How to parse an invalid date with lubridate?

Question

I need to parse dates and have a cases like "31/02/2018":

library(lubridate)
> dmy("31/02/2018", quiet = T)
[1] NA

This makes sense as the 31st of Feb does not exist. Is there a way to parse the string "31/02/2018" to e.g. 2018-02-28 ? So not to get an NA, but an actual date?

Thanks.

No, there isn't (because it is not an actual date). You can subset and replace it. — Roland, Nov 09 '18 at 12:52
Is this a [XY problem](https://meta.stackexchange.com/questions/66377/what-is-the-xy-problem)? — Rui Barradas, Nov 09 '18 at 13:03
@RuiBarradas blame the one, who filled the survey... or the one who did not set up a format filter. your choice. ;) — c0bra, Nov 09 '18 at 17:37
OK, not solvable at the source :(. But solvable, there's already an accepted solution. — Rui Barradas, Nov 09 '18 at 17:40

score 1 · Accepted Answer · answered Nov 09 '18 at 13:05

We can write a function assuming you would only have dates which could be higher than the actual date and would have the same format always.

library(lubridate)

get_correct_date <- function(example_date) {
  #Split vector on "/" and get 3 components (date, month, year)
  vecs <- as.numeric(strsplit(example_date, "\\/")[[1]])

  #Check number of days in that month
  last_day_of_month <-  days_in_month(vecs[2])

  #If the input date is higher than actual number of days in that month
  #replace it with last day of that month
  if (vecs[1] > last_day_of_month)
    vecs[1] <- last_day_of_month

  #Paste the date components together to get new modified date
  dmy(paste0(vecs, collapse = "/"))
}


get_correct_date("31/02/2018")
#[1] "2018-02-28"

get_correct_date("31/04/2018")
#[1] "2018-04-30"

get_correct_date("31/05/2018")
#[1] "2018-05-31"

With small modification you can adjust the dates if they have different format or even if some dates are smaller than the first date.

Thanks for the solution. I also checked last_day_of_month for NA as some dates even had a faulty month and could not be recovered. — c0bra, Nov 12 '18 at 11:21
@c0bra I think month would need a separate treatment. This works only for dates. You can add a new question for that. — Ronak Shah, Nov 12 '18 at 11:45

How to parse an invalid date with lubridate?

1 Answers1