DEPR Deprecate mapper argument in rename?

[MarcoGorelli]

xref #40979 (comment)

thanks @MarcoGorelli i realize now why this is convoluted, will be glad to remove this in 2.0

@jreback not sure I understand what you mean - as in, to remove mapper in 2.0 and only support the index= and columns= calls? (and hence, introduce a deprecation warning in time for 1.4)

[jreback]

no what u mean is to fix the structure of the impl here
maybe we just need to make the structure in generic a private function to avoid having to handle the signature like this

[MarcoGorelli]

OK, so rename rename in pandas/core/generic.py to _rename, and go back to the old signature for Series.rename, i.e.

    def rename(
        self,
        index=None,
        *,
        axis=None,
        copy=True,
        inplace=False,
        level=None,
        errors="ignore",
    ):

? Like this, mypy wouldn't complain about the incompatibility.
Wish I'd thought of this actually, it seems like a better solution than what I did.

If so, I think that could be done straight away, rather than having to wait for 2.0, it wouldn't alter anything user-facing

[jreback]

yeah i think that's right. its not worth trying to shoe-horn this in for methods which have a different signatture in series vs dataframe just to placate mypy, better to just have the impl be private.

[Varun270]

@MarcoGorelli I tried to understand this issue and so far I understood that in pandas/core/generic.py we have to make the rename function private and make it _rename. Is that all? I am sure I am missing out on a lot.

[MarcoGorelli]

Hey @Varun270 !

So, there's a few things to do:

1. rename rename in pandas/core/generic.py to _rename
2. update rename (in both pandas/core/frame.py and pandas/core/series.py to call super()._rename, rather than super().rename)
3. revert some of the changes I made in TYP: Signature of "rename" incompatible with supertype "NDFrame" #40979 so that the signature of Series.rename becomes:

    def rename(
        self,
        index=None,
        *,
        axis=None,
        copy=True,
        inplace=False,
        level=None,
        errors="ignore",
    ) ->  Series | None:

If you open a PR (even if you're not sure about everything) and tag me I can help you along with this

[Varun270]

Hey @Varun270 !

So, there's a few things to do:

1. rename rename in pandas/core/generic.py to _rename
2. update rename (in both pandas/core/frame.py and pandas/core/series.py to call super()._rename, rather than super().rename)
3. revert some of the changes I made in TYP: Signature of "rename" incompatible with supertype "NDFrame" #40979 so that the signature of Series.rename becomes:

    def rename(
        self,
        index=None,
        *,
        axis=None,
        copy=True,
        inplace=False,
        level=None,
        errors="ignore",
    ) ->  Series | None:

If you open a PR (even if you're not sure about everything) and tag me I can help you along with this

I understood the first two points but will require your guidance on 3rd point as I don't know how can I revert changes done by someone else.

[Varun270]

Currently in Series.rename the function rename looks like this -

def rename(
        self,
        index=None,
        *,
        axis=None,
        copy=True,
        inplace=False,
        level=None,
        errors="ignore",
    ):

So after reverting it would look like this -

    def rename(
        self,
        index=None,
        *,
        axis=None,
        copy=True,
        inplace=False,
        level=None,
        errors="ignore",
    ) ->  Series | None:

Is that what you are saying?

[MarcoGorelli]

Currently in Series.rename the function rename looks like this -

Make sure to pull the latest changes, it doesn't quite look like that at the moment. Anyway,

So after reverting it would look like this -

Yes, it should look like that at the end

[Varun270]

Currently in Series.rename the function rename looks like this -

Make sure to pull the latest changes, it doesn't quite look like that at the moment. Anyway,

So after reverting it would look like this -

Yes, it should look like that at the end

Okay, so I guess I would have to revert the changes first then rename the rename functions. How am I supposed to do that? Like should I do the usual git revert commit_id?

[MarcoGorelli]

it's up to you how you do it

[Varun270]

Alright, will do it!

[Varun270]

@MarcoGorelli I found the commit that needed to be reverted but you have done some other changes in that commit apart from adding args to function rename. Here have a look at this - 40ecf6e
so should I proceed ahead since reverting it would revert some additional changes as well?

[MarcoGorelli]

What can be reverted are:

• changes to Series.rename (can just revert them all, and change super().rename to super()._rename
• the extra tests in pandas/tests/series/methods/test_rename.py are no longer necessary

Might be easier to do this manually rather than reverting a commit using git. Up to you though, if you open a PR I'll take a look

[Varun270]

What can be reverted are:

• changes to Series.rename (can just revert them all, and change super().rename to super()._rename
• the extra tests in pandas/tests/series/methods/test_rename.py are no longer necessary

Might be easier to do this manually rather than reverting a commit using git. Up to you though, if you open a PR I'll take a look

Yeah, even I was thinking that manually doing these changes would be much easier for me. Will create a PR.