I have a df that looks like this:
pd.DataFrame.from_dict({'master_feature':['ab',float('NaN'),float('NaN')], 'feature':[float('NaN'),float('NaN'),'pq'], 'epic':[float('NaN'),'fg',float('NaN')]})
I want to create a new column named promoted
from the columns master_feature, epic, and feature:
value of promoted
will be :
master feature
if adjacentmaster_feature
column value is not null.feature
if adjacentfeature
column value is not null ,and likewise forepic
something like:
df.promoted = 'master feature' if not pd.isnull(df.master_feature) elseif 'feature' if not pd.isnull(df.feature) elseif 'epic' pd.isnull(df.epic) else 'Na'
how can I achieve this using a df.apply
?
is it much more efficient if I use np.select
?
Advertisement
Answer
np.select
is the way to go. Try below . . . I think I got the logic correct based on your question. Also, there is some discrepancy in your logic: “feature if adjacent feature column value is not null ,and likewise for epic” is not the same as “elseif ‘epic’ pd.isnull(df.epic)” So I went with if df['epic'] is not null then 'epic'
Let me know if that is correct.
cond = [~df['master_feature'].isna(), # if master_feater is not null then 'master feater' ~df['feature'].isna(), # if feature is not null then 'feature ~df['epic'].isna()] # if epic is not null then 'epic' choice = ['master feature', 'feature', 'epic'] df['promoted'] = np.select(cond, choice, np.nan) master_feature feature epic promoted 0 ab NaN NaN master feature 1 NaN NaN fg epic 2 NaN pq NaN feature