为什么sum（DF）与DF.sum（）的行为不同？

Question

In pandas 0.14, sum(DF) returns the sum of the headers, rather than the sum of the entries. Thus, one gets, for example:

>df = pandas.DataFrame([1,2,3])
>print(df)
   0
0  1
1  2
2  3
>sum(df)
0

Is there a logical reason, a use case for this?

Answer 1

buildin function sum() work as:

sum(list(iter(df)))

and iter(df) get an iterator of column names.

you should not use buildin sum() for DataFrame.

Answer 2

Or use numpy.sum :

In [43]:

df = pd.DataFrame([1,2,3])
print 'numpy:', np.sum(df)
print '.sum() method:',df.sum()
print 'buid-in:', sum(df)
numpy: 0    6
dtype: int64
.sum() method: 0    6
dtype: int64
buid-in: 0