How can I group a pandas dataframe based on a string hierarchy?

December 2, 2021

Given a dataframe like this:

Client	Priority
Client 1	Low
Client 1	Medium
Client 2	Low
Client 2	High
Client 3	Low

I want to group by client and select the highest priority (Low, Medium, High) for each client. It should look something like this:

Client	Priority
Client 1	Medium
Client 2	High
Client 3	Low

>Solution :

convert priority into a categorical column before grouping:

dtype = pd.CategoricalDtype(['High', 'Medium', 'Low'], ordered = True)

df['Priority'] = df.Priority.astype(dtype)

df.sort_values(['Client', 'Priority']).groupby('Client', sort = False).head(1)

     Client Priority
1  Client 1   Medium
3  Client 2     High
4  Client 3      Low

if you do not want to convert to category, you can map the positions instead:

(df.assign(temp = df.Priority.map({'Low':3,'Medium':2, 'High':1}))
   .sort_values(['Client', 'temp'])
   .drop(columns='temp')
   .groupby('Client')
   .head(1)
)
     Client Priority
1  Client 1   Medium
3  Client 2     High
4  Client 3      Low