Print Row and Column Header if column/row is not NaN Pandas-CodePudding

It's a weird question - but can y'all think of a good way to just print the rows or a list of the rows and their corresponding column headers if the dataframe cell is not NaN?

Imagine a dataframe like this:

     col1   col2    col3    col4
1    1      NaN     2       NaN
2    NaN    NaN     1       2
3    2      NaN     NaN     1

Result should look something like this:

1    [col1: 1, col3: 2]
2    [col3: 1, col4: 2]
3    [col1: 2, col4: 1]

Thanks in advance!

CodePudding user response：

You can transpose the dataframe, and for each row, drop NaNs and convert to dict:

out = df.T.apply(lambda x: dict(x.dropna().astype(int)))

Output:

>>> out
1    {'col1': 1, 'col3': 2}
2    {'col3': 1, 'col4': 2}
3    {'col1': 2, 'col4': 1}
dtype: object

CodePudding user response：

Let us try stack

df.stack().reset_index(level=0).groupby('level_0')[0].agg(dict)
Out[184]: 
level_0
1    {'col1': 1.0, 'col3': 2.0}
2    {'col3': 1.0, 'col4': 2.0}
3    {'col1': 2.0, 'col4': 1.0}
Name: 0, dtype: object

CodePudding user response：

combine agg(dict) and list comprehension

d = [{k:v for k, v in x.items() if v == v } for x in df.agg(dict,1)]

[{'col1': 1.0, 'col3': 2.0},
 {'col3': 1.0, 'col4': 2.0},
 {'col1': 2.0, 'col4': 1.0}]