%D \module
%D   [       file=lang-mis,
%D        version=1997.03.20, % used to be supp-lan.tex
%D          title=\CONTEXT\ Language Macros,
%D       subtitle=Compounds,
%D         author=Hans Hagen,
%D           date=\currentdate,
%D      copyright={PRAGMA ADE \& \CONTEXT\ Development Team}]
%C
%C This module is part of the \CONTEXT\ macro||package and is
%C therefore copyrighted by \PRAGMA. See mreadme.pdf for
%C details.

\writestatus{loading}{ConTeXt Language Macros / Compounds}

%D More or less replaced.

%D \gdef\starttest#1\stoptest{\starttabulate[|l|l|p|]#1\stoptabulate}
%D \gdef\test     #1{\NC\detokenize{#1}\NC\hyphenatedword{#1}\NC#1\NC\NR}

\unprotect

%D One of \TEX's strong points in building paragraphs is the way
%D hyphenations are handled. Although for real good hyphenation
%D of non||english languages some extensions to the program are
%D needed, fairly good results can be reached with the standard
%D mechanisms and an additional macro, at least in Dutch.

%D \CONTEXT\ originates in the wish to typeset educational
%D materials, especially in a technical environment. In
%D production oriented environments, a lot of compound words
%D are used. Because the Dutch language poses no limits on
%D combining words, we often favor putting dashes between those
%D words, because it facilitates reading, at least for those
%D who are not that accustomed to it.
%D
%D In \TEX\ compound words, separated by a hyphen, are not
%D hyphenated at all. In spite of the multiple pass paragraph
%D typesetting this can lead to parts of words sticking into
%D the margin. The solution lays in saying \type
%D {spoelwater||terugwinunit} instead of \type
%D {spoelwater-terugwinunit}. By using a one character command
%D like \type {|}, delimited by the same character \type {|},
%D we get ourselves both a decent visualization (in \TEXEDIT\
%D and colored verbatim we color these commands yellow) and an
%D efficient way of combining words.
%D
%D The sequence \type{||} simply leads to two words connected by
%D a hyphen. Because we want to distinguish such a hyphen from
%D the one inserted when \TEX\ hyphenates a word, we use a bit
%D longer one.
%D
%D \hyphenation {spoel-wa-ter te-rug-win-unit}
%D
%D \starttest
%D \test {spoelwater||terugwinunit}
%D \stoptest
%D
%D As we already said, the \type{|} is a command. This commands
%D accepts an optional argument before it's delimiter, which is
%D also a \type{|}.
%D
%D \hyphenation {po-ly-meer che-mie}
%D
%D \starttest
%D \test {polymeer|*|chemie}
%D \stoptest
%D
%D Arguments like \type{*} are not interpreted and inserted
%D directly, in contrary to arguments like:
%D
%D \starttest
%D \test {polymeer|~|chemie}
%D \test {|(|polymeer|)|chemie}
%D \test {polymeer|(|chemie|)| }
%D \stoptest
%D
%D Although such situations seldom occur |<|we typeset thousands
%D of pages before we encountered one that forced us to enhance
%D this mechanism|>| we also have to take care of comma's.
%D
%D \hyphenation {uit-stel-len}
%D
%D \starttest
%D \test {op||, in|| en uitstellen}
%D \stoptest
%D
%D The next special case (concerning quotes) was brought to my
%D attention by Piet Tutelaers, one of the driving forces
%D behind rebuilding hyphenation patterns for the dutch
%D language.\footnote{In 1996 the spelling of the dutch
%D language has been slightly reformed which made this topic
%D actual again.} We'll also take care of this case.
%D
%D \starttest
%D \test {AOW|'|er}
%D \test {cd|'|tje}
%D \test {ex|-|PTT|'|er}
%D \test {rock|-|'n|-|roller}
%D \stoptest
%D
%D Tobias Burnus pointed out that I should also support
%D something like
%D
%D \starttest
%D \test {well|_|known}
%D \stoptest
%D
%D to stress the compoundness of hyphenated words.
%D
%D Of course we also have to take care of the special case:
%D
%D \starttest
%D \test {text||color and ||font}
%D \stoptest

%D \macros
%D   {installdiscretionaries}
%D
%D The mechanism described here is one of the older inner parts
%D of \CONTEXT. The most recent extensions concerns some
%D special cases as well as the possibility to install other
%D characters as delimiters. The prefered way of specifying
%D compound words is using \type{||}, which is installed by:
%D
%D \starttyping
%D \installdiscretionary | -
%D \stoptyping
%D
%D Some alternative definitions are:
%D
%D \startbuffer
%D \installdiscretionary * -
%D \installdiscretionary + -
%D \installdiscretionary / -
%D \installdiscretionary ~ -
%D \stopbuffer
%D
%D \typebuffer
%D
%D after which we can say:
%D
%D \start \getbuffer
%D \starttest
%D \test {test**test**test}
%D \test {test++test++test}
%D \test {test//test//test}
%D \test {test~~test~~test}
%D \stoptest
%D \stop

%D \macros
%D   {compoundhyphen,
%D    beginofsubsentence,endofsubsentence}
%D
%D Now let's go to the macros. First we define some variables.
%D In the main \CONTEXT\ modules these can be tuned by a setup
%D command. Watch the (maybe) better looking compound hyphen.

\ifx\compoundhyphen    \undefined \unexpanded\def\compoundhyphen    {\hbox{-\kern-.25ex-}} \fi
\ifx\beginofsubsentence\undefined \unexpanded\def\beginofsubsentence{\hbox{\emdash}}       \fi
\ifx\endofsubsentence  \undefined \unexpanded\def\endofsubsentence  {\hbox{\emdash}}       \fi

%D The last two variables are needed for subsentences
%D |<|like this one|>| which we did not yet mention.
%D
%D We want to enable breaking but at the same time don't want
%D compound characters like |-| or || to be separated from the
%D words. \TEX\ hackers will recognise the next two macro's:

\ifx\prewordbreak \undefined \unexpanded\def\prewordbreak    {\penalty\plustenthousand\hskip\zeropoint\relax} \fi
\ifx\postwordbreak\undefined \unexpanded\def\postwordbreak   {\penalty\zerocount      \hskip\zeropoint\relax} \fi
\ifx\hspaceamount \undefined            \def\hspaceamount#1#2{.16667\emwidth}                                 \fi % language specific

\unexpanded\def\permithyphenation{\ifhmode\prewordbreak\fi} % doesn't remove spaces

%D \macros
%D   {beginofsubsentencespacing,endofsubsentencespacing}
%D
%D In the previous macros we provided two hooks which can be
%D used to support nested sub||sentences. In \CONTEXT\ these
%D hooks are used to insert a small space when needed.

\ifx\beginofsubsentencespacing\undefined \let\beginofsubsentencespacing\relax \fi
\ifx\endofsubsentencespacing  \undefined \let\endofsubsentencespacing  \relax \fi

%D The following piece of code is a torture test compound
%D hndling. The \type {\relax} before the \type {\ifmmode} is
%D needed because of the alignment scanner (in \ETEX\ this
%D problem is not present because there a protected macro is
%D not expanded. Thanks to Tobias Burnus for providing this
%D example.
%D
%D \startformula
%D   \left|f(x_n)-{1\over2}\right| =
%D      {\cases{|{1\over2}-x_n| &for $0\le x_n < {1\over2}$\cr
%D              |x_n-{1\over2}| &for ${1\over2}<x_n\le1$   \cr}}
%D \stopformula

\installcorenamespace{discretionaryaction}
\installcorenamespace{discretionarytext}
\installcorenamespace{discretionarymath}
\installcorenamespace{discretionaryboth}

\unexpanded\def\installdiscretionary#1#2%
  {\setevalue{\??discretionarymath\detokenize{#1}}{\detokenize{#1}}% ?
   \setvalue {\??discretionarytext\detokenize{#1}}{#2}%
   \setvalue {\??discretionaryboth\detokenize{#1}}{\lang_discretionaries_command#1}%
   \scratchcounter\expandafter`\detokenize{#1}%
   \expandafter\uedcatcodecommand\expandafter\ctxcatcodes\expandafter\scratchcounter\csname\??discretionaryboth\detokenize{#1}\endcsname}

\unexpanded\def\handlemathmodediscretionary#1{\executeifdefined{\??discretionarymath\detokenize{#1}}\donothing}
\unexpanded\def\handletextmodediscretionary#1{\executeifdefined{\??discretionarytext\detokenize{#1}}\donothing}

\unexpanded\def\installdiscretionaries#1#2{\writestatus\m!system{use \string \installdiscretionary}} % obsolete

\setnewconstant\discretionarymode\plusone

\unexpanded\def\ignorediscretionaries
  {\discretionarymode\zerocount}

\def\lang_discretionaries_command
  {% if direct if, we need \relax for lookahead in math mode
   \csname
     \ifcase\discretionarymode
       \strippedcsname\lang_discretionaries_process_none
     \else\ifmmode
       \strippedcsname\lang_discretionaries_process_math
     \else
       \strippedcsname\lang_discretionaries_process_text
     \fi\fi
   \endcsname}

\def\lang_discretionaries_process_none#1%
  {\detokenize{#1}}

%D The macro \type{\lang_discretionaries_check_before} takes care
%D of loners like \type{||word}, while it counterpart \type
%D {\lang_discretionaries_check_after} is responsible for handling
%D the comma.

\newsignal\compoundbreakpoint

\newconditional\punctafterdiscretionary
\newconditional\spaceafterdiscretionary

\def\lang_discretionaries_check_before
  {\ifvmode
     \dontleavehmode
   \fi
   \ifhmode
     \begingroup
     \setbox\scratchbox\lastbox
     \ifzeropt\wd\scratchbox
       \let\postwordbreak\prewordbreak
     \fi
     \box\scratchbox\relax
     \endgroup
  \fi}

\def\lang_discretionaries_check_after
  {\setfalse\punctafterdiscretionary
   \setfalse\spaceafterdiscretionary
   \ifx\blankspace\nextnext \settrue \spaceafterdiscretionary \else
   \ifx\space     \nextnext \settrue \spaceafterdiscretionary \else
   \ifx          .\nextnext \settrue \punctafterdiscretionary \else
   \ifx          ,\nextnext \settrue \punctafterdiscretionary \else
   \ifx          :\nextnext \settrue \punctafterdiscretionary \else
   \ifx          ;\nextnext \settrue \punctafterdiscretionary \fi\fi\fi\fi\fi\fi}

\let\lang_discretionaries_process_math\handlemathmodediscretionary

\def\lang_discretionaries_process_text#1% grouped !
  {\bgroup
   \let\nextnextnext\egroup
   \def\next##1#1%
     {\def\next{\activedododotextmodediscretionary#1{##1}}%
      \futurelet\nextnext\next}%
   \next}

\unexpanded\def\activedododotextmodediscretionary#1#2%
  {\edef\discretionarytoken{\detokenize{#2}}%
   \def\textmodediscretionary{\handletextmodediscretionary{#1}}%
   \lang_discretionaries_check_after
   \ifx\discretionarytoken\empty
     \ifx#1\nextnext % takes care of ||| and +++ and ......
       \ifcsname\??discretionaryaction\string#1\endcsname
         \csname\??discretionaryaction\string#1\endcsname
       \else\ifconditional\spaceafterdiscretionary
         \prewordbreak\hbox{\string#1}\relax
       \else\ifconditional\punctafterdiscretionary
         \prewordbreak\hbox{\string#1}\relax
       \else
         \prewordbreak\hbox{\string#1}\prewordbreak
       \fi\fi\fi
       \def\nextnextnext{\afterassignment\egroup\let\next=}%
     \else
       \lang_discretionaries_check_before
       % the next line has been changed (20050203)
       % \prewordbreak\hbox{\textmodediscretionary\nextnext}\allowbreak\postwordbreak
       % but an hbox blocks a possible \discretionary
       \ifcsname\??discretionaryaction\endcsname
         \csname\??discretionaryaction\endcsname
       \else\ifconditional\spaceafterdiscretionary
         \prewordbreak\textmodediscretionary\relax
       \else\ifconditional\punctafterdiscretionary
         \prewordbreak\textmodediscretionary\relax
       \else
         \prewordbreak\textmodediscretionary\prewordbreak
       \fi\fi\fi
     %  \prewordbreak\textmodediscretionary\nextnext\allowbreak\postwordbreak
     \fi
   \else\ifcsname\??discretionaryaction\discretionarytoken\endcsname
     \csname\??discretionaryaction\discretionarytoken\endcsname
   \else
     \lang_discretionaries_check_before
     \ifconditional\spaceafterdiscretionary
       \prewordbreak\hbox{#2}\relax
     \else\ifconditional\punctafterdiscretionary
       \prewordbreak\hbox{#2}\relax
     \else
       \prewordbreak\discretionary{\hbox{#2}}{}{\hbox{#2}}\allowbreak\postwordbreak
     \fi\fi
   \fi\fi
   \nextnextnext} % lookahead in commands

%D \macros
%D   {directdiscretionary}
%D
%D In those situations where the nature of characters is
%D less predictable, we can use the more direct approach:

\unexpanded\def\directdiscretionary
  {\csname
     \ifcase\discretionarymode
       \strippedcsname\lang_discretionaries_process_none
     \else
       \strippedcsname\lang_discretionaries_process_direct
     \fi
   \endcsname}

\unexpanded\def\indirectdiscretionary
  {\csname
     \ifcase\discretionarymode
       \strippedcsname\lang_discretionaries_process_none
     \else
       \strippedcsname\lang_discretionaries_process_indirect
     \fi
   \endcsname}

\unexpanded\def\lang_discretionaries_process_direct#1%
  {\edef\discretionarytoken{\detokenize{#1}}%
   \let\textmodediscretionary\compoundhyphen
   \executeifdefined{\??discretionaryaction\discretionarytoken}{\indirectdiscretionary{#1}}}

\unexpanded\unexpanded\def\lang_discretionaries_process_indirect#1%
  {\prewordbreak\discretionary{\hbox{#1}}{}{\hbox{#1}}\allowbreak\postwordbreak}

\unexpanded\def\definetextmodediscretionary #1
  {\setvalue{\??discretionaryaction\detokenize{#1}}}

% \start \hsize 1mm
% test |||test test|||, test\blank
% test test|-|, test|-| and test|-|test\blank
% test test|_|, test|_| and test|_|test\blank
% test cd|'|tje\blank
% test |(|test test|)|, test\blank
% test test test|x|, test\blank
% test|~|test
% test|^|test
% \stop

% x\discretionary{1}{2}{3}xxxxxxx
% xxxxxxx\discretionary{1}{2}{3}x
%
% xxx3xxx
% xxx1<newline>2xxx

\def\lang_discretionaries_hyphen_like#1#2%
  {\ifconditional\spaceafterdiscretionary
     \prewordbreak\hbox{#1}\relax
   \else\ifconditional\punctafterdiscretionary
     \prewordbreak\hbox{#1}\relax
   \else
     \prewordbreak#2\postwordbreak % was prewordbreak
   \fi\fi}

\definetextmodediscretionary {}
  {\lang_discretionaries_hyphen_like\textmodehyphen\textmodehyphendiscretionary}

\definetextmodediscretionary -
  {\lang_discretionaries_hyphen_like\normalhyphen\normalhyphendiscretionary}

\definetextmodediscretionary _
  {\lang_discretionaries_hyphen_like\composedhyphen\composedhyphendiscretionary}

\definetextmodediscretionary )
  {\lang_discretionaries_hyphen_like{)}{\discretionary{-)}{}{)}}}

\definetextmodediscretionary (
  {\ifdim\lastskip>\zeropoint
     (\prewordbreak
   \else
     \prewordbreak\discretionary{}{(-}{(}\prewordbreak
   \fi}

\definetextmodediscretionary ~
  {\prewordbreak\discretionary{-}{}{\thinspace}\postwordbreak}

\definetextmodediscretionary '
  {\prewordbreak\discretionary{-}{}{'}\postwordbreak}

\definetextmodediscretionary ^
  {\prewordbreak\discretionary{\hbox{$|$}}{}{\hbox{$|$}}%
   \allowbreak\postwordbreak} % bugged

\definetextmodediscretionary <
  {\beginofsubsentence\prewordbreak\beginofsubsentencespacing}

\definetextmodediscretionary >
  {\endofsubsentencespacing\prewordbreak\endofsubsentence}

\definetextmodediscretionary =
  {\prewordbreak\midsentence\prewordbreak} % {\prewordbreak\compoundhyphen}

% french

\definetextmodediscretionary : {\removeunwantedspaces\prewordbreak\kern\hspaceamount\empty{:}:}
\definetextmodediscretionary ; {\removeunwantedspaces\prewordbreak\kern\hspaceamount\empty{;};}
\definetextmodediscretionary ? {\removeunwantedspaces\prewordbreak\kern\hspaceamount\empty{?}?}
\definetextmodediscretionary ! {\removeunwantedspaces\prewordbreak\kern\hspaceamount\empty{!}!}

\definetextmodediscretionary *
  {\prewordbreak\discretionary{-}{}{\kern.05em}\prewordbreak}

% spanish

\definetextmodediscretionary ?? {\prewordbreak\questiondown}
\definetextmodediscretionary !! {\prewordbreak\exclamdown}

% \ifx\normalcompound\undefined \let\normalcompound=| \fi

%D \installdiscretionary | +
%D \installdiscretionary + =

\def\defaultdiscretionaryhyphen{\compoundhyphen}

\installdiscretionary | \defaultdiscretionaryhyphen % installs in ctx and prt will fall back on it

%D \macros
%D   {fakecompoundhyphen}
%D
%D In headers and footers as well as in active pieces of text
%D we need a dirty hack. Try to imagine what is needed to
%D savely break the next text across a line and at the same
%D time make the words interactive.
%D
%D \starttyping
%D \goto{Some||Long||Word}
%D \stoptyping

\unexpanded\def\fakecompoundhyphen
  {\def\|{\mathortext\vert\lang_compounds_fake_hyphen}}

\def\lang_compounds_fake_hyphen
  {\def##1|%
     {\doifelsenothing{##1}\compoundhyphen{##1}%
      \kern\compoundbreakpoint\allowbreak}}

%D \macros
%D   {midworddiscretionary}
%D
%D If needed, one can add a discretionary hyphen using \type
%D {\midworddiscretionary}. This macro does the same as
%D \PLAIN\ \TEX's \type {\-}, but, like the ones implemented
%D earlier, this one also looks ahead for spaces and grouping
%D tokens.

\unexpanded\def\midworddiscretionary
  {\futurelet\nexttoken\lang_discretionaries_mid_word}

\def\lang_discretionaries_mid_word
  {\ifx\nexttoken\blankspace\else
   \ifx\nexttoken\bgroup    \else
   \ifx\nexttoken\egroup    \else
     \discretionary{-}{}{}%
   \fi\fi\fi}

%D \macros
%D   {installcompoundcharacter}
%D
%D When Tobias Burnus started translating the dutch manual of
%D \PPCHTEX\ into german, he suggested to let \CONTEXT\ support
%D the \type{german.sty} method of handling compound
%D characters, especially the umlaut. This package is meant for
%D use with \PLAIN\ \TEX\ as well as \LATEX.
%D
%D I decided to implement compound character support as
%D versatile as possible. As a result one can define his own
%D compound character support, like:
%D
%D \starttyping
%D \installcompoundcharacter "a {\"a}
%D \installcompoundcharacter "e {\"e}
%D \installcompoundcharacter "i {\"i}
%D \installcompoundcharacter "u {\"u}
%D \installcompoundcharacter "o {\"o}
%D \installcompoundcharacter "s {\SS}
%D \stoptyping
%D
%D or even
%D
%D \starttyping
%D \installcompoundcharacter "ck {\discretionary {k-}{k}{ck}}
%D \installcompoundcharacter "ff {\discretionary{ff-}{f}{ff}}
%D \stoptyping
%D
%D The support is not limited to alphabetic characters, so the
%D next definition is also valid.
%D
%D \starttyping
%D \installcompoundcharacter ". {.\doifnextcharelse{\spacetoken}{}{\kern.125em}}
%D \stoptyping
%D
%D The implementation looks familiar and uses the same tricks as
%D mentioned earlier in this module. We take care of two
%D arguments, which complicates things a bit.

\installcorenamespace{compoundnormal}
\installcorenamespace{compoundsingle}
\installcorenamespace{compoundmultiple}
\installcorenamespace{compounddefinition}

%D When we started working on MK IV code, we needed a different
%D approach for defining the active character itself. In MK II as
%D well as in MK IV we now use the catcode vectors.

\setnewconstant\compoundcharactermode\plusone

\newcount\c_lang_compounds_character

\def\installcompoundcharacter #1#2#3 #4% {#4} no grouping
  {\ifcase\compoundcharactermode
      % ignore mode
   \else
     \chardef\c_lang_compounds_character`#1%
     \expandafter\chardef\csname\??compoundnormal\string#1\endcsname\c_lang_compounds_character
     \def\!!stringa{#3}%
     \expandafter\def\csname\ifx\!!stringa\empty\??compoundsingle\else\??compoundmultiple\fi\detokenize{#1#2#3}\endcsname{#4}%
     \setevalue{\??compounddefinition\detokenize{#1}}{\noexpand\lang_compounds_handle_character{\detokenize{#1}}}% beter nr's
     \expandafter\letcatcodecommand\expandafter\ctxcatcodes\expandafter\c_lang_compounds_character\csname\??compounddefinition\detokenize{#1}\endcsname
   \fi}

%D We can also ignore definitions (needed in for instance \XML). Beware,
%D this macro is supposed to be used grouped!

\def\ignorecompoundcharacter
  {\compoundcharactermode\zerocount}

%D In handling the compound characters we have to take care of
%D \type{\bgroup} and \type{\egroup} tokens, so we end up with
%D a multi||step interpretation macro. We look ahead for a
%D \type{\bgroup}, \type{\egroup} or \type{\blankspace}. Being
%D no user of this mechanism, the credits for testing them goes
%D to Tobias Burnus, the first german user of \CONTEXT.
%D
%D We define these macros as \type{\long} because we can
%D expect \type{\par} tokens. We need to look into the future
%D with \type{\futurelet} to prevent spaces from
%D disappearing.

\def\lang_compounds_handle_character#1%
  {\def\lang_compounds_handle_character_finish{\lang_compounds_handle_character_finish_indeed{#1}}%
   \futurelet\nexttoken\xhandlecompoundcharacter}

\def\lang_compounds_handle_character_finish_indeed
  {\ifx\nexttoken\bgroup
    %\expandafter\lang_compounds_handle_character_pickup % handle "{ee} -> \"ee
    %\expandafter\gobbleoneargument                          % forget "{ee} -> ee
     \expandafter\lang_compounds_handle_character_one    % ignore "{ee} -> "ee
   \else\ifx\nexttoken\egroup
     \doubleexpandafter\lang_compounds_handle_character_normal
   \else\ifx\nexttoken\blankspace
     \tripleexpandafter\lang_compounds_handle_character_normal
   \else
     \tripleexpandafter\lang_compounds_handle_character_pickup
   \fi\fi\fi}

\def\lang_compounds_handle_character_normal#1%
  {\csname\??compoundnormal\string#1\endcsname}

\def\lang_compounds_handle_character_pickup#1#2% preserve space
  {\def\lang_compounds_handle_character_finish{\lang_compounds_handle_character_finish_indeed#1#2}%
   \futurelet\nexttoken\lang_compounds_handle_character_finish}

\def\lang_compounds_handle_character_finish_indeed
  {\ifx\nexttoken\bgroup
     \expandafter\lang_compounds_handle_character_one
   \else\ifx\nexttoken\egroup
     \doubleexpandafter\lang_compounds_handle_character_one
   \else\ifx\nexttoken\blankspace
     \tripleexpandafter\lang_compounds_handle_character_one
   \else
     \tripleexpandafter\lang_compounds_handle_character_two
   \fi\fi\fi}

%D Besides taken care of the grouping and space tokens, we have
%D to deal with three situations. First we look if the next
%D character equals the first one, if so, then we just insert
%D the original. Next we look if indeed a compound character is
%D defined. We either execute the compound character or just
%D insert the first. So we have
%D
%D \starttyping
%D <key><key>  <key><known>  <key><unknown>
%D \stoptyping
%D
%D In later modules we will see how these commands are used.

\def\lang_compounds_handle_character_one#1#2%
  {\if\string#1\string#2% was: \ifx#1#2%
     \def\next{\csname\??compoundnormal\string#1\endcsname}%
   \else\ifcsname\??compoundsingle\string#1\string#2\endcsname
     \def\next{\csname\??compoundsingle\string#1\string#2\endcsname}%
   \else
     \def\next{\csname\??compoundnormal\string#1\endcsname#2}%
   \fi\fi
   \next}

\def\lang_compounds_handle_character_two#1#2#3%
  {\if\string#1\string#2%
     \def\next{\csname\??compoundnormal\string#1\endcsname#3}%
   \else\ifcsname\??compoundmultiple\string#1\string#2\string#3\endcsname
     \def\next{\csname\??compoundmultiple\string#1\string#2\string#3\endcsname}%
   \else\ifcsname\??compoundsingle\string#1\string#2\endcsname
     \def\next{\csname\??compoundsingle\string#1\string#2\endcsname#3}%
   \else
     \def\next{\csname\??compoundnormal\string#1\endcsname#2#3}%
   \fi\fi\fi
   \next}

%D For very obscure applications (see for an application \type
%D {lang-sla.tex}) we provide:

\def\simplifiedcompoundcharacter#1#2%
  {\ifcsname\??compoundsingle\string#1\string#2\endcsname
     \doubleexpandafter\firstofoneargument\csname\??compoundsingle\string#1\string#2\endcsname
   \else
     #2%
   \fi}

%D \macros
%D   {disablediscretionaries,disablecompoundcharacter}
%D
%D Occasionally we need to disable this mechanism. For the
%D moment we assume that \type {|} is used.

\let\disablediscretionaries   \ignorediscretionaries
\let\disablecompoundcharacters\ignorecompoundcharacter

%D \macros
%D   {normalcompound}
%D
%D Handy in for instance XML. (Kind of obsolete)

\ifdefined\normalcompound \else \let\normalcompound=| \fi

%D \macros
%D   {compound}
%D
%D We will overload the already active \type {|} so we have
%D to save its meaning in order to be able to use this handy
%D macro.
%D
%D \starttyping
%D so test\compound{}test can be used instead of test||test
%D \stoptyping

\bgroup

    \catcode\barasciicode\activecatcode

    \unexpanded\gdef\compound#1{|#1|}

    \doglobal \appendtoks
        \def|#1|{\ifx#1\empty\empty-\else#1\fi}%
    \to \everysimplifycommands

\egroup

%D Here we hook some code into the clean up mechanism needed
%D for verbatim data.

\appendtoks
    \disablecompoundcharacters
    \disablediscretionaries
\to \everycleanupfeatures

\protect \endinput