Inconsistent date parsing when year is 2- or 4-digits and order "mdy"

[malwinare]

Hi all,

For dates with mixed 2- and 4-digit year, the date-parsing behavior of parse_date_time() is inconsistent when orders = "mdy". But, it works fine when, for example, order = "ymd".

Examples

parse_date_time("apr.12.50", orders = "mdy")   # "2050-04-12" 

# inconsistent
parse_date_time(c("apr.12.50","apr.2.2016"), orders = "mdy")  #  "0050-04-12"  "2016-04-02"

# works fine:
parse_date_time(c("50.apr.12","2016.apr.2"), orders = "ymd")  # "2050-04-12" "2016-04-02"

R version: 3.4.0
lubridate: lubridate_1.6.0.9009

Thanks in advance.
Best,
Malwina

[cderv]

I tried to find where it could come from but it is not obvious.

Training of lubridate found two formats %b.%d.%Y and %b.%d.%y. lubridate:::.select_formats prioritizes the first one. Then lubridate::.striptime pass parsing through base::strptime that do not gives back NA on "apr.12.50" when used with format %b.%d.%Y. So %b.%d.%y is never used and %b.%d.%Y is used on both.

It is not the case with orders = "ymd". Difference is it uses c_parser with %y.%Om.%d and %Y.%Om.%d formats. with orders = "mdy", guess_format does not replace m with Om because of grepl("[^O][mbB]", orders)

Hope this little investigation could help fix the bug.

[vspinu]

Thanks @cderv for spotting this. This was a corner case bug due to the scoring in .select_formats indeed. When y occurred at the end of the format, it wasn't detected by that regexp. Now fixed!