fpc/docs/packages/mdwtools/mdwmath.dtx
1998-09-21 10:15:33 +00:00

934 lines
29 KiB
TeX

% \begin{meta-comment}
%
% $Id$
%
% Various nicer mathematical things
%
% (c) 1996 Mark Wooding
%
%----- Revision history -----------------------------------------------------
%
% $Log$
% Revision 1.1 1998-09-21 10:19:01 michael
% Initial implementation
%
% Revision 1.1 1996/11/19 20:53:21 mdw
% Initial revision
%
%
% \end{meta-comment}
%
% \begin{meta-comment} <general public licence>
%%
%% mdwmath package -- various nicer mathematical things
%% Copyright (c) 1996 Mark Wooding
%%
%% This program is free software; you can redistribute it and/or modify
%% it under the terms of the GNU General Public License as published by
%% the Free Software Foundation; either version 2 of the License, or
%% (at your option) any later version.
%%
%% This program is distributed in the hope that it will be useful,
%% but WITHOUT ANY WARRANTY; without even the implied warranty of
%% MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
%% GNU General Public License for more details.
%%
%% You should have received a copy of the GNU General Public License
%% along with this program; if not, write to the Free Software
%% Foundation, Inc., 675 Mass Ave, Cambridge, MA 02139, USA.
%%
% \end{meta-comment}
%
% \begin{meta-comment} <Package preamble>
%<+package>\NeedsTeXFormat{LaTeX2e}
%<+package>\ProvidesPackage{mdwmath}
%<+package> [1996/04/11 1.1 Nice mathematical things]
%<+oldeqnarray>\NeedsTeXFormat{LaTeX2e}
%<+oldeqnarray>\ProvidesPackage{eqnarray}
%<+oldeqnarray> [1996/04/11 1.1 Old enhanced eqnarray]
% \end{meta-comment}
%
% \CheckSum{259}
% \begin{old-eqnarray}
% \CheckSum{484}
% \end{old-eqnarray}
%% \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 \~}
%%
%
% \begin{meta-comment}
%
%<*driver>
\input{mdwtools}
\describespackage{mdwmath}
% \describespackage{eqnarray}
\ignoreenv{old-eqnarray}
% \unignoreenv{old-eqnarray}
\mdwdoc
%</driver>
%
% \end{meta-comment}
%
% \section{User guide}
%
% \subsection{Square root typesetting}
%
% \DescribeMacro{\sqrt}
% The package supplies a star variant of the |\sqrt| command which omits the
% vinculum over the operand (the line over the top). While this is most
% useful in simple cases like $\sqrt*{2}$ it works for any size of operand.
% The package also re-implements the standard square root command so that it
% positions the root number rather better.
%
% \begin{figure}
% \begin{demo}[w]{Examples of the new square root command}
%\[ \sqrt*{2} \quad \mbox{rather than} \quad \sqrt{2} \]
%\[ \sqrt*[3]{2} \quad \mbox{ rather than } \quad \sqrt[3]{2} \]
%\[ \sqrt{x^3 + \sqrt*[y]{\alpha}} - \sqrt*[n+1]{a} \]
%\[ x = \sqrt*[3]{\frac{3y}{7}} \]
%\[ q = \frac{2\sqrt*{2}}{5}+\sqrt[\frac{n+1}{2}]{2x^2+3xy-y^2} \]
% \end{demo}
% \end{figure}
%
% [Note that omission of the vinculum was originally a cost-cutting exercise
% because the radical symbol can just fit in next to its operand and
% everything ends up being laid out along a line. However, I find that the
% square root without vinculum is less cluttered, so I tend to use it when
% it doesn't cause ambiguity.]
%
% \subsection{Some maths symbols you already have}
%
% Having just tried to do some simple things, I've found that there are maths
% symbols missing. Here they are, in all their glory:
%
% \begin{center} \unverb\| \begin{tabular}{cl|cl|cl}
% $\&$ & "\&" & $\bitor$ & "\bitor" & $\dbland$ & "\dbland" \\
% $\bitand$ & "\bitand" & $\dblor$ & "\dblor" &
% \end{tabular} \end{center}
%
% \begin{ignore}
% There used to be an eqnarray here, but that's migrated its way into the
% \package{mdwtab} package. Maybe the original version, without dependency
% on \package{mdwtab} ought to be releasable separately. I'll keep it around
% just in case.
%
% The following is the documentation for the original version. There's an
% updated edition in \package{mdwtab}.
% \end{ignore}
%
% \begin{old-eqnarray}
%
% \subsection{A new \env{eqnarray} environment}
%
% \LaTeX's built-in \env{eqnarray} is horrible -- it puts far too much space
% between the items in the array. This environment is rather nearer to the
% \env{amsmath} \env{align} environments, although rather less capable.
%
% \bigskip
% \DescribeEnv{eqnarray}
% {\synshorts
% \setbox0\hbox{"\\begin{eqnarray}["<preamble>"]" \dots "\\end{eqnarray}"}
% \leavevmode \hskip-\parindent \fbox{\box0}
% }
% \smallskip
%
% The new version of \env{eqnarray} tries to do everything which you really
% want it to. The \synt{preamble} string allows you to define the column
% types in a vaguely similar way to the wonderful \env{tabular} environment.
% The types provided (and it's easy-ish to add more) are:
%
% \def\ch{\char`}
% \begin{description} \def\makelabel{\hskip\labelsep\normalfont\ttfamily}
% \item [r] Right aligned equation
% \item [c] Centre-aligned equation
% \item [l] Left aligned equation
% \item [\textrm{\texttt{Tr}, \texttt{Tc} and \texttt{Tl}}] Right, centre and
% left aligned text (not maths)
% \item [L] Left aligned zero-width equation
% \item [x] Centred entire equation
% \item [:] Big gap separating sets of equations
% \item [q] Quad space
% \item [>\ch\{\synt{text}\ch\}] Insert text before column
% \item [<\ch\{\synt{text}\ch\}] Insert text after column
% \end{description}
%
% Some others are also defined: don't use them because they do complicated
% things which are hard to explain and they aren't much use anyway.
%
% The default preamble, if you don't supply one of your own, is \lit{rcl}.
% Most of the time, \lit{rl} is sufficient, although compatibility is more
% important to me.
%
% By default, there is no space between columns, which makes formul\ae\ in an
% \env{eqnarray} environment look just like formul\ae\ typeset on their own,
% except that things get aligned in columns. This is where the default
% \env{eqnarray} falls down: it leaves |\arraycolsep| space between each
% column making the thing look horrible.
%
% An example would be good here, I think. This one's from exercise 22.9 of
% the \textit{\TeX book}.
%
% \begin{demo}[w]{Simultaneous equations}
%\begin{eqnarray}[rcrcrcrl]
% 10w & + & 3x & + & 3y & + & 18z & = 1 \\
% 6w & - & 17x & & & - & 5z & = 2
%\end{eqnarray}
% \end{demo}
%
% Choosing a more up-to-date example, here's one demonstrating the \lit{:}
% column specifier from the \textit{\LaTeX\ Companion}.
%
% \begin{demo}[w]{Lots of equations}
%\begin{eqnarray}[rl:rl:l]
% V_i &= v_i - q_i v_j, & X_i &= x_i - q_i x_j, &
% U_i = u_i, \qquad \mbox{for $i \ne j$} \label{eq:A} \\
% V_j &= v_j, & X_j &= x_j &
% U_j u_j + \sum_{i \ne j} q_i u_i.
%\end{eqnarray}
% \end{demo}
%
% We can make things more interesting by adding a plain text column. Here we
% go:
%
% \begin{demo}[w]{Plain text column}
%\begin{eqnarray}[rlqqTl]
% x &= y & by (\ref{eq:A}) \\
% x' &= y' & by definition \\
% x + x' &= y + y' & by Axiom~1
%\end{eqnarray}
% \end{demo}
%
% The new features also mean that you don't need to mess about with
% |\lefteqn| any more. This is handled by the \lit{L} column type:
%
% \begin{demo}{Splitting example}
%\begin{eqnarray*}[Ll]
% w+x+y+z = \\
% & a+b+c+d+e+ \\
% & f+g+h+i+j
%\end{eqnarray*}
% \end{demo}
%
% Finally, just to prove that the spacing's right at last, here's another one
% from the \textit{Companion}.
%
% \begin{demo}{Spacing demonstration}
%\begin{equation}
% x^2 + y^2 = z^2
%\end{equation}
%\begin{eqnarray}[rl]
% x^2 + y^2 &= z^2 \\
% y^2 &< z^2
%\end{eqnarray}
% \end{demo}
%
% Well, that was easy enough. Now on to numbering. As you've noticed, the
% equations above are numbered. You can use the \env{eqnarray$*$}
% environment to turn off the numbering in the whole environment, or say
% |\nonumber| on a line to suppress numbering of that one in particular.
% More excitingly, you can say \syntax{"\\nonumber["<text>"]"} to choose
% what text to display.
%
% A note for cheats: you can use the sparkly new \env{eqnarray} for simple
% equations simply by specifying \lit{x} as the column description. Who
% needs \AmSTeX? |;-)|
%
% \end{old-eqnarray}
%
% \implementation
%
% \section{Implementation}
%
% This isn't really complicated (honest) although it is a lot hairier than I
% think it ought to be.
%
% \begin{macrocode}
%<*package>
% \end{macrocode}
%
% \subsection{Square roots}
%
% \subsubsection{Where is the square root sign?}
%
% \LaTeX\ hides the square root sign away somewhere without telling anyone
% where it is. I extract it forcibly by peeking inside the |\sqrtsign| macro
% and scrutinising the contents. Here we go: prepare for yukkiness.
%
% \begin{macrocode}
\newcount\sq@sqrt
\begingroup
\catcode`\|0 \catcode`\\12
|def|sq@readrad#1"#2\#3|relax{|global|sq@sqrt"#2|relax}
|expandafter|sq@readrad|meaning|sqrtsign|relax
|endgroup
\def\sq@delim{\delimiter\sq@sqrt\relax}
% \end{macrocode}
%
% \subsubsection{Drawing fake square root signs}
%
% \TeX\ absolutely insists on drawing square root signs with a vinculum over
% the top. In order to get the same effect, we have to attempt to emulate
% \TeX's behaviour.
%
% \begin{macro}{\sqrtdel}
%
% This does the main job of typesetting a vinculum-free radical.\footnote{^^A
% Note for chemists: this is nothing to do with short-lived things which
% don't have their normal numbers of electrons. And it won't reduce the
% appearance of wrinkles either.}
% It's more or less a duplicate of what \TeX\ does internally, so it might be
% a good plan to have a copy of Appendix~G open while you examine this.
%
% We start off by using |\mathpalette| to help decide how big things should
% be.
%
% \begin{macrocode}
\def\sqrtdel{\mathpalette\sqrtdel@i}
% \end{macrocode}
%
% Read the contents of the radical into a box, so we can measure it.
%
% \begin{macrocode}
\def\sqrtdel@i#1#2{%
\setbox\z@\hbox{$\m@th#1#2$}% %%% Bzzzt -- uncramps the mathstyle
% \end{macrocode}
%
% Now try and sort out the values needed in this calculation. We'll assume
% that $\xi_8$ is 0.6\,pt, the way it usually is. Next try to work out the
% value of $\varphi$.
%
% \begin{macrocode}
\ifx#1\displaystyle%
\@tempdima1ex%
\else%
\@tempdima.6\p@%
\fi%
% \end{macrocode}
%
% That was easy. Now for $\psi$.
%
% \begin{macrocode}
\@tempdimb.6\p@%
\advance\@tempdimb.25\@tempdima%
% \end{macrocode}
%
% Build the `delimiter' in a box of height $h(x)+d(x)+\psi+\xi_8$, as
% requested. Box~2 will do well for this purpose.
%
% \begin{macrocode}
\dimen@.6\p@%
\advance\dimen@\@tempdimb%
\advance\dimen@\ht\z@%
\advance\dimen@\dp\z@%
\setbox\tw@\hbox{%
$\left\sq@delim\vcenter to\dimen@{}\right.\n@space$%
}%
% \end{macrocode}
%
% Now we need to do some more calculating (don't you hate it?). As far as
% Appendix~G is concerned, $\theta=h(y)=0$, because we want no rule over the
% top.
%
% \begin{macrocode}
\@tempdima\ht\tw@%
\advance\@tempdima\dp\tw@%
\advance\@tempdima-\ht\z@%
\advance\@tempdima-\dp\z@%
\ifdim\@tempdima>\@tempdimb%
\advance\@tempdima\@tempdimb%
\@tempdimb.5\@tempdima%
\fi%
% \end{macrocode}
%
% Work out how high to raise the radical symbol. Remember that Appendix~G
% thinks that the box has a very small height, although this is untrue here.
%
% \begin{macrocode}
\@tempdima\ht\z@%
\advance\@tempdima\@tempdimb%
\advance\@tempdima-\ht\tw@%
% \end{macrocode}
%
% Build the output (finally). The brace group is there to turn the output
% into a mathord, one of the few times that this is actually desirable.
%
% \begin{macrocode}
{\raise\@tempdima\box\tw@\vbox{\kern\@tempdimb\box\z@}}%
}
% \end{macrocode}
%
% \end{macro}
%
% \subsubsection{The new square root command}
%
% This is where we reimplement all the square root stuff. Most of this stuff
% comes from the \PlainTeX\ macros, although some is influenced by \AmSTeX\
% and \LaTeXe, and some is original. I've tried to make the spacing vaguely
% automatic, so although it's not configurable like \AmSTeX's version, the
% output should look nice more of the time. Maybe.
%
% \begin{macro}{\sqrt}
%
% \LaTeX\ says this must be robust, so we make it robust. The first thing to
% do is to see if there's a star and pass the appropriate squareroot-drawing
% command on to the rest of the code.
%
% \begin{macrocode}
\DeclareRobustCommand\sqrt{\@ifstar{\sqrt@i\sqrtdel}{\sqrt@i\sqrtsign}}
% \end{macrocode}
%
% Now we can sort out an optional argument to be displayed on the root.
%
% \begin{macrocode}
\def\sqrt@i#1{\@ifnextchar[{\sqrt@ii{#1}}{\sqrt@iv{#1}}}
% \end{macrocode}
%
% Stages~2 and~3 below are essentially equivalents of \PlainTeX's
% |\root|\dots|\of| and |\r@@t|. Here we also find the first wrinkle: the
% |\rootbox| used to store the number is spaced out on the left if necessary.
% There's a backspace after the end so that the root can slip underneath, and
% everything works out nicely. Unfortunately size is fixed here, although
% doesn't actually seem to matter.
%
% \begin{macrocode}
\def\sqrt@ii#1[#2]{%
\setbox\rootbox\hbox{$\m@th\scriptscriptstyle{#2}$}%
\ifdim\wd\rootbox<6\p@%
\setbox\rootbox\hb@xt@6\p@{\hfil\unhbox\rootbox}%
\fi%
\mathpalette{\sqrt@iii{#1}}%
}
% \end{macrocode}
%
% Now we can actually build everything. Note that the root is raised by its
% depth -- this prevents a common problem with letters with descenders.
%
% \begin{macrocode}
\def\sqrt@iii#1#2#3{%
\setbox\z@\hbox{$\m@th#2#1{#3}$}%
\dimen@\ht\z@%
\advance\dimen@-\dp\z@%
\dimen@.6\dimen@%
\advance\dimen@\dp\rootbox%
\mkern-3mu%
\raise\dimen@\copy\rootbox%
\mkern-10mu%
\box\z@%
}
% \end{macrocode}
%
% Finally handle a non-numbered root. We read the rooted text in as an
% argument, to stop problems when people omit the braces. (\AmSTeX\ does
% this too.)
%
% \begin{macrocode}
\def\sqrt@iv#1#2{#1{#2}}
% \end{macrocode}
%
% \end{macro}
%
% \begin{macro}{\root}
%
% We also re-implement \PlainTeX's |\root| command, just in case someone uses
% it, and supply a star-variant. This is all very trivial.
%
% \begin{macrocode}
\def\root{\@ifstar{\root@i\sqrtdel}{\root@i\sqrtsign}}
\def\root@i#1#2\of{\sqrt@ii{#1}[#2]}
% \end{macrocode}
%
% \end{macro}
%
% \subsection{Some magic new maths characters}
%
% This is all really easy.
%
% \begin{macrocode}
\DeclareMathSymbol{&}{\mathbin}{operators}{`\&}
\DeclareMathSymbol{\bitand}{\mathbin}{operators}{`\&}
\def\bitor{\mathbin\mid}
\def\dblor{\mathbin{\mid\mid}}
\def\dbland{\mathbin{\mathrel\bitand\mathrel\bitand}}
% \end{macrocode}
%
% \subsection{Biggles}
%
% Now for some user-controlled delimiter sizing. The standard bigness of
% plain \TeX's delimiters are all right, but it's a little limiting.
%
% The biggness of delimiters is based on the size of the current |\strut|,
% which \LaTeX\ keeps up to date all the time. This will make the various
% delimiters grow in proportion when the text gets bigger. Actually, I'm
% not sure that this is exactly right -- maybe it should be nonlinear,
%
% \begin{macro}{\bbigg}
% \begin{macro}{\bbiggl}
% \begin{macro}{\bbiggr}
% \begin{macro}{\bbiggm}
%
% This is where the bigness is done. This is more similar to the plain \TeX\
% big delimiter stuff than to the \package{amsmath} stuff, although there's
% not really a lot of difference.
%
% The two arguments are a multiplier for the delimiter size, and a small
% increment applied \emph{before} the multiplication (which is optional).
%
% This is actually a front for a low-level interface which can be called
% directly for efficiency.
%
% \begin{macrocode}
\def\bbigg{\@bbigg\mathord}
\def\bbiggl{\@bbigg\mathopen}
\def\bbiggr{\@bbigg\mathclose}
\def\bbiggm{\@bbigg\mathrel}
% \end{macrocode}
%
% \end{macro}
% \end{macro}
% \end{macro}
% \end{macro}
%
% \begin{macro}{\@bbigg}
%
% This is an optional argument parser providing a front end for the main
% macro |\bbigg@|.
%
% \begin{macrocode}
\def\@bbigg#1{\@ifnextchar[{\@bigg@i{#1}}{\@bigg@i{#1}[\z@]}}
\def\@bigg@i#1[#2]#3#4{#1{\bbigg@{#2}{#3}{#4}}}
% \end{macrocode}
%
% \end{macro}
%
% \begin{macro}{\bbigg@}
%
% This is it, at last. The arguments are as described above: an addition
% to be made to the strut height, and a multiplier. Oh, and the delimiter,
% of course.
%
% This is a bit messy. The smallest `big' delimiter, |\big|, is the same
% height as the current strut box. Other delimiters are~$1\frac12$, $2$
% and~$2\frac12$ times this height. I'll set the height of the delimiter by
% putting in a |\vcenter| of the appropriate size.
%
% Given an extra height~$x$, a multiplication factor~$f$ and a strut
% height~$h$ and depth~$d$, I'll create a vcenter with total height
% $f(h+d+x)$. Easy, isn't it?
%
% \begin{macrocode}
\def\bbigg@#1#2#3{%
\hbox{$%
\dimen@\ht\strutbox\advance\dimen@\dp\strutbox%
\advance\dimen@#1%
\dimen@#2\dimen@%
\left#3\vcenter to\dimen@{}\right.\n@space%
$}%
}
% \end{macrocode}
%
% \end{macro}
%
% \begin{macro}{\big}
% \begin{macro}{\Big}
% \begin{macro}{\bigg}
% \begin{macro}{\Bigg}
%
% Now for the easy macros.
%
% \begin{macrocode}
\def\big{\bbigg@\z@\@ne}
\def\Big{\bbigg@\z@{1.5}}
\def\bigg{\bbigg@\z@\tw@}
\def\Bigg{\bbigg@\z@{2.5}}
% \end{macrocode}
%
% \end{macro}
% \end{macro}
% \end{macro}
% \end{macro}
%
%
% \begin{ignore}
% The following is the original definition of the enhanced eqnarray
% environment. It's not supported, although if you can figure out how to
% extract it, it's all yours.
% \end{ignore}
%
% \begin{old-eqnarray}
%
% \subsection{The sparkly new \env{eqnarray}}
%
% Start off by writing a different package.
%
% \begin{macrocode}
%</package>
%<*oldeqnarray>
% \end{macrocode}
%
% \subsubsection{Options handling}
%
% We need to be able to cope with \textsf{fleqn} and \textsf{leqno} options.
% This will adjust our magic modified \env{eqnarray} environment
% appropriately.
%
% \begin{macrocode}
\newif\if@fleqn
\newif\if@leqno
\DeclareOption{fleqn}{\@fleqntrue}
\DeclareOption{leqno}{\@leqnotrue}
\ProcessOptions
% \end{macrocode}
%
% This is all really different to the \LaTeX\ version. I've looked at the
% various \env{tabular} implementations, the original \env{eqnarray} and the
% \textit{\TeX book} to see how best to do this, and then went my own way.
% If it doesn't work it's all my fault.
%
% \subsubsection{Some useful registers}
%
% The old \LaTeX\ version puts the equation numbers in by keeping a count of
% where it is in the alignment. Since I don't know how may columns there are
% going to be, I'll just use a switch in the preamble to tell me to stop
% tabbing.
%
% \begin{macrocode}
\newif\if@eqalast
% \end{macrocode}
%
% Now define some useful length parameters. First allocate them:
%
% \begin{macrocode}
\newskip\eqaopenskip
\newskip\eqacloseskip
\newskip\eqacolskip
\newskip\eqainskip
% \end{macrocode}
%
% Now assign some default values. Users can play with these if they really
% want although I can't see the point myself.
%
% \begin{macrocode}
\if@fleqn
\AtBeginDocument{\eqaopenskip\leftmargini}
\else
\eqaopenskip\@centering
\fi
\eqacloseskip\@centering
\eqacolskip\@centering
\eqainskip\z@
% \end{macrocode}
%
% We allow the user to play with the style if this is really wanted. I dunno
% why, really. Maybe someone wants very small alignments.
%
% \begin{macrocode}
\let\eqa@style\displaystyle
% \end{macrocode}
%
% \subsubsection{The main environments}
%
% We define the toplevel commands here. They just add in default arguments
% and then call |\@eqnarray| with a preamble string. The only difference is
% the last column they add in -- \env{eqnarray$*$} throws away the last
% column by sticking it in box~0. (I used to |\@gobble| it but that caused
% the |\cr| to be lost.)
%
% \begin{macrocode}
\def\eqnarray{\@ifnextchar[\eqnarray@i{\eqnarray@i[rcl]}}
\def\eqnarray@i[#1]{%
\@eqnarray{#1!{\hb@xt@\z@{\hss##}\tabskip\z@}}
}
\@namedef{eqnarray*}{\@ifnextchar[\eqnarray@s@i{\eqnarray@s@i[rcl]}}
\def\eqnarray@s@i[#1]{%
\@eqnarray{#1!{\nonumber\setbox\z@\hbox{##}\tabskip\z@}}%
}
% \end{macrocode}
%
% \subsubsection{Set up the initial display}
%
% \begin{macro}{\@eqnarray}
%
% The |\@eqnarray| command does most of the initial work. It sets up some
% flags and things, builds the |\halign| preamble, and returns.
%
% \begin{macrocode}
\def\@eqnarray#1{%
% \end{macrocode}
%
% Start playing with the counter here. The original does some icky internal
% playing, which isn't necessary. The |\if@eqnsw| switch is |true| if the
% user hasn't supplied an equation number. The |\if@eqalast| switch is
% |true| in the final equation-number column.
%
% \begin{macrocode}
\refstepcounter{equation}%
\@eqalastfalse%
\global\@eqnswtrue%
\m@th%
% \end{macrocode}
%
% Set things up for the |\halign| which is coming up.
%
% \begin{macrocode}
\openup\jot%
\tabskip\eqaopenskip%
\let\\\@eqncr%
\everycr{}%
$$%
% \end{macrocode}
%
% We'll build the real |\halign| and preamble in a token register. All we
% need to do is stuff the header in the token register, clear a switch
% (that'll be explained later), parse the preamble and then expand the
% tokens we collected. Easy, no?
%
% \begin{macrocode}
\toks@{\halign to\displaywidth\bgroup}%
\@tempswafalse%
\eqa@preamble#1\end%
\the\toks@\cr%
}
% \end{macrocode}
%
% \end{macro}
%
% \subsubsection{Parsing the preamble}
%
% All this actually involves is reading the next character and building a
% command from it. That can pull off an argument if it needs it. Just make
% sure we don't fall off the end and we'll be OK.
%
% \begin{macrocode}
\def\eqa@preamble#1{%
\ifx\end#1\else\csname eqa@char@#1\expandafter\endcsname\fi%
}
% \end{macrocode}
%
% Adding stuff to the preamble tokens is a simple matter of using
% |\expandafter| in the correct way.\footnote{^^A
% I have no idea why \LaTeX\ uses \cmd\edef\ for building its preamble. It
% seems utterly insane to me -- the amount of bodgery that \env{tabular}
% has to go through to make everything expand at the appropriate times is
% scary. Maybe Messrs~Lamport and Mittelbach just forgot about token
% registers when they were writing the code. Maybe I ought to rewrite the
% thing properly some time. Sigh.
%
% As a sort of postscript to the above, I \emph{have} rewritten the
% \env{tabular} environment, and made a damned fine job of it, in my
% oh-so-humble opinion. All this \env{eqnarray} stuff has been remoulded
% in terms of the generic column-defining things in \package{mdwtab}.
% You're reading the documentation of the old version, which isn't
% supported any more, so any bugs here are your own problem.}
%
% \begin{macrocode}
\def\eqa@addraw#1{\expandafter\toks@\expandafter{\the\toks@#1}}
% \end{macrocode}
%
% Now for some cleverness again. In order to put all the right bits of
% |\tabskip| glue in the right places we must \emph{not} terminate each
% column until we know what the next one is. We set |\if@tempswa| to be
% |true| if there's a column waiting to be closed (so it's initially
% |false|). The following macro adds a column correctly, assuming we're in
% a formula. Other column types make their own arrangements.
%
% \begin{macrocode}
\def\eqa@add#1{%
\if@tempswa%
\eqa@addraw{\tabskip\eqainskip&#1}%
\else%
\eqa@addraw{#1}%
\fi%
\@tempswatrue%
}
% \end{macrocode}
%
% Now to defining column types. Let's define a macro which allows us to
% define column types:
%
% \begin{macrocode}
\def\eqa@def#1{\expandafter\def\csname eqa@char@#1\endcsname}
% \end{macrocode}
%
% Now we can define the column types. Each column type must loop back to
% |\eqa@preamble| once it's finished, to read the rest of the preamble
% string. Note the positioning of ord atoms in the stuff below. This will
% space out relations and binops correctly when they occur at the edges of
% columns, and won't affect ord atoms at the edges, because ords pack
% closely.
%
% First the easy onces. Just stick |\hfil| in the right places and
% everything will be all right.
%
% \begin{macrocode}
\eqa@def r{\eqa@add{\hfil$\eqa@style##{}$}\eqa@preamble}
\eqa@def c{\eqa@add{\hfil$\eqa@style{}##{}$\hfil}\eqa@preamble}
\eqa@def l{\eqa@add{$\eqa@style{}##$\hfil}\eqa@preamble}
\eqa@def x{\eqa@add{\hfil$\eqa@style##$\hfil}\eqa@preamble}
% \end{macrocode}
%
% Now for the textual ones. This is also fairly easy.
%
% \begin{macrocode}
\eqa@def T#1{%
\eqa@add{}%
\if#1l\else\eqa@addraw{\hfil}\fi%
\eqa@addraw{##}%
\if#1r\else\eqa@addraw{\hfil}\fi%
\eqa@preamble%
}
% \end{macrocode}
%
% Sort of split types of equations. I mustn't use |\rlap| here, or
% everything goes wrong -- |\\| doesn't get noticed by \TeX\ in the same way
% as |\cr| does.
%
% \begin{macrocode}
\eqa@def L{\eqa@add{\hb@xt@\z@{$\eqa@style##$\hss}\qquad}\eqa@preamble}
% \end{macrocode}
%
% The \lit{:} column type is fairly simple. We set |\tabskip| up to make
% lots of space and close the current column, because there must be one.^^A
% \footnote{This is an assumption.}
%
% \begin{macrocode}
\eqa@def :{%
\eqa@addraw{\tabskip\eqacolskip&}\@tempswafalse\eqa@preamble%
}
\eqa@def q{\eqa@add{\quad}\@tempswafalse\eqa@preamble}
% \end{macrocode}
%
% The other column types just insert given text in an appropriate way.
%
% \begin{macrocode}
\eqa@def >#1{\eqa@add{#1}\@tempswafalse\eqa@preamble}
\eqa@def <#1{\eqa@addraw{#1}\eqa@preamble}
% \end{macrocode}
%
% Finally, the magical \lit{!} column type, which sets the equation number.
% We set up the |\tabskip| glue properly, tab on, and set the flag which
% marks the final column.
%
% \begin{macrocode}
\eqa@def !#1{%
\eqa@addraw{\tabskip\eqacloseskip&\@eqalasttrue#1}\eqa@preamble%
}
% \end{macrocode}
%
% \subsubsection{Newline codes}
%
% Newline sequences (|\\|) get turned into calls of |\@eqncr|. The job is
% fairly simple, really. However, to avoid reading `|&|' characters
% prematurely, we set up a magic brace (from the \package{array} package --
% this avoids creating ord atoms and other nastyness).
%
% \begin{macrocode}
\def\@eqncr{%
\iffalse{\fi\ifnum0=`}\fi%
\@ifstar{\eqacr@i{\@M}}{\eqacr@i{\interdisplaylinepenalty}}%
}
\def\eqacr@i#1{\@ifnextchar[{\eqacr@ii{#1}}{\eqacr@ii{#1}[\z@]}}
\def\eqacr@ii#1[#2]{%
\ifnum0=`{}\fi%
\eqa@eqnum%
\noalign{\penalty#1\vskip#2\relax}%
}
% \end{macrocode}
%
% \subsubsection{Setting equation numbers}
%
% Before we start, we need to generalise the flush-left number handling bits.
% The macro |\eqa@eqpos| will put its argument in the right place.
%
% \begin{macrocode}
\if@leqno
\def\eqa@eqpos#1{%
\hb@xt@.01\p@{}\rlap{\normalfont\normalcolor\hskip-\displaywidth#1}%
}
\else
\def\eqa@eqpos#1{\normalfont\normalcolor#1}
\fi
% \end{macrocode}
%
% First we need to move into the right column. Then we just set the equation
% number appropriately. There is some subtlety here, ish. The |\relax| is
% important, to delay expansion of the |\if|\dots\ until the new column has
% been started. The two helper macros are important too, to hide `|&|'s and
% `|\cr|'s from \TeX's scanner until the right time.
%
% \begin{macrocode}
\def\eqa@eqnum{%
\relax%
\if@eqalast\expandafter\eqa@eqnum@i\else\expandafter\eqa@eqnum@ii\fi%
}
\def\eqa@eqnum@i{%
\if@eqnsw%
\eqa@eqpos{(\theequation)}\stepcounter{equation}%
\else%
\eqa@eqpos\eqa@number%
\fi%
\global\@eqnswtrue%
\cr%
}
\def\eqa@eqnum@ii{&\eqa@eqnum}
% \end{macrocode}
%
% \subsubsection{Numbering control}
%
% This is trivial. We set the |\if@eqnsw| flag to be |false| and store the
% text in a macro.
%
% \begin{macrocode}
\let\nonumber\relax
\newcommand\nonumber[1][]{\global\@eqnswfalse\global\def\eqa@number{#1}}
% \end{macrocode}
%
% \subsubsection{Closing the environments off}
%
% This is really easy. Set the final equation number, close the |\halign|,
% tidy up the equation counter (it's been stepped once too many times) and
% close the display.
%
% \begin{macrocode}
\def\endeqnarray{%
\eqa@eqnum%
\egroup%
\global\advance\c@equation\m@ne%
$$%
\global\@ignoretrue%
}
\expandafter\let\csname endeqnarray*\endcsname\endeqnarray
% \end{macrocode}
%
% Now start up the other package again.
%
% \begin{macrocode}
%</oldeqnarray>
%<*package>
% \end{macrocode}
%
% \end{old-eqnarray}
%
% That's all there is. Byebye.
%
% \begin{macrocode}
%</package>
% \end{macrocode}
%
% \hfill Mark Wooding, \today
%
% \Finale
\endinput