def test_join_on_same_col_multiple_times():
data = data_frame(ii=[1, 2, 3], jj=[1, 2, 9])
df_a = backend.load_df(data)
df_b = backend.load_df(data_frame(ii=[1, 2, 3]))
out = inner_join(df_a, df_b, {("ii", "jj"): "ii"}) >> collect()
# keeps all but last row
assert_frame_sort_equal(out, data.iloc[:2, ])
def test_join_on_list_arg(backend):
# TODO: how to validate how cols are being matched up?
data = DF1.assign(jj=lambda d: d.ii)
df_a = backend.load_df(data)
df_b = backend.load_df(DF2.assign(jj=lambda d: d.ii))
out = inner_join(df_a, df_b, ["ii", "jj"]) >> collect()
assert_frame_sort_equal(out, data.iloc[:2, :].assign(y=["a", "b"]))
def test_anti_join_arg_sql_on(backend, df1, df2):
cond = lambda lhs, rhs: lhs.ii > rhs.ii
# collect sql result
out = anti_join(df1, df2, sql_on=cond) >> collect()
# for target, do full cartesian product, then filter based on cond
target = data_frame(ii=[1], x=["a"])
assert_frame_sort_equal(out, target)
def test_left_join_arg_sql_on(backend, df1, df2):
cond = lambda lhs, rhs: lhs.ii > rhs.ii
# collect sql result
out = left_join(df1, df2, sql_on=cond) >> collect()
# for target, do full cartesian product, then filter based on cond
target = data_frame(ii_x=[1, 2, 3, 3, 4, 4],
ii_y=[None, 1, 1, 2, 1, 2],
x=["a", "b", "c", "c", "d", "d"],
y=[None, "a", "a", "b", "a", "b"])
# TODO: SQL columns seem to be returned in random order, so sort
# not sure why it's happening, look into in SqlAlchemy?
assert_frame_sort_equal(out.sort_index(axis=1), target)
def test_basic_left_join(df1, df2):
out = left_join(df1, df2, {"ii": "ii"}) >> collect()
target = DF1.assign(y=["a", "b", None, None])
assert_frame_sort_equal(out, target)
def test_join_suffixes_dupe_names(df1):
out = inner_join(df1, df1, {"ii": "ii"}) >> collect()
non_index_cols = DF1.columns[DF1.columns != "ii"]
assert all((non_index_cols + "_x").isin(out))
assert all((non_index_cols + "_y").isin(out))
def test_join_on_str_arg(df1, df2):
out = inner_join(df1, df2, "ii") >> collect()
target = DF1.iloc[:2, ].assign(y=["a", "b"])
assert_frame_sort_equal(out, target)
def test_join_diff_vars_keeps_left(backend, df1, df2_jj):
out = inner_join(df1, df2_jj, {"ii": "jj"}) >> collect()
assert out.columns.tolist() == ["ii", "x", "y"]
def test_basic_anti_join(backend, df1, df2):
assert_frame_sort_equal(
anti_join(df1, df2, on={
"ii": "ii",
"x": "y"
}) >> collect(), DF1.iloc[2:, ])
def test_semi_join_no_cross(backend, df1, df2):
df_ii = backend.load_df(data_frame(ii=[1, 1]))
assert_frame_sort_equal(
semi_join(df1, df_ii, {"ii": "ii"}) >> collect(), DF1.iloc[:1, ])
def test_basic_semi_join(backend, df1, df2):
assert_frame_sort_equal(
semi_join(df1, df2, {"ii": "ii"}) >> collect(), DF1.iloc[:2, ])
def test_basic_full_join(skip_backend, backend, df1, df2):
out = full_join(df1, df2, {"ii": "ii"}) >> collect()
target = DF1.merge(DF2, on="ii", how="outer")
assert_frame_sort_equal(out, target)
def test_basic_inner_join(df1, df2):
out = inner_join(df1, df2, {"ii": "ii"}) >> collect()
target = DF1.iloc[:2, :].assign(y=["a", "b"])
assert_frame_sort_equal(out, target)
def test_basic_right_join(backend, df1, df2):
# same as left join, but flip df arguments
out = right_join(df2, df1, {"ii": "ii"}) >> collect()
target = DF1.assign(y=["a", "b", None, None])
assert_frame_sort_equal(out, target)
def test_collect(df):
assert isinstance(collect(df), pd.DataFrame)
assert isinstance(df >> collect(), pd.DataFrame)
assert isinstance(collect(), Pipeable)
def test_basic_anti_join_on_str(backend, df1, df2):
assert_frame_sort_equal(
anti_join(df1, df2, on="ii") >> collect(), DF1.iloc[2:, ])
