def test_grouping_is_preserved():
# In siuba this actually tests 3 things:
# 1) Can we pipe a grouped DataFrame in to `pivot_longer`? - Not yet
# 2) Does it retain grouping? - Not yet
# 3) Can we get the names of grouping _columns_ ie variables? - Not yet?
df = data_frame(g = [1, 2], x1 = [1, 2], x2 = [3, 4])
# Breaks; `pivot_longer` needs singledispatch for grouped DataFrames
out = (
df
>> group_by(_.g)
>> pivot_longer(_["x1":"x2"], names_to = "x", values_to = "v")
)
# Breaks, as group_vars does not exist yet.
# For now, in pandas it is probably better to check if the DataFrame remains
# grouped, and if it matches the expected output
# assert group_vars(out) == "g"
expected = data_frame(
g = [1, 1, 2, 2],
x = ["x1", "x2", "x1", "x2"],
v = [1, 3, 2, 4],
_index = [0, 0, 1, 1]
).groupby("g")
# assert_frame_equal does not work with DataFrameGroupBy.
isinstance(out, expected.__class__)
assert_frame_equal(out.obj, expected.obj)
def test_spec_add_multi_columns():
df = data_frame(x = [1,2], y = [3,4])
# TODO: is this the right format for a spec
sp = data_frame(_name = ["x", "y"], _value = "v", a = 1, b = 2)
pv = pivot_longer_spec(df, spec = sp)
assert pv.columns.tolist() == ["a", "b", "v"]
def test_pivot_all_cols_to_long():
"can pivot all cols to long"
src = data_frame(x = [1,2], y = [3,4])
dst = data_frame(name = ["x", "y", "x", "y"], value = [1, 3, 2, 4])
res = pivot_longer(src, _["x":"y"])
assert_frame_equal(res.reset_index(drop=True), dst)
def test_spec_can_pivot_to_multi_measure_cols():
df = data_frame(x = "x", y = 1)
sp = data_frame(_name = ["x", "y"], _value = ["X", "Y"], row = [1, 1])
pv = pivot_longer_spec(df, sp)
assert pv.columns.tolist() == ["row", "X", "Y"]
assert pv["X"] == "x"
assert pv["Y"] == 1
def test_original_col_order_is_preserved():
df = data_frame(id = ["A", "B"],
z_1 = [1, 7], y_1 = [2, 8], x_1 = [3, 9],
z_2 = [4, 10], y_2 = [5, 11], x_2 = [6, 12]
)
pv = pivot_longer(df, -_.id, names_to = ("_value", "n"), names_sep = "_")
assert pv.columns.tolist() == ["id", "n", "z", "y", "x"]
def test_preserves_original_keys():
df = data_frame(x = [1,2], y = [2,2], z = [1,2])
pv = pivot_longer(df, _["y":"z"])
assert pv.columns.tolist() == ["x", "name", "value"]
assert_series_equal(
pv["x"],
pd.Series(df["x"].repeat(2))
)
def test_can_pivot_duplicate_names_to_value():
df = data_frame(x = 1, a_1 = 1, a_2 = 2, b_1 = 3, b_2 = 4)
pv1 = pivot_longer(df, -_.x, names_to = ("_value", np.nan), names_sep = "_")
pv2 = pivot_longer(df, -_.x, names_to = ("_value", np.nan), names_pattern = "(.)_(.)")
pv3 = pivot_longer(df, -_.x, names_to = "_value", names_pattern = "(.)_.")
assert pv1.columns.tolist() == ["x", "a", "b"]
assert pv1["a"].tolist() == [1, 2]
assert_frame_equal(pv2, pv1)
assert_frame_equal(pv3, pv1)
def test_can_handle_missing_combinations():
df = data_frame(id = ["A", "B"], x_1 = [1, 3], x_2 = [2, 4], y_2 = ["a", "b"])
pv = pivot_longer(df, -_.id, names_to = ("_value", "n"), names_sep = "_")
pv_expected = pd.Series([np.nan, "a", np.nan, "b"],
index = [0, 0, 1, 1],
name = 'y')
assert pv.columns.tolist() == ["id", "n", "x", "y"]
assert pv["x"].tolist() == [1, 2, 3, 4]
pd.testing.assert_series_equal(pv["y"], pv_expected)
def test_type_error_message_uses_var_names():
# Error handling is by default 'better' in python than R
# This test is tricky, as python doesn't care when stacking data of different types.
df = data_frame(abc = 1, xyz = "b")
try:
# This should by default pivot everything, as with tidyr
pivot_longer(df, _[:])
except:
# Ideally we'd print an error message here and compare if the keys are
# printed correctly, but `pivot_longer` doesn't brake in python as in R
# when stacking different data types. Not sure if we should 'make' it brake?
print(err)
def test_value_can_be_any_pos_in_names_to():
samp = data_frame(
i = np.arange(1, 5),
y_t1 = np.random.standard_normal(4),
y_t2 = np.random.standard_normal(4),
z_t1 = [3] * 4,
z_t2 = [-2] * 4,
)
value_first = pivot_longer(samp, -_.i,
names_to = ("_value", "time"), names_sep = "_")
samp2 = samp.rename(columns={"y_t1": "t1_y", "y_t2": "t2_y",
"z_t1": "t1_z", "z_t2": "t2_z"})
value_second = pivot_longer(samp2, -_.i,
names_to = ("time", "_value"), names_sep = "_")
assert_frame_equal(value_first, value_second)
import pytest
from siuba.tests.helpers import data_frame
import pandas as pd
from siuba.experimental.pd_groups.translate import method_agg_op, method_el_op, method_el_op2
from siuba.experimental.pd_groups.groupby import broadcast_agg
#TODO:
# - what if they have mandatory, non-data args?
# - support accessor methods like _.x.str.upper()
# - support .expanding and .rolling
data_dt = data_frame(g=['a', 'a', 'b', 'b'],
x=pd.to_datetime([
"2019-01-01 01:01:01", "2020-04-08 02:02:02",
"2021-07-15 03:03:03", "2022-10-22 04:04:04"
]))
data_str = data_frame(g=['a', 'a', 'b', 'b'], x=['abc', 'cde', 'fg', 'h'])
data_default = data_frame(g=['a', 'a', 'b', 'b'],
x=[10, 11, 12, 13],
y=[1, 2, 3, 4])
data = {'dt': data_dt, 'str': data_str, None: data_default}
# Test translator =============================================================
from pandas.testing import assert_frame_equal, assert_series_equal
from siuba.experimental.pd_groups.groupby import GroupByAgg, SeriesGroupBy
f_min = method_agg_op('min', is_property=False, accessor=None)
def test_can_override_default_output_col_type():
df = data_frame(x = "x", y = 1)
pv = pivot_longer(df, _["x":"y"], values_transform = {"value": list})
assert pv["value"].tolist() == [["x"], [1]]
def test_mixed_columns_are_auto_coerced():
# TODO: pandas stack (and melt) coerces categorical data when stacking.
df = data_frame(x = pd.Categorical(["a"]), y = pd.Categorical(["b"]))
pv = pivot_longer(df, _["x":"y"])
assert_series_equal(pv["value"], pd.Categorical(['a', 'b']))
def test_can_drop_missing_values():
df = data_frame(x = [1, np.nan], y = [np.nan, 2])
pv = pivot_longer(df, _["x":"y"], values_drop_na=True)
assert pv["name"].tolist() == ["x", "y"]
assert pv["value"].tolist() == [1, 2]
def test_values_interleaved_correctly():
# TODO: fix order issue
df = data_frame(x = [1,2], y = [10, 20], z = [100, 200])
pv = pivot_longer(df, _[0:3])
assert pv["value"].tolist() == [1, 10, 100, 2, 20, 200]
