# How can I tell LaTeX that an \input file is in a different encoding from inputenc?

I'm currently writing a .tex report with my wife. She's on a Windows machine and I'm running Linux over here; we've decided to encode our report in ISO-8859-1 (i.e., we use \usepackage[latin1]{inputenc} on the preamble).

We're also generating lots of tables in R using xtable() and some other customized functions to output LaTeX. Whenever I sink an R output, I use this command: sink(file("filename.tex", encoding = "ISO-8856-9")). There is, however, this one long table that contains lots of special characters. Whenever I try to sink this table as ISO I get this message:

Warning message:
invalid char string in output conversion

Even though it's just a warning, it prevents our LaTeX compiler from being able to correctly parse the command \input{filename.tex}.

The workaround we have for this is using encoding = "UTF-8" in R, then opening the file on a text editor and saving it with ISO encoding. Then LaTeX compiles the file correctly with apparently all special characters.

Is there a command by which I can tell LaTeX that a file I'm inputting has a different encoding from the main LaTeX file? In my mind, it would look something like this \input[utf8]{filename.tex} (but unfortunately \input or it's fellow \include don't take options).

-

## migrated from stackoverflow.comFeb 22 '12 at 20:23

This question came from our site for professional and enthusiast programmers.

Are you certain that the encoding should be ISO-8856-9? I've never heard of that one. Maybe you mean ISO-8859-6 (Arabic) or ISO-8859-9 (Turkish)?

Let's assume that you meant ISO-8859-9. Then you just write

\usepackage[latin1]{inputenc}

in the beginning of your code,

\inputencoding{latin5}

just before you include the file and

\inputencoding{latin1}

to restore the original encoding.

EDIT:

UTF-8 might cause some problems. I'd solve it by stating that the whole document should be in UTF-8 and then switching back. Here's an example:

The following file is "main-file.tex", and should be saved with Latin 1.

\documentclass{article}
\usepackage[utf8]{inputenc}
\inputencoding{latin1}
\usepackage[russian,brazil]{babel}
\usepackage[T1]{fontenc}
\begin{document}
\section{First section}
Nunca vi socó coçar sem bico, nunca vi sem bico socó coçar. Um socó, dois socós, três socós, quatro socós, quantos socós pra um só socó coçar.

\inputencoding{utf8}
\selectlanguage{russian}
\input{external-file}
\inputencoding{latin1}
\selectlanguage{brazil}

\section{Third section}
A sábia não sabia que o sábio sabia que o sabiá sabia que o sábio não sabia que o sabiá não sabia que a sábia não sabia que o sabiá sabia assobiar.
\end{document}

Also create the file "external-file.tex", and save it with UTF-8:

\section{Second section}
Карл у Клары украл кораллы, а Клара у Карла украла кларнет.

This works for me without any warnings at all. Let me know the specific warnings and errors if you still can't get it to work.

-
Thanks for the help, but that doesn't work (we're writing the document in Brazilian Portuguese, BTW, so we're using \usepackage[brazil]{babel} and \usepackage[T1]{fontenc} in the preamble. I've also tried ambracing my \input{} statement between \inputencoding{utf8} and \inputencoding{latin1}, but it didn't work. –  wleoncio Feb 21 '12 at 21:54
@wleoncio It strange that it doesn't work, because it's working for me when I'm attempting a configuration that I believe is similar to yours. Could you try the example above and let me know if it works? Thanks. –  aliquis Feb 22 '12 at 4:23
That did work, no warnings or errors whatsoever. Thanks for the help! –  Waldir Leoncio Feb 24 '12 at 13:26

If the only problem is in fact the warning message produced by the call to sink(file()), you could use suppressWarnings() to tell R that you don't want to hear any grumbling from that particular bit of code.

I can't exactly recreate your problem without the files you're working on, so here's an example that uses another warning-producing call. Here's how you could suppress the warning produced by an attempt to convert the character string "A" to a numeric vector:

sink(file("otherEgFile.txt", encoding = "ISO-8856-9"))
suppressWarnings(as.numeric("A"))
sink()
-
Unfortunately, it doesn't. Maybe I'm doing this wrong, but when I use supressWarnings(), what it does is echo what should have been sunk. –  wleoncio Feb 21 '12 at 21:03
@wleoncio Just to clarify, did you wrap the call that actually produces the warnings with suppressWarnings()? I realize I gave you the wrong advice, since the call to sink() is not where you are getting your warning. I'll edit my answer above accordingly. –  Josh O'Brien Feb 21 '12 at 21:11
Thanks, Josh, your tip helped me solve the problem. What suppressing the warnings for the table inside the sink command actually did was generate a .tex file with an incomplete table. This table was interrupted just before the mischievous character. Since it only happens twice in the table, I can replace it inside the R script, before creating the table. BTW, here's the prick: ’ –  wleoncio Feb 21 '12 at 22:10
@wleoncio -- Superb. I'm glad to hear that helped you isolate the problem. Cheers. –  Josh O'Brien Feb 21 '12 at 22:21