Question

I have a bunch of author names from foreign countries in a CSV which R reads in just fine. I’m trying to clean them for upload to Mechanical Turk (which really doesn’t like even a single internationalized character). In so doing, I have a question (to be posted later), but I can’t even dput them in a sensible way:

> dput(df[306,"primauthfirstname"])
"Gwena\xeblle M"
> test <- "Gwena\xeblle M"
<simpleError in nchar(val): invalid multibyte string 1>

In other words, dput works just fine, but pasting the result in fails. Why doesn’t dput output the necessary information to allow copy/pasting back into R (presumably all it needs to do is add the encoding attributes the a structure statement?). How do I get it to do so?

Note that \xeb is a valid character as far as R is concerned:

> gsub("\xeb","", turk.df[306,"primauthfirstname"] )
[1] "Gwenalle M"

But that you can’t evaluate the characters individually–it’s hex code \x## or nothing:

> gsub("\\x","", turk.df[306,"primauthfirstname"] )
[1] "Gwena\xeblle M"

Answer 1

dput()‘s helppage says: “Writes an ASCII text representation of an R object”. So if your object contains non-ASCII characters, these cannot be represented and have to be converted somehow.

So I would suggest you use iconv() to convert your vector before dputing. One approach is:

> test <- "Gwena\xeblle M"
> out <- iconv(test, from="latin1", to="ASCII", sub="byte")
> out
[1] "Gwena<eb>lle M"
> gsub('<eb>', 'ë', out)
[1] "Gwenaëlle M"

which, as you see, works both ways. You can later use gsub() to back-convert bytes into characters (if your encoding supports it, e.g. utf-8).

The second approach is simpler (and I guess preferable for your needs), but works one-way and your libiconv may not support it:

> test <- "Gwena\xeblle M"
> iconv(test, from="latin1", to="ASCII//TRANSLIT")
[1] "Gwenaelle M"

Hope this helps!