TODO

   1 1. Make it work on a cartesian product of cones in the correct order.
   2
   3 2. Make it work on a cartesian product of cones in the wrong order
   4    (apply a perm utation before/after).
   5
   6 3. Make sure we have the dimensions of the PSD cone correct.
   7
   8 4. Come up with a fast heuristic (like making nu huge and taking e1 as
   9    our point) that finds a primal feasible point.
  10
  11 5. Fix the solve failures that we get in the translation tests. For example,
  12
  13   ERROR: test_translation_orthant (test.symmetric_linear_game_test.
  14                                    SymmetricLinearGameTest)
  15   ----------------------------------------------------------------------
  16   Traceback (most recent call last):
  17     File "/home/mjo/src/dunshire/test/symmetric_linear_game_test.py",
  18     line 374, in test_translation_orthant
  19       self.assert_translation_works(L, K, e1, e2)
  20     File "/home/mjo/src/dunshire/test/symmetric_linear_game_test.py",
  21       line 361, in assert_translation_works
  22       value2 = game2.solution().game_value()
  23     File "/home/mjo/src/dunshire/dunshire/games.py", line 458, in solution
  24       raise GameUnsolvableException(self, soln_dict)
  25   dunshire.errors.GameUnsolvableException: Solution failed with result
  26   "unknown."
  27   The linear game (L, K, e1, e2) where
  28     L = [352.0763359 267.0812248 300.8004888 307.8135853]
  29         [429.8303135 324.8322824 361.6866231 372.1748983]
  30         [390.6592961 286.8039007 320.7409227 330.1854235]
  31         [316.0538913 247.7440818 276.9063990 274.9871772],
  32     K = Nonnegative orthant in the real 4-space,
  33     e1 = [7.7040001]
  34          [9.4324457]
  35          [8.3882819]
  36          [6.8908420],
  37     e2 = [8.5054325]
  38          [6.4738132]
  39          [7.2452437]
  40          [7.3307357].
  41   CVXOPT returned:
  42     dual infeasibility: 0.053819211766446585
  43     dual objective: -5.369636805607942
  44     dual slack: 2.105806354638527e-17
  45     gap: 2.6823510532777825e-16
  46     iterations: 11
  47     primal infeasibility: 4.71536776301359e-15
  48     primal objective: -5.3799616179161
  49     primal slack: 1.0328930392495263e-17
  50     relative gap: 4.985818196816016e-17
  51     residual as dual infeasibility certificate: 0.18587493201993227
  52     residual as primal infeasibility certificate: None
  53     s:
  54       [0.0115539]
  55       [0.0000000]
  56       [0.0000000]
  57       [0.1230066]
  58       [0.4837410]
  59       [0.0000000]
  60       [0.0000000]
  61       [0.4044349]
  62     status: unknown
  63     x:
  64       [ 5.3799616]
  65       [ 0.0115539]
  66       [-0.0000000]
  67       [-0.0000000]
  68       [ 0.1230066]
  69     y:
  70       [5.3696368]
  71     z:
  72       [0.0000000]
  73       [0.4176330]
  74       [0.6007564]
  75       [0.0000000]
  76       [0.0000000]
  77       [0.0889310]
  78       [0.0191076]
  79       [0.0000000]
  80
  81 6. Fix the math domain errors that sometimes pop up:
  82
  83   ERROR: test_scaling_icecream (test.symmetric_linear_game_test
  84                                 .SymmetricLinearGameTest)
  85   ----------------------------------------------------------------------
  86   Traceback (most recent call last):
  87     File "/home/mjo/src/dunshire/test/symmetric_linear_game_test.py",
  88     line 336, in test_scaling_icecream
  89       self.assert_scaling_works(L, K, e1, e2)
  90     File "/home/mjo/src/dunshire/test/symmetric_linear_game_test.py",
  91     line 317, in assert_scaling_works
  92       value2 = game2.solution().game_value()
  93     File "/home/mjo/src/dunshire/dunshire/games.py", line 428, in solution
  94       soln_dict = solvers.conelp(c, G, h, C.cvxopt_dims(), A, b)
  95     File "/usr/lib64/python3.4/site-packages/cvxopt/coneprog.py", line 1395,
  96     in conelp
  97       misc.update_scaling(W, lmbda, ds, dz)
  98     File "/usr/lib64/python3.4/site-packages/cvxopt/misc.py", line 510,
  99     in update_scaling
 100       ln = jnrm2(lmbda, n = m, offset = ind)
 101     File "/usr/lib64/python3.4/site-packages/cvxopt/misc.py", line 856, in jnrm2
 102       return math.sqrt(x[offset] - a) * math.sqrt(x[offset] + a)
 103   ValueError: math domain error