CODEWITHSUNDEEP
pythonpandas
noobtalgic
noobtalgic

Reputation: 60

Search elements of list in a Pandas column and return the elements to a new column if matched

I am working towards cleaning a pandas DataFrame column. The column has words which I would like find and extract from a list.

Below is what I 've got. But it does not return multiple matches. Below is an example.

data = {'A':['abc 1 foo','def 1,bar','abc 2','def 2', 'abc 1/def 1 baz', 'abc 1,def 1']}
l = ['abc 1', 'def 1']

df = pd.DataFrame(data)

for idx, row in df.iterrows():
    for x in l:
        if x in row.A:
            df.loc[idx, 'new_col'] = x```

Actual output:

A            new_col
abc 1            abc 1
def 1            def 1
abc 2            NaN
def 2            NaN
abc 1/def 1      def 1
abc 1,def 1      def 1

Expected output:

A            new_col
abc 1            abc 1
def 1            def 1
abc 2            NaN
def 2            NaN
abc 1/def 1      abc 1,def 1
abc 1,def 1      abc 1,def 1

Note: the seperator in col A could be anything('/', ';') but seperator in new_col should be fixed.

Upvotes: 2

Views: 80

Answers (1)

jezrael
jezrael

Reputation: 862611

Use str.findall with Series.str.join join values of list with pattern joined by | for regex OR and \b for word boundaries:

pat = '|'.join(r"\b{}\b".format(x) for x in l)
df['new_col'] = df['A'].str.findall(pat).str.join(',')
print (df)
                 A      new_col
0        abc 1 foo        abc 1
1        def 1,bar        def 1
2            abc 2             
3            def 2             
4  abc 1/def 1 baz  abc 1,def 1
5      abc 1,def 1  abc 1,def 1

If need NaNs instead empty strings use numpy.where:

pat = '|'.join(r"\b{}\b".format(x) for x in l)
s = df['A'].str.findall(pat)
df['new_col'] = np.where(s.astype(bool), s.str.join(','), np.nan)
print (df)
                 A      new_col
0        abc 1 foo        abc 1
1        def 1,bar        def 1
2            abc 2          NaN
3            def 2          NaN
4  abc 1/def 1 baz  abc 1,def 1
5      abc 1,def 1  abc 1,def 1

Upvotes: 1

Related Questions