Htykuut

I have a pandas dataframe:

street_name        eircode

Malborough Road    BLT12

123 Fake Road      NaN

My Street          NaN

I would like to create another column called unique based on these conditions:

1. If it has eircode, return 'yes' in the unique column, THEN


2. If it doesn't have an eircode, check the first string in the street_name:
   
   
   
   
   • if the first string is a digit, return 'yes' in the unique column
   
   
   • if it is not, return 'no' in the unique column
   
   
   
   

I came up with this solution where:

1. I changed the data types to string for both columns street_name and eircode
   


2. Get the first string using a lambda function


3. Defined a tagging function to be applied to the data frame

# change data types

df['eircode'] = df['eircode'].astype('str')
df['street_name'] = df['street_name'].astype('str')

# get the first string from street_name column
df['first_str'] = df['street_name'].apply(lambda x: x.split()[0])

def tagging(x):

if x['eircode'] != 'nan':

    return 'yes'

elif x['first_str'].isdigit() == True:

    return 'yes'

else:

    return 'no'



df['unique'] = df.apply(tagging, axis=1)

The issue with this is that I have to change the data type and then have to make separate column. Is there a more elegant way or a more concise way to achieve the same result?

You can supply those separate conditions using the | operator, then map the resulting Boolean array to yes and no. The first condition just looks wether eircode is null, and the second uses a regex to check that street_name starts with a digit:

df['unique'] = ((~df.eircode.isnull()) | (df.street_name.str.match('^[0-9]'))).map({True:'yes',False:'no'})

>>> df

       street_name eircode unique

0  Malborough Road   BLT12    yes

1    123 Fake Road     NaN    yes

2        My Street     NaN     no

With Pandas, it's best to use column-wise calculations; apply with a custom function represents an inefficient, Python-level row-wise loop.

df = pd.DataFrame({'street_name': ['Malborough Road', '123 Fake Road', 'My Street'],

                   'eircode': ['BLT12', None, None]})



cond1 = df['eircode'].isnull()

cond2 = ~df['street_name'].str.split(n=1).str[0].str.isdigit()



df['unique'] = np.where(cond1 & cond2, 'no', 'yes')



print(df)



  eircode      street_name unique

0   BLT12  Malborough Road    yes

1    None    123 Fake Road    yes

2    None        My Street     no