Pandas SettingWithCopyWarning when modifying a column added to df after read_csv

Code Sample, a copy-pastable example if possible

# Your code here
df = pd.read_csv('package_infos.csv')
m = df.shape[0]
df['etags'] = pd.Series(dtype=str, index=range(m))
df['etags'][0] = "Hello world"

Download package_infos.csv from here: https://raw.githubusercontent.com/jamesqo/nuget_recommender/b05ef43a01e54b61851f51708cdd6dce3b2afe33/package_infos.csv

Doing df.loc[0, 'etags'] = "Hello world" seems to remove the warning, but it's slowing down my app significantly.

Comment From: jamesqo

I forgot to mention, this does not seem to happen if the df is created manually instead of calling read_csv.

Comment From: ZhuBaohe

df['etags'] = ' ' df.loc[0, 'etags'] = "Hello world"

Comment From: jreback

this is well documented: http://pandas.pydata.org/pandas-docs/stable/indexing.html#returning-a-view-versus-a-copy

it IS slower because indexing must determie if you are chained indexing, which requires garbage collection (now it normally doesn't even touch this code unless it fails a quick test, where more comprehensive checks happen). so if you avoid chained indexing you won't have a problem.