ENH: add errors='raise' option to rename

[max-sixty]

Is there a reason that .rename doesn't raise if keys are supplied that don't exist? This caught me out a couple of times:

In [14]: import pandas as pd

In [15]: df=pd.DataFrame({'a': [1,2], 'b': [3,4]})

In [16]: df
Out[16]: 
   a  b
0  1  3
1  2  4

In [17]: df.rename(columns={'a': 'c'})
Out[17]: 
   c  b
0  1  3
1  2  4
# ok

In [18]: df.rename(columns={'d': 'c'})
Out[18]: 
   a  b
0  1  3
1  2  4

# this should raise?

[TomAugspurger]

I suspect it is intentional. Either way, the behavior should be clarified in the docs. Missing keys (in columns but not in the dict) are explicitly documented.

I know I've used this in the past. For example a bunch of CSVs with variations on the same items. I'll make one dictionary with all the mappings

renamer = {'zip': 'zip_code', 'postal code': 'zip_code', ...}`

And use that same renamer on many dataframes.

[max-sixty]

OK. You know this better than me - do you think this is too much vs the increase in explicitness?

df.rename(columns={col: renamer[col] for col in renamer if col in df.columns})

[jorisvandenbossche]

For DataFrame.drop, the errors keyword was added some time ago:

errors : {'ignore', 'raise'}, default 'raise'
    If 'ignore', suppress error and existing labels are dropped.

    .. versionadded:: 0.16.1

We could maybe add this to rename as well to have similar functionality? (although the default value would be different for rename)

[TomAugspurger]

Having a keyword to control this would be great.

Another option is to revisit the "tell people to use relabel discussion". If we implemented it there we could have the default be to raise.

On Jun 30, 2016, at 18:02, Joris Van den Bossche notifications@github.com wrote:

For DataFrame.drop, the errors keyword was added some time ago:

errors : {'ignore', 'raise'}, default 'raise'
If 'ignore', suppress error and existing labels are dropped.

.. versionadded:: 0.16.1

We could maybe add this to rename as well to have similar functionality? (although the default value would be different for rename)

—
You are receiving this because you commented.
Reply to this email directly, view it on GitHub, or mute the thread.

[ghost]

Is there a particular reason why the default value would be "ignore" for rename? If I'm renaming keys in a mapping but one isn't there, I normally anticipate a KeyError. A deprecation period would be necessary, of course -- I think a similar change has occurred recently with missing columns for loc.

[WillAyd]

Historical reasons. I think raising would be preferable though correct would require deprecation cycle

[TomAugspurger]

To be clear, I don't think there's a consensus for deprecating the current behavior. Just adding the option to raise when not present.

[bala-analytics]

This issue is closed but the problem still persists.

[TomAugspurger]

@bala-analytics the original example correctly raises with df.rename(columns={'d': 'c'}, errors="raise").

[bala-analytics]

Thanks. Sorry missed that errors = ‘raise’ option From: Tom Augspurger <notifications@github.com> Sent: 16 June 2020 21:27 To: pandas-dev/pandas <pandas@noreply.github.com> Cc: Bhaskaran, Balakrishnan GSUK-PTX/D/S <B.Bhaskaran@shell.com>; Mention <mention@noreply.github.com> Subject: Re: [pandas-dev/pandas] ENH: add errors='raise' option to rename (#13473) @bala-analytics<https://eur02.safelinks.protection.outlook.com/?url=https%3A%2F%2Fgithub.com%2Fbala-analytics&data=02%7C01%7Cb.bhaskaran%40shell.com%7Cbc50fd93bf194f3efdfe08d81233ae88%7Cdb1e96a8a3da442a930b235cac24cd5c%7C0%7C0%7C637279360500483631&sdata=CZVLf7jrfBzbxr4IkiXVQNGkABTv2g9bmxodCM2fa9o%3D&reserved=0> the original example correctly raises with df.rename(columns={'d': 'c'}, errors="raise"). — You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub<https://eur02.safelinks.protection.outlook.com/?url=https%3A%2F%2Fgithub.com%2Fpandas-dev%2Fpandas%2Fissues%2F13473%23issuecomment-644994081&data=02%7C01%7Cb.bhaskaran%40shell.com%7Cbc50fd93bf194f3efdfe08d81233ae88%7Cdb1e96a8a3da442a930b235cac24cd5c%7C0%7C0%7C637279360500493620&sdata=ZTTB3a%2F86Fc8%2BlJSuHWyjNiWS6gSjGyMQEkCqRxuMIU%3D&reserved=0>, or unsubscribe<https://eur02.safelinks.protection.outlook.com/?url=https%3A%2F%2Fgithub.com%2Fnotifications%2Funsubscribe-auth%2FAO6KXJRJB5BLM5FQQZ3JGG3RW7ISZANCNFSM4CHCIO6A&data=02%7C01%7Cb.bhaskaran%40shell.com%7Cbc50fd93bf194f3efdfe08d81233ae88%7Cdb1e96a8a3da442a930b235cac24cd5c%7C0%7C0%7C637279360500493620&sdata=UeQR5zHkJKZMO9szJhmIE82ebBcwnCDDtDr9J%2BR9ONU%3D&reserved=0>.