\input tugboat.sty

%\input tugboat.dates
%\input tb75pages
%\pageno=\TBcasa

% EXTRA CODE

\ifx\plaindefaults\undefined \let\next=\relax
\else \let\next=\plaindefaults \fi \next

\input metatex
\MTline{input delay.mf;}

\def\MTeX{{\manual META}\-\TeX}
\font\smallsf=cmssq8
\font\sf=cmss10

% END EXTRA CODE


\title * {\MTeX} *
\author * Ram\'on Casares *
\address * Telef\'onica de Espa\~na *
\netaddress * r.casares@computer.org *

\article %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%


\head * Abstract *

{\MTeX} is a set of {\plain} {\TeX} and {\MF} macros that you can use to
define both the text and the figures in a single source file. Because
{\MTeX} sets up two way communication, from {\TeX} to {\MF} and back
from {\MF} to {\TeX}, drawing dimensions can be controlled by {\TeX} and
labels can be located by {\MF}. Only standard features of {\TeX} and
{\MF} are used, but two runs of {\TeX} and one of {\MF} are needed.


\head * Overview *

Together, {\TeX} and {\MF} define the page layout to the pixel.  This
means that nothing more is needed, not even a means of including figures
in a {\TeX} document.  To prove this is the aim of this paper.

To split the typesetting process in two parts, one to define and draw
the characters and the other to arrange the characters in paragraphs and
pages, is surely the best way to reduce the complexity of the
typesetting task, provided it needs simplification (see Figure~1).  But
this method makes it difficult, for example, to integrate labels with
graphics in figures, because while {\TeX} is best suited to typeset the
labels, {\MF} is the appropriate tool to draw the graphics.  And, of
course, labels should be located in accordance with the graphics.

Therefore, the true successor of {\TeX} has to include in a single
program both the capabilities of {\TeX} and {\MF}.  Then the typesetting
engine would include a powerful graphic tool, a grid in which to typeset if
required, and it could take into account the shapes of the characters to
determine, for example, kernings or italic corrections.  The other way
around is also possible.  It could be seen as a graphic engine with a
powerful typesetting tool. From this point of view, the page would be a
graphic object that could contain paragraphs of different shapes
requested from the typesetting tool.

\MTeX, although it does not fulfill the requirements of such a
successor, can be seen as an early sign of its possibilities.  For the
moment, {\MTeX} takes advantage of {\MF}'s equation solving capabilities
to locate objects, including the labels, which are typeset by {\TeX}.  The
cost of this nice feature is that two {\TeX} passes are required.

During the first {\TeX} pass a {\MF} file is written.  As it is {\TeX}
itself who writes the {\MF} file, any dimension controlled by {\TeX} can
be used and incorporated in {\MF}'s calculations.  For example, the label
sizes, as they will be typeset by {\TeX}, are made known to {\MF}.

After the first {\TeX} pass, {\MF} draws the graphic figures and
writes the label locations in its log file.  So it is {\MF}'s
responsibility to locate the labels.  Note that, depending on the style
of {\MF} programming, this can be completely determined from {\TeX}. In
other words, you can relate the label location to the location and size
of other {\MF} objects, or not.

When {\TeX} executes its second pass, it takes the graphics
from the new font, reads the location of labels from the
{\MF} log file, and then everything is complete.

Because labels are just |\hbox|es typeset by \TeX, every macro currently
defined for text automatically applies also to figures. For example, if
a macro |\person| is defined to write its argument in a small caps font
and save it to an index file, the same happens whenever it is used
inside a figure label.


\head * Methods *

{\MTeX} allows the source file to include, in addition to the customary
{\TeX} macros to control the text, other commands to generate
figures with {\MF}.

\subhead * Steps *

In order to use {\MTeX} the following three steps are to be executed:

\list[\numbered]

\item The {\MTeX} file, suppose it is |filename.ext|, is first
processed by |TEX|, with the {\plain} format, during which a {\MF} file
named |auxiliar.mf| is created.  This {\MF} file contains information
provided by {\TeX} concerning the size of the labels, so the |MF|
program can delete this area from the figure if requested. If the output
file |filename.dvi| were typeset now, it would have blanks in place of
the figures, but otherwise be the same as the final document.

\item Then |MF|, with the {\plain} base, is run on |auxiliar.mf|.
As a result, information specifying where to typeset the labels is
written in the log file, |auxiliar.log|.  In addition, the metric file,
|auxiliar.tfm|, and the generic format bitmap font, |auxiliar.329gf|,
are created.  On my system I have to process this |gf| file to get a
|pk| file that my drivers can read, so I execute the program |GFtoPK| on
it, getting the packed bitmap font |auxiliar.329pk|. Please note three
points.  i)~The number |329|, referring to the resolution, varies
according to the {\MF} mode.  ii)~The |tfm| and |pk| files must be in or
moved to directories where programs can find them. iii)~{\MTeX} sets the
{\MF} mode to |localfont|, thus assuming that |localfont| is assigned
the appropriate name.

\item Lastly, |filename.ext| is again run through |TEX|.  During this
second run, both the font |auxiliar| containing the figures and the
information explaining where to locate the labels are available, so the
document is complete.

\endlist

The figures fill exactly the same area in both the first and second
|TEX| program runs, so indices, tables of contents, and other references
that also need two passes to be resolved can take advantage of the two
runs needed by {\MTeX}.

\subhead * Use *

To use the {\MTeX} macros, they must be imported by writing in the
source file:  ||\input metatex||

\noindent This has to be written after |\mag| has been given its final
value. When |metatex.tex| is read, {\MTeX} checks whether the file
|auxiliar.mf| exists. If it does not exist, then things are set up for
the first pass; for example, |auxiliar.mf| is opened for writing.  If it
does exist, then things are set for the second pass; for example,
|auxiliar.log| is opened for reading.  This means that if
|auxiliar.mf| is not deleted, then step~3, the second |TEX| program
pass, is executed directly.  This saves time when only the text in file
|filename.ext|, but not the figures, were modified.

\subhead * User macros *

The {\MTeX} user macros are:

\list

\item |\MTbeginchar(wd,ht,dp);| states that a figure sized as given
(width |wd|, height |ht|, depth |dp|) will be created.  These values
should be known both by {\TeX} and by {\MF}, so for example |12pt|,
|6cm|, |\the\hsize| or |\the\dimen0|, always without |#|, are allowed.
During the {\it first pass}, {\TeX} writes in |auxiliar.mf| the {\MF}
macro |beginchar| assigning character codes sequentially, and box
|\MTbox| is made empty but sized as specified by the arguments of this
macro. During the {\it second pass}, {\TeX} puts the corresponding
character of the font |auxiliar| in box |\MTbox|.  The size of |\MTbox| is
that specified and not affected by the character dimensions.

\item |\MTendchar;| finishes the figure definition.  During the
{\it first pass}, {\TeX} writes the {\MF} macro |endchar;| in file
|auxiliar.mf|. During the {\it second pass}, box |\MTbox| contains the
complete figure, including labels.  Something like |\box\MTbox| is
used to typeset the figure.

\item |\MTlabel*(s)cc"Text";| adds a label to the current figure.
The parameter between quotes, |Text| in the example, is the label
content; it will be put inside an |\hbox| and therefore could be
anything that {\TeX} allows inside an |\hbox|. The optional asterisk
after |\MTlabel| instructs {\MTeX} to erase the area of the figure
already drawn that it is under the label.

The label will be located at {\MF} point |z.s|, where |s| is the parameter
between parentheses. The reference point is further specified by the
optional parameter after the right parenthesis, |cc| in the example.
This parameter is composed of exactly two letters: the first can be |t|
meaning top, |c| meaning center or |b| meaning bottom; and the second
letter can be |l| meaning left, |c| meaning center or |r| meaning right.
So, for example, |tl| means that the label reference point is its top
left corner. The default value for the reference point is |cc|, that is,
its center.

|\MTlabel| should only be used between |\MTbeginchar| and |\MTendchar|.
During the {\it first pass}, it writes the following three elements in
|auxiliar.mf|: i)~the {\MF} macros which in turn cause |MF| to write the
label reference point location to its log file, |auxiliar.log|; ii)~the
four label sides, which are by this means made available to the
following {\MF} code for the figure, notated as |y.s.t| for the top side,
|y.s.b| for the bottom side, |x.s.l| for the left side and |x.s.r| for the right
side; and iii)~the code to delete, if requested, the figure area already
drawn that is under the rectangle occupied by the label. During the {\it
second pass}, it adds the label to the box |\MTbox| in the place that
reads from file |auxiliar.log|, making no modification to the dimensions of
|\MTbox|, even if the label is typeset outside the box.

\endlist

There are three more macros for passing information to \MF, that is, for
writing general text in |auxiliar.mf|: |\MT:|, |\MTcode| and |\MTline|.
This happens only during the first pass; during the second pass, these
macros do nothing.

\list

\item |\MT:| writes in file |auxiliar.mf| everything till the end of
line. It writes verbatim except for the character |\|, which keeps its
normal {\TeX} |\catcode| of |0|.  Spaces are {\it not\/} ignored after
macros. The sequence |\\| writes a single |\| in file |auxiliar.mf|.

\item |\MTcode| writes in file |auxiliar.mf| everything until it finds a
line equal (including |\catcode|s) to the current value of |\MTendmark|.
By default, this is a blank line, thus, |\def\MTendmark{}|.  As with
|\MT:|, it writes verbatim except for |\|, which still operates as an
escape character.  The control sequence |\\| writes a single |\| in file
|auxiliar.mf|.

\item |\MTline{text}| writes its parameter to |auxiliar.mf|,
|text| in the example.  It does not change the |\catcode|s in the
argument, so it does not perform verbatim writing.  But all {\plain}
special characters can be written prefixing them by the escape character
|\|. The {\plain} special characters are (not including the first colon
nor the final period): |\{}$&#^_~%|.  For example, |\#| results in
|#|.

\endlist

When defining {\TeX} macros that write to |auxiliar.mf|, |\MTline|
should generally be used in preference to |\MT:| or |\MTcode|, because
the latter two use
the end of line in a special way that is not usually available when
{\TeX} is reading a macro.

{\TeX} dimensions can be included using any of these three writing
macros.  For example, |\the\hsize| will be expanded to {\tt \the\hsize}
(for the present article),
and written as such to the {\MF} file |auxiliar.mf|. Note that the character
|\| keeps its escape |\catcode| in all three writing macros.  In the
case of |\MTline|, braces |{}| also keep their |\catcode|s and therefore
macros with parameters can be used normally.


%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%


\head * Examples *

\subhead * Diagram *

First a typical example of {\MTeX} usage, showing the file formats,
programs, and their relationships.  The figure width is exactly
|\hsize|, but what is more important is that the same code will adapt
itself to any value for the measure.  Well, of course, not to {\it any\/}
width but to any width between, let's say, $8\,$cm and $25\,$cm.

\figure[\caption{Figure 1: One column diagram}]
 \input diagram
\endfigure

%\twocolfigure[\caption{Figure 2: Two columns diagram}]
% \input diagram
%\endfigure

Figure~1 is the one column version, and Figure~2 (above the appendix) is
the two column version, generated by the same source.


%%%%%%%%%%%%%%%%%%%

\input shadow

\def\beginshadowpar{\setbox0=\vbox\bgroup \kern3pt
 \advance\parindent by -3pt \leftskip=3pt\rightskip=3pt }
\def\endshadowpar{\par\kern3pt\egroup\framebox0\shadowbox0\box0 }

% \subhead * Shadowing *
\bigskip % \subhead skip

\beginshadowpar\noindent
{\bf Shadowing.} Both {\TeX} and {\MF} are ill suited to creating
shadows. In {\TeX}, one straightforward technique is double use of
|\leaders|, but in practice this results in huge |dvi| files.  In {\MF},
drawing lots of tiny points easily exceeds the capacity of the program.
The solution is to coordinate the work of both programs.
\endshadowpar

To create a large rectangular shadow we divide it into an array of $n
\times m$ smaller rectangles. The smaller rectangles are all identical,
so it is enough for {\MF} to draw one shadow character and then for
{\TeX} to typeset a solid area repeating it.

To simplify the tasks of both {\TeX} and {\MF}, the size of the shadow
character should be similar to that of normal characters, because neither
program was designed to work well with extraordinarily large (or small)
characters. So a
good approach is to make the shadow character as big as possible but
never wider nor higher than 16$\,$pt.

For {\MTeX}, each figure is a character. This causes problems with {\MF}
when the figure is big and the resolution is high, because it cannot
draw areas bigger than $4095\times4095$ pixels. This is not usually a
problem working at 300$\,$dpi. (It is never a problem with {\MP}, see
the following section on {\PS}. Another advantage of using {\PS} is that
you get a shadow simply by drawing a grey rule, and none of the above
machinations are necessary.)

%%%%%%%%%%%%%%%%%%%

\subhead * Keys *

After the following {\MTeX} macros:

\verbatim[\numbered\outputtofile{keys.tex}]
\MTcode
def keybox =
 pickup pencircle scaled 0.8pt;
 x1 = x3 = 1pt;
 x2 = x4 = w - 1pt;
 x5 = 0; x6 = w;
 y1 = y2 = -d;
 y3 = y4 = h;
 y5 = y6 = (h - d)/2;
 draw z1 -- z2 .. z6{up} ..
    z4 -- z3 .. z5{down} .. cycle;
 z0 = (x1,0);
enddef;

\def\defkey#1#2{\setbox0=\hbox{\sf#2}%
 \dimen0=\wd0\advance\dimen0 by 2pt
 \dimen2=\ht0\advance\dimen2 by 1pt
 \dimen4=\dp0\advance\dimen4 by 1pt
 \MTbeginchar(\the\dimen0,%
              \the\dimen2,%
              \the\dimen4);%
  \MTline{keybox;}%
  \MTlabel(0)bl"\sf #2";%
 \MTendchar;%
 \expandafter\newbox
  \csname\string#1box\endcsname
 \expandafter\setbox
  \csname\string#1box\endcsname
  =\vtop{\unvbox\MTbox}%
 \def#1{\expandafter\copy
  \csname\string#1box\endcsname}}

\def\makekey#1{\expandafter\defkey%
 \csname#1\endcsname{#1}}
\endverbatim

\input keys

Then, we can declare |\makekey{Alt}|\makekey{Alt} to
typeset {\Alt} simply via |\Alt|.  It is also possible to
declare |\defkey\escape{\tt\char92}|\defkey\escape{\tt\char92} and then
|\escape| results in \escape.

%%%%%%%%%%%%%%%%%%%

\subhead * Baroque tables *

Baroque periods are the result of new technical achievements providing
unexplored possibilities and hence the urgent need to experiment with them,
frequently far away from what discretion might recommend. This explains
the time of baroque software that we live in, and increases the value of
\MTeX, because it provides the means to easily draw baroque tables. I am not
a baroque man, so my baroque table example is not baroque but, and this
is the point, it is at least not built with straight lines.

\MTcode
def tablebox =
 pickup pencircle scaled 1.6pt;
 x0 = 5pt; y0 = 5pt;
 draw (0,h) ... (w/2,h+y0) ... (w,h) ... (w+x0,(h-d)/2) ... (w,-d) ...
      (w/2,-d-y0) ... (0,-d) ... (-x0,(h-d)/2) ... cycle;
enddef;

\def\beginbaroquetable{\setbox0=\vbox\bgroup
 \halign\bgroup##\hfil&&\quad##\hfil\cr}
\def\endbaroquetable{\crcr\egroup\egroup
 \MTbeginchar(\the\wd0,\the\ht0,\the\dp0);%
  \MTline{ tablebox;}%
 \MTendchar;%
 $$\rlap{\box\MTbox}\box0$$}

\beginbaroquetable
 This is not a straight table\cr
 but it's only an example\cr
 and therefore not so ample\cr
 of what's {\MTeX}-able!
\endbaroquetable

%%%%%%%%%%%%%%%%%%%

\MTcode
def leftbrace =
 pickup penrazor scaled 2pt rotated 30;
 x3 = 0; x0 = x6 = w; x1 = x5 = w/5; x2 = x4 = w-x1;
 y0 = 0; y1 = h/8; y2 = h/3; y3 = h/2;
 y4 + y2 = h; y1 + y5 = h; y6 = h;
 draw z0 .. z1 .. z2 .. z3;
 draw z6 .. z5 .. z4 .. z3;
enddef;

\def\beginbracedpar{\setbox0=\vbox\bgroup
 \leftskip=20pt }
\def\endbracedpar{\strut\par\egroup
 \MTbeginchar(20pt,\the\ht0,0pt);%
  \MTline{ leftbrace;}%
 \MTendchar;%
 \hbox to\hsize{\rlap{\box\MTbox}\box0\hss}}

%\subhead * Ornate paragraphs *
\bigskip % \subhead skip

\beginbracedpar\noindent
{\bf Ornate paragraphs.} Only if you are truly baroque can you get the
most from {\MTeX}. If, for example, you like ornate paragraphs, you are
in your element. Just put the material in a |\vbox| to get the height
and the depth, and pass these dimensions to {\MF} to draw a right sized
embellishment. Ah!, but be aware that Computer Modern is a neoclassical
font, so it won't mix well with your elaborations.
\endbracedpar

%%%%%%%%%%%%%%%%%%%

% To fill unused space!
\MTcode
 def tribar(expr alpha) =
  pickup pencircle scaled 0.8pt;
  % draw (0,0)--(w,0)--(w,h)--(0,h)--cycle;
  save u, v; u = w/2; v = 12;
  z1 = (w/2,h/2) + u*(right rotated alpha);
  z2 = (w/2,h/2) + u*(right rotated (alpha+120));
  z3 = (w/2,h/2) + u*(right rotated (alpha-120));
  z1r = (w/2,h/2) + u*(right rotated (alpha+v));
  z2r = (w/2,h/2) + u*(right rotated (alpha+120+v));
  z3r = (w/2,h/2) + u*(right rotated (alpha-120+v));
  draw z1 .. z1r; draw z2 .. z2r; draw z3 .. z3r;
  draw z1 .. z3r; draw z2 .. z1r; draw z3 .. z2r;
  z1m = whatever[z1,z2r]; z1m = whatever[z1r,z3];
  z2m = whatever[z2,z3r]; z2m = whatever[z2r,z1];
  z3m = whatever[z3,z1r]; z3m = whatever[z3r,z2];
  draw z1 .. z2m; draw z2 .. z3m; draw z3 .. z1m;
  z1ra = z1 reflectedabout (z1r,z1m);
  z2ra = z2 reflectedabout (z2r,z2m);
  z3ra = z3 reflectedabout (z3r,z3m);
  draw z1m .. z1ra; draw z2m .. z2ra; draw z3m .. z3ra;
  z3x - z1x = whatever*(z3-z1m);
  z2x - z3x = whatever*(z2-z3m);
  z1x - z2x = whatever*(z1-z2m);
  z1ra - z1x = whatever*(z3-z1m);
  z3ra - z3x = whatever*(z2-z3m);
  z2ra - z2x = whatever*(z1-z2m);
  z1y = whatever[z1m,z3]; z1y = whatever[z1x,z2x];
  z2y = whatever[z2m,z1]; z2y = whatever[z2x,z3x];
  z3y = whatever[z3m,z2]; z3y = whatever[z3x,z1x];
  draw z1x .. z3y;
  draw z2x .. z1y;
  draw z3x .. z2y;
 enddef;

\def\tribar#1#2{% #1=ht, #2=angle
 \MTbeginchar(#1,#1,0pt);\MTline{tribar(#2);}\MTendchar;\box\MTbox}

\newcount\angle \angle=-36
\newdimen\tribarht \tribarht=75pt

%\subhead * Exercice *
\bigskip % \subhead skip

 \noindent
 {\bf Exercise.}\footnote*{The real exercise is to fill the rest of the
 page with tribars. A clue: let {\TeX} to calculate how many are needed,
 so you can concentrate your efforts in drawing the figure.}
 Take three equal bars and build as shown.
 Ask Roger Penrose if you don't find the solution.
 \vskip1pc

 \loop
  \vfil
  \line{\tribar{\the\tribarht}{\the\angle}\hfil \global\advance\angle5
        \tribar{\the\tribarht}{\the\angle}\hfil \global\advance\angle5
        \tribar{\the\tribarht}{\the\angle}\global\advance\angle5 }
  \dimen0=\pagegoal \advance\dimen0 by-\pagetotal
 \ifdim\dimen0>\tribarht
 \repeat


%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%


\head * {\PS} *

By taking advantage of {\MP}, we can make {\PS} versions of the
{\MTeX} files.  Simply execute |mpost &plain auxiliar.mf| instead of
{\MF} and, after {\TeX}'s second pass, execute |dvips| ({\MTeX} uses
|dvips| specials). This works because |auxiliar.mf| is valid code
for both {\MF} and {\MP}, and because, in the second {\TeX} pass,
{\MTeX} checks which one was used and adapts itself to the situation.

And thanks to {\PDF\TeX} and the Con{\TeX}t files |supp-mis.tex| and
|supp-pdf.tex|, it is also possible to get \acro{PDF} output. Just
execute {\PDF\TeX} twice, instead of {\TeX}, and once {\MP}, instead of
{\MF}. (In practice, {\MTeX} uses its own macro file |mtmp2pdf.tex| for
this, instead of the Con{\TeX}t files. I extracted all that {\MTeX}
needs from the Con{\TeX}t files into |mtmp2pdf.tex|.)

This works because, if {\MTeX} determines that {\MP} was employed to
draw the figures, it then checks which program, {\TeX} or {\PDF\TeX}, is
executing. If {\TeX}, then it includes the files produced by {\MP} using
the |dvips| specials. If {\PDF\TeX}, it translates from |ps| to |pdf|.

Therefore, the same {\MTeX} source file generates at will any of the
three output formats---|dvi|, |ps|, or |pdf|---just by running the
appropriate programs. In the appendix, as an example, there is a
\acro{DOS} batch file that shows how to get the |pdf| version of a file
|filename.ext|.


\head * Other graphical tools *

There are other tools to include pictures in a {\TeX} document.
{\LaTeX}'s |picture| environment, {\PiCTeX}, and |mfpic|, are three of
them. {\MTeX} is similar to |mfpic|, in that both use {\MF} to draw.

The aim of |mfpic| is to overcome the difficulties of {\LaTeX}'s
|picture| environment and of {\PiCTeX}. {\LaTeX}'s |picture| environment
uses four pre-cooked special fonts, and its drawings are just
compositions of these characters (as well as {\TeX}'s builtin |\hrule|s and
|\vrule|s). {\PiCTeX} only uses a tiny point to compose the pictures, so
it is more general. But letting {\TeX} to draw the figures setting point
after point is painful, as noted above. The |mfpic| solution uses a
better tool for drawing: {\MF}.

With these origins, for |mfpic|, {\MF} is a hidden back-end processor,
and |mfpic| imposes only two requirements on its users: to know {\TeX},
and to know |mfpic|. On the other hand, {\MTeX}'s approach is
minimalist, at the cost of being more demanding with its users.  A
{\MTeX} user has to know {\TeX}, {\MF}, and {\MTeX}\Dash although this
last requirement is small, because {\MTeX} only builds the necessary
bridges to use {\TeX} and {\MF} in a cooperative way.

A feature that shows the different strategies employed in designing
|mfpic| and {\MTeX} is label positioning. {\MTeX} labels are located by
{\MF}, so {\TeX} has to read the {\MF} log file to learn where to
typeset them. For |mfpic|, {\TeX} itself locates the labels,
but by doing so |mfpic| has to give up some nice {\MF} characteristics,
such as its equation solving capabilities.

% |mfpic| is to {\LaTeX} as {\MTeX} is to |plain| {\TeX}.
In summary, |mfpic|'s aim is to draw pictures in {\TeX} documents in a
better way than using {\LaTeX}'s |picture| environment or {\PiCTeX};
while {\MTeX}'s intention is to coordinate the work of {\TeX} and
{\MF}. In this way, {\MTeX} provides the full raw power of {\TeX} and
{\MF}, and it is up to you to harness them.


\head * Final remarks *

I have been using {\MTeX} for some years.  The first version was dated
1994, but it has been used only for personal purposes. For this reason,
it is not truly a straightforward end-user tool, as for example {\LaTeX} packages
should be.  It has to be used knowledgeably and with care. And though
most tasks can be automated, by chaining {\TeX} and {\MF} errors are
even more difficult to pinpoint than in {\TeX} or {\MF} alone.

Nevertheless, {\MTeX} serves to validate the feasibility of a closer
collaboration between {\TeX} and {\MF} and to appraise the interest of
such a collaboration. And, of course, if you dare, you can get lots of
fun, and at least an equal amount of frustration, using {\MTeX}. Try it!

The {\MTeX} package is available from {\CTAN} in
\acro{CTAN}|:/macros/plain/contrib/metatex|.

Happy {\MTeX}ing!


\makesignature


\newpage


\onecol % Annex in one column, can't switch to \twocol on the same page
        % with current plain tugboat macros.

\figure[\caption{Figure 2: Two-column diagram}]
\input diagram
\endfigure

\head * Appendix: Pseudo-batch example *

This is an example based on \acro{DOS} batch files that can be adapted for
other operating systems. It processes the file |filename.ext|, generating
|filename.pdf|. We will comment each line of pseudo-code.

\list[\numbered]
\item We go to the directory where our working files are.
 ||cd c:\dir\subdir\mydir||
\item We set environment variables (if necessary). In this case we are
using a |web2c| system, so it is enough to set one. In |web2c|, by
default, all programs look for files in the current dir, ``|.|'', and
that is enough for us.
 ||set TEXMFCNF=c:\tex\texmf.local\web2c;c:\tex\texmf\web2c;d:\texmf\web2c||
\item We tell the operating system where to find the programs.
 ||path=$path$;c:\tex\bin\dos||
\item After the settings, we force the first {\TeX} pass.
 ||if exist auxiliar.mf del auxiliar.mf||
\item Then, we execute the first {\TeX} pass (in this case, it is {\PDF\TeX}).
 ||pdftex &plain filename.ext||
\item If {\MTeX} was not used, and no |auxiliar.mf| was written, then we
are done.
 ||if not exist auxiliar.mf goto end||
\item Otherwise we run {\MP} with its |&plain| memory (format),
also known as |&mpost|.
 ||mpost &plain auxiliar.mf||
\item Finally, we execute the second {\TeX} pass.
 ||pdftex &plain filename.ext||
\item We now have the complete |filename.pdf| file.
 ||:end||

\endlist

\MTline{end.} % METATeX changes \bye
\endarticle


% To check labels (ink intensive example!)
\MTbeginchar(5cm,5cm,0cm);
\MT: fill (0,0)--(w,0)--(w,h)--(0,h)--cycle;
\MT: z0 = (w/2,h/2);
\MT: z1 = (w/2,3h/4); z11 = z1;
\MT: z2 = (w/2,h/4);
\MT: z3 = (0,h/2);
\MT: z4 = (0,h/4);
\MTlabel*(0)"Center";
\MTlabel*(1)tl".";
\MTlabel*(11)br".";
\MTlabel*(2)":";
\MTlabel(3)"Out";
\MTlabel*(4)"Intruder";
\MTendchar;
$$\box\MTbox$$