CODEWITHSUNDEEP
pythonpandas
asPlankBridge
asPlankBridge

Reputation: 1082

How to add timestamps in a Pandas.Series?

I am trying to create a Pandas Series object with a timestamp field start:

a = pd.Series(index=['preceding_id', 'file', 'start'],  dtype=[np.int, np.str, np.datetime64], )

it fails with a 

TypeError: data type not understood

Can someone explain me what I am doing wrong?
I've been looking for dates and datetime objects in pandas, but the documentation only says how to use it as index - which is not what I want...

Thank you!

Upvotes: 1

Views: 2713

Answers (2)

David Hagan
David Hagan

Reputation: 1166

Are you sure you don't want a dataframe?

If so, it would look something like:

data = {'preceeding_id': [list of ids],
        'file': [list of files], 
        'start': [list of timestamps]}

df = pd.DataFrame(data)
df.start = pd.to_datetime(df.start)

Or, if you are reading the data in from a file or something, you could easily use parse_dates=True for most of pandas' I/O functions. Pandas is actually pretty fantastic at assigning the correct dtype inherently.

Upvotes: 0

Dan Allan
Dan Allan

Reputation: 35235

A Series can only have one data type. If you want to store multiple types in one Series, the Series' type will be object, the generic Python type.

In [12]: Series([1, 'some string', pd.to_datetime('2014-01-01')])
Out[12]: 
0                      1
1            some string
2    2014-01-01 00:00:00
dtype: object

This is no problem. The types of the constituent elements are retained. For example, the Timestamp in the Series above is still a Timestamp, as we can see by accessing it.

In [13]: Series([1, 'some string', pd.to_datetime('2014-01-01')])[2]
Out[13]: Timestamp('2014-01-01 00:00:00', tz=None)

So, in conclusion, don't specify the datatypes. In general, they'll be handled properly without your help.

Upvotes: 4

Related Questions