def test_dataframe_setitem_new_columns(df, arg, value):
gdf = DataFrame.from_pandas(df)
cudf_replace_value = value
if isinstance(cudf_replace_value, pd.DataFrame):
cudf_replace_value = DataFrame.from_pandas(value)
df[arg] = value
gdf[arg] = cudf_replace_value
assert_eq(df, gdf, check_dtype=True)
def test_kernel_shallow_copy():
pdf = pd.DataFrame([[1, 2, 3], [4, 5, 6], [7, 8, 9]],
columns=["a", "b", "c"])
gdf = DataFrame.from_pandas(pdf)
cdf = gdf.copy(deep=False)
sr = gdf["a"]
add_one[1, len(sr)](sr.to_gpu_array())
assert_eq(gdf, cdf)
def test_cudf_dataframe_copy(copy_fn, ncols, data_type):
pdf = pd.DataFrame()
for i in range(ncols):
pdf[chr(i + ord("a"))] = pd.Series(np.random.randint(
0, 1000, 20)).astype(data_type)
df = DataFrame.from_pandas(pdf)
copy_df = copy_fn(df)
assert_eq(df, copy_df)
def test_kernel_deep_copy():
pdf = pd.DataFrame([[1, 2, 3], [4, 5, 6], [7, 8, 9]],
columns=["a", "b", "c"])
gdf = DataFrame.from_pandas(pdf)
cdf = gdf.copy(deep=True)
sr = gdf["b"]
add_one[1, len(sr)](sr._column.data_array_view)
assert not gdf.to_string().split() == cdf.to_string().split()
def test_dataframe_copy_shallow():
pdf = pd.DataFrame([[1, 2, 3], [4, 5, 6], [7, 8, 9]],
columns=["a", "b", "c"])
gdf = DataFrame.from_pandas(pdf)
copy_pdf = pdf.copy(deep=False)
copy_gdf = gdf.copy(deep=False)
copy_pdf["b"] = [0, 0, 0]
copy_gdf["b"] = [0, 0, 0]
assert_eq(pdf["b"], copy_pdf["b"])
assert_eq(gdf["b"], copy_gdf["b"])
def test_kernel_deep_copy():
pdf = pd.DataFrame([[1, 2, 3], [4, 5, 6], [7, 8, 9]],
columns=["a", "b", "c"])
gdf = DataFrame.from_pandas(pdf)
cdf = gdf.copy(deep=True)
sr = gdf["b"]
# column.to_gpu_array calls to_dense_buffer which returns a copy
# need to access buffer directly and then call gpu_array
add_one[1, len(sr)](sr.data.to_gpu_array())
assert not gdf.to_string().split() == cdf.to_string().split()
def test_dataframe_deep_copy_and_insert(copy_parameters):
pdf = pd.DataFrame([[1, 2, 3], [4, 5, 6], [7, 8, 9]],
columns=["a", "b", "c"])
gdf = DataFrame.from_pandas(pdf)
copy_pdf = copy_parameters["fn"](pdf)
copy_gdf = copy_parameters["fn"](gdf)
copy_pdf["b"] = [0, 0, 0]
copy_gdf["b"] = [0, 0, 0]
pdf_is_equal = np.array_equal(pdf["b"].values, copy_pdf["b"].values)
gdf_is_equal = np.array_equal(gdf["b"].to_array(),
copy_gdf["b"].to_array())
assert pdf_is_equal == copy_parameters["expected_equality"]
assert gdf_is_equal == copy_parameters["expected_equality"]
def test_cudf_dataframe_copy_then_insert(copy_fn, ncols, data_type):
pdf = pd.DataFrame()
for i in range(ncols):
pdf[chr(i + ord("a"))] = pd.Series(np.random.randint(
0, 1000, 20)).astype(data_type)
df = DataFrame.from_pandas(pdf)
copy_df = copy_fn(df)
copy_pdf = copy_fn(pdf)
copy_df["aa"] = pd.Series(np.random.randint(0, 1000, 20)).astype(data_type)
copy_pdf["aa"] = pd.Series(np.random.randint(0, 1000,
20)).astype(data_type)
assert not copy_pdf.to_string().split() == pdf.to_string().split()
assert not copy_df.to_string().split() == df.to_string().split()
def test_setitem_dataframe_series_inplace(df):
pdf = df
gdf = DataFrame.from_pandas(pdf)
pdf["a"].replace(1, 500, inplace=True)
gdf["a"].replace(1, 500, inplace=True)
assert_eq(pdf, gdf)
psr_a = pdf["a"]
gsr_a = gdf["a"]
psr_a.replace(500, 501, inplace=True)
gsr_a.replace(500, 501, inplace=True)
assert_eq(pdf, gdf)
def test_dataframe_setitem_bool_mask_scaler(df, arg, value):
gdf = DataFrame.from_pandas(df)
df[arg] = value
gdf[arg] = value
assert_eq(df, gdf)