I have a dataframe that looks like the following image.

This is what df.head().df.dict() produces:

The Time Occurred field is a time field in the HHMM format but datatype is int. I'd like to create a new field "Hour" from the "Time Occurred" field which only shows the hour. For instance, where '450' I'd have '4' in the hour field. I tried the following code:

s = df['Time Occurred'].astype(str)
df['Hour'] = pd.to_timedelta(s.str[:2]   ':'   s.str[2:], unit='h')

But I get the following error: ValueError: unit must not be specified if the input contains a str

I found several posts on stack that addressed extracting time from datetime but this isn't the issue I'm trying to solve for

CodePudding user response：

pandas.to_datetime will help. Its option format enables to convert a string with a given format into a datetime. We parse "HHMM" using format='%H%M (see datetime package).

To answer the question:

# a toy dataframe
df = pd.DataFrame({'Time Occurred': {0:'450', 1:'2320', 2:'545'}})
# parse the date
date = pd.to_datetime(df['Time Occurred'], format='%H%M')
# make new columns
df['Hours'] = date.dt.hour
df['Minutes'] = date.dt.minute

print(df)

  Time Occurred  Hours  Minutes
0           450      4       50
1          2320     23       20
2           545      5       45

Note that you can also parse date strings into datetime objects using datetime.strptime.

CodePudding user response：

One way using pandas.Series.str.zfill:

s = pd.Series([1, 450, 2320, 545, 350, 2100])
pd.to_datetime(s.astype(str).str.zfill(4), format="%H%M")

Output:

0   1900-01-01 00:01:00
1   1900-01-01 04:50:00
2   1900-01-01 23:20:00
3   1900-01-01 05:45:00
4   1900-01-01 03:50:00
5   1900-01-01 21:00:00

Explain:

• str.zfill(n) pads zeros on the left of its inputs; so your ints become a string with at most 4 digits.

  0    0001
  1    0450
  2    2320
  3    0545
  4    0350
  5    2100
  

• pd.to_datetime then uses the string to parse based on format, i.e. HHMM (%H%M).