Interval boundaries are not exactly equal

[Niklas191]

Fore some reason it occasionally occurs that the the tstart value of an interval is not exactly equal to the tend value of the previous interval but differs slightly (in the range <1e-18), see example below. Interestingly, this only happens for one interval in the whole dataset.

Unfortunately, this leads to an error when using get_intervals() and, consequentially, prevents pec::pec() from working.

Example (example dataset is attached):
df.csv

> df <- read.csv("df.csv", header = TRUE)
> head(df)
      time event        x1        x2
1 0.143619     1 0.6961045 0.8430649
2 0.124192     1 0.2825347 0.5019340
3 0.038448     1 0.6041976 0.6403361
4 0.536073     1 0.5275156 0.2009848
5 0.345269     1 0.0115978 0.4454234
6 0.194007     1 0.8846513 0.0103127
> unique(sapply(df$time, nchar) - 2) # all values of time have at most 6 digits (-2 for for "." and leading 0)
[1] 6 5 4
> ped_df <- as_ped(df, Surv(time = time, event = event)~., cut = unique(df$time), id = "id")
> table(sort(unique(ped_df$tstart)) == lag(sort(unique(ped_df$tend))))

FALSE  TRUE 
    1   398 
> which(sort(unique(ped_df$tstart)) != lag(sort(unique(ped_df$tend))))
[1] 6
> head(ped_df)
  id   tstart     tend            interval    offset ped_status        x1        x2
1  1 0.000000 0.000849        (0,0.000849] -7.071451          0 0.6961045 0.8430649
2  1 0.000849 0.001239 (0.000849,0.001239] -7.849364          0 0.6961045 0.8430649
3  1 0.001239 0.001301 (0.001239,0.001301] -9.688376          0 0.6961045 0.8430649
4  1 0.001301 0.001592 (0.001301,0.001592] -8.142187          0 0.6961045 0.8430649
5  1 0.001592 0.005539 (0.001592,0.005539] -5.534799          0 0.6961045 0.8430649
6  1 0.005539 0.007474 (0.005539,0.007474] -6.247648          0 0.6961045 0.8430649
> ped_df$tend[5] == ped_df$tstart[6]
[1] FALSE
> ped_df$tend[5] - ped_df$tstart[6]
[1] -8.673617e-19