# Input encoding error after upgrading from Biber 1.9 to Biber 2.1

Under biber 2.1 my biblatex produces input encoding errors. The same procedures worked smoothly under biblatex 2.9a/biber 1.9, now that I upgraded to biblatex 3.0/biber 2.1 I get the following message when pdflatex runs after biber:

Package inputenc Error: Unicode char \u8:╠ü not set up for use with LaTeX.

Here is a minimum working example:

\documentclass{article}
\usepackage{lmodern}
\usepackage[T1]{fontenc}
\usepackage[utf8]{inputenc}
\usepackage[backend=biber]{biblatex}
\bibliography{test.bib}
\usepackage{filecontents}

\begin{filecontents}{test.bib}
@article{rodr,
author = {Rodr{\'{\i}}guez, A},
year = {1999} }
\end{filecontents}

\begin{document}
Rodr\'{\i}guez
\textcite{rodr}
\printbibliography
\end{document}


Any help appreciated!

• github.com/plk/biber/issues/65 – Johannes_B Jun 20 '15 at 14:33
• Just out of curiosity: Is there a reason why you don't input the accented character directly as í? – Mico Jun 20 '15 at 16:48
• I have a large library in Mendeley, and this is how the export-to-bib function deals with the character. – cth Jun 20 '15 at 19:07
• Typing \'i for getting í has been in LaTeX for about 20 years. – egreg Jun 21 '15 at 14:43

run biber with

biber --output-safechars <file>


• Great! Works perfect! – cth Jun 20 '15 at 19:10
• This is the only solution that works for me. – jhin Nov 16 '18 at 16:17

Writing Rodr{\'{\i}}guez seems needlessly complicated. I suggest you write Rodr{\'i}guez instead. Not only is it easier to do so, it also makes the biber/inputenc issue go away automatically.

Incidentally, outside the bib file, i.e., in the body of tex file, I'd write Rodr\'iguez.

• I'd write the unicode character in all instances ;-) – Johannes_B Jun 20 '15 at 15:38
• @Johannes_B - If it were my bib file, I would too. :-) I'm assuming, though, that the OP has a good reason for not entering the character as í. That's why I suggested he/she input it as {\'i}. – Mico Jun 20 '15 at 15:40
• I think the real issue here is ref-manager export. See also golatex.de/i-mit-diaresis-in-bibtex-t15369.html – Johannes_B Jun 20 '15 at 15:48
• @Johannes_B - That may well be the case. For the OP's MWE, though, the problem arises even without a ref-manager (bibdesk? jabref?) being involved. – Mico Jun 20 '15 at 15:57

This is a change in biber 2.1 with \i in particular. Now biber properly encodes this as a dotless i (ı - 0x0131) with a combining accent. Even though biber always converts to precomposed (NFC) form on output and is therefore generally as friendly as possible to inputenc, there is no precomposed form of this combination. There is a precomposed form for a normal 'i' followed by the same combining char but this is incorrect and it is a completely different thing, causing problems in some fonts and needing a special case in the decoding. \'i is generally a better choice because most fonts know that the "i" in this shouldn't have both a dot and an accent (or alternatively you simply can't see the dot because of the accent if you are lucky). inputenc also knows about normal ISO8859-1 literal í characters so you can just use these without any macros.

If you get your data from some external source and therefore don't get to choose the form in which you get your accented "i"s, put in a biber sourcemap to fix up the code if you have inputenc or font issues with dotless "i"s:

\DeclareSourcemap{
\maps[datatype=bibtex]{
\map[overwrite]{
\step[fieldsource=author,
match=\regexp{\x{0131}\x{0301}},
replace=\regexp{\x{00ED}}]
}
}
}


Since macro decoding into UTF8 is done before source mapping, it's already in UTF-8 NFD form by the time mappings are applied. The above simply changes the dotless i (U+0131) followed by combining acute accent (U+0301) into a standard ISO 8859-1 lower case i with acute (U+00ED) which is supported by inputenc.

Or you can just use the biber --output-safechars option to force encoding of UTF-8 into LaTeX macros when writing the .bbl.

• There is \DeclareTextComposite{\'}{T1}{i}{237} in t1enc.def and similar code for the dot above, grave, dieresis and circumflex accents. This is the reason why \'i is good. It's not a font property. However, other accents such as breve or macron could give problems because of this change to the behavior of \i, I guess. – egreg Jun 21 '15 at 14:42
• Yes, it's a general thing with \i, regardless of the accent. There used to be all sorts of special cases just for this which was making recoding hard because really, you are applying an accent to a dotless i. It's just that these inputenc semantics were easier in days gone by. – PLK Jun 21 '15 at 14:49
• Well, just so whomever made that call knows I just wasted about an hour combing through my .bib file for non-unicode characters, converting it into hex and looking for cc81 until someone finally showed me \DeclareUnicodeCharacter{0301}{HERE}. So thanks for making biber generate unsafe output for pdflatex by default. Wouldn't that be better as the non-default option? – Canageek Apr 28 '16 at 21:29

If you typeset directly the í, it runs fine:

\documentclass{article}
\usepackage{lmodern}
\usepackage[T1]{fontenc}
\usepackage[utf8]{inputenc}
\usepackage[backend=biber]{biblatex}
\bibliography{test.bib}
\usepackage{filecontents}
%
\begin{filecontents}{test.bib}
@article{rodr,
author = {Rodríguez, A},
year = {1999} }
\end{filecontents}

\begin{document}
Rodríguez
\textcite{rodr}
\printbibliography
\end{document}