CODEWITHSUNDEEP
pythonpandasgroup-by
Kvothe
Kvothe

Reputation: 1393

combine rows into single cell

I currently have a dataframe (df) like this:

name    info
alpha   foo,bar
alpha   bar,foo
beta    foo,bar
beta    bar,foo
beta    baz,qux

I'm looking to create a dataframe like this:

name    info
alpha   (foo,bar),(bar,foo)
beta    (foo,bar),(bar,foo),(baz,qux)

I'm getting close with groupby.apply(list). Eg.

new_df=df.groupby('name')['info'].apply(list)

However, I can't seem to figure out how to get the output in the original dataframe format. (i.e with two columns (like the example)

I think I need reset_index and unstack? Appreciate any help!

Upvotes: 3

Views: 1676

Answers (3)

rnso
rnso

Reputation: 24593

Try following using for loop:

uniqnames = df.name.unique() # get unique names
newdata = []                 # data list for output dataframe
for u in uniqnames:          # for each unique name
    subdf = df[df.name == u] # get rows with this unique name
    s = ""
    for i in subdf['info']:
        s += "("+i+"),"      # join all info cells for that name
    newdata.append([u, s[:-1]]) # remove trailing comma from infos & add row to data list

newdf = pd.DataFrame(data=newdata, columns=['name','info'])
print(newdf)

Output is exactly as desired:

    name                           info
0  alpha            (foo,bar),(bar,foo)
1   beta  (foo,bar),(bar,foo),(baz,qux)

Upvotes: 2

BENY
BENY

Reputation: 323316

IIUC

df.assign(info='('+df['info']+')').groupby('name')['info'].apply(','.join).to_frame('info')
Out[267]: 
                                info
name                                
alpha            (foo,bar),(bar,foo)
beta   (foo,bar),(bar,foo),(baz,qux)



#df.assign(info='('+df['info']+')')# adding the ( and ) for your single string to match with the out put 
#groupby('name')# group by the name, you need merge info under the same name
#apply(','.join).to_frame('info') # this will combine each info into one string under the same group

Upvotes: 1

Scott Boston
Scott Boston

Reputation: 153500

IIUC:

df = pd.DataFrame({'name':['alpha']*2+['beta']*3,
                   'info':[['foo','bar'],['bar','foo'],
                           ['foo','bar'],['bar','foo'],
                           ['baz','qux']]})
print(df)

Inuput:

         info   name
0  [foo, bar]  alpha
1  [bar, foo]  alpha
2  [foo, bar]   beta
3  [bar, foo]   beta
4  [baz, qux]   beta

Now, groupby and apply then reset_index() to get back to dataframe:

new_df = df.groupby('name')['info'].apply(list)

new_df = new_df.reset_index()

print(new_df)

Output:

    name                                  info
0  alpha              [[foo, bar], [bar, foo]]
1   beta  [[foo, bar], [bar, foo], [baz, qux]]

Upvotes: 0

Related Questions