How to combine multiple rows of pandas dataframe into one between two other row values python3? - python-3.x

Question

I have a pandas dataframe with a single column that contains name, address, and phone info separated by blank or na rows like this:
data
0 Business name one
1 1234 address ln
2 Town, ST 55655
3 [phone]
4 nan
5 Business name two
6 5678 address dr
7 New Town, ST 55677
8 nan
9 Business name three
10 nan
and so on...
What I want is this:
Name Addr1 Addr2 Phone
0 Business name one 1234 address ln Town, ST 55655 [phone]
1 Business name two 5678 address dr New Town, ST 55677
2 Business name three
I am using python 3 and have been stuck, any help is much appreciated!

Answer 1

You can use:
create groups for each row with isnull and cumsum
for align with non NaN rows add reindex
remove NaNs by dropna, set_index to MultiIndex with cumcount
reshape by unstack
a = df['data'].isnull().cumsum().reindex(df.dropna().index)
print (a)
0 0
1 0
2 0
3 0
5 1
6 1
7 1
9 2
Name: data, dtype: int32
df = df.dropna().set_index([a, a.groupby(a).cumcount()])['data'].unstack()
df.columns = ['Name','Addr1','Addr2','Phone']
print (df)
Name Addr1 Addr2 Phone
data
0 Business name one 1234 address ln Town, ST 55655 [phone]
1 Business name two 5678 address dr New Town, ST 55677 None
2 Business name three None None None
If there is multiple address is possible create columns dynamically:
df.columns = ['Name'] +
['Addr{}'.format(x+1) for x in range(len(df.columns) - 2)] +
['Phone']

Answer 2

df['group']=df['data'].str.contains('Business').cumsum().replace({True:1}).ffill()
df1=df.groupby('group')['data'].apply(list).apply(pd.Series).dropna(axis=1,thresh =1)
df1.columns=['Name','Addr1','Addr2','Phone']
df1
Out[1221]:
Name Addr1 Addr2 \
group
1.0 Business name one 1234 address ln Town, ST 55655
2.0 Business name two 5678 address dr New Town, ST 55677
3.0 Business name three NaN NaN
Phone
group
1.0 [phone]
2.0 NaN
3.0 NaN