Return localized weekdays for wday()

[alberthkcheng]

wday() always returns English names of weekdays. I made the following changes using date_names_lang in readr, to overcome this. Is there solution exists without using readr?

Also, the same applies to month.

wday.numeric <- function (x, label = FALSE, abbr = TRUE, language="en") 
{
  if (!label) 
    return(x)
  if (abbr) {
    labels <- date_names_lang(language)$day_ab
  }
  else {
    labels <- date_names_lang(language)$day
  }
  ordered(x, levels = 1:7, labels = labels)
}

wday(ymd(160408),label=TRUE,language="ko")
# [1] 금
# Levels: 일 < 월 < 화 < 수 < 목 < 금 < 토

[vspinu]

Yes. It's relatively easy to implement. This applies to month as well.

[hadley]

Where would you get the localisation databases from? I think it's better to use non-operating system sources so code works the same way across platforms.

[vspinu]

You don't need a data base. I had in mind something along the following:

> enc2utf8(unique(format(lubridate:::.date_template, format = "%a@%A")))
[1] "Thu@Thursday"  "Mon@Monday"    "Tue@Tuesday"   "Sun@Sunday"    "Fri@Friday"    "Wed@Wednesday" "Sat@Saturday" 

[hadley]

Oh, so the user would supply the localisation date names, not a language name?

[vspinu]

Yes. By default wday would return names in current locale using the trick above. The new argument locale will accept a valid locale name on user's OS, or a shortcut for 100 common languages such that wday(2, locale = "en") would also work.

[vspinu]

100 is a bit too much but a small data set of common 10-20 languages should be doable.

[hadley]

readr has 185 ;)

[alberthkcheng]

Thank you for the fix :)

The one in Korean is good now. There is still one issue:
For traditional chinese, the format are different from the one in readr (which is more commonly used in HK, etc)

# lubridate
> enc2utf8(unique(format(.date_template, format = format)))
 [1] "四@周四@ 1@1月@上午@"  "一@周一@ 2@2月@下午@"  "二@周二@ 3@3月@上午@"  "日@周日@ 4@4月@下午@" 
 [5] "五@周五@ 5@5月@上午@"  "二@周二@ 6@6月@下午@"  "五@周五@ 7@7月@上午@"  "三@周三@ 8@8月@下午@" 
 [9] "二@周二@ 9@9月@上午@"  "五@周五@10@10月@下午@" "二@周二@11@11月@上午@" "六@周六@12@12月@下午@"

# readr
> locale("zh")
<locale>
Numbers:  123,456.78
Formats:  %Y%.%m%.%d / %H:%M
Timezone: UTC
Encoding: UTF-8
<date_names>
Days:   星期日 (周日), 星期一 (周一), 星期二 (周二), 星期三 (周三), 星期四 (周四), 星期五 (周五), 星期六
        (周六)
Months: 一月 (1月), 二月 (2月), 三月 (3月), 四月 (4月), 五月 (5月), 六月 (6月), 七月 (7月), 八月 (8月), 九月
        (9月), 十月 (10月), 十一月 (11月), 十二月 (12月)
AM/PM:  上午/下午

any idea @hadley @vspinu ?
I am willing to help to fix it if needed ;)

[alberthkcheng]

Also, AM/PM translated correctly in traditional chinese but not in Japanese and Korean.
The one in readr locale() are more completed for all the attributes.

[vspinu]

Also, AM/PM translated correctly in traditional chinese but not in Japanese and Korean. The one in readr locale() are more completed for all the attributes.

I guess it's because readr is using its own data base for this. lubridate relies on system's libraries and compiles attributes like PM/AM on the fly. If thse are missing that probably means they are missing at system level.

[alberthkcheng]

I see. Thanks. Then I think let's leave it to #396 ... :)