Issue with to_long_format()

I have a dataset which looks like this:

transforming it into the long format with to_long_format(data, duration_col='time2') produces this:

Note how the start column has zeros. Either there is an issue here or I misunderstood how this function works.

[CamDavidsonPilon]

You’re right, this is confusing. The purpose of to_long_format needs to be expanded, gimme a few hours to make this better.

[sursu]

OK, my case can be covered like this:

data.sort_values(['id','time2'], inplace=True)
data.rename(columns={'time2':'stop'}, inplace=True)
data['start'] = data.groupby('id')['stop'].shift().fillna(0)
cols = [col for col in  data.columns if col != 'stop'] + ['stop']
data = data[cols]

The example you have mentioned above can be solved using these steps:

dd = pd.merge(df, cv, how='left', on='id')
dd['start'] = dd.groupby('id')['t_y'].shift().fillna(0)
dd['stop'] = dd[['t_x', 't_y']].max(axis=1)
dd.loc[set(dd.index) - set(dd.groupby('id')['E'].tail(1).index),'E'] = 0
dd.drop(['t_x','t_y'], axis=1)

But you need to check this. For instance, I have assumed that the time in df cannot be lower than the one in cv. Also, proper sorting needs to be done in case cv is given unsorted.