BUG: In `main`, using `resample().interpolate(inplace=True)` raises an exception about pandas HOT 7 OPEN

as in, remove it from interpolate. if any intermediate function (which is also used in other places) requires inplace, just pass inplace=False

from pandas.

@Dr-Irv Thank you for reporting this problem! I am able to reproduce this. The problem is purely related to the inplace=True option. The interpolation result here becomes None and internally the _update_inplace method call here does not yet possess the final result. There is still the step of index correction missing, which is only applied afterwards here.

The problem is that the context in pandas.core.generic.NDFrame.interpolate does not know about the correct final index. So far, I am unable to figure out how I can update the state of the parent data frame that led to the pandas.core.resample.Resampler.interpolate call. I tried to manipulate self._selected_obj but it does not propagate. If somebody can point me at the object that needs to be updated I can easily fix this.

from pandas.

I could use some input here @MarcoGorelli. As we are introducing a breaking change anyway, we could think about removing the inplace option. I think it is counter-intuitive that the second method in a chain of methods modifies the original data frame "in place". Just think about the completely valid option to split the calls like this:

df = ...
resampler = df.resample("ME")   # creates a resampler instance, does not modify df
resampler.interpolate(inplace=True)  # modifies df, returns None

from pandas.

we could think about removing the inplace option.

yup - this looks like one of the places where inplace was a lie anyway? Agree, "in for a penny, in for a pound" - if we gonna break, let's break. It's for the better anyway

from pandas.

@MarcoGorelli 🙌 I'll take care of it and ping you in the PR.

from pandas.

@MarcoGorelli I looked into this but it is extremely convoluted 😢 The inplace option is passed on in multiple steps, partially to overloaded methods or abstract methods with many implementations. There are dozens of tests that use inplace=True or even test this in particular.

Before I spend a lot of time on this, could you indicate a safe way to approach it? Shall I really drop the inplace argument in all those methods, delete all the tests against the feature and switch to not-inplace in the tests that use it?

from pandas.

yeah i'd be ok with that

from pandas.