def setUp(self):
self.graph = Graph()
self.operator_split = OpSplitMaskArray(graph=self.graph)
self.operator_split.Input.meta.axistags = vigra.AxisTags("txyzc")
class TestOpSplitMaskArray(object):
def setUp(self):
self.graph = Graph()
self.operator_split = OpSplitMaskArray(graph=self.graph)
self.operator_split.Input.meta.axistags = vigra.AxisTags("txyzc")
def test1(self):
# Generate a random dataset and see if it we get the right masking from the operator.
data = numpy.random.random((4, 5, 6, 7, 3)).astype(numpy.float32)
mask = numpy.zeros(data.shape, dtype=bool)
# Mask borders of the expected output.
left_slicing = (mask.ndim - 1) * (slice(None),) + (slice(None, 1),)
right_slicing = (mask.ndim - 1) * (slice(None),) + (slice(-1, None),)
for i in xrange(mask.ndim):
left_slicing = left_slicing[-1:] + left_slicing[:-1]
right_slicing = right_slicing[-1:] + right_slicing[:-1]
mask[left_slicing] = True
mask[right_slicing] = True
data = numpy.ma.masked_array(
data,
mask=mask,
fill_value=data.dtype.type(numpy.nan),
shrink=False
)
# Provide input read all output.
self.operator_split.Input.setValue(data)
output_array = self.operator_split.OutputArray[None].wait()
output_mask = self.operator_split.OutputMask[None].wait()
output_fill_value = self.operator_split.OutputFillValue[None].wait()
assert not isinstance(output_array, numpy.ma.masked_array)
assert not isinstance(output_mask, numpy.ma.masked_array)
assert not isinstance(output_fill_value, numpy.ma.masked_array)
assert output_array.dtype.type == data.dtype.type
assert output_mask.dtype.type == data.mask.dtype
assert output_fill_value.dtype.type == data.dtype.type
assert (
(data.data == output_array) |
(numpy.isnan(data.data) & numpy.isnan(output_array))
).all()
assert (data.mask == output_mask).all()
assert (
(data.fill_value[()] == output_fill_value) |
(numpy.isnan(data.fill_value[()]) & numpy.isnan(output_fill_value))
).all()
def test2(self):
# Generate a dataset and grab chunks of it from the operator. The result should be the same as above.
data = numpy.random.random((4, 5, 6, 7, 3)).astype(numpy.float32)
mask = numpy.zeros(data.shape, dtype=bool)
# Mask borders of the expected output.
left_slicing = (mask.ndim - 1) * (slice(None),) + (slice(None, 1),)
right_slicing = (mask.ndim - 1) * (slice(None),) + (slice(-1, None),)
for i in xrange(mask.ndim):
left_slicing = left_slicing[-1:] + left_slicing[:-1]
right_slicing = right_slicing[-1:] + right_slicing[:-1]
mask[left_slicing] = True
mask[right_slicing] = True
data = numpy.ma.masked_array(
data,
mask=mask,
fill_value=data.dtype.type(numpy.nan),
shrink=False
)
# Create array to store results. Don't keep original data.
output_array = data.data.copy()
output_array[:] = 0
output_mask = data.mask.copy()
output_mask[:] = False
# Provide input and grab chunks.
self.operator_split.Input.setValue(data)
output_array[:2] = self.operator_split.OutputArray[:2].wait()
output_mask[:2] = self.operator_split.OutputMask[:2].wait()
output_array[2:] = self.operator_split.OutputArray[2:].wait()
output_mask[2:] = self.operator_split.OutputMask[2:].wait()
output_fill_value = self.operator_split.OutputFillValue[None].wait()
assert not isinstance(output_array, numpy.ma.masked_array)
assert not isinstance(output_mask, numpy.ma.masked_array)
assert not isinstance(output_fill_value, numpy.ma.masked_array)
assert output_array.dtype.type == data.dtype.type
assert output_mask.dtype.type == data.mask.dtype
assert output_fill_value.dtype.type == data.dtype.type
assert (
(data.data == output_array) |
(numpy.isnan(data.data) & numpy.isnan(output_array))
).all()
assert (data.mask == output_mask).all()
assert (
(data.fill_value[()] == output_fill_value) |
(numpy.isnan(data.fill_value[()]) & numpy.isnan(output_fill_value))
).all()
def tearDown(self):
# Take down operators
self.operator_split.Input.disconnect()
self.operator_split.OutputArray.disconnect()
self.operator_split.OutputMask.disconnect()
self.operator_split.OutputFillValue.disconnect()
self.operator_split.cleanUp()
