path: root/Master/texmf-dist/doc/context/manuals/reference/en/co-language.tex

\startcomponent co-language

\environment contextref-env
\product contextref

\def\ShowComposed #1
  {\handletokens#1\with\type\VL#1\VL\hyphenatedword{#1}}

\startbuffer[hyph-1]
\starttable[|l|l|l|]
\HL
\VL \bf input \VL \bf normal \VL \bf hyphenated \VL\SR
\HL
\VL \ShowComposed intra||word   \VL\FR
\VL \ShowComposed intra|-|word  \VL\MR
\VL \ShowComposed intra|(|word) \VL\MR
\VL \ShowComposed (intra|)|word \VL\MR
\VL \ShowComposed intra|--|word \VL\MR
\VL \ShowComposed intra|~|word  \VL\LR
\HL
\stoptable
\stopbuffer

\startbuffer[messian]
The French composer {\fr Olivier Messiaen} wrote \quote {\fr Quatuor pour
la fin du temps} during the World War II in a concentration camp. This
may well be one of the most moving musical pieces of that period.
\stopbuffer

\chapter[hyphenation,languages]{Language specific issues}

\section{Introduction}

One of the more complicated corners of \CONTEXT\ is the
department that deals with languages. Fortunately users will
seldom notice this, but each language has its own demands
and we put quite some effort in making sure that most of the
issues on hyphenation rules and accented and non latin
characters could be dealt with. For as long as it does not
violate the \CONTEXT\ user interface, we also support
existing input schemes. 

In the early days \TEX\ was very American oriented, but
since \TEX\ version~3 there is (simultaneous) support for
multiple languages. The input of languages with many accents
|<|sometimes more accents per character|>| may look rather
complicated, depending on the use of dedicated input
encodings or special \TEX\ commands.  

The situation is further complicated by the fact that
specific input does not have a one||to||one relation with
the position of a glyph in a font. We discussed this in \in
{section}[encoding]. It is important to make the right
choices for input and font encoding. 

In this chapter we will deal with hyphenation and language
specific labels. More details can be found in the language
definition files (\type {lang-<<xxx>>}), the font files
(\type{font-<<xxx>>}) and the encoding files (\type
{enco-<<xxx>>}). There one can find details on how to define
commands that deal with accents and special characters as
covered in a previous chapter, sorting indexes, providing
support for \UNICODE, and more. 

\section{Automatic hyphenating}
\index{languages}
\index{hyphenation}
\macro{\tex{taal}}
\macro{\tex{nl}}
\macro{\tex{en}}
\macro{\tex{fr}}
\macro{\tex{de}}
\macro{\tex{sp}}

Each language has its own hyphenation rules. As soon as you
switch to another language, \CONTEXT\ will activate the
appropriate set of hyphenation patterns for that language.
Languages are identified by their official two character
identifiers, like: Dutch (\type{nl}), English (\type{en}),
German (\type{de}) and French (\type{fr}). A language is
chosen with the following command: \footnote{In case of any
doubt please check if the hyphenation patterns are included
in the \type {fmt}||file.}

\showsetup{language}

Some short cut  commands are also available. They can be used
enclosed in braces: 

\startexample
\starttyping
\nl  \en  \de  \fr  \sp  \uk  \pl  \cz  ...
\stoptyping
\stopexample

The command \type {\language[nl]} can be compared with
\type {\nl}. The first command is more transparant. The two
character commands may conflict with existing commands.
Take, for example, Italian and the code for {\it italic}
type setting. For this reason we use capitals for commands
that may cause any conflicts. One may also use the full 
names, like \type {czech}. 

At any instance you can switch to another language. In the
example below we switch from English to French and vice
versa. 

\startexample
\typebuffer[messian]
\stopexample

We use these language switching commands if we cannot be
certain that an alternative hyphenation pattern is
necessary. 

\startcolumns[n=4,,rule=on]
\getbuffer[messian]
\stopcolumns

How far do we go in changing languages. Borrowed words like
perestrojka and glasnost are often hyphenated okay, since
these are Russian words used in an English context. When
words are incorrectly hyphenated you can define an
hyphenation pattern with the \TEX||command: 

\startexample
\starttyping
\hyphenation{<<ab-bre-via-tion>>}
\stoptyping
\stopexample

You can also influence the hyphenation in a text by 
indicating the allowed hyphenation pattern in the word:
at the right locations the command \type{\-}
is added: \type{al}\type{\-}\type{lo}\type{\-}\type{wed}.

\section{Definitions and setups}
\macro{\tex{installlanguage}}
\macro{\tex{setuplanguage}}

When a format file is generated the hyphenation pattern one
needs should be added to this file. The definition and
installation of a language is therefore not transparant for
the user. We show the process to give some insight in the
mechanism. An example: \footnote {The somewhat strange name
\type {\upperleftsinglesixquote} is at least telling us what
the quote will look like.} 
 
\starttyping
\installlanguage
  [en]
  [spacing=broad,
   leftsentence=---,
   rightsentence=---,
   leftsubsentence=---,
   rightsubsentence=---,
   leftquote=\upperleftsinglesixquote,
   rightquote=\upperrightsingleninequote,
   leftquotation=\upperleftdoublesixquote,
   rightquotation=\upperrightdoubleninequote,
   date={month,\ ,day,{,\ },year},
   default=en,
   state=stop] 
\stoptyping

and:

\starttyping
\installlanguage
  [uk]
  [default=en,
   state=stop]
\stoptyping

With the first definition you define the language component.
You can view this definition in the file \type
{lang-ger.tex}, the german languages. Languages are arranged
in language groups. This arrangement is of no further
significance at the moment. Since language definitions are 
preloaded, users should not bother about setting up such 
files. 

The second definition inherits its set up from the English
installation. In both definitions \type {state} is set at
\type {stop}. This means that no patterns are loaded yet.
That is done in the files \type{cont-<<xx>>}, the language
and interface specific \CONTEXT\ versions. As soon as \type
{state} is set at \type {start}, a new pattern is loaded,
which can only be done during the generation of a format
file. 

We use some conventions in the file names of the patterns
\type {lang-xx.pat} and the exceptions \type {lang-xx.hyp}.
Normally a language is installed with a two character code.
However there are three character codes, like {deo} for
hyphenating \quote {old deutsch} and \type {nlx} the Dutch
extended characterset, or 8||bit encoding. On distributions 
that come with patterns, the filenames mentioned can be mapped 
onto the ones available on the system. This happens in the 
file \type {cont-usr.tex}.  

After installation you are not bound to the two character
definitions. Default the longer (English) equivalents are
defined: 

\starttyping
\installlanguage[german][de]
\stoptyping

\showsetup{installlanguage}

\showsetup{setuplanguage}

The setup in these commands relate to the situations that are
shown below. 

\startbuffer
\currentdate
|<|all right there we go|>|
|<| |<|all right|>| there we go|>|
|<|all right |<|there|>| we go|>|
\quote{all right there we go}
\quotation{all right there we go}
\quotation{\quote{all right} there we go}
\quotation{all right \quote{there} we go}
\stopbuffer

\typebuffer

This becomes: 

\startlines
\getbuffer 
\stoplines

We will discuss \type{||} in one of the next sections.

\section[datum]{Date}
\index{date}
\macro{\tex{currentdate}}
\macro{\tex{date}}

Typesetting a date is also language specific so we have to
pay some attention to dates here. When the computer runs at
the actual time and date the system date can be recalled
with: 

\showsetup{currentdate}

The sequence in which \type {day}, \type {month} and \type
{year} are given is not mandatory. The pattern \type
{[day,month,year]} results in \currentdate [day,month,year].
We use \type {\currentdate[weekday,month,day,{,},year]} to
obtain \currentdate[weekday,month,day,{,},year]. 

A short cut looks like this: \type {[dd,mm,yy]} and will
result in \currentdate [dd,mm,yy]. Something like \type
{[d,m,y]} would result in \currentdate [d,m,y] and with
\type {[referral]} you will get a \currentdate [referral].
Combinations are also possible. Characters can also be added
to the date pattern. The date \currentdate [dd,--,mm,--,yy] is
generated by the pattern \type {[dd,--,mm,--,yy]}. 

A date can be (type|)|set with the command:

\showsetup{date}

The first (optional) argument is used to specify the date: 

\startbuffer
\date[d=10,m=3,y=1996][weekday,month,day, year]
\stopbuffer

\startexample
\typebuffer
\stopexample

When no argument is given you will obtain the actual date.
When the second argument is left out the result equals that
of \type {\currentdate}. The example results in: 

\startreality
\getbuffer
\stopreality

\section[labels]{Labels and heads}
\index{labels}
\index{heads}
\macro{\tex{mainlanguage}}
\macro{\tex{setuplabeltext}}
\macro{\tex{setupheadtext}}
\macro{\tex{labeltext}}
\macro{\tex{headtext}}

In some cases \CONTEXT\ will generate text labels
automatically, for example the word {\bf Figure} is
generated automatically when a caption is placed under a
figure. These kind of words are called textlabels. Labels
can be set with the command: 

\showsetup{setuplabeltext}

Relevant labels are: \type {table}, \type {figure}, \type
{chapter}, \type {appendix} and comparable text elements. An
example of such a set up is: 

\startexample
\starttyping
\setuplabeltext[en][chapter=Chapter ]
\setuplabeltext[nl][hoofdstuk=Hoofdstuk ]
\stoptyping
\stopexample

The space after \type{Chapter} is essential, because
otherwise the chapternumber will be placed right after the
word Chapter (Chapter1 instead of Chapter~1). A labeltext
can recalled with: 

\showsetup{labeltext}

Some languages, like Chinese, use split labels. These can be 
passed as a comma separated list, like  \type
{chapter={left,right}}. 

Titleheads for special sections of a document, like
abbreviations and appendices are set up with: 

\showsetup{setupheadtext}

Examples of titleheads are \type{Content}, \type{Tables},
\type{Figures}, \type{Abbreviations}, \type{Index} etc. An
example definition looks like: 

\startexample
\starttyping
\setupheadtext[content=Content]
\stoptyping
\stopexample

A header can be recalled with:

\showsetup{headtext}

Labels and titleheads are defined in the file \type
{lang-<<xxx>>}. You should take a look in these files to
understand the use of titleheads and labels. 

The actual language that is active during document
generation does not have to be the same language that is
used for the labels. For this reason next to \type
{\language} we have: 

\showsetup{mainlanguage}

When typesetting a document, there is normally one main
language, say \type {\mainlanguage[en]}. A temporary switch
to another language is then accomplished by for instance
\type {\language[nl]}, since this does not influence the
labels and titles. language. 

\section{Language specific commands}
\index{german}

German \TEX\ users are accustomed to entering \type {"e} and
getting \"e typeset in return. This and a lot more are
defined in \type {lang-ger} using the compound character
mechanism built in \CONTEXT. Certain two or three character
combinations result in one glyph or proper hyphenation. The
example below illustrates this. Some macros are used that
will not be explained here. Normally, users can stick to 
simply using the already defined commands. 

\starttyping
\startlanguagespecifics[de]
  \installcompoundcharacter "a  {\moveaccent{-.1ex}\"a\midworddiscretionary}
  \installcompoundcharacter "s  {\SS}
  .....
  \installcompoundcharacter "U  {\smashaccent\"U}
  \installcompoundcharacter "Z  {SZ}
  .....
  \installcompoundcharacter "ck {\discretionary {k-}{k}{ck}}
  \installcompoundcharacter "TT {\discretionary{TT-}{T}{TT}}
  .....
  \installcompoundcharacter "`  {\handlequotation\c!leftquotation}
\stoplanguagespecifics
\stoptyping

The command \type {\installcompoundcharacter} takes care of
the German type setting, \type {"a} is converted to {\de
"a}, \type {"U} in {\de "U}, \type {"ck} for the right
hyphenation, etc. One can add more definitions, but this will
violate portability. In a Polish \CONTEXT\ the \type {/} is 
used instead of a \type {"}. 

\section{Automatic translation}
\index{translate}
\macro{\tex{translate}} 

It is possible to translate a text automatically in the
actual language. This may be comfortable when typesetting
letterheads. The example below illustrates this. 

\showsetup{translate}

\startbuffer
It depends on the actual language whether a labeltext is type set in 
English {\en as an \translate [en=example, fr=exemple], \fr or in French
as an \translate}.
\stopbuffer

\startexample
\typebuffer
\stopexample

The second command call \type{\translate} uses the applied
values. That is, \type {\translate} with no options uses the 
options of the last call to \type {\translate}.  

\startreality
\getbuffer
\stopreality

\section{Composed words}
\index{hyphen}
\macro{\tex{setuphyphenmark}}

Words consisting of two separate words are often separated
by an intra word dash, as in x||axis. This dash can be
placed between \type{| |}, for example \type{|--|}. This
command, which does not begin with a \texescape, serves
several purposes. When \type{||} is typed the default intra
word dash is used, which is \type{--}. This dash is set up
with: 

\showsetup{setuphyphenmark}

The \type {| |} is also used in word combinations like
(intra)word, which is typed as \type {(intra|)|word}. The
mechanism is not foolproof but it serves most purposes. In
case the hyphenation is incorrect you can hyphenate the
first word of the composed one by hand: \type
{(in\-tra|)|word}. 

\placetable
  {Hyphenation of composed words.}
  {\getbuffer[hyph-1]}

The main reason behind this mechanism is that \TEX\ doesn't
really know how to hyphenate composed words and how to
handle subsentences. \TEX\ know a lot about math, but far
less about normal texts. Using this command not only serves
consistency, but also makes sure that \TEX\ can break
compound words at the right places. It also keeps boundary
characters at the right place when a breakpoint is inserted. 

\section{Pattern files manual}

\todo{A large part of this section is obsolete}

\TEX\ has two mysterious commands that the average user will never
or seldom meet:

\starttyping
\hyphenation{as-so-ciates}
\patterns   {.ach4}
\stoptyping

Both commands can take multiple strings, so in fact both commands
should be plural. The first command can be given any time and can be
used to tell \TEX\ that a word should be hyphenated in a certain
way. The second command can only be issued when \TEX\ is in virgin
mode, i.e.\ starting with a clean slate. Normally this only happens
when a format is generated.

The second command is more mysterious than the first one and its
entries are a compact way to tell \TEX\ between what character
sequences it may hyphenate words. The numbers represent weights and
the (often long) lists of such entries are generated with a special
program called \type {patgen}. Since making patterns is work for
specialists, we will not go into the nasty details here.

In the early stage of \CONTEXT\ development it came with its own
pattern files. Their names started with \type{lang-} and their
suffixes were \type {pat} and \type {hyp}.

However, when \CONTEXT\ went public, I was convinced to drop those
files and use the files already available in distributions. This was
achieved by using the \CONTEXT\ filename remapping mechanism. Although
those files are supposed to be generic, this is not always the case,
and it remains a gamble if they work with \CONTEXT. Even worse,
their names are not consistent and the names of some files as well
as locations in the tree keep changing. The price \CONTEXT\ users
pay for this is lack of hyphenation until such changes are noticed
and taken care of. Because constructing the files is an uncoordinated
effort, all pattern files have their own characteristics, most
noticably their encoding.

After the need to adapt the name mapping once again, I decided to
get back to providing \CONTEXT\ specific pattern files. Pattern
cooking is a special craft and \TEX\ users may call themselves
lucky that it's taken care of. So, let's start with thanking all
those \TEX\ experts who dedicate their time and effort to get
their language hyphenated. It's their work we will build (and keep
building) upon.

In the process of specific \CONTEXT\ support, we will take care of:

\startitemize
\item consistent naming, i.e.\ using language codes when possible as
      a prelude to a more sophisticated naming scheme, taking versions
      into account
\item consistent splitting of patterns and hyphenation exceptions in
      files that can be recognized by their suffix
\item making the files encoding independent using named glyphs
\item providing a way to use those patterns in plain \TEX\ as well
\stopitemize

Instead of using a control sequence for the named glyphs, we use a
different notation:

\starttyping
[ssharp] [zcaron] [idiaeresis]
\stoptyping

The advantage of this notation is that we don't have to mess with
spacing so that parsing and cleanup with scripts becomes more robust. The
names conform to the \CONTEXT\ way of naming glyphs and the names
and reverse mappings are taken from the encoding files in the
\CONTEXT\ distribution, so you need to have \CONTEXT\ installed.

The \CONTEXT\ pattern files are generated by a \RUBY\ script. Although
the converting is rather straightforward, some languages need special
treatment, but a script is easily adapted. If you want a whole bunch of
pattern files, just say:

\starttyping
ctxtools --patterns all
\stoptyping

or, if you want one language:

\starttyping
ctxtools --patterns nl
\stoptyping

If for some reason this program does not start, try:

\starttyping
texmfstart ctxtools --patterns nl
\stoptyping

When things run well, this will give you four files:

\starttabulate[|l|p|]
\NC \type{lang-nl.pat} \NC the patterns in an encoding indepent format \NC \NR
\NC \type{lang-nl.hyp} \NC the hyphenation exceptions \NC \NR
\NC \type{lang-nl.log} \NC the conversion log (can be deleted afterwards) \NC \NR
\NC \type{lang-nl.rme} \NC the preambles of the files used (copyright notices and such) \NC \NR
\stoptabulate

If you redistribute the files, it makes sense to bundle the \type
{rme} files as well, unless the originals are already in the
distribution. It makes no sense to keep the log files on your
system. When the file \type {lang-all.xml} is present, the info
from that file will be used and added to the pattern and
hyphenation files. In that case no \type {rme} and \type {log}
file will be generated, unless \type {--log} is provided.

In the Dutch pattern file you will notice entries like the
following:

\starttyping
e[ediaeresis]n3
\stoptyping

So, instead of those funny (encoding specific) \type {^^fc} or
(format specific) \type {\"e} we use names. Although this looks
\CONTEXT\ dependent it is rather easy to map those names back to
characters, especially when one takes into account that most
languages only have a few of those special characters and we only
have to deal with lower case instances.

The \CONTEXT\ support module \type {supp-pat.tex} is quite generic and
contains only a few lines of code. Actually, most of the code is
dedicated to the simple \XML\ handler. Loading a pattern meant
for EC encoded fonts in another system than \CONTEXT\ is done as
follows:

\starttyping
\bgroup

  \input supp-pat

  \lccode"E4="E4  \definepatterntoken adiaeresis ^^e4
  \lccode"F6="F6  \definepatterntoken odiaeresis ^^f6
  \lccode"FC="FC  \definepatterntoken ediaeresis ^^fc
  \lccode"FF="FF  \definepatterntoken ssharp     ^^ff

  \enablepatterntokens
  \enablepatternxml

  \input lang-de.pat
  \input lang-de.hyp

\egroup
\stoptyping

In addition to this one may want to set additional lower and
uppercase codes. In \ETEX\ these are stored with the language.

Just for completeness we provide the magic command to generate
the \XML\ variants:

\starttyping
ctxtools --patterns --xml all
\stoptyping

This will give you files like:

\starttyping
<?xml version='1.0' standalone='yes'?>

<!-- some comment -->

<patterns>
... e&ediaeresis;n3 ...
</patterns>
\stoptyping

This is also accepted as input but for our purpose it's probably
best to stick to the normal method. The pattern language is a \TEX\
specific one anyway.

\section{Installing languages}

Installing a language in \CONTEXT\ should not take too much effort
assuming the language is supported. Language specific labels are
grouped in \type {lang-*} files, like \type {lang-ger.tex} for the
germanic languages.

Patterns will be loaded from the files in the general \TEX\
distribution unless \type {lang-nl.pat} is found, in which case
\CONTEXT\ assumes that you prefer the \CONTEXT\ patterns. In that
case, run

\starttyping
ctxtools --patterns all
\stoptyping

You need to move the files to the \CONTEXT\ base path that you can
locate with:

\starttyping
textools --find context.tex
\stoptyping

You can also use \type {kpsewhich}, but the above method does an
extensive search. Of course you can also generate the files on a
temporary location. Now it's time to generate the formats:

\starttyping
texexec --make --all
\stoptyping

Since \XETEX\ needs patterns in \UTF-8 encoding, we provide a switch
for achieving that:

\starttyping
texexec --make --all --utf8
\stoptyping

Beware: you need to load patterns for each language and encoding
combination you are going to use. You can configure your local
\type {cont-usr} file to take care of this. When an encoding does
not have the characters that are needed, you will get an error.
When using the non \CONTEXT\ versions of teh patterns this may go
unnoticed because the encoding is hard coded in the file. Of
course it will eventually get noticed when the hyphenations come
out wrong.

The \CONTEXT\ distribution has a file lang-all.xml that holds the
copyright and other notes of the patterns. A discription looks like:

\starttyping
<description language='nl'>
  <sourcefile>nehyph96.tex</sourcefile>
  <title>TeX hyphenation patterns for the Dutch language</title>
  <copyright>
    <year>1996</year>
    <owner> Piet Tutelaers (P.T.H.Tutelaers@tue.nl)</owner>
    <comment>8-bit hyphenation patterns for TeX based upon the new
      Dutch spelling, officially since 1 August 1996. These
      patterns follow the new hyphenation rules in the
      `Woordenlijst Nederlandse Taal, SDU Uitgevers, Den Haag
      1995' (the so called `Groene Boekje') described in
      section 5.2 (Het afbreekteken)</comment>
  </copyright>
</description>
\stoptyping

{\em This file is \quote {work in process}: more details will be added
and comments will be enriched.}

\section {Commands}

You can at any moment add additional hyphenation exceptions to the
language specific dictionaries. For instance:

\starttyping
\language[nl] \hyphenation{pa-tiën-ten}
\stoptyping

Switching to another language is done with the \type {\language}
command. The document language is set with \type {\mainlanguage}.

If you want to let \TEX\ know that a word should be hyphenated in a
special way, you use the \type {\-} command, for instance:

\starttyping
Con\-TeXt
\stoptyping

Compound words are not recognized by the hyphenation engine, so there
you need to add directives, like:

\starttyping
the ConTeXt|-|system
\stoptyping

If you are using \XML\ as input format, you need to load the
hyphenation filter module. Here we assume that \UTF\ encoding
is used:

\starttyping
\useXMLfilter[utf,hyp]
\stoptyping

In your \XML\ file you can now add:

\starttyping
<hyphenations language='nl' regime='utf'>
  <hyphenation>pa-tiën-ten</hyphenation>
  <hyphenation>pa-tiën-ten-or-ga-ni-sa-tie</hyphenation>
  <hyphenation>pa-tiën-ten-plat-form</hyphenation>
</hyphenations>
\stoptyping

This filter also defines some auxiliary elements. Explicit
hyphenation points can be inserted as follows:

\starttyping
Zullen we hier af<hyphenate/>bre<hyphenate/>ken of niet?
\stoptyping

The compound token can be anything, but keep in mind that some
tokens are treated special (see other manuals).

\starttyping
Wat is eigenlijk een patiënten<compound token="-"/>platform?
\stoptyping

A language is set with:

\starttyping
nederlands <language code="en">english</language> nederlands
\stoptyping

If you set attribute \type {scope} to \type {global}, labels (as
used for figure captions and such) adapt to the language switch.
This option actually invokes \type {\mainlanguage}.

\section {Languages}

When users in a specific language area use more than one font
encoding, patterns need to be loaded multiple times. In theory
this means that one can end up with more instances than \TEX\ can
host. However, the number of sensible font encodings is limited as
is the number of languages that need hyphenation. Now that memory
is cheap and machines are fast, preloading a lot of pattern files
is no problem. The following table shows the patterns that are
preloaded in the version of \CONTEXT\ that is used to process this
file.

% \showpatterns

\fixme{\type{\showpatterns} doesn't exist anymore}

{\em In the (near) future the somewhat arcane \type {pl0} and \type
{il2} encodings will go away since they are only used for Polish
and Czech/Slovak computer modern fonts, which can be replaced by
Latin Modern alternatives. Also, a new dense encoding may find its
way into this list.}

\section{Hyphenation}

If you want to know what patterns are used, you can try to
hyphenate a word with \type {\showhyphenations}.

\showhyphenations{abracadabra}



While hypenating, \TEX\ has to deal with ligatures as well. While
Thomas, Taco and I were discussing the best ways to neutralize the
ancient greek patterns, Taco Hoekwater came up with the following
explanation.\footnote {Thomas Schmitz is responsible
for the associated third party module.}

\placefigure
  [nonumber]
  {The most common ligatures.}
  {\startcombination[4*1]
     {\scale[height=3\lineheight]{\color[AltColor]{fi}}}  {}
     {\scale[height=3\lineheight]{\color[AltColor]{fl}}}  {}
     {\scale[height=3\lineheight]{\color[AltColor]{ffi}}} {}
     {\scale[height=3\lineheight]{\color[AltColor]{ffl}}} {}
   \stopcombination}

Any direct use of a ligature (as accessed by \type {\char} or
through active characters) is wrong and will create faulty
hypenation. Normally, when TeX sees \quote {office}, it has the
six tokens \type {office} and it knows from the patterns that it
can hyphenate  between the \type {ff}. It will build an
internal list of four nodes, like this:

\starttyping
[char, o  , ffi ]
[lig , ffi, c   ,[f,f,i]]
[char, c  , e   ]
[char, e  , NULL]
\stoptyping

As you can see from the \type {ffi} line, it has remembered the
original characters. While hyphenating, it temporarily changes
back to that, then re-instates the ligature afterwards.

If you feed it the ligature directly, like so:

\starttyping
[char, o   , ffi ]
[char, ffi , c   ]
[char, c   , e   ]
[char, e   , NULL]
\stoptyping

it cannot do that. It tries to hyphenate as if the \type{ffi}
was a character, and the result is wrong hyphenation.


\stopcomponent