Add century parameter when parsing "yy"

[msberends]

I have patients that are, according to my data, born in "27-01-44". Like most West-European countries, we're used to dd-mm-yyyy or dd-mm-yy formats in the Netherlands. No problem for lubridate, just:

dmy("27-01-44")
#> [1] "2044-01-27"

Wait, what? Why is the year higher than that of Sys.Date()?

This has been a problem for years, proven by this question from 2012 until this answer from September 2019.

So what about base R?

as.Date("27-01-44", "%d-%m-%y")
#> [1] "2044-01-27"

So, maybe not a problem in lubridate. But I think (and others on StackOverflow do too) that it would be great if lubridate could bring a solution to this.

I would suggest to add a new argument to dmy() (and probably others as well), to add a reference date that should be the maximum of the year determination, for example Sys.Date().

[vspinu]

You can use parse_date_time2 with cutoff_2000=0.

> parse_date_time2("27-01-44", "dmy", cutoff_2000 = 0)
[1] "1944-01-27 UTC"

Unfortunately similar argument cannot be incorporated in other functions because in some cases we rely on R's strptime which doesn't have a configuration for this parameter.

This part could be improved when parser is extracted from lubridate (lubridate rewrite).

[msberends]

I think if so many people are running into this problem, there should be a more convenient way.

Unfortunately similar argument cannot be incorporated in other functions because in some cases we rely on R's strptime which doesn't have a configuration for this paramete

Sounds like solvable by an ifelse() 😉 Perhaps use another backend than strptime if people use e.g.

dmy("27-01-44", century_max_year = 19)

?

[vspinu]

In principle, strptime is needed only for parsing non-numeric months and week days in non english locales, so indeed, a half baked solution to this might be reasonable as it covers 99% of use cases.