Pythonic way to set entire column to value Pandas (SettingWithCopyWarning)

Question

I want to set an entire column to a single string value. When doing so I get the (ever so popular) SettingWithCopy. I've tried to search SO before posting about this specific issue.

import pandas as pd
import numpy as np
dfp = pd.DataFrame({'A' : [1,21,8,44,np.NaN,6,75,8,44,999], 
                    'B' : [1,1,3,5,0,0,np.NaN,9,np.NaN,0], 
                    'C' : ['AA1233445','AA1233445', 'rmacy','Idaho Rx','Ab123455','TV192837','RX','Ohio Drugs','RX12345','USA Pharma'], 
                    'D' : [123456,123456,1234567,12345678,12345,12345,12345678,123456789,1234567,np.NaN],
                    'E' : ['Assign','Assign','Hello','Ugly','Appreciate','Undo','Testing','Unicycle','Pharma','Unicorn',]})
print(dfp)

new_df_to_show_copy = dfp.loc[(dfp['A']>100) |(dfp['E']=='Unicorn')]
new_df_to_show_copy['Reason'] = 'what is with the copy warning'

Now I can get rid of the warning with

new_df_to_show_copy = dfp.loc[(dfp['A']>100) |(dfp['E']=='Unicorn')].copy() <--Notice copy()
new_df_to_show_copy['Reason'] = 'what is with the copy warning'

And I know I can get rid of the warning with pd.options.mode.chained_assignment = None but I feel like that's "cheating". I'm looking at the documentation, but can't find a minimal way for setting an entire column to a single value without adding .copy() or suppressing the warning. What's the best way to do that?

It doesn't look to me like setting the entire column has anything to do with the problem. The problem is setting values at all on a copy created in a way that people often don't realize makes a copy. — user2357112, Jun 09 '17 at 20:48
I agree that setting the entire column is not the problem. Was just wondering if there was a way to do so without enacting the warning. I'm not even saying the warning is a bad thing, but just curious! — MattR, Jun 09 '17 at 20:59
There's an answer to a similar question here: https://stackoverflow.com/a/39798005/7306999 — Xukrao, Jun 09 '17 at 21:21
I'm not sitting in front of a computer right now, but what if you use `new_df_to_show_copy.loc[:,'Reason'] = '...'` — elPastor, Jun 10 '17 at 06:37
Sorry I have not responded lately. Been busy on more important projects. @pshep123, your example returns the same warning. I thought that would work as well! — MattR, Jun 16 '17 at 12:56

score 0 · Accepted Answer · answered Aug 25 '17 at 15:58

I was working on a similar requirement earlier today and I came across assign. I got rid of the copy warning without using pd.options.mode.chained_assignment = None. Here was my solution:

new_df_to_show_copy = dfp.loc[(dfp['A']>100) |(dfp['E']=='Unicorn')]
new_df_to_show_copy = new_df_to_show_copy.assign(Reason = 'what is with the copy warning')

# output:

       A    B           C   D        E                         Reason
9  999.0  0.0  USA Pharma NaN  Unicorn  what is with the copy warning

assign will leave a copy of the dataframe, and there is no inplace=True parameter. so just reassigning the value worked for me.

Pythonic way to set entire column to value Pandas (SettingWithCopyWarning)

1 Answers1

Linked