def test_group_split_bind_rows_round_trip():
iris["Species"] = iris["Species"].astype("category")
setosa = iris >> filter(f.Species == "setosa")
chunks = setosa >> group_split.list(f.Species)
assert len(chunks) == 1
assert bind_rows(chunks).equals(setosa)
chunks = setosa >> group_split.list(f.Species, _drop=False)
assert len(chunks) == 3
assert_frame_equal(chunks[0], setosa)
def test_group_list_respects_empty_groups():
tbl = tibble(
x=[1, 2, 3, 4],
g=factor(rep(["a", "b"], each=2), levels=["a", "b", "c"]),
)
res = group_split.list(tbl, f.g)
assert res[0].equals(tbl.iloc[:2, :])
assert res[1].equals(tbl.iloc[[2, 3], :].reset_index(drop=True))
res = group_split.list(tbl, f.g, _drop=False)
assert res[0].equals(tbl.iloc[:2, :])
assert res[1].equals(tbl.iloc[[2, 3], :].reset_index(drop=True))
assert res[2].equals(tbl.iloc[[], :])
def test_group_split_on_a_rowwise_df_returns_a_list_of_tibbles():
df = tibble(x=[1, 2])
rdf = rowwise(df)
out = group_split.list(rdf)
assert len(out) == 2
assert out[0].equals(df.iloc[[0], :])
assert out[1].equals(df.iloc[[1], :].reset_index(drop=True))
def test_group_split_on_a_grouped_df_returns_a_list_of_tibbles():
df = tibble(x=[1, 2])
gdf = group_by(df, f.x)
out = group_split.list(gdf)
assert len(out) == 2
assert out[0].equals(df.iloc[[0], :])
assert out[1].equals(df.iloc[[1], :].reset_index(drop=True))
def test_group_split_works_with_subclasses_implementing_group_by_ungroup():
# test_that("group_split() works with subclasses implementing group_by()
# / ungroup()", {
class TibbleGrouped1(TibbleGrouped):
...
df = TibbleGrouped1.from_groupby(DataFrame(dict(x=[1, 2, 2])).groupby("x"))
out = group_split.list(df, f.x)
assert len(out) == 2
assert out[0].equals(df.iloc[[0], :])
assert out[1].equals(df.iloc[[1, 2], :].reset_index(drop=True))
def test_group_split_keep_false_does_not_tryto_remove_virtual_grouping_cols():
# test_that("group_split(keep=FALSE) does not try to
# remove virtual grouping columns (#4045)", {
iris3 = iris.head(4).copy()
df = group_by(iris3, _bootstrap=[0, 1, 0, 1])
rows = [[0, 2], [1, 3]]
res = group_split.list(df, _keep=False)
iris3 = select(iris3, ~f._bootstrap)
assert len(res) == 2
assert_frame_equal(res[0], iris3.iloc[rows[0], :].reset_index(drop=True))
assert_frame_equal(res[1], iris3.iloc[rows[1], :].reset_index(drop=True))
def test_group_split_respects__drop():
# test_that("group_split() respects .drop", {
chunks = tibble(f=factor(["b"], levels=list("abc"))) >> group_split.list(
f.f, _drop=True)
assert len(chunks) == 1
def test_group_split_works_if_no_grouping_column():
out = group_split.list(iris)
assert len(out) == 1
assert out[0].equals(iris)
def test_group_split_can_discard_grouping_vars_by__keep_eqs_false():
tbl = tibble(x=[1, 2, 3, 4], g=factor(rep(["a", "b"], each=2)))
res = group_split.list(tbl, f.g, _keep=False)
assert res[0].equals(tbl.iloc[:2, [0]])
assert res[1].equals(tbl.iloc[[2, 3], [0]].reset_index(drop=True))