January is not parsed correctly

[vspinu]

> mdy("January131973")
[1] NA

Reported first at stacoverflow.

[Deleetdk]

Locale problem? I had a similar issue where October and May wouldn't parse. These turned out to be the only two for which the name differed in the 3 letter abbreviation between my locale and English.

#366

[cderv]

here's some code I tried to replicate the issue in another locale, then I tried different thing. And clearly I do not understand the behaviour. It juste seems that it is not a simple parsing problem.

• issue is the same in French.
• it appears only with a length one character vector beginning by "January" or "janvier"'
• With a vector with another date to parse, it is working.

library(lubridate)
Sys.setlocale("LC_TIME", "French_France.1252")
#> [1] "French_France.1252"
vdate <- paste0(format(dmy("13/01/1973") + (0:11)*months(1), "%B"), "131973")
vdate
#>  [1] "janvier131973"   "février131973"   "mars131973"     
#>  [4] "avril131973"     "mai131973"       "juin131973"     
#>  [7] "juillet131973"   "août131973"      "septembre131973"
#> [10] "octobre131973"   "novembre131973"  "décembre131973"

mdy(vdate)
#>  [1] "1973-01-13" "1973-02-13" "1973-03-13" "1973-04-13" "1973-05-13"
#>  [6] "1973-06-13" "1973-07-13" "1973-08-13" "1973-09-13" "1973-10-13"
#> [11] "1973-11-13" "1973-12-13"
mdy(vdate[1])
#> Warning: All formats failed to parse. No formats found.
#> [1] NA
mdy(vdate[2])
#> [1] "1973-02-13"
mdy(vdate[1:2])
#> [1] "1973-01-13" "1973-02-13"
mdy(vdate[c(1,1)])
#> Warning: All formats failed to parse. No formats found.
#> [1] NA NA

Sys.setlocale("LC_TIME", "C")
#> [1] "C"

vdate <- paste0(format(dmy("13/01/1973") + (0:11)*months(1), "%B"), "131973")
vdate
#>  [1] "January131973"   "February131973"  "March131973"    
#>  [4] "April131973"     "May131973"       "June131973"     
#>  [7] "July131973"      "August131973"    "September131973"
#> [10] "October131973"   "November131973"  "December131973"

mdy(vdate)
#>  [1] "1973-01-13" "1973-02-13" "1973-03-13" "1973-04-13" "1973-05-13"
#>  [6] "1973-06-13" "1973-07-13" "1973-08-13" "1973-09-13" "1973-10-13"
#> [11] "1973-11-13" "1973-12-13"
mdy(vdate[1])
#> Warning: All formats failed to parse. No formats found.
#> [1] NA
mdy(vdate[2])
#> [1] "1973-02-13"
mdy(vdate[1:2])
#> [1] "1973-01-13" "1973-02-13"
mdy(vdate[c(1,1)])
#> Warning: All formats failed to parse. No formats found.
#> [1] NA NA

Will continue to investigate...

[cderv]

Playing with debugmode, I find that guess_format is not working for "Janvier131973, and this function is called in the process of mdy("Janvier1973").

I think I found a problem into the regex expression stored into the environnement .locale_reg_cache, that is called at the beggining of guess_format.

(((?<b_m_e>>janv\\.|févr\\. and (?<B_m_e>>janvier|février . I think there is an extra >.
If I delete it, it becomes (((?<b_m_e>janv\\.|févr\\. and (?<B_m_e>janvier|février).
and if I replace the object in .locale_reg_cache, as I tried, it works.

It is a bit long to post here and I will try to make a gist to show you the example I made

I think it may be a problem in guess.r file when creating the object strored in .locale_reg_cache but I do not fully search how .get_locale_regs work to create the regex.

I will try to find later. Hope this helps!