hdf5 cache

[marscher]

======================================================================
FAIL: test_tica_cached_input (chainsaw.tests.test_cache.TestCache)
----------------------------------------------------------------------
Traceback (most recent call last):
  File "/home/travis/miniconda/envs/_test/lib/python2.7/site-packages/chainsaw-0.1a0+3.g3c67aeb-py2.7-linux-x86_64.egg/chainsaw/tests/test_cache.py", line 87, in test_tica_cached_input
    np.testing.assert_allclose(tica_cache_inp.cov, tica_without_cache.cov, atol=1e-10)
  File "/home/travis/miniconda/envs/_test/lib/python2.7/site-packages/numpy/testing/utils.py", line 1395, in assert_allclose
    verbose=verbose, header=header)
  File "/home/travis/miniconda/envs/_test/lib/python2.7/site-packages/numpy/testing/utils.py", line 737, in assert_array_compare
    raise AssertionError(msg)
AssertionError: 
Not equal to tolerance rtol=1e-07, atol=1e-10
(mismatch 22.2222222222%)
 x: array([[ 0.083683, -0.000466,  0.000317],
       [-0.000466,  0.083363, -0.000487],
       [ 0.000317, -0.000487,  0.083347]])
 y: memmap([[ 0.083683, -0.000466,  0.000317],
       [-0.000466,  0.083363, -0.000487],
       [ 0.000317, -0.000487,  0.083347]])

It seems that the output of the Cache does not fully reproduce the input in a numerical precise way.... sigh