Converting multiple UTC timestamps columns to US/Pacific in pandas dataframe-CodePudding

I have a dataset containing 2 timestamp columns with string format YYYY-mm-dd HH:mm:ss. The timezone of these timestamps is not included in the string, but they are in UTC and I want to convert them to US/Pacific.

My dataset is in a csv file, so I'm parsing those 2 timestamp columns using

df = pd.read_csv("my_dataset.csv", parse_dates=['timestamp_col1', 'timestamp_col2'])

After reading some posts here it seems the below (example for one column only) could be a possible solution

pd.to_datetime(df.timestamp_col1, utc=True).astimezone(timezone('US/Pacific'))

However, I'm getting the error below. Any insights into how to fix this or is there a better way to do this?

'Series' object has no attribute 'astimezone'

CodePudding user response：

It seems you are looking for dt.tz_convert:

pd.to_datetime(df.timestamp_col, utc=True)
#0   2022-10-10 10:10:00 00:00
#Name: timestamp_col, dtype: datetime64[ns, UTC]

pd.to_datetime(df.timestamp_col, utc=True).dt.tz_convert(tz="US/Pacific")
#0   2022-10-10 03:10:00-07:00
#Name: timestamp_col, dtype: datetime64[ns, US/Pacific]

Data:

df = pd.read_csv(StringIO("""timestamp_col
2022-10-10 10:10:00
"""), parse_dates=['timestamp_col'])

df
        timestamp_col
0 2022-10-10 10:10:00

CodePudding user response：

The pandas.Timestamp object has its native timezone feature. Localize it with the tz argument and convert it simply with the tz_convert() method. So no need to mingle packages

pd.Timestamp.now(tz='UTC').tz_convert('US/Pacific')

This post covered it also.

You can also convert all timestamps of a series by using apply

sr = pd.Series([pd.Timestamp.now(tz='UTC'), pd.Timestamp.now(tz='UTC')])
sr.apply(lambda x: x.tz_convert('US/Pacific'))

One needs to use this workaround since pandas.Series.tz_conver() expects a pandas.DatetimeIndex so that it converts the index and not the values.