Non-converging test case

[jason-neal]

Need to work out better the convergence bound on the true_anomaly computation. One test is failing with K=0.1 and ecc=0.999. (Does not converge before 10000 iterations)

Here is the travis output.
This will be a fail randomly until fixed.

____________________________ test_RV_from_params ______________________________

    @settings(deadline=400)  # double deadline for this test

>   @given(st.floats(min_value=0.01, max_value=1e6, allow_nan=False, allow_infinity=False),

           st.floats(min_value=0.1, max_value=1e5, allow_nan=False, allow_infinity=False),

           st.floats(min_value=0.1, max_value=1e6, allow_nan=False, allow_infinity=False),

           st.floats(min_value=0, max_value=360, allow_nan=False, allow_infinity=False),

           st.floats(min_value=0, max_value=0.999, allow_nan=False, allow_infinity=False))

    def test_RV_from_params(k1, period, tau, omega, ecc):

tests/test_old_code.py:29: 

_ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ 


...

    ta = RV.true_anomaly(ma, ecc)

_ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ 

ma = array([ 0.        ,  0.0315738 ,  0.06314759,  0.09472139,  0.12629518,

        0.15786898,  0.18944277,  0.22101657, ...059494,  6.06216874,  6.09374253,  6.12531633,

        6.15689013,  6.18846392,  6.22003772,  6.25161151,  6.28318531])

ecc = 0.9989999999963652, niterationmax = 10000

    @staticmethod

    def true_anomaly(ma, ecc, niterationmax=10000):

        # type: (Any, float, int) -> Any

       ...
       

        if not isinstance(ma, (int, float)):

            ea = ma

        else:

            ea = np.array([ma, ], dtype=np.float)

    

        if isinstance(ea, list):

            raise TypeError("Unsupported type 'list', input a numpy array or an int/float.")

        if len(ea) == 0:

            raise ValueError("A empty array was given.")

    

        # Initialise at ea0 = ma

        niteration = 0

        ea0 = ma

    

        while np.linalg.norm(ea - ea0, ord=1) > 1e-5 or niteration == 0:

            ea0 = ea

    

            ff = ea - ecc * np.sin(ea) - ma  # Function

            dff = 1 - ecc * np.cos(ea)  # Derivative

    

            # Use Newton method

            ea = ea0 - ff / dff

    

            # Increase iteration number; if above limit, break with exception.

            niteration += 1

            if niteration >= niterationmax:

>               raise RuntimeError('Eccentric anomaly computation '

                                   'not converged.')

E               RuntimeError: Eccentric anomaly computation not converged.

/home/travis/miniconda/envs/test-env/lib/python3.6/site-packages/ObservationTools-0.1-py3.6.egg/utils/rv_utils.py:201: RuntimeError

---------------------------------- Hypothesis ----------------------------------

Falsifying example: test_RV_from_params(k1=0.01, period=12524.503774713052, tau=62500.093750000015, omega=0.0, ecc=0.9989999999963652)

You can add @seed(151146495245698969667814574467189155361) to this test or run pytest with --hypothesis-seed=151146495245698969667814574467189155361 to reproduce this failure.

=============== 1 failed, 134 passed, 3 xfailed in 67.64 seconds ===============