def test_downsampling_consistency():
d = np.arange(1, 24)
s = channel.Slice(channel.Continuous(d, with_offset(0), 10))
# Multiple of 5 should downsample to the same irrespective of the method
# Source frequency was 1e9 / 10 Hz. So we go to .2e8 Hz.
s1 = s.downsampled_to(.2e8)
s2 = s.downsampled_by(5)
np.testing.assert_allclose(s1.data, s2.data)
np.testing.assert_equal(s1.timestamps, s2.timestamps)
d = np.arange(1, 24)
s = channel.Slice(channel.Continuous(d, with_offset(5), 10))
# Multiple of 5 should downsample to the same irrespective of the method
# Source frequency was 1e9 / 10 Hz. So we go to .2e8 Hz.
s1 = s.downsampled_to(.2e8)
s2 = s.downsampled_by(5)
np.testing.assert_allclose(s1.data, s2.data)
np.testing.assert_equal(s1.timestamps, s2.timestamps)
# Multiple of 5 should downsample to the same irrespective of the method
# Source frequency was 1e9 / 7 Hz.
d = np.arange(1, 24)
s = channel.Slice(channel.Continuous(d, with_offset(0), 7))
s1 = s.downsampled_to(int(1e9 / 7 / 5))
s2 = s.downsampled_by(5)
np.testing.assert_allclose(s1.data, s2.data)
np.testing.assert_equal(s1.timestamps, s2.timestamps)
def test_calibration_continuous_channels():
time_field = 'Stop time (ns)'
mock_calibration = ForceCalibration(time_field=time_field,
items=[
{'Calibration Data': 50, time_field: 50},
{'Calibration Data': 20, time_field: 20},
{'Calibration Data': 30, time_field: 30},
{'Calibration Data': 40, time_field: 40},
{'Calibration Data': 80, time_field: 80},
{'Calibration Data': 90, time_field: 90},
{'Calibration Data': 120, time_field: 120},
])
# Channel should have calibration points 40, 50 since this is the only area that has force data.
cc = channel.Slice(channel.Continuous([14, 15, 16, 17], 40, 10), calibration=mock_calibration)
assert len(cc.calibration) == 2
assert cc.calibration[0]["Calibration Data"] == 40
assert cc.calibration[1]["Calibration Data"] == 50
# Channel should have calibration points 40, 50, 80, 90, 120
# and time points 40, 50, ... 120
cc = channel.Slice(channel.Continuous(np.arange(14, 23, 1), 40, 10), calibration=mock_calibration)
assert len(cc.calibration) == 5
calibration = cc[50:80].calibration
assert len(calibration) == 1
assert calibration[0]["Calibration Data"] == 50
calibration = cc[50:90].calibration
assert len(calibration) == 2
assert calibration[0]["Calibration Data"] == 50
assert calibration[1]["Calibration Data"] == 80
# Check whether slice nesting works for calibration data
# :120 => keeps 40, 50, 80, 90
nested_slice = cc[:120]
assert len(nested_slice.calibration) == 4
assert nested_slice.calibration[0]["Calibration Data"] == 40
assert nested_slice.calibration[1]["Calibration Data"] == 50
assert nested_slice.calibration[2]["Calibration Data"] == 80
assert nested_slice.calibration[3]["Calibration Data"] == 90
nested_slice = nested_slice[:90]
assert len(nested_slice.calibration) == 3
# 120 and up results in calibration point 120.
# This case would be 80 if calibration data would be sliced every time, rather than filtered only when requested.
nested_slice = nested_slice[120:]
assert len(nested_slice.calibration) == 1
assert nested_slice.calibration[0]["Calibration Data"] == 120
def test_correlation():
cc = channel.Slice(channel.Continuous(np.arange(10, 80, 2), 10, 2))
# Test image stack without dead time
fake_tiff = TiffStack(
MockTiff([["10", "20"], ["20", "30"], ["30", "40"], ["40", "50"],
["50", "60"], ["60", "70"]]))
stack = CorrelatedStack.from_data(fake_tiff)
assert (np.allclose(
np.hstack([cc[x.start:x.stop].data for x in stack[2:4]]),
np.arange(30, 50, 2)))
# Test image stack with dead time
fake_tiff = TiffStack(
MockTiff([["10", "18"], ["20", "28"], ["30", "38"], ["40", "48"],
["50", "58"], ["60", "68"]]))
stack = CorrelatedStack.from_data(fake_tiff)
assert (np.allclose(
np.hstack([cc[x.start:x.stop].data for x in stack[2:4]]),
np.hstack([np.arange(30, 38, 2),
np.arange(40, 48, 2)])))
# Unit test which tests whether we obtain an appropriately downsampled time series when ask for downsampling of a
# slice based on a stack.
ch = cc.downsampled_over(stack[0:3].timestamps)
assert (np.allclose(ch.data, [
np.mean(np.arange(10, 18, 2)),
np.mean(np.arange(20, 28, 2)),
np.mean(np.arange(30, 38, 2))
]))
assert (np.allclose(ch.timestamps, [(10 + 18) / 2, (20 + 28) / 2,
(30 + 38) / 2]))
ch = cc.downsampled_over(stack[1:4].timestamps)
assert (np.allclose(ch.data, [
np.mean(np.arange(20, 28, 2)),
np.mean(np.arange(30, 38, 2)),
np.mean(np.arange(40, 48, 2))
]))
assert (np.allclose(ch.timestamps, [(20 + 28) / 2, (30 + 38) / 2,
(40 + 48) / 2]))
with pytest.raises(TypeError):
cc.downsampled_over(stack[1:4])
with pytest.raises(ValueError):
cc.downsampled_over(stack[1:4].timestamps, where='up')
with pytest.raises(AssertionError):
cc["0ns":"20ns"].downsampled_over(stack[3:4].timestamps)
with pytest.raises(AssertionError):
cc["40ns":"70ns"].downsampled_over(stack[0:1].timestamps)
assert (stack[0].raw.start == 10)
assert (stack[1].raw.start == 20)
assert (stack[1:3][0].raw.start == 20)
assert (stack[1:3].raw[0].start == 20)
assert (stack[1:3].raw[1].start == 30)
def test_downsampling():
s = channel.Slice(channel.Continuous([14, 15, 16, 17], start=with_offset(0), dt=10))
assert s.sample_rate == 1e8
s2 = s.downsampled_by(2)
np.testing.assert_allclose(s2.data, 14.5, 16.5)
np.testing.assert_equal(s2.timestamps, with_offset([5, 25]))
assert s2.sample_rate == 0.5e8
s4 = s.downsampled_by(4)
np.testing.assert_allclose(s4.data, 15.5)
np.testing.assert_equal(s4.timestamps, with_offset([15]))
assert s4.sample_rate == 0.25e8
s3 = s.downsampled_by(3)
np.testing.assert_allclose(s3.data, 15)
np.testing.assert_equal(s3.timestamps, with_offset([10]))
assert s3.sample_rate == 33333333
s22 = s2.downsampled_by(2)
np.testing.assert_allclose(s22.data, 15.5)
np.testing.assert_equal(s22.timestamps, with_offset([15]))
assert s22.sample_rate == 0.25e8
with pytest.raises(ValueError):
s.downsampled_by(-1)
with pytest.raises(TypeError):
s.downsampled_by(1.5)
def test_consistency_downsampled_to():
d = np.arange(1, 41)
s = channel.Slice(channel.Continuous(d, with_offset(50), 10))
one_step = s.downsampled_to(.1e8)
two_step = s.downsampled_to(.2e8).downsampled_to(.1e8)
np.testing.assert_allclose(one_step.data, two_step.data)
np.testing.assert_allclose(one_step.timestamps, two_step.timestamps)
def test_downsampling_over_subset():
d = np.arange(1, 24)
s = channel.Slice(channel.Continuous(d, with_offset(0), 10))
sd = s.downsampled_over([*with_offset([(20, 40), (40, 60), (60, 80)])])
# Data starts at 1, timestamps start at 0. 20-40 corresponds to data [3, 4], 40-60 to [5,6] etc.
np.testing.assert_allclose(sd.data, [(3 + 4) / 2, (5 + 6) / 2, (7 + 8) / 2])
np.testing.assert_equal(
sd.timestamps, with_offset([(20 + 30) // 2, (40 + 50) // 2, (60 + 70) // 2])
)
def test_start_stop():
s = channel.Slice(channel.TimeSeries([14, 15, 16, 17], [4, 6, 8, 10]))
np.testing.assert_allclose(s.start, 4)
np.testing.assert_allclose(s.stop, 10 + 1)
s = channel.Slice(channel.Continuous([14, 15, 16, 17], 4, 2))
np.testing.assert_allclose(s.start, 4)
np.testing.assert_allclose(s.stop, 12)
s = channel.Slice(channel.TimeTags([14, 15, 16, 17]))
np.testing.assert_allclose(s.start, 14)
np.testing.assert_allclose(s.stop, 17 + 1)
s = channel.Slice(channel.TimeTags([14, 15, 16, 17], 4, 30))
np.testing.assert_allclose(s.start, 4)
np.testing.assert_allclose(s.stop, 30)
def test_plot_correlated_smaller_channel():
from matplotlib.backend_bases import MouseEvent
# Regression test for a bug where the start index was added twice. In the regression, this lead to an out of range
# error.
fake_tiff = TiffStack(
MockTiffFile(
data=[
np.zeros((3, 3)),
np.ones((3, 3)),
np.ones((3, 3)) * 2,
np.ones((3, 3)) * 3,
np.ones((3, 3)) * 4,
np.ones((3, 3)) * 5,
],
times=make_frame_times(7, step=10),
),
align_requested=False,
)
# Add test for when there's only a subset in terms of channel data
cc = channel.Slice(channel.Continuous(np.arange(10, 80, 2), 30, 2), {
"y": "mock",
"title": "mock"
})
with pytest.warns(UserWarning):
CorrelatedStack.from_dataset(fake_tiff).plot_correlated(cc)
def mock_click(fig, data_position):
pos = fig.axes[0].transData.transform(data_position)
fig.canvas.callbacks.process(
"button_press_event",
MouseEvent("button_press_event", fig.canvas, pos[0], pos[1], 1))
images = [
obj for obj in mpl.pyplot.gca().get_children()
if isinstance(obj, mpl.image.AxesImage)
]
assert len(images) == 1
return images[0].get_array()
np.testing.assert_allclose(mock_click(mpl.pyplot.gcf(), np.array([0, 40])),
np.ones((3, 3)) * 2)
np.testing.assert_allclose(
mock_click(mpl.pyplot.gcf(), np.array([10.1e-9, 40])),
np.ones((3, 3)) * 3)
def test_slice_properties():
size = 5
s = channel.Slice(channel.TimeSeries(np.random.rand(size), np.random.rand(size)))
assert len(s) == size
assert s.sample_rate is None
s = channel.Slice(channel.Continuous(np.random.rand(size), start=0, dt=1))
assert len(s) == size
assert s.sample_rate == 1e9
size = 10
s = channel.Slice(channel.TimeTags(np.arange(0, size, dtype=np.int64)))
assert len(s) == size
assert s.sample_rate is None
s = channel.empty_slice
assert len(s) == 0
assert s.sample_rate is None
def test_downsampling_like():
d = [1, 1, 2, 2, 3, 3, 4, 4, 5, 5, 5, 5, 5, 5, 6, 6, 6, 7, 7, 7, 8, 8, 8, 9, 9, 9]
s = channel.Slice(channel.Continuous(d, with_offset(0), 2))
t_downsampled = with_offset([0, 4, 8, 12, 16, 34, 40, 46, 50, 54])
y_downsampled = np.array([0, 1, 2, 3, 4, 6, 7, 8, 9, 10])
reference = channel.Slice(channel.TimeSeries(y_downsampled, t_downsampled))
ds, ref_out = s.downsampled_like(reference)
np.testing.assert_equal(ds.timestamps, ref_out.timestamps)
np.testing.assert_equal(t_downsampled[1:-1], ds.timestamps)
np.testing.assert_allclose(y_downsampled[1:-1], ds.data)
with pytest.raises(NotImplementedError):
reference.downsampled_like(reference)
with pytest.raises(AssertionError):
s.downsampled_like(s)
def test_channel_plot():
def testLine(x, y):
data = [obj for obj in mpl.pyplot.gca().get_children() if isinstance(obj, mpl.lines.Line2D)]
assert len(data) == 1
line = data[0].get_data()
np.testing.assert_allclose(line[0], x)
np.testing.assert_allclose(line[1], y)
d = np.arange(1, 24)
s = channel.Slice(channel.Continuous(d, int(5e9), int(10e9)))
s.plot()
testLine(np.arange(0, 230, 10), d)
mpl.pyplot.gca().clear()
s.plot(start=0)
testLine(np.arange(5, 230, 10), d)
mpl.pyplot.gca().clear()
s.plot(start=100e9)
testLine(np.arange(0, 230, 10) - 100 + 5, d)
def test_continuous_idexing():
s = channel.Slice(channel.Continuous([14, 15, 16, 17], 4, 1))
np.testing.assert_equal(s[0:5].data, [14])
np.testing.assert_equal(s[0:5].timestamps, [4])
np.testing.assert_equal(s[4:5].data, [14])
np.testing.assert_equal(s[4:5].timestamps, [4])
np.testing.assert_equal(s[4:6].data, [14, 15])
np.testing.assert_equal(s[4:6].timestamps, [4, 5])
np.testing.assert_equal(s[4:10].data, [14, 15, 16, 17])
np.testing.assert_equal(s[4:10].timestamps, [4, 5, 6, 7])
s = channel.Slice(channel.Continuous([14, 15, 16, 17], 4, 2))
np.testing.assert_equal(s[0:5].data, [14])
np.testing.assert_equal(s[0:5].timestamps, [4])
np.testing.assert_equal(s[4:5].data, [14])
np.testing.assert_equal(s[4:5].timestamps, [4])
np.testing.assert_equal(s[4:8].data, [14, 15])
np.testing.assert_equal(s[4:8].timestamps, [4, 6])
np.testing.assert_equal(s[4:14].data, [14, 15, 16, 17])
np.testing.assert_equal(s[4:14].timestamps, [4, 6, 8, 10])
with pytest.raises(IndexError) as exc:
assert s[1]
assert str(exc.value) == "Scalar indexing is not supported, only slicing"
with pytest.raises(IndexError) as exc:
assert s[1:2:3]
assert str(exc.value) == "Slice steps are not supported"
s = channel.Slice(channel.TimeSeries([], []))
assert len(s[1:2].data) == 0
assert len(s[1:2].timestamps) == 0
# Regression test for slicing within timestep
s = channel.Slice(channel.Continuous([14, 15, 16, 17], 4, 2))
assert s[5:15].timestamps[0] == 6
s = channel.Slice(channel.Continuous([14, 15, 16, 17], -4, 2))
assert s[-3:].timestamps[0] == -2
s = channel.Slice(channel.Continuous([14, 15, 16, 17], 4, 3))
assert s[3:15].timestamps[0] == 4
s = channel.Slice(channel.Continuous([14, 15, 16, 17], 4, 3))
assert s[4:15].timestamps[0] == 4
assert s[5:15].timestamps[0] == 7
assert s[6:15].timestamps[0] == 7
assert s[7:15].timestamps[0] == 7
assert s[8:15].timestamps[0] == 10
assert s[6:14].timestamps[-1] == 13
assert s[6:13].timestamps[-1] == 10
def test_plot_correlated():
cc = channel.Slice(channel.Continuous(np.arange(10, 80, 2), 10, 2), {
"y": "mock",
"title": "mock"
})
# Regression test for a bug where the start index was added twice. In the regression, this lead to an out of range
# error.
fake_tiff = TiffStack(
MockTiffFile(
data=[
np.zeros((3, 3)),
np.ones((3, 3)),
np.ones((3, 3)) * 2,
np.ones((3, 3)) * 3,
np.ones((3, 3)) * 4,
np.ones((3, 3)) * 5,
],
times=make_frame_times(7, step=10),
),
align_requested=False,
)
CorrelatedStack.from_dataset(fake_tiff)[3:5].plot_correlated(cc)
imgs = [
obj for obj in mpl.pyplot.gca().get_children()
if isinstance(obj, mpl.image.AxesImage)
]
assert len(imgs) == 1
np.testing.assert_allclose(imgs[0].get_array(), np.ones((3, 3)) * 3)
CorrelatedStack.from_dataset(fake_tiff)[3:5].plot_correlated(cc, frame=1)
imgs = [
obj for obj in mpl.pyplot.gca().get_children()
if isinstance(obj, mpl.image.AxesImage)
]
assert len(imgs) == 1
np.testing.assert_allclose(imgs[0].get_array(), np.ones((3, 3)) * 4)
def test_continuous_downsampling_to():
d = np.arange(1, 24)
s = channel.Slice(channel.Continuous(d, with_offset(0), 500)) # 2 MHz
# to 1000 ns step
s2a = s.downsampled_to(1e6, where='left')
np.testing.assert_equal(s2a.timestamps, with_offset(np.arange(0, 11000, 1000)))
np.testing.assert_allclose(
s2a.data, [1.5, 3.5, 5.5, 7.5, 9.5, 11.5, 13.5, 15.5, 17.5, 19.5, 21.5]
)
s2b = s.downsampled_to(1e6, where="center")
np.testing.assert_equal(
s2b.timestamps, with_offset((np.arange(0, 10500, 1000) + np.arange(500, 10501, 1000)) // 2)
)
np.testing.assert_allclose(
s2a.data, [1.5, 3.5, 5.5, 7.5, 9.5, 11.5, 13.5, 15.5, 17.5, 19.5, 21.5]
)
# upsampling
with pytest.raises(ValueError):
s.downsampled_to(3e6, where="left")
# non-integer ratio
with pytest.raises(ValueError):
s.downsampled_to(3e5, where="left")
# non-integer ratio
s3a = s.downsampled_to(3e5, where='left', method="ceil")
np.testing.assert_equal(s3a.timestamps, with_offset([0, 3000, 6000]))
np.testing.assert_allclose(s3a.data, [3.5, 9.5, 15.5])
s3b = s.downsampled_to(3e5, where='center', method="ceil")
np.testing.assert_equal(
s3b.timestamps, with_offset([(0 + 2500) // 2, (3000 + 5500) // 2, (6000 + 8500) // 2])
)
np.testing.assert_allclose(s3a.data, [3.5, 9.5, 15.5])
def test_correlation(shape):
cc = channel.Slice(channel.Continuous(np.arange(10, 80, 2), 10, 2))
# Test image stack without dead time
fake_tiff = TiffStack(
MockTiffFile(data=[np.ones(shape)] * 6,
times=make_frame_times(6, step=10)),
align_requested=False,
)
stack = CorrelatedStack.from_dataset(fake_tiff)
np.testing.assert_allclose(
np.hstack([cc[x.start:x.stop].data for x in stack[2:4]]),
np.arange(30, 50, 2))
# Test image stack with dead time
fake_tiff = TiffStack(MockTiffFile(data=[np.ones(shape)] * 6,
times=make_frame_times(6)),
align_requested=False)
stack = CorrelatedStack.from_dataset(fake_tiff)
np.testing.assert_allclose(
np.hstack([cc[x.start:x.stop].data for x in stack[2:4]]),
np.hstack([np.arange(30, 38, 2),
np.arange(40, 48, 2)]),
)
# Unit test which tests whether we obtain an appropriately downsampled time series when ask for downsampling of a
# slice based on a stack.
ch = cc.downsampled_over(stack[0:3].frame_timestamp_ranges)
np.testing.assert_allclose(
ch.data,
[
np.mean(np.arange(10, 18, 2)),
np.mean(np.arange(20, 28, 2)),
np.mean(np.arange(30, 38, 2)),
],
)
np.testing.assert_allclose(ch.timestamps, [(10 + 16) / 2, (20 + 26) / 2,
(30 + 36) / 2])
ch = cc.downsampled_over(stack[1:4].frame_timestamp_ranges)
np.testing.assert_allclose(
ch.data,
[
np.mean(np.arange(20, 28, 2)),
np.mean(np.arange(30, 38, 2)),
np.mean(np.arange(40, 48, 2)),
],
)
np.testing.assert_allclose(ch.timestamps, [(20 + 26) / 2, (30 + 36) / 2,
(40 + 46) / 2])
with pytest.raises(TypeError):
cc.downsampled_over(stack[1:4])
with pytest.raises(ValueError):
cc.downsampled_over(stack[1:4].frame_timestamp_ranges, where="up")
with pytest.raises(AssertionError):
cc["0ns":"20ns"].downsampled_over(stack[3:4].frame_timestamp_ranges)
with pytest.raises(AssertionError):
cc["40ns":"70ns"].downsampled_over(stack[0:1].frame_timestamp_ranges)
assert stack[0]._get_frame(0).start == 10
assert stack[1]._get_frame(0).start == 20
assert stack[1:3]._get_frame(0).start == 20
assert stack[1:3]._get_frame(0).start == 20
assert stack[1:3]._get_frame(1).start == 30
# Regression test downsampled_over losing precision due to reverting to double rather than int64.
cc = channel.Slice(
channel.Continuous(np.arange(10, 80, 2), 1588267266006287100, 1000))
ch = cc.downsampled_over([(1588267266006287100, 1588267266006287120)],
where="left")
assert int(ch.timestamps[0]) == 1588267266006287100
d = np.arange(1, 24)
s = channel.Slice(channel.Continuous(d, int(5e9), int(10e9)))
s.plot()
testLine(np.arange(0, 230, 10), d)
mpl.pyplot.gca().clear()
s.plot(start=0)
testLine(np.arange(5, 230, 10), d)
mpl.pyplot.gca().clear()
s.plot(start=100e9)
testLine(np.arange(0, 230, 10) - 100 + 5, d)
def test_regression_lazy_loading(channel_h5_file):
ch = channel.Continuous.from_dataset(channel_h5_file["Force HF"]["Force 1x"])
assert type(ch._src._src_data) == h5py.Dataset
@pytest.mark.parametrize(
"data, new_data",
[
(channel.Continuous([1, 2, 3, 4, 5], start=1, dt=1), np.array([5, 6, 7, 8, 9])),
(channel.TimeSeries([1, 2, 3, 4, 5], [2, 3, 4, 5, 6]), np.array([5, 6, 7, 8, 9])),
],
)
def test_with_data(data, new_data):
np.testing.assert_allclose(data._with_data(new_data).data, new_data)
old_timestamps = data.timestamps
np.testing.assert_allclose(data._with_data(new_data).timestamps, old_timestamps)
def test_seconds_property():
s = channel.Slice(channel.Continuous([14, 15, 16, 17], start=40, dt=1e9))
np.testing.assert_allclose(s.seconds, [0, 1, 2, 3])
s = channel.Slice(channel.TimeSeries([14, 15, 16, 17], [40e9, 41e9, 42e9, 43e9]))
np.testing.assert_allclose(s.seconds, [0, 1, 2, 3])