Fill matrix by data from pandas.DataFrame, skip NaN

Question

I want to fill the matrix ref using pd.DataFrame xxx but skip the NaN .

print xxx
OUT >> 
   intensity name  rowtype1  rowtype2
0        100    A         1       4.0
1        200    A         2       NaN
2        300    B         3       5.0

Then I fill the matrix by ref[rowtype,col] = intensity where I have 2 rowtype .

ref = np.zeros(shape=(7,4))
for idx, inte, name, r1, r2 in xxx.itertuples():
    ref[r1,idx] = inte
    ref[r2,idx] = inte # error because of NaN in rowtype2

print ref

How can I skip NaN here? I know one way to use drop.na() but it has to create new dataframe which has rowtype2 and intensity . I would like to have quick simple way, like just jump across NaN with intensity = 200 to next rowtype2 = 5 with intensity = 300 .

---

Additional info:

1) Here is how to create xxx

prot = ['A','A','B']
calc_m = [1,2,3]
calc_m2 = [4, np.nan,5]
inte = [100,200,300]
xxx = pd.DataFrame({'name' : pd.Series(prot),
                    'rowtype1': pd.Series(calc_m),
                    'rowtype2': pd.Series(calc_m2),
                    'intensity': pd.Series(inte)
                    })

Answer 1

You could use this option using melt , and then setting the index of ref using numpy's indexing vs. using a for loop

set = xxx.reset_index().melt(['intensity','index'],['rowtype1','rowtype2']).dropna()

ref[set.value.astype(int).values,set['index'].values] = set.intensity.values

which gives you

array([[   0.,    0.,    0.,    0.],
       [ 100.,    0.,    0.,    0.],
       [   0.,  200.,    0.,    0.],
       [   0.,    0.,  300.,    0.],
       [ 100.,    0.,    0.,    0.],
       [   0.,    0.,  300.,    0.],
       [   0.,    0.,    0.,    0.]])

Answer 2

I'm not sure I fully understand what behavior you are looking for, but the pandas dropna() command has the "subset" argument... for example, dropping all rows with NaN in the rowtype2 column could be done with

xxx.dropna(subset=['rowtype2'],inplace=True)

That way, you would drop only rows with NaN in the rowtype2 column.