def test_ifelse_wrong_type():
DT = dt.Frame(A=range(10))
DT["B"] = dt.str32(f.A)
msg = r"The condition argument in ifelse\(\) must be a boolean column"
with pytest.raises(TypeError, match=msg):
DT[:, ifelse(f.A, f.A, f.A)]
with pytest.raises(TypeError, match=msg):
DT[:, ifelse(f.B, f.A, f.A)]
def test_ifelse_with_groupby():
DT = dt.Frame(A=[2, 5, 2, 5, 2, 2], B=range(6))
R1 = DT[:, ifelse(f.A == 2, dt.min(f.B), dt.max(f.B)), by(f.A)]
R2 = DT[:, ifelse(f.A == 2, f.B, dt.max(f.B)), by(f.A)]
R3 = DT[:, ifelse(f.A == 2, dt.min(f.B), f.B), by(f.A)]
R4 = DT[:, ifelse(f.B > 2, dt.min(f.B), f.B), by(f.A)]
assert_equals(R1, dt.Frame(A=[2, 5], C0=[0, 3]))
assert_equals(R2, dt.Frame(A=[2, 2, 2, 2, 5, 5], C0=[0, 2, 4, 5, 3, 3]))
assert_equals(R3, dt.Frame(A=[2, 2, 2, 2, 5, 5], C0=[0, 0, 0, 0, 1, 3]))
assert_equals(R4, dt.Frame(A=[2, 2, 2, 2, 5, 5], C0=[0, 2, 0, 0, 1, 1]))
def test_ifelse_columnsets():
DT = dt.Frame(A=range(10), B=[7] * 10, C=list('abcdefghij'))
msg = r"Multi-column expressions are not supported in ifelse\(\) function"
with pytest.raises(TypeError, match=msg):
DT[:, ifelse(f[:], 0, 1)]
with pytest.raises(TypeError, match=msg):
DT[:, ifelse(f.A > 3, f[:], f.A)]
with pytest.raises(TypeError, match=msg):
DT[:, ifelse(f.A > 3, f.A, f[:])]
# We could in theory make this work...
with pytest.raises(TypeError, match=msg):
DT[:, ifelse(f[int] > 3, 3, f[int])]
def test_ifelse_bad_signature():
DT = dt.Frame(A=range(10))
msg = r"Function datatable\.ifelse\(\) requires at least 3 arguments"
with pytest.raises(TypeError, match=msg):
DT[:, ifelse()]
with pytest.raises(TypeError, match=msg):
DT[:, ifelse(f.A > 0)]
with pytest.raises(TypeError, match=msg):
DT[:, ifelse(f.A > 0, f.A)]
msg = r"Missing the required default argument in function " \
r"datatable\.ifelse\(\)"
with pytest.raises(TypeError, match=msg):
DT[:, ifelse(f.A > 0, f.A, f.A, f.A)]
def test_ifelse_bad_signature():
DT = dt.Frame(A=range(10))
msg = r"Function ifelse\(\) requires 3 arguments"
with pytest.raises(TypeError, match=msg):
DT[:, ifelse()]
with pytest.raises(TypeError, match=msg):
DT[:, ifelse(f.A > 0)]
with pytest.raises(TypeError, match=msg):
DT[:, ifelse(f.A > 0, f.A)]
msg = r"ifelse\(\) takes at most 3 positional arguments, " \
r"but 4 were given"
with pytest.raises(TypeError, match=msg):
DT[:, ifelse(f.A > 0, f.A, f.A, f.A)]
def test_countries():
# inspired by @knapply's example
countries = dt.Frame(
name=["Czech Republic", "Czecho-Slovakia", "Mexico", "Czech Republic",
"Canada", "Czechoslovakia", "USA", "Britain"],
year=[1918] + list(range(1990, 1997))
)
def is_czech_name(x):
return ((x == "Czechoslovak Republic") |
(x == "Czechoslovakia") |
(x == "Czech Republic") |
(x == "Czecho-Slovakia"))
name, year = countries.export_names()
RES = countries[:, {"historic_name": ifelse(
is_czech_name(name) & (year <= 1938), "Czechoslovak Republic",
is_czech_name(name) & (year <= 1992), "Czechoslovakia",
is_czech_name(name) & (year >= 1993), "Czech Republic",
(name == "USA"), "United States of America",
(name == "Britain"), "United Kingdom",
name
)}]
assert_equals(RES,
dt.Frame(historic_name=["Czechoslovak Republic", "Czechoslovakia",
"Mexico", "Czechoslovakia", "Canada", "Czech Republic",
"United States of America", "United Kingdom"]))
def test_ifelse_multi_different_grouplevels():
DT = dt.Frame(A=[1, 2, 3, 4, 5, 6])
RES = DT[:, ifelse(f.A <= 2, dt.min(f.A),
f.A >= 5, dt.max(f.A),
f.A == 0, 1000000,
f.A)]
assert_equals(RES, dt.Frame([1, 1, 3, 4, 6, 6]))
def addRiskColumn(table, incidencePrognosisColumn, newColumn, darkFactor):
incidence = dt.f[incidencePrognosisColumn]
newTable = table[:, dt.f[:].extend({
newColumn:
dt.ifelse(incidence <= 0, 99999, 100000 / (incidence * darkFactor))
})]
#print(newTable)
return newTable
def test_ifelse_columnsets():
DT = dt.Frame(A=range(10), B=[7]*10, C=list('abcdefghij'))
msg = r"The condition1 argument in ifelse\(\) cannot be a multi-column " \
r"expression"
with pytest.raises(TypeError, match=msg):
DT[:, ifelse(f[:], 0, 1)]
with pytest.raises(TypeError, match=msg):
DT[:, ifelse(f[int] > 3, 3, f[int])]
msg = r"The value1 argument in ifelse\(\) cannot be a multi-column " \
r"expression"
with pytest.raises(TypeError, match=msg):
DT[:, ifelse(f.A > 3, f[:], f.A)]
msg = r"The value2 argument in ifelse\(\) cannot be a multi-column " \
r"expression"
with pytest.raises(TypeError, match=msg):
DT[:, ifelse(f.A > 3, f.A, f[:])]
def arreglar_infinitos(dataset: Frame) -> Frame:
for column in dataset.names:
if column != 'clase_ternaria':
dataset[column] = dataset[:,
ifelse(
math.isinf(f[column]) ==
1, None, f[column])]
return dataset
def test_example():
DT = dt.Frame(domestic_income=[4500, 2500, 1500, 4000],
internationaL_income=[2000, 5000, 1000, 4500])
DT["profit_loss"] = ifelse(f.domestic_income > f.internationaL_income,
"profit", "loss")
assert_equals(
DT,
dt.Frame(domestic_income=[4500, 2500, 1500, 4000],
internationaL_income=[2000, 5000, 1000, 4500],
profit_loss=["profit", "loss", "profit", "loss"]))
def calcular_ganancia(data: Frame, votos: int) -> float:
datos_votos = data[:, 'votos'].to_numpy().flatten() >= votos
ganancias = data[:, ifelse(f.clase_ternaria ==
'BAJA+2', 29250, -750)].to_numpy().flatten()
return np.dot(datos_votos, ganancias)
def test_condition_with_NAs():
DT = dt.Frame(A=[True, False, None], B=[5, 7, 9])
RES = DT[:, ifelse(f.A, f.B, -f.B)]
assert_equals(RES, dt.Frame([5, -7, None]))
def test_different_stypes():
DT = dt.Frame(A=[3], B=[7.1])
RES = DT[:, ifelse(f.A > 0, f.A, f.B)]
assert_equals(RES, dt.Frame([3.0]))
def test_ifelse_with_scalars():
DT = dt.Frame(A=range(10))
RES = DT[:, ifelse(f.A % 2 == 0, "even", "odd")]
assert_equals(RES, dt.Frame(["even", "odd"] * 5))
def test_ifelse_simple():
DT = dt.Frame(A=range(10))
DT["B"] = ifelse(f.A > 5, f.A - 5, f.A + 5)
assert_equals(DT, dt.Frame(A=range(10), B=[5, 6, 7, 8, 9, 10, 1, 2, 3, 4]))
"""Transform dataset features"""
# Map and create new features by adding new columns or with in-place update.
# For example, use for mapping multi-valued key to single column or
# any other types of map (row by row) transformations.
#
# Specification:
# Inputs:
# X: datatable - primary dataset
# Parameters:
# transformations: map - map with datatable transformation in the form of key: value pairs where key
# is new / existing column name and value is datatable expression for this column.
# Output:
# dataset containing original and transformed features
from datatable import f, isna, ifelse
transformations = {'title_with_type': f['primaryTitle'] + '-' + f['titleType'], # concatentate 2 columns
'startYear': ifelse(f['startYear']=='\\N', None, f['startYear']), # override empty value with NULL
'endYear': ifelse(f['endYear']=='\\N', None, f['endYear']), # override empty value with NULL in another column
'spanYears': ifelse((f['startYear']=='\\N') | (f['endYear']=='\\N'),
0, dt.int32(f['endYear']) - dt.int32(f['startYear'])) # compute the different between two columns
}
X[:, dt.update(**transformations)]
return {"temp_to_delte": X}
np.random.seed(TRAIN_PARAMS['seed'])
if __name__ == '__main__':
args = parser.parse_args()
TRAIN_PARAMS['binaria_especial'] = args.binaria_especial
TRAIN_PARAMS['model'] = args.model
TRAIN_PARAMS['file_data'] = f'../datasets/datos_fe_hist_v{args.version}.gz'
dataset_original = fread(TRAIN_PARAMS['file_data'])
dataset = dataset_original[f.foto_mes <= TRAIN_PARAMS['max_foto_mes_entero'], :]
dataset['azar'] = np.random.uniform(size=dataset.shape[0])
dataset['clase01'] = dataset[:, ifelse(f.clase_ternaria == 'CONTINUA', 0, 1)]
if args.binaria_especial:
dataset['target'] = dataset[:, f.clase_ternaria != 'CONTINUA']
dataset['weight'] = dataset[:, ifelse(f.clase_ternaria == 'BAJA+2', 1.0000001, 1)]
campos_buenos = f[:].remove([f.clase_ternaria, f.target, f.azar, f.clase01, f.weight])
else:
dataset['target'] = dataset[:, f.clase_ternaria == 'BAJA+2']
campos_buenos = f[:].remove([f.clase_ternaria, f.target, f.azar, f.clase01])
X = dataset[(f.foto_mes <= TRAIN_PARAMS['max_foto_mes_train']) & ((f.clase01 == 1) | (f.azar < 0.1)), campos_buenos]
y = dataset[(f.foto_mes <= TRAIN_PARAMS['max_foto_mes_train']) & ((f.clase01 == 1) | (f.azar < 0.1)), f.target]
weights = None
if args.binaria_especial:
weights = dataset[(f.foto_mes <= TRAIN_PARAMS['max_foto_mes_train']) & ((f.clase01 == 1) | (f.azar < 0.1)), f.weight]
# any other types of map (row by row) transformations.
#
# Specification:
# Inputs:
# X: datatable - primary dataset
# Parameters:
# transformations: map - map with datatable transformation in the form of key: value pairs where key
# is new / existing column name and value is datatable expression for this column.
# Output:
# dataset containing original and transformed features
from datatable import f, isna, ifelse
transformations = {
'title_with_type':
f['primaryTitle'] + '-' + f['titleType'], # concatentate 2 columns
'startYear':
ifelse(f['startYear'] == '\\N', None,
f['startYear']), # override empty value with NULL
'endYear':
ifelse(f['endYear'] == '\\N', None, f['endYear']),
# override empty value with NULL in another column
'spanYears':
ifelse((f['startYear'] == '\\N') | (f['endYear'] == '\\N'), 0,
dt.int32(f['endYear']) - dt.int32(f['startYear']))
# compute the different between two columns
}
X[:, dt.update(**transformations)]
return {"temp_to_delte": X}
TRAIN_PARAMS['file_data'] = f'../datasets/datos_fe_hist_v{args.version}.gz'
dataset = fread(TRAIN_PARAMS['file_data'])
for experimento in args.experimentos:
for file in os.listdir(f'../experimentos/{experimento}/'):
if file.endswith('stacking_apply.csv'):
stacking = fread(f'../experimentos/{experimento}/{file}')
dataset[f'{experimento}_prob'] = stacking['prob']
break
dapply_kaggle = dataset[f.foto_mes == TRAIN_PARAMS['foto_mes_kaggle'],
f[:].remove([f.clase_ternaria])]
dataset = dataset[f.foto_mes <= TRAIN_PARAMS['max_foto_mes_entero'], :]
dataset['azar'] = np.random.uniform(size=dataset.shape[0])
dataset['clase01'] = dataset[:,
ifelse(f.clase_ternaria == 'CONTINUA', 0, 1)]
if args.binaria_especial:
dataset['target'] = dataset[:, f.clase_ternaria != 'CONTINUA']
dataset['weight'] = dataset[:,
ifelse(f.clase_ternaria ==
'BAJA+2', 1.0000001, 1)]
campos_buenos = f[:].remove(
[f.clase_ternaria, f.target, f.azar, f.clase01, f.weight])
else:
dataset['target'] = dataset[:, f.clase_ternaria == 'BAJA+2']
campos_buenos = f[:].remove(
[f.clase_ternaria, f.target, f.azar, f.clase01])
X = dataset[(f.foto_mes <= TRAIN_PARAMS['max_foto_mes_train']) &
((f.clase01 == 1) | (f.azar < 0.1)), campos_buenos]
def test_ifelse_multi():
DT = dt.Frame(A=['fox', 'cat', 'jay', 'cow'])
RES = DT[:, ifelse(f.A == 'fox', 3,
f.A == 'dog', 7,
f.A == 'cow', 2, -1)]
assert_equals(RES, dt.Frame([3, -1, -1, 2]))
