def test_feature_with_nan():
df = pd.DataFrame({
"y": [-1, 1, 1],
"x1": [1, 2, None],
"x2": [3, None, 2]
})
conv = DFtoVW(df=df, features=[Feature("x1"), Feature("x2")], label=SimpleLabel("y"))
lines = conv.convert_df()
assert lines == ['-1 | x1:1.0 x2:3.0', '1 | x1:2.0 ', '1 | x2:2.0']
def test_as_type_in_features():
df = pd.DataFrame({"y": [1], "a": [2], "b": [3], "c": ["4"]})
features = [
Feature("a", as_type="categorical"),
Feature("b"),
Feature("c", as_type="numerical"),
]
conv = DFtoVW(label=SimpleLabel("y"), features=features, df=df)
first_line = conv.convert_df()[0]
assert first_line == "1 | a=2 b:3 c:4"
def test_multiple_named_namespaces_multiple_features_multiple_lines():
df = pd.DataFrame({"y": [1, -1], "a": [2, 3], "b": ["x1", "x2"], "c": [36.4, 47.8]})
ns1 = Namespace(name="FirstNameSpace", features=Feature("a"))
ns2 = Namespace(name="DoubleIt", value=2, features=[Feature("b"), Feature("c")])
label = SimpleLabel("y")
conv = DFtoVW(df=df, label=label, namespaces=[ns1, ns2])
lines_list = conv.convert_df()
assert lines_list == [
"1 |FirstNameSpace a:2 |DoubleIt:2 b=x1 c:36.4",
"-1 |FirstNameSpace a:3 |DoubleIt:2 b=x2 c:47.8",
]
def test_multiple_named_namespaces():
df = pd.DataFrame({"y": [1], "a": [2], "b": [3]})
conv = DFtoVW(
df=df,
label=SimpleLabel("y"),
namespaces=[
Namespace(name="FirstNameSpace", features=Feature("a")),
Namespace(name="DoubleIt", value=2, features=Feature("b")),
],
)
first_line = conv.convert_df()[0]
assert first_line == "1 |FirstNameSpace a:2 |DoubleIt:2 b:3"
def test_mixed_type_features():
df = pd.DataFrame({"y": [1], "x1": ["a"], "x2": [2]})
conv = DFtoVW(
label=SimpleLabel("y"), features=[Feature(value=colname) for colname in ["x1", "x2"]], df=df
)
first_line = conv.convert_df()[0]
assert first_line == "1 | x1=a x2:2"
def test_multiclasslabel():
df = pd.DataFrame({"a": [1], "b": [0.5], "c": [-3]})
conv = DFtoVW(df=df,
label=MulticlassLabel(label="a", weight="b"),
features=Feature("c"))
first_line = conv.convert_df()[0]
assert first_line == "1 0.5 | c:-3"
def test_dirty_colname_feature():
df = pd.DataFrame(
{"target :": [1], " my first feature:": ["x"], "white space at the end ": [2]}
)
features = [Feature(colname) for colname in [" my first feature:", "white space at the end "]]
conv = DFtoVW(label=SimpleLabel("target :"), features=features, df=df)
first_line = conv.convert_df()[0]
assert first_line == "1 | my_first_feature=x white_space_at_the_end:2"
def test_multiple_lines_with_weight():
df = pd.DataFrame({
"y": [1, 2, -1],
"w": [2.5, 1.2, 3.75],
"x": ["a", "b", "c"]
})
conv = DFtoVW(df=df, label=SimpleLabel(label="y", weight="w"), features=Feature("x"))
lines_list = conv.convert_df()
assert lines_list == ['1 2.5 | x=a', '2 1.2 | x=b', '-1 3.75 | x=c']
def test_multiclasslabel_negative_weight_error():
df = pd.DataFrame({"y": [1], "w": [-0.5], "x": [2]})
with pytest.raises(ValueError) as value_error:
DFtoVW(
df=df,
label=MulticlassLabel(label="y", weight="w"),
features=Feature("x"),
)
expected = "In argument 'weight' of 'MulticlassLabel', column 'w' must be >= 0."
assert expected == str(value_error.value)
def test_multiclasslabel_non_positive_label_error():
df = pd.DataFrame({"a": [0], "b": [0.5], "c": ["x"]})
with pytest.raises(ValueError) as value_error:
DFtoVW(
df=df,
label=MulticlassLabel(label="a", weight="b"),
features=Feature("c"),
)
expected = "In argument 'label' of 'MulticlassLabel', column 'a' must be >= 1."
assert expected == str(value_error.value)
def test_absent_col_error():
with pytest.raises(ValueError) as value_error:
df = pd.DataFrame({"a": [1]})
DFtoVW(
df=df,
label=SimpleLabel("a"),
features=[Feature(col) for col in ["a", "c", "d"]],
)
expected = "In 'Feature': column(s) 'c', 'd' not found in dataframe."
assert expected == str(value_error.value)
def test_multilabel_non_positive_label_error():
df = pd.DataFrame({"y": [0], "b": [1]})
with pytest.raises(ValueError) as value_error:
DFtoVW(
df=df,
label=MultiLabel(label="y"),
features=Feature("b"),
)
expected = "In argument 'label' of 'MultiLabel', column 'y' must be >= 1."
assert expected == str(value_error.value)
def test_feature_column_renaming_and_tag():
df = pd.DataFrame({"idx": ["id_1"], "y": [1], "x": [2]})
conv = DFtoVW(
label=SimpleLabel("y"),
tag="idx",
features=Feature(value="x", rename_feature="col_x"),
df=df,
)
first_line = conv.convert_df()[0]
assert first_line == "1 id_1| col_x:2"
def test_contextualbanditlabel_non_positive_action():
df = pd.DataFrame({"a": [0], "c": [-0.5], "p": [0.5], "x": [1]})
with pytest.raises(ValueError) as value_error:
DFtoVW(
df=df,
label=ContextualbanditLabel("a", "c", "p"),
features=Feature("x"),
)
expected = "In argument 'action' of 'ContextualbanditLabel', column 'a' must be >= 1."
assert expected == str(value_error.value)
def test_listlabel_mixed_label_error():
with pytest.raises(TypeError) as type_error:
df = pd.DataFrame({"a": [1], "c": [-0.5], "p": [0.1], "b": [2], "x": [1]})
conv = DFtoVW(
df=df,
label=[ContextualbanditLabel("a", "c", "p"), MultiLabel("b")],
features=Feature("x"),
)
expected = "The list passed in 'label' has mixed label types."
assert expected == str(type_error.value)
def test_contextualbanditlabel_non_float_proba_error():
df = pd.DataFrame({"a": [1], "c": [-0.5], "p": [1], "x": [1]})
with pytest.raises(TypeError) as value_error:
DFtoVW(
df=df,
label=ContextualbanditLabel("a", "c", "p"),
features=Feature("x"),
)
expected = "In argument 'probability' of 'ContextualbanditLabel', column 'p' should be either of the following type(s): 'float'."
assert expected == str(value_error.value)
def test_contextualbanditlabel_multiple_label():
df = pd.DataFrame(
{"a1": [1], "c1": [-0.5], "p1": [0.1], "a2": [2], "c2": [-1.5], "p2": [0.6], "x": [1]}
)
conv = DFtoVW(
df=df,
label=[ContextualbanditLabel("a1", "c1", "p1"), ContextualbanditLabel("a2", "c2", "p2")],
features=Feature("x"),
)
first_line = conv.convert_df()[0]
assert first_line == "1:-0.5:0.1 2:-1.5:0.6 | x:1"
def test_listlabel_not_allowed_label_error():
with pytest.raises(TypeError) as type_error:
df = pd.DataFrame({"a": [1], "b": [2], "x": [1]})
conv = DFtoVW(
df=df,
label=[SimpleLabel("a"), SimpleLabel("b")],
features=Feature("x"),
)
expected = (
"The only labels that can be used with list are 'ContextualbanditLabel', 'MultiLabel'."
)
assert expected == str(type_error.value)
def test_contextualbanditlabel_negative_proba_error():
df = pd.DataFrame({"a": [1], "c": [-0.5], "p": [-0.1], "x": [1]})
with pytest.raises(ValueError) as value_error:
DFtoVW(
df=df,
label=ContextualbanditLabel("a", "c", "p"),
features=Feature("x"),
)
expected = (
"In argument 'probability' of 'ContextualbanditLabel', column 'p' must be >= 0 and <= 1."
)
assert expected == str(value_error.value)
def test_multiple_lines():
df = pd.DataFrame({"y": [1, -1], "x": [1, 2]})
conv = DFtoVW(label=SimpleLabel("y"), features=Feature(value="x"), df=df)
lines_list = conv.convert_df()
assert lines_list == ["1 | x:1", "-1 | x:2"]
def test_namespace_with_value_but_no_name():
with pytest.raises(ValueError) as value_error:
Namespace(features=Feature("a"), value=10)
expected = "Namespace can't have a 'value' argument without a 'name' argument or an empty string 'name' argument"
assert expected == str(value_error.value)
def test_no_label_rename_feature():
df = pd.DataFrame({"a": [1]})
first_line = DFtoVW(df=df, features=Feature("a", rename_feature="")).convert_df()[0]
assert first_line == "| :1"
def test_wrong_weight_type_error():
df = pd.DataFrame({"y": [1], "x": [2], "w": ["a"]})
with pytest.raises(TypeError) as type_error:
DFtoVW(df=df, label=SimpleLabel(label="y", weight="w"), features=Feature("x"))
expected = "In argument 'weight' of 'SimpleLabel', column 'w' should be either of the following type(s): 'int', 'float'."
assert expected == str(type_error.value)
def test_contextualbanditlabel_one_label():
df = pd.DataFrame({"a": [1], "c": [-0.5], "p": [0.1], "x": [1]})
conv = DFtoVW(df=df, label=ContextualbanditLabel("a", "c", "p"), features=Feature("x"))
first_line = conv.convert_df()[0]
assert first_line == "1:-0.5:0.1 | x:1"
def test_non_default_index():
df = pd.DataFrame({"y": [0], "x": [1]}, index=[1])
conv = DFtoVW(df=df, label=SimpleLabel("y"), features=Feature("x"))
first_line = conv.convert_df()[0]
assert first_line == "0 | x:1"
def test_multilabel_list_of_len_1():
df = pd.DataFrame({"y": [1], "x": [2]})
conv1 = DFtoVW(df=df, label=MultiLabel(["y"]), features=Feature("x"))
conv2 = DFtoVW(df=df, label=MultiLabel("y"), features=Feature("x"))
assert conv1.convert_df()[0] == conv2.convert_df()[0]
def test_multilabel_with_listlabel_builder():
df = pd.DataFrame({"y1": [1], "y2": [2], "x": [3]})
conv = DFtoVW(df=df, label=[MultiLabel("y1"), MultiLabel("y2")], features=Feature("x"))
first_line = conv.convert_df()[0]
assert first_line == "1,2 | x:3"
def test_multilabel():
df = pd.DataFrame({"y1": [1], "y2": [2], "x": [3]})
conv = DFtoVW(df=df, label=MultiLabel(["y1", "y2"]), features=Feature("x"))
first_line = conv.convert_df()[0]
assert first_line == "1,2 | x:3"
def test_no_label_multiple_features():
df = pd.DataFrame({"a": [2], "b": [3]})
conv = DFtoVW(df=df, features=[Feature(col) for col in ["a", "b"]])
first_line = conv.convert_df()[0]
assert first_line == "| a:2 b:3"
def test_non_numerical_error():
df = pd.DataFrame({"y": ["a"], "x": ["featX"]})
with pytest.raises(TypeError) as type_error:
DFtoVW(df=df, label=SimpleLabel(label="y"), features=Feature("x"))
expected = "In argument 'label' of 'SimpleLabel', column 'y' should be either of the following type(s): 'int', 'float'."
assert expected == str(type_error.value)
