def _get_data():
    # Read raw data
    raw = Raw(raw_fname)
    raw.info['bads'] = ['MEG 2443', 'EEG 053']  # 2 bads channels

    # Set picks
    picks = mne.fiff.pick_types(raw.info, meg=True, eeg=False, eog=False,
                                stim=False, exclude='bads')

    # Read several epochs
    event_id, tmin, tmax = 1, -0.2, 0.5
    events = mne.read_events(event_fname)[0:100]
    epochs = mne.Epochs(raw, events, event_id, tmin, tmax, proj=True,
                        picks=picks, baseline=(None, 0), preload=True,
                        reject=dict(grad=4000e-13, mag=4e-12))

    # Create an epochs object with one epoch and one channel of artificial data
    event_id, tmin, tmax = 1, 0.0, 1.0
    epochs_sin = mne.Epochs(raw, events[0:5], event_id, tmin, tmax, proj=True,
                            picks=[0], baseline=(None, 0), preload=True,
                            reject=dict(grad=4000e-13))
    freq = 10
    epochs_sin._data = np.sin(2 * np.pi * freq
                              * epochs_sin.times)[None, None, :]
    return epochs, epochs_sin

def test_raw_index_as_time():
    """ Test index as time conversion"""
    raw = Raw(fif_fname, preload=True)
    t0 = raw.index_as_time([0], True)[0]
    t1 = raw.index_as_time([100], False)[0]
    t2 = raw.index_as_time([100], True)[0]
    assert_true((t2 - t1) == t0)

def test_cov_estimation_on_raw_segment():
    """Test estimation from raw on continuous recordings (typically empty room)
    """
    raw = Raw(raw_fname, preload=False)
    cov = compute_raw_data_covariance(raw)
    cov_mne = read_cov(erm_cov_fname)
    assert_true(cov_mne.ch_names == cov.ch_names)
    assert_true(linalg.norm(cov.data - cov_mne.data, ord='fro')
                / linalg.norm(cov.data, ord='fro') < 1e-4)

    # test IO when computation done in Python
    cov.save(op.join(tempdir, 'test-cov.fif'))  # test saving
    cov_read = read_cov(op.join(tempdir, 'test-cov.fif'))
    assert_true(cov_read.ch_names == cov.ch_names)
    assert_true(cov_read.nfree == cov.nfree)
    assert_array_almost_equal(cov.data, cov_read.data)

    # test with a subset of channels
    picks = pick_channels(raw.ch_names, include=raw.ch_names[:5])
    cov = compute_raw_data_covariance(raw, picks=picks)
    assert_true(cov_mne.ch_names[:5] == cov.ch_names)
    assert_true(linalg.norm(cov.data - cov_mne.data[picks][:, picks],
                ord='fro') / linalg.norm(cov.data, ord='fro') < 1e-4)
    # make sure we get a warning with too short a segment
    raw_2 = raw.crop(0, 1)
    with warnings.catch_warnings(record=True) as w:
        cov = compute_raw_data_covariance(raw_2)
        assert_true(len(w) == 1)

def test_io_complex():
    """Test IO with complex data types
    """
    dtypes = [np.complex64, np.complex128]

    raw = Raw(fif_fname, preload=True)
    picks = np.arange(5)
    start, stop = raw.time_as_index([0, 5])

    data_orig, _ = raw[picks, start:stop]

    for di, dtype in enumerate(dtypes):
        imag_rand = np.array(1j * np.random.randn(data_orig.shape[0],
                             data_orig.shape[1]), dtype)

        raw_cp = raw.copy()
        raw_cp._data = np.array(raw_cp._data, dtype)
        raw_cp._data[picks, start:stop] += imag_rand
        # this should throw an error because it's complex
        with warnings.catch_warnings(record=True) as w:
            raw_cp.save(op.join(tempdir, 'raw.fif'), picks, tmin=0, tmax=5)
            # warning only gets thrown on first instance
            assert_equal(len(w), 1 if di == 0 else 0)

        raw2 = Raw(op.join(tempdir, 'raw.fif'))
        raw2_data, _ = raw2[picks, :]
        n_samp = raw2_data.shape[1]
        assert_array_almost_equal(raw2_data[:, :n_samp],
                                  raw_cp._data[picks, :n_samp])
        # with preloading
        raw2 = Raw(op.join(tempdir, 'raw.fif'), preload=True)
        raw2_data, _ = raw2[picks, :]
        n_samp = raw2_data.shape[1]
        assert_array_almost_equal(raw2_data[:, :n_samp],
                                  raw_cp._data[picks, :n_samp])

def test_preload_modify():
    """ Test preloading and modifying data
    """
    for preload in [False, True, 'memmap.dat']:
        raw = Raw(fif_fname, preload=preload)

        nsamp = raw.last_samp - raw.first_samp + 1
        picks = pick_types(raw.info, meg='grad', exclude='bads')

        data = np.random.randn(len(picks), nsamp / 2)

        try:
            raw[picks, :nsamp / 2] = data
        except RuntimeError as err:
            if not preload:
                continue
            else:
                raise err

        tmp_fname = op.join(tempdir, 'raw.fif')
        raw.save(tmp_fname)

        raw_new = Raw(tmp_fname)
        data_new, _ = raw_new[picks, :nsamp / 2]

        assert_array_almost_equal(data, data_new)

def test_preload_modify():
    """ Test preloading and modifying data
    """
    for preload in [False, True, "memmap.dat"]:
        raw = Raw(fif_fname, preload=preload)

        nsamp = raw.last_samp - raw.first_samp + 1
        picks = pick_types(raw.info, meg="grad", exclude="bads")

        data = np.random.randn(len(picks), nsamp / 2)

        try:
            raw[picks, : nsamp / 2] = data
        except RuntimeError as err:
            if not preload:
                continue
            else:
                raise err

        tmp_fname = op.join(tempdir, "raw.fif")
        raw.save(tmp_fname, overwrite=True)

        raw_new = Raw(tmp_fname)
        data_new, _ = raw_new[picks, : nsamp / 2]

        assert_allclose(data, data_new)

def test_copy_append():
    """Test raw copying and appending combinations
    """
    raw = Raw(fif_fname, preload=True).copy()
    raw_full = Raw(fif_fname)
    raw_full.append(raw)
    data = raw_full[:, :][0]
    assert_true(data.shape[1] == 2 * raw._data.shape[1])

def test_drop_channels_mixin():
    """Test channels-dropping functionality
    """
    raw = Raw(fif_fname, preload=True)
    drop_ch = raw.ch_names[:3]
    ch_names = raw.ch_names[3:]
    raw.drop_channels(drop_ch)
    assert_equal(ch_names, raw.ch_names)
    assert_equal(len(ch_names), raw._data.shape[0])

def test_as_data_frame():
    """Test Pandas exporter"""
    raw = Raw(fif_fname, preload=True)
    df = raw.as_data_frame()
    assert_true((df.columns == raw.ch_names).all())
    df = raw.as_data_frame(use_time_index=False)
    assert_true('time' in df.columns)
    assert_array_equal(df.values[:, 1], raw._data[0] * 1e13)
    assert_array_equal(df.values[:, 3], raw._data[2] * 1e15)

def test_hilbert():
    """ Test computation of analytic signal using hilbert """
    raw = Raw(fif_fname, preload=True)
    picks_meg = pick_types(raw.info, meg=True, exclude='bads')
    picks = picks_meg[:4]

    raw2 = raw.copy()
    raw.apply_hilbert(picks)
    raw2.apply_hilbert(picks, envelope=True, n_jobs=2)

    env = np.abs(raw._data[picks, :])
    assert_array_almost_equal(env, raw2._data[picks, :])

def test_hilbert():
    """ Test computation of analytic signal using hilbert """
    raw = Raw(fif_fname, preload=True)
    picks_meg = pick_types(raw.info, meg=True, exclude="bads")
    picks = picks_meg[:4]

    raw2 = raw.copy()
    raw.apply_hilbert(picks)
    raw2.apply_hilbert(picks, envelope=True, n_jobs=2)

    env = np.abs(raw._data[picks, :])
    assert_allclose(env, raw2._data[picks, :], rtol=1e-2, atol=1e-13)

def test_equalize_channels():
    """Test equalization of channels
    """
    raw1 = Raw(fif_fname)

    raw2 = raw1.copy()
    ch_names = raw1.ch_names[2:]
    raw1.drop_channels(raw1.ch_names[:1])
    raw2.drop_channels(raw2.ch_names[1:2])
    my_comparison = [raw1, raw2]
    equalize_channels(my_comparison)
    for e in my_comparison:
        assert_equal(ch_names, e.ch_names)

def test_raw_to_nitime():
    """ Test nitime export """
    raw = Raw(fif_fname, preload=True)
    picks_meg = pick_types(raw.info, meg=True, exclude='bads')
    picks = picks_meg[:4]
    raw_ts = raw.to_nitime(picks=picks)
    assert_true(raw_ts.data.shape[0] == len(picks))

    raw = Raw(fif_fname, preload=False)
    picks_meg = pick_types(raw.info, meg=True, exclude='bads')
    picks = picks_meg[:4]
    raw_ts = raw.to_nitime(picks=picks)
    assert_true(raw_ts.data.shape[0] == len(picks))

    raw = Raw(fif_fname, preload=True)
    picks_meg = pick_types(raw.info, meg=True, exclude='bads')
    picks = picks_meg[:4]
    raw_ts = raw.to_nitime(picks=picks, copy=False)
    assert_true(raw_ts.data.shape[0] == len(picks))

    raw = Raw(fif_fname, preload=False)
    picks_meg = pick_types(raw.info, meg=True, exclude='bads')
    picks = picks_meg[:4]
    raw_ts = raw.to_nitime(picks=picks, copy=False)
    assert_true(raw_ts.data.shape[0] == len(picks))

def test_raw_copy():
    """ Test Raw copy"""
    raw = Raw(fif_fname, preload=True)
    data, _ = raw[:, :]
    copied = raw.copy()
    copied_data, _ = copied[:, :]
    assert_array_equal(data, copied_data)
    assert_equal(sorted(raw.__dict__.keys()), sorted(copied.__dict__.keys()))

    raw = Raw(fif_fname, preload=False)
    data, _ = raw[:, :]
    copied = raw.copy()
    copied_data, _ = copied[:, :]
    assert_array_equal(data, copied_data)
    assert_equal(sorted(raw.__dict__.keys()), sorted(copied.__dict__.keys()))

def test_compute_proj_raw():
    """Test SSP computation on raw"""
    # Test that the raw projectors work
    raw_time = 2.5  # Do shorter amount for speed
    raw = Raw(raw_fname, preload=True).crop(0, raw_time, False)
    for ii in (0.25, 0.5, 1, 2):
        with warnings.catch_warnings(True) as w:
            projs = compute_proj_raw(raw, duration=ii - 0.1, stop=raw_time,
                                     n_grad=1, n_mag=1, n_eeg=0)
            assert_true(len(w) == 1)

        # test that you can compute the projection matrix
        projs = activate_proj(projs)
        proj, nproj, U = make_projector(projs, raw.ch_names, bads=[])

        assert_true(nproj == 2)
        assert_true(U.shape[1] == 2)

        # test that you can save them
        raw.info['projs'] += projs
        raw.save(op.join(tempdir, 'foo_%d_raw.fif' % ii))

    # Test that purely continuous (no duration) raw projection works
    with warnings.catch_warnings(True) as w:
        projs = compute_proj_raw(raw, duration=None, stop=raw_time,
                                 n_grad=1, n_mag=1, n_eeg=0)
        assert_true(len(w) == 1)

    # test that you can compute the projection matrix
    projs = activate_proj(projs)
    proj, nproj, U = make_projector(projs, raw.ch_names, bads=[])

    assert_true(nproj == 2)
    assert_true(U.shape[1] == 2)

    # test that you can save them
    raw.info['projs'] += projs
    raw.save(op.join(tempdir, 'foo_rawproj_continuous_raw.fif'))

    # test resampled-data projector, upsampling instead of downsampling
    # here to save an extra filtering (raw would have to be LP'ed to be equiv)
    raw_resamp = cp.deepcopy(raw)
    raw_resamp.resample(raw.info['sfreq'] * 2, n_jobs=2)
    projs = compute_proj_raw(raw_resamp, duration=None, stop=raw_time,
                             n_grad=1, n_mag=1, n_eeg=0)
    projs = activate_proj(projs)
    proj_new, _, _ = make_projector(projs, raw.ch_names, bads=[])
    assert_array_almost_equal(proj_new, proj, 4)

def test_output_formats():
    """Test saving and loading raw data using multiple formats
    """
    formats = ['short', 'int', 'single', 'double']
    tols = [1e-4, 1e-7, 1e-7, 1e-15]

    # let's fake a raw file with different formats
    raw = Raw(fif_fname, preload=True)
    raw.crop(0, 1, copy=False)

    temp_file = op.join(tempdir, 'raw.fif')
    for ii, (format, tol) in enumerate(zip(formats, tols)):
        # Let's test the overwriting error throwing while we're at it
        if ii > 0:
            assert_raises(IOError, raw.save, temp_file, format=format)
        raw.save(temp_file, format=format, overwrite=True)
        raw2 = Raw(temp_file)
        raw2_data = raw2[:, :][0]
        assert_allclose(raw2_data, raw._data, rtol=tol, atol=1e-25)
        assert_true(raw2.orig_format == format)

def test_compensation():
    """Test compensation
    """
    raw = Raw(ctf_comp_fname, compensation=None)
    comp1 = make_compensator(raw.info, 3, 1, exclude_comp_chs=False)
    assert_true(comp1.shape == (340, 340))
    comp2 = make_compensator(raw.info, 3, 1, exclude_comp_chs=True)
    assert_true(comp2.shape == (311, 340))

    # make sure that changing the comp doesn't modify the original data
    raw2 = Raw(ctf_comp_fname, compensation=2)
    assert_true(get_current_comp(raw2.info) == 2)
    fname = op.join(tempdir, 'ctf-raw.fif')
    raw2.save(fname)
    raw2 = Raw(fname, compensation=None)
    data, _ = raw[:, :]
    data2, _ = raw2[:, :]
    assert_allclose(data, data2, rtol=1e-9, atol=1e-20)
    for ch1, ch2 in zip(raw.info['chs'], raw2.info['chs']):
        assert_true(ch1['coil_type'] == ch2['coil_type'])

def test_crop():
    """Test cropping raw files
    """
    # split a concatenated file to test a difficult case
    raw = Raw([fif_fname, fif_fname], preload=True)
    split_size = 10.  # in seconds
    sfreq = raw.info['sfreq']
    nsamp = (raw.last_samp - raw.first_samp + 1)

    # do an annoying case (off-by-one splitting)
    tmins = np.r_[1., np.round(np.arange(0., nsamp - 1, split_size * sfreq))]
    tmins = np.sort(tmins)
    tmaxs = np.concatenate((tmins[1:] - 1, [nsamp - 1]))
    tmaxs /= sfreq
    tmins /= sfreq
    raws = [None] * len(tmins)
    for ri, (tmin, tmax) in enumerate(zip(tmins, tmaxs)):
        raws[ri] = raw.crop(tmin, tmax, True)
    all_raw_2 = concatenate_raws(raws, preload=True)
    assert_true(raw.first_samp == all_raw_2.first_samp)
    assert_true(raw.last_samp == all_raw_2.last_samp)
    assert_array_equal(raw[:, :][0], all_raw_2[:, :][0])

    tmins = np.round(np.arange(0., nsamp - 1, split_size * sfreq))
    tmaxs = np.concatenate((tmins[1:] - 1, [nsamp - 1]))
    tmaxs /= sfreq
    tmins /= sfreq

    # going in revere order so the last fname is the first file (need it later)
    raws = [None] * len(tmins)
    for ri, (tmin, tmax) in enumerate(zip(tmins, tmaxs)):
        raws[ri] = raw.copy()
        raws[ri].crop(tmin, tmax, False)
    # test concatenation of split file
    all_raw_1 = concatenate_raws(raws, preload=True)

    all_raw_2 = raw.crop(0, None, True)
    for ar in [all_raw_1, all_raw_2]:
        assert_true(raw.first_samp == ar.first_samp)
        assert_true(raw.last_samp == ar.last_samp)
        assert_array_equal(raw[:, :][0], ar[:, :][0])

def test_compensation_raw():
    raw1 = Raw(ctf_comp_fname, compensation=None)
    assert_true(raw1.comp is None)
    data1, times1 = raw1[:, :]
    raw2 = Raw(ctf_comp_fname, compensation=3)
    data2, times2 = raw2[:, :]
    assert_true(raw2.comp is None)  # unchanged (data come with grade 3)
    assert_array_equal(times1, times2)
    assert_array_equal(data1, data2)
    raw3 = Raw(ctf_comp_fname, compensation=1)
    data3, times3 = raw3[:, :]
    assert_true(raw3.comp is not None)
    assert_array_equal(times1, times3)
    # make sure it's different with a different compensation:
    assert_true(np.mean(np.abs(data1 - data3)) > 1e-12)
    assert_raises(ValueError, Raw, ctf_comp_fname, compensation=33)

    # Try IO with compensation
    temp_file = op.join(tempdir, "raw.fif")

    raw1.save(temp_file, overwrite=True)
    raw4 = Raw(temp_file)
    data4, times4 = raw4[:, :]
    assert_array_equal(times1, times4)
    assert_array_equal(data1, data4)

    # Now save the file that has modified compensation
    # and make sure we can the same data as input ie. compensation
    # is undone
    raw3.save(temp_file, overwrite=True)
    raw5 = Raw(temp_file)
    data5, times5 = raw5[:, :]
    assert_array_equal(times1, times5)
    assert_allclose(data1, data5, rtol=1e-12, atol=1e-22)

def test_rank_estimation():
    """Test raw rank estimation
    """
    raw = Raw(fif_fname)
    n_meg = len(pick_types(raw.info, meg=True, eeg=False, exclude="bads"))
    n_eeg = len(pick_types(raw.info, meg=False, eeg=True, exclude="bads"))
    raw = Raw(fif_fname, preload=True)
    assert_array_equal(raw.estimate_rank(), n_meg + n_eeg)
    raw = Raw(fif_fname, preload=False)
    raw.apply_proj()
    n_proj = len(raw.info["projs"])
    assert_array_equal(raw.estimate_rank(tstart=10, tstop=20), n_meg + n_eeg - n_proj)