage: _rho(K2) == _rho(K3)
+        sage: K2_S == K3_S
         True
 
     TESTS:
@@ -127,8 +134,7 @@ def _rho(K, K2=None):
 
         sage: set_random_seed()
         sage: K = random_cone(max_ambient_dim = 8)
-        sage: K_S = _rho(K)
-        sage: K_S.is_solid()
+        sage: _restrict_to_space(K, K.span()).is_solid()
         True
 
     And the resulting cone should live in a space having the same
@@ -136,22 +142,22 @@ def _rho(K, K2=None):
 
         sage: set_random_seed()
         sage: K = random_cone(max_ambient_dim = 8)
-        sage: K_S = _rho(K, K.dual() )
-        sage: K_S.lattice_dim() == K.dual().dim()
+        sage: K_P = _restrict_to_space(K, K.dual().span())
+        sage: K_P.lattice_dim() == K.dual().dim()
         True
 
     This function should not affect the dimension of a cone::
 
         sage: set_random_seed()
         sage: K = random_cone(max_ambient_dim = 8)
-        sage: K.dim() == _rho(K).dim()
+        sage: K.dim() == _restrict_to_space(K,K.span()).dim()
         True
 
     Nor should it affect the lineality of a cone::
 
         sage: set_random_seed()
         sage: K = random_cone(max_ambient_dim = 8)
-        sage: K.lineality() == _rho(K).lineality()
+        sage: K.lineality() == _restrict_to_space(K, K.span()).lineality()
         True
 
     No matter which space we restrict to, the lineality should not
@@ -159,56 +165,50 @@ def _rho(K, K2=None):
 
         sage: set_random_seed()
         sage: K = random_cone(max_ambient_dim = 8)
-        sage: K.lineality() >= _rho(K).lineality()
+        sage: S = K.span(); P = K.dual().span()
+        sage: K.lineality() >= _restrict_to_space(K,S).lineality()
         True
-        sage: K.lineality() >= _rho(K, K.dual()).lineality()
+        sage: K.lineality() >= _restrict_to_space(K,P).lineality()
         True
 
     If we do this according to our paper, then the result is proper::
 
         sage: set_random_seed()
         sage: K = random_cone(max_ambient_dim = 8)
-        sage: K_S = _rho(K)
-        sage: K_SP = _rho(K_S.dual()).dual()
+        sage: K_S = _restrict_to_space(K, K.span())
+        sage: K_SP = _restrict_to_space(K_S.dual(), K_S.dual().span()).dual()
         sage: K_SP.is_proper()
         True
-        sage: K_SP = _rho(K_S, K_S.dual())
+        sage: K_SP = _restrict_to_space(K_S, K_S.dual().span())
         sage: K_SP.is_proper()
         True
 
     Test the proposition in our paper concerning the duals and
     restrictions. Generate a random cone, then create a subcone of
-    it. The operation of dual-taking should then commute with rho::
+    it. The operation of dual-taking should then commute with
+    _restrict_to_space::
 
         sage: set_random_seed()
         sage: J = random_cone(max_ambient_dim = 8)
         sage: K = Cone(random_sublist(J.rays(), 0.5), lattice=J.lattice())
-        sage: K_W_star = _rho(K, J).dual()
-        sage: K_star_W = _rho(K.dual(), J)
+        sage: K_W_star = _restrict_to_space(K, J.span()).dual()
+        sage: K_star_W = _restrict_to_space(K.dual(), J.span())
         sage: _basically_the_same(K_W_star, K_star_W)
         True
 
     """
-    if K2 is None:
-        K2 = K
-
-    # First we project K onto the span of K2. This will explode if the
-    # rank of ``K2.lattice()`` doesn't match ours.
-    span_K2 = Cone(K2.rays() + (-K2).rays(), lattice=K.lattice())
-    K = K.intersection(span_K2)
-
-    # Cheat a little to get the subspace span(K2). The paper uses the
-    # rays of K2 as a basis, but everything is invariant under linear
-    # isomorphism (i.e. a change of basis), and this is a little
-    # faster.
-    W = span_K2.linear_subspace()
+    # First we want to intersect ``K`` with ``W``. The easiest way to
+    # do this is via cone intersection, so we turn the subspace ``W``
+    # into a cone.
+    W_cone = Cone(W.basis() + [-b for b in W.basis()], lattice=K.lattice())
+    K = K.intersection(W_cone)
 
     # We've already intersected K with the span of K2, so every
     # generator of K should belong to W now.
-    W_rays = [ W.coordinate_vector(r) for r in K.rays() ]
+    K_W_rays = [ W.coordinate_vector(r) for r in K.rays() ]
 
-    L = ToricLattice(K2.dim())
-    return Cone(W_rays, lattice=L)
+    L = ToricLattice(W.dimension())
+    return Cone(K_W_rays, lattice=L)
 
 
 
@@ -635,8 +635,8 @@ def lyapunov_rank(K):
         sage: set_random_seed()
         sage: K = random_cone(max_ambient_dim=8)
         sage: actual = lyapunov_rank(K)
-        sage: K_S = _rho(K)
-        sage: K_SP = _rho(K_S.dual()).dual()
+        sage: K_S = _restrict_to_space(K, K.span())
+        sage: K_SP = _restrict_to_space(K_S.dual(), K_S.dual().span()).dual()
         sage: l = K.lineality()
         sage: c = K.codim()
         sage: expected = lyapunov_rank(K_SP) + K.dim()*(l + c) + c**2
@@ -671,7 +671,7 @@ def lyapunov_rank(K):
 
     if m < n:
         # K is not solid, restrict to its span.
-        K = _rho(K)
+        K = _restrict_to_space(K, K.span())
 
         # Non-solid reduction lemma.
         beta += (n - m)*n
@@ -680,7 +680,7 @@ def lyapunov_rank(K):
         # K is not pointed, restrict to the span of its dual. Uses a
         # proposition from our paper, i.e. this is equivalent to K =
         # _rho(K.dual()).dual().
-        K = _rho(K, K.dual())
+        K = _restrict_to_space(K, K.dual().span())
 
         # Non-pointed reduction lemma.
         beta += l * m