Python utils.pandas_sort函数代码示例

    def test_isolated_continuous_random_walks(self):
        # Two 2D random walks
        np.random.seed(0)
        N = 30
        Y = 250
        M = 20 # margin, because negative values raise OutOfHash
        a = DataFrame({'x': M + random_walk(N), 'y': M + random_walk(N), 'frame': np.arange(N)})
        b = DataFrame({'x': M + random_walk(N - 1), 'y': M + Y + random_walk(N - 1), 'frame': np.arange(1, N)})
        f = pd.concat([a, b])
        expected = f.copy().reset_index(drop=True)
        expected['particle'] = np.concatenate([np.zeros(N), np.ones(N - 1)])
        pandas_sort(expected, ['particle', 'frame'], inplace=True)
        actual = self.link_df(f, 5)
        assert_frame_equal(actual, expected)
        actual_iter = self.link_df_iter(f, 5, hash_size=(2*M, Y + 2*M))
        assert_frame_equal(actual_iter, expected)

        # Many 2D random walks
        np.random.seed(0)
        initial_positions = [(100, 100), (200, 100), (100, 200), (200, 200)]
        import itertools
        c = itertools.count()
        def walk(x, y): 
            i = next(c)
            return DataFrame({'x': x + random_walk(N - i), 
                              'y': y + random_walk(N - i),
                             'frame': np.arange(i, N)})
        f = pd.concat([walk(*pos) for pos in initial_positions])
        expected = f.copy().reset_index(drop=True)
        expected['particle'] = np.concatenate([i*np.ones(N - i) for i in range(len(initial_positions))])
        pandas_sort(expected, ['particle', 'frame'], inplace=True)
        actual = self.link_df(f, 5)
        assert_frame_equal(actual, expected)
        actual_iter = self.link_df_iter(f, 5, hash_size=(200 + M, 200 + M))
        assert_frame_equal(actual_iter, expected)

    def test_memory_on_one_gap(self):
        N = 5
        Y = 2
        # Begin second feature one frame later than the first, so the particle labeling (0, 1) is
        # established and not arbitrary.
        a = DataFrame({'x': np.arange(N), 'y': np.ones(N), 'frame': np.arange(N)})
        b = DataFrame({'x': np.arange(1, N), 'y': Y + np.ones(N - 1), 'frame': np.arange(1, N)})
        a = a.drop(3).reset_index(drop=True)
        f = pd.concat([a, b])
        expected = f.copy().reset_index(drop=True)
        expected['particle'] = np.concatenate([np.array([0, 0, 0, 0]), np.ones(N - 1)])
        pandas_sort(expected, ['particle', 'frame'], inplace=True)
        expected.reset_index(drop=True, inplace=True)
        actual = self.link_df(f, 5, memory=1)
        assert_traj_equal(actual, expected)

        # Sort rows by frame (normal use)
        actual = self.link_df(pandas_sort(f, 'frame'), 5, memory=1)
        assert_traj_equal(actual, expected)

        # Shuffle rows (crazy!)
        np.random.seed(0)
        f1 = f.reset_index(drop=True)
        f1.reindex(np.random.permutation(f1.index))
        actual = self.link_df(f1, 5, memory=1)
        assert_traj_equal(actual, expected)

    def test_two_isolated_steppers(self):
        N = 5
        Y = 25
        # Begin second feature one frame later than the first, so the particle labeling (0, 1) is
        # established and not arbitrary.
        a = DataFrame({'x': np.arange(N), 'y': np.ones(N), 'frame': np.arange(N)})
        b = DataFrame({'x': np.arange(1, N), 'y': Y + np.ones(N - 1), 'frame': np.arange(1, N)})
        f = pd.concat([a, b])
        expected = f.copy().reset_index(drop=True)
        expected['particle'] = np.concatenate([np.zeros(N), np.ones(N - 1)])
        pandas_sort(expected, ['particle', 'frame'], inplace=True)
        actual = self.link_df(f, 5)
        assert_frame_equal(actual, expected)
        actual_iter = self.link_df_iter(f, 5, hash_size=(50, 50))
        assert_frame_equal(actual_iter, expected)

        # Sort rows by frame (normal use)
        actual = self.link_df(pandas_sort(f, 'frame'), 5)
        assert_frame_equal(actual, expected)
        actual_iter = self.link_df_iter(pandas_sort(f, 'frame'), 5, hash_size=(50, 50))
        assert_frame_equal(actual_iter, expected)

        # Shuffle rows (crazy!)
        np.random.seed(0)
        f1 = f.reset_index(drop=True)
        f1.reindex(np.random.permutation(f1.index))
        actual = self.link_df(f1, 5)
        assert_frame_equal(actual, expected)
        actual_iter = self.link_df_iter(f1, 5, hash_size=(50, 50))
        assert_frame_equal(actual_iter, expected)

def assert_traj_equal(actual, expected, pos_atol=1):
    assert_equal(len(actual), len(expected))
    actual = pandas_sort(actual, 'frame').reset_index(drop=True)
    expected = pandas_sort(expected, 'frame').reset_index(drop=True)
    actual_order = []
    for frame_no in expected['frame'].unique():
        actual_f = actual[actual['frame'] == frame_no]
        expected_f = expected[expected['frame'] == frame_no]
        assert_equal(len(actual_f), len(expected_f),
                     err_msg='Actual and expected numbers of features '
                             'differ in frame %i' % frame_no)
        tree = cKDTree(actual_f[['y', 'x']].values)
        devs, argsort = tree.query(expected_f[['y', 'x']].values)
        assert_allclose(devs, 0., atol=pos_atol)
        actual_order.extend(actual_f.index[argsort].tolist())

    actual = actual.loc[actual_order].reset_index(drop=True, inplace=False)
    for p_actual in actual.particle.unique():
        actual_ind = actual.index[actual['particle'] == p_actual]
        p_expected = expected.loc[actual_ind[0], 'particle']
        expected_ind = expected.index[expected['particle'] == p_expected]
        assert_array_equal(actual_ind, expected_ind,
                           err_msg='Actual and expected linking results '
                           'differ for actual particle %i/expected particle %i'
                           '' % (p_actual, p_expected))

    def test_topn(self):
        self.check_skip()
        L = 21
        dims = (L, L + 2)  # avoid square images in tests
        cols = ['y', 'x']
        PRECISION = 0.1

        # top 2
        pos1 = np.array([7, 7])
        pos2 = np.array([14, 14])
        pos3 = np.array([7, 14])
        image = np.ones(dims, dtype='uint8')
        draw_point(image, pos1, 100)
        draw_point(image, pos2, 90)
        draw_point(image, pos3, 80)
        actual = tp.locate(image, 5, 1, topn=2, preprocess=False,
                           engine=self.engine)[cols]
        actual = pandas_sort(actual, ['x', 'y'])  # sort for reliable comparison
        expected = pandas_sort(DataFrame([pos1, pos2], columns=cols), ['x', 'y'])
        assert_allclose(actual, expected, atol=PRECISION)

        # top 1
        actual = tp.locate(image, 5, 1, topn=1, preprocess=False,
                           engine=self.engine)[cols]
        actual = pandas_sort(actual, ['x', 'y'])  # sort for reliable comparison
        expected = pandas_sort(DataFrame([pos1], columns=cols), ['x', 'y'])
        assert_allclose(actual, expected, atol=PRECISION)

    def test_two_isolated_steppers_one_gapped(self):
        N = 5
        Y = 25
        # Begin second feature one frame later than the first, 
        # so the particle labeling (0, 1) is established and not arbitrary.
        a = DataFrame({'x': np.arange(N), 'y': np.ones(N), 
                      'frame': np.arange(N)})
        a = a.drop(3).reset_index(drop=True)
        b = DataFrame({'x': np.arange(1, N), 'y': Y + np.ones(N - 1), 
                      'frame': np.arange(1, N)})
        f = pd.concat([a, b])
        expected = f.copy()
        expected['particle'] = np.concatenate([np.array([0, 0, 0, 2]), np.ones(N - 1)])
        pandas_sort(expected, ['particle', 'frame'], inplace=True)
        expected.reset_index(drop=True, inplace=True)
        actual = self.link_df(f, 5)
        assert_traj_equal(actual, expected)
        # link_df_iter() tests not performed, because hash_size is
        # not knowable from the first frame alone.

        # Sort rows by frame (normal use)
        actual = self.link_df(pandas_sort(f, 'frame'), 5)
        assert_traj_equal(actual, expected)

        # Shuffle rows (crazy!)
        np.random.seed(0)
        f1 = f.reset_index(drop=True)
        f1.reindex(np.random.permutation(f1.index))
        actual = self.link_df(f1, 5)
        assert_traj_equal(actual, expected)

def compare(shape, count, radius, noise_level, engine):
    radius = tp.utils.validate_tuple(radius, len(shape))
    # tp.locate ignores a margin of size radius, take 1 px more to be safe
    margin = tuple([r + 1 for r in radius])
    diameter = tuple([(r * 2) + 1 for r in radius])
    draw_range = tuple([d * 3 for d in diameter])
    cols = ['x', 'y', 'z'][:len(shape)][::-1]
    pos = gen_nonoverlapping_locations(shape, count, draw_range, margin)
    image = draw_spots(shape, pos, draw_range, noise_level)
    f = tp.locate(image, diameter, engine=engine)
    actual = pandas_sort(f[cols], cols)
    expected = pandas_sort(DataFrame(pos, columns=cols), cols)
    return actual, expected

    def test_penalty(self):
        """A test case of two particles, spaced 8 and each moving by 8 down
        and 7 to the right. We have two likely linking results:

        1. two links, total squared displacement = 2*(8**2 + 7**2) = 226
        2. one link, total squared displacement = (8**2 + 1**2) + sr**2

        Case 2 gets a penalty for not linking, which equals the search range
        squared. We vary this in this test.

        With a penalty of 13, case 2 has a total cost of 234 and we expect case
        1. as the result.

        With a penalty of 12, case 2. will have a total cost of 209 and we
        expect case 2. as the result.
        """
        f = pd.DataFrame({'x': [0, 8, 7, 8 + 7],
                          'y': [0, 0, 8, 8],
                          'frame': [0, 0, 1, 1]})
        case1 = f.copy()
        case1['particle'] = np.array([0, 1, 0, 1])
        case2 = f.copy()
        case2['particle'] = np.array([0, 1, 1, 2])

        actual = self.link_df(f, 13)
        pandas_sort(case1, ['x'], inplace=True)
        pandas_sort(actual, ['x'], inplace=True)
        assert_array_equal(actual['particle'].values.astype(np.int),
                           case1['particle'].values.astype(np.int))

        actual = self.link_df(f, 12)
        pandas_sort(case2, ['x'], inplace=True)
        pandas_sort(actual, ['x'], inplace=True)
        assert_array_equal(actual['particle'].values.astype(np.int),
                           case2['particle'].values.astype(np.int))

    def test_nearby_continuous_random_walks(self):
        # Two 2D random walks
        np.random.seed(0)
        N = 30
        Y = 250
        M = 20 # margin, because negative values raise OutOfHash
        a = DataFrame({'x': M + random_walk(N),
                       'y': M + random_walk(N),
                       'frame': np.arange(N)})
        b = DataFrame({'x': M + random_walk(N - 1),
                       'y': M + Y + random_walk(N - 1),
                       'frame': np.arange(1, N)})
        f = pd.concat([a, b])
        expected = f.copy().reset_index(drop=True)
        expected['particle'] = np.concatenate([np.zeros(N), np.ones(N - 1)])
        pandas_sort(expected, ['particle', 'frame'], inplace=True)
        actual = self.link_df(f, 5)
        assert_frame_equal(actual, expected)
        actual = self.link_df_iter(f, 5, hash_size=(2*M, 2*M + Y))
        assert_frame_equal(actual, expected)

        # Several 2D random walks
        np.random.seed(0)
        initial_positions = [(10, 11), (10, 18), (14, 15), (20, 21), (13, 13),
                             (10, 10), (17, 19)]
        import itertools
        c = itertools.count()
        def walk(x, y):
            i = next(c)
            return DataFrame({'x': x + random_walk(N - i),
                              'y': y + random_walk(N - i),
                              'frame': np.arange(i, N)})
        f = pd.concat([walk(*pos) for pos in initial_positions])
        expected = f.copy().reset_index(drop=True)
        expected['particle'] = np.concatenate([i*np.ones(N - i) for i in range(len(initial_positions))])
        pandas_sort(expected, ['particle', 'frame'], inplace=True)
        actual = self.link_df(f, 5)
        assert_frame_equal(actual, expected)
        actual = self.link_df_iter(f, 5, hash_size=(2*M, 2*M))
        assert_frame_equal(actual, expected)

        # Shuffle rows (crazy!)
        np.random.seed(0)
        f1 = f.reset_index(drop=True)
        f1.reindex(np.random.permutation(f1.index))
        actual = self.link_df(f1, 5)
        assert_frame_equal(actual, expected)
        actual = self.link_df_iter(f1, 5, hash_size=(2*M, 2*M))
        assert_frame_equal(actual, expected)

    def link(self, f, search_range, *args, **kwargs):
        if 'pos_columns' in kwargs:
            raise nose.SkipTest('Skipping find_link tests with custom pos_columns.')
        # the minimal spacing between features in f is assumed to be 1.

        # from scipy.spatial import cKDTree
        # mindist = 1e7
        # for _, _f in f.groupby('frame'):
        #     dists, _ = cKDTree(_f[['y', 'x']].values).query(_f[['y', 'x']].values, k=2)
        #     mindist = min(mindist, dists[:, 1].min())
        # print("Minimal dist is {0:.3f}".format(mindist))

        kwargs = dict(self.linker_opts, **kwargs)
        size = 3
        separation = kwargs['separation']
        f = f.copy()
        f[['y', 'x']] *= separation
        topleft = (f[['y', 'x']].min().values - 4 * separation).astype(np.int)
        f[['y', 'x']] -= topleft
        shape = (f[['y', 'x']].max().values + 4 * separation).astype(np.int)
        reader = CoordinateReader(f, shape, size)
        if kwargs.get('adaptive_stop', None) is not None:
            kwargs['adaptive_stop'] *= separation
        result = find_link(reader,
                           search_range=search_range*separation,
                           *args, **kwargs)
        result = pandas_sort(result, ['particle', 'frame']).reset_index(drop=True)
        result[['y', 'x']] += topleft
        result[['y', 'x']] /= separation
        return result

 def link_df_iter(self, *args, **kwargs):
     kwargs.update(self.linker_opts)
     kwargs['diagnostics'] = self.do_diagnostics
     args = list(args)
     features = args.pop(0)
     res = pd.concat(tp.link_df_iter(
         (df for fr, df in features.groupby('frame')), *args, **kwargs))
     return pandas_sort(res, ['particle', 'frame']).reset_index(drop=True)

def conformity(df):
    """ Organize toy data to look like real data. Be strict about dtypes:
    particle is a float and frame is an integer."""
    df['frame'] = df['frame'].astype(np.int)
    df['particle'] = df['particle'].astype(np.float)
    df['x'] = df['x'].astype(np.float)
    df['y'] = df['y'].astype(np.float)
    df.set_index('frame', drop=False, inplace=True)
    return pandas_sort(df, by=['frame', 'particle'])

    def link(self, f, search_range, *args, **kwargs):

        def df_iter(f, first_frame, last_frame):
            """ link_df_iter requires a generator of dataframes """
            for t in range(first_frame, last_frame + 1):
                yield f[f['frame'] == t]

        res_iter = link_df_iter(df_iter(f, 0, int(f['frame'].max())),
                                search_range, *args, **kwargs)
        res = pd.concat(res_iter)
        return pandas_sort(res, ['particle', 'frame']).reset_index(drop=True)

    def test_two_nearby_steppers(self):
        N = 5
        Y = 2
        # Begin second feature one frame later than the first, so the particle labeling (0, 1) is
        # established and not arbitrary.
        a = DataFrame({'x': np.arange(N), 'y': np.ones(N), 'frame': np.arange(N)})
        b = DataFrame({'x': np.arange(1, N), 'y': Y + np.ones(N - 1), 'frame': np.arange(1, N)})
        f = pd.concat([a, b])
        expected = f.copy().reset_index(drop=True)
        expected['particle'] = np.concatenate([np.zeros(N), np.ones(N - 1)])
        pandas_sort(expected, ['particle', 'frame'], inplace=True)
        actual = self.link_df(f, 5)
        assert_frame_equal(actual, expected)
        actual_iter = self.link_df_iter(f, 5, hash_size=(50, 50))
        assert_frame_equal(actual_iter, expected)

        # Sort rows by frame (normal use)
        actual = self.link_df(pandas_sort(f, 'frame'), 5)
        assert_frame_equal(actual, expected)
        actual_iter = self.link_df_iter(pandas_sort(f, 'frame'), 5, hash_size=(50, 50))
        assert_frame_equal(actual_iter, expected)

        # Shuffle rows (crazy!)
        np.random.seed(0)
        f1 = f.reset_index(drop=True)
        f1.reindex(np.random.permutation(f1.index))
        actual = self.link_df(f1, 5)
        assert_frame_equal(actual, expected)
        actual_iter = self.link_df_iter(f1, 5, hash_size=(50, 50))
        assert_frame_equal(actual_iter, expected)

        if self.do_diagnostics:
            assert 'diag_subnet' in self.diag.columns
            assert 'diag_subnet_size' in self.diag.columns
            # Except for frame in which they appear, all particles should have
            # been labeled with a search_range
            assert not any(self.diag['diag_search_range'][
                               actual_iter.frame > 1].isnull())
            # The number of loop iterations is reported by the numba linker only
            if self.linker_opts['link_strategy'] == 'numba':
                assert 'diag_subnet_iterations' in self.diag.columns

    def setUp(self):
        N = 10
        Y = 1
        a = DataFrame({'x': np.zeros(N), 'y': np.zeros(N),
                       'frame': np.arange(N), 'particle': np.zeros(N)})
        b = DataFrame({'x': np.zeros(N - 1), 'y': Y + np.zeros(N - 1),
                       'frame': np.arange(1, N), 'particle': np.ones(N - 1)})
        self.dead_still = conformity(pd.concat([a, b]))
        pandas_sort(self.dead_still, ['frame', 'particle'], inplace=True)

        P = 1000 # particles
        A = 0.00001 # step amplitude
        np.random.seed(0)
        particles = [DataFrame({'x': A*random_walk(N), 'y': A*random_walk(N),
                                'frame': np.arange(N), 'particle': i})
                     for i in range(P)]
        self.many_walks = conformity(pd.concat(particles))

        a = DataFrame({'x': np.arange(N), 'y': np.zeros(N),
                       'frame': np.arange(N), 'particle': np.zeros(N)})
        b = DataFrame({'x': np.arange(1, N), 'y': Y + np.zeros(N - 1),
                       'frame': np.arange(1, N), 'particle': np.ones(N - 1)})
        self.steppers = conformity(pd.concat([a, b]))