Rev 81018 | Blame | Compare with Previous | Last modification | View Log | Download | RSS feed
#### Regression test of translation not working outside packages for R < 4.1.0.#### We try French (set in Makefile.common).### First off, message translation needs to be supported.if (!capabilities("NLS")) { ## e.g. when R was configured with --disable-nlsmessage("no natural language support")q("no")}#### Report locale and charsetSys.getlocale()l10n_info()#### Skip locales that do not support French (especially C)OK <- l10n_info()[["UTF-8"]] || l10n_info()[["Latin-1"]]if (!OK) {if(.Platform$OS.type == "windows") {OK <- l10n_info()[["codepage"]] == 28605 ## Latin-9} else {z <- l10n_info()[["codeset"]]## macOS and Solaris have the first, Linux the second.OK <- tolower(z) %in% c("iso8859-15", "iso-8859-15", "iso885915")}}if( !OK ) {message("The locale encoding is not known to support French")q("no")}## Translation domain for a function not in a package: PR#17998tryCEmsg <- function(expr) tryCatch(expr, error = conditionMessage)tryCWmsg <- function(expr) tryCatch(expr, warning = conditionMessage)chk0 <- function(x) stopifnot(x == 0)nsSt <- asNamespace("stats")(m1 <- tryCEmsg(chk0(1))) # (not translated in R < 4.1.0)## switch back to English (if possible) for final report.Sys.setenv(LANGUAGE="en")m2 <- "x == 0 n'est pas TRUE"if(m1 != m2) stop("message was not translated to French")## More -- for PR#18902 (<--> PR#17998, part 2)enTxt <- "incompatible dimensions"deTxt <- "inkompatible Dimensionen"Sys.setenv(LANGUAGE="de")stopifnot(identical(deTxt, gettext(enTxt, domain="R-stats")))f <- function(...) stop(enTxt)environment(f) <- nsStstopifnot(identical(deTxt, tryCEmsg(f()))) # failed in R <= 4.1.x## 2nd example (base vs stats):enTxt <- "namespace is already attached"deTxt <- "Namensraum ist bereits angehÃĪngt"Encoding(deTxt) <- "UTF-8" # e.g. on Windows where it was "latin1"all.equal(gettext(enTxt, domain="R-stats"), enTxt)(trTxt <- gettext(enTxt, domain="R-base")); Encoding(trTxt) # unknownall.equal(trTxt, deTxt)f <- function(...) warning(enTxt)(trTxtSt <- {environment(f) <- nsSt; tryCWmsg(f())})stopifnot(exprs = {identical(trTxt, deTxt)identical(gettext(enTxt, domain="R-stats"), enTxt)identical({environment(f) <- .BaseNamespaceEnv; tryCWmsg(f())}, deTxt)identical({environment(f) <- nsSt; tryCWmsg(f())}, enTxt)# not in R <= 4.1.xidentical(trTxtSt, enTxt) # (not in 4.1.x, but in 4.0.5)})# in all cases: not present in stats => not translated## Functions not *from* package namespace, but "as if" (PR#17998, from Comment 35):enT <- "empty model supplied"(deT <- gettext(enT, domain="R-stats"))# "leeres Modell angegeben"isD <- function(tx) identical(deT, tx)stopifnot(exprs = {## 1-4: translated in R 4.0.z *and* 4.1.zisD(evalq(function() gettext(enT), nsSt)())isD(evalq(function() do.call(gettext, list(enT)), nsSt)())isD(evalq(function() evalq(gettext(enT)), nsSt)())isD(evalq(function() local(gettext(enT)), nsSt)())## 5-7: not translated in R 4.0.*; translated in R 4.1.* (incl. R-patched)## ditto in R-devel *after* the Oct.20 (2021) patch:isD(evalq(local(gettext(enT)), nsSt))isD(evalq(gettext(enT), nsSt))isD(do.call("gettext", list(enT), envir=nsSt))## 8-11: in comment #37, Suharto added " Other cases: "isD(evalq(function() (function() gettext(enT))(), nsSt)())isD(evalq(function() function() gettext(enT), nsSt)()())isD(evalq((function() function() gettext(enT))(), nsSt)())isD(evalq(local(function() gettext(enT)), nsSt)())require(compiler) ## and more cases with byte compiler considerationisD(cmpfun(evalq(function() gettext(enT), nsSt))())isD(cmpfun(evalq(function() do.call("gettext", list(enT)), nsSt))())isD(cmpfun(evalq(function() evalq(gettext(enT)), nsSt))())isD(cmpfun(evalq(function() local(gettext(enT)), nsSt))())isD(eval(compile(quote(local(gettext(enT)))), nsSt))isD(eval(compile(quote(gettext(enT))), nsSt))})