fwrite: pre-encode strings and factor levels

Previously, fwrite() deferred encoding of the strings to
fwriteR.c:getString,getCategString called from OpenMP threads. Calling
translateChar[UTF8] to encode a string results in memory allocation
unless it is already in the desired encoding, which is unsafe to perform
on a non-main thread.

Fixes: #6883

Author: aitap

R/fwrite.R

@@ -111,6 +111,15 @@ fwrite = function(x, file="", append=FALSE, quote="auto",
   }
   # nocov end
   file = enc2native(file) # CfwriteR cannot handle UTF-8 if that is not the native encoding, see #3078.
+  # pre-encode any strings or factor levels to avoid translateChar trying to allocate from OpenMP threads
+  if (encoding %chin% c("UTF-8", "native")) {
+    enc = switch(encoding, "UTF-8" = enc2utf8, "native" = enc2native)
+    x = lapply(x, function(x) {
+      if (is.character(x)) x = enc(x)
+      if (is.factor(x)) levels(x) = enc(levels(x))
+      x
+    })
+  }
   .Call(CfwriteR, x, file, sep, sep2, eol, na, dec, quote, qmethod=="escape", append,
         row.names, col.names, logical01, scipen, dateTimeAs, buffMB, nThread,
         showProgress, is_gzip, compressLevel, bom, yaml, verbose, encoding)

inst/tests/tests.Rraw

@@ -21113,3 +21113,7 @@ test(2309.09, as.data.table(df, keep.rownames=TRUE), data.table(rn = c("a","b"),
 as.data.frame.no.reset = function(x) x
 DF = structure(list(a = 1:2), class = c("data.frame", "no.reset"), row.names = c(NA, -2L))
 test(2310.01, as.data.table(DF), data.table(a=1:2))
+
+# avoid translateChar*() in OpenMP threads, #6883
+DF = list(rep(iconv(strrep("\uf8", 100), to = "latin1"), 100000))
+test(2311, fwrite(DF, nullfile(), encoding = "UTF-8"), NULL)