def test_split_into_nhot_view():
f0 = dt.Frame(A=["cat,dog,mouse", "mouse", None, "dog, cat"])
f1 = dt.split_into_nhot(f0[::-1, :])
f2 = dt.split_into_nhot(f0[3, :])
assert set(f1.names) == {"cat", "dog", "mouse"}
assert f1[:, ["cat", "dog", "mouse"]].to_list() == \
[[1, 0, 0, 1], [1, 0, 0, 1], [0, 0, 1, 1]]
assert set(f2.names) == {"cat", "dog"}
assert f2[:, ["cat", "dog"]].to_list() == [[1], [1]]
def test_split_into_nhot_sep():
f0 = dt.Frame(["a|b|c", "b|a", "a|c"])
f1 = dt.split_into_nhot(f0, sep="|")
assert set(f1.names) == {"a", "b", "c"}
fr = dt.Frame(a=[1, 1, 1], b=[1, 1, 0], c=[1, 0, 1])
assert set(f1.names) == set(fr.names)
assert f1.to_list() == fr[:, f1.names].to_list()
def test_split_into_nhot0():
f0 = dt.Frame([
"cat, dog, mouse, peacock, frog", "armadillo, fox, hedgehog",
"dog, fox, mouse, cat, peacock", "horse, raccoon, cat, frog, dog"
])
f1 = dt.split_into_nhot(f0)
f1.internal.check()
fr = dt.Frame({
"cat": [1, 0, 1, 1],
"dog": [1, 0, 1, 1],
"mouse": [1, 0, 1, 0],
"peacock": [1, 0, 1, 0],
"frog": [1, 0, 0, 1],
"armadillo": [0, 1, 0, 0],
"fox": [0, 1, 1, 0],
"hedgehog": [0, 1, 0, 0],
"horse": [0, 0, 0, 1],
"raccoon": [0, 0, 0, 1]
})
assert set(f1.names) == set(fr.names)
fr = fr[:, f1.names]
assert f1.names == fr.names
assert f1.stypes == (dt.stype.bool8, ) * f1.ncols
assert f1.shape == fr.shape
assert f1.to_list() == fr.to_list()
def test_split_into_nhot_long(seed, st):
random.seed(seed)
n = int(random.expovariate(0.0001) + 100)
col1 = [random.getrandbits(1) for _ in range(n)]
col2 = [random.getrandbits(1) for _ in range(n)]
col3 = [random.getrandbits(1) for _ in range(n)]
col4 = [0] * n
data = [
",".join(["liberty"] * col1[i] + ["equality"] * col2[i] +
["justice"] * col3[i]) for i in range(n)
]
# Introduce 1% of None's, making sure we preserve data at freedom_index
na_indices = random.sample(range(n), n // 100)
freedom_index = random.randint(0, n - 1)
for i in na_indices:
if i == freedom_index: continue
col1[i] = None
col2[i] = None
col3[i] = None
col4[i] = None
data[i] = None
col4[freedom_index] = 1
data[freedom_index] += ", freedom"
f0 = dt.Frame(data, stype=st)
assert f0.stypes == (st, )
assert f0.shape == (n, 1)
f1 = dt.split_into_nhot(f0)
assert f1.shape == (n, 4)
fr = dt.Frame(liberty=col1, equality=col2, justice=col3, freedom=col4)
assert set(f1.names) == set(fr.names)
f1 = f1[..., fr.names]
assert f1.to_list() == fr.to_list()
def test_split_into_nhot1():
f0 = dt.Frame([
" meow \n", "[ meow]", "['meow' ,purr]", '(\t"meow", \'purr\')',
"{purr}"
])
f1 = dt.split_into_nhot(f0)
f1.internal.check()
fr = dt.Frame(meow=[1, 1, 1, 1, 0], purr=[0, 0, 1, 1, 1])
assert set(f1.names) == set(fr.names)
fr = fr[..., f1.names]
assert f1.shape == fr.shape == (5, 2)
assert f1.stypes == (dt.stype.bool8, dt.stype.bool8)
assert f1.to_list() == fr.to_list()
def test_split_into_nhot_long(seed, st):
random.seed(seed)
n = int(random.expovariate(0.0001) + 100)
col1 = [random.getrandbits(1) for _ in range(n)]
col2 = [random.getrandbits(1) for _ in range(n)]
col3 = [random.getrandbits(1) for _ in range(n)]
col4 = [0] * n
data = [",".join(["liberty"] * col1[i] +
["equality"] * col2[i] +
["justice"] * col3[i]) for i in range(n)]
i = random.randint(0, n - 1)
col4[i] = 1
data[i] += ", freedom"
f0 = dt.Frame(data, stype=st)
assert f0.stypes == (st,)
assert f0.shape == (n, 1)
f1 = dt.split_into_nhot(f0)
assert f1.shape == (n, 4)
fr = dt.Frame(liberty=col1, equality=col2, justice=col3, freedom=col4)
assert set(f1.names) == set(fr.names)
f1 = f1[..., fr.names]
assert f1.to_list() == fr.to_list()
def test_split_into_nhot_bad():
f0 = dt.Frame([[1.25], ["foo"], ["bar"]])
with pytest.raises(ValueError) as e:
dt.split_into_nhot(f0)
assert ("Function split_into_nhot() may only be applied to a single-column "
"Frame of type string; got frame with 3 columns" == str(e.value))
with pytest.raises(TypeError) as e:
dt.split_into_nhot(f0[:, 0])
assert ("Function split_into_nhot() may only be applied to a single-column "
"Frame of type string; received a column of type float64" ==
str(e.value))
with pytest.raises(ValueError) as e:
dt.split_into_nhot(f0[:, 1], sep=",;-")
assert ("Parameter `sep` in split_into_nhot() must be a single character"
in str(e.value))
def test_split_into_nhot_quotes():
f0 = dt.split_into_nhot(dt.Frame(['foo, "bar, baz"']))
f1 = dt.split_into_nhot(dt.Frame(['foo, "bar, baz']))
assert set(f0.names) == {"foo", "bar, baz"}
assert set(f1.names) == {"foo", '"bar', "baz"}
def test_split_into_nhot_empty():
f0 = dt.split_into_nhot(dt.Frame(["", None]))
assert_equals(f0, dt.Frame())
def test_split_into_nhot_none():
f0 = dt.split_into_nhot(None)
assert (f0 == None)
def test_split_into_nhot_noarg():
with pytest.raises(ValueError) as e:
noop(dt.split_into_nhot())
assert ("Required parameter `frame` is missing" == str(e.value))
def test_split_into_nhot_deprecated():
DT = dt.Frame(["a, b, c"])
with pytest.warns(FutureWarning):
RES = dt.split_into_nhot(DT)
EXP = dt.Frame([[1], [1], [1]], names=["a", "b", "c"], stype=dt.bool8)
assert_equals(RES, EXP)
