Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

ENH: support downcasting of nullable EAs in pd.to_numeric #38746

Merged
merged 20 commits into from
Dec 30, 2020
Merged
Show file tree
Hide file tree
Changes from 3 commits
Commits
File filter

Filter by extension

Filter by extension

Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
25 changes: 25 additions & 0 deletions pandas/core/tools/numeric.py
Original file line number Diff line number Diff line change
Expand Up @@ -7,6 +7,9 @@
ensure_object,
is_datetime_or_timedelta_dtype,
is_decimal,
is_extension_array_dtype,
is_float_dtype,
is_integer_dtype,
is_number,
is_numeric_dtype,
is_scalar,
Expand All @@ -15,6 +18,8 @@
from pandas.core.dtypes.generic import ABCIndex, ABCSeries

import pandas as pd
from pandas.core.arrays.numeric import NumericArray
from pandas.core.construction import extract_array


def to_numeric(arg, errors="raise", downcast=None):
Expand Down Expand Up @@ -118,10 +123,14 @@ def to_numeric(arg, errors="raise", downcast=None):
is_series = False
is_index = False
is_scalars = False
is_numeric_extension_dtype = False

if isinstance(arg, ABCSeries):
is_series = True
values = arg.values
if is_extension_array_dtype(arg) and isinstance(values, NumericArray):
is_numeric_extension_dtype = True
values = extract_array(arg)
Copy link
Member

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Is this extract_array line needed? (the values = arg.values from above should have worked fine, I think)

Copy link
Member Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

It's not, I reverted this bit

elif isinstance(arg, ABCIndex):
is_index = True
if needs_i8_conversion(arg.dtype):
Expand All @@ -142,6 +151,12 @@ def to_numeric(arg, errors="raise", downcast=None):
else:
values = arg

if is_numeric_extension_dtype or (
is_extension_array_dtype(arg) and isinstance(values, NumericArray)
):
is_numeric_extension_dtype = True
mask, values = values._mask, values._data
Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

you likely need to just take the masked values only for the following block of code

Copy link
Member Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Done


values_dtype = getattr(values, "dtype", None)
if is_numeric_dtype(values_dtype):
pass
Expand Down Expand Up @@ -188,6 +203,16 @@ def to_numeric(arg, errors="raise", downcast=None):
if values.dtype == dtype:
break

if is_numeric_extension_dtype:
Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

L194 might need to handle the mask

Copy link
Member Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Do you mean?

            float_32_ind = typecodes.index(float_32_char)

I feel like there's a testcase I haven't looked at if yes (the ones I have work as is)

if is_integer_dtype(values):
from pandas.core.arrays import IntegerArray

values = IntegerArray(values, mask)
elif is_float_dtype(values):
from pandas.core.arrays import FloatingArray

values = FloatingArray(values, mask)

if is_series:
return arg._constructor(values, index=arg.index, name=arg.name)
elif is_index:
Expand Down
16 changes: 16 additions & 0 deletions pandas/tests/tools/test_to_numeric.py
Original file line number Diff line number Diff line change
Expand Up @@ -725,3 +725,19 @@ def test_to_numeric_from_nullable_string(values, expected):
s = Series(values, dtype="string")
result = to_numeric(s)
tm.assert_series_equal(result, expected)


@pytest.mark.parametrize(
"data, input_dtype, downcast, expected_dtype",
(
([1, 1], "Int64", "integer", "Int8"),
([450, 300], "Int64", "integer", "Int16"),
([1, 1], "Float64", "float", "Float32"),
([1, 1], "Float64", "integer", "Int8"),
arw2019 marked this conversation as resolved.
Show resolved Hide resolved
Copy link
Member

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Nit: maybe move this one up to the other "integer" downcast cases

Copy link
Member Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Done

),
)
def test_downcast_nullable_numeric(data, input_dtype, downcast, expected_dtype):
arr = pd.array(data, dtype=input_dtype)
result = pd.to_numeric(arr, downcast=downcast)
expected = pd.array(data, dtype=expected_dtype)
tm.assert_extension_array_equal(result, expected)