DEPS: Added tz keyword to to_datetime function, deprecates utc #17413
Conversation
|
Hello @harpolea! Thanks for updating the PR. Cheers ! There are no PEP8 issues in this Pull Request. 🍻 Comment last updated on September 01, 2017 at 13:50 Hours UTC |
…ame rather than the timezone
gfyoung
added
Compat
| @@ -328,6 +327,10 @@ Deprecations | |||
|
|
|||
| - ``pd.options.html.border`` has been deprecated in favor of ``pd.options.display.html.border`` (:issue:`15793`). | |||
|
|
|||
|
|
|||
| - :func: `to_datetime` now takes ``tz`` keyword argument, ``utc`` argument is deprecated (:issue:`13712`) | |||
There was a problem hiding this comment.
Say something like (don't copy, make sure to format): The "utc" argument in "to_datetime" has been deprecated in favor of "tz"
| @@ -270,6 +270,23 @@ def test_to_datetime_utc_is_true(self): | |||
| expected = pd.DatetimeIndex(data=date_range) | |||
| tm.assert_index_equal(result, expected) | |||
|
|
|||
| def test_to_datetime_tz_kw(self): | |||
There was a problem hiding this comment.
Make to test the utc kwarg as being deprecated too.
|
note that we just merged: #17109 so rebase |
|
|
||
| .. deprecated: 0.21.0 | ||
|
|
||
| tz : pytz.timezone or dateutil.tz.tzfile, default None |
| @@ -431,6 +441,10 @@ def _convert_listlike(arg, box, format, name=None, tz=tz): | |||
| result = arg | |||
|
|
|||
| if result is None and (format is None or infer_datetime_format): | |||
| if tz == 'utc' or tz == 'UTC': | |||
| @@ -182,9 +183,12 @@ def _guess_datetime_format_for_array(arr, **kwargs): | |||
| return _guess_datetime_format(arr[non_nan_elements[0]], **kwargs) | |||
|
|
|||
|
|
|||
| @deprecate_kwarg(old_arg_name='utc', new_arg_name='tz', | |||
There was a problem hiding this comment.
instead of doing this, simply translate utc=True to tz='UTC' in the code itself (and issue a deprecation warning). Further you will need to raise if utc=True and tz is not None, so need a bit more handling (and add a test for this).
|
|
||
| result = pd.to_datetime(data, format='%Y%m%d %H%M%S', tz=tz) | ||
| expected = pd.DatetimeIndex(data=date_range) | ||
| tm.assert_numpy_array_equal(result.values, expected.values) |
There was a problem hiding this comment.
compare index using tm.assert_index_equal
There was a problem hiding this comment.
@jreback for some context, the reason I suggested to do it like this is because assert_index_equal fails because timezone localization on single timestamp or on datetimeindex results in a slightly different tz object. This is a separate issue as the the one solved in this pr, therefore suggested this workaround which at least test this is correct (but didn't have time to open an issue, and from phone now)
There was a problem hiding this comment.
we never compare like this - too easy to get wrong; if there is an issue pls show
There was a problem hiding this comment.
I can't test it now, but if I remember correctly: compare pd.Timestamp('2012-01-01', tz='US/Eastern').tz with pd.to_datetime(['2012-01-01']).tz_localize('US/Eastern').tz. Both attributes are not equal
|
In general, I am wondering whether it is actually needed to deprecate the keyword / worth the trouble for all impacted users. |
|
why don't we put a DeprecationWanring on utc rather than a FutureWarninhg kind of a soft deprecation |
|
can you rebase / update |
|
can u rebase and move note to 0.22 |
|
can you rebase / move note to 0.22.0 |
|
closing as stale, but if you'd like to keep working, ping and we can reopen |
git diff upstream/master -u -- "*.py" | flake8 --diff