Convert a single row to a sparse matrix of NaNs with pandas

user_4909087 0
Viewed 231

Given a dataframe of the form:

   col1  col2  col3  col4
0     1     2     3     4

I want to move each element at position [0, i] in the current dataframe to position [i, i] in the new one. So, for this input, my desired output will be:

   col1   col2  col3  col4    
0   1.0    NaN   NaN   NaN
1   NaN    2.0   NaN   NaN
2   NaN    NaN   3.0   NaN
3   NaN    NaN   NaN   4.0

Note that the position has nothing to do with the values of the elements themselves.

I know I can do this with numpy by calling np.diag, but that returns a 2D array of 0s, and I cannot replace 0s with NaNs, because some of the diagonal values might also be replaced, if they are also 0s.

What's the simplest way to do this with pandas?

pandas dataframe nan python
5 Answers

IIUIC, you could use a comprehension generator.

In [1115]: df
Out[1115]:
   col1  col2  col3  col4
0     1     2     3     4

In [1116]: pd.DataFrame({c: v} for c, v in df.iloc[0].items())
Out[1116]:
   col1  col2  col3  col4
0   1.0   NaN   NaN   NaN
1   NaN   2.0   NaN   NaN
2   NaN   NaN   3.0   NaN
3   NaN   NaN   NaN   4.0
user_21372550
Related