BUG: Series construction sharing data with DTI #22113
Conversation
doc/source/whatsnew/v0.24.0.txt
Outdated
| @@ -434,7 +434,7 @@ Indexing | |||
| - Fixed ``DataFrame[np.nan]`` when columns are non-unique (:issue:`21428`) | |||
| - Bug when indexing :class:`DatetimeIndex` with nanosecond resolution dates and timezones (:issue:`11679`) | |||
| - Bug where indexing with a Numpy array containing negative values would mutate the indexer (:issue:`21867`) | |||
|
|
|||
| - Bug when :class:`Series` was created from :class:`DatetimeIndex`, the series shares the same underneath data with that index (:issue:`21907`) | |||
There was a problem hiding this comment.
don't need the word underneath.
| @@ -206,7 +206,11 @@ def __init__(self, data=None, index=None, dtype=None, name=None, | |||
| data = data.astype(dtype) | |||
| else: | |||
| # need to copy to avoid aliasing issues | |||
| data = data._values.copy() | |||
There was a problem hiding this comment.
wouldn't passing deep=True simply solve this?
There was a problem hiding this comment.
I think the problem only occurs with tz-aware cases.
There was a problem hiding this comment.
this is not a good solution here, to special case things. pass deep=True here; if this doesn't work, then the copy constructor is broken.
There was a problem hiding this comment.
I guess this bug is caused by the return value of _values. In datetimes.py, _values returns the DatetimeIndex itself for tz-aware and ndarray for tz-naive.
@property
def _values(self):
# tz-naive -> ndarray
# tz-aware -> DatetimeIndex
if self.tz is not None:
return self
else:
return self.values
That 's why this problem only occurs with tz-aware cases as jbrockmendel said.
In [2]: dti1 = pd.date_range('2016-01-01', periods=10, tz='US/Pacific')
In [3]: dti2 = pd.date_range('2016-01-01', periods=10)
In [4]: ser1 = pd.Series(dti1)
In [5]: ser2 = pd.Series(dti2)
In [6]: ser1[::3]=pd.NaT
In [7]: ser2[::3]=pd.NaT
In [8]: dti1
Out[8]:
DatetimeIndex([ 'NaT', '2016-01-02 00:00:00-08:00',
'2016-01-03 00:00:00-08:00', 'NaT',
'2016-01-05 00:00:00-08:00', '2016-01-06 00:00:00-08:00',
'NaT', '2016-01-08 00:00:00-08:00',
'2016-01-09 00:00:00-08:00', 'NaT'],
dtype='datetime64[ns, US/Pacific]', freq='D')
In [9]: dti2
Out[9]:
DatetimeIndex(['2016-01-01', '2016-01-02', '2016-01-03', '2016-01-04',
'2016-01-05', '2016-01-06', '2016-01-07', '2016-01-08',
'2016-01-09', '2016-01-10'],
dtype='datetime64[ns]', freq='D')
jreback
changed the title
|
also pls for all indices via the indices fixture, we DO test construction on all indices, so maybe you can just add onto these tests, look in pandas/tests/series/test_constructor.py |
|
can you rebase and update |
makbigc
force-pushed
the
bug-fix-GH21907
branch
from
ab8376b to
fd9c8a6
Compare
|
Hello @makbigc! Thanks for updating the PR.
|
makbigc
force-pushed
the
bug-fix-GH21907
branch
from
fd9c8a6 to
388a21a
Compare
Codecov Report
@@ Coverage Diff @@
## master #22113 +/- ##
==========================================
+ Coverage 92.23% 92.28% +0.05%
==========================================
Files 161 161
Lines 51408 51451 +43
==========================================
+ Hits 47416 47483 +67
+ Misses 3992 3968 -24
Continue to review full report at Codecov.
|
|
closing as superseded by #24096 |
add a type check for DatetimeIndex and then pass a deep copy to create a Series object
closes #21907