CODEWITHSUNDEEP
pythonpandas
NineWasps
NineWasps

Reputation: 2253

Pandas: using groupby with condition

I have df

i,Unnamed: 0,ID,active_seconds,subdomain,search_term,period,code,buy    
0,56574,08cd0141663315ce71e0121e3cd8d91f,6,market.yandex.ru,None,515,100.0,1.0  
1,56576,08cd0141663315ce71e0121e3cd8d91f,26,market.yandex.ru,None,515,100.0,1.0 
2,56578,08cd0141663315ce71e0121e3cd8d91f,14,market.yandex.ru,None,515,100.0,1.0 
3,56579,08cd0141663315ce71e0121e3cd8d91f,2,market.yandex.ru,None,515,100.0,1.0  
4,56581,08cd0141663315ce71e0121e3cd8d91f,8,market.yandex.ru,None,515,100.0,1.0  
5,56582,08cd0141663315ce71e0121e3cd8d91f,32,market.yandex.ru,None,515,100.0,1.0 
6,56583,08cd0141663315ce71e0121e3cd8d91f,16,market.yandex.ru,None,515,100.0,1.0 
7,56584,7602962fb83ac2e2a0cb44158ca88464,4,market.yandex.ru,None,515,100.0,2.0  
8,56585,7602962fb83ac2e2a0cb44158ca88464,10,market.yandex.ru,None,515,100.0,2.0 
9,56639,7602962fb83ac2e2a0cb44158ca88464,2,market.yandex.ru,None,516,100.0,2.0

I need to count sum of active_seconds to every ID, 

df.groupby(['ID', 'buy']).agg({'active_seconds': sum}).rename(columns={'active_seconds': 'count_sec'}).reset_index()

But I need do it, if buy == 2 or buy == 3, if buy == 1, I need to print date from this df.

ID  date    buy
7602962fb83ac2e2a0cb44158ca88464    01.01.2016  1
bc8a731e4c7e6f6b96e56ebe7f766bcd    10.02.2016  1
a703114aa8a03495c3e042647212fa63    20.02.2016  2

How can I do that?

Upvotes: 0

Views: 110

Answers (1)

Shahram
Shahram

Reputation: 814

If I understand your question correctly, you want to join with a different data frame when buy == 1. Assuming the first data frame is named df and the second data frame that contains dates is named df2 then this is my proposed solution:

df.groupby(['ID', 'buy']).agg({'active_seconds': sum}).rename(columns={'active_seconds': 'count_sec'}).reset_index().merge(df2, how='left', on=['ID','buy']).apply(lambda x: x['date'] if x['buy']==1 else x['count_sec'],axis=1)

Upvotes: 1

Related Questions