将每个第一个匹配行中的列值设置为 0

Question

I'm attempting to detect sequences of time where the difference between timestamps is below some threshold. To perform this I transform a list of timestamps into seconds and measure the difference between each time stamp. I've written code to perform this but the issue is when I measure differences between days the time difference value should be zero for the first row in each day. So the value 86390 in below dataframe should be 0. This is just a contrived example. For multiple groups how to set the first entry in each group to 0 ?

Code :

import pandas as pd

arr= []
df = pd.DataFrame(
    {'date': ['2019-01-01 00:02:48.714000' , '2019-01-01 00:02:58.714000' , '2019-01-02 00:02:48.714000' , '2019-01-02 00:04:48.714000'],
     'id': [1 , 2 , 3 , 4],

    })
df['date'] = pd.to_datetime(df['date'])

for d in df['date'] : 
    arr.append(d.timestamp())

df.sort_values(by=['date'])
df['TIME_IN_SEC'] = arr
df['TIME_IN_SEC_SHIFT'] = df.TIME_IN_SEC.shift(1)
df['TIME_DIFF'] = df["TIME_IN_SEC"] - df["TIME_IN_SEC_SHIFT"]

list_values = []

for g in df.groupby(pd.Grouper(key='date',freq='D')) : 
    list_values.append(sum(g[1]['TIME_DIFF']) / len(g[1]))

df

renders :

Answer 1

To set the first row of each day to zero, you can group by the date column but extract the actual date value, then aggregate to the 'first' row. Create a series from this for the 'id' column. (I'm assuming they are unique values.)

id_filt = df.groupby(df.date.dt.date).first()['id']

Then use loc to return only rows with the 'id' values, then set the columns equal to zero.

df.loc[df["id"].isin(id_filt.values), ["TIME_IN_SEC_SHIFT", "TIME_DIFF"]] = 0

                     date  id   TIME_IN_SEC  TIME_IN_SEC_SHIFT  TIME_DIFF
0 2019-01-01 00:02:48.714   1  1.546301e+09       0.000000e+00        0.0
1 2019-01-01 00:02:58.714   2  1.546301e+09       1.546301e+09       10.0
2 2019-01-02 00:02:48.714   3  1.546387e+09       0.000000e+00        0.0
3 2019-01-02 00:04:48.714   4  1.546387e+09       1.546387e+09      120.0

Of course you could combine these together to get:

df.loc[
    df["id"].isin(df.groupby(df.date.dt.date).first()["id"].values),
    ["TIME_IN_SEC_SHIFT", "TIME_DIFF"],
] = 0