DEPR: Deprecate box kwarg for to_timedelta and to_datetime

Author: Chris Bertinato

doc/source/whatsnew/v0.24.0.rst

@@ -1326,6 +1326,7 @@ Deprecations
 - Passing an integer to :meth:`Series.fillna` and :meth:`DataFrame.fillna` with ``timedelta64[ns]`` dtypes is deprecated, will raise ``TypeError`` in a future version.  Use ``obj.fillna(pd.Timedelta(...))`` instead (:issue:`24694`)
 - ``Series.cat.categorical``, ``Series.cat.name`` and ``Sersies.cat.index`` have been deprecated. Use the attributes on ``Series.cat`` or ``Series`` directly. (:issue:`24751`).
 - Passing a dtype without a precision like ``np.dtype('datetime64')`` or ``timedelta64`` to :class:`Index`, :class:`DatetimeIndex` and :class:`TimedeltaIndex` is now deprecated. Use the nanosecond-precision dtype instead (:issue:`24753`).
+- The functions :func:`pandas.to_datetime` and :func:`pandas.to_timedelta` have deprecated the ``box`` keyword. Use :attr:`Series.values` and :meth:`Timestamp.to_datetime64`/:meth:`Timedelta.to_timedelta64` instead to get an ndarray of values or ``numpy.timestamp64``/``numpy.timedelta64``, respectively (:issue:`24416`).
 
 .. _whatsnew_0240.deprecations.datetimelike_int_ops:
 

pandas/core/dtypes/cast.py

@@ -794,10 +794,10 @@ def soft_convert_objects(values, datetime=True, numeric=True, timedelta=True,
         # Immediate return if coerce
         if datetime:
             from pandas import to_datetime
-            return to_datetime(values, errors='coerce', box=False)
+            return to_datetime(values, errors='coerce').to_numpy()
         elif timedelta:
             from pandas import to_timedelta
-            return to_timedelta(values, errors='coerce', box=False)
+            return to_timedelta(values, errors='coerce').to_numpy()
         elif numeric:
             from pandas import to_numeric
             return to_numeric(values, errors='coerce')

pandas/core/indexes/datetimelike.py

@@ -300,7 +300,12 @@ def asobject(self):
         return self.astype(object)
 
     def _convert_tolerance(self, tolerance, target):
-        tolerance = np.asarray(to_timedelta(tolerance, box=False))
+        tolerance = to_timedelta(tolerance)
+        if isinstance(tolerance, ABCIndexClass):
+            tolerance = tolerance.to_numpy()
+        else:
+            tolerance = np.asarray(tolerance.to_timedelta64())
+
         if target.size != tolerance.size and tolerance.size > 1:
             raise ValueError('list-like tolerance size must match '
                              'target index size')

pandas/core/tools/datetimes.py

@@ -20,6 +20,8 @@
 from pandas import compat
 from pandas.core import algorithms
 
+from pandas.util._decorators import deprecate_kwarg
+
 
 def _guess_datetime_format_for_array(arr, **kwargs):
     # Try to guess the format based on the first non-NaN element
@@ -398,6 +400,7 @@ def _adjust_to_origin(arg, origin, unit):
     return arg
 
 
+@deprecate_kwarg(old_arg_name='box', new_arg_name=None)
 def to_datetime(arg, errors='raise', dayfirst=False, yearfirst=False,
                 utc=None, box=True, format=None, exact=True,
                 unit=None, infer_datetime_format=False, origin='unix',
@@ -444,6 +447,12 @@ def to_datetime(arg, errors='raise', dayfirst=False, yearfirst=False,
 
         - If True returns a DatetimeIndex or Index-like object
         - If False returns ndarray of values.
+
+        .. deprecated:: 0.24.0
+            Use :meth:`.to_numpy` or :meth:`Timestamp.to_datetime64`
+            instead to get an ndarray of values or numpy.datetime64,
+            respectively.
+
     format : string, default None
         strftime to parse time, eg "%d/%m/%Y", note that "%f" will parse
         all the way up to nanoseconds.

pandas/core/tools/timedeltas.py

@@ -14,7 +14,10 @@
 
 from pandas.core.arrays.timedeltas import sequence_to_td64ns
 
+from pandas.util._decorators import deprecate_kwarg
 
+
+@deprecate_kwarg(old_arg_name='box', new_arg_name=None)
 def to_timedelta(arg, unit='ns', box=True, errors='raise'):
     """
     Convert argument to timedelta.
@@ -40,6 +43,12 @@ def to_timedelta(arg, unit='ns', box=True, errors='raise'):
         - If True returns a Timedelta/TimedeltaIndex of the results.
         - If False returns a numpy.timedelta64 or numpy.darray of
           values of dtype timedelta64[ns].
+
+        .. deprecated:: 0.24.0
+            Use :meth:`.to_numpy` or :meth:`Timedelta.to_timedelta64`
+            instead to get an ndarray of values or numpy.timedelta64,
+            respectively.
+
     errors : {'ignore', 'raise', 'coerce'}, default 'raise'
         - If 'raise', then invalid parsing will raise an exception.
         - If 'coerce', then invalid parsing will be set as NaT.

pandas/io/parsers.py

@@ -14,10 +14,12 @@
 
 import numpy as np
 
+from pandas import DatetimeIndex
+
 import pandas._libs.lib as lib
 import pandas._libs.ops as libops
 import pandas._libs.parsers as parsers
-from pandas._libs.tslibs import parsing
+from pandas._libs.tslibs import parsing, Timestamp
 import pandas.compat as compat
 from pandas.compat import (
     PY3, StringIO, lrange, lzip, map, range, string_types, u, zip)
@@ -3164,11 +3166,11 @@ def converter(*date_cols):
                 return tools.to_datetime(
                     ensure_object(strs),
                     utc=None,
-                    box=False,
                     dayfirst=dayfirst,
                     errors='ignore',
                     infer_datetime_format=infer_datetime_format
-                )
+                ).to_numpy()
+
             except ValueError:
                 return tools.to_datetime(
                     parsing.try_parse_dates(strs, dayfirst=dayfirst))

pandas/tests/indexes/datetimes/test_tools.py

@@ -221,12 +221,13 @@ def test_to_datetime_format_weeks(self, cache):
     def test_to_datetime_parse_tzname_or_tzoffset(self, box, const,
                                                   fmt, dates, expected_dates):
         # GH 13486
-        result = pd.to_datetime(dates, format=fmt, box=box)
-        expected = const(expected_dates)
-        tm.assert_equal(result, expected)
+        with tm.assert_produces_warning(FutureWarning):
+            result = pd.to_datetime(dates, format=fmt, box=box)
+            expected = const(expected_dates)
+            tm.assert_equal(result, expected)
 
-        with pytest.raises(ValueError):
-            pd.to_datetime(dates, format=fmt, box=box, utc=True)
+            with pytest.raises(ValueError):
+                pd.to_datetime(dates, format=fmt, box=box, utc=True)
 
     @pytest.mark.parametrize('offset', [
         '+0', '-1foo', 'UTCbar', ':10', '+01:000:01', ''])
@@ -256,7 +257,7 @@ def test_to_datetime_dtarr(self, tz):
         result = to_datetime(arr)
         assert result is arr
 
-        result = to_datetime(arr, box=True)
+        result = to_datetime(arr)
         assert result is arr
 
     def test_to_datetime_pydatetime(self):
@@ -364,7 +365,7 @@ def test_to_datetime_array_of_dt64s(self, cache):
         # Assuming all datetimes are in bounds, to_datetime() returns
         # an array that is equal to Timestamp() parsing
         tm.assert_numpy_array_equal(
-            pd.to_datetime(dts, box=False, cache=cache),
+            pd.to_datetime(dts, cache=cache).to_numpy(),
             np.array([Timestamp(x).asm8 for x in dts])
         )
 
@@ -376,8 +377,8 @@ def test_to_datetime_array_of_dt64s(self, cache):
             pd.to_datetime(dts_with_oob, errors='raise')
 
         tm.assert_numpy_array_equal(
-            pd.to_datetime(dts_with_oob, box=False, errors='coerce',
-                           cache=cache),
+            pd.to_datetime(dts_with_oob, errors='coerce',
+                           cache=cache).to_numpy(),
             np.array(
                 [
                     Timestamp(dts_with_oob[0]).asm8,
@@ -392,8 +393,8 @@ def test_to_datetime_array_of_dt64s(self, cache):
         # are converted to their .item(), which depending on the version of
         # numpy is either a python datetime.datetime or datetime.date
         tm.assert_numpy_array_equal(
-            pd.to_datetime(dts_with_oob, box=False, errors='ignore',
-                           cache=cache),
+            pd.to_datetime(dts_with_oob, errors='ignore',
+                           cache=cache).to_numpy(),
             np.array(
                 [dt.item() for dt in dts_with_oob],
                 dtype='O'
@@ -628,10 +629,14 @@ def test_to_datetime_cache(self, utc, format, box, constructor):
         date = '20130101 00:00:00'
         test_dates = [date] * 10**5
         data = constructor(test_dates)
-        result = pd.to_datetime(data, utc=utc, format=format, box=box,
-                                cache=True)
-        expected = pd.to_datetime(data, utc=utc, format=format, box=box,
-                                  cache=False)
+
+        with tm.assert_produces_warning(FutureWarning):
+            result = pd.to_datetime(data, utc=utc, format=format, box=box,
+                                    cache=True)
+
+        with tm.assert_produces_warning(FutureWarning):
+            expected = pd.to_datetime(data, utc=utc, format=format, box=box,
+                                      cache=False)
         if box:
             tm.assert_index_equal(result, expected)
         else:
@@ -684,7 +689,10 @@ def test_iso_8601_strings_with_same_offset(self):
     def test_iso_8601_strings_same_offset_no_box(self):
         # GH 22446
         data = ['2018-01-04 09:01:00+09:00', '2018-01-04 09:02:00+09:00']
-        result = pd.to_datetime(data, box=False)
+
+        with tm.assert_produces_warning(FutureWarning):
+            result = pd.to_datetime(data, box=False)
+
         expected = np.array([
             datetime(2018, 1, 4, 9, 1, tzinfo=pytz.FixedOffset(540)),
             datetime(2018, 1, 4, 9, 2, tzinfo=pytz.FixedOffset(540))
@@ -753,6 +761,16 @@ def test_timestamp_utc_true(self, ts, expected):
         result = to_datetime(ts, utc=True)
         assert result == expected
 
+    def test_to_datetime_box_deprecated(self):
+        expected = np.datetime64('2018-09-09')
+
+        # Deprecated - see GH24416
+        with tm.assert_produces_warning(FutureWarning):
+            pd.to_datetime(expected, box=False)
+
+        result = pd.to_datetime(expected).to_datetime64()
+        assert result == expected
+
 
 class TestToDatetimeUnit(object):
     @pytest.mark.parametrize('cache', [True, False])
@@ -891,7 +909,7 @@ def test_unit_rounding(self, cache):
     def test_unit_ignore_keeps_name(self, cache):
         # GH 21697
         expected = pd.Index([15e9] * 2, name='name')
-        result = pd.to_datetime(expected, errors='ignore', box=True, unit='s',
+        result = pd.to_datetime(expected, errors='ignore', unit='s',
                                 cache=cache)
         tm.assert_index_equal(result, expected)
 
@@ -1052,7 +1070,10 @@ def test_dataframe_box_false(self):
         df = pd.DataFrame({'year': [2015, 2016],
                            'month': [2, 3],
                            'day': [4, 5]})
-        result = pd.to_datetime(df, box=False)
+
+        with tm.assert_produces_warning(FutureWarning):
+            result = pd.to_datetime(df, box=False)
+
         expected = np.array(['2015-02-04', '2016-03-05'],
                             dtype='datetime64[ns]')
         tm.assert_numpy_array_equal(result, expected)
@@ -1069,8 +1090,7 @@ def test_dataframe_utc_true(self):
 
     def test_to_datetime_errors_ignore_utc_true(self):
         # GH 23758
-        result = pd.to_datetime([1], unit='s', box=True, utc=True,
-                                errors='ignore')
+        result = pd.to_datetime([1], unit='s', utc=True, errors='ignore')
         expected = DatetimeIndex(['1970-01-01 00:00:01'], tz='UTC')
         tm.assert_index_equal(result, expected)
 

pandas/tests/indexes/timedeltas/test_tools.py

@@ -19,15 +19,18 @@ def conv(v):
 
         d1 = np.timedelta64(1, 'D')
 
-        assert (to_timedelta('1 days 06:05:01.00003', box=False) ==
-                conv(d1 + np.timedelta64(6 * 3600 + 5 * 60 + 1, 's') +
-                     np.timedelta64(30, 'us')))
-        assert (to_timedelta('15.5us', box=False) ==
-                conv(np.timedelta64(15500, 'ns')))
+        with tm.assert_produces_warning(FutureWarning):
+            assert (to_timedelta('1 days 06:05:01.00003', box=False) ==
+                    conv(d1 + np.timedelta64(6 * 3600 + 5 * 60 + 1, 's') +
+                         np.timedelta64(30, 'us')))
 
-        # empty string
-        result = to_timedelta('', box=False)
-        assert result.astype('int64') == iNaT
+        with tm.assert_produces_warning(FutureWarning):
+            assert (to_timedelta('15.5us', box=False) ==
+                    conv(np.timedelta64(15500, 'ns')))
+
+            # empty string
+            result = to_timedelta('', box=False)
+            assert result.astype('int64') == iNaT
 
         result = to_timedelta(['', ''])
         assert isna(result).all()
@@ -37,10 +40,11 @@ def conv(v):
         expected = pd.Index(np.array([np.timedelta64(1, 's')]))
         tm.assert_index_equal(result, expected)
 
-        # ints
-        result = np.timedelta64(0, 'ns')
-        expected = to_timedelta(0, box=False)
-        assert result == expected
+        with tm.assert_produces_warning(FutureWarning):
+            # ints
+            result = np.timedelta64(0, 'ns')
+            expected = to_timedelta(0, box=False)
+            assert result == expected
 
         # Series
         expected = Series([timedelta(days=1), timedelta(days=1, seconds=1)])
@@ -53,16 +57,18 @@ def conv(v):
         expected = to_timedelta([0, 10], unit='s')
         tm.assert_index_equal(result, expected)
 
-        # single element conversion
-        v = timedelta(seconds=1)
-        result = to_timedelta(v, box=False)
-        expected = np.timedelta64(timedelta(seconds=1))
-        assert result == expected
+        with tm.assert_produces_warning(FutureWarning):
+            # single element conversion
+            v = timedelta(seconds=1)
+            result = to_timedelta(v, box=False)
+            expected = np.timedelta64(timedelta(seconds=1))
+            assert result == expected
 
-        v = np.timedelta64(timedelta(seconds=1))
-        result = to_timedelta(v, box=False)
-        expected = np.timedelta64(timedelta(seconds=1))
-        assert result == expected
+        with tm.assert_produces_warning(FutureWarning):
+            v = np.timedelta64(timedelta(seconds=1))
+            result = to_timedelta(v, box=False)
+            expected = np.timedelta64(timedelta(seconds=1))
+            assert result == expected
 
         # arrays of various dtypes
         arr = np.array([1] * 5, dtype='int64')
@@ -90,22 +96,26 @@ def conv(v):
         expected = TimedeltaIndex([np.timedelta64(1, 'D')] * 5)
         tm.assert_index_equal(result, expected)
 
-        # Test with lists as input when box=false
-        expected = np.array(np.arange(3) * 1000000000, dtype='timedelta64[ns]')
-        result = to_timedelta(range(3), unit='s', box=False)
-        tm.assert_numpy_array_equal(expected, result)
+        with tm.assert_produces_warning(FutureWarning):
+            # Test with lists as input when box=false
+            expected = np.array(np.arange(3) * 1000000000, dtype='timedelta64[ns]')
+            result = to_timedelta(range(3), unit='s', box=False)
+            tm.assert_numpy_array_equal(expected, result)
 
-        result = to_timedelta(np.arange(3), unit='s', box=False)
-        tm.assert_numpy_array_equal(expected, result)
+        with tm.assert_produces_warning(FutureWarning):
+            result = to_timedelta(np.arange(3), unit='s', box=False)
+            tm.assert_numpy_array_equal(expected, result)
 
-        result = to_timedelta([0, 1, 2], unit='s', box=False)
-        tm.assert_numpy_array_equal(expected, result)
+        with tm.assert_produces_warning(FutureWarning):
+            result = to_timedelta([0, 1, 2], unit='s', box=False)
+            tm.assert_numpy_array_equal(expected, result)
 
-        # Tests with fractional seconds as input:
-        expected = np.array(
-            [0, 500000000, 800000000, 1200000000], dtype='timedelta64[ns]')
-        result = to_timedelta([0., 0.5, 0.8, 1.2], unit='s', box=False)
-        tm.assert_numpy_array_equal(expected, result)
+        with tm.assert_produces_warning(FutureWarning):
+            # Tests with fractional seconds as input:
+            expected = np.array(
+                [0, 500000000, 800000000, 1200000000], dtype='timedelta64[ns]')
+            result = to_timedelta([0., 0.5, 0.8, 1.2], unit='s', box=False)
+            tm.assert_numpy_array_equal(expected, result)
 
     def test_to_timedelta_invalid(self):
 
@@ -173,3 +183,13 @@ def test_to_timedelta_on_missing_values(self):
 
         actual = pd.to_timedelta(pd.NaT)
         assert actual.value == timedelta_NaT.astype('int64')
+
+    def test_to_timedelta_box_deprecated(self):
+        result = np.timedelta64(0, 'ns')
+
+        # Deprecated - see GH24416
+        with tm.assert_produces_warning(FutureWarning):
+            to_timedelta(0, box=False)
+
+        expected = to_timedelta(0).to_timedelta64()
+        assert result == expected