summaryrefslogtreecommitdiff
diff options
context:
space:
mode:
authorKarl Berry <karl@freefriends.org>2011-10-20 22:33:29 +0000
committerKarl Berry <karl@freefriends.org>2011-10-20 22:33:29 +0000
commit8e6d9e31498d059bac74c8942b5856090dab6ec4 (patch)
tree9d1a6f041fc8cb9302c22e0b777eb33d49bbb8b5
parent5afa6e8568e4a59832204d0e5ccfa9a551c73087 (diff)
translation-array-fr (20oct11)
git-svn-id: svn://tug.org/texlive/trunk@24344 c570f23f-e606-0410-a88d-b1316a301751
-rw-r--r--Master/texmf-dist/doc/latex/translation-array-fr/Copyright25
-rw-r--r--Master/texmf-dist/doc/latex/translation-array-fr/README23
-rw-r--r--Master/texmf-dist/doc/latex/translation-array-fr/f-array.dtx2850
-rw-r--r--Master/texmf-dist/doc/latex/translation-array-fr/f-array.pdfbin0 -> 642738 bytes
-rw-r--r--Master/texmf-dist/doc/latex/translation-array-fr/ltxdoc.cfg10
-rwxr-xr-xMaster/tlpkg/bin/tlpkg-ctan-check1
-rwxr-xr-xMaster/tlpkg/libexec/ctan2tds12
-rw-r--r--Master/tlpkg/tlpsrc/collection-documentation-french.tlpsrc1
-rw-r--r--Master/tlpkg/tlpsrc/translation-array-fr.tlpsrc0
9 files changed, 2917 insertions, 5 deletions
diff --git a/Master/texmf-dist/doc/latex/translation-array-fr/Copyright b/Master/texmf-dist/doc/latex/translation-array-fr/Copyright
new file mode 100644
index 00000000000..85385c24452
--- /dev/null
+++ b/Master/texmf-dist/doc/latex/translation-array-fr/Copyright
@@ -0,0 +1,25 @@
+All those translations are under the LaTeX Project Public
+License. Thus they are freely distributable, as long as the code of
+the packages that is in the document (which does not belong to us) is
+itself freely distributable. If you're allowed to use a package in a
+given way, then you're allowed to use the french translation of its
+documentation in exactly the sane way. The only restrictions (imposed
+by the LPPL) are on the translation itself.
+
+In short words: do not modify a document if you find a mistake, just
+contact the right person (the author if the package, the person who
+translated it, the translation team, etc).
+
+---
+
+Toutes ces traductions sont régies par la License Publique du Projet
+LaTeX (LPPL). Elles sont donc librement distribuables, du moment que
+le code des packages inclus dans les documentations l'est (il n'est
+pas de notre ressort). Si vous êtes autorisé à utiliser un package
+d'une certaines façon, alors vous êtes autorisé à faire la même chose
+avec la version française de sa documentation. Les seules restrictions
+(imposées par la LPPL) sont sur la traduction elle-même.
+
+En quelques mots: ne modifiez pas un document si vous trouvez une
+faute, indiquez-la plutôt a la bonne personne (l'auteur du package, le
+traducteur, l'équipe de traduction, etc).
diff --git a/Master/texmf-dist/doc/latex/translation-array-fr/README b/Master/texmf-dist/doc/latex/translation-array-fr/README
new file mode 100644
index 00000000000..cfc4fb63b9d
--- /dev/null
+++ b/Master/texmf-dist/doc/latex/translation-array-fr/README
@@ -0,0 +1,23 @@
+***********************************************************************
+* *
+* About the french translations of LaTeX documentations *
+* *
+***********************************************************************
+
+You'll find here french translations of as many LaTeX documentations
+as possible. As all those documents are in french and for french
+users, all the informations files are in french. We guess it will not
+hurt anybody. If it does, just tell us about it and we will write a
+few informative files in english.
+
+If you think the idea is interesting and you want to reproduce the
+same project in another language, do not hesitate to contact us, we
+can help, at least by giving you feedback about the way we work and
+how we think it can be improved.
+
+For copyright notice, see the file "copyright".
+
+Happy TeXing,
+
+ the translation team.
+
diff --git a/Master/texmf-dist/doc/latex/translation-array-fr/f-array.dtx b/Master/texmf-dist/doc/latex/translation-array-fr/f-array.dtx
new file mode 100644
index 00000000000..861e8bee9d3
--- /dev/null
+++ b/Master/texmf-dist/doc/latex/translation-array-fr/f-array.dtx
@@ -0,0 +1,2850 @@
+% \iffalse meta-comment
+%
+% Copyright 1993 1994 1995 1996 1997 1998 1999
+% The LaTeX3 Project and any individual authors listed elsewhere
+% in this file.
+%
+% This file is part of the Standard LaTeX `Tools Bundle'.
+% -------------------------------------------------------
+%
+% It may be distributed and/or modified under the
+% conditions of the LaTeX Project Public License, either version 1.2
+% of this license or (at your option) any later version.
+% The latest version of this license is in
+% http://www.latex-project.org/lppl.txt
+% and version 1.2 or later is part of all distributions of LaTeX
+% version 1999/12/01 or later.
+%
+% The list of all files belonging to the LaTeX `Tools Bundle' is
+% given in the file `manifest.txt'.
+%
+% \fi
+%\iffalse % this is a METACOMMENT !
+%
+%<*dtx>
+ \ProvidesFile{f-array.dtx}
+%</dtx>
+%% Package `array' to use with LaTeX 2e
+%% Copyright (C) 1989-1998 Frank Mittelbach, all rights reserved.
+%<+package>\NeedsTeXFormat{LaTeX2e}[1995/06/01]
+%<+package>\ProvidesPackage{array}
+% \fi
+% \ProvidesFile{f-array.dtx}
+ [1998/05/13 v2.3m Tabular extension package (FMi)]
+%
+%
+% \CheckSum{1108}
+%% \CharacterTable
+%% {Upper-case \A\B\C\D\E\F\G\H\I\J\K\L\M\N\O\P\Q\R\S\T\U\V\W\X\Y\Z
+%% Lower-case \a\b\c\d\e\f\g\h\i\j\k\l\m\n\o\p\q\r\s\t\u\v\w\x\y\z
+%% Digits \0\1\2\3\4\5\6\7\8\9
+%% Exclamation \! Double quote \" Hash (number) \#
+%% Dollar \$ Percent \% Ampersand \&
+%% Acute accent \' Left paren \( Right paren \)
+%% Asterisk \* Plus \+ Comma \,
+%% Minus \- Point \. Solidus \/
+%% Colon \: Semicolon \; Less than \<
+%% Equals \= Greater than \> Question mark \?
+%% Commercial at \@ Left bracket \[ Backslash \\
+%% Right bracket \] Circumflex \^ Underscore \_
+%% Grave accent \` Left brace \{ Vertical bar \|
+%% Right brace \} Tilde \~}
+%%
+%
+%
+% \changes{v2.3c}{1995/11/02}{(DPC) changement mineur dans la documentation}
+%
+% \changes{v2.3a}{1994/10/16}{ajout du code pour \cs{firsthline} et
+% assimil\'es}
+%
+% \changes{v2.2c}{1994/03/14}{suppression du test du bug de \cs{@tfor}}
+%
+% \changes{v1.0b}{1987/06/04}{`@classi (plus rapide),
+% `@classvi (nouveau) A dans le pr\'eambule
+% signifie \&\& dans `halign.}
+%
+% \changes{v1.1a}{1987/07/05}{Nouveau concept:
+% preamblechar: c,l,r,C,L,R,A,p,t,{\tt !|},@,!!}
+% \changes{v1.1b}{1987/09/21}{p reprend son r\^ole original comme dans
+% \LaTeX{} et z pour une `parbox centr\'ee.}
+%
+% \changes{v1.2a}{1987/09/27}{Impl\'ementatrion compl\`etement nouvelle.}
+% \changes{v1.2b}{1987/10/06}{{\tt !|} ne g\'en\`ere plus d'espace au
+% d\'ebut ou \`a la fin du pr\'eambule. Sinon
+% `hline est trop longue.}
+%\selectlanguage{english}
+% \changes{v1.2b}{1987/10/06}{Enlarged `@arstrutbox by 1pt (Test-Impl)
+% with dimen `@strutheight.}
+% \changes{v1.2c}{1987/10/22}{Nouvelle dimension `extrarowheight
+% (d\'efaut: 0pt).}
+% \changes{v1.2c}{1987/10/22}{Augmente `@arstrutbox de `extrarowheight.
+% Thus you may avoid large characters to
+% overprint a `hline.}
+% \changes{v1.2c}{1987/10/22}{Introduced `m@th in `@array to allow
+% non-zero values of `mathsurround.}
+% \changes{v1.2d}{1987/11/02}{Completed the documentation.}
+% \changes{v1.2e}{1987/11/03}{Bug fixed: A at start of preamble resulted
+% in an error since `@mkpream generated
+% `@arstrut \& ... as a preamble.}
+% \changes{v1.2f}{1987/11/09}{`@testpach documented.}
+%
+% \changes{v1.3a}{1987/11/11}{Again a new implementation, with a new
+% concept (cf. the documentation).}
+% \changes{v1.3b}{1988/03/17}{`@decl expands now into `@empty, i.e., it
+% disappears when the preamble is generated,
+% except when the user specifies A\{\} or
+% B\{\}.}
+%
+% \changes{v1.4a}{1988/03/18}{Test implementation of use of token
+% registers in order to do without `protect.}
+% \changes{v1.4b}{1988/03/19}{Changed erroneous class numbers:
+% 5 -!> 6
+% 6 -!> 7
+% 7 -!> 5
+% Corresponding changes in the macros.}
+% \changes{v1.4c}{1988/03/19}{Everything except p,z now works with token
+% registers.}
+%
+% \changes{v1.9a}{1988/03/20}{Last (so I hope) major change: 1) Options
+% B,A now called !>,<. These options now point
+% to the column they modify.}
+% \changes{v1.9a}{1988/03/20}{2) `protect is no longer necessary. But
+% still the macro `@expast needs top be
+% modified. `multicolumn still does not work.}
+% \changes{v1.9b}{1988/04/29}{inserted missing `fi in `@testpach.
+% Corrected \LaTeX bug in `@tfor.}
+% \changes{v1.9c}{1988/05/07}{Re-introduced `@endpbox.
+% `multicolumn now works!! Version number still
+% 1.9 since the documentation is still not
+% finished.}
+% \changes{v1.9c}{1988/05/07}{1) `def `the@toks \{`the ...\} remaining
+% only in `@mkpream. 2) Removed `@classiii and
+% replaced by `save@decl.}
+% \changes{v1.9c}{1988/05/07}{3) `insert@column contains only `@tempcnta
+% and `count@ counters. 4) `@@startpbox and
+% `@@endpbox now totally obsolete.}
+% \changes{v1.9d}{1988/05/10}{Replaced `number by `the where the `toks
+% registers' contents are used.}
+% \changes{v1.9e}{1988/05/11}{Re-introduced `@xargarraycr and
+% `@yargarraycr, since `endtemplate seems to
+% be `outer.}
+% \changes{v1.9f}{1988/05/20}{Small changes finally carried out:
+% 1) `par!=`@empty.
+% 2) \{..ifnum0!=!`\}... $\to$ `bgroup and
+% analoguously `egroup.}
+% \changes{v1.9g}{1988/02/24}{Inserted again \{..ifnum0!=!`\}..,
+% c.f. Appendix D of the \protect\TeX{}book.}
+% \changes{v1.9h}{1988/06/28}{No longer necessary to read in the file
+% twice.}
+% \changes{v1.9i}{1988/06/28}{Corrected typo in german version.}
+% \changes{v1.9j}{1988/11/23}{In a `r' column an extra `kern`z@ is
+% needed.}
+% \changes{v1.9j}{1988/11/23}{Otherwise the `hfil on the left side
+% will be removed by the `unskip in
+% `insert@column if the entry is empty.}
+% \changes{v1.9k}{1988/06/28}{Corrected typo in german version.}
+% \changes{v1.9k}{1989/01/16}{`begin{Macro} changed to `begin{macro} in
+% documentation.}
+%
+% \changes{v2.0a}{1989/05/12}{{\tt\textbackslash @thetoks} changed to
+% {\tt\textbackslash the@toks}.}
+% \changes{v2.0a}{1989/05/12}{source changed to reflect new doc.sty
+% conventions.}
+% \changes{v2.0a}{1989/05/12}{t option renamed to p to be compatible to
+% the original.}
+% \changes{v2.0a}{1989/05/12}{File renamed from arraye.sty to
+% array.sty.}
+% \changes{v2.0b}{1989/05/17}{Three forgotten end macro added.}
+% \changes{v2.0b}{1989/05/17}{All lines shortened to 72 or less.}
+% \changes{v2.2a}{1994/02/03}{Upgrade to \LaTeXe}
+%\selectlanguage{french}
+%
+% \DoNotIndex{\@depth,\@ehc,\@fortmp,\@height,\@ifnextchar,\@ifstar}
+% \DoNotIndex{\@ifundefined,\@ne,\@nil,\@tempa,\@tempb}
+% \DoNotIndex{\@tempcnta,\@tempd,\@tempdima,\@whilenum,\@width,\\}
+% \DoNotIndex{\@tforloop}
+% \DoNotIndex{\advance}
+% \DoNotIndex{\baselineskip,\begingroup,\bgroup}
+% \DoNotIndex{\cr,\crcr,\csname}
+% \DoNotIndex{\def,\do,\docdate,\dp}
+% \DoNotIndex{\edef,\egroup,\else,\endcsname,\endinput,\expandafter}
+% \DoNotIndex{\fi,\filedate,\fileversion}
+% \DoNotIndex{\gdef}
+% \DoNotIndex{\hbox,\hfil,\hsize,\hskip,\ht}
+% \DoNotIndex{\if,\ifcase,\ifdim,\ifnum,\ifx,\ignorespaces}
+% \DoNotIndex{\kern}
+% \DoNotIndex{\leavevmode,\let,\lineskip}
+% \DoNotIndex{\m@ne,\multispan}
+% \DoNotIndex{\newcount,\newdimen,\noalign}
+% \DoNotIndex{\or}
+% \DoNotIndex{\relax}
+% \DoNotIndex{\setbox,\space,\strutbox}
+% \DoNotIndex{\tabskip,\thr@@,\the,\toks,\toks@,\tw@,\typeout}
+% \DoNotIndex{\unhcopy,\unskip}
+% \DoNotIndex{\vbox,\vcenter,\vline,\vrule,\vtop,\vskip}
+% \DoNotIndex{\xdef}
+% \DoNotIndex{\z@}
+%
+% \GetFileInfo{f-array.dtx}
+%
+% \title{Une nouvelle \'ecriture des environnements
+% \textsf{tabular} et \textsf{array} de \LaTeX\thanks{Ce fichier
+% a le num\'ero de version \fileversion, r\'evis\'e
+% le \filedate.}}
+%
+% \author{Frank Mittelbach
+% \and
+% David Carlisle\thanks{David a accept\'e gr\^acieusement l'inclusion
+% de code auparavant dans \texttt{newarray.sty} pour l'\'ecriture
+% de \texttt{\textbackslash{}newcolumntype}} \and
+% \makebox[0.9\linewidth]{Traduction fran\c{c}aise
+% par Jean-Pierre Drucbert\thanks{Derni\`ere mise \`a
+% jour le 20/01/2000}}}
+%
+% \date{Imprim\'e le \today}
+%
+% \maketitle
+%
+% \MakeShortVerb{\=}
+%
+% \begin{abstract}
+% Ce package est une nouvelle impl\'ementation des environnements
+% \texttt{tabular} et \texttt{array} de \LaTeX. Les principaux avantages
+% de cette impl\'ementation sont de nouvelles options pour mettre en
+% forme les colonnes et que les commandes <<~fragiles~>> de \LaTeX\
+% n'ont plus besoin d'\^etre prot\'eg\'ees par =\protect= \`a
+% l'int\'erieur de ces environnements.
+%
+% La majeure partie du code de ce package date de 1988---il en est
+% de m\^eme de sa documentation.
+% \end{abstract}
+%
+%
+%
+% \section{Introduction}
+%
+% Cette nouvelle impl\'ementation des environnements \texttt{tabular} et
+% \texttt{array} fait partie d'un projet plus large, dans lequel nous
+% essayons d'am\'eliorer le code de \LaTeX\ et de le rendre encore plus
+% facile \`a manipuler.
+%
+% Le lecteur devrait \^etre familier avec les environnements
+% mentionn\'es ci-dessus. De plus amples informations sont disponibles
+% dans \cite{bk:lamport} et \cite{bk:GMS94}. Les options
+% suppl\'ementaires qui peuvent \^etre utilis\'ees dans le pr\'eambule
+% ainsi que celles qui ont un sens l\'eg\`erement diff\'erent sont
+% d\'ecrites dans la table~\ref{tab+opt}.
+%
+% \DescribeMacro\extrarowheight
+% Il y a un nouveau param\`etre, la longueur =\extrarowheight=, qui, si
+% elle est positive, est ajout\'ee \`a la hauteur normale de chaque
+% ligne de la table, alors que la profondeur reste la m\^eme. Ceci est
+% important pour les tables comportant des lignes horizontales car
+% normalement ces lignes touchent les lettres majuscules.
+% Par exemple, nous avons utilis\'e =\setlength{\extrarowheight}{1pt}=
+% dans la table~\ref{tab+opt}.
+%
+% \begin{table}[t]
+% \begin{center}
+% \setlength{\extrarowheight}{1pt}
+% \begin{tabular}{|>{\tt}c|m{9cm}|}
+% \hline
+% \multicolumn{2}{|c|}{Options inchang\'ees}\\
+% \hline
+% l & Colonne cadr\'ee \`a gauche. \\
+% c & Colonne centr\'ee. \\
+% r & Colonne cadr\'ee \`a droite. \\
+% p\{width\} & \'Equivalente \`a =\parbox[t]{width}=. \\
+% @\{decl.\} & Supprime l'espace entre les colonnes et le
+% remplace par \texttt{decl.}. \\
+% \hline
+% \multicolumn{2}{|c|}{Options nouvelles}\\
+% \hline
+% m\{width\} & D\'efinit une colonne de largeur \texttt{width}.
+% Chaque entr\'ee sera centr\'ee en proportion du
+% reste de la ligne. Ceci est assez semblable \`a
+% =\parbox{width}=. \\
+% \hline
+% b\{width\} & \'Equivalente \`a =\parbox[b]{width}=. \\
+% \hline
+% >\{decl.\} & Peut \^etre utilis\'ee avant une option \texttt{l}, \texttt{r},
+% \texttt{c}, \texttt{p}, \texttt{m} ou
+% \texttt{b}. Ceci ins\`ere \texttt{decl.}
+% directement en t\^ete de l'entr\'ee dans la colonne. \\
+% \hline
+% <\{decl.\} & Peut \^etre utilis\'ee apr\`es une option \texttt{l}, \texttt{r},
+% \texttt{c}, \texttt{p}, \texttt{m} ou
+% \texttt{b}. Ceci ins\`ere \texttt{decl.}
+% juste apr\`es l'entr\'ee dans la colonne. \\
+% \hline
+% | & Ins\`ere une ligne verticale. La distance entre
+% deux colonnes sera augment\'ee de l'\'epaisseur
+% de la ligne, contrairement \`a la d\'efinition
+% originale de \LaTeX.\\
+% \hline
+% !\{decl.\} & Peut \^etre utilis\'ee n'importe o\`u et correspond \`a l'option
+% \texttt{|}. La diff\'erence est que
+% \texttt{decl.} sera ins\'er\'ee \`a la place
+% d'une ligne verticale, donc
+% cette option ne supprime pas l'espace
+% normalement ins\'er\'e entre
+% les colonnes, contrairement \`a =@{...}=.\\
+% \hline
+% \end{tabular}
+% \end{center}
+% \caption{Les options du pr\'eambule.} \label{tab+opt}
+% \end{table}
+%
+%
+% Nous allons pr\'esenter quelques exemples des nouvelles options du
+% pr\'eambule.
+% \begin{itemize}
+% \item
+% Si vous d\'esirez utiliser une fonte sp\'eciale (par exemple
+% =\bfseries=) dans une colonne cadr\'ee \`a gauche, ceci peut se
+% faire avec =>{\bfseries}l=. Vous n'\^etes plus oblig\'es de
+% commencer chaque entr\'ee par =\bfseries=.
+% \item
+% Dans les colonnes qui ont \'et\'e engendr\'ees par \texttt{p},
+% \texttt{m} ou \texttt{b}, la valeur par d\'efaut de =\parindent=
+% est 0pt. Ceci peut \^etre modifi\'e par:\\
+% =>{\setlength{\parindent}{1cm}}p=.
+% \item
+% les options \texttt{<} et \texttt{>} avaient \'et\'e
+% d\'evelopp\'ees \`a l'origine pour l'application suivante:
+% =>{$}c<{$}= engendre une colonne en mode math\'ematique
+% dans un environnement \textsf{tabular}. Si vous utilisez ce type
+% de pr\'eambule dans un environnement \textsf{array}, vous
+% obtenez une colonne en mode~LR car les dollars suppl\'ementaires
+% annulent ceux d\'ej\`a pr\'esents.
+% \item
+% Vous pouvez aussi penser \`a des applications plus complexes. Un
+% probl\`eme classique peut facilement \^etre r\'esolu avec la
+% s\'equence
+% =>{\centerdots}c=\linebreak[0]=<{\endcenterdots}=.
+% Pour centrer des nombres sur leurs points d\'ecimaux, il
+% suffit\,\footnote{C'est Frank Mittelbach qui le
+% dit!}~(juste~?) de d\'efinir les macros suivantes:
+% \begin{verbatim}
+%{\catcode`\.\active\gdef.{\egroup\setbox2\hbox\bgroup}}
+%\def\centerdots{\catcode`\.\active\setbox0\hbox\bgroup}
+%\def\endcenterdots{\egroup\ifvoid2 \setbox2\hbox{0}\fi
+% \ifdim \wd0>\wd2 \setbox2\hbox to\wd0{\unhbox2\hfill}\else
+% \setbox0\hbox to\wd2{\hfill\unhbox0}\fi
+% \catcode`\.12 \box0.\box2}
+%\end{verbatim}
+% Attention: ces macros ne marchent pas si une cellule comporte
+% plusieurs points d\'ecimaux ou si le tableau est utilis\'e en argument
+% d'une autre commande. Une approche bien meilleure est d'utiliser le
+% package \texttt{dcolumn.sty} \'ecrit par David Carlisle.
+% \item
+% En utilisant =c!{\hspace{1cm}}c= vous \'elargissez d'un
+% centim\`etre l'espace entre deux colonnes, tandis que
+% =c@{\hspace{1cm}}c= vous donne exactement un centim\`etre
+% d'espace entre deux colonnes.
+% \end{itemize}
+%
+% \subsection{D\'efinir de nouveaux types de colonnes}
+%
+% \DeleteShortVerb{\=}
+% \MakeShortVerb{\"}
+% \DescribeMacro{\newcolumntype}
+% Alors qu'il est pratique de taper
+% \begin{quote}
+% ">{"\meta{quelques d\'eclarations}"}{c}<{"\meta{quelques
+% autres d\'eclarations}"}"
+% \end{quote}
+% si vous avez une colonne particuli\`ere dans une seule table, cela
+% devient peu pratique si vous utilisez souvent des colonnes de cette
+% forme. La nouvelle version vous permet de d\'efinir un nouveau
+% sp\'ecificateur de colonne, par exemple \texttt{x}, qui remplacera
+% plusieurs commandes de sp\'ecificateurs de colonnes\footnote{Cette
+% commande \'etait appel\'ee \texttt{\textbackslash{}newcolumn} dans
+% \texttt{newarray.sty}. Actuellement, la commande
+% \texttt{\textbackslash{}newcolumn} est toujours valide (mais affiche
+% un message d'avertissement). Dans les versions ult\'erieures, elle
+% dispara\^{\i}tra.}.
+% Nous pouvons donc d\'efinir:
+% \begin{quote}
+% "\newcolumntype{x}{>{"\meta{ d\'eclarations}"}{c}<{"\meta{autres
+% d\'eclarations}"}}"\hspace*{-3cm} ^^A no overfull from this line
+% \end{quote}
+% On peut alors utiliser le sp\'ecificateur de colonne \texttt{x} dans
+% les arguments de pr\'eambule de tous les environnements \texttt{array}
+% ou \texttt{tabular} dans lesquels vous d\'esirez ce format.
+%
+% Il est habituel d'avoir besoin de colonnes en mode math\'ematique
+% et en mode~LR
+% dans le m\^eme tableau. Si nous d\'efinissons:
+% \begin{quote}
+% "\newcolumntype{C}{>{$}c<{$}}" \\
+% "\newcolumntype{L}{>{$}l<{$}}" \\
+% "\newcolumntype{R}{>{$}r<{$}}"
+% \end{quote}
+% alors nous pouvons utiliser \texttt{C} pour avoir un mode~LR centr\'e
+% dans un \texttt{array}, ou un mode math\'ematique centr\'e dans
+% un \texttt{tabular}.
+%
+% L'exemple donn\'e plus haut pour les <<~points d\'ecimaux centr\'es~>>
+% pourrait \^etre affect\'e \`a un sp\'ecificateur \texttt{d} par la
+% commande suivante:
+% \begin{quote}
+% "\newcolumntype{d}{>{\centerdots}c<{\endcenterdots}}"
+% \end{quote}
+%
+% Cette solution centre toujours le point dans la colonne. Ceci n'est
+% pas tr\`es joli si la colonne est form\'ee de grands nombres, mais
+% avec peu de d\'ecimales. Une autre solution d'une colonne \texttt{d}
+% est:
+% \begin{quote}
+% "\newcolumntype{d}[1]{>{\rightdots{#1}}r<{\endrightdots}}"
+% \end{quote}
+% o\`u les macros ad\'equates seraient\footnote{Le package
+% \texttt{dcolumn.sty} contient des commandes plus robustes bas\'ees sur
+% ces id\'ees.}:
+% \begin{verbatim}
+% \def\coldot{.}% Ou si vous pr\'ef\'erez, \def\coldot{\cdot}
+% {\catcode`\.=\active
+% \gdef.{$\egroup\setbox2=\hbox to \dimen0 \bgroup$\coldot}}
+% \def\rightdots#1{%
+% \setbox0=\hbox{$1$}\dimen0=#1\wd0
+% \setbox0=\hbox{$\coldot$}\advance\dimen0 \wd0
+% \setbox2=\hbox to \dimen0 {}%
+% \setbox0=\hbox\bgroup\mathcode`\.="8000 $}
+% \def\endrightdots{$\hfil\egroup\box0\box2}
+%\end{verbatim}
+% Notez que "\newcolumntype" accepte le m\^eme argument optionnel que
+% "\newcommand" pour d\'eclarer le nombre d'arguments pour le
+% sp\'ecificateur de colonne. Maintenant vous pouvez
+% sp\'ecifier "d{2}" dans votre pr\'eambule pour une colonne de
+% nombres ayant au plus deux d\'ecimales.
+%
+% Une utilisation assez diff\'erente de "\newcolumntype"
+% tire parti du fait que le texte de remplacement dans cette commande
+% peut si nécessaire faire r\'ef\'erence \`a plus d'une colonne.
+% Supposons maintenant que notre document contienne de nombreux
+% environnements \texttt{tabular} utilisant le m\^eme pr\'eambule,
+% mais que vous vouliez en essayer d'autres. La d\'efinition originale
+% dans \LaTeX\ (par Leslie Lamport) vous permettait de faire (bien que
+% ce soit sans doute un usage incorrect du syst\`eme):
+% \begin{quote}
+% "\newcommand{\X}{clr}"\\
+% "\begin{tabular}{\X}" \ldots
+% \end{quote}
+% \texttt{array.sty} fait bien attention de ne pas d\'evelopper le
+% pr\'eambule, et donc la solution ci-dessus ne marche pas avec le
+% nouveau sch\'ema. Cette fonctionnalit\'e est obtenue avec:
+% \begin{quote}
+% "\newcolumntype{X}{clr}"\\
+% "\begin{tabular}{X}" \ldots
+% \end{quote}
+%
+% Le texte de remplacement dans une commande "\newcolumntype" peut faire
+% r\'ef\'erence \`a chacune des primitives de \texttt{array.sty} (voir
+% la table~\ref{tab+opt} en page~\pageref{tab+opt}), ainsi qu'aux
+% nouvelles lettres d\'efinies dans d'autres commandes "\newcolumntype".
+%
+% \DescribeMacro{\showcols}Une liste de tous les sp\'ecificateurs
+% d\'efinis par "\newcolumntype" est affich\'ee \`a l'\'ecran et dans le
+% fichier de logs lorsque la commande "\showcols" est utilis\'ee.
+%
+%
+% \subsection{Variantes sp\'eciales de \texttt{\textbackslash hline}}
+%
+% La famille des environnements \texttt{tabular} permet un
+% positionnement vertical bas\'e sur la ligne de base du texte dans
+% lequel l'environnement apparait. Par d\'efaut, l'environnement est
+% centr\'e verticalement, mais cela peut \^etre modifi\'e pour l'aligner
+% sur la premi\`ere ou la derni\`ere ligne l'environnement en
+% sp\'ecifiant la valeur \texttt{t} ou \texttt{b} \`a l'argument
+% optionnel de positionnement. Cependant, cela ne marche pas si le
+% premier ou le dernier \'el\'ement de l'environnement est une commande
+% "\hline" ; dans ce cas, l'environnement est align\'e sur la barre
+% horizontale.
+%
+% \pagebreak[3]
+%
+% Voici un example:
+% \begin{center}
+% \begin{minipage}[t]{.4\linewidth}
+% Tables
+% \begin{tabular}[t]{l}
+% sans \\ commande \\ hline
+% \end{tabular} et \\ tables
+% \begin{tabular}[t]{|l|}
+% \hline
+% avec \\ commandes \\ hline \\
+% \hline
+% \end{tabular}
+% \end{minipage}
+% \begin{minipage}[t]{.5\linewidth}
+% \begin{verbatim}
+% Tables
+% \begin{tabular}[t]{l}
+% sans \\ commande \\ hline
+% \end{tabular} et \\ tables
+% \begin{tabular}[t]{|l|}
+% \hline
+% avec \\ commandes \\ hline \\
+% \hline
+% \end{tabular}
+% \end{verbatim}
+% \end{minipage}
+% \end{center}
+%
+% \DescribeMacro\firsthline
+% \DescribeMacro\lasthline
+% L'utilisation de "\firsthline" et "\lasthline" va r\'esoudre ce
+% probl\`eme, et les tables seront correctement align\'ees tant que
+% leur premi\`ere ou derni\`ere ligne ne contient pas d'objets
+% excessivement larges.
+% \begin{center}
+% \begin{minipage}[t]{.4\linewidth}
+% Tables
+% \begin{tabular}[t]{l}
+% sans \\ commande \\ ...line
+% \end{tabular} et \\ tables
+% \begin{tabular}[t]{|l|}
+% \firsthline
+% avec \\ commandes \\ ...line \\
+% \lasthline
+% \end{tabular}
+% \end{minipage}
+% \begin{minipage}[t]{.5\linewidth}
+% \begin{verbatim}
+% Tables
+% \begin{tabular}[t]{l}
+% sans \\ commande \\ ...line
+% \end{tabular} et \\ tables
+% \begin{tabular}[t]{|l|}
+% \firsthline
+% avec \\ commandes \\ ...line \\
+% \lasrhline
+% \end{tabular}
+% \end{verbatim}
+% \end{minipage}
+% \end{center}
+%
+% \DescribeMacro\extratabsurround
+% L'\'ecriture de ces deux commandes contient une longueur
+% suppl\'ementaire, appel\'ee "\extratabsurround", pour ajouter une
+% espace au d\'ebut ou \`a la fin de ces environnements.
+% C'est tr\`es utile quand les tables sont embo\^{\i}t\'ees.
+%
+% \section{Derniers commentaires}
+%
+% \subsection{Manipulation des barres de s\'eparation}
+%
+% Il y a deux approches pour la manipulation des barres horizontales et
+% verticales \`a l'int\'erieur des tables:
+% \begin{enumerate}
+% \item ces barres peuvent \^etre plac\'ees dans l'espace disponible
+% sans augmenter la taille de la table, ou
+% \item elles peuvent \^etre plac\'ees entre les colonnes et les
+% lignes, et par l\`a m\^eme agrandissent la table.
+% \end{enumerate}
+% \texttt{array.sty} impl\'emente la seconde approche, alors que
+% la premi\`ere est utilis\'ee dans le noyau de \LaTeX.
+% Les deux ont chacunes leurs avantages, mais il faut faire attention aux
+% implications suivantes:
+% \begin{itemize}
+% \item
+% Avec le \LaTeX{} standard, ajouter des barres dans une table ne
+% modifiera ni la largeur ni la hauteur de la table (\`a moins
+% d'utiliser des doubles barres) ; par exemple changer le
+% pr\'eambule de \verb=lll= en \verb=l|l|l= ne modifie pas le
+% document autrement qu'en ajoutant les barres. Au contraire, avec
+% \texttt{array.sty}, une table qui rentre juste dans la largeur
+% \verb=\textwidth= peut produire un d\'ebordement horizontal.
+% \item
+% Avec le \LaTeX{} standard, modifier l'\'epaisseur des barres peut
+% d\'et\'eriorer l'apparence des tables si on ne modifie pas en m\^eme
+% temps les param\`etres \verb=\tabcolsep=, etc, l'espace entre la
+% barre et la colonne pouvant alors devenir trop petit (ou trop
+% grand). En fait, la surimpression de texte est m\^eme possible.
+% Au contraire, avec \texttt{array.sty}, la modification de cette
+% largeur marche souvent sans probl\`eme parce que les espaces de
+% s\'eparation (comme \verb=\tabcolsep=, etc) ne d\'ependent pas de
+% l'\'epaisseur des barres.
+% \item
+% Avec le \LaTeX{} standard, les tableaux encadr\'es pr\'esentent des
+% coins bizarres parce que les barres horizontales se terminent au
+% milieu des barres verticales. Quand le param\`etre
+% \verb=\arrayrulewidth= est grand, l'apparence de la table est tr\`es
+% d\'esagr\'eable. Dans ce cas, une simple table comme
+%\begin{verbatim}
+%\setlength{\arrayrulewidth}{5pt}
+%\begin{tabular}{|l|}
+% \hline A \\ \hline
+%\end{tabular}
+%\end{verbatim}
+% donnera quelque chose comme
+% \begin{center}
+%\setlength{\arrayrulewidth}{5pt}
+%\begin{tabular}{@{}l@{}}
+% \hline \null\hskip-.5\arrayrulewidth\vline
+% \hskip\tabcolsep
+% A\hskip\tabcolsep
+% \vline\hskip-.5\arrayrulewidth\null \\ \hline
+%\end{tabular}
+% \quad
+% au lieu de
+% \quad
+%\begin{tabular}{|l|}
+% \hline A \\ \hline
+%\end{tabular}
+% \end{center}
+% \end{itemize}
+%
+% \subsection{Comparaisons avec les anciennes versions de \texttt{array.sty}}
+%
+% Il y a plusieurs diff\'erences dans la fa\c{c}on dont la version 2.1
+% traite les donn\'ees incorrectes, m\^eme si le fichier source ne
+% semble pas utiliser les fonctionnalit\'es suppl\'ementaires de cette
+% nouvelle version.
+% \begin{itemize}
+% \item Un pr\'eambule de la forme "{wx*{0}{abc}yz}" \'etait trait\'ee
+% dans les versions ant\'erieures \`a 2.1 comme "{wx}". La version 2.1
+% le traite comme "{wxyz}"
+% \item Un param\`etre de positionnement erron\'e comme \texttt{[Q]}
+% \'etait consid\'er\'e comme \texttt{[c]} par \texttt{array.sty}, mais
+% il est maintenant trait\'e comme \texttt{[t]}.
+% \item Un pr\'eambule tel que "{cc*{2}}" avec une erreur dans
+% l'argument de l'\'etoile~$*$ engendrera des erreurs diff\'erentes dans
+% la nouvelle version. Dans les deux cas, les messages d'erreur ne sont
+% pas d'une aide tr\`es utile pour l'utilisateur ordinaire.
+% \item Des r\'ep\'etitions de constructions avec \texttt{<} ou \texttt{>}
+% engendraient une erreur dans les versions ant\'erieures, mais sont
+% maintenant autoris\'ees. Par exemple,
+% ">{"\meta{decl1}"}>{"\meta{decl2}"}" est \'equivalent \`a
+% ">{"\meta{decl2}\meta{decl1}"}".
+% \item La commande "\extracolsep" ne fonctionne pas avec les anciennes
+% versions de \texttt{array.sty} (lire les commentaires dans
+% \texttt{array.bug}). Avec la version 2.1, "\extracolsep" peut de
+% nouveau \^etre utilis\'ee dans des expressions \verb=@{...}= comme avec
+% le \LaTeX{} standard; elle peut aussi se trouver dans une commande
+% \verb=!{...}=, mais il convient de lire la note ci-dessous.
+% \end{itemize}
+%
+%
+% \subsection{Bogues et fonctionnalit\'es}
+%
+% \begin{itemize}
+% \item Les messages d'erreur engendr\'es en d\'ecodant les
+% sp\'ecifications des colonnes font r\'ef\'erence \`a l'argument du
+% pr\'eambule \textbf{apr\`es} qu'il a \'et\'e r\'e\'ecrit par la
+% commande "\newcolumntype", et non pas au pr\'eambule tel qu'il a
+% \'et\'e \'ecrit par l'utilisateur. Cela semble in\'evitable pour
+% tout syst\`eme utilisant un pr\'e-traitement, et est donc
+% consid\'er\'e comme une \textbf{fonctionnalit\'e}.
+%
+% \item Le traitement de la d\'eclaration de plusieurs \texttt{<} ou
+% \texttt{>} peut sembler \'etrange de prime abord. Les anciennes
+% versions consid\'eraient ">{"\meta{decl1}"}>{"\meta{decl2}"}"
+% comme \'equivalant \`a ">{"\meta{decl1}\meta{decl2}"}".
+% Cependant, cela emp\^eche l'utilisateur de remplacer les
+% param\`etres d'une commande "\newcolumntype" d\'efinie en utilisant
+% ces d\'eclarations. Par exemple, supposons que nous utilisons un
+% sp\'ecificateur de colonne \texttt{C} (d\'efini au-dessus) dans un
+% environnement \texttt{array}. Le \texttt{C} indique du texte
+% centr\'e dans la colonne, alors que ">{\bfseries}C", qui est
+% r\'e\'ecrit en ">{\bfseries}>{$}c<{$}" ne peut pas sp\'ecifier une
+% colonne de texte en gras, parce que le pr\'eambule ressemblera en
+% fait \`a "\hfil$\bfseries$#$ $\hfil" et donc le texte dans la
+% colonne n'est pas affect\'e par la commande "\bfseries"\,! La
+% version actuelle inverse l'ordre de ces d\'eclarations, et ainsi
+% l'exemple pr\'ec\'edent fournit un pr\'eambule de la forme "\hfil$"
+% "$\bfseries#$" "$\hfil", les dollars s'annulant mutuellement pour
+% donner l'effet escompt\'e.
+%
+% \item L'utilisation de "\extracolsep" est sujette aux deux
+% restrictions suivantes. Il doit y a voir au plus une commande
+% "\extracolsep" par sp\'ecificateur "@" ou "!", et cette commande
+% doit \^etre entr\'ee telle quelle, elle ne doit pas \^etre le
+% r\'esultat du d\'eveloppement d'une commande.
+%
+% Ainsi, "\newcommand{\ef}{\extracolsep{\fill}}" \ldots "@{\ef}" ne
+% marche pas avec ce package. Cependant, il est possible d'utiliser
+% \`a la place une construction comme
+% "\newcolumntype{e}{@{\extracolsep{\fill}}".
+%
+% \item Comme indiqu\'e dans le \LaTeX{} book, afin de d\'efinir la
+% commande "\multicolumn", chaque colonne \`a l'exception de la
+% premi\`ere est compos\'ee du texte compris dans la cellule et de ce
+% qui se trouve dans l'espace entre les colonnes situ\'e
+% \emph{apr\`es} cette colonne. Cela signifie que dans un tableau avec
+% comme pr\'eambule "|l|l|l|l|", une entr\'ee telle que
+% "\multicolumn{2}{|c|}" est incorrecte si elle n'intervient pas dans la
+% premi\`ere colonne.
+%
+% Dans le \LaTeX{} standard, cette erreur n'est pas tellement visible
+% parce que cette version ajoute des espaces n\'egatifs, si bien que
+% chaque "|" n'occupe pas de place horizontalement. Mais dans ce
+% package, les lignes verticales conservent leur \'epaisseur, on voit
+% alors deux lignes si deux barres ont \'et\'e sp\'ecifi\'ees.
+% \end{itemize}
+%
+%
+%
+% \changes{v2.2b}{1994/02/04}{Suppression de l'invite en mode interactif}
+%
+% \StopEventually{
+%
+%
+% \begin{thebibliography}{1}
+% \bibitem{bk:GMS94} \textsc{M.~Goossens}, \textsc{F.~Mittelbach}
+% et \textsc{A.~Samarin}.
+% \newblock The \LaTeX{} Companion.
+% \newblock
+% Addison-Wesley, Reading, Massachusetts, 1994.
+% \bibitem{bk:knuth} \textsc{D. E. Knuth}.
+% \newblock The \TeX{}book (Computers \& Typesetting Volume A).
+% \newblock
+% Addison-Wesley, Reading, Massachusetts, 1986.
+% \bibitem{bk:lamport} \textsc{L. Lamport}.
+% \newblock
+% \LaTeX\ --- A Document Preparation System.
+% \newblock
+% Addison-Wesley, Reading, Massachusetts, 1986.
+% \end{thebibliography}
+%
+% } ^^A end of \StopEventually
+%
+%
+%\selectlanguage{english}
+%
+%
+% \section{The documentation driver file}
+%
+% The first bit of code contains the documentation driver file for
+% \TeX{}, i.e., the file that will produce the documentation you are
+% currently reading. It will be extracted from this file by the
+% \texttt{docstrip} program.
+% \begin{macrocode}
+%<*driver>
+\NeedsTeXFormat{LaTeX2e}[1995/12/01]
+\documentclass{ltxdoc}
+
+\AtBeginDocument{\DeleteShortVerb{\|}} % undo the default is not used
+
+ \usepackage{array}
+%</driver>
+% \end{macrocode}
+% \iffalse
+%<*driver>
+ \usepackage[english,frenchb]{babel}
+ \usepackage[T1]{fontenc}
+ \usepackage{lmodern}
+%</driver>
+% \fi
+% \begin{macrocode}
+%<*driver>
+
+ % Allow large table at bottom
+ \renewcommand{\bottomfraction}{0.7}
+
+\EnableCrossrefs
+ %\DisableCrossrefs % Say \DisableCrossrefs if index is ready
+
+\RecordChanges % Gather update information
+
+\CodelineIndex % Index code by line number
+
+ %\OnlyDescription % comment out for implementation details
+ %\OldMakeindex % use if your MakeIndex is pre-v2.9
+\begin{document}
+ \DocInput{f-array.dtx}
+\end{document}
+%</driver>
+% \end{macrocode}
+%
+%
+% \section{The construction of the preamble}
+%
+% \DeleteShortVerb{\"}
+% \MakeShortVerb{\=}
+%
+% It is obvious that those environments will consist mainly of an
+% =\halign=, because \TeX\ typesets tables using this primitive.
+% That is why we will now take a look at the algorithm which determines
+% a preamble for a =\halign= starting with a given user preamble
+% using the options mentioned above.
+%
+%
+% The current version is defined at the top of the file looking
+% something like this
+% \begin{macrocode}
+%<*package>
+%\NeedsTeXFormat{LaTeX2e}[1994/05/13]
+%\ProvidesPackage{array}[\filedate\space version\fileversion]
+% \end{macrocode}
+%
+% The most interesting macros of this implementation are without doubt
+% those which are responsible for the construction of the preamble for
+% the =\halign=. The underlying algorithm was developed by
+% \textsc{Lamport} (resp.\ \textsc{Knuth}, see texhax V87\#??), and it
+% has been extended and improved.
+%
+% The user preamble will be read \textsf{token} by \textsf{token}. A
+% \textsf{token} is a single character like \texttt{c} or a block
+% enclosed in ={...}=. For example the preamble of
+% =\begin{tabular}=\linebreak[0]={lc||c@{\hspace{1cm}}}= consists of
+% the \textsf{token} \texttt{l}, \texttt{c}, \texttt{|}, \texttt{|},
+% \texttt{@} and =\hspace{1cm}=.
+%
+% The currently used \textsf{token} and the one, used before, are needed
+% to decide on how the construction of the preamble has to be
+% continued.
+% In the example mentioned above the \texttt{l} causes the preamble
+% to begin with =\hskip\tabcolsep=. Furthermore
+% =# \hfil= would be appended to define a flush left column.
+% The next \textsf{token} is a \texttt{c}. Because it was preceded by an
+% \texttt{l} it generates a new column. This is done with
+% =\hskip \tabcolsep & \hskip \tabcolsep=. The column which is to
+% be centered will be appended with =\hfil # \hfil=.
+% The \textsf{token} \texttt{|} would then add a space of
+% =\hskip \tabcolsep=
+% and a vertical line because the last
+% \textsf{tokens} was a \texttt{c}.
+% The following \textsf{token} \texttt{|} would only add a space
+% =\hskip \doublerulesep= because it was preceded by the
+% \textsf{token} \texttt{|}. We will not discuss our example further but
+% rather take a look at the general case of constructing preambles.
+%
+% The example shows that the desired preamble for the
+% =\halign= can be constructed as soon as the action of all
+% combinations
+% of the preamble \textsf{tokens} are specified. There are 18 such
+% \textsf{tokens}
+% so we have $19 \cdot 18 \string= 342$ combinations if we count the
+% beginning of
+% the preamble as a special \textsf{token}. Fortunately, there are many
+% combinations which generate the same spaces, so we can define
+% \textsf{token} classes. We will identify a
+% \textsf{token} within a class with a number, so we can insert the
+% formatting (for example of a column).
+% Table~\ref{tab+Klassen} lists all \textsf{token} classes and
+% their corresponding numbers.
+% \begin{table}[ht]
+% \begin{center}
+% \begin{tabular}[t]{>{\ttfamily}ccc}
+% \textsf{token} & =\@chclass= & =\@chnum= \\[2mm]
+% c & 0 & 0 \\
+% l & 0 & 1 \\
+% r & 0 & 2 \\
+% p-arg & 0 & 3 \\
+% t-arg & 0 & 4 \\
+% b-arg & 0 & 5 \\
+% | & 1 & 0 \\
+% !-arg & 1 & 1 \\
+% <-arg & 2 & --- \\
+% >-arg & 3 & ---
+% \end{tabular}
+% \kern3mm \vrule \kern3mm%
+% \begin{tabular}[t]{>{\ttfamily}ccc}
+% \textsf{token} & =\@chclass= & =\@chnum= \\[2mm]
+% Start & 4 & --- \\
+% @-arg & 5 & --- \\
+% ! & 6 & --- \\
+% @ & 7 & --- \\
+% < & 8 & --- \\
+% > & 9 & --- \\
+% p & 10 & 3 \\
+% t & 10 & 4 \\
+% b & 10 & 5
+% \end{tabular}
+% \end{center}
+% \caption{Classes of preamble \textsf{tokens}}
+% \label{tab+Klassen}
+% \end{table}
+%
+%
+% \begin{macro}{\@chclass}
+% \begin{macro}{\@chnum}
+% \begin{macro}{\@lastchclass}
+% The class and the number of the current \textsf{token} are saved in
+% the
+% \textsf{count} registers =\@chclass=
+% and =\@chnum=, while the class of the previous
+% \textsf{token} is stored in the
+% \textsf{count} register =\@lastchclass=.
+% All of the mentioned registers are already allocated in
+% \texttt{latex.tex},
+% which is the reason why the following three lines of code are
+% commented out.
+% Later throughout the text I will not mention it again explicitely
+% whenever I use a =%= sign. These parts are already defined in
+% \texttt{latex.tex}.
+% \begin{macrocode}
+% \newcount \@chclass
+% \newcount \@chnum
+% \newcount \@lastchclass
+% \end{macrocode}
+% \end{macro}
+% \end{macro}
+% \end{macro}
+%
+%
+%
+% \begin{macro}{\@addtopreamble}
+% We will save the already constructed preamble for
+% the =\halign=
+% in the global macro =\@preamble=. This will then be
+% enlarged with
+% the command =\@addtopreamble=.
+% \begin{macrocode}
+\def\@addtopreamble#1{\xdef\@preamble{\@preamble #1}}
+% \end{macrocode}
+% \end{macro}
+%
+%
+%
+%
+%
+% \subsection{The character class of a \textsf{token}}
+%
+% \begin{macro}{\@testpach}
+% \changes{v2.0a}{1989/05/12}{p option renamed to m (middle).}
+% \changes{v2.0a}{1989/05/12}{t option renamed to p to be compatible to
+% the original.}
+% With the help of =\@lastchclass= we can now define a macro
+% which determines the class and the number of a given preamble
+% \textsf{token}
+% and assigns them to the registers
+% =\@chclass= and =\@chnum=.
+% \changes{v2.0f}{1992/02/29}{Argument removed since implicitly known}
+% \begin{macrocode}
+\def\@testpach{\@chclass
+% \end{macrocode}
+% First we deal with the cases in which the \textsf{token}
+% (=#1=) is the argument of \texttt{!}, \texttt{@}, \texttt{<} or
+% \texttt{>}. We can see this from the value of =\@lastchclass=:
+% \begin{macrocode}
+ \ifnum \@lastchclass=6 \@ne \@chnum \@ne \else
+ \ifnum \@lastchclass=7 5 \else
+ \ifnum \@lastchclass=8 \tw@ \else
+ \ifnum \@lastchclass=9 \thr@@
+% \end{macrocode}
+% Otherwise we will assume that the \textsf{token} belongs to the
+% class $0$
+% and assign the corresponding number to =\@chnum= if our
+% assumption is correct.
+% \begin{macrocode}
+ \else \z@
+% \end{macrocode}
+% If the last \textsf{token} was a \texttt{p}, \texttt{m} or a
+% \texttt{b}, =\@chnum= already has the right value. This is the
+% reason for the somewhat curious choice of the \textsf{token}
+% numbers in class $10$.
+% \begin{macrocode}
+ \ifnum \@lastchclass = 10 \else
+% \end{macrocode}
+% Otherwise we will check if =\@nextchar= is either a \texttt{c},
+% \texttt{l} or an \texttt{r}. Some applications change the
+% catcodes of certain characters like ``\texttt{@}'' in
+% \texttt{amstex.sty}. As a result the tests below would fail since
+% they assume non-active character tokens. Therefore we evaluate
+% =\@nextchar= once thereby turning the first token of its
+% replacement text into a char. At this point here this should have
+% been the only char present in =\@nextchar= which put into via a
+% =\def=.
+% \changes{v2.0f}{1992/02/29}{Ensure to test a char which is not active}
+% \begin{macrocode}
+ \edef\@nextchar{\expandafter\string\@nextchar}%
+ \@chnum
+ \if \@nextchar c\z@ \else
+ \if \@nextchar l\@ne \else
+ \if \@nextchar r\tw@ \else
+% \end{macrocode}
+% If it is a different \textsf{token}, we know that the class was
+% not $0$. We assign the value $0$ to =\@chnum= because this value
+% is needed for the \texttt{|}--\textsf{token}. Now we must check
+% the remaining classes. Note that the value of =\@chnum= is
+% insignificant here for most classes.
+% \begin{macrocode}
+ \z@ \@chclass
+ \if\@nextchar |\@ne \else
+ \if \@nextchar !6 \else
+ \if \@nextchar @7 \else
+ \if \@nextchar <8 \else
+ \if \@nextchar >9 \else
+% \end{macrocode}
+% The remaining permitted \textsf{tokens} are \texttt{p},
+% \texttt{m} and \texttt{b} (class $10$).
+% \begin{macrocode}
+ 10
+ \@chnum
+ \if \@nextchar m\thr@@\else
+ \if \@nextchar p4 \else
+ \if \@nextchar b5 \else
+% \end{macrocode}
+% Now the only remaining possibility is a forbidden \textsf{token},
+% so we choose class $0$ and number $0$ and give an error message.
+% Then we finish the macro by closing all =\if='s.
+% \begin{macrocode}
+ \z@ \@chclass \z@ \@preamerr \z@ \fi \fi \fi \fi
+ \fi \fi \fi \fi \fi \fi \fi \fi \fi \fi \fi \fi}
+% \end{macrocode}
+% \end{macro}
+%
+%
+%
+%
+%
+% \subsection{Multiple columns ($*$--form)}
+%
+% \begin{macro}{\@xexpast}
+% \begin{macro}{\the@toks}
+% \begin{macro}{\the@toksz}
+% \label{@xexpast} Now we discuss the macro that deletes all forms
+% of type =*{=\textit{N\/}=}{=\textit{String\/}=}= from a user
+% preamble and replaces them with \textit{N} copies of
+% \textit{String}. Nested $*$--expressions are dealt with
+% correctly, that means $*$--expressions are not substituted if
+% they are in explicit braces, as in =@{*}=.
+%
+% This macro is called via
+% =\@xexpast=\meta{preamble}=*0x\@@=.
+% The $*$--expression =*0x= is being used to terminate the
+% recursion,
+% as we shall see later, and =\@@= serves as an argument
+% delimiter. =\@xexpast= has four arguments. The first
+% one is the part of the
+% user preamble before the first $*$--expression while the second
+% and third ones are the arguments of the first $*$--expression
+% (that is \textit{N} and \textit{String} in the notation mentioned
+% above).
+% The fourth argument is the rest of the preamble.
+% \begin{macrocode}
+\def\@xexpast#1*#2#3#4\@@{%
+% \end{macrocode}
+% The number of copies of \textit{String} (=#2=) that are to be
+% produced will be saved in a \textsf{count} register.
+% \begin{macrocode}
+ \@tempcnta #2
+% \end{macrocode}
+% We save the part of the preamble which does not
+% contain a $*$--form (=#1=)
+% in a \PlainTeX\ \textsf{token} register.
+% We also save \textit{String} (=#3=) using a \LaTeX\
+% \textsf{token} register.
+% \begin{macrocode}
+ \toks@={#1}\@temptokena={#3}%
+% \end{macrocode}
+% Now we have to use a little trick to produce \textit{N} copies of
+% \textit{String}.
+% We could try =\def\@tempa{#1}= and then
+% \textit{N} times =\edef\@tempa{\@tempa#3}=. This would have the
+% undesired effect that all macros within =#1= and =#3=
+% would be expanded, although, for example, constructions like
+% =@{..}= are not supposed to be changed.
+% That is why we =\let= two control sequences to
+% be equivalent to =\relax=.
+% \begin{macrocode}
+ \let\the@toksz\relax \let\the@toks\relax
+% \end{macrocode}
+% Then we ensure that =\@tempa= contains
+% ={\the@toksz\the@toks...\the@toks}= (the macro
+% =\the@toks= exactly \textit{N\/} times) as substitution text.
+% \begin{macrocode}
+ \def\@tempa{\the@toksz}%
+ \ifnum\@tempcnta >0 \@whilenum\@tempcnta >0\do
+ {\edef\@tempa{\@tempa\the@toks}\advance \@tempcnta \m@ne}%
+% \end{macrocode}
+% If \textit{N\/} was greater than zero we prepare for another call
+% of =\@xexpast=. Otherwise we assume we have reached the end of
+% the user preamble, because we had appended =*0x\@@= when we first
+% called =\@xexpast=. In other words: if the user inserts
+% =*{0}{..}= in his preamble, \LaTeX\ ignores the rest of it.
+% \begin{macrocode}
+ \let \@tempb \@xexpast \else
+ \let \@tempb \@xexnoop \fi
+% \end{macrocode}
+% Now we will make sure that the part of the user preamble, which
+% was already dealt with, will be saved again in =\@tempa=.
+% \begin{macrocode}
+ \def\the@toksz{\the\toks@}\def\the@toks{\the\@temptokena}%
+ \edef\@tempa{\@tempa}%
+% \end{macrocode}
+% We have now evaluated the first $*$--expression, and the user
+% preamble up to this point
+% is saved in =\@tempa=. We will put the contents of
+% =\@tempa= and the rest of the user preamble together and work
+% on the result with =\@tempb=. This macro either corresponds
+% to =\@xexpast=, so that the next
+% $*$--expression is handled, or to the macro =\@xexnoop=,
+% which only ends the recursion by deleting its argument.
+% \begin{macrocode}
+ \expandafter \@tempb \@tempa #4\@@}
+% \end{macrocode}
+% \end{macro}
+% \end{macro}
+% \end{macro}
+%
+% \begin{macro}{\@xexnoop}
+% So the first big problem is solved. Now it is easy to
+% specify =\@xexnoop=.
+% Its argument is delimited by =\@@= and it simply expands to
+% nothing.
+% \begin{macrocode}
+% \def\@xexnoop#1\@@{}
+% \end{macrocode}
+% \end{macro}
+%
+%
+%
+%
+% \section{The insertion of declarations
+% (\texttt{>}, \texttt{<}, \texttt{!}, \texttt{@})}
+%
+%
+% The preamble will be enlarged with the help of =\xdef=, but the
+% arguments of \texttt{>}, \texttt{<},~\texttt{!}\ and \texttt{@} are
+% not supposed to be expanded during the construction (we want an
+% implementation that doesn't need a =\protect=). So we have to find a
+% way to inhibit the expansion of those arguments.
+%
+% We will solve this problem with \textsf{token} registers. We need
+% one register for every \texttt{!}\ and \texttt{@}, while we need two
+% for every \texttt{c}, \texttt{l}, \texttt{r}, \texttt{m}, \texttt{p}
+% or \texttt{b}. This limits the number of columns of a table because
+% there are only 256 \textsf{token} registers. But then, who needs
+% tables with more than 100 columns?
+%
+% One could also find a solution which only needs two or three
+% \textsf{token} registers by proceeding similarly as in the macro
+% =\@xexpast= (see page \pageref{@xexpast}). The advantage of our
+% approach is the fact that we avoid some of the problems that arise
+% with the other method\footnote{Maybe there are also historical
+% reasons.}.
+%
+% So how do we proceed? Let us assume that we had =!{foo}= in the
+% user preamble and say we saved \texttt{foo} in
+% \textsf{token} register $5$. Then we call
+% =\@addtopreamble{\the@toks5}= where
+% =\the@toks= is defined in a way that it does not expand
+% (for example it could be equivalent to =\relax=). Every
+% following call
+% of =\@addtopreamble= leaves =\the@toks5= unchanged in
+% =\@preamble=. If the construction of the preamble is completed
+% we change the definition of =\the@toks= to
+% =\the\toks= and expand =\@preamble= for the last time.
+% During this process all parts of the form
+% =\the@toks=\meta{Number}
+% will be substituted by the contents of the respective \textsf{token}
+% registers.
+%
+% As we can see from this informal discussion the construction of the
+% preamble has to take place within a group, so that the
+% \textsf{token} registers we use will be freed later on. For that
+% reason we keep all assignments to =\@preamble= global; therefore the
+% replacement text of this macro will remain the same after we leave
+% the group.
+%
+% \begin{macro}{\count@}
+% We further need a \textsf{count} register to remember which
+% \textsf{token} register is to be used next. This will be
+% initialized with $-1$ if we want to begin with the \textsf{token}
+% register $0$. We use the \PlainTeX\ scratch register =\count@=
+% because everything takes place locally. All we have to do is
+% insert =\the@toks= =\the= =\count@= into the preamble.
+% =\the@toks= will remain unchanged and =\the\count@= expands into
+% the saved number.
+% \end{macro}
+%
+% \begin{macro}{\prepnext@tok}
+% The macro =\prepnext@tok= is in charge of preparing the next
+% \textsf{token} register. For that purpose we increase
+% =\count@= by $1$:
+% \begin{macrocode}
+\def\prepnext@tok{\advance \count@ \@ne
+% \end{macrocode}
+% Then we locally delete any contents the
+% \textsf{token} register might have.
+% \begin{macrocode}
+ \toks\count@{}}
+% \end{macrocode}
+% \end{macro}
+%
+% \begin{macro}{\save@decl}
+% During the construction of the preamble the current
+% \textsf{token} is always saved in the macro =\@nextchar= (see the
+% definition of =\@mkpream= on page \pageref{@mkpream}). The macro
+% =\save@decl= saves it into the next free \textsf{token} register,
+% i.e.\ in =\toks\count@=.
+% \changes{v2.0c}{1990/08/14}{\cs{relax} removed and added elsewhere.}
+% \begin{macrocode}
+\def\save@decl{\toks\count@ \expandafter{\@nextchar}}
+% \end{macrocode}
+% The reason for the use of =\relax= is the following
+% hypothetical situation in the preamble:
+% \quad =..\the\toks1\the\toks2..= \quad \TeX\ expands
+% =\the\toks2= first in order to find out if the digit =1=
+% is followed by other digits. E.g.\ a =5= saved in the
+% \textsf{token} register $2$ would lead \TeX\ to insert the contents
+% of \textsf{token} register $15$ instead of $1$ later on.
+%
+% The example above referred to an older version of =\save@decl= which
+% inserted a =\relex= inside the token register. This is now moved to
+% the places where the actual token registers are inserted (look for
+% =\the@toks=) because the old version would still make =@=
+% expressions to moving arguments since after expanding the second
+% register while looking for the end of the number the contents of the
+% token register is added so that later on the whole register will be
+% expanded. This serious bug was found after nearly two years
+% international use of this package by Johannes Braams.
+% \end{macro}
+%
+%
+%
+% How does the situation look like, if we want to add another column
+% to the preamble, i.e.\ if we have found a \texttt{c}, \texttt{l},
+% \texttt{r}, \texttt{p}, \texttt{m} or \texttt{b} in the user
+% preamble? In this case we have the problem of the \textsf{token}
+% register from =>{..}= and =<{..}= having to be inserted at this
+% moment because formating instructions like =\hfil= have to be set
+% around them. On the other hand it is not known yet, if any =<{..}=
+% instruction will appear in the user preamble at all.
+%
+% We solve this problem by adding two \textsf{token} registers at a
+% time. This explains, why we have freed the \textsf{token} registers
+% in =\prepnext@tok=.
+%
+% \begin{macro}{\insert@column}
+% \begin{macro}{\@sharp}
+% We now define the macro =\insert@column= which will do
+% this work for us.
+% \begin{macrocode}
+\def\insert@column{%
+% \end{macrocode}
+% Here, we assume that the \textsf{count} register
+% =\@tempcnta= has saved the value $=\count@= - 1$.
+% \begin{macrocode}
+ \the@toks \the \@tempcnta
+% \end{macrocode}
+% Next follows the =#= sign which specifies the place
+% where the text of the column shall be inserted. To avoid
+% errors during the expansions in
+% =\@addtopreamble= we hide this sign in the command
+% =\@sharp= which is temporarily occupied with
+% =\relax= during the build-up of the preamble.
+% To remove unwanted spaces before and after the column text, we set
+% an =\ignorespaces= in front and a =\unskip= afterwards.
+% \changes{v2.0e}{1991/02/07}{Added \{\} around \cs{@sharp} for new ftsel}
+% \changes{v2.0h}{1992/06/22}{Removed \{\} again in favour of
+% \cs{d@llarbegin}}
+% \begin{macrocode}
+ \ignorespaces \@sharp \unskip
+% \end{macrocode}
+% Then the second \textsf{token} register follows whose number should
+% be saved in =\count@=.
+% We make sure that there will be no further expansion after reading
+% the number, by finishing with =\relax=. The case above is not
+% critical since it is ended by =\ignorespaces=.
+% \changes{v2.0c}{1990/08/14}{\cs{relax} added to avoid problem
+% \cs{the}\cs{toks0}\cs{the}\cs{toks1}.}
+% \begin{macrocode}
+ \the@toks \the \count@ \relax}
+% \end{macrocode}
+% \end{macro}
+% \end{macro}
+%
+%
+%
+%
+% \subsection{The separation of columns}
+%
+% \begin{macro}{\@addamp}
+% In the preamble a =&= has to be inserted between any two columns;
+% before the first column there should not be a =&=. As the user
+% preamble may start with a \texttt{|} we have to remember somehow
+% if we have already inserted a =#= (i.e.\ a column). This is done
+% with the boolean variable =\if@firstamp= that we test in
+% =\@addamp=, the macro that inserts the =&=.
+% \begin{macrocode}
+% \newif \@iffirstamp
+% \def\@addamp{\if@firstamp \@firstampfalse
+% \else \@addtopreamble &\fi}
+% \end{macrocode}
+% \end{macro}
+%
+% \begin{macro}{\@acol}
+% \begin{macro}{\@acolampacol}
+% \begin{macro}{\col@sep}
+% We will now define some abbreviations for the extensions,
+% appearing most often in the preamble build-up.
+% Here =\col@sep= is a \textsf{dimen} register which is set
+% equivalent to =\arraycolsep= in an \textsf{array}--environment,
+% otherwise it is set equivalent to =\tabcolsep=.
+% \begin{macrocode}
+\newdimen\col@sep
+\def\@acol{\@addtopreamble{\hskip\col@sep}}
+% \def\@acolampacol{\@acol\@addamp\@acol}
+% \end{macrocode}
+% \end{macro}
+% \end{macro}
+% \end{macro}
+%
+%
+% \subsection{The macro \texttt{\textbackslash @mkpream}}
+%
+% \begin{macro}{\@mkpream}
+% \begin{macro}{\the@toks}
+% \label{@mkpream}
+% Now we can define the macro which builds up the preamble for the
+% =\halign=.
+% First we initialize =\@preamble=, =\@lastchclass=
+% and the boolean variable =\if@firstamp=.
+% \begin{macrocode}
+\def\@mkpream#1{\gdef\@preamble{}\@lastchclass 4 \@firstamptrue
+% \end{macrocode}
+% During the build-up of the preamble we cannot directly use the
+% =#= sign; this would lead to an error message in the next
+% =\@addtopreamble= call.
+% Instead, we use the command =\@sharp= at places where later
+% a =#= will be.
+% This command is at first given the meaning =\relax=;
+% therefore it will not be expanded when the preamble
+% is extended.
+% In the macro =\@array=, shortly before the =\halign=
+% is carried out, =\@sharp= is given its final meaning.
+%
+% In a similar way,
+% we deal with the commands =\@startpbox= and
+% =\@endpbox=, although the reason is different here: these
+% macros expand in many \textsf{tokens} which would delay the
+% build-up of the preamble.
+% \begin{macrocode}
+ \let\@sharp\relax \let\@startpbox\relax \let\@endpbox\relax
+% \end{macrocode}
+% Now we remove possible $*$-forms in the user preamble with the
+% command =\@xexpast=. As we already know, this command saves
+% its result in the macro =\@tempa=.
+% \begin{macrocode}
+ \@xexpast #1*0x\@@
+% \end{macrocode}
+% Afterwards we initialize all registers and macros, that we need
+% for the build-up of the preamble.
+% Since we want to start with the \textsf{token} register $0$,
+% =\count@= has to contain the value $-1$.
+% \begin{macrocode}
+ \count@\m@ne
+ \let\the@toks\relax
+% \end{macrocode}
+% Then we call up =\prepnext@tok= in order to prepare the
+% \textsf{token} register $0$ for use.
+% \begin{macrocode}
+ \prepnext@tok
+% \end{macrocode}
+% To evaluate the user preamble (without stars) saved in
+% =\@tempa= we use the \LaTeX--macro =\@tfor=.
+% The strange appearing construction with =\expandafter= is
+% based on the fact that we have to put the replacement text of
+% =\@tempa= and not the macro =\@tempa= to this
+% \LaTeX--macro.
+% \begin{macrocode}
+ \expandafter \@tfor \expandafter \@nextchar
+ \expandafter :\expandafter =\@tempa \do
+% \end{macrocode}
+% The body of this loop (the group after the =\do=)
+% is executed for one \textsf{token} at a time, whereas
+% the current \textsf{token} is saved in =\@nextchar=.
+% At first we evaluate the current \textsf{token} with the already
+% defined macro =\@testpach=, i.e.\ we assign to
+% =\@chclass= the character class and to =\@chnum=
+% the character number of this \textsf{token}.
+% \changes{v2.0f}{1992/02/29}{\cs{@testpach} now without arg}
+% \begin{macrocode}
+ {\@testpach
+% \end{macrocode}
+% Then we branch out depending on the value of =\@chclass= into
+% different macros that extend the preamble respectively.
+% \begin{macrocode}
+ \ifcase \@chclass \@classz \or \@classi \or \@classii
+ \or \save@decl \or \or \@classv \or \@classvi
+ \or \@classvii \or \@classviii \or \@classix
+ \or \@classx \fi
+% \end{macrocode}
+% Two cases deserve our special attention: Since the current
+% \textsf{token} cannot have the character class $4$ (start) we
+% have skipped this possibility. If the character class is $3$,
+% only the content of =\@nextchar= has to be saved into the current
+% \textsf{token} register; therefore we call up =\save@decl=
+% directly and save a macro name. After the preamble has been
+% extended we assign the value of =\@chclass= to the counter
+% =\@lastchclass= to assure that this information will be available
+% during the next run of the loop.
+% \begin{macrocode}
+ \@lastchclass\@chclass}%
+% \end{macrocode}
+% After the loop has been finished space must still be added to
+% the created preamble, depending on the last \textsf{token}.
+% Depending on the value of =\@lastchclass= we perform
+% the necessary operations.
+% \begin{macrocode}
+ \ifcase\@lastchclass
+% \end{macrocode}
+% If the last class equals $0$ we add a
+% =\hskip \col@sep=.
+% \begin{macrocode}
+ \@acol \or
+% \end{macrocode}
+% If it equals $1$ we do not add any additional space so that the
+% horizontal lines do not exceed the vertical ones.
+% \begin{macrocode}
+ \or
+% \end{macrocode}
+% Class $2$ is treated like class $0$ because a =<{...}= can
+% only directly follow after class $0$.
+% \begin{macrocode}
+ \@acol \or
+% \end{macrocode}
+% Most of the other possibilities can only appear if the user
+% preamble was defective. Class $3$ is not allowed since after a
+% =>{..}= there must always follow a \texttt{c}, \texttt{l},
+% \texttt{r}, \texttt{p},\texttt{m} or \texttt{b}. We report an
+% error and ignore the declaration given by ={..}=.
+% \begin{macrocode}
+ \@preamerr \thr@@ \or
+% \end{macrocode}
+% If =\@lastchclass= is $4$ the user preamble has been empty.
+% To continue, we insert a =#= in the preamble.
+% \begin{macrocode}
+ \@preamerr \tw@ \@addtopreamble\@sharp \or
+% \end{macrocode}
+% Class $5$ is allowed again. In this case
+% (the user preamble ends with =@{..}=) we need not
+% do anything.
+% \begin{macrocode}
+ \or
+% \end{macrocode}
+% Any other case means that the arguments to =@=, \texttt{!},
+% \texttt{<}, \texttt{>}, \texttt{p}, \texttt{m} or \texttt{b} have
+% been forgotten. So we report an error and ignore the last
+% \textsf{token}.
+% \begin{macrocode}
+ \else \@preamerr \@ne \fi
+% \end{macrocode}
+% Now that the build-up of the preamble is almost finished we can
+% insert the \textsf{token} registers and therefore redefine
+% =\the@toks=. The actual insertion, though, is performed
+% later.
+% \begin{macrocode}
+ \def\the@toks{\the\toks}}
+% \end{macrocode}
+% \end{macro}
+% \end{macro}
+%
+%
+%
+% \section{The macros \texttt{\textbackslash @classz}
+% to \texttt{\textbackslash @classx}}
+%
+% The preamble is extended by the macros =\@classz= to
+% =\@classx= which are called by =\@mkpream=
+% depending on =\@lastchclass=
+% (i.e. the character class of the last \textsf{token}).
+% \begin{macro}{\@classx}
+% First we define =\@classx= because of its important r\^ole.
+% When it is called we find that the current
+% \textsf{token} is \texttt{p}, \texttt{m} or \texttt{b}.
+% That means that a new column has to start.
+% \begin{macrocode}
+\def\@classx{%
+% \end{macrocode}
+% Depending on the value of =\@lastchclass= different actions
+% must take place:
+% \begin{macrocode}
+ \ifcase \@lastchclass
+% \end{macrocode}
+% If the last character class was $0$ we separate the columns by
+% =\hskip\col@sep= followed by =&= and another
+% =\hskip\col@sep=.
+% \begin{macrocode}
+ \@acolampacol \or
+% \end{macrocode}
+% If the last class was class $1$ --- that means that a vertical
+% line was
+% drawn, --- before this line a =\hskip\col@sep= was inserted.
+% Therefore there has to be only a =&= followed by
+% =\hskip\col@sep=. But this =&= may be inserted only
+% if this is not the first column. This process is controlled
+% by =\if@firstamp= in the macro =\addamp=.
+% \begin{macrocode}
+ \@addamp \@acol \or
+% \end{macrocode}
+% Class $2$ is treated like class $0$ because =<{...}= can only
+% follow after class $0$.
+% \begin{macrocode}
+ \@acolampacol \or
+% \end{macrocode}
+% Class $3$ requires no actions because all things necessary have
+% been done by the preamble \textsf{token} \texttt{>}.
+% \begin{macrocode}
+ \or
+% \end{macrocode}
+% Class $4$ means that we are at the beginning of the preamble.
+% Therefore we start the preamble with =\hskip\col@sep= and
+% then call =\@firstampfalse=. This makes sure that a later
+% =\@addamp= inserts the character
+% =&= into the preamble.
+% \begin{macrocode}
+ \@acol \@firstampfalse \or
+% \end{macrocode}
+% For class $5$ \textsf{tokens} only the character =&= is inserted
+% as a column separator. Therefore we call =\@addamp=.
+% \begin{macrocode}
+ \@addamp
+% \end{macrocode}
+% Other cases are impossible. For an example
+% $=\@lastchclass= \string= 6$---as it might appear in a
+% preamble of the form =...!p...=---\texttt{p} would have
+% been taken as an argument of \texttt{!}\ by =\@testpach=.
+% \begin{macrocode}
+ \fi}
+% \end{macrocode}
+% \end{macro}
+%
+%
+% \begin{macro}{\@classz}
+% If the character class of the last \textsf{token} is $0$ we have
+% \texttt{c}, \texttt{l}, \texttt{r} or an argument of \texttt{m},
+% \texttt{b} or\ \texttt{p}. In the first three cases the preamble
+% must be extended the same way as if we had class $10$. The
+% remaining two cases do not require any action because the space
+% needed was generated by the last \textsf{token} (i.e.\
+% \texttt{m}, \texttt{b} or \texttt{p}). Since =\@lastchclass= has
+% the value $10$ at this point nothing happens when =\@classx= is
+% called. So the macro =\@chlassz= may start like this:
+% \begin{macrocode}
+\def\@classz{\@classx
+% \end{macrocode}
+% According to the definition of =\insert@column= we must store
+% the number of the \textsf{token} register in which a preceding
+% =>{..}= might have stored its argument into
+% =\@tempcnta=.
+% \begin{macrocode}
+ \@tempcnta \count@
+% \end{macrocode}
+% To have $=\count@= \string= =\@tmpcnta= + 1$ we prepare
+% the next \textsf{token} register.
+% \begin{macrocode}
+ \prepnext@tok
+% \end{macrocode}
+% Now the preamble must be extended with the column whose format
+% can be determinated by =\@chnum=.
+% \begin{macrocode}
+ \@addtopreamble{\ifcase \@chnum
+% \end{macrocode}
+% If =\@chnum= has the value $0$ a centered column has to be
+% generated.
+% So we begin with stretchable space.
+% \begin{macrocode}
+ \hfil
+% \end{macrocode}
+% The command =\d@llarbegin= follows expanding into =\begingroup=
+% (in the \textsf{tabular}--environment) or into =$=. Doing this
+% (provided an appropriate setting of =\d@llarbegin=) we achieve
+% that the contents of the columns of an \textsf{array}--environment
+% are set in math mode while those of a \textsf{tabular}--environment
+% are set in LR mode.
+% \begin{macrocode}
+ \d@llarbegin
+% \end{macrocode}
+% Now we insert the contents of the two \textsf{token} registers
+% and the symbol
+% for the column entry (i.e.\ =#= or
+% more precise =\@sharp=) using =\insert@column=.
+% \begin{macrocode}
+ \insert@column
+% \end{macrocode}
+% We end this case with =\d@llarend= and =\hfil= where =\d@llarend=
+% again is either =$= or =\endgroup=.
+% \begin{macrocode}
+ \d@llarend \hfil \or
+% \end{macrocode}
+% The templates for \texttt{l} and \texttt{r} (i.e.\ =\@chnum= $1$
+% or $2$) are generated the same way. Since one =\hfil= is
+% missing the text is moved to the relevant side.
+% The =\kern\z@= is needed in case of an empty column
+% entry. Otherwise
+% the =\unskip= in =\insert@column= removes the
+% =\hfil=. Changed to =\hskip1sp= so that it interacts better with
+% =\@bsphack=.
+% \changes{v2.3f}{1996/04/22}
+% {(DPC) Extra \cs{kern} keeps tabcolsep in empty l columns
+% internal/2122}
+% \changes{v2.3i}{1996/06/14}
+% {Change both \cs{kern}\cs{z@} to \cs{hskip}1sp for latex/2160}
+% \begin{macrocode}
+ \hskip1sp\d@llarbegin \insert@column \d@llarend \hfil \or
+ \hfil\hskip1sp\d@llarbegin \insert@column \d@llarend \or
+% \end{macrocode}
+% The templates for \texttt{p}, \texttt{m} and \texttt{b} mainly
+% consist of a \textsf{box}. In case of \texttt{m} it is generated
+% by =\vcenter=. This command is allowed only in math
+% mode. Therefore we start with a~=$=.
+% \begin{macrocode}
+ $\vcenter
+% \end{macrocode}
+% The part of the templates which is the same in all three cases
+% (\texttt{p}, \texttt{m} and \texttt{b})
+% is built by the macros =\@startpbox= and
+% =\@endpbox=. =\@startpbox= has an argument:
+% the width of the column which is stored in the current
+% \textsf{token} (i.e.\ =\@nextchar=).
+% Between these two macros we find the well known
+% =\insert@column=.
+% \begin{macrocode}
+ \@startpbox{\@nextchar}\insert@column \@endpbox $\or
+% \end{macrocode}
+% The templates for \texttt{p} and \texttt{b} are generated in the
+% same way though we do not need the =$= characters because we use
+% =\vtop= or =\vbox=.
+% \begin{macrocode}
+ \vtop \@startpbox{\@nextchar}\insert@column \@endpbox \or
+ \vbox \@startpbox{\@nextchar}\insert@column \@endpbox
+% \end{macrocode}
+% Other values for =\@chnum= are impossible. Therefore we
+% end the arguments to =\@addtopreamble= and =\ifcase=.
+% Before we come to the end of =\@classz= we have to
+% prepare the next \textsf{token} register.
+% \begin{macrocode}
+ \fi}\prepnext@tok}
+% \end{macrocode}
+% \end{macro}
+%
+%
+% \begin{macro}{\@classix}
+% In case of class $9$ (\texttt{>}--\textsf{token}) we first check
+% if the character class of the last
+% \textsf{token} was $3$. In this case we have a
+% user preamble of the form =..>{...}>{...}..= which
+% is not allowed. We only give an error message and continue.
+% So the declarations defined by the first =>{...}=
+% are ignored.
+% \begin{macrocode}
+\def\@classix{\ifnum \@lastchclass = \thr@@
+ \@preamerr \thr@@ \fi
+% \end{macrocode}
+% Furthermore, we call up =\@class10= because afterwards always a
+% new column is started by \texttt{c}, \texttt{l}, \texttt{r},
+% \texttt{p}, \texttt{m} or \texttt{b}.
+% \begin{macrocode}
+ \@classx}
+% \end{macrocode}
+% \end{macro}
+%
+%
+%
+% \begin{macro}{\@classviii}
+% If the current \textsf{token} is a \texttt{<} the last character
+% class must be $0$. In this case it is not necessary to extend the
+% preamble. Otherwise we output an error message, set =\@chclass=
+% to $6$ and call =\@classvi=. By doing this we achieve that
+% \texttt{<} is treated like \texttt{!}.
+% \begin{macrocode}
+\def\@classviii{\ifnum \@lastchclass >\z@
+ \@preamerr 4\@chclass 6 \@classvi \fi}
+% \end{macrocode}
+% \end{macro}
+%
+% \begin{macro}{\@arrayrule}
+% There is only one incompatibility with the original definition:
+% the definition of =\@arrayrule=. In the original a line without
+% width\footnote{So the space between \texttt{cc} and \texttt{c|c}
+% is equal.} is created by multiple insertions of
+% =\hskip .5\arrayrulewidth=.
+% We only insert a vertical line into the
+% preamble. This is done to prevent problems with \TeX's main
+% memory when generating tables with many vertical lines in them
+% (especially in the case of \textsf{floats}).
+% \begin{macrocode}
+\def\@arrayrule{\@addtopreamble \vline}
+% \end{macrocode}
+% \end{macro}
+%
+% \begin{macro}{\@classvii}
+% As a consequence it follows that in case of class $7$
+% (=@= \textsf{token}) the preamble need not to be extended.
+% In the original definition $=\@lastchclass= \string= 1$
+% is treated by inserting =\hskip .5\arrayrulewidth=.
+% We only check if the last \textsf{token} was of class $3$ which is
+% forbidden.
+% \begin{macrocode}
+\def\@classvii{\ifnum \@lastchclass = \thr@@
+% \end{macrocode}
+% If this is true we output an error message and
+% ignore the declarations stored
+% by the last =>{...}=, because these are overwritten
+% by the argument of \texttt{@}.
+% \begin{macrocode}
+ \@preamerr \thr@@ \fi}
+% \end{macrocode}
+% \end{macro}
+%
+%
+% \begin{macro}{\@classvi}
+% If the current \textsf{token} is a regular \texttt{!}\ and the
+% last class was $0$ or $2$ we extend the preamble with
+% =\hskip\col@sep=. If the last \textsf{token} was of class $1$
+% (for instance \texttt{|}) we extend with =\hskip \doublerulesep=
+% because the construction =!{...}= has to be treated like
+% \texttt{|}.
+% \begin{macrocode}
+\def\@classvi{\ifcase \@lastchclass
+ \@acol \or
+ \@addtopreamble{\hskip \doublerulesep}\or
+ \@acol \or
+% \end{macrocode}
+% Now =\@preamerr...= should follow because a
+% user preamble of the form =..>{..}!.= is not allowed.
+% To save memory we call =\@classvii= instead which also
+% does what we want.
+% \begin{macrocode}
+ \@classvii
+% \end{macrocode}
+% If =\@lastchclass= is $4$ or $5$ nothing has to be done.
+% Class $6$ to $10$ are not possible.
+% So we finish the macro.
+% \begin{macrocode}
+ \fi}
+% \end{macrocode}
+% \end{macro}
+%
+% \begin{macro}{\@classii}
+% \begin{macro}{\@classiii}
+% In the case of character classes $2$ and $3$ (i.e.\ the argument
+% of \texttt{<} or \texttt{>}) we only have to store the current
+% \textsf{token} (=\@nextchar=) into the corresponding
+% \textsf{token} register since the preparation and
+% insertion of these registers
+% are done by the macro =\@classz=.
+% This is equivalent to calling =\save@decl= in the case of
+% class $3$. To save command identifiers we do this call up
+% in the macro =\@mkpream=.
+%
+% Class $2$ exhibits a more complicated situation: the
+% \textsf{token} registers have already been inserted by
+% =\@classz=. So the value of =\count@= is too high
+% by one. Therefore we decrease =\count@= by $1$.
+% \begin{macrocode}
+\def\@classii{\advance \count@ \m@ne
+% \end{macrocode}
+% Next we store the current \textsf{token} into the correct
+% \textsf{token} register by calling =\save@decl= and then
+% increase the value of =\count@= again. At this point we
+% can save memory once more (at the cost of time) if we use the
+% macro =\prepnext@tok=.
+% \begin{macrocode}
+ \save@decl\prepnext@tok}
+% \end{macrocode}
+% \end{macro}
+% \end{macro}
+%
+%
+% \begin{macro}{\@classv}
+% If the current \textsf{token} is of class $5$ then it is an
+% argument of a \texttt{@} \textsf{token}. It must be stored into a
+% \textsf{token} register.
+% \begin{macrocode}
+\def\@classv{\save@decl
+% \end{macrocode}
+% We extend the preamble with a command which inserts this
+% \textsf{token} register into the preamble when its construction
+% is finished. The user expects that this argument is worked out in
+% math mode if it was used in an
+% \textsf{array}--environment. Therefore we surround it with
+% =\d@llar...='s.
+% \changes{v2.0c}{1990/08/14}{\cs{relax} added to avoid problem
+% `the`toks0`the`toks1.}
+% \begin{macrocode}
+ \@addtopreamble{\d@llarbegin\the@toks\the\count@\relax\d@llarend}%
+% \end{macrocode}
+% Finally we must prepare the next \textsf{token} register.
+% \begin{macrocode}
+ \prepnext@tok}
+% \end{macrocode}
+% \end{macro}
+%
+% \begin{macro}{\@classi}
+% In the case of class $0$ we were able to generate the necessary
+% space between columns by using the macro =\@classx=.
+% Analogously the macro =\@classvi= can be used for class $1$.
+% \begin{macrocode}
+\def\@classi{\@classvi
+% \end{macrocode}
+% Depending on =\@chnum= a vertical line
+% \begin{macrocode}
+ \ifcase \@chnum \@arrayrule \or
+% \end{macrocode}
+% or (in case of =!{...}=) the current \textsf{token} --- stored
+% in =\@nextchar= --- has to be inserted into the preamble.
+% This corresponds to calling =\@classv=.
+% \begin{macrocode}
+ \@classv \fi}
+% \end{macrocode}
+% \end{macro}
+%
+%
+%
+% \begin{macro}{\@startpbox}
+% In =\@classz= the macro =\@startpbox= is used.
+% The width of the \textsf{parbox} is passed as an argument.
+% =\vcenter=, =\vtop= or =\vbox= are already in the
+% preamble. So we start with the braces for the wanted box.
+% \begin{macrocode}
+\def\@startpbox#1{\bgroup
+% \end{macrocode}
+% The argument is the width of the box. This information has to be
+% assigned to =\hsize=.
+% Then we assain default values to several parameters used in a
+% \textsf{parbox}.
+% \changes{v2.3k}{1998/05/12}{Use \cs{setlength} to set \cs{hsize},
+% so that the calc package can be applied here (pr/2793)}
+% \begin{macrocode}
+ \setlength\hsize{#1}\@arrayparboxrestore
+% \end{macrocode}
+% Our main problem is to obtain the same distance between succeeding
+% lines of the \textsf{parbox}.
+% We have to remember that the distance between two \textsf{parboxes}
+% should be defined by =\@arstrut=. That means that it can be
+% greater than the distance in a \textsf{parbox}.
+% Therefore it is not enough to set a =\@arstrut= at the
+% beginning and at the end of the \textsf{parbox}. This would
+% dimension the distance
+% between first and second line and the distance between the two
+% last lines of the \textsf{parbox} wrongly.
+% To prevent this we set an invisible rule of height
+% =\@arstrutbox=
+% at the beginning of the \textsf{parbox}. This has no effect on the
+% depth of the first line. At the end of the \textsf{parbox} we set
+% analogously another invisible rule which only affects the depth
+% of the last line. It is necessary to wait inserting this strut
+% until the paragraph actually starts to allow for things like
+% =\parindent= changes via =>{...}=.
+% \changes{v2.1c}{1992/12/14}{Use `everypar to insert strut}
+% \begin{macrocode}
+ \everypar{%
+ \vrule \@height \ht\@arstrutbox \@width \z@
+ \everypar{}}%
+ }
+% \end{macrocode}
+% \end{macro}
+%
+% \begin{macro}{\@endpbox}
+% If there are any declarations defined by =>{...}=
+% and =<{...}=
+% they now follow in the macro =\@classz= --- the contents
+% of the column in between.
+% So the macro =\@endpbox= must insert the \textsf{specialstrut}
+% mentioned earlier and then close the group opened by
+% =\@startpbox=.
+% \changes{v2.2d}{1994/05/16}{Use \LaTeXe \cs{@finalstrut}}
+% \changes{v2.3g}{1996/05/07}{Add \cs{hfil} for tools/2120}
+% \begin{macrocode}
+\def\@endpbox{\@finalstrut\@arstrutbox \egroup\hfil}
+% \end{macrocode}
+% \end{macro}
+%
+%
+% \section{Building and calling \texttt{\textbackslash halign}}
+%
+% \begin{macro}{\@array}
+% After we have discussed the macros needed for the evaluation
+% of the user preamble we can define the macro =\@array=
+% which uses these macros to create a =\halign=.
+% It has two arguments. The first one is a position argument
+% which can be \texttt{t}, \texttt{b} or \texttt{c}; the
+% second one describes the wanted preamble,
+% e.g.\ it has the form =|c|c|c|=.
+% \begin{macrocode}
+\def\@array[#1]#2{%
+% \end{macrocode}
+% First we define a \textsf{strut} whose size basically corresponds
+% to a normal \textsf{strut} multiplied by the factor
+% =\arraystretch=.
+% This \textsf{strut} is then inserted into every row and enforces
+% a minimal distance between two rows.
+% Nevertheless, when using horizontal lines, large letters
+% (like accented capital letters) still collide with such lines.
+% Therefore at first we add to the height of a normal \textsf{strut}
+% the value of the parameter =\extrarowheight=.
+% \begin{macrocode}
+ \@tempdima \ht \strutbox
+ \advance \@tempdima by\extrarowheight
+ \setbox \@arstrutbox \hbox{\vrule
+ \@height \arraystretch \@tempdima
+ \@depth \arraystretch \dp \strutbox
+ \@width \z@}%
+% \end{macrocode}
+% Then we open a group, in which the user preamble is evaluated by
+% the macro =\@mkpream=. As we know this must happen locally.
+% This macro creates a preamble for a =\halign= and saves
+% its result globally in the control sequence =\@preamble=.
+% \begin{macrocode}
+ \begingroup
+ \@mkpream{#2}%
+% \end{macrocode}
+% We again redefine =\@preamble= so that a call up of =\@preamble=
+% now starts the =\halign=. Thus also the arguments of \texttt{>},
+% \texttt{<}, \texttt{@} and \texttt{!}, saved in the
+% \textsf{token} registers are inserted into the preamble. The
+% =\tabskip= at the beginning and end of the preamble is set to
+% \textsf{0pt} (in the beginning by the use of =\ialign=). Also the
+% command =\@arstrut= is build in, which inserts the
+% =\@arstrutbox=, defined above. Of course, the opening brace after
+% =\ialign= has to be implicit as it will be closed in =\endarray=
+% or another macro.
+% \changes{v2.3m}{1998/12/31}{Added \cs{noexpand} in front of \cs{ialign}
+% to guard against interesting :-) changes to \cs{halign} done to support
+% text glyphs in math}
+%
+% The =\noexpand= in front of =\ialign= does no harm in standard \LaTeX{}
+% and was added since some experimental support for using text glyphs in math
+% redefines =\halign= with the result that is becomes expandable with
+% disastrous results in cases like this.
+% In the kernel definition for this macro the problem does
+% not surface because there =\protect= is set (which is not necessary in this
+% implementation as there is no arbitrary user input that can get expanded) and
+% the experimental code made the redefinition robust. Whether this is the right
+% approach is open to question; consider the =\noexpand= a curtesy to allow an
+% unsupported redefinition of a \TeX{} primitive for the moment (as people rely
+% on that experimental code).
+% \begin{macrocode}
+ \xdef\@preamble{\noexpand \ialign \@halignto
+ \bgroup \@arstrut \@preamble
+ \tabskip \z@ \cr}%
+% \end{macrocode}
+% What we have not explained yet is the macro =\@halignto=
+% that was just used. Depending on its replacement text the
+% =\halign= becomes a =\halign= \texttt{to} \meta{dimen}.
+% Now we close the group again. Thus
+% =\@startpbox= and =\@endpbox= as well as all
+% \textsf{token} registers get their former meaning back.
+% \begin{macrocode}
+ \endgroup
+% \end{macrocode}
+% To support the \texttt{delarray.sty} package we include a hook
+% into this part of the code which is a no-op in the main package.
+% \changes{v2.1a}{1992/07/03}{Hook for delarray added}
+% \begin{macrocode}
+ \@arrayleft
+% \end{macrocode}
+% Now we decide depending on the position argument in which
+% \textsf{box} the =\halign= is to be put. (=\vcenter= may be used
+% because we are in math mode.)
+% \changes{v2.1a}{1992/07/03}{Wrong spec is now equiv to [t]}
+% \begin{macrocode}
+ \if #1t\vtop \else \if#1b\vbox \else \vcenter \fi \fi
+% \end{macrocode}
+% Now another implicit opening brace appears; then definitions
+% which shall stay local follow. While constructing the
+% =\@preamble= in =\@mkpream= the =#= sign must be
+% hidden in the macro =\@sharp= which is =\let= to
+% =\relax= at that moment (see definition of =\@mkpream=
+% on page~\pageref{@mkpream}).
+% All these now get their actual meaning.
+% \begin{macrocode}
+ \bgroup
+ \let \@sharp ##\let \protect \relax
+% \end{macrocode}
+% With the above defined \textsf{struts} we fix down the distance
+% between rows by setting =\lineskip= and =\baselineskip=
+% to \textsf{0pt}. Since there have to be set =$='s
+% around every column in the \textsf{array}--environment
+% the parameter =\mathsurround= should
+% also be set to \textsf{0pt}. This prevents additional space between
+% the rows. The
+% \PlainTeX--macro =\m@th= does this.
+% \begin{macrocode}
+ \lineskip \z@
+ \baselineskip \z@
+ \m@th
+% \end{macrocode}
+% Beside, we have to assign a special meaning (which we still have
+% to specify) to the line separator =\\=. We also have to
+% redefine the command =\par= in such a way that empty lines in
+% =\halign= cannot do any damage. We succeed in doing so
+% by choosing something that will disappear when expanding.
+% After that we only have to call up =\@preamble= to
+% start the wanted =\halign=.
+% \changes{1994/12/08}{v2.3b}{add \cs{tabularnewline}}
+% \begin{macrocode}
+ \let\\\@arraycr \let\tabularnewline\\\let\par\@empty \@preamble}
+% \end{macrocode}
+% \end{macro}
+%
+%
+% \begin{macro}{\extrarowheight}
+% The \textsf{dimen} parameter used above also needs to be
+% allocated. As a default value we use \textsf{0pt}, to ensure
+% compatibility with standard \LaTeX.
+% \begin{macrocode}
+\newdimen \extrarowheight
+\extrarowheight=0pt
+% \end{macrocode}
+% \end{macro}
+%
+% \begin{macro}{\@arstrut}
+% Now the insertion of =\@arstrutbox= through =\@arstut=
+% is easy since we know exactly in which mode \TeX\ is while working
+% on the =\halign= preamble.
+% \begin{macrocode}
+\def\@arstrut{\unhcopy\@arstrutbox}
+% \end{macrocode}
+% \end{macro}
+%
+%
+% \section{The line separator \texttt{\textbackslash\textbackslash}}
+%
+% \begin{macro}{\@arraycr}
+% In the macro =\@array= the line separator =\\= is
+% =\let= to the command =\@arraycr=.
+% Its definition starts with a special brace which I have directly
+% copied from the original definition. It is
+% necessary, because the =\futurlet= in =\@ifnextchar=
+% might
+% expand a following =&= \textsf{token} in a construction like
+% =\\ &=. This would otherwise end the alignment template at a
+% wrong time. On the other hand we have to be careful to avoid
+% producing a real group, i.e.\ ={}=, because the command will also
+% be used for the array environment, i.e.\ in math mode. In that
+% case an extra ={}= would produce an ord atom which could mess up
+% the spacing. For this reason we use a combination that does not
+% really produce a group at all but modifies the master counter so
+% that a =&= will not be considered belonging to the current
+% =\halign= while we are looking for a =*= or =[=.
+% For further information see
+% \cite[Appendix D]{bk:knuth}.
+% \changes{v2.3c}{1995/04/23}{Avoid adding an ord atom in math}
+% \begin{macrocode}
+\def\@arraycr{\relax\iffalse{\fi\ifnum 0=`}\fi
+% \end{macrocode}
+% Then we test whether the user is using the star form and ignore
+% a possible star (I also disagree with this procedure, because a
+% star does not make any sense here).
+% \begin{macrocode}
+ \@ifstar \@xarraycr \@xarraycr}
+% \end{macrocode}
+% \end{macro}
+%
+% \begin{macro}{\@xarraycr}
+% In the command =\@xarraycr= we test if an optional argument
+% exists.
+% \begin{macrocode}
+\def\@xarraycr{\@ifnextchar [%
+% \end{macrocode}
+% If it does, we branch out into the macro =\@argarraycr= if
+% not we close the special brace (mentioned above) and end the row
+% of the =\halign= with a =\cr=.
+% \changes{v2.3c}{1995/04/23}{Avoid adding an ord atom in math}
+% \begin{macrocode}
+ \@argarraycr {\ifnum 0=`{}\fi\cr}}
+% \end{macrocode}
+% \end{macro}
+%
+%
+% \begin{macro}{\@argarraycr}
+% If additional space is requested by the user this case is treated
+% in the macro =\@argarraycr=. First we close the special brace
+% and then we test if the additional space is positive.
+% \changes{v2.3c}{1995/04/23}{Avoid adding an ord atom in math}
+% \begin{macrocode}
+\def\@argarraycr[#1]{\ifnum0=`{}\fi\ifdim #1>\z@
+% \end{macrocode}
+% If this is the case we create an invisible vertical rule with
+% depth =\dp\@arstutbox=${}+{}$\meta{wanted\ space}.
+% Thus we achieve that all vertical lines specified
+% in the user preamble by a \texttt{|} are now
+% generally drawn.
+% Then the row ends with a =\cr=.
+%
+% If the space is negative we end the row at once with a =\cr=
+% and move back up with a =\vskip=.
+%
+% While testing these macros I found out that the
+% =\endtemplate=
+% created by =\cr= and =&= is something like an
+% =\outer= primitive and therefore it should not appear in
+% incomplete =\if= statements. Thus the following solution was
+% chosen which hides the =\cr= in other macros when \TeX\
+% is skipping conditional text.
+% \changes{v2.3c}{1995/04/23}{Use \cs{expandafter}'s in conditional}
+% \begin{macrocode}
+ \expandafter\@xargarraycr\else
+ \expandafter\@yargarraycr\fi{#1}}
+% \end{macrocode}
+% \end{macro}
+%
+% \begin{macro}{\@xargarraycr}
+% \begin{macro}{\@yargarraycr}
+% The following macros were already explained above.
+% \begin{macrocode}
+\def\@xargarraycr#1{\unskip
+ \@tempdima #1\advance\@tempdima \dp\@arstrutbox
+ \vrule \@depth\@tempdima \@width\z@ \cr}
+\def\@yargarraycr#1{\cr\noalign{\vskip #1}}
+% \end{macrocode}
+% \end{macro}
+% \end{macro}
+%
+%
+%
+%
+% \section{Spanning several columns}
+%
+% \begin{macro}{\multicolumn}
+% If several columns should be held together with a special format
+% the command =\multicolumn= must be used. It has three
+% arguments: the number of columns to be covered; the format for
+% the result column and the actual column entry.
+% \changes{v2.3j}{1998/01/29}{Command made \cs{long} to match
+% kernel change for pr/2180}
+% \begin{macrocode}
+\long\def\multicolumn#1#2#3{%
+% \end{macrocode}
+% First we combine the given number of columns into a single one;
+% then we start a new block so that the following definition is kept
+% local.
+% \begin{macrocode}
+ \multispan{#1}\begingroup
+% \end{macrocode}
+% Since a =\multicolumn= should only describe the format of a
+% result column, we redefine =\@addamp= in such a way that one gets
+% an error message if one uses more than one \texttt{c},
+% \texttt{l}, \texttt{r}, \texttt{p}, \texttt{m} or \texttt{b} in
+% the second argument. One should consider that this definition is
+% local to the build-up of the preamble; an \textsf{array}-- or
+% \textsf{tabular}--environment in the third argument of the
+% =\multicolumn= is therefore worked through correctly as well.
+% \begin{macrocode}
+ \def\@addamp{\if@firstamp \@firstampfalse \else
+ \@preamerr 5\fi}%
+% \end{macrocode}
+% Then we evaluate the second argument with the help of
+% =\@mkpream=.
+% Now we still have to insert the contents of the \textsf{token}
+% register into the =\@preamble=, i.e.\ we have to say
+% =\xdef\@preamble{\@preamble}=. This is achieved shorter by
+% writing:
+% \begin{macrocode}
+ \@mkpream{#2}\@addtopreamble\@empty
+% \end{macrocode}
+% After the =\@preamble= is created we forget all local
+% definitions and occupations of the \textsf{token} registers.
+% \begin{macrocode}
+ \endgroup
+% \end{macrocode}
+% In the special situation of =\multicolumn= =\@preamble=
+% is not needed as preamble for a =\halign= but it is directly
+% inserted into our table. Thus instead of =\sharp=
+% there has to be the column entry (=#3=) wanted by the user.
+% \begin{macrocode}
+ \def\@sharp{#3}%
+% \end{macrocode}
+% Now we can pass the =\@preamble= to \TeX\ . For safety
+% we start with an =\@arstrut=. This should usually be in the
+% template for the first column however we do not know if this
+% template was overwritten by our =\multicolumn=.
+% We also add a =\null= at the right end to prevent any following
+% =\unskip= (for example from =\\[..]=) to remove the =\tabcolsep=.
+% \changes{v2.2e}{1994/06/01}{Added \cs{null}}
+% \begin{macrocode}
+ \@arstrut \@preamble
+ \null
+ \ignorespaces}
+% \end{macrocode}
+% \end{macro}
+%
+%
+%
+% \section{The Environment Definitions}
+%
+% After these preparations we are able to define the environments. They
+% only differ in the initialisations of =\d@llar...=, =\col@sep=
+% and =\@halignto=.
+%
+% \begin{macro}{\@halignto}
+% \begin{macro}{\d@llarbegin}
+% \begin{macro}{\d@llarend}
+% In order to relieve the \textsf{save stack} we assign the
+% replacement texts for =\@halignto= globally. =\d@llar= has to be
+% local since otherwise nested \textsf{tabular} and \textsf{array}
+% environments (via =\multicolumn=) are impossible.
+% \changes{v2.0g}{1992/06/18}{`d@llarbegin defined on toplevel.}
+% When the new font selection scheme is in force we have to
+% we surround all =\halign= entries
+% with braces. See remarks in TUGboat 10\#2. Actually we are going
+% to use =\begingroup= and =\endgroup=. However, this is only
+% necessary when we are in text mode. In math the surrounding
+% dollar signs will already serve as the necessary extra grouping
+% level. Therefore we switch the settings of =\d@llarbegin= and
+% =\d@llarend= between groups and dollar signs.
+% \begin{macrocode}
+\let\d@llarbegin\begingroup
+\let\d@llarend\endgroup
+% \end{macrocode}
+% \end{macro}
+% \end{macro}
+% \end{macro}
+%
+%
+% \begin{macro}{\array}
+% Our new definition of =\array= then reads:
+% \changes{v2.0d}{1990/08/20}{`d@llar local to preamble.}
+% \begin{macrocode}
+\def\array{\col@sep\arraycolsep
+ \def\d@llarbegin{$}\let\d@llarend\d@llarbegin\gdef\@halignto{}%
+% \end{macrocode}
+% Since there might be an optional argument we call another
+% macro which is also used by the other environments.
+% \begin{macrocode}
+ \@tabarray}
+% \end{macrocode}
+% \end{macro}
+%
+% \begin{macro}{\@tabarray}
+% This macro tests for a optional bracket and then calls up
+% =\@array= or =\@array[c]= (as default).
+% \begin{macrocode}
+\def\@tabarray{\@ifnextchar[{\@array}{\@array[c]}}
+% \end{macrocode}
+% \end{macro}
+%
+%
+% \begin{macro}{\tabular}
+% \begin{macro}{\tabular*}
+% The environments \textsf{tabular} and \textsf{tabular$*$} differ
+% only in the initialisation of the command =\@halignto=. Therefore
+% we define
+% \begin{macrocode}
+\def\tabular{\gdef\@halignto{}\@tabular}
+% \end{macrocode}
+% and analoguesly for the star form. We evalute the argument first
+% using =\setlength= so that users of the \texttt{calc} package can
+% write code like\\ =\begin{tabular*}{(\columnwidth-1cm)/2}...=
+% \changes{v2.3l}{1998/05/13}{Use \cs{setlength} evaluate arg
+% so that the calc package can be applied here (pr/2793)}
+% \begin{macrocode}
+\expandafter\def\csname tabular*\endcsname#1{%
+ \setlength\dimen@{#1}%
+ \xdef\@halignto{to\the\dimen@}\@tabular}
+% \end{macrocode}
+% \end{macro}
+% \end{macro}
+%
+% \begin{macro}{\@tabular}
+% The rest of the job is carried out by the =\@tabular= macro:
+% \begin{macrocode}
+\def\@tabular{%
+% \end{macrocode}
+% First of all we have to make sure that we start out in
+% \textsf{hmode}. Otherwise we might find our table dangling by
+% itself on a line.
+% \begin{macrocode}
+ \leavevmode
+% \end{macrocode}
+% It should be taken into consideration that the macro =\@array=
+% must be called in math mode. Therefore we open a \textsf{box},
+% insert a =$= and then assign the correct values to =\col@sep= and
+% =\d@llar...=.
+% \changes{v2.0d}{1990/08/20}{`d@llar local to preamble.}
+% \begin{macrocode}
+ \hbox \bgroup $\col@sep\tabcolsep \let\d@llarbegin\begingroup
+ \let\d@llarend\endgroup
+% \end{macrocode}
+% Now everything \textsf{tabular} specific is done and we are able to
+% call the =\@tabarray= macro.
+% \begin{macrocode}
+ \@tabarray}
+% \end{macrocode}
+% \end{macro}
+%
+% \begin{macro}{\endarray}
+% When the processing of \textsf{array} is finished we have to
+% close the =\halign=
+% and afterwards the surrounding \textsf{box} selected by
+% =\@array=. To save \textsf{token} space we then redefine
+% =\@preamble=
+% because its replacement text isn't longer needed.
+% \begin{macrocode}
+\def\endarray{\crcr \egroup \egroup \gdef\@preamble{}}
+% \end{macrocode}
+% \end{macro}
+%
+% \begin{macro}{\endtabular}
+% \begin{macro}{\endtabular*}
+% To end a \textsf{tabular} or \textsf{tabular$*$} environment we
+% call up =\endarray=, close the math mode and then the surrounding
+% =\hbox=.
+% \begin{macrocode}
+\def\endtabular{\endarray $\egroup}
+\expandafter\let\csname endtabular*\endcsname=\endtabular
+% \end{macrocode}
+% \end{macro}
+% \end{macro}
+%
+%
+%
+% \section{Last minute definitions}
+%
+%
+% If this file is used as a package file we should =\let= all macros
+% to =\relax= that were used in the original but are no longer
+% necessary.
+% \begin{macrocode}
+\let\@ampacol=\relax \let\@expast=\relax
+\let\@arrayclassiv=\relax \let\@arrayclassz=\relax
+\let\@tabclassiv=\relax \let\@tabclassz=\relax
+\let\@arrayacol=\relax \let\@tabacol=\relax
+\let\@tabularcr=\relax \let\@@endpbox=\relax
+\let\@argtabularcr=\relax \let\@xtabularcr=\relax
+% \end{macrocode}
+%
+% \begin{macro}{\@preamerr}
+% We also have to redefine the error routine =\@preamerr= since
+% new kind of errors are possible.
+% The code for this macro is not perfect yet;
+% it still needs too much memory.
+% \begin{macrocode}
+\def\@preamerr#1{\def\@tempd{{..} at wrong position: }%
+ \PackageError{array}{%
+ \ifcase #1 Illegal pream-token (\@nextchar): `c' used\or %0
+ Missing arg: token ignored\or %1
+ Empty preamble: `l' used\or %2
+ >\@tempd token ignored\or %3
+ <\@tempd changed to !{..}\or %4
+ Only one column-spec. allowed.\fi}\@ehc} %5
+% \end{macrocode}
+% \end{macro}
+%
+%
+%
+% \section
+% [Defining your own column specifiers]
+% {Defining your own column specifiers\footnotemark}
+%
+% \footnotetext{The code and the documentation in this section was
+% written by David. So far only the code from newarray was plugged
+% into array so that some parts of the documentation still claim
+% that this is newarray and even worse, some parts of the code are
+% unnecessarily doubled. This will go away in a future release. For
+% the moment we thought it would be more important to bring both
+% packages together.}
+% \changes{v2.1a}{1992/07/03}{Newcolumn stuff added}
+%
+% \DeleteShortVerb{\=}
+% \MakeShortVerb{\"}
+%
+% \begin{macro}{\newcolumn}
+% In \texttt{newarray.sty} the macro for specifying new columns was
+% named "\newcolumn". When the functionality was added to
+% \texttt{array.sty} the command was renamed "\newcolumntype".
+% Initially both names were supported, but now (In versions of this
+% package distributed for \LaTeXe) the old name is not defined.
+% \changes{v2.2a}{1994/02/03}{Now made `newcolumn an error}
+% \changes{v2.2a}{1994/02/04}{Removed `newcolumn}
+% \begin{macrocode}
+%<*ncols>
+% \end{macrocode}
+% \end{macro}
+%
+% \begin{macro}{\newcolumntype}
+% \changes{v2.1b}{1992/06/07}{Macro renamed from `newcolumn} As
+% described above, the "\newcolumntype" macro gives users the chance
+% to define letters, to be used in the same way as the primitive
+% column specifiers, `c' `p' etc.
+% \begin{macrocode}
+\def\newcolumntype#1{%
+% \end{macrocode}
+% "\NC@char" was added in V2.01 so that active characters, like "@" in
+% AMS\LaTeX\ may be used. This trick was stolen from \texttt{array.sty}
+% 2.0h. Note that we need to use the possibly active token,
+% "#1", in several places, as that is the token that actually
+% appears in the preamble argument.
+% \begin{macrocode}
+ \edef\NC@char{\string#1}%
+% \end{macrocode}
+% First we check whether there is already a definition for this column.
+% Unlike "\newcommand" we give a warning rather than an error if it is
+% defined. If it is a new column, add "\NC@do" \meta{column} to
+% the list "\NC@list".
+% \begin{macrocode}
+ \@ifundefined{NC@find@\NC@char}%
+ {\@tfor\next:=<>clrmbp@!|\do{\if\noexpand\next\NC@char
+ \PackageWarning{array}%
+ {Redefining primitive column \NC@char}\fi}%
+ \NC@list\expandafter{\the\NC@list\NC@do#1}}%
+ {\PackageWarning{array}{Column \NC@char\space is already defined}}%
+% \end{macrocode}
+% Now we define a macro with an argument delimited by the new column
+% specifier, this is used to find occurences of this specifier in the
+% user preamble.
+% \begin{macrocode}
+ \@namedef{NC@find@\NC@char}##1#1{\NC@{##1}}%
+% \end{macrocode}
+% If an optional argument was not given, give a default argument of 0.
+% \begin{macrocode}
+ \@ifnextchar[{\newcol@{\NC@char}}{\newcol@{\NC@char}[0]}}
+% \end{macrocode}
+% \end{macro}
+% \begin{macro}{\newcol@}
+% We can now define the macro which does the rewriting,
+% "\@reargdef" takes the same arguments as "\newcommand", but
+% does not check that the command is new. For a column, say `D' with
+% one argument, define a command "\NC@rewrite@D" with one
+% argument, which recursively calls "\NC@find" on the user preamble
+% after replacing the first token or group with the replacement text
+% specified in the "\newcolumntype" command. "\NC@find" will find the
+% next occurrence of `D' as it will be "\let" equal to
+% "\NC@find@D" by "\NC@do".
+% \begin{macrocode}
+\def\newcol@#1[#2]#3{\expandafter\@reargdef
+ \csname NC@rewrite@#1\endcsname[#2]{\NC@find#3}}
+% \end{macrocode}
+% \end{macro}
+% \begin{macro}{\NC@}
+% Having found an occurence of the new column, save the preamble
+% before the column in "\@temptokena", then check to see if we
+% are at the end of the preamble. (A dummy occurrence of the column
+% specifier will be placed at the end of the preamble by "\NC@do".
+% \begin{macrocode}
+\def\NC@#1{%
+ \@temptokena\expandafter{\the\@temptokena#1}\futurelet\next\NC@ifend}
+% \end{macrocode}
+% \end{macro}
+% \begin{macro}{\NC@ifend}
+% We can tell that we are at the end as "\NC@do" will place a "\relax"
+% after the dummy column.
+% \begin{macrocode}
+\def\NC@ifend{%
+% \end{macrocode}
+% If we are at the end, do nothing. (The whole preamble will now be in
+% "\@temptokena".)
+% \begin{macrocode}
+ \ifx\next\relax
+% \end{macrocode}
+% Otherwise set the flag "\if@tempswa", and rewrite the column.
+% "\expandafter" introduced 1n V2.01
+% \begin{macrocode}
+ \else\@tempswatrue\expandafter\NC@rewrite\fi}
+% \end{macrocode}
+% \end{macro}
+% \begin{macro}{\NC@do}
+% If the user has specified `C' and `L' as new columns, the list of
+% rewrites (in the token register "\NC@list") will look like
+% "\NC@do *" "\NC@do C" "\NC@do L".
+% So we need to define "\NC@do" as a one argument macro which
+% initialises the rewriting of the specified column. Let us assume that
+% `C' is the argument.
+% \begin{macrocode}
+\def\NC@do#1{%
+% \end{macrocode}
+% First we let "\NC@rewrite" and "\NC@find" be
+% "\NC@rewrite@C" and "\NC@find@C" respectively.
+% \begin{macrocode}
+ \expandafter\let\expandafter\NC@rewrite
+ \csname NC@rewrite@\string#1\endcsname
+ \expandafter\let\expandafter\NC@find
+ \csname NC@find@\string#1\endcsname
+% \end{macrocode}
+% Clear the token register "\@temptokena" after putting the present
+% contents of the register in front of the token "\NC@find". At the
+% end we place the tokens `"C\relax"' which "\NC@ifend" will use
+% to detect the end of the user preamble.
+% \begin{macrocode}
+ \expandafter\@temptokena\expandafter{\expandafter}%
+ \expandafter\NC@find\the\@temptokena#1\relax}
+% \end{macrocode}
+% \end{macro}
+%
+% \begin{macro}{\showcols}
+% This macro is useful for debugging "\newcolumntype" specifications,
+% it is the equivalent of the primitive "\show" command for macro
+% definitions. All we need to do is locally redefine "\NC@do" to take
+% its argument (say `C') and then "\show" the (slightly modified)
+% definition of "\NC@rewrite@C". Actually as the the list always
+% starts off with "\NC@do *" and we do not want to print the
+% definition of the $*$-form, define "\NC@do" to throw away the first
+% item in the list, and then redefine itsef to print the rest of the
+% definitions.
+% \begin{macrocode}
+\def\showcols{{\def\NC@do##1{\let\NC@do\NC@show}\the\NC@list}}
+% \end{macrocode}
+% \end{macro}
+% \begin{macro}{\NC@show}
+% If the column `C' is defined as above, then
+% "\show\NC@rewrite@C" would output\\
+% "\long macro: ->\NC@find >{$}c<{$}".
+% We want to strip the "long macro: ->" and the "\NC@find". So first we
+% use "\meaning" and then apply the macro "\NC@strip" to the tokens so
+% produced and then "\typeout" the required string.
+% \begin{macrocode}
+\def\NC@show#1{%
+ \typeout{Column #1\expandafter\expandafter\expandafter\NC@strip
+ \expandafter\meaning\csname NC@rewrite@#1\endcsname\@@}}
+% \end{macrocode}
+% \end{macro}
+% \begin{macro}{\NC@strip}
+% Delimit the arguments to "\NC@strip" with `\texttt{:}', `\texttt{->}',
+% a space, and "\@@" to pull out the required parts of the output from
+% "\meaning".
+% \begin{macrocode}
+\def\NC@strip#1:#2->#3 #4\@@{#2 -> #4}
+% \end{macrocode}
+% \end{macro}
+% \begin{macro}{\NC@list}
+% Allocate the token register used for the rewrite list.
+% \begin{macrocode}
+\newtoks\NC@list
+% \end{macrocode}
+% \end{macro}
+%
+% \subsection{The $*$--form}
+% We view the $*$-form as a slight generalisation of the system
+% described in the previous subsection. The idea is to define a $*$
+% column by a command of the form:
+% \begin{verbatim}
+% \newcolumntype{*}[2]{%
+% \count@=#1\ifnum\count@>0
+% \advance\count@ by -1 #2*{\count@}{#2}\fi}
+% \end{verbatim}
+% \begin{macro}{\NC@rewrite@*}\label{NC@rewrite@*}
+% This does not work however as "\newcolumntype" takes great care not
+% to expand anything in the preamble, and so the "\if" is never
+% expanded. "\newcolumntype" sets up various other parts of the
+% rewrite correctly though so we can define:
+% \begin{macrocode}
+\newcolumntype{*}[2]{}
+% \end{macrocode}
+% Now we must correct the definition of "\NC@rewrite@*". The
+% following is probably more efficient than a direct translation of
+% the idea sketched above, we do not need to put a $*$ in the preamble
+% and call the rewrite recursively, we can just put "#1" copies of
+% "#2" into "\@temptokena". (Nested $*$ forms will be expanded
+% when the whole rewrite list is expanded again, see "\@mkpream")
+% \begin{macrocode}
+\long\@namedef{NC@rewrite@*}#1#2{%
+% \end{macrocode}
+% Store the number.
+% \begin{macrocode}
+ \count@#1
+% \end{macrocode}
+% Put "#1" copies of "#2" in the token register.
+% \begin{macrocode}
+ \loop
+ \ifnum\count@>\z@
+ \advance\count@\m@ne
+ \@temptokena\expandafter{\the\@temptokena#2}%
+ \repeat
+% \end{macrocode}
+% "\NC@do" will ensure that "\NC@find" is "\let" equal
+% to "\NC@find@*".
+% \begin{macrocode}
+ \NC@find}
+% \end{macrocode}
+% \end{macro}
+%
+% \subsection{Modifications to internal macros of \texttt{array.sty}}
+%
+% \begin{macro}{\@xexpast}
+% \begin{macro}{\@xexnoop}
+% These macros are used to expand $*$-forms in
+% \texttt{array.sty}. "\let" them to "\relax" to save space.
+% \begin{macrocode}
+\let\@xexpast\relax
+\let\@xexnoop\relax
+% \end{macrocode}
+% \end{macro}
+% \end{macro}
+%
+% \begin{macro}{\save@decl}
+% We do not assume that the token register is free, we add the new
+% declarations to the front of the register. This is to allow user
+% preambles of the form, ">{foo}>{bar}..". Users are not encouraged to
+% enter such expressions directly, but they may result from the
+% rewriting of "\newcolumntype"'s.
+% \begin{macrocode}
+\def\save@decl{\toks \count@ = \expandafter\expandafter\expandafter
+ {\expandafter\@nextchar\the\toks\count@}}
+% \end{macrocode}
+% \end{macro}
+% \begin{macro}{\@mkpream}
+% The main modification to "\@mkpream" is to replace the call to
+% "\@xexpast" (which expanded $*$-forms) by a loop which expands
+% all "\newcolumntype" specifiers.
+% \begin{macrocode}
+\def\@mkpream#1{\gdef\@preamble{}\@lastchclass 4 \@firstamptrue
+ \let\@sharp\relax \let\@startpbox\relax \let\@endpbox\relax
+% \end{macrocode}
+% Now we remove possible $*$-forms and user-defined column
+% specifiers in the user preamble by repeatedly executing the list
+% "\NC@list" until the re-writes have no more effect. The
+% expanded preamble will then be in the token register
+% "\@temptokena". Actually we need to know at this point that
+% this is not "\toks0".
+% \begin{macrocode}
+ \@temptokena{#1}\@tempswatrue
+ \@whilesw\if@tempswa\fi{\@tempswafalse\the\NC@list}%
+% \end{macrocode}
+% Afterwards we initialize all registers and macros, that we need
+% for the build-up of the preamble.
+% \begin{macrocode}
+ \count@\m@ne
+ \let\the@toks\relax
+ \prepnext@tok
+% \end{macrocode}
+% Having expanded all tokens defined using "\newcolumntype" (including
+% "*"), we evaluate the remaining tokens, which are saved in
+% "\@temptokena". We use the \LaTeX--macro "\@tfor" to inspect each
+% token in turn.
+% \begin{macrocode}
+ \expandafter \@tfor \expandafter \@nextchar
+ \expandafter :\expandafter =\the\@temptokena \do
+% \end{macrocode}
+% "\@testpatch" does not take an argument since \texttt{array.sty} 2.0h.
+% \begin{macrocode}
+ {\@testpach
+ \ifcase \@chclass \@classz \or \@classi \or \@classii
+ \or \save@decl \or \or \@classv \or \@classvi
+ \or \@classvii \or \@classviii
+% \end{macrocode}
+% In \texttt{newarray.sty} class 9 is equivalent to class 10.
+% \begin{macrocode}
+ \or \@classx
+ \or \@classx \fi
+ \@lastchclass\@chclass}%
+ \ifcase\@lastchclass
+ \@acol \or
+ \or
+ \@acol \or
+ \@preamerr \thr@@ \or
+ \@preamerr \tw@ \@addtopreamble\@sharp \or
+ \or
+ \else \@preamerr \@ne \fi
+ \def\the@toks{\the\toks}}
+% \end{macrocode}
+% \end{macro}
+%
+% \begin{macro}{\@classix}
+% \texttt{array.sty} does not allow repeated \texttt{>}
+% declarations for the same column. This is allowed in
+% \texttt{newarray.sty} as documented in the introduction. Removing
+% the test for this case makes class 9 equivalent to class 10, and
+% so this macro is redundant. It is "\let" to "\relax" to save
+% space.
+% \begin{macrocode}
+\let\@classix\relax
+% \end{macrocode}
+% \end{macro}
+%
+% \begin{macro}{\@classviii}
+% In \texttt{newarray.sty} explicitly allow class 2, as repeated
+% \texttt{<} expressions are accepted by this package.
+% \begin{macrocode}
+\def\@classviii{\ifnum \@lastchclass >\z@\ifnum\@lastchclass=\tw@\else
+ \@preamerr 4\@chclass 6 \@classvi \fi\fi}
+% \end{macrocode}
+% \end{macro}
+%
+% \begin{macro}{\@classv}
+% Class 5 is \texttt{@}-expressions (and is also called by class 1)
+% This macro was incorrect in Version~1. Now we do not expand the
+% "@"-expression, but instead explicitly replace an
+% "\extracolsep" command by an assignment to "\tabskip" by a
+% method similar to the "\newcolumntype" system described above.
+% "\d@llarbegin" "\d@llarend" were introduced in V2.01 to match
+% \texttt{array.sty} 2.0h.
+% \begin{macrocode}
+\def\@classv{\save@decl
+ \expandafter\NC@ecs\@nextchar\extracolsep{}\extracolsep\@@@
+ \@addtopreamble{\d@llarbegin\the@toks\the\count@\relax\d@llarend}%
+ \prepnext@tok}
+% \end{macrocode}
+% \end{macro}
+%
+% \begin{macro}{\NC@ecs}
+% Rewrite the first occurrence of "\extracolsep{1in}" to
+% "\tabskip1in\relax". As a side effect discard any tokens after a
+% second "\extracolsep", there is no point in the user entering two of
+% these commands anyway, so this is not really a restriction.
+% \begin{macrocode}
+\def\NC@ecs#1\extracolsep#2#3\extracolsep#4\@@@{\def\@tempa{#2}%
+ \ifx\@tempa\@empty\else\toks\count@={#1\tabskip#2\relax#3}\fi}
+%</ncols>
+% \end{macrocode}
+% \end{macro}
+%
+%
+% \subsection{Support for the \texttt{delarray.sty}}
+%
+% The \texttt{delarray.sty} package extends the array syntax by
+% supporting the notation of delimiters. To this end we extend the
+% array parsing mechanism to include a hook which can be used by this
+% (or another) package to do some additional parsing.
+%
+% \begin{macro}{\@tabarray}
+% This macro tests for an optional bracket and then calls up
+% "\@@array" or "\@@array[c]" (as default).
+% \begin{macrocode}
+%<*package>
+\def\@tabarray{\@ifnextchar[{\@@array}{\@@array[c]}}
+% \end{macrocode}
+% \end{macro}
+% \begin{macro}{\@@array}
+% This macro tests could then test an optional delimiter before the
+% left brace of the main preamble argument. Here in the main package
+% it simply is let to be "\@array".
+% \begin{macrocode}
+\let\@@array\@array
+% \end{macrocode}
+% \end{macro}
+%
+% \begin{macro}{\endarray}
+% \begin{macro}{\@arrayright}
+% We have to declare the hook we put into "\@array" above.
+% A similar hook `"\@arrayright"' will be inserted into the
+% "\endarray" to gain control. Both defaults to empty.
+% \begin{macrocode}
+\def\endarray{\crcr \egroup \egroup \@arrayright \gdef\@preamble{}}
+\let\@arrayleft\@empty
+\let\@arrayright\@empty
+% \end{macrocode}
+% \end{macro}
+% \end{macro}
+%
+% \subsection{Support for \texttt{\textbackslash firsthline} and
+% \texttt{\textbackslash lasthline}}
+%
+% The Companion~\cite[p.137]{bk:GMS94} suggests two additional
+% commands to control the allignments in case of tabulars with
+% horizontal lines. They are now added to this package.
+%
+% \begin{macro}{\extratabsurround}
+% The extra space around a table when "\firsthline" or "\lasthline"
+% are used.
+% \begin{macrocode}
+\newlength{\extratabsurround}
+\setlength{\extratabsurround}{2pt}
+% \end{macrocode}
+% \end{macro}
+%
+% \begin{macro}{\backup@length}
+% This register will be used internally by "\firsthline" and
+% "\lasthline".
+% \begin{macrocode}
+\newlength{\backup@length}
+% \end{macrocode}
+% \end{macro}
+%
+% \begin{macro}{\firsthline}
+% \changes{v2.3h}{1996/05/25}{Complete reimplementation}
+% This code can probably be improved but for the moment it should
+% serve.
+%
+% We start by producing a single tabular row without any visible
+% content that will produce the external reference point in case
+% "[t]" is used.
+% \begin{macrocode}
+\newcommand{\firsthline}{%
+ \multicolumn1c{%
+% \end{macrocode}
+% Within this row we calculate "\backup@length" to be the height
+% plus depth of a standard line. In addition we have to add the
+% width of the "\hline", something that was forgotten in the
+% original definition.
+% \begin{macrocode}
+ \global\backup@length\ht\@arstrutbox
+ \global\advance\backup@length\dp\@arstrutbox
+ \global\advance\backup@length\arrayrulewidth
+% \end{macrocode}
+% Finally we do want to make the height of this first line be a bit
+% larger than usual, for this we place the standard array strut
+% into it but raised by "\extratabsurround"
+% \begin{macrocode}
+ \raise\extratabsurround\copy\@arstrutbox
+% \end{macrocode}
+% Having done all this we end the line and back up by the value of
+% "\backup@length" and then finally place our "\hline". This should
+% place the line exactly at the right place but keep the reference
+% point of the whole tabular at the baseline of the first row.
+% \begin{macrocode}
+ }\\[-\backup@length]\hline
+}
+% \end{macrocode}
+% \end{macro}
+%
+% \begin{macro}{\lasthline}
+% \changes{v2.3h}{1996/05/25}{Complete reimplementation}
+% For "\lasthline" the situation is even worse and I got it
+% completely wrong initially.
+%
+% The problem in this case is that if the optional argument "[b]"
+% is used we do want the reference point of the tabular be at the
+% baseline of the last row but at the same time do want the the
+% depth of this last line increased by "\extratabsurround" without
+% changing the placement "\hline".
+%
+% We start by placing the rule followed by an invisible row.
+% \begin{macrocode}
+\newcommand{\lasthline}{\hline\multicolumn1c{%
+% \end{macrocode}
+% We now calculate "\backup@length" to be the height and depth of
+% two lines plus the width of the rule.
+% \begin{macrocode}
+ \global\backup@length2\ht\@arstrutbox
+ \global\advance\backup@length2\dp\@arstrutbox
+ \global\advance\backup@length\arrayrulewidth
+% \end{macrocode}
+% This will bring us back to the baseline of the second last row:
+% \begin{macrocode}
+ }\\[-\backup@length]%
+% \end{macrocode}
+% Thus if we now add another invisible row the reference point of
+% that row will be at the baseline of the last row (and will be the
+% reference for the whole tabular). Since this row is invisible we
+% can enlarge its depth by the desired amount.
+% \begin{macrocode}
+ \multicolumn1c{%
+ \lower\extratabsurround\copy\@arstrutbox
+ }%
+}
+% \end{macrocode}
+% \end{macro}
+%
+%
+% \subsection{Getting the spacing around rules right}
+%
+% Beside a larger functionality \texttt{array.sty} has one
+% important difference to the standard \texttt{tabular} and
+% \texttt{array} environments: horizontal and vertical rules make a
+% table larger or wider, e.g., \verb=\doublerulesep= really denotes
+% the space between two rules and isn't measured from the middle of
+% the rules.
+%
+% \begin{macro}{\@xhline}
+% For vertical rules this is implemented by the definitions above,
+% for horizontal rules we have to take out the backspace.
+% \changes{v2.3d}{1995/11/19}{fix space between double rules pr/1945}
+% \begin{macrocode}
+\CheckCommand*\@xhline{\ifx\reserved@a\hline
+ \vskip\doublerulesep
+ \vskip-\arrayrulewidth
+ \fi
+ \ifnum0=`{\fi}}
+\renewcommand*\@xhline{\ifx\reserved@a\hline
+ \vskip\doublerulesep
+ \fi
+ \ifnum0=`{\fi}}
+%</package>
+% \end{macrocode}
+% \end{macro}
+%
+% \PrintIndex
+% \PrintChanges
+%
+% \Finale
+%
+\endinput
+
diff --git a/Master/texmf-dist/doc/latex/translation-array-fr/f-array.pdf b/Master/texmf-dist/doc/latex/translation-array-fr/f-array.pdf
new file mode 100644
index 00000000000..1e07dcb69fc
--- /dev/null
+++ b/Master/texmf-dist/doc/latex/translation-array-fr/f-array.pdf
Binary files differ
diff --git a/Master/texmf-dist/doc/latex/translation-array-fr/ltxdoc.cfg b/Master/texmf-dist/doc/latex/translation-array-fr/ltxdoc.cfg
new file mode 100644
index 00000000000..6da8e4c2c68
--- /dev/null
+++ b/Master/texmf-dist/doc/latex/translation-array-fr/ltxdoc.cfg
@@ -0,0 +1,10 @@
+\ifx\traductionchanges\undefined\else
+ \AtBeginDocument{%
+ \def\HistoryName{Historique des changements}
+ \def\generalname{G\'en\'eral}
+ \GlossaryPrologue{\section*{{\HistoryName}}%
+ \markboth{{\HistoryName}}{{\HistoryName}}%
+ }}
+\fi
+\AtBeginDocument{\RecordChanges}
+\AtEndDocument{\PrintChanges}
diff --git a/Master/tlpkg/bin/tlpkg-ctan-check b/Master/tlpkg/bin/tlpkg-ctan-check
index 0262f94cdbf..192196254f3 100755
--- a/Master/tlpkg/bin/tlpkg-ctan-check
+++ b/Master/tlpkg/bin/tlpkg-ctan-check
@@ -384,6 +384,7 @@ my @TLP_working = qw(
tokenizer toolbox tools topfloat totcount totpages toptesi
tqft
trajan
+ translation-array-fr
translation-arsclassica-de translation-biblatex-de translation-chemsym-de
translation-enumitem-de translation-europecv-de translation-evc-de
translation-filecontents-de translation-moreverb-de translation-tabbing-fr
diff --git a/Master/tlpkg/libexec/ctan2tds b/Master/tlpkg/libexec/ctan2tds
index 5503eb8647e..40bc208b4cc 100755
--- a/Master/tlpkg/libexec/ctan2tds
+++ b/Master/tlpkg/libexec/ctan2tds
@@ -1276,8 +1276,9 @@ $standardtex='\.(.bx|cfg|sty|clo|ldf|cls|def|fd|cmap|4ht)$';
'tkz-orm', 'tkz.*\.sty', # no pgfmanual.sty
'tools', '\.cfg|\.sty|\.clo|\.cls|\.def|\.fd$|.?\.tex|^\.tex$',
'translator', '\.dict|translator-language-mappings.tex|' . $standardtex,
- 'translation-europecv-de','NULL', # doc package
- 'translation-tabbing-fr', 'NULL', # doc package
+ 'translation-array-fr', 'NULL', # doc
+ 'translation-europecv-de', 'NULL', # doc
+ 'translation-tabbing-fr', 'NULL', # doc
'truncate', '^..[^s].*\.sty', # not miscdoc.sty
'tugboat-plain','\.cmn|' . $standardtex,
'underscore', '^..[^s].*\.sty', # not miscdoc.sty
@@ -1663,9 +1664,10 @@ $standardsource='\.(c|drv|dtx|fea|fdd|ins|sfd)$|configure.*|install-sh';
'pgfplots', 'NULL', # manual.install.tex in doc/
'rcs', 'rcs.el|src|' . $standardsource,
'thailatex', 'NULL', # no gain in splitting up
- 'translation-chemsym-de', 'NULL', # doc package
- 'translation-filecontents-de', 'NULL', # doc package
- 'translation-tabbing-fr', 'NULL', # doc package
+ 'translation-array-fr', 'NULL', # doc
+ 'translation-chemsym-de', 'NULL', # doc
+ 'translation-filecontents-de', 'NULL', # doc
+ 'translation-tabbing-fr', 'NULL', # doc
'tkz-orm', 'NULL', # no gain ...
'xlop', 'manual.sty|' . $standardsource,
);
diff --git a/Master/tlpkg/tlpsrc/collection-documentation-french.tlpsrc b/Master/tlpkg/tlpsrc/collection-documentation-french.tlpsrc
index 05dd46428b1..a3ba1ee3d35 100644
--- a/Master/tlpkg/tlpsrc/collection-documentation-french.tlpsrc
+++ b/Master/tlpkg/tlpsrc/collection-documentation-french.tlpsrc
@@ -6,5 +6,6 @@ depend impatient-fr
depend l2tabu-french
depend lshort-french
depend texlive-fr
+depend translation-array-fr
depend translation-tabbing-fr
depend collection-documentation-base
diff --git a/Master/tlpkg/tlpsrc/translation-array-fr.tlpsrc b/Master/tlpkg/tlpsrc/translation-array-fr.tlpsrc
new file mode 100644
index 00000000000..e69de29bb2d
--- /dev/null
+++ b/Master/tlpkg/tlpsrc/translation-array-fr.tlpsrc