Pandas find first nan value by rows and return column name

Question

I have a dataframe like this

>>df1 = pd.DataFrame({'A': ['1', '2', '3', '4','5'],
              'B': ['1', '1', '1', '1','1'],
              'C': ['c', 'A1', None, 'c3',None],
              'D': ['d0', 'B1', 'B2', None,'B4'],
              'E': ['A', None, 'S', None,'S'],
              'F': ['3', '4', '5', '6','7'],
              'G': ['2', '2', None, '2','2']})
>>df1

   A  B     C     D     E  F     G
0  1  1     c    d0     A  3     2
1  2  1    A1    B1  None  4     2
2  3  1  None    B2     S  5  None
3  4  1    c3  None  None  6     2
4  5  1  None    B4     S  7     2

and I drop the rows which contain nan values df2 = df1.dropna()

   A  B     C     D     E  F     G   
1  2  1    A1    B1  None  4     2
2  3  1  None    B2     S  5  None
3  4  1    c3  None  None  6     2
4  5  1  None    B4     S  7     2

This is a dropped dataframe due to those rows contain nan values. However,I wanna know why they be dropped? Which column is the "first nan value column" made the row been dropped ? I need a dropped reason for report.

the output should be

['E','C','D','C']

I know I can do dropna by each column then record it as the reason but it's really non-efficient.

Is any more efficient way to solve this problem? Thank you

Answer 1

I think you can create boolean dataframe by DataFrame.isnull , then filter by boolean indexing with mask where are at least one True by any and last idxmax - you get column names of first True values of DataFrame :

booldf = df1.isnull()
print (booldf)
       A      B      C      D      E      F      G
0  False  False  False  False  False  False  False
1  False  False  False  False   True  False  False
2  False  False   True  False  False  False   True
3  False  False  False   True   True  False  False
4  False  False   True  False  False  False  False

print (booldf.any(axis=1))
0    False
1     True
2     True
3     True
4     True
dtype: bool

print (booldf[booldf.any(axis=1)].idxmax(axis=1))
1    E
2    C
3    D
4    C
dtype: object

Answer 2

I would use a combination of itertools and numpy.where , along with pd.DataFrame.isnull :

>>> df1.isnull()
       A      B      C      D      E      F      G
0  False  False  False  False  False  False  False
1  False  False  False  False   True  False  False
2  False  False   True  False  False  False   True
3  False  False  False   True   True  False  False
4  False  False   True  False  False  False  False
>>> from itertools import *
>>> r,c = np.where(df1.isnull().values)
>>> first_cols = [next(g)[1] for _, g in groupby(izip(r,c), lambda t:t[0])]
>>> df1.columns[first_cols]
Index([u'E', u'C', u'D', u'C'], dtype='object')
>>> 

For Python 2, use izip from itertools , and in Python 3 simply use built-in zip .