How to find out the rows or columns all filled with zeros and delete them in pandas.DataFrame in python?

Question

I have such pandas.DataFrame() object like this:

In [11]: df
Out[11]: 
   a  b
0  0  1
1  0  1
2  0  0
3  0  0
4  1  1

[5 rows x 2 columns]

I want to delete the rows all filled with zeros: the rows with indexes 2 and 3 here.

Desired output:

In [12]: magic_func(df)
Out[12]: 
   a  b
0  0  1
1  0  1
4  1  1

[3 rows x 2 columns]

Answer 1

df.loc[~(df == 0).all(axis=1)]

---

Demo:

In [92]: df = pd.DataFrame({'a':[0,0,0,0,1], 'b':[1,1,0,0,1]})

In [93]: df
Out[93]: 
   a  b
0  0  1
1  0  1
2  0  0
3  0  0
4  1  1

[5 rows x 2 columns]

In [94]: (df == 0).all(axis=1)
Out[94]: 
0    False
1    False
2     True
3     True
4    False
dtype: bool

In [95]: df.loc[~(df == 0).all(axis=1)]
Out[95]: 
   a  b
0  0  1
1  0  1
4  1  1

[3 rows x 2 columns]

---

df[~df.isin([0]).all(axis=1)] also works:

In [108]: df[~df.isin([0]).all(axis=1)]
Out[108]: 
   a  b
0  0  1
1  0  1
4  1  1

but it may be slower for large dataframes:

In [106]: df2 = pd.concat([df]*10000)

In [109]: %timeit df2.loc[~(df2 == 0).all(axis=1)]
100 loops, best of 3: 5.19 ms per loop

In [110]: %timeit df2[~df2.isin([0]).all(axis=1)]
10 loops, best of 3: 50.2 ms per loop

isin is useful when you need to test membership against a large set of values, but for only one value it isn't surprising df == 0 , being more direct, is faster.