R: Translate Text Messages

gettext {base}

R Documentation

Translate Text Messages

Description

Translation of text messages typically from calls to stop(), warning(), or message() happens when Native Language Support (NLS) was enabled in this build of R as it is almost always, see also the bindtextdomain() example.

The functions documented here are the low level building blocks used explicitly or implicitly in almost all such message producing calls and they attempt to translate character vectors or set where the translations are to be found.

Usage

gettext(..., domain = NULL, trim = TRUE)

ngettext(n, msg1, msg2, domain = NULL)

bindtextdomain(domain, dirname = NULL)

Sys.setLanguage(lang, unset = "en", C.vs.en = c("msg", "warn", "silent"),
                force = FALSE)

Arguments

`...`	one or more character vectors.
`trim`	logical indicating if the white space trimming in `gettext()` should happen. `trim = FALSE` may be needed for compiled code (C / C++) messages which often end with `\n`.
`domain`	the ‘domain’ for the translation, a `character` string, or `NULL`; see ‘Details’.
`n`	a non-negative integer.
`msg1`	the message to be used in English for `n = 1`.
`msg2`	the message to be used in English for `n = 0, 2, 3, ...`.
`dirname`	the directory in which to find translated message catalogs for the domain.
`lang`	a `character` string specifying a language for which translations should be sought.
`unset`	a string, specifying the default language assumed to be current in the case `Sys.getenv("LANGUAGE")` is unset or empty.
`C.vs.en`	a string specifying how a current (sometimes “hard”) setting of language `"C"` is treated for `lang = "en"`. When `"silent"`, nothing happens, when `"msg"`, a `message()` and if `"warn"` a `warning()` is signalled.
`force`	logical indicating if in case of a “hard” set `"C"` locale, `Sys.setLanguage()` should try modifying the locale (to `"en"`) in order to subsequently set the desired language `lang`, currently only when `.Platform[["OS.type"]]` is `"unix"`.

Details

If domain is NULL (the default) in gettext or ngettext, the domain is inferred. If gettext or ngettext is called from a function in the namespace of package pkg including called via stop(), warning(), or message() from the function, or, say, evaluated as if called from that namespace, see the evalq() example, the domain is set to "R-pkg". Otherwise there is no default domain and messages are not translated.

Setting domain = NA in gettext or ngettext suppresses any translation.

"" does not match any domain. In gettext or ngettext, domain = "" is effectively the same as domain = NA.

If the domain is found, each character string is offered for translation, and replaced by its translation into the current language if one is found.

The language to be used for message translation is determined by your OS default and/or the locale setting at R's startup, see Sys.getlocale(), and notably the LANGUAGE environment variable, and also Sys.setLanguage() here.

Conventionally the domain for R warning/error messages in package pkg is "R-pkg", and that for C-level messages is "pkg".

For gettext, when trim is true as by default, leading and trailing whitespace is ignored (“trimmed”) when looking for the translation.

ngettext is used where the message needs to vary by a single integer. Translating such messages is subject to very specific rules for different languages: see the GNU Gettext Manual. The string will often contain a single instance of %d to be used in sprintf. If English is used, msg1 is returned if n == 1 and msg2 in all other cases.

bindtextdomain is typically wrapper for the C function of the same name: your system may have a man page for it. With a non-NULL dirname it specifies where to look for message catalogues: with dirname = NULL it returns the current location. If NLS is not enabled, bindtextdomain(*,*) returns NULL. The special case bindtextdomain(NULL) calls C level textdomain(textdomain(NULL)) for the purpose of flushing (i.e., emptying) the cache of already translated strings; it returns TRUE when NLS is enabled.

The utility Sys.setLanguage(lang) combines setting the LANGUAGE environment variable with flushing the translation cache by bindtextdomain(NULL).

Value

For gettext, a character vector, one element per string in .... If translation is not enabled or no domain is found or no translation is found in that domain, the original strings are returned.

For ngettext, a character string.

For bindtextdomain, a character string giving the current base directory, or NULL if setting it failed.

For Sys.setLanguage(), the previous LANGUAGE setting with attribute attr(*, "ok"), a logical indicating success. Note that currently, using a non-existing language lang is still set and no translation will happen, without any message.

Examples

bindtextdomain("R")  # non-null if and only if NLS is enabled

for(n in 0:3)
    print(sprintf(ngettext(n, "%d variable has missing values",
                              "%d variables have missing values"),
                  n))

## Not run: ## for translation, those strings should appear in R-pkg.pot as
msgid        "%d variable has missing values"
msgid_plural "%d variables have missing values"
msgstr[0] ""
msgstr[1] ""

## End(Not run)

miss <- "One only" # this line, or the next for the ngettext() below
miss <- c("one", "or", "another")
cat(ngettext(length(miss), "variable", "variables"),
    paste(sQuote(miss), collapse = ", "),
    ngettext(length(miss), "contains", "contain"), "missing values\n")

## better for translators would be to use
cat(sprintf(ngettext(length(miss),
                     "variable %s contains missing values\n",
                     "variables %s contain missing values\n"),
            paste(sQuote(miss), collapse = ", ")))

thisLang <- Sys.getenv("LANGUAGE", unset = NA) # so we can reset it
if(is.na(thisLang) || !nzchar(thisLang)) thisLang <- "en" # "factory" default
enT <- "empty model supplied"
Sys.setenv(LANGUAGE = "de") # may not always 'work'
gettext(enT, domain="R-stats")# "leeres Modell angegeben" (if translation works)
tget <- function() gettext(enT)
tget() # not translated as fn tget() is not from "stats" pkg/namespace
evalq(function() gettext(enT), asNamespace("stats"))() # *is* translated

## Sys.setLanguage()  -- typical usage --
Sys.setLanguage("en") -> oldSet # does set LANGUAGE env.var
errMsg <- function(expr) tryCatch(expr, error=conditionMessage)
(errMsg(1 + "2") -> err)
Sys.setLanguage("fr")
errMsg(1 + "2")
Sys.setLanguage("de")
errMsg(1 + "2")
## Usually, you would reset the language to "previous" via
Sys.setLanguage(oldSet)

## A show off of translations -- platform (font etc) dependent:
## The translation languages available for "base" R in this version of R:
if(capabilities("NLS")) withAutoprint({
  langs <- list.files(bindtextdomain("R"),
		      pattern = "^[a-z]{2}(_[A-Z]{2}|@quot)?$")
  langs
  txts <- sapply(setNames(,langs),
		 function(lang) { Sys.setLanguage(lang)
				 gettext("incompatible dimensions", domain="R-stats") })
  cbind(txts)
  (nTrans <- length(unique(txts)))
  (not_translated <- names(txts[txts == txts[["en"]]]))
})

## Here, we reset to the *original* setting before the full example started:
if(nzchar(thisLang)) { ## reset to previous and check
  Sys.setLanguage(thisLang)
  stopifnot(identical(errMsg(1 + "2"), err))
} # else staying at 'de' ..

[Package base version 4.5.0 Index]