def test_replace_inplace():
data = np.array([5, 1, 2, 3, 4])
sr = Series(data)
psr = pd.Series(data)
sr_copy = sr.copy()
psr_copy = psr.copy()
assert_eq(sr, psr)
assert_eq(sr_copy, psr_copy)
sr.replace(5, 0, inplace=True)
psr.replace(5, 0, inplace=True)
assert_eq(sr, psr)
assert_eq(sr_copy, psr_copy)
sr = Series(data)
psr = pd.Series(data)
sr_copy = sr.copy()
psr_copy = psr.copy()
assert_eq(sr, psr)
assert_eq(sr_copy, psr_copy)
sr.replace({5: 0, 3: -5})
psr.replace({5: 0, 3: -5})
assert_eq(sr, psr)
assert_eq(sr_copy, psr_copy)
srr = sr.replace()
psrr = psr.replace()
assert_eq(srr, psrr)
psr = pd.Series(["one", "two", "three"], dtype="category")
sr = Series.from_pandas(psr)
sr_copy = sr.copy()
psr_copy = psr.copy()
assert_eq(sr, psr)
assert_eq(sr_copy, psr_copy)
sr.replace("one", "two", inplace=True)
psr.replace("one", "two", inplace=True)
assert_eq(sr, psr)
assert_eq(sr_copy, psr_copy)
pdf = pd.DataFrame({"A": [0, 1, 2, 3, 4], "B": [5, 6, 7, 8, 9]})
gdf = DataFrame.from_pandas(pdf)
pdf_copy = pdf.copy()
gdf_copy = gdf.copy()
assert_eq(pdf, gdf)
assert_eq(pdf_copy, gdf_copy)
pdf.replace(5, 0, inplace=True)
gdf.replace(5, 0, inplace=True)
assert_eq(pdf, gdf)
assert_eq(pdf_copy, gdf_copy)
pds = pd.Series([1, 2, 3, 45])
gds = Series.from_pandas(pds)
vals = np.array([]).astype(int)
assert_eq(pds.replace(vals, -1), gds.replace(vals, -1))
pds.replace(vals, 77, inplace=True)
gds.replace(vals, 77, inplace=True)
assert_eq(pds, gds)
pdf = pd.DataFrame({"a": [1, 2, 3, 4, 5, 666]})
gdf = DataFrame.from_pandas(pdf)
assert_eq(pdf.replace({"a": 2}, {"a": -33}),
gdf.replace({"a": 2}, {"a": -33}))
assert_eq(
pdf.replace({"a": [2, 5]}, {"a": [9, 10]}),
gdf.replace({"a": [2, 5]}, {"a": [9, 10]}),
)
assert_eq(
pdf.replace([], []),
gdf.replace([], []),
)
assert_exceptions_equal(
lfunc=pdf.replace,
rfunc=gdf.replace,
lfunc_args_and_kwargs=([], {
"to_replace": -1,
"value": []
}),
rfunc_args_and_kwargs=([], {
"to_replace": -1,
"value": []
}),
compare_error_message=False,
)
def test_replace_inplace():
data = np.array([5, 1, 2, 3, 4])
sr = Series(data)
psr = pd.Series(data)
sr_copy = sr.copy()
psr_copy = psr.copy()
assert_eq(sr, psr)
assert_eq(sr_copy, psr_copy)
sr.replace(5, 0, inplace=True)
psr.replace(5, 0, inplace=True)
assert_eq(sr, psr)
assert_eq(sr_copy, psr_copy)
sr = Series(data)
psr = pd.Series(data)
sr_copy = sr.copy()
psr_copy = psr.copy()
assert_eq(sr, psr)
assert_eq(sr_copy, psr_copy)
sr.replace({5: 0, 3: -5})
psr.replace({5: 0, 3: -5})
assert_eq(sr, psr)
assert_eq(sr_copy, psr_copy)
srr = sr.replace()
psrr = psr.replace()
assert_eq(srr, psrr)
psr = pd.Series(["one", "two", "three"], dtype="category")
sr = Series.from_pandas(psr)
sr_copy = sr.copy()
psr_copy = psr.copy()
assert_eq(sr, psr)
assert_eq(sr_copy, psr_copy)
sr.replace("one", "two", inplace=True)
psr.replace("one", "two", inplace=True)
assert_eq(sr, psr)
assert_eq(sr_copy, psr_copy)
pdf = pd.DataFrame({"A": [0, 1, 2, 3, 4], "B": [5, 6, 7, 8, 9]})
gdf = DataFrame.from_pandas(pdf)
pdf_copy = pdf.copy()
gdf_copy = gdf.copy()
assert_eq(pdf, gdf)
assert_eq(pdf_copy, gdf_copy)
pdf.replace(5, 0, inplace=True)
gdf.replace(5, 0, inplace=True)
assert_eq(pdf, gdf)
assert_eq(pdf_copy, gdf_copy)
pds = pd.Series([1, 2, 3, 45])
gds = Series.from_pandas(pds)
vals = np.array([]).astype(int)
assert_eq(pds.replace(vals, -1), gds.replace(vals, -1))
pds.replace(vals, 77, inplace=True)
gds.replace(vals, 77, inplace=True)
assert_eq(pds, gds)
pdf = pd.DataFrame({"a": [1, 2, 3, 4, 5, 666]})
gdf = DataFrame.from_pandas(pdf)
assert_eq(pdf.replace({"a": 2}, {"a": -33}),
gdf.replace({"a": 2}, {"a": -33}))
assert_eq(
pdf.replace({"a": [2, 5]}, {"a": [9, 10]}),
gdf.replace({"a": [2, 5]}, {"a": [9, 10]}),
)
assert_eq(
pdf.replace([], []),
gdf.replace([], []),
)
with pytest.raises(TypeError):
pdf.replace(-1, [])
with pytest.raises(TypeError):
gdf.replace(-1, [])
class GpuArrowNodeReader(object):
def __init__(self, table, index):
self._table = table
self._field = table.schema[index]
self._series = Series(column.as_column(table.column(index)))
self._series.name = self.name
def __len__(self):
return len(self._series)
@property
def schema(self):
return self._table.schema
@property
def field_schema(self):
return self._field
@property
def is_dictionary(self):
return pa.types.is_dictionary(self._field.type)
@property
def null_count(self):
return self._series.null_count
@property
def dtype(self):
return arrow_to_pandas_dtype(self._field.type)
@property
def index_dtype(self):
return self._field.type.index_type.to_pandas_dtype()
@property
def name(self):
return self._field.name
@property
def data(self):
"""
Return the data as the expected dtype
and with the padding bytes truncated.
"""
if self.data_raw is not None:
return self.data_raw.view(
self.dtype if not self.is_dictionary else self.index_dtype)
@property
def null(self):
"""
Return the null mask with the padding bytes truncated.
"""
if self.null_raw is not None:
bits = mask_bitsize
itemsize = mask_dtype.itemsize
end = ((len(self) + bits - 1) // bits) * itemsize
return self.null_raw[:end].view(mask_dtype)
@property
def data_raw(self):
"Accessor for the data buffer as a device array"
return self._series._column.data_array_view
@property
def null_raw(self):
"Accessor for the null buffer as a device array"
return self._series._column.mask_array_view
def make_series(self):
"""Make a Series object out of this node
"""
return self._series.copy(deep=False)
def _make_dictionary_series(self):
"""Make a dictionary-encoded series from this node
"""
assert self.is_dictionary
return self._series.copy(deep=False)
