pandas：基于 NaN 的切片数据帧

Question

I have following dataframe df

prod_id prod_ref
10      ef3920
12      bovjhd
NaN     lkbljb
NaN     jknnkn
30      kbknkn

I am trying the following:

df[df['prod_id'] != np.nan]

but I get exactly the same dataframe.

I would like to display

prod_id prod_ref
10      ef3920
12      bovjhd
30      kbknkn

What am I doing wrong?

Answer 1

Use function notna or inverting isna :

print (df[df.prod_id.notna()])
   prod_id prod_ref
0     10.0   ef3920
1     12.0   bovjhd
4     30.0   kbknkn

print (df[~df.prod_id.isna()])

   prod_id prod_ref
0     10.0   ef3920
1     12.0   bovjhd
4     30.0   kbknkn

Another solution is dropna , but need specify column for check NaN :

print (df.dropna(subset=['prod_id']))
   prod_id prod_ref
0     10.0   ef3920
1     12.0   bovjhd
4     30.0   kbknkn

If in another columns are not NaN values, use Alberto Garcia-Raboso's solution .

Answer 2

The problem is that np.nan != np.nan is True (alternatively, np.nan == np.nan is False ). Pandas provides the .dropna() method to do what you want:

df.dropna()

Output:

   prod_id prod_ref
0     10.0   ef3920
1     12.0   bovjhd
4     30.0   kbknkn

By default, .dropna() will drop any row that has a NaN in any column. You can tweak this behavior in two ways:

• check only some columns using the subset argument, and
• require that the row contains NaN in all columns (in the subset , if you are using it) using how='all' — the default is how='any' .

You can check the documentation .