Assigning a timestamp value to a series with pandas creates an int instead

Question

In Python, Pandas:

g = pd.Series(dict(a = 5, b =datetime(2018, 1,1)))
g['datetime'] = pd.Timestamp('2018-01-02')

g returns:

a                             5
b           2018-01-01 00:00:00
datetime    1514851200000000000
dtype: object

Anyone has an idea why the timestamp is converted to its int value here, and how to avoid the problem and properly attach a timestamp to the Series?

Answer 1

I agree with @MrE, as he says:

I think it makes sense: 5 is not a datetime object, so pandas is parsing the data and find that you have both a number and a datetime object, and it parses both as 'object' string, because the dtype needs to be consistent across all data in the Series. The string representation of a datetime is the date string, not timestamp. When you then add to it, you can modify the dtype of a cell

I really agree, it has to find a dtype and found int, if you drop the 'a' index, it will work (this is basically checking if the above is true or not):

g = pd.Series(dict(b =datetime(2018, 1,1)))
g['datetime'] = pd.Timestamp('2018-01-02')
print(g)

Output:

b          2018-01-01
datetime   2018-01-02
dtype: datetime64[ns]

So it is true.

To solve the issue:

Simply add an extra line at the end:

g['datetime']=pd.to_datetime(g['datetime'])

And now:

print(g)

Is:

a                             5
b           2018-01-01 00:00:00
datetime    2018-01-02 00:00:00
dtype: object