ENH: Add inplace argument to pandas.DataFrame.insert

[bhvieira]

Feature Type

• Adding new functionality to pandas

• Changing existing functionality in pandas

• Removing existing functionality in pandas

Problem Description

Currently, most methods only make in-place alterations when explicitly told to (default is inplace = False). pandas.DataFrame.insert is an exception, working in true OOP-way. There is not even an inplace argument for it.

I don't know if there is any justification for this choice, but I would expect this method to work as the others.

If this is desirable, I can work on this as well.

Feature Description

Add a new inplace parameter for the pandas.DataFrame.insert. Initally, its default will be inplace = True, with a deprecation warning. In a future breaking change, it could be changed to inplace = False

df = df.insert(1, "A", col, inplace=False) 

Alternative Solutions

I'm not aware of an alternative solution, pandas.DataFrame.insert returns None, like other methods with the inplace=True parameter.

Additional Context

No response

[phofl]

Hi, thanks for the issue.

I am -1 on this. What would be the advantage of making a copy here? There are other ways of emulating an insert if you want to do method chaining

[bhvieira]

@phofl the advantage is consistency to the rest of the methods, imo

[jreback]

-1 as well

and we r actively going to be deprecating inplace soon

[bhvieira]

@jreback I see, but then insert will continue to do inplace operations (while users often don't expect it)

[jreback]

it's been the case since the very beginning of time (update as well)

[jbrockmendel]

agreed on not adding inplace, but id be +1 on adding copy and returning a DataFrame object, xref #48141

[topper-123]

-1 on this as well.

There seem to be relative agreement among the core devs in not going forward with this, so I'm closing this. Thanks for the proposal though @bhvieira.