def anyarray(
draw,
min_dims: int = 0,
max_dims: int = 2,
include_complex_numbers: bool = True,
dtype: Optional[np.dtype] = None,
):
if dtype is None:
if include_complex_numbers:
dtype = one_of(
integer_dtypes(), floating_dtypes(), complex_number_dtypes()
)
else:
dtype = one_of(integer_dtypes(), floating_dtypes())
arr = draw(
arrays(
dtype=dtype,
shape=array_shapes(min_dims=min_dims, max_dims=max_dims),
)
)
assume(not np.any(np.isnan(arr)))
assume(np.all(np.isfinite(arr)))
return arr
def test_weighted_negative_log_likelihood_vs_softmax_cross_entropy(
data: st.DataObject, labels_as_tensor: bool):
s = data.draw(
hnp.arrays(
shape=hnp.array_shapes(min_side=1,
max_side=10,
min_dims=2,
max_dims=2),
dtype=float,
elements=st.floats(-100, 100),
))
y_true = data.draw(
hnp.arrays(
shape=(s.shape[0], ),
dtype=hnp.integer_dtypes(),
elements=st.integers(min_value=0, max_value=s.shape[1] - 1),
).map(Tensor if labels_as_tensor else lambda x: x))
weights = data.draw(
hnp.arrays(
shape=(s.shape[1], ),
dtype=float,
elements=st.floats(1e-8, 100),
))
scores = Tensor(s)
weights = Tensor(weights)
for score, y in zip(scores, y_true):
score = mg.log(mg.nnet.softmax(score.reshape(1, -1)))
y = y.reshape(-1)
nll = negative_log_likelihood(score, y)
weighted_nll = negative_log_likelihood(score, y, weights=weights)
assert np.isclose(weighted_nll.data, weights[y.data].data * nll.data)
def test_softmax_crossentropy(data: st.DataObject, labels_as_tensor: bool):
s = data.draw(
hnp.arrays(
shape=hnp.array_shapes(max_side=10, min_dims=2, max_dims=2),
dtype=float,
elements=st.floats(-100, 100),
))
y_true = data.draw(
hnp.arrays(
shape=(s.shape[0], ),
dtype=hnp.integer_dtypes(),
elements=st.integers(min_value=0, max_value=s.shape[1] - 1),
).map(Tensor if labels_as_tensor else lambda x: x))
scores = Tensor(s)
softmax_cross = softmax_crossentropy(scores, y_true, constant=False)
softmax_cross.backward()
mygrad_scores = Tensor(s)
probs = softmax(mygrad_scores)
correct_labels = (range(len(y_true)),
y_true.data if labels_as_tensor else y_true)
truth = np.zeros(mygrad_scores.shape)
truth[correct_labels] = 1
mygrad_cross = (-1 / s.shape[0]) * (log(probs) * truth).sum()
mygrad_cross.backward()
assert_allclose(softmax_cross.data,
mygrad_cross.data,
atol=1e-5,
rtol=1e-5)
assert_allclose(scores.grad, mygrad_scores.grad, atol=1e-5, rtol=1e-5)
def test_multiclass_hinge(data):
""" Test the built-in implementation of multiclass hinge against the pure pygrad version"""
s = data.draw(
hnp.arrays(
shape=hnp.array_shapes(max_side=10, min_dims=2, max_dims=2),
dtype=float,
elements=st.floats(-100, 100),
))
l = data.draw(
hnp.arrays(
shape=(s.shape[0], ),
dtype=hnp.integer_dtypes(),
elements=st.integers(min_value=0, max_value=s.shape[1] - 1),
))
hinge_scores = Tensor(s)
hinge_loss = multiclass_hinge(hinge_scores, l, constant=False)
hinge_loss.backward()
pygrad_scores = Tensor(s)
correct_labels = (range(len(l)), l)
correct_class_scores = pygrad_scores[correct_labels] # Nx1
Lij = pygrad_scores - correct_class_scores[:,
np.newaxis] + 1.0 # NxC margins
Lij[Lij <= 0] = 0
Lij[correct_labels] = 0
pygrad_loss = Lij.sum() / pygrad_scores.shape[0]
pygrad_loss.backward()
assert_allclose(hinge_loss.data, pygrad_loss.data)
assert_allclose(pygrad_scores.grad, hinge_scores.grad)
def test_softmax_crossentropy(data):
""" Test the built-in implementation of multiclass hinge against the pure pygrad version"""
s = data.draw(
hnp.arrays(
shape=hnp.array_shapes(max_side=10, min_dims=2, max_dims=2),
dtype=float,
elements=st.floats(-100, 100),
))
l = data.draw(
hnp.arrays(
shape=(s.shape[0], ),
dtype=hnp.integer_dtypes(),
elements=st.integers(min_value=0, max_value=s.shape[1] - 1),
))
scores = Tensor(s)
softmax_cross = softmax_crossentropy(scores, l, constant=False)
softmax_cross.backward()
pygrad_scores = Tensor(s)
probs = softmax(pygrad_scores)
correct_labels = (range(len(l)), l)
truth = np.zeros(pygrad_scores.shape)
truth[correct_labels] = 1
pygrad_cross = (-1 / s.shape[0]) * (log(probs) * truth).sum()
pygrad_cross.backward()
assert_allclose(softmax_cross.data,
pygrad_cross.data,
atol=1e-5,
rtol=1e-5)
assert_allclose(scores.grad, pygrad_scores.grad, atol=1e-5, rtol=1e-5)
def test_ranked_margin(shape, margin, data):
x1 = data.draw(
hnp.arrays(shape=shape, dtype=float, elements=st.floats(-1000, 1000)),
label="x1",
)
x2 = data.draw(
hnp.arrays(shape=shape, dtype=float, elements=st.floats(-1000, 1000)),
label="x2",
)
y = data.draw(
st.sampled_from((-1, 1))
| hnp.arrays(
shape=shape[:1],
dtype=hnp.integer_dtypes(),
elements=st.sampled_from((-1, 1)),
),
label="y",
)
x1_copy = np.copy(x1)
x2_copy = np.copy(x2)
y_copy = np.copy(y)
x1_dum = mg.Tensor(x1)
x2_dum = mg.Tensor(x2)
x1_real = mg.Tensor(x1)
x2_real = mg.Tensor(x2)
loss_dum = simple_loss(x1_dum, x2_dum, y, margin)
loss_real = margin_ranking_loss(x1_real, x2_real, y, margin)
assert_allclose(actual=loss_real.data,
desired=loss_dum.data,
err_msg="losses don't match")
assert_array_equal(x1, x1_copy, err_msg="`x1` was mutated by forward")
assert_array_equal(x2, x2_copy, err_msg="`x2` was mutated by forward")
if isinstance(y, np.ndarray):
assert_array_equal(y, y_copy, err_msg="`y` was mutated by forward")
loss_dum.backward()
loss_real.backward()
assert_allclose(actual=x1_real.grad,
desired=x1_dum.grad,
err_msg="x1.grad doesn't match")
assert_allclose(actual=x2_real.grad,
desired=x2_dum.grad,
err_msg="x2.grad doesn't match")
assert_array_equal(x1, x1_copy, err_msg="`x1` was mutated by backward")
assert_array_equal(x2, x2_copy, err_msg="`x2` was mutated by backward")
if isinstance(y, np.ndarray):
assert_array_equal(y, y_copy, err_msg="`y` was mutated by backward")
loss_real.null_gradients()
assert x1_real.grad is None
assert x2_real.grad is None
def get_scalar_dtype_strategy(exclude=None):
"""
A `hypothesis` strategy yielding
"""
possible_strategies = {
"datetime": hyp_np.datetime64_dtypes(max_period="ms", min_period="ns"),
"uint": hyp_np.unsigned_integer_dtypes(),
"int": hyp_np.integer_dtypes(),
"float": hyp_np.floating_dtypes(),
"byte": hyp_np.byte_string_dtypes(),
"unicode": hyp_np.unicode_string_dtypes(),
}
if exclude is None:
exclude = {}
elif not isinstance(exclude, list):
exclude = [exclude]
for ex in exclude:
if ex in possible_strategies:
del possible_strategies[ex]
else:
raise ValueError(
"Strategy {} unknown. Possible values are {}".format(
ex, possible_strategies.keys()
)
)
return hyp_st.one_of(*list(possible_strategies.values()))
def test_can_restrict_endianness(data, byteorder):
dtype = data.draw(
nps.integer_dtypes(endianness=byteorder, sizes=(16, 32, 64)))
if byteorder == ("<" if sys.byteorder == "little" else ">"):
assert dtype.byteorder == "="
else:
assert dtype.byteorder == byteorder
def test_negative_log_likelihood_vs_softmax_cross_entropy(
data: st.DataObject, labels_as_tensor: bool):
s = data.draw(
hnp.arrays(
shape=hnp.array_shapes(max_side=10, min_dims=2, max_dims=2),
dtype=float,
elements=st.floats(-100, 100),
))
y_true = data.draw(
hnp.arrays(
shape=(s.shape[0], ),
dtype=hnp.integer_dtypes(),
elements=st.integers(min_value=0, max_value=s.shape[1] - 1),
).map(Tensor if labels_as_tensor else lambda x: x))
scores = Tensor(s)
nll = negative_log_likelihood(mg.log(mg.nnet.softmax(scores)), y_true)
nll.backward()
cross_entropy_scores = Tensor(s)
ce = softmax_crossentropy(cross_entropy_scores, y_true)
ce.backward()
assert_allclose(nll.data, ce.data, atol=1e-5, rtol=1e-5)
assert_allclose(scores.grad,
cross_entropy_scores.grad,
atol=1e-5,
rtol=1e-5)
def ints_floats_datetimes_and_timedeltas(draw):
dtypes = (
unsigned_integer_dtypes(endianness="="),
integer_dtypes(endianness="="),
floating_dtypes(endianness="="),
datetime64_dtypes(endianness="="),
timedelta64_dtypes(endianness="="),
)
return draw(one_of(dtypes))
def ints_floats_complex_or_booleans(draw):
dtypes = (
unsigned_integer_dtypes(endianness="="),
integer_dtypes(endianness="="),
floating_dtypes(endianness="="),
# complex_number_dtypes(endianness="="),
boolean_dtypes(),
)
return draw(one_of(dtypes))
def ints_floats_or_complex_dtypes(draw):
# Endianness needs to be specified for now, otherwise the byte-order may get flipped
# https://jira/browse/SOQTEST-6478
dtypes = (
unsigned_integer_dtypes(endianness="="),
integer_dtypes(endianness="="),
floating_dtypes(endianness="="),
# complex_number_dtypes(endianness="="),
)
return draw(one_of(dtypes))
def ints_or_floats_dtypes(draw):
# Endianness needs to be specified for now, otherwise the byte-order may get flipped
# https://jira/browse/SOQTEST-6478
# Half floats are not supported.
dtypes = (
unsigned_integer_dtypes(endianness="="),
integer_dtypes(endianness="="),
floating_dtypes(endianness="=", sizes=(32, 64)),
)
return draw(one_of(dtypes))
def array_with_two_entries(draw, array_length=10_000):
length = draw(integers(1, max_value=array_length))
arr = draw(
arrays(
dtype=one_of(integer_dtypes(), floating_dtypes()),
shape=(length, 2),
))
assume(not np.any(np.isnan(arr)))
assume(np.all(np.isfinite(arr)))
return arr
def gen_codes_for_fmt(fmt, include_nas=False, valid_only=True):
elems = None if not valid_only \
else hs.sampled_from(
sorted(fmt.keys()) + [np.nan] if include_nas \
else sorted(fmt.keys()))
dtypes = hnp.floating_dtypes(endianness='=') \
if include_nas or valid_only \
else hs.one_of(hnp.floating_dtypes(endianness='='),
hnp.integer_dtypes(endianness='='))
return hnp.arrays(dtype=dtypes,
shape=hnp.array_shapes(max_dims=1, max_side=1000),
elements=elems)
def test_generate_arbitrary_indices(data):
min_size = data.draw(st.integers(0, 10), "min_size")
max_size = data.draw(st.none() | st.integers(min_size, min_size + 10),
"max_size")
unique = data.draw(st.booleans(), "unique")
dtype = data.draw(
st.one_of(
npst.boolean_dtypes(),
npst.integer_dtypes(endianness="="),
npst.floating_dtypes(endianness="="),
npst.complex_number_dtypes(endianness="="),
npst.datetime64_dtypes(endianness="="),
npst.timedelta64_dtypes(endianness="="),
).filter(supported_by_pandas),
"dtype",
)
pass_elements = data.draw(st.booleans(), "pass_elements")
converted_dtype = pandas.Index([], dtype=dtype).dtype
try:
inferred_dtype = pandas.Index([data.draw(npst.from_dtype(dtype))
]).dtype
if pass_elements:
elements = npst.from_dtype(dtype)
dtype = None
else:
elements = None
index = data.draw(
pdst.indexes(
elements=elements,
dtype=dtype,
min_size=min_size,
max_size=max_size,
unique=unique,
))
except Exception as e:
if type(e).__name__ == "OutOfBoundsDatetime":
# See https://github.com/HypothesisWorks/hypothesis-python/pull/826
reject()
else:
raise
if dtype is None:
assert index.dtype == inferred_dtype
else:
assert index.dtype == converted_dtype
if unique:
assert len(set(index.values)) == len(index)
def get_nll_args(*arrs):
(s, ) = arrs
y_true = hnp.arrays(
shape=(s.shape[0], ),
dtype=hnp.integer_dtypes(),
elements=st.integers(min_value=0, max_value=s.shape[1] - 1),
)
weights = st.none() | hnp.arrays(
shape=(s.shape[1], ),
dtype=float,
elements=st.floats(1e-8, 100),
)
return st.fixed_dictionaries(dict(y_true=y_true, weights=weights))
def test_eager_flat_strict_upper_int_verification(self, data):
with eager_mode():
# sample square matrix
n = data.draw(st.integers(min_value=3, max_value=10))
input_ = data.draw(
np_st.arrays(dtype=np_st.integer_dtypes(), shape=(n, n)))
output = np.array(utils.flat_strict_upper(input_))
# verify
k = 0
for i in range(n):
for j in range(i + 1, n):
assert output[k] == input_[i, j]
k += 1
def one_of_supported_dtypes(draw):
# A strategy that selects a dtype that riptable is known to handle.
# dtype size 16-bit is not supported
# little endian is not supported
return one_of(
boolean_dtypes(),
integer_dtypes(endianness="=", sizes=(8, 32, 64)),
unsigned_integer_dtypes(endianness="=", sizes=(8, 32, 64)),
floating_dtypes(endianness="=", sizes=(32, 64)),
byte_string_dtypes(endianness="="),
unicode_string_dtypes(endianness="="),
# the following dtypes are not supported
# complex_number_dtypes(),
# datetime64_dtypes(),
# timedelta64_dtypes(),
)
def numpy_number(draw, min_val, max_val):
dtype = draw(st.one_of(nps.integer_dtypes(),
nps.unsigned_integer_dtypes(),
nps.floating_dtypes()))
if 'f' in dtype.str:
if min_val < np.finfo(dtype).min:
min_val = np.finfo(dtype).min
if max_val > np.finfo(dtype).max:
max_val = np.finfo(dtype).max
number = draw(st.floats(min_val, max_val, allow_nan=False,
allow_infinity=False))
else:
if min_val < np.iinfo(dtype).min:
min_val = np.iinfo(dtype).min
if max_val > np.iinfo(dtype).max:
max_val = np.iinfo(dtype).max
number = draw(st.integers(min_val, max_val))
return np.array([number], dtype)[0]
def numpy_number(draw, min_val, max_val):
dtype = draw(
st.one_of(nps.integer_dtypes(), nps.unsigned_integer_dtypes(),
nps.floating_dtypes()))
if 'f' in dtype.str:
if min_val < np.finfo(dtype).min:
min_val = np.finfo(dtype).min
if max_val > np.finfo(dtype).max:
max_val = np.finfo(dtype).max
number = draw(
st.floats(min_val, max_val, allow_nan=False, allow_infinity=False))
else:
min_val, max_val = np.ceil(min_val), np.floor(max_val)
if min_val < np.iinfo(dtype).min:
min_val = np.iinfo(dtype).min
if max_val > np.iinfo(dtype).max:
max_val = np.iinfo(dtype).max
number = draw(st.integers(min_val, max_val))
return np.array([number], dtype)[0]
def dataframe(draw):
n_cols = draw(integers(min_value=1, max_value=20))
dtypes = draw(
lists(
one_of(
np_strategies.floating_dtypes(),
np_strategies.integer_dtypes(),
np_strategies.unicode_string_dtypes(),
),
min_size=n_cols,
max_size=n_cols,
))
colnames = draw(
lists(text() | integers(),
min_size=n_cols,
max_size=n_cols,
unique=True))
return draw(
data_frames(columns=[
column(name=name, dtype=dtype)
for dtype, name in zip(dtypes, colnames)
]))
except ValueError:
try:
slow_result = slow_func(arr, axis=axis)
assert False
except ValueError:
return
slow_result = slow_func(arr, axis=axis)
assert_array_almost_equal(bn_result, slow_result)
@pytest.mark.parametrize("func", (bn.nanmin, bn.nanmax, bn.anynan, bn.allnan),
ids=lambda x: x.__name__)
@hypothesis.given(array=hy_arrays(
dtype=one_of(integer_dtypes(sizes=(32, 64)),
floating_dtypes(sizes=(32, 64))),
shape=array_shapes(),
))
def test_reduce_hypothesis(func, array):
_hypothesis_helper(func, array)
@pytest.mark.parametrize("func", (bn.nanargmin, bn.nanargmax),
ids=lambda x: x.__name__)
@hypothesis.given(array=hy_arrays(
dtype=one_of(integer_dtypes(sizes=(32, 64)),
floating_dtypes(sizes=(32, 64))),
shape=array_shapes(),
))
def test_reduce_hypothesis_errata(func, array):
from hypothesis import given
import hypothesis.strategies as st
import hypothesis.extra.numpy as hnp
import numpy as np
import pytest
from mygrad import Tensor
from mygrad.nnet.initializers import normal
@given(dtype=hnp.unsigned_integer_dtypes() | hnp.integer_dtypes()
| hnp.complex_number_dtypes())
def test_normal_dtype_validation(dtype):
with pytest.raises(ValueError):
normal(1, dtype=dtype)
@given(std=st.floats(-1000, 0, exclude_max=True))
def test_normal_std_validation(std):
with pytest.raises(ValueError):
normal(1, std=std)
_array_shapes = ((1000000, ), (1000, 100, 10), (10, 10, 10, 10, 10, 10))
@given(
shape=st.sampled_from(_array_shapes),
mean=st.floats(-100, 100),
std=st.floats(0, 5),
)
import hypothesis.extra.numpy as npst
import hypothesis.strategies as st
from hypothesis import given, settings
import xarray as xr
# Run for a while - arrays are a bigger search space than usual
settings.register_profile("ci", deadline=None)
settings.load_profile("ci")
an_array = npst.arrays(
dtype=st.one_of(
npst.unsigned_integer_dtypes(),
npst.integer_dtypes(),
npst.floating_dtypes(),
),
shape=npst.array_shapes(max_side=3), # max_side specified for performance
)
@given(st.data(), an_array)
def test_CFMask_coder_roundtrip(data, arr):
names = data.draw(st.lists(st.text(), min_size=arr.ndim,
max_size=arr.ndim, unique=True).map(tuple))
original = xr.Variable(names, arr)
coder = xr.coding.variables.CFMaskCoder()
roundtripped = coder.decode(coder.encode(original))
xr.testing.assert_identical(original, roundtripped)
def test_can_restrict_endianness(data, byteorder):
dtype = data.draw(nps.integer_dtypes(byteorder, sizes=(16, 32, 64)))
if byteorder == ('<' if sys.byteorder == 'little' else '>'):
assert dtype.byteorder == '='
else:
assert dtype.byteorder == byteorder
def test_can_turn_off_subarrays(dt):
for field, _ in dt.fields.values():
assert field.shape == ()
@pytest.mark.parametrize('byteorder', ['<', '>'])
@given(data=st.data())
def test_can_restrict_endianness(data, byteorder):
dtype = data.draw(nps.integer_dtypes(byteorder, sizes=(16, 32, 64)))
if byteorder == ('<' if sys.byteorder == 'little' else '>'):
assert dtype.byteorder == '='
else:
assert dtype.byteorder == byteorder
@given(nps.integer_dtypes(sizes=8))
def test_can_specify_size_as_an_int(dt):
assert dt.itemsize == 1
@given(st.data())
def test_can_draw_shapeless_from_scalars(data):
dt = data.draw(nps.scalar_dtypes())
result = data.draw(nps.arrays(dtype=dt, shape=()))
assert isinstance(result, dt.type)
@given(st.data())
def test_unicode_string_dtypes_generate_unicode_strings(data):
dt = data.draw(nps.unicode_string_dtypes())
result = data.draw(nps.from_dtype(dt))
from functools import partial
import numpy as np
import pandas as pd
import pytest
import xarray as xr
pytest.importorskip("hypothesis")
import hypothesis.extra.numpy as npst # isort:skip
import hypothesis.extra.pandas as pdst # isort:skip
import hypothesis.strategies as st # isort:skip
from hypothesis import given # isort:skip
numeric_dtypes = st.one_of(npst.unsigned_integer_dtypes(),
npst.integer_dtypes(), npst.floating_dtypes())
numeric_series = numeric_dtypes.flatmap(lambda dt: pdst.series(dtype=dt))
an_array = npst.arrays(
dtype=numeric_dtypes,
shape=npst.array_shapes(max_dims=2), # can only convert 1D/2D to pandas
)
@st.composite
def datasets_1d_vars(draw) -> xr.Dataset:
"""Generate datasets with only 1D variables
Suitable for converting to pandas dataframes.
"""
def test_can_turn_off_subarrays(dt):
for field, _ in dt.fields.values():
assert field.shape == ()
@pytest.mark.parametrize("byteorder", ["<", ">"])
@given(data=st.data())
def test_can_restrict_endianness(data, byteorder):
dtype = data.draw(nps.integer_dtypes(byteorder, sizes=(16, 32, 64)))
if byteorder == ("<" if sys.byteorder == "little" else ">"):
assert dtype.byteorder == "="
else:
assert dtype.byteorder == byteorder
@given(nps.integer_dtypes(sizes=8))
def test_can_specify_size_as_an_int(dt):
assert dt.itemsize == 1
@given(st.data())
def test_can_draw_arrays_from_scalars(data):
dt = data.draw(nps.scalar_dtypes())
result = data.draw(nps.arrays(dtype=dt, shape=()))
assert isinstance(result, np.ndarray)
assert result.dtype == dt
@given(st.data())
def test_can_cast_for_scalars(data):
# Note: this only passes with castable datatypes, certain dtype
def test_array_dtypes_may_have_field_titles():
find_any(nps.array_dtypes(), lambda dt: len(dt.fields) > len(dt.names))
@pytest.mark.parametrize("byteorder", ["<", ">"])
@given(data=st.data())
def test_can_restrict_endianness(data, byteorder):
dtype = data.draw(nps.integer_dtypes(endianness=byteorder, sizes=(16, 32, 64)))
if byteorder == ("<" if sys.byteorder == "little" else ">"):
assert dtype.byteorder == "="
else:
assert dtype.byteorder == byteorder
@given(nps.integer_dtypes(sizes=8))
def test_can_specify_size_as_an_int(dt):
assert dt.itemsize == 1
@given(st.data())
def test_can_draw_arrays_from_scalars(data):
dt = data.draw(nps.scalar_dtypes())
result = data.draw(nps.arrays(dtype=dt, shape=()))
assert isinstance(result, np.ndarray)
assert result.dtype == dt
@given(st.data())
def test_can_cast_for_arrays(data):
# Note: this only passes with castable datatypes, certain dtype
from hypothesis import given
from hypothesis.extra.numpy import from_dtype, integer_dtypes
from hypothesis.strategies import data, floats, integers
@given(floats(width=32))
def test_float32_exactly_representable(x):
clipped = np.dtype("float32").type(x)
if np.isnan(x):
assert np.isnan(clipped)
else:
assert x == float(clipped)
@given(floats(width=16))
def test_float16_exactly_representable(x):
clipped = np.dtype("float16").type(x)
if np.isnan(x):
assert np.isnan(clipped)
else:
assert x == float(clipped)
@given(data=data(), dtype=integer_dtypes())
def test_floor_ceil_lossless(data, dtype):
# Regression test for issue #1667; ceil converting numpy integers
# to float and back to int with loss of exact value.
x = data.draw(from_dtype(dtype))
assert data.draw(integers(x, x)) == x
def test_minimise_array_strategy():
smallest = minimal(nps.arrays(
nps.nested_dtypes(max_itemsize=settings.default.buffer_size // 3**3),
nps.array_shapes(max_dims=3, max_side=3)))
assert smallest.dtype == np.dtype(u'bool') and not smallest.any()
@given(nps.array_dtypes(allow_subarrays=False))
def test_can_turn_off_subarrays(dt):
for field, _ in dt.fields.values():
assert field.shape == ()
@given(nps.integer_dtypes(endianness='>'))
def test_can_restrict_endianness(dt):
if dt.itemsize == 1:
assert dt.byteorder == '|'
else:
assert dt.byteorder == '>'
@given(nps.integer_dtypes(sizes=8))
def test_can_specify_size_as_an_int(dt):
assert dt.itemsize == 1
@given(st.data())
def test_can_draw_shapeless_from_scalars(data):
dt = data.draw(nps.scalar_dtypes())
pytest.importorskip("hypothesis")
import hypothesis.extra.numpy as npst
import hypothesis.strategies as st
from hypothesis import given, settings
import xarray as xr
# Run for a while - arrays are a bigger search space than usual
settings.register_profile("ci", deadline=None)
settings.load_profile("ci")
an_array = npst.arrays(
dtype=st.one_of(
npst.unsigned_integer_dtypes(), npst.integer_dtypes(), npst.floating_dtypes()
),
shape=npst.array_shapes(max_side=3), # max_side specified for performance
)
@pytest.mark.slow
@given(st.data(), an_array)
def test_CFMask_coder_roundtrip(data, arr):
names = data.draw(
st.lists(st.text(), min_size=arr.ndim, max_size=arr.ndim, unique=True).map(
tuple
)
)
original = xr.Variable(names, arr)
coder = xr.coding.variables.CFMaskCoder()
from typing import Optional
import hypothesis.extra.numpy as hnp
import hypothesis.strategies as st
import numpy as np
import pytest
from hypothesis import given, settings
from numpy.testing import assert_array_equal
from mygrad import Tensor
real_types = (hnp.integer_dtypes() | hnp.unsigned_integer_dtypes()
| hnp.floating_dtypes())
@given(
tensor=st.tuples(
hnp.arrays(shape=hnp.array_shapes(), dtype=real_types),
st.booleans(),
).map(lambda x: Tensor(x[0], constant=x[1])),
dest_type=real_types,
constant=st.booleans() | st.none(),
)
def test_astype(tensor: Tensor, dest_type: type, constant: Optional[bool]):
tensor = tensor * 1 # give tensor a creator
new_tensor = tensor.astype(dest_type, constant=constant)
assert new_tensor.constant is (tensor.constant
if constant is None else constant)
assert tensor.creator is not None
assert new_tensor.creator is None
floats,
integers,
lists,
one_of,
text,
tuples,
)
from superintendent.distributed.dbqueue import DatabaseQueue
guaranteed_dtypes = one_of(
np_strategies.scalar_dtypes(),
np_strategies.unsigned_integer_dtypes(),
np_strategies.datetime64_dtypes(),
np_strategies.floating_dtypes(),
np_strategies.integer_dtypes(),
)
@composite
def dataframe(draw):
n_cols = draw(integers(min_value=1, max_value=20))
dtypes = draw(
lists(
one_of(
np_strategies.floating_dtypes(),
np_strategies.integer_dtypes(),
np_strategies.unicode_string_dtypes(),
),
min_size=n_cols,
max_size=n_cols,
dictionaries,
floats,
integers,
lists,
one_of,
recursive,
text,
)
import superintendent.prioritisation
from superintendent import SemiSupervisor
primitive_strategy = text() | integers() | floats(allow_nan=False) | booleans()
guaranteed_dtypes = (boolean_dtypes()
| integer_dtypes()
| floating_dtypes()
| unicode_string_dtypes())
container_strategy = dictionaries(
text(), primitive_strategy) | lists(primitive_strategy)
nested_strategy = recursive(
container_strategy,
lambda children: lists(children) | dictionaries(text(), children),
)
container_strategy = dictionaries(
text(), primitive_strategy) | lists(primitive_strategy)
nested_strategy = recursive(
with Tiff(filename, "w") as handle:
handle.write(img, method="tile")
with Tiff(filename) as handle:
data = handle[:]
assert np.all(img == data[:, :, :3])
with Tiff(filename, "w") as handle:
handle.write(img, method="scanline")
with Tiff(filename) as handle:
data = handle[:]
assert np.all(img == data[:, :, :3])
# scanline integer tests
@settings(buffer_size=11000000)
@given(data=hnp.arrays(dtype=st.one_of(hnp.integer_dtypes(endianness="="), hnp.unsigned_integer_dtypes(endianness="=")),
shape=hnp.array_shapes(min_dims=2, max_dims=2, min_side=10, max_side=50)))
def test_write_int_scanline(data, tmpdir_factory):
filename = str(tmpdir_factory.mktemp("write").join("int_img.tif"))
with Tiff(filename, "w") as handle:
handle.write(data, method="scanline")
with tifffile.TiffFile(filename) as handle:
img = handle.asarray()
np.testing.assert_array_equal(data, img)
@settings(buffer_size=11000000)
@given(data=hnp.arrays(dtype=st.one_of(hnp.integer_dtypes(endianness="="), hnp.unsigned_integer_dtypes(endianness="=")),
shape=hnp.array_shapes(min_dims=2, max_dims=2, min_side=10, max_side=50)))
def test_write_int_scanline_set_rows_per_strip(data, tmpdir_factory):
filename = str(tmpdir_factory.mktemp("write").join("int_img.tif"))