The pd.Series.prod() function

问题

This should be probably elementary but still I can not figure it out.

I am reading the documentation on pd.Series and doing simple exercises.

My code is the following:

import pandas as pd
import numpy as np    
pd.Series([2, 4, 6]).prod()
Out[7]: 48

a = pd.Series(np.arange(1, 100, 3))

a
Out[9]: 
0      1
1      4
2      7
3     10
4     13
5     16
6     19
7     22
8     25
9     28
10    31
11    34
12    37
13    40
14    43
15    46
16    49
17    52
18    55
19    58
20    61
21    64
22    67
23    70
24    73
25    76
26    79
27    82
28    85
29    88
30    91
31    94
32    97
dtype: int32

a.prod()
Out[10]: 0

a = pd.Series(np.arange(1, 100, 2))

a.prod()
Out[15]: -373459037

type(a)
Out[18]: pandas.core.series.Series

My question is why this erratic -to my eyes-- behavior? Why a.prod() the first time evaluates to 0 and then evaluates to a negative integer?

Your advice will be appreciated.

回答1:

We can directly use numpy's np.prod with specified dtype to overcome int overflow :

np.prod(a.values,dtype=np.int64)
Out[938]: 5196472710489536419

回答2:

it's a int32 overflow:

In [340]: a = pd.Series(np.arange(1, 100, 3)).astype(np.int64)
#  NOTE:                 --------------->    ^^^^^^^^^^^^^^^^^

In [341]: a.prod()
Out[341]: 8624389262030143488

来源：https://stackoverflow.com/questions/47335608/the-pd-series-prod-function