Index in Swedish sorting order ( ...YZÅÄÖ...) with xindy and LaTeX

Question

I still haven't succeed to sort Swedish letters in the index list with class article in the correct order.

I've tried it with specifying a sort key, i.e. \index{g~ra@göra}, but the result is not perfect. There will not be any empty lines between the groups of Å, Ä, and Ö sorting. And the extra coding job seems not be up to date 2014.

This way is quite cumbersome. Is there a better way?

Xindy should do the work (easy?), but how do I setup xindy to create an index with Swedish sorting order?

My comment: All texts are not written in English. I like the LaTeX system, it's a lot better than Word in my case, but the sorting troubles me.

Answer 1

The problem is that with inputenc and utf8 option, index files are written in a way that xindy cannot process. This sample:

\documentclass[11pt]{article} 
\usepackage[T1]{fontenc}
\usepackage[swedish]{babel}
\usepackage[utf8]{inputenc} 

\usepackage{makeidx}
\makeindex

\begin{document}
This kind of index in text:

\index{Säker plats|textbf}Säker plats
foo\index{Aäö}\index{Aäö}
bar\index{äö}\index{aø}\index{ohne}\index{øæ}
\printindex
\end{document}

produces this .idx file:

\indexentry{S\IeC {\"a}ker plats|textbf}{1}
\indexentry{A\IeC {\"a}\IeC {\"o}}{1}
\indexentry{A\IeC {\"a}\IeC {\"o}}{1}
\indexentry{\IeC {\"a}\IeC {\"o}}{1}
\indexentry{a\IeC {\o }}{1}
\indexentry{ohne}{1}
\indexentry{\IeC {\o }\IeC {\ae }}{1} 

you need to convert this code to utf8. It is possible to do this using my script iec2utf. Download the file iec2utf.lua to the directory with your document and make batch script swedxindy:

#!/usr/bin/env sh
texlua iec2utf.lua T1 < `basename $1 .tex`.idx | texindy -i -M lang/swedish/utf8-lang -o `basename $1 .tex`.ind

I don't use texshop so I don't know how to add this script to the menu, but you can call it from command line (you will have to make it executable, I think it can be made with command chmod -x swedxindy like in Linux).

Edit

I simplified the process - there is now script called utftexindy in the iec2utf repository. Process described above can be now simplified to:

texlua utftexindy.lua -L swedish sample.idx

Result:

Answer 2

In his answer michal.h21 has identified the problem and found the problem in the way, UTF-8 characters are written, if \usepackage[utf8]{inputenc} is used:

\IeC{<LICR>}% LICR = LaTeX Internal Character Representation

Since package inputenc makes the 8-bit bytes active, they can be redefined to print themselves instead of the \IeC stuff.

Also \index uses verbatim category codes for its argument. LaTeX does not include 8-bit characters in its verbatim category codes, because it has to map the UTF-8 byte sequence to a character slot of a font encoding to get the correct character.

The following example, based on michal.h21's example, patches \index that it does not write expanded UTF-8 characters:

\documentclass[11pt]{article}
\usepackage[T1]{fontenc}
\usepackage[swedish]{babel}
\usepackage[utf8]{inputenc}

\usepackage{makeidx}
\makeindex

\usepackage{etoolbox}
\makeatletter
\patchcmd\index{\@sanitize}{\@sanitize\index@sanitize}{}{%
  \errmessage{Patching \noexpand\index failed}%
}
\let\index@sanitize\@empty
\begingroup
  \count@=127
  \@whilenum\count@<255 \do{%
    \advance\count@\@ne
    \lccode`\*=\count@
    \lccode`\~=\count@
    \lowercase{%
      \expandafter
      \g@addto@macro\expandafter\index@sanitize\expandafter{%
        % verbatim catcode
        \expandafter\@makeother\csname *\endcsname
        % active character expands to non-expandable itself
        \def~{*}%
      }%

    }%  
  }
\endgroup
\makeatother

\begin{document}
This kind of index in text:

\index{Säker plats|textbf}Säker plats
foo\index{Aäö}\index{Aäö}
bar\index{äö}\index{aø}\index{ohne}\index{øæ}

\textbf{foo\index{Aäö}\index{Aäö}}
\printindex
\end{document}

The following .idx file is written:

\indexentry{Säker plats|textbf}{1}
\indexentry{Aäö}{1}
\indexentry{Aäö}{1}
\indexentry{äö}{1}
\indexentry{aø}{1}
\indexentry{ohne}{1}
\indexentry{øæ}{1}
\indexentry{Aäö}{1}
\indexentry{Aäö}{1}

Answer 3

Here's an alternative (requires at least version 4.02 of glossaries):

% arara: pdflatex
% arara: makeglossaries
% arara: pdflatex
\documentclass[11pt]{article}
\usepackage[T1]{fontenc}
\usepackage[swedish]{babel}
\usepackage[utf8]{inputenc}

\usepackage[index,xindy]{glossaries}

\makeglossaries

\newterm[name={Säker plats}]{Saker plats}
\newterm[name={Aäö}]{Aao}
\newterm[name={äö}]{ao}
\newterm[name={aø}]{aoslash}
\newterm{ohne}
\newterm[name={øæ}]{oae}

\begin{document}
This kind of index in text:

\gls[format=textbf]{Saker plats}
foo\glsadd{Aao}\glsadd{Aao}
bar\glsadd{ao}\glsadd{aoslash}\glsadd{ohne}\glsadd{oae}

\printindex[style=indexgroup]
\end{document}

This produces:

This works because the default behaviour of glossaries is to sanitize the sort key before writing it to the external indexing file, so the .idx file for the above looks like:

(indexentry :tkey (("Säker plats" "\\glossentry{Saker plats}") ) :locref "{}{1}" :attr "pagetextbf" )
(indexentry :tkey (("Aäö" "\\glossentry{Aao}") ) :locref "{}{1}" :attr "pageglsnumberformat" )
(indexentry :tkey (("Aäö" "\\glossentry{Aao}") ) :locref "{}{1}" :attr "pageglsnumberformat" )
(indexentry :tkey (("äö" "\\glossentry{ao}") ) :locref "{}{1}" :attr "pageglsnumberformat" )
(indexentry :tkey (("aø" "\\glossentry{aoslash}") ) :locref "{}{1}" :attr "pageglsnumberformat" )
(indexentry :tkey (("ohne" "\\glossentry{ohne}") ) :locref "{}{1}" :attr "pageglsnumberformat" )
(indexentry :tkey (("øæ" "\\glossentry{oae}") ) :locref "{}{1}" :attr "pageglsnumberformat" )

so xindy is able to sort the entries correctly.

The drawback is that each indexed entry must be first defined using \newterm and the label can't contain unicode characters such as ø. (This is because the label forms the names of control sequences that store the entry's information.)