CODEWITHSUNDEEP
pythonpandasdataframe
Webcy
Webcy

Reputation: 181

Get first occurrence value in each column

Here is my df. I want to get the first value in each column which contains (F)

>>> d = {0: ['1', '2(F)', '6', '8', '5'], 
    1: ['8(F)', '6', '8', '4(F)', '4'], 
    2: ['1', '6', '8(F)', '4(F)', '5'],
    3: ['1', '8', '8', '1', '5']}
>>> df = pd.DataFrame(data=d)
>>> df
      0     1     2  3
0     1  8(F)     1  1
1  2(F)     6     6  8
2     6     8  8(F)  8
3     8  4(F)  4(F)  1
4     5     4     5  5

And the result should look like this

0    2(F)
1    8(F)
2    8(F)
3     NaN

But when I used the code below, I received some errors

>>> mask = df.apply(lambda x: x.str.contains('F'))
>>> a = mask.idxmax().where(mask.any())
>>> print(df[a])

KeyError: '[nan] not in index'

Upvotes: 4

Views: 541

Answers (4)

Rajat Jain
Rajat Jain

Reputation: 2022

Here's, a one-liner but it doesn't give out answer for 4th row:

df.replace("\d$", np.nan, regex=True).dropna(how='all', axis=1).apply(lambda x: x.dropna().iloc[0], 0)

It clears all elements other than \F, then for each column it finds out first non-empty elements.

Upvotes: 0

piRSquared
piRSquared

Reputation: 294218

applymap, lookup

mask = df.applymap(lambda x: '(F)' in x)
vals = df[mask].lookup(mask.idxmax(), df.columns)
pd.Series(vals, df.columns)

0    2(F)
1    8(F)
2    8(F)
3     NaN
dtype: object

Numpy Variant

Over engineered

from numpy.core.defchararray import find

v = df.values.astype(str)
m = find(v, '(F)') >= 0
i = m.argmax(0)
j = np.arange(v.shape[1])

pd.Series(np.where(m[i, j], v[i, j], np.nan), df.columns)

Upvotes: 4

BENY
BENY

Reputation: 323226

Here is one way 

mask = df.applymap(lambda x: '(F)' in x)

df[mask].bfill().iloc[0,]
Out[624]: 
0    2(F)
1    8(F)
2    8(F)
3     NaN
Name: 0, dtype: object

Upvotes: 5

jezrael
jezrael

Reputation: 862511

Use numpy indexing for get values by idxmax and last add where:

mask = df.apply(lambda x: x.str.contains('F', na=False))
a = mask.idxmax()   
s = pd.Series(df.values[a, a.index]).where(mask.any())
print(s)
0    2(F)
1    8(F)
2    8(F)
3     NaN
dtype: object

Another solution with reshape by DataFrame.stack, filtering and get first value by GroupBy.first, last add non exist values by Series.reindex:

s = df.stack()
s = s[s.str.contains('F', na=False)].groupby(level=1).first().reindex(df.columns)
print (s)
0    2(F)
1    8(F)
2    8(F)
3     NaN
dtype: object

Upvotes: 4

Related Questions