CLN: remove _maybe_update_attributes #27896
Conversation
| def __neg__(self): | ||
| if self.freq is not None: | ||
| return type(self)(-self._data, freq=-self.freq) | ||
| return type(self)(-self._data) | ||
|
|
||
| def __pos__(self): | ||
| return type(self)(self._data, freq=self.freq) |
There was a problem hiding this comment.
This is new? Or where was it done before?
There was a problem hiding this comment.
In _add_numeric_methods_unary.
| @@ -163,6 +164,20 @@ def values(self): | |||
| def asi8(self): | |||
| return self._data.asi8 | |||
|
|
|||
| def __array_wrap__(self, result, context=None): | |||
There was a problem hiding this comment.
sidenote (not for this PR): we should move to __array_ufunc__ here instead
There was a problem hiding this comment.
agreed on this, can you create an issue @jbrockmendel
| def __neg__(self): | ||
| if self.freq is not None: | ||
| return type(self)(-self._data, freq=-self.freq) | ||
| return type(self)(-self._data) | ||
|
|
||
| def __pos__(self): | ||
| return type(self)(self._data, freq=self.freq) |
There was a problem hiding this comment.
In _add_numeric_methods_unary.
|
can you rebase |
|
Took a glance at the ASV portal this might have caused a lot of performance regressions: https://pandas.pydata.org/speed/pandas/#/regressions @jbrockmendel do you have a quick way of validating that? |
Not a quick way, no. I'll take a look, but likely won't get to it today. |
|
Oh yea no rush - thanks! |
|
Might have misread but thought that was the commit that preceded this one on master
…Sent from my iPhone
On Oct 18, 2019, at 7:06 PM, jbrockmendel ***@***.***> wrote:
@WillAyd looking at this it isn't obvious which regressions you think may have been caused here. I see a bunch from August 15 linking to 15b11c8; are those the ones you have in mind?
—
You are receiving this because you were mentioned.
Reply to this email directly, view it on GitHub, or unsubscribe.
|
|
It's indeed this commit that asv reports for the biggest regressions. But there was also a gap in asv runs between August 2 and this commit (August 15), so probably it can be anything in that window |
|
@TomAugspurger it also seems that the last runs were in September? |
|
Will check today. We had some issues with a bad ASV env failing builds.
… On Oct 19, 2019, at 07:05, Joris Van den Bossche ***@***.***> wrote:
@TomAugspurger it also seems that the last runs were in September?
—
You are receiving this because you were mentioned.
Reply to this email directly, view it on GitHub, or unsubscribe.
|
It is only non-trivial for DTI and TDI, and only actually needed in
__array_wrap__. It is cleaner to implement the relevant patch inDatetimeIndexOpsMixin.__array_wrap__and avoid the need for maybe_update_attributes altogether.