CODEWITHSUNDEEP
pythonpandasdataframetimestamp
N.Foe
N.Foe

Reputation: 79

Create new timestamp in pandas based on occurence in other column

I have a dataframe-column with random timestamps and NaT values in between them:

timestamp             
01-01-2018 13:12:48
NaT
NaT
NaT
04-01-2018 08:15:12
NaT
Nat

I want to create another column that starts counting from 0 once there is a new timestamp in the timestamp column (col_A). I wouldnt mind if that column was a timestamp object but without the date (col_B). Is that possible?

timestamp            col_A  col_B             
01-01-2018 13:12:48  0      00:00:00
NaT                  1      00:01:00
NaT                  2      00:02:00
NaT                  3      00:03:00
04-01-2018 08:15:12  0      00:00:00
NaT                  1      00:01:00
Nat                  2      00:02:00

Upvotes: 1

Views: 43

Answers (1)

Code Different
Code Different

Reputation: 93151

It's an island-and-gap problem: every time timestamp is not null, it creates a new island. You usually solve these problems with a cumulative sum of some kind.

Try this:

islands = df['timestamp'].notnull().cumsum()
df['col_A'] = df.groupby(islands).cumcount()
df['col_B'] = pd.to_timedelta(df['col_A'], unit='minute')

Upvotes: 1

Related Questions