This caught us yesterday when a string that we assumed to be in UTF-8 was
actually using
CP1252. (This came from an internal web based service, so the root cause is not
R's
fault.) The help page for iconv states that the result of an invalid conversion
is NA
only when the toRaw argument is TRUE, but this appears to be true in general.
Example:
test1 <- "M?ni?re's disease" # the offending string (it
was buried in a 13000
character result string)
test2 <- iconv(test1, to="CP1252") # create a version of the
string that is in
Window-1252 coding
iconv(test2, from="UTF-8") # reprise our error
[1] NA
Note that Encoding(test2) returns "latin-1", which is also not quite
in alignment with the
help page.
