texk/web2c/dvicopy.web

% Copyright (C) 1990--2014 Peter Breitenlohner (peb@@mppmu.mpg.de)
%
% 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 1, or (at your option)
% any later version.
%
% 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.
%
% Version 0.9 was finished May 21, 1990.
% Version 1.0 pixel rounding for real devices (August 6, 1990).
% Version 1.1 major rearrangements for DVIprint (October 7, 1990).
% Version 1.2 fixed some bugs, page selection (February 13, 1991).
% Version 1.3 several more changes, command line options,
%             don't load fonts that are never used (August 25, 1992).
% Version 1.4 fixed a typo (March 28, 1995).
% Version 1.5 avoided cur_name_length identifier conflict (October 15, 1995).
% Version 1.6 minor cleanup: avoid unused or uninitialized variables,
%             diagnose impossible cases (September 2009).
%             bug fix (not for Web2C) and some typos (May 2014)
%               from Udo Wermuth (u.wermuth@@icloud.com).

% Here is TeX material that gets inserted after \input webmac
\def\hang{\hangindent 3em\indent\ignorespaces}
\font\ninerm=cmr9
\let\mc=\ninerm % medium caps for names like SAIL
\def\PASCAL{Pascal}
\font\logo=manfnt % font used for the METAFONT logo
\def\MF{{\logo META}\-{\logo FONT}}
\mathchardef\RA="3221 % right arrow

\def\(#1){} % this is used to make section names sort themselves better
\def\9#1{} % this is used for sort keys in the index

\def\title{DVI\lowercase{copy}} % don't change this line!
\def\contentspagenumber{1}
\def\topofcontents{\null
  \def\titlepage{F} % include headline on the contents page
  \def\rheader{\mainfont\hfil \contentspagenumber}
  \vfill
  \centerline{\titlefont The {\ttitlefont DVIcopy} processor}
  \vskip 5pt
  \centerline{Copyright (C) 1990--2014 Peter Breitenlohner}
  \centerline{Distributed under terms of GNU General Public License}
  \vskip 15pt
  \centerline{(Version 1.6, September 2009)}
  \vfill}
\def\botofcontents{\vfill
  \centerline{\hsize 5in\baselineskip9pt
    \vbox{\ninerm\noindent
    This program was developed at the Max-Planck-Institut f\"ur Physik
    (Werner-Heisenberg-Institut), Munich, Germany.
    `\TeX' is a trademark of the American Mathematical Society.
    `{\logo hijklmnj}\kern1pt' is a trademark of Addison-Wesley
    Publishing Company.}}}
\pageno=\contentspagenumber \advance\pageno by 1

@* Introduction.
The \.{DVIcopy} utility program copies (selected pages of) binary
device-independent (``\.{DVI}'') files that are produced by document
compilers such as \TeX, and replaces all references to characters from
virtual fonts by the typesetting instructions specified for them in
binary virtual-font (``\.{VF}'') files.
This program has two chief purposes: (1)~It can be used as preprocessor
for existing \.{DVI}-related software in cases where this software is
unable to handle virtual fonts or (given suitable \.{VF} files) where
this software cannot handle fonts with more than 128~characters;
and (2)~it serves as an example of a program that reads \.{DVI} and
\.{VF} files correctly, for system programmers who are developing
\.{DVI}-related software.

Goal number (1) is important since quite a few existing programs have
to be adapted to the extened capabilities of Version~3 of \TeX\ which
will require some time. Moreover some existing programs are `as is' and
the source code is, unfortunately, not available.
Goal number (2) needs perhaps a bit more explanation. Programs for
typesetting need to be especially careful about how they do arithmetic; if
rounding errors accumulate, margins won't be straight, vertical rules
won't line up, and so on (see the documentaion of \.{DVItype} for more
details). This program is written as if it were a \.{DVI}-driver for a
hypothetical typesetting device |out_file|, the output file receiving
the copy of the input |dvi_file|. In addition all code related to
|out_file| is concentrated in two chapters at the end of this program
and quite independent of the rest of the code concerned with the
decoding of \.{DVI} and \.{VF} files and with font substitutions. Thus
it should be relatively easy to replace the device dependent code of
this program by the corresponding code required for a real typesetting
device. Having this in mind \.{DVItype}'s pixel rounding algorithms are
included as conditional code not used by \.{DVIcopy}.

The |banner| and |preamble_comment| strings defined here should be
changed whenever \.{DVIcopy} gets modified.

@d banner=='This is DVIcopy, Version 1.6' {printed when the program starts}
@d title=='DVIcopy' {the name of this program, used in some messages}
@d copyright=='Copyright (C) 1990,2009 Peter Breitenlohner'
@#
@d preamble_comment=='DVIcopy 1.6 output from '
@d comm_length=24 {length of |preamble_comment|}
@d from_length=6 {length of its |' from '| part}

@ This program is written in standard \PASCAL, except where it is necessary
to use extensions; for example, \.{DVIcopy} must read files whose names
are dynamically specified, and that would be impossible in pure \PASCAL.
All places where nonstandard constructions are used have been listed in
the index under ``system dependencies.''
@!@^system dependencies@>

One of the extensions to standard \PASCAL\ that we shall deal with is the
ability to move to a random place in a binary file; another is to
determine the length of a binary file. Such extensions are not necessary
for reading \.{DVI} files; since \.{DVIcopy} is (a model for) a
production program it should, however, be made as efficient as possible
for a particular system. If \.{DVIcopy} is being used with
\PASCAL s for which random file positioning is not efficiently available,
the following definition should be changed from |true| to |false|; in such
cases, \.{DVIcopy} will not include the optional feature that reads the
postamble first.

@d random_reading==true {should we skip around in the file?}

@ The program begins with a fairly normal header, made up of pieces that
@^system dependencies@>
will mostly be filled in later. The \.{DVI} input comes from file
|dvi_file|, the \.{DVI} output goes to file |out_file|, and messages
go to \PASCAL's standard |output| file.
The \.{TFM} and \.{VF} files are defined later since their external
names are determined dynamically.

If it is necessary to abort the job because of a fatal error, the program
calls the `|jump_out|' procedure, which goes to the label |final_end|.

@d final_end = 9999 {go here to wrap it up}

@p @t\4@>@<Compiler directives@>@/
program DVI_copy(@!dvi_file,@!out_file,@!output);
label final_end;
const @<Constants in the outer block@>@/
type @<Types in the outer block@>@/
var @<Globals in the outer block@>@/
@<Error handling procedures@>@/
procedure initialize; {this procedure gets things started properly}
  var @<Local variables for initialization@>@/
  begin print_ln(banner);@/
  print_ln(copyright);
  print_ln('Distributed under terms of GNU General Public License');@/
  @<Set initial values@>@/
  end;

@ The definition of |max_font_type| should be adapted to the number of
font types used by the program; the first three values have a fixed
meaning:  |defined_font=0| indicates that a font has been defined,
|loaded_font=1| indicates that the \.{TFM} file has been loaded but the
font has not yet been used, and |vf_font_type=2| indicates a virtual
font.  Font type values |>=real_font=3| indicate real fonts and
different font types are used to distinguish various kinds of font files
(\.{GF} or \.{PK} or \.{PXL}).  \.{DVIcopy} uses |out_font_type=3| for
fonts that appear in the output \.{DVI} file.
@!@^font types@>

@d defined_font=0 {this font has been defined}
@d loaded_font=1 {this font has been defined and loaded}
@d vf_font_type=2 {this font is a virtual font}
@d real_font=3 {smallest font type for real fonts}
@#
@d out_font_type=3 {this font appears in the output file}
@d max_font_type=3

@ The following parameters can be changed at compile time to extend or
reduce \.{DVIcopy}'s capacity.

@d max_select=10 {maximum number of page selection ranges}

@<Constants...@>=
@!max_fonts=100; {maximum number of distinct fonts}
@!max_chars=10000; {maximum number of different characters among all fonts}
@!max_widths=3000; {maximum number of different characters widths}
@!max_packets=5000; {maximum number of different characters packets;
  must be less than 65536}
@!max_bytes=30000; {maximum number of bytes for characters packets}
@!max_recursion=10; {\.{VF} files shouldn't recurse beyond this level}
@!stack_size=100; {\.{DVI} files shouldn't |push| beyond this depth}
@!terminal_line_length=150; {maximum number of characters input in a single
  line of input from the terminal}
@!name_length=50; {a file name shouldn't be longer than this}

@ As mentioned above, \.{DVIcopy} has two chief purposes: (1)~It produces
a copy of the input \.{DVI} file with all references to characters from
virtual fonts replaced by their expansion as specified in the character
packets of \.{VF} files; and (2)~it serves as an example of a program
that reads \.{DVI} and \.{VF} files correctly, for system programmers
who are developing \.{DVI}-related software.

In fact, a very large section of code (starting with the second chapter
`Introduction (continued)' and ending with the fifteenth chapter
`The main program') is used in identical form in \.{DVIcopy} and in
\.{DVIprint}, a prototype \.{DVI}-driver.  This has been made possible
mostly by using several \.{WEB} coding tricks, such as not to make the
resulting \PASCAL\ program inefficient in any way.

Parts of the program that are needed in \.{DVIprint} but not in
\.{DVIcopy} are delimited by the codewords `$|device|\ldots|ecived|$';
these are mostly the pixel rounding algorithms used to convert the
\.{DVI} units of a \.{DVI} file to the raster units of a real output
device and have been copied more or less verbatim from \.{DVItype}.

@d device==@{ {change this to `$\\{device}\equiv\null$' when output
  for a real device is produced}
@d ecived==@t@>@} {change this to `$\\{ecived}\equiv\null$' when output
  for a real device is produced}
@f device==begin
@f ecived==end

@* Introduction (continued).
On some systems it is necessary to use various integer subrange types
in order to make \.{\title} efficient; this is true in particular for
frequently used variables such as loop indices. Consider an integer
variable |x| with values in the range |0..255|: on most small systems
|x| should be a one or two byte integer whereas on most large systems
|x| should be a four byte integer.
Clearly the author of a program knows best which range of values is
required for each variable; thus \.{\title} never uses \PASCAL's |integer|
type. All integer variables are declared as one of the integer subrange
types defined below as \.{WEB} macros or \PASCAL\ types; these definitions
can be used without system-dependent changes, provided the signed 32~bit
integers are a subset of the standard type |integer|, and the compiler
automatically uses the optimal representation for integer subranges
(both conditions need not be satisfied for a particular system).
@^system dependencies@>

The complementary problem of storing large arrays of integer type
variables as compactly as possible is addressed differently; here
\.{\title} uses a \PASCAL\ |type|~declaration for each kind of array
element.

Note that the primary purpose of these definitions is optimizations, not
range checking. All places where optimization for a particular system is
highly desirable have been listed in the index under ``optimization.''
@!@^optimization@>

@d int_32 == integer {signed 32~bit integers}

@<Types...@>=
@!int_31 = 0..@"7FFFFFFF; {unsigned 31~bit integer}
@!int_24u = 0..@"FFFFFF; {unsigned 24~bit integer}
@!int_24 = -@"800000..@"7FFFFF; {signed 24~bit integer}
@!int_23 = 0..@"7FFFFF; {unsigned 23~bit integer}
@!int_16u = 0..@"FFFF; {unsigned 16~bit integer}
@!int_16 = -@"8000..@"7FFF; {signed 16~bit integer}
@!int_15 = 0..@"7FFF; {unsigned 15~bit integer}
@!int_8u = 0..@"FF; {unsigned 8~bit integer}
@!int_8 = -@"80..@"7F; {signed 8~bit integer}
@!int_7 = 0..@"7F; {unsigned 7~bit integer}

@ Some of this code is optional for use when debugging only;
such material is enclosed between the delimiters |debug| and $|gubed|$.
Other parts, delimited by |stat| and $|tats|$, are optionally included
if statistics about \.{\title}'s memory usage are desired.

@d debug==@{ {change this to `$\\{debug}\equiv\null$' when debugging}
@d gubed==@t@>@} {change this to `$\\{gubed}\equiv\null$' when debugging}
@f debug==begin
@f gubed==end
@#
@d stat==@{ {change this to `$\\{stat}\equiv\null$'
  when gathering usage statistics}
@d tats==@t@>@} {change this to `$\\{tats}\equiv\null$'
  when gathering usage statistics}
@f stat==begin
@f tats==end

@ The \PASCAL\ compiler used to develop this program has ``compiler
directives'' that can appear in comments whose first character is a dollar sign.
In production versions of \.{\title} these directives tell the compiler that
@^system dependencies@>
it is safe to avoid range checks and to leave out the extra code it inserts
for the \PASCAL\ debugger's benefit, although interrupts will occur if
there is arithmetic overflow.

@<Compiler directives@>=
@{@&$C-,A+,D-@} {no range check, catch arithmetic overflow, no debug overhead}
@!debug @{@&$C+,D+@}@+ gubed {but turn everything on when debugging}

@ Labels are given symbolic names by the following definitions. We insert
the label `|exit|:' just before the `\ignorespaces|end|\unskip' of a
procedure in which we have used the `|return|' statement defined below;
the label `|restart|' is occasionally used at the very beginning of a
procedure; and the label `|reswitch|' is occasionally used just prior to
a \&{case} statement in which some cases change the conditions and we wish to
branch to the newly applicable case.
Loops that are set up with the \&{loop} construction defined below are
commonly exited by going to `|done|' or to `|found|' or to `|not_found|',
and they are sometimes repeated by going to `|continue|'.

@d exit=10 {go here to leave a procedure}
@d restart=20 {go here to start a procedure again}
@d reswitch=21 {go here to start a case statement again}
@d continue=22 {go here to resume a loop}
@d done=30 {go here to exit a loop}
@d found=31 {go here when you've found it}
@d not_found=32 {go here when you've found something else}

@ The term |print| is used instead of |write| when this program writes on
|output|, so that all such output could easily be redirected if desired;
the term |d_print| is used for conditional output if we are debugging.

@d print(#)==write(output,#)
@d print_ln(#)==write_ln(output,#)
@d new_line==write_ln(output) {start new line}
@d print_nl(#)==  {print information starting on a new line}
  begin new_line; print(#);
  end
@#
@d d_print(#)==@!debug print(#) @; @+ gubed
@d d_print_ln(#)==@! debug print_ln(#) @; @+ gubed

@ Here are some macros for common programming idioms.

@d incr(#) == #:=#+1 {increase a variable by unity}
@d decr(#) == #:=#-1 {decrease a variable by unity}
@#
@d Incr_Decr_end(#)==#
@d Incr(#)==#:=#+Incr_Decr_end {we use |Incr(a)(b)| to increase \dots}
@d Decr(#)==#:=#-Incr_Decr_end {\dots\ and |Decr(a)(b)| to decrease
  variable |a| by |b|; this can be optimized for some compilers}
@#
@d loop == @+ while true do@+ {repeat over and over until a |goto| happens}
@d do_nothing == {empty statement}
@d return == goto exit {terminate a procedure call}
@f return == nil
@f loop == xclause

@ We assume that |case| statements may include a default case that applies
if no matching label is found. Thus, we shall use constructions like
@^system dependencies@>
$$\vbox{\halign{#\hfil\cr
|case x of|\cr
1: $\langle\,$code for $x=1\,\rangle$;\cr
3: $\langle\,$code for $x=3\,\rangle$;\cr
|othercases| $\langle\,$code for |x<>1| and |x<>3|$\,\rangle$\cr
|endcases|\cr}}$$
since most \PASCAL\ compilers have plugged this hole in the language by
incorporating some sort of default mechanism. For example, the compiler
used to develop \.{WEB} and \TeX\ allows `|others|:' as a default label,
and other \PASCAL s allow syntaxes like `\ignorespaces|else|\unskip' or
`\&{otherwise}' or `\\{otherwise}:', etc. The definitions of |othercases|
and |endcases| should be changed to agree with local conventions. (Of
course, if no default mechanism is available, the |case| statements of
this program must be extended by listing all remaining cases.
Donald~E. Knuth, the author of the \.{WEB} system program \.{TANGLE},
@^Knuth, Donald Ervin@>
would have taken the trouble to modify \.{TANGLE} so that such extensions
were done automatically, if he had not wanted to encourage \PASCAL\
compiler writers to make this important change in \PASCAL, where it belongs.)

@d othercases == others: {default for cases not listed explicitly}
@d endcases == @+end {follows the default case in an extended |case| statement}
@f othercases == else
@f endcases == end

@* The character set.
Like all programs written with the  \.{WEB} system, \.{\title} can be
used with any character set. But it uses ASCII code internally, because
the programming for portable input-output is easier when a fixed internal
code is used, and because \.{DVI} and \.{VF} files use ASCII code for
file names and certain other strings.

The next few sections of \.{\title} have therefore been copied from the
analogous ones in the \.{WEB} system routines. They have been considerably
simplified, since \.{\title} need not deal with the controversial
ASCII codes less than @'40 or greater than @'176.
If such codes appear in the \.{DVI} file,
they will be printed as question marks.

@<Types...@>=
@!ASCII_code=" ".."~"; {a subrange of the integers}

@ The original \PASCAL\ compiler was designed in the late 60s, when six-bit
character sets were common, so it did not make provision for lower case
letters. Nowadays, of course, we need to deal with both upper and lower case
alphabets in a convenient way, especially in a program like \.{\title}.
So we shall assume that the \PASCAL\ system being used for \.{\title}
has a character set containing at least the standard visible characters
of ASCII code (|"!"| through |"~"|).

Some \PASCAL\ compilers use the original name |char| for the data type
associated with the characters in text files, while other \PASCAL s
consider |char| to be a 64-element subrange of a larger data type that has
some other name.  In order to accommodate this difference, we shall use
the name |text_char| to stand for the data type of the characters in the
output file.  We shall also assume that |text_char| consists of
the elements |chr(first_text_char)| through |chr(last_text_char)|,
inclusive. The following definitions should be adjusted if necessary.
@^system dependencies@>

@d text_char == char {the data type of characters in text files}
@d first_text_char=0 {ordinal number of the smallest element of |text_char|}
@d last_text_char=127 {ordinal number of the largest element of |text_char|}

@<Types...@>=
@!text_file=packed file of text_char;

@ @<Local variables for init...@>=
@!i:int_16; {loop index for initializations}

@ The \.{\title} processor converts between ASCII code and
the user's external character set by means of arrays |xord| and |xchr|
that are analogous to \PASCAL's |ord| and |chr| functions.

@<Globals...@>=
@!xord: array [text_char] of ASCII_code;
  {specifies conversion of input characters}
@!xchr: array [0..255] of text_char;
  {specifies conversion of output characters}

@ Under our assumption that the visible characters of standard ASCII are
all present, the following assignment statements initialize the
|xchr| array properly, without needing any system-dependent changes.

@<Set init...@>=
for i:=0 to @'37 do xchr[i]:='?';
xchr[@'40]:=' ';
xchr[@'41]:='!';
xchr[@'42]:='"';
xchr[@'43]:='#';
xchr[@'44]:='$';
xchr[@'45]:='%';
xchr[@'46]:='&';
xchr[@'47]:='''';@/
xchr[@'50]:='(';
xchr[@'51]:=')';
xchr[@'52]:='*';
xchr[@'53]:='+';
xchr[@'54]:=',';
xchr[@'55]:='-';
xchr[@'56]:='.';
xchr[@'57]:='/';@/
xchr[@'60]:='0';
xchr[@'61]:='1';
xchr[@'62]:='2';
xchr[@'63]:='3';
xchr[@'64]:='4';
xchr[@'65]:='5';
xchr[@'66]:='6';
xchr[@'67]:='7';@/
xchr[@'70]:='8';
xchr[@'71]:='9';
xchr[@'72]:=':';
xchr[@'73]:=';';
xchr[@'74]:='<';
xchr[@'75]:='=';
xchr[@'76]:='>';
xchr[@'77]:='?';@/
xchr[@'100]:='@@';
xchr[@'101]:='A';
xchr[@'102]:='B';
xchr[@'103]:='C';
xchr[@'104]:='D';
xchr[@'105]:='E';
xchr[@'106]:='F';
xchr[@'107]:='G';@/
xchr[@'110]:='H';
xchr[@'111]:='I';
xchr[@'112]:='J';
xchr[@'113]:='K';
xchr[@'114]:='L';
xchr[@'115]:='M';
xchr[@'116]:='N';
xchr[@'117]:='O';@/
xchr[@'120]:='P';
xchr[@'121]:='Q';
xchr[@'122]:='R';
xchr[@'123]:='S';
xchr[@'124]:='T';
xchr[@'125]:='U';
xchr[@'126]:='V';
xchr[@'127]:='W';@/
xchr[@'130]:='X';
xchr[@'131]:='Y';
xchr[@'132]:='Z';
xchr[@'133]:='[';
xchr[@'134]:='\';
xchr[@'135]:=']';
xchr[@'136]:='^';
xchr[@'137]:='_';@/
xchr[@'140]:='`';
xchr[@'141]:='a';
xchr[@'142]:='b';
xchr[@'143]:='c';
xchr[@'144]:='d';
xchr[@'145]:='e';
xchr[@'146]:='f';
xchr[@'147]:='g';@/
xchr[@'150]:='h';
xchr[@'151]:='i';
xchr[@'152]:='j';
xchr[@'153]:='k';
xchr[@'154]:='l';
xchr[@'155]:='m';
xchr[@'156]:='n';
xchr[@'157]:='o';@/
xchr[@'160]:='p';
xchr[@'161]:='q';
xchr[@'162]:='r';
xchr[@'163]:='s';
xchr[@'164]:='t';
xchr[@'165]:='u';
xchr[@'166]:='v';
xchr[@'167]:='w';@/
xchr[@'170]:='x';
xchr[@'171]:='y';
xchr[@'172]:='z';
xchr[@'173]:='{';
xchr[@'174]:='|';
xchr[@'175]:='}';
xchr[@'176]:='~';
for i:=@'177 to 255 do xchr[i]:='?';

@ The following system-independent code makes the |xord| array contain a
suitable inverse to the information in |xchr|.

@<Set init...@>=
for i:=first_text_char to last_text_char do xord[chr(i)]:=@'40;
for i:=" " to "~" do xord[xchr[i]]:=i;

@* Reporting errors to the user.
The \.{\title} processor does not verify that every single bit read from
one of its binary input files is meaningful and consistent; there are
other programs, e.g., \.{DVItype}, \.{TFtoPL}, and \.{VFtoPL}, specially
designed for that purpose.

On the other hand, \.{\title} is designed to avoid unpredictable results
due to undetected arithmetic overflow, or due to violation of integer
subranges or array bounds under {\it all\/} circumstances. Thus a fair
amount of checking is done when reading and analyzing the input data,
even in cases where such checking reduces the efficiency of the program
to some extent.

@ A global variable called |history| will contain one of four values
at the end of every run: |spotless| means that no unusual messages were
printed; |harmless_message| means that a message of possible interest
was printed but no serious errors were detected; |error_message| means that
at least one error was found; |fatal_message| means that the program
terminated abnormally. The value of |history| does not influence the
behavior of the program; it is simply computed for the convenience
of systems that might want to use such information.

@d spotless=0 {|history| value for normal jobs}
@d harmless_message=1 {|history| value when non-serious info was printed}
@d error_message=2 {|history| value when an error was noted}
@d fatal_message=3 {|history| value when we had to stop prematurely}
@#
@d mark_harmless==@t@>@+if history=spotless then history:=harmless_message
@d mark_error==history:=error_message
@d mark_fatal==history:=fatal_message

@<Glob...@>=@!history:spotless..fatal_message; {how bad was this run?}

@ @<Set init...@>=history:=spotless;

@ If an input (\.{DVI}, \.{TFM}, \.{VF}, or other) file is badly malformed,
the whole process must be aborted; \.{\title} will give up, after issuing
an error message about what caused the error. These messages will, however,
in most cases just indicate which input file caused the error. One of the
programs \.{DVItype}, \.{TFtoPL} or \.{VFtoVP} should then be used to
diagnose the error in full detail.

Such errors might be discovered inside of subroutines inside of subroutines,
so a procedure called |jump_out| has been introduced. This procedure, which
transfers control to the label |final_end| at the end of the program,
contains the only non-local |@!goto| statement in \.{\title}.
@^system dependencies@>
Some \PASCAL\ compilers do not implement non-local |goto| statements. In
such cases the |goto final_end| in |jump_out| should simply be replaced
by a call on some system procedure that quietly terminates the program.
@^system dependencies@>

@d abort(#)==begin print_ln(' ',#,'.'); jump_out;
    end

@<Error handling...@>=
@<Basic printing procedures@>@;
procedure close_files_and_terminate; forward;
@#
procedure jump_out;
begin mark_fatal; close_files_and_terminate;
goto final_end;
end;

@ Sometimes the program's behavior is far different from what it should
be, and \.{\title} prints an error message that is really for the
\.{\title} maintenance person, not the user. In such cases the program
says |confusion(|indication of where we are|)|.

@<Error handling...@>=
procedure confusion(@!p:pckt_pointer);
begin print(' !This can''t happen ('); print_packet(p); print_ln(').');
@.This can't happen@>
jump_out;
end;

@ An overflow stop occurs if \.{\title}'s tables aren't large enough.

@<Error handling...@>=
procedure overflow(@!p:pckt_pointer;@!n:int_16u);
begin print(' !Sorry, ',title,' capacity exceeded ['); print_packet(p);
@.Sorry, {\title} capacity exceeded@>
print_ln('=',n:1,'].');
jump_out;
end;

@* Binary data and binary files.
A detailed description of the \.{DVI} file format can be found in the
documentation of \TeX, \.{DVItype}, or \.{GFtoDVI}; here we just define
symbolic names for some of the \.{DVI} command bytes.

@d set_char_0=0 {typeset character 0 and move right}
@d set1=128 {typeset a character and move right}
@d set_rule=132 {typeset a rule and move right}
@d put1=133 {typeset a character}
@d put_rule=137 {typeset a rule}
@d nop=138 {no operation}
@d bop=139 {beginning of page}
@d eop=140 {ending of page}
@d push=141 {save the current positions}
@d pop=142 {restore previous positions}
@d right1=143 {move right}
@d w0=147 {move right by |w|}
@d w1=148 {move right and set |w|}
@d x0=152 {move right by |x|}
@d x1=153 {move right and set |x|}
@d down1=157 {move down}
@d y0=161 {move down by |y|}
@d y1=162 {move down and set |y|}
@d z0=166 {move down by |z|}
@d z1=167 {move down and set |z|}
@d fnt_num_0=171 {set current font to 0}
@d fnt1=235 {set current font}
@d xxx1=239 {extension to \.{DVI} primitives}
@d xxx4=242 {potentially long extension to \.{DVI} primitives}
@d fnt_def1=243 {define the meaning of a font number}
@d pre=247 {preamble}
@d post=248 {postamble beginning}
@d post_post=249 {postamble ending}
@#
@d dvi_id=2 {identifies \.{DVI} files}
@d dvi_pad=223 {pad bytes at end of \.{DVI} file}

@ A \.{DVI}, \.{VF}, or \.{TFM} file is a sequence of 8-bit bytes.
The bytes appear physically in what is called a `|packed file of 0..255|'
in \PASCAL\ lingo. One, two, three, or four consecutive bytes are often
interpreted as (signed or unsigned) integers.
We might as well define the corresponding data types.
@!@^system dependencies@>

@<Types...@>=
@!signed_byte=-@"80..@"7F; {signed one-byte quantity}
@!eight_bits=0..@"FF; {unsigned one-byte quantity}
@!signed_pair=-@"8000..@"7FFF; {signed two-byte quantity}
@!sixteen_bits=0..@"FFFF; {unsigned two-byte quantity}
@!signed_trio=-@"800000..@"7FFFFF; {signed three-byte quantity}
@!twentyfour_bits=0..@"FFFFFF; {unsigned three-byte quantity}
@!signed_quad=int_32; {signed four-byte quantity}

@ Packing is system dependent, and many \PASCAL\ systems fail to implement
such files in a sensible way (at least, from the viewpoint of producing
good production software).  For example, some systems treat all
byte-oriented files as text, looking for end-of-line marks and such
things. Therefore some system-dependent code is often needed to deal with
binary files, even though most of the program in this section of
\.{\title} is written in standard \PASCAL.
@^system dependencies@>

One common way to solve the problem is to consider files of |integer|
numbers, and to convert an integer in the range $-2^{31}\L x<2^{31}$ to
a sequence of four bytes $(a,b,c,d)$ using the following code, which
avoids the controversial integer division of negative numbers:
$$\vbox{\halign{#\hfil\cr
|if x>=0 then a:=x div @'100000000|\cr
|else begin x:=(x+@'10000000000)+@'10000000000; a:=x div @'100000000+128;|\cr
\quad|end|\cr
|x:=x mod @'100000000;|\cr
|b:=x div @'200000; x:=x mod @'200000;|\cr
|c:=x div @'400; d:=x mod @'400;|\cr}}$$
The four bytes are then kept in a buffer and output one by one. (On 36-bit
computers, an additional division by 16 is necessary at the beginning.
Another way to separate an integer into four bytes is to use/abuse
\PASCAL's variant records, storing an integer and retrieving bytes that are
packed in the same place; {\sl caveat implementor!\/}) It is also desirable
in some cases to read a hundred or so integers at a time, maintaining a
larger buffer.

@ We shall stick to simple \PASCAL\ in the standard version of this program,
for reasons of clarity, even if such simplicity is sometimes unrealistic.

@<Types...@>=
@!byte_file=packed file of eight_bits; {files that contain binary data}

@ For some operating systems it may be convenient or even necessary to
close the input files.

@d close_in(#)==do_nothing {close an input file}

@ Character packets extracted from \.{VF} files will be stored in a large
array |byte_mem|. Other packets of bytes, e.g., character packets
extracted from a \.{GF} or \.{PK} or \.{PXL} file could be stored in the
same way. A `|pckt_pointer|' variable, which signifies a packet,
is an index into another array |pckt_start|. The actual sequence of bytes
in the packet pointed to by |p| appears in positions |pckt_start[p]| to
|pckt_start[p+1]-1|, inclusive, in |byte_mem|.

Packets will also be used to store sequences of |ASCII_code|s; in this
respect the |byte_mem| array is very similar to \TeX's string pool and
part of the following code has, in fact, been copied more or less
verbatim from \TeX.

In other respects the packets resemble the identifiers used by
\.{TANGLE} and \.{WEAVE} (also stored in an array called |byte_mem|)
since there is, in general, at most one packet with a given contents;
thus part of the code below has been adapted from the corresponding code
in these programs.

Some \PASCAL\ compilers won't pack integers into a single byte unless the
integers lie in the range |-128..127|. To accommodate such systems we
access the array |byte_mem| only via macros that can easily be redefined.
@^system dependencies@>

@d bi(#) == # {convert from |eight_bits| to |packed_byte|}
@d bo(#) == # {convert from |packed_byte| to |eight_bits|}

@<Types...@>=
@!packed_byte = eight_bits; {elements of |byte_mem| array}
@!byte_pointer = 0..max_bytes; {an index into |byte_mem|}
@!pckt_pointer = 0..max_packets; {an index into |pckt_start|}

@ The global variable |byte_ptr| points to the first unused location in
|byte_mem| and |pckt_ptr| points to the first unused location in
|pckt_start|.

@<Globals...@>=
@!byte_mem: packed array [byte_pointer] of packed_byte; {bytes of packets}
@!pckt_start: array [pckt_pointer] of byte_pointer;
  {directory into |byte_mem|}
@!byte_ptr: byte_pointer;
@!pckt_ptr: pckt_pointer;

@ Several of the elementary operations with packets are performed using
\.{WEB} macros instead of \PASCAL\ procedures, because many of the
operations are done quite frequently and we want to avoid the
overhead of procedure calls. For example, here is
a simple macro that computes the length of a packet.
@.WEB@>

@d pckt_length(#)==(pckt_start[#+1]-pckt_start[#]) {the number of bytes
  in packet number \#}

@ Packets are created by appending bytes to |byte_mem|.
The |append_byte| macro, defined here, does not check to see if the
value of |byte_ptr| has gotten too high; this test is supposed to be
made before |append_byte| is used. There is also a |flush_byte|
macro, which erases the last byte appended.

To test if there is room to append |l| more bytes to |byte_mem|,
we shall write |pckt_room(l)|, which aborts \.{\title} and gives an
apologetic error message if there isn't enough room.

@d append_byte(#) == {put byte \# at the end of |byte_mem|}
begin byte_mem[byte_ptr]:=bi(#); incr(byte_ptr);
end
@d flush_byte == decr(byte_ptr) {forget the last byte in |byte_mem|}
@d pckt_room(#) == {make sure that |byte_mem| hasn't overflowed}
  if max_bytes-byte_ptr<# then overflow(str_bytes,max_bytes)
@#
@d append_one(#) ==
begin pckt_room(1); append_byte(#);
end

@ The length of the current packet is called |cur_pckt_length|:

@d cur_pckt_length == (byte_ptr - pckt_start[pckt_ptr])

@ Once a sequence of bytes has been appended to |byte_mem|, it
officially becomes a packet when the |make_packet| function is called.
This function returns as its value the identification number of either
an existing packet with the same contents or, if no such packet exists,
of the new packet. Thus two packets have the same contents if and only
if they have the same identification number. In order to locate the
packet with a given contents, or to find out that no such packet exists,
we need a hash table. The hash table is kept by the method of simple
chaining, where the heads of the individual lists appear in the |p_hash|
array. If |h| is a hash code, the hash table list starts at |p_hash[h]|
and proceeds through |p_link| pointers.

@d hash_size=353 {should be prime, must be |>256|}

@<Types...@>=
@!hash_code=0..hash_size;

@ @<Glob...@>=
@!p_link:array[pckt_pointer] of pckt_pointer; {hash table}
@!p_hash:array[hash_code] of pckt_pointer;

@ Initially |byte_mem| and all the hash lists are empty; |empty_packet|
is the empty packet.

@d empty_packet=0 {the empty packet}
@d invalid_packet==max_packets {used when there is no packet}

@<Set init...@>=
pckt_ptr:=1; byte_ptr:=1;
pckt_start[0]:=1; pckt_start[1]:=1;
for h:=0 to hash_size-1 do p_hash[h]:=0;

@ @<Local variables for init...@>=
@!h:hash_code; {index into hash-head arrays}

@ Here now is the |make_packet| function used to create packets (and
strings).

@p function make_packet:pckt_pointer;
label found;
var i,@!k:byte_pointer; {indices into |byte_mem|}
@!h:hash_code; {hash code}
@!s,@!l:byte_pointer; {start and length of the given packet}
@!p:pckt_pointer; {where the packet is being sought}
begin s:=pckt_start[pckt_ptr]; l:=byte_ptr-s; {compute start and length}
if l=0 then p:=empty_packet
else  begin @<Compute the packet hash code |h|@>;
  @<Compute the packet location |p|@>;
  if pckt_ptr=max_packets then overflow(str_packets,max_packets);
  incr(pckt_ptr); pckt_start[pckt_ptr]:=byte_ptr;
  end;
found:make_packet:=p;
end;

@ A simple hash code is used: If the sequence of bytes is
$b_1b_2\ldots b_n$, its hash value will be
$$(2^{n-1}b_1+2^{n-2}b_2+\cdots+b_n)\,\bmod\,|hash_size|.$$

@<Compute the packet hash...@>=
h:=bo(byte_mem[s]); i:=s+1;
while i<byte_ptr do
  begin h:=(h+h+bo(byte_mem[i])) mod hash_size; incr(i);
  end

@ If the packet is new, it will be placed in position |p=pckt_ptr|,
otherwise |p| will point to its existing location.

@<Compute the packet location...@>=
p:=p_hash[h];
while p<>0 do
  begin if pckt_length(p)=l then
      @<Compare packet |p| with current packet, |goto found| if equal@>;
  p:=p_link[p];
  end;
p:=pckt_ptr; {the current packet is new}
p_link[p]:=p_hash[h]; p_hash[h]:=p {insert |p| at beginning of hash list}

@ @<Compare packet |p|...@>=
begin i:=s; k:=pckt_start[p];
while (i<byte_ptr)and(byte_mem[i]=byte_mem[k]) do
  begin incr(i); incr(k);
  end;
if i=byte_ptr then {all bytes agree}
  begin byte_ptr:=pckt_start[pckt_ptr]; goto found;
  end;
end

@ Some packets are initialized with predefined strings of |ASCII_code|s;
a few macros permit us to do the initialization with a compact program.
Since this initialization is done when |byte_mem| is still empty, and
since |byte_mem| is supposed to be large enough for all the predefined
strings, |pckt_room| is used only if we are debugging.

@d pid0(#)==#:=make_packet
@d pid1(#)==byte_mem[byte_ptr-1]:=bi(#); pid0
@d pid2(#)==byte_mem[byte_ptr-2]:=bi(#); pid1
@d pid3(#)==byte_mem[byte_ptr-3]:=bi(#); pid2
@d pid4(#)==byte_mem[byte_ptr-4]:=bi(#); pid3
@d pid5(#)==byte_mem[byte_ptr-5]:=bi(#); pid4
@d pid6(#)==byte_mem[byte_ptr-6]:=bi(#); pid5
@d pid7(#)==byte_mem[byte_ptr-7]:=bi(#); pid6
@d pid8(#)==byte_mem[byte_ptr-8]:=bi(#); pid7
@d pid9(#)==byte_mem[byte_ptr-9]:=bi(#); pid8
@d pid10(#)==byte_mem[byte_ptr-10]:=bi(#); pid9
@#
@d pid_init(#)==
  @!debug pckt_room(#); @+ gubed @;
  Incr(byte_ptr)(#)
@#
@d id1==pid_init(1); pid1
@d id2==pid_init(2); pid2
@d id3==pid_init(3); pid3
@d id4==pid_init(4); pid4
@d id5==pid_init(5); pid5
@d id6==pid_init(6); pid6
@d id7==pid_init(7); pid7
@d id8==pid_init(8); pid8
@d id9==pid_init(9); pid9
@d id10==pid_init(10); pid10

@ Here we initialize some strings used as argument of the |overflow| and
|confusion| procedures.

@<Initialize predefined strings@>=
id5("f")("o")("n")("t")("s")(str_fonts);
id5("c")("h")("a")("r")("s")(str_chars);
id6("w")("i")("d")("t")("h")("s")(str_widths);
id7("p")("a")("c")("k")("e")("t")("s")(str_packets);
id5("b")("y")("t")("e")("s")(str_bytes);
id9("r")("e")("c")("u")("r")("s")("i")("o")("n")(str_recursion);
id5("s")("t")("a")("c")("k")(str_stack);
id10("n")("a")("m")("e")("l")("e")("n")("g")("t")("h")(str_name_length);

@ @<Glob...@>=
@!str_fonts,@!str_chars,@!str_widths,@!str_packets,@!str_bytes,
@!str_recursion,@!str_stack,@!str_name_length:pckt_pointer;

@ Some packets, e.g., the preamble comments of \.{DVI} and \.{VF} files,
are needed only temporarily. In such cases |new_packet| is used to
create a packet (which might duplicate an existing packet) and
|flush_packet| is used to discard it; the calls to |new_packet| and
|flush_packet| must occur in balanced pairs, without any intervening
calls to |make_packet|.

@p function new_packet: pckt_pointer;
begin if pckt_ptr=max_packets then overflow(str_packets,max_packets);
new_packet:=pckt_ptr; incr(pckt_ptr); pckt_start[pckt_ptr]:=byte_ptr;
end;
@#
procedure flush_packet;
begin decr(pckt_ptr); byte_ptr:=pckt_start[pckt_ptr];
end;

@ The |print_packet| procedure prints the contents of a packet; such a
packet should, of course, consists of a sequence of |ASCII_code|s.

@<Basic printing...@>=
procedure print_packet(p:pckt_pointer);
var k:byte_pointer;
begin for k:=pckt_start[p] to pckt_start[p+1]-1 do
  print(xchr[bo(byte_mem[k])]);
end;

@ When we interpret a packet we will use two (global or local) variables:
|cur_loc| will point to the byte to be used next, and |cur_limit| will
point to the start of the next packet. The macro |pckt_extract| will be
used to extract one byte; it should, however, never be used with
|cur_loc>=cur_limit|.

@d pckt_extract(#) ==
@!debug if cur_loc>=cur_limit then confusion(str_packets) @+ else @/
gubed @;
  begin #:=bo(byte_mem[cur_loc]); incr(cur_loc); @+ end

@<Globals...@>=
@!cur_pckt: pckt_pointer; {the current packet}
@!cur_loc: byte_pointer; {current location in a packet}
@!cur_limit: byte_pointer; {start of next packet}

@ We will need routines to extract one, two, three, or four bytes from
|byte_mem|, from the \.{DVI} file, or from a \.{VF} file and assemble
them into (signed or unsigned) integers and these routines should be
optimized for efficiency. Here we define \.{WEB} macros to be used for
the body of these routines; thus the changes for system dependent
optimization have to be applied only once.
@^system dependencies@>
@^optimization@>

In addition we demonstrates how these macros can be used to define
functions that extract one, two, three, or four bytes from a character
packet and assemble them into signed or unsigned integers (assuming that
|cur_loc| and |cur_limit| are initialized suitably).

@d begin_byte(#) ==
var a:eight_bits;
begin #(a)
@d comp_sbyte(#) == if a<128 then #:=a @+ else #:=a-256
@d comp_ubyte(#) == #:=a
@f begin_byte == begin

@p function pckt_sbyte:int_8; {returns the next byte, signed}
@!begin_byte(pckt_extract); comp_sbyte(pckt_sbyte);
end;
@#
function pckt_ubyte:int_8u; {returns the next byte, unsigned}
@!begin_byte(pckt_extract); comp_ubyte(pckt_ubyte);
end;

@ @d begin_pair(#) ==
var a,@!b:eight_bits;
begin #(a); #(b)
@d comp_spair(#) == if a<128 then #:=a*256+b @+ else #:=(a-256)*256+b
@d comp_upair(#) == #:=a*256+b
@f begin_pair == begin

@p function pckt_spair:int_16; {returns the next two bytes, signed}
@!begin_pair(pckt_extract); comp_spair(pckt_spair);
end;
@#
function pckt_upair:int_16u; {returns the next two bytes, unsigned}
@!begin_pair(pckt_extract); comp_upair(pckt_upair);
end;

@ @d begin_trio(#) ==
var a,@!b,@!c:eight_bits;
begin #(a); #(b); #(c)
@d comp_strio(#) ==
if a<128 then #:=(a*256+b)*256+c @+ else #:=((a-256)*256+b)*256+c
@d comp_utrio(#) == #:=(a*256+b)*256+c
@f begin_trio == begin

@p function pckt_strio:int_24; {returns the next three bytes, signed}
@!begin_trio(pckt_extract); comp_strio(pckt_strio);
end;
@#
function pckt_utrio:int_24u; {returns the next three bytes, unsigned}
@!begin_trio(pckt_extract); comp_utrio(pckt_utrio);
end;

@ @d begin_quad(#) ==
var a,@!b,@!c,@!d:eight_bits;
begin #(a); #(b); #(c); #(d)
@d comp_squad(#) ==
if a<128 then #:=((a*256+b)*256+c)*256+d
else #:=(((a-256)*256+b)*256+c)*256+d
@f begin_quad == begin

@p function pckt_squad:int_32; {returns the next four bytes, signed}
@!begin_quad(pckt_extract); comp_squad(pckt_squad);
end;

@ A similar set of routines is needed for the inverse task of
decomposing a \.{DVI} command into a sequence of bytes to be appended
to |byte_mem| or, in the case of \.{DVIcopy}, to be written to the
output file. Again we define \.{WEB} macros to be used for the body
of these routines; thus the changes for system dependent optimization
have to be applied only once.
@^system dependencies@>
@^optimization@>

First, the |pckt_one| outputs one byte, negative values are represented
in two's complement notation.

@d begin_one == begin
@d comp_one(#) ==
if x<0 then Incr(x)(256);
#(x)
@f begin_one == begin

@p @!device
procedure pckt_one(@!x:int_32); {output one byte}
@!begin_one; pckt_room(1); comp_one(append_byte);
end;
ecived

@ The |pckt_two| outputs two bytes, negative values are represented in
two's complement notation.

@d begin_two == begin
@d comp_two(#) ==
if x<0 then Incr(x)(@"10000);
#(x div @"100); #(x mod @"100)
@f begin_two == begin

@p @!device
procedure pckt_two(@!x:int_32); {output two byte}
@!begin_two; pckt_room(2); comp_two(append_byte);
end;
ecived

@ The |pckt_four| procedure outputs four bytes in two's complement
notation, without risking arithmetic overflow.

@d begin_four == begin
@d comp_four(#) ==
if x>=0 then #(x div @"1000000)
else  begin Incr(x)(@"40000000); Incr(x)(@"40000000);
  #((x div @"1000000) + 128);
  end;
x:=x mod @"1000000; #(x div @"10000);
x:=x mod @"10000; #(x div @"100);
#(x mod @"100)
@f begin_four == begin

@p procedure pckt_four(@!x:int_32); {output four bytes}
@!begin_four; pckt_room(4); comp_four(append_byte);
end;

@ Next, the |pckt_char| procedure outputs a |set_char| or \\{set} command
or, if |upd=false|, a |put| command.

@d begin_char ==
var o:eight_bits; {|set1| or |put1|}
begin
@d comp_char(#) ==
if (not upd)or(res>127)or(ext<>0) then
  begin o:=dvi_char_cmd[upd]; {|set1| or |put1|}
  if ext<0 then Incr(ext)(@"1000000);
  if ext=0 then #(o) @+ else @;
    begin if ext<@"100 then #(o+1) @+ else @;
      begin if ext<@"10000 then #(o+2) @+ else @;
        begin #(o+3); #(ext div @"10000); ext:=ext mod @"10000;
        end;
      #(ext div @"100); ext:=ext mod @"100;
      end;
    #(ext);
    end;
  end;
#(res)
@f begin_char == begin

@p procedure pckt_char(@!upd:boolean;@!ext:int_32;@!res:eight_bits);
  {output \\{set} or |put|}
@!begin_char; pckt_room(5); comp_char(append_byte);
end;

@ Then, the |pckt_unsigned| procedure outputs a |fnt| or |xxx|
command with its first parameter (normally unsigned); a |fnt| command
is converted into |fnt_num| whenever this is possible.

@d begin_unsigned == begin
@d comp_unsigned(#) ==
if (x<@"100)and(x>=0) then
  if (o=fnt1)and(x<64) then Incr(x)(fnt_num_0) @+ else #(o)
else
  begin if (x<@"10000)and(x>=0) then #(o+1) @+ else @;
    begin if (x<@"1000000)and(x>=0) then #(o+2) @+ else @;
      begin #(o+3);
      if x>=0 then #(x div @"1000000)
      else  begin Incr(x)(@"40000000); Incr(x)(@"40000000);
        #((x div @"1000000) + 128);
        end;
      x:=x mod @"1000000;
      end;
    #(x div @"10000); x:=x mod @"10000;
    end;
  #(x div @"100); x:=x mod @"100;
  end;
#(x)
@f begin_unsigned == begin

@p procedure pckt_unsigned(@!o:eight_bits;@!x:int_32);
  {output |fnt_num|, |fnt|, or |xxx|}
@!begin_unsigned; pckt_room(5); comp_unsigned(append_byte);
end;

@ Finally, the |pckt_signed| procedure outputs a movement (|right|, |w|,
|x|, |down|, |y|, or |z|) command with its (signed) parameter.

@d begin_signed ==
var xx:int_31; {`absolute value' of |x|}
begin
@d comp_signed(#) ==
if x>=0 then xx:=x @+ else xx:=-(x+1);
if xx<@"80 then
  begin #(o); @+ if x<0 then Incr(x)(@"100); @+ end
else  begin if xx<@"8000 then
    begin #(o+1); @+ if x<0 then Incr(x)(@"10000); @+ end
  else  begin if xx<@"800000 then
      begin #(o+2); @+ if x<0 then Incr(x)(@"1000000); @+ end
    else  begin #(o+3);
      if x>=0 then #(x div @"1000000)
      else  begin x:=@"7FFFFFFF-xx; #((x div @"1000000) + 128); @+ end;
      x:=x mod @"1000000;
      end;
    #(x div @"10000); x:=x mod @"10000;
    end;
  #(x div @"100); x:=x mod @"100;
  end;
#(x)
@f begin_signed == begin

@p procedure pckt_signed(@!o:eight_bits;@!x:int_32);
  {output |right|, |w|, |x|, |down|, |y|, or |z|}
@!begin_signed; pckt_room(5); comp_signed(append_byte);
end;

@ The |hex_packet| procedure prints the contents of a packet in
hexadecimal form.

@<Basic printing...@>=
@!debug procedure hex_packet(@!p:pckt_pointer); {prints a packet in hex}
var j,@!k,@!l:byte_pointer; {indices into |byte_mem|}
@!d:int_8u;
begin j:=pckt_start[p]-1; k:=pckt_start[p+1]-1;
print_ln(' packet=',p:1,' start=',j+1:1,' length=',k-j:1);
for l:=j+1 to k do
  begin d:=(bo(byte_mem[l])) div 16;
  if d<10 then print(xchr[d+"0"]) @+ else print(xchr[d-10+"A"]);
  d:=(bo(byte_mem[l])) mod 16;
  if d<10 then print(xchr[d+"0"]) @+ else print(xchr[d-10+"A"]);
  if (l=k)or(((l-j) mod 16)=0) then new_line
  else if ((l-j) mod 4)=0 then print('  ')
  else print(' ');
  end;
end;
gubed

@* File names.
The structure of file names is different for different systems; therefore
this part of the program will, in most cases, require system dependent
modifications. Here we assume that a file name consists of three parts:
an area or directory specifying where the file can be found, a name
proper and an extension; \.{\title} assumes that these three parts appear
in order stated above but this need not be true in all cases.

The font names extracted from \.{DVI} and \.{VF} files consist of an area
part and a name proper; these are stored as packets consisting of the
length of the area part followed by the area and the name proper.
When we print an external font name we simple print the area and the name
contained in the `file name packet' without delimiter between them.
This may need to be modified for some systems.
@^system dependencies@>

@<Basic printing...@>=
procedure print_font(@!f:font_number);
var p:pckt_pointer; {the font name packet}
@!k:byte_pointer; {index into |byte_mem|}
@!m:int_31; {font magnification}
begin print(' = '); p:=font_name(f);
for k:=pckt_start[p]+1 to pckt_start[p+1]-1 do
  print(xchr[bo(byte_mem[k])]);
m:=round((font_scaled(f)/font_design(f))*out_mag);
if m<>1000 then print(' scaled ',m:1);
end;

@ Before a font file can be opened for input we must build a string
with its external name.

@<Glob...@>=
@!cur_name:packed array[1..name_length] of char; {external name,
  with no lower case letters}
@!l_cur_name:int_15; {this many characters are actually relevant in
  |cur_name|}

@ For \.{TFM} and \.{VF} files we just append the appropriate extension
to the file name packet; in addition a system dependent area part
(usually different for \.{TFM} and \.{VF} files) is prepended if
the file name packet contains no area part.
@^system dependencies@>

@d append_to_name(#)==
  if l_cur_name<name_length then
    begin incr(l_cur_name); cur_name[l_cur_name]:=#;
    end
  else overflow(str_name_length,name_length)
@d make_font_name_end(#)==
  append_to_name(#[l]); make_name
@d make_font_name(#)==
  l_cur_name:=0; for l:=1 to # do make_font_name_end

@ For files with character raster data (e.g., \.{GF} or \.{PK} files) the
extension and\slash or area part will in most cases depend on the
resolution of the output device (corrected for font magnification).
If the special character |res_char| occurs in the extension and\slash or
default area, a character string representing the device resolution will
be substituted.
@^system dependencies@>

@d res_char=='?' {character to be replaced by font resolution}
@d res_ASCII="?" {|xord[res_char]|}
@#
@d append_res_to_name(#)==
  begin c:=#;
  @!device if c=res_char then
    for ll:=n_res_digits downto 1 do append_to_name(res_digits[ll])
  else ecived@;@/
  append_to_name(c);
  end
@d make_font_res_end(#)==
  append_res_to_name(#[l]); make_name
@d make_font_res(#)==
  make_res; l_cur_name:=0; for l:=1 to # do make_font_res_end

@ @<Glob...@>=
@!device
@!f_res:int_16u; {font resolution}
@!res_digits:array [1..5] of char;
@!n_res_digits:int_7; {number of significant characters in |res_digits|}
ecived

@ The |make_res| procedure creates a sequence of characters representing
to the font resolution |f_res|.

@p @!device procedure make_res;
var r:int_16u;
begin n_res_digits:=0; r:=f_res;
repeat incr(n_res_digits);
  res_digits[n_res_digits]:=xchr["0"+(r mod 10)]; r:=r div 10;
until r=0;
end;
ecived

@ The |make_name| procedure used to build the external file name. The
global variable |l_cur_name| contains the length of a default area
which has been copied to |cur_name| before |make_name| is called.
@^system dependencies@>

@p procedure make_name(@!e:pckt_pointer);
var b:eight_bits; {a byte extracted from |byte_mem|}
@!n:pckt_pointer; {file name packet}
@!cur_loc,@!cur_limit:byte_pointer; {indices into |byte_mem|}
@!device
@!ll:int_15; {loop index}
ecived@;@/
@!c:char; {a character to be appended to |cur_name|}
begin n:=font_name(cur_fnt);
cur_loc:=pckt_start[n]; cur_limit:=pckt_start[n+1];
pckt_extract(b); {length of area part}
if b>0 then l_cur_name:=0;
while cur_loc<cur_limit do
  begin pckt_extract(b);
  if (b>="a")and(b<="z") then Decr(b)(("a"-"A")); {convert to upper case}
  append_to_name(xchr[b]);
  end;
cur_loc:=pckt_start[e]; cur_limit:=pckt_start[e+1];
while cur_loc<cur_limit do
  begin pckt_extract(b); append_res_to_name(xchr[b]);
  end;
while l_cur_name<name_length do
  begin incr(l_cur_name); cur_name[l_cur_name]:=' ';
  end;
end;

@* Font data.
\.{DVI} file format does not include information about character widths, since
that would tend to make the files a lot longer. But a program that reads
a \.{DVI} file is supposed to know the widths of the characters that appear
in \\{set\_char} commands. Therefore \.{\title} looks at the font metric
(\.{TFM}) files for the fonts that are involved.
@.TFM {\rm files}@>

The character-width data appears also in other files (e.g., in \.{VF} files
or in \.{GF} and \.{PK} files that specify bit patterns for digitized
characters); thus, it is usually possible for \.{DVI} reading programs
to get by with accessing only one file per font. For \.{VF} reading
programs there is, however, a problem: (1)~when reading the character
packets from a \.{VF} file the \.{TFM} width for its local fonts should
be known in order to analyze and optimize the packets (e.g., determine
if a packet must indeed be enclosed with |push| and |pop| as implied by
the \.{VF} format); and (2)~ in order to avoid infinite recursion such
programs must not try to read a \.{VF} file for a font before a
character from that font is actually used. Thus \.{\title} reads the
\.{TFM} file whenever a new font is encountered and delays the decision
whether this is a virtual font or not.

@ First of all we need to know for each font~|f| such things as its
external name, design and scaled size, and the approximate size of
inter-word spaces. In addition we need to know the range |bc..ec| of
valid characters for this font, and for each character~|c| in~|f|  we
need to know if this character exists and if so what is the width of~|c|.
Depending on the font type of~|f| we may want to know a few other things
about character~|c| in~|f| such as the character packet from a \.{VF}
file or the raster data from a \.{PK} file.
@^font types@>

In \.{\title} we want to be able to handle the full range
|@t$-2^{31}$@><=c<@t$2^{31}$@>| of character codes; each character code
is decomposed into a character residue |0<=res<256| and character
extension |@t$-2^{23}$@><=ext<@t$2^{23}$@>| such that |c=256*ext+res|.
At present \.{VFtoVP}, \.{VPtoVF}, and the standard version of \TeX\ use
only characters in the range |0<=c<256| (i.e., |ext=0|), there are,
however, extensions of \TeX\ which use characters with |ext<>0|.
In any case characters with |ext<>0| will be used rather infrequently
and we want to handle this possibility without too much overhead.

Some of the data for each character~|c| depend only on its residue:
first of all its width and escapement; others, such as \.{VF} packets or
raster data will also depend on its extension. The later will be stored
as packets in |byte_mem|, and the packets for characters with the same
residue but different extension will be chained.

Thus we have to maintain several variables for each character
residue~|bc<=res<=ec| from each font~|f|; we store each type of variable
in a large array such that the array index |font_chars(f)+res| points to
the value for characters with residue |res| from font~|f|.

@ Quite often a particular width value is shared by several characters in
a font or even by characters from different fonts; the later will
probably occur in particular for virtual fonts and the local fonts used
by them. Thus the array |widths| is used to store all different \.{TFM}
width values of all legal characters in all fonts; a variable of type
|width_pointer| is an index into |widths| or is zero if a characters does
not exist.

In order to locate a given width value we use again a hash
table with simple chaining; this time the heads of the individual lists
appear in the |w_hash| array and the lists proceed through |w_link|
pointers.

@<Types...@>=
@!width_pointer=0..max_widths; {an index into |widths|}

@ @<Glob...@>=
@!widths:array[width_pointer] of int_32; {the different width values}
@!w_link:array[width_pointer] of width_pointer; {hash table}
@!w_hash:array[hash_code] of width_pointer;
@!n_widths:width_pointer; {first unoccupied position in |widths|}

@ Initially the |widths| array and all the hash lists are empty, except
for one entry: the width value zero; in addition we set |widths[0]:=0|.

@d invalid_width=0 {width pointer for invalid characters}
@d zero_width=1 {a width pointer to the value zero}

@<Set init...@>=
w_hash[0]:=1; w_link[1]:=0; widths[0]:=0; widths[1]:=0; n_widths:=2;
for h:=1 to hash_size-1 do w_hash[h]:=0;

@ The |make_width| function returns an index into |widths| and, if
necessary, adds a new width value; thus two characters will have the
same |width_pointer| if and only if their widths agree.

@p function make_width(@!w:int_32):width_pointer;
label found;
var h:hash_code; {hash code}
@!p:width_pointer; {where the identifier is being sought}
@!x:int_16; {intermediate value}
begin widths[n_widths]:=w;
@<Compute the width hash code |h|@>;
@<Compute the width location |p|, |goto| found unless the value is new@>;
if n_widths=max_widths then overflow(str_widths,max_widths);
incr(n_widths);
found:make_width:=p;
end;

@ A simple hash code is used: If the width value consists of the four
bytes $b_0b_1b_2b_3$, its hash value will be
$$(8*b_0+4*b_1+2*b_2+b_3)\,\bmod\,|hash_size|.$$

@<Compute the width hash...@>=
if w>=0 then x:=w div @"1000000
else  begin w:=w+@"40000000; w:=w+@"40000000; x:=(w div @"1000000)+@"80;
  end;
w:=w mod @"1000000; x:=x+x+(w div @"10000);
w:=w mod @"10000; x:=x+x+(w div @"100);
h:=(x+x+(w mod @"100)) mod hash_size

@ If the width is new, it has been placed into position |p=n_widths|,
otherwise |p| will point to its existing location.

@<Compute the width location...@>=
p:=w_hash[h];
while p<>0 do
  begin if widths[p]=widths[n_widths] then goto found;
  p:=w_link[p];
  end;
p:=n_widths; {the current width is new}
w_link[p]:=w_hash[h]; w_hash[h]:=p {insert |p| at beginning of hash list}

@ The |char_widths| array is used to store the |width_pointer|s for all
different characters among all fonts.  The |char_packets| array is used
to store the |pckt_pointer|s for all different characters among all
fonts; they can point to character packets from \.{VF} files or, e.g.,
raster packets from \.{PK} files.

@<Types...@>=
@!char_offset=-255..max_chars; {|char_pointer| offset for a font}
@!char_pointer=0..max_chars; {index into |char_widths| or similar arrays}

@ @<Glob...@>=
@!char_widths:array[char_pointer] of width_pointer; {width pointers}
@!char_packets:array[char_pointer] of pckt_pointer; {packet pointers}
@!n_chars:char_pointer; {first unused position in |char_widths|}

@ @<Set init...@>=
n_chars:=0;

@ The current number of known fonts is |nf|; each known font has an
internal number |f|, where |0<=f<nf|.  For the moment we need for each
known font:  |font_check|, |font_scaled|, |font_design|, |font_name|,
|font_bc|, |font_ec|, |font_chars|, and |font_type|.  Here |font_scaled|
and |font_design| are measured in \.{DVI} units and |font_chars| is of
type |char_offset|:  the width pointer for character~|c| of the font is
stored in |char_widths[char_offset+c]| (for |font_bc<=c<=font_ec|).
Later on we will need additional information depending on the font type:
\.{VF} or real (\.{GF}, \.{PK}, or \.{PXL}).

@<Types...@>=
@!f_type=defined_font..max_font_type; {type of a font}
@!font_number=0..max_fonts;

@ @<Glob...@>=
@!nf:font_number;

@ These data are stored in several arrays and we use \.{WEB} macros
to access the various fields. Thus it would be simple to store the
data in an array of record structures and adapt the \.{WEB} macros
accordingly.

We will say, e.g., |font_name(f)| for the name field of font~|f|, and
|font_width(f)(c)| for the width pointer of character~|c| in font~|f|
and |font_packet(f)(c)| for its character packet (this character
exists provided |font_bc(f)<=c<=font_ec(f)| and
|font_width(f)(c)<>invalid_width|). The actual width of character~|c| in
font~|f| is stored in |widths[font_width(f)(c)]|.

@d font_check(#)==fnt_check[#] {checksum}
@d font_scaled(#)==fnt_scaled[#] {scaled or `at' size}
@d font_design(#)==fnt_design[#] {design size}
@d font_name(#)==fnt_name[#] {area plus name packet}
@d font_bc(#)==fnt_bc[#] {first character}
@d font_ec(#)==fnt_ec[#] {last character}
@d font_chars(#)==fnt_chars[#] {character info offset}
@d font_type(#)==fnt_type[#] {type of this font}
@d font_font(#)==fnt_font[#] {use depends on |font_type|}
@#
@d font_width_end(#)==#]
@d font_width(#)==char_widths[font_chars(#)+font_width_end
@d font_packet(#)==char_packets[font_chars(#)+font_width_end

@<Glob...@>=
@!fnt_check:array [font_number] of int_32; {checksum}
@!fnt_scaled:array [font_number] of int_31; {scaled size}
@!fnt_design:array [font_number] of int_31; {design size}
@!device @<Declare device dependent font data arrays@>@; @+ ecived @; @/
@!fnt_name:array [font_number] of pckt_pointer; {pointer to area plus
  name packet}
@!fnt_bc:array [font_number] of eight_bits; {first character}
@!fnt_ec:array [font_number] of eight_bits; {last character}
@!fnt_chars:array [font_number] of char_offset; {character info offset}
@!fnt_type:array [font_number] of f_type; {type of font}
@!fnt_font:array [font_number] of font_number; {use depends on |font_type|}

@ @d invalid_font==max_fonts {used when there is no valid font}

@<Set init...@>=
@!device @<Initialize device dependent font data@>@; @+ ecived @;@/
nf:=0;

@ A \.{VF}, or \.{GF}, or \.{PK} file may contain information for
several characters with the same residue but with different extension;
all except the first of the corresponding packets in |byte_mem| will
contain a pointer to the previous one and |font_packet(f)(res)|
identifies the last such packet.

A character packet in |byte_mem| starts with a flag byte
$$\hbox{|flag=@"40*ext_flag+@"20*chain_flag+type_flag|}$$
with |0<=ext_flag<=3|, |0<=chain_flag<=1|, |0<=type_flag<=@"1F|,
followed by |ext_flag| bytes with the character extension for this
packet and, if |chain_flag=1|, by a two byte packet pointer to the
previous packet for the same font and character residue. The actual
character packet follows after these header bytes and the
interpretation of the |type_flag| depends on whether this is a \.{VF}
packet or a packet for raster data.

The empty packet is interpreted as a special case of a packet with
|flag=0|.

@d ext_flag=@"40
@d chain_flag=@"20

@<Types...@>=
@!type_flag=0..chain_flag-1; {the range of values for the |type_flag|}

@ The global variable |cur_fnt| is the internal font number of the
currently selected font, or equals |invalid_font| if no font has
been selected; |cur_res| and |cur_ext| are the residue and extension
part of the current character code. The type of a character packet
located by the |find_packet| function defined below is |cur_type|.
While building a character packet for a character, |pckt_ext| and
|pckt_res| are the extension and residue of this character; |pckt_dup|
indicates whether a packet for this extension exists already.

@<Glob...@>=
@!cur_fnt:font_number; {the currently selected font}
@!cur_ext:int_24; {the current character extension}
@!cur_res:int_8u; {the current character residue}
@!cur_type:type_flag; {type of the current character packet}
@!pckt_ext:int_24; {character extension for the current character packet}
@!pckt_res:int_8u; {character residue for the current character packet}
@!pckt_dup:boolean; {is there a previous packet for the same extension?}
@!pckt_prev:pckt_pointer; {a previous packet for the same extension}
@!pckt_m_msg,@!pckt_s_msg,@!pckt_d_msg:int_7; {counts for various character
  packet error messages}

@ @<Set init...@>=
cur_fnt:=invalid_font; pckt_m_msg:=0; pckt_s_msg:=0; pckt_d_msg:=0;

@ The |find_packet| functions is used to locate the character packet for
the character with residue~|cur_res| and extension~|cur_ext| from
font~|cur_fnt| and returns |false| if no packet exists for any extension;
otherwise the result is |true| and the global variables |cur_packet|,
|cur_type|, |cur_loc|, and |cur_limit| are initialized. In case none of
the character packets has the correct extension, the last one in the
chain (the one defined first) is used instead and |cur_ext| is changed
accordingly.

@p function find_packet:boolean;
label found,exit;
var p,@!q:pckt_pointer; {current and next packet}
@!f:eight_bits; {a flag byte}
@!e:int_24; {extension for a packet}
begin q:=font_packet(cur_fnt)(cur_res);
if q=invalid_packet then
  begin if pckt_m_msg<10 then {stop telling after first 10 times}
    begin print_ln('---missing character packet for character ',cur_res:1,
@.missing character packet...@>
      ' font ',cur_fnt:1);
    incr(pckt_m_msg); mark_error;
    if pckt_m_msg=10 then print_ln('---further messages suppressed.');
    end;
  find_packet:=false; return;
  end;
@<Locate a character packet and |goto found| if found@>;
if pckt_s_msg<10 then {stop telling after first 10 times}
  begin print_ln('---substituted character packet with extension ',
@.substituted character packet...@>
    e:1,' instead of ',cur_ext:1,' for character ',cur_res:1,
    ' font ',cur_fnt:1);
  incr(pckt_s_msg); mark_error;
  if pckt_s_msg=10 then print_ln('---further messages suppressed.');
  end;
cur_ext:=e;
found: cur_pckt:=p; cur_type:=f; find_packet:=true;
exit: end;

@ @<Locate a character packet and |goto found| if found@>=
repeat p:=q; q:=invalid_packet;
  cur_loc:=pckt_start[p]; cur_limit:=pckt_start[p+1];
  if p=empty_packet then
    begin e:=0; f:=0;
    end
  else  begin pckt_extract(f);
    case (f div ext_flag) of
    0: e:=0;
    1: e:=pckt_ubyte;
    2: e:=pckt_upair;
    othercases e:=pckt_strio; {|f div ext_flag = 3|}
    endcases;
    if (f mod ext_flag)>=chain_flag then q:=pckt_upair;
    f:=f mod chain_flag;
    end;
  if e=cur_ext then goto found;
until q=invalid_packet

@ The |start_packet| procedure is used to create the header bytes of a
character packet for the character with residue~|cur_res| and
extension~|cur_ext| from font~|cur_fnt|; if a previous such packet
exists, we try to build an exact duplicate, i.e., use the chain field of
that previous packet.

@p procedure start_packet(@!t:type_flag);
label found,not_found;
var p,@!q:pckt_pointer; {current and next packet}
@!f:int_8u; {a flag byte}
@!e:int_32; {extension for a packet}
@!cur_loc: byte_pointer; {current location in a packet}
@!cur_limit: byte_pointer; {start of next packet}
begin q:=font_packet(cur_fnt)(cur_res);
if q<>invalid_packet then @<Locate a character packet...@>;
q:=font_packet(cur_fnt)(cur_res); pckt_dup:=false; goto not_found;
found: pckt_dup:=true; pckt_prev:=p;
not_found: pckt_ext:=cur_ext; pckt_res:=cur_res; pckt_room(6);
@!debug if byte_ptr<>pckt_start[pckt_ptr] then confusion(str_packets);
gubed @;@/
if q=invalid_packet then f:=t @+ else f:=t+chain_flag;
e:=cur_ext;
if e<0 then Incr(e)(@"1000000);
if e=0 then append_byte(f) @+ else @;
  begin if e<@"100 then append_byte(f+ext_flag) @+ else @;
    begin if e<@"10000 then append_byte(f+ext_flag+ext_flag) @+ else @;
      begin append_byte(f+ext_flag+ext_flag+ext_flag);
      append_byte(e div @"10000); e:=e mod @"10000;
      end;
    append_byte(e div @"100); e:=e mod @"100;
    end;
  append_byte(e);
  end;
if q<>invalid_packet then
  begin append_byte(q div @"100); append_byte(q mod @"100);
  end;
end;

@ The |build_packet| procedure is used to finish a character packet.
If a previous packet for the same character extension exists, the new
one is discarded; if the two packets are identical, as it occasionally
occurs for raster files, this is done without an error message.

@p procedure build_packet;
var k,@!l:byte_pointer; {indices into |byte_mem|}
begin if pckt_dup then
  begin k:=pckt_start[pckt_prev+1]; l:=pckt_start[pckt_ptr];
  if (byte_ptr-l)<>(k-pckt_start[pckt_prev]) then pckt_dup:=false;
  while pckt_dup and(byte_ptr>l) do
    begin flush_byte; decr(k);
    if byte_mem[byte_ptr]<>byte_mem[k] then pckt_dup:=false;
    end;
  if (not pckt_dup)and(pckt_d_msg<10) then {stop telling after first 10 times}
    begin print('---duplicate packet for character ',pckt_res:1);
@.duplicate packet for character...@>
    if pckt_ext<>0 then print('.',pckt_ext:1);
    print_ln(' font ',cur_fnt:1);
    incr(pckt_d_msg); mark_error;
    if pckt_d_msg=10 then print_ln('---further messages suppressed.');
    end;
  byte_ptr:=l;
  end
else font_packet(cur_fnt)(pckt_res):=make_packet;
end;

@* Defining fonts.
A detailed description of the \.{TFM} file format can be found in the
documentation of \TeX, \MF, or \.{TFtoPL}.  In order to read \.{TFM}
files the program uses the binary file variable |tfm_file|.

@<Glob...@>=
@!tfm_file:byte_file; {a \.{TFM} file}
@!tfm_ext:pckt_pointer; {extension for \.{TFM} files}

@ @<Initialize predefined strings@>=
id4(".")("T")("F")("M")(tfm_ext); {file name extension for \.{TFM} files}

@ If no font directory has been specified, \.{\title} is supposed to use
the default \.{TFM} directory, which is a system-dependent place where
the \.{TFM} files for standard fonts are kept.
The string variable |TFM_default_area| contains the name of this area.
@^system dependencies@>

@d TFM_default_area_name=='TeXfonts:' {change this to the correct name}
@d TFM_default_area_name_length=9 {change this to the correct length}

@<Glob...@>=
@!TFM_default_area:packed array[1..TFM_default_area_name_length] of char;

@ @<Set init...@>=
TFM_default_area:=TFM_default_area_name;

@ If a \.{TFM} file is badly malformed, we say |bad_font|; for a \.{TFM}
file the |bad_tfm| procedure is used to give an error message which
refers the user to \.{TFtoPL} and \.{PLtoTF}, and terminates \.{\title}.

@<Error handling...@>=
procedure bad_tfm;
begin print('Bad TFM file'); print_font(cur_fnt); print_ln('!');
@.Bad TFM file@>
abort('Use TFtoPL/PLtoTF to diagnose and correct the problem');
@.Use TFtoPL/PLtoTF@>
end;
@#
procedure bad_font;
begin new_line;
case font_type(cur_fnt) of
  defined_font: confusion(str_fonts);
  loaded_font: bad_tfm;
  @<Cases for |bad_font|@>@;@/
  othercases abort('internal error');
  endcases;
end;

@ To prepare |tfm_file| for input we |reset| it.

@<TFM: Open |tfm_file|@>=
make_font_name(TFM_default_area_name_length)(TFM_default_area)(tfm_ext);
reset(tfm_file,cur_name);
if eof(tfm_file) then
@^system dependencies@>
  abort('---not loaded, TFM file can''t be opened!')
@.TFM file can\'t be opened@>

@ It turns out to be convenient to read four bytes at a time, when we
are inputting from \.{TFM} files. The input goes into global variables
|tfm_b0|, |tfm_b1|, |tfm_b2|, and |tfm_b3|, with |tfm_b0| getting the
first byte and |tfm_b3| the fourth.

@<Glob...@>=
@!tfm_b0,@!tfm_b1,@!tfm_b2,@!tfm_b3: eight_bits; {four bytes input at once}

@ Reading a \.{TFM} file should be done as efficient as possible for a
particular system; on many systems this means that a large number of
bytes from |tfm_file| is read into a buffer and will then be extracted
from that buffer. In order to simplify such system dependent changes
we use the \.{WEB} macro |tfm_byte| to extract the next \.{TFM} byte;
this macro and |eof(tfm_file)| are used only in the |read_tfm_word|
procedure which sets |tfm_b0| through |tfm_b3| to the next four bytes
in the current \.{TFM} file. Here we give simple minded definitions in
terms of standard \PASCAL.
@^system dependencies@>
@^optimization@>

@d tfm_byte(#)==read(tfm_file,#) {read next \.{TFM} byte}

@p procedure read_tfm_word;
begin tfm_byte(tfm_b0); tfm_byte(tfm_b1);
tfm_byte(tfm_b2); tfm_byte(tfm_b3);
if eof(tfm_file) then bad_font;
end;

@ Here are three procedures used to check the consistency of font files:
First, the |check_check_sum| procedure compares two check sum values: a
warning is given if they differ and are both non-zero; if the second
value is not zero it may replace the first one.
Next, the |check_design_size| procedure compares two design size
values: a warning is given if they differ by more than a small amount.
Finally, the |check_width| function compares the character width value
for character |cur_res| read from a \.{VF} or raster file for font
|cur_fnt| with the value previously read from the \.{TFM} file and
returns the width pointer for that value; a warning is given if the two
values differ.

@p procedure check_check_sum(@!c:int_32;@!u:boolean);
  {compare |font_check(cur_fnt)| with |c|}
begin if (c<>font_check(cur_fnt))and(c<>0) then
  begin
  if font_check(cur_fnt)<>0 then
    begin new_line; print_ln('---beware: check sums do not agree!   (',
@.beware: check sums do not agree@>
@.check sums do not agree@>
      c:1,' vs. ',font_check(cur_fnt):1,')');
    mark_harmless;
    end;
  if u then font_check(cur_fnt):=c;
  end;
end;
@#
procedure check_design_size(@!d:int_32);
  {compare |font_design(cur_fnt)| with |d|}
begin if abs(d-font_design(cur_fnt))>2 then
  begin new_line; print_ln('---beware: design sizes do not agree!   (',
@.beware: design sizes do not agree@>
@.design sizes do not agree@>
    d:1,' vs. ',font_design(cur_fnt):1,')');
  mark_error;
  end;
end;
@#
function check_width(w:int_32):width_pointer;
  {compare |widths[font_width(cur_fnt)(cur_res)]| with |w|}
var wp:width_pointer; {pointer to \.{TFM} width value}
begin if (cur_res>=font_bc(cur_fnt))and(cur_res<=font_ec(cur_fnt)) then
  wp:=font_width(cur_fnt)(cur_res)
else wp:=invalid_width;
if wp=invalid_width then
  begin print_nl('Bad char ',cur_res:1);
@.Bad char c@>
  if cur_ext<>0 then print('.',cur_ext:1);
  print(' font ',cur_fnt:1); print_font(cur_fnt);
  abort(' (compare TFM file)');
  end;
if w<>widths[wp] then
  begin new_line; print_ln('---beware: char widths do not agree!   (',
@.beware: char widths do not agree@>
@.char widths do not agree@>
    w:1,' vs. ',widths[wp]:1,')');
  mark_error;
  end;
check_width:=wp;
end;

@ The |load_font| procedure reads the \.{TFM} file for a font and puts
the data extracted into position |cur_fnt| of the font data arrays.

@p procedure load_font; {reads a \.{TFM} file}
var l:int_16; {loop index}
@!p:char_pointer; {index into |char_widths|}
@!q:width_pointer; {index into |widths|}
@!bc,@!ec:int_15; {first and last character in this font}
@!lh:int_15; {length of header in four byte words}
@!nw:int_15; {number of words in width table}
@!w:int_32; {a four byte integer}
@<Variables for scaling computation@>@;
begin print('TFM: font ',cur_fnt:1); print_font(cur_fnt);
font_type(cur_fnt):=loaded_font;
@<TFM: Open |tfm_file|@>;
@<TFM: Read past the header data@>;
@<TFM: Store character-width indices@>;
@<TFM: Read and convert the width values@>;
@<TFM: Convert character-width indices to character-width pointers@>;
close_in(tfm_file);
@!device @<Initialize device dependent data for a font@>@; @+ ecived @; @/
d_print(' loaded at ',font_scaled(cur_fnt):1,' DVI units');
print_ln('.');
end;

@ @<Glob...@>=
@!tfm_conv:real; {\.{DVI} units per absolute \.{TFM} unit}

@ We will use the following \.{WEB} macros to construct integers from
two or four of the four bytes read by |read_tfm_word|.
@^system dependencies@>

@d tfm_b01(#)== {|tfm_b0..tfm_b1| as non-negative integer}
if tfm_b0>127 then bad_font
else #:=tfm_b0*256+tfm_b1
@d tfm_b23(#)== {|tfm_b2..tfm_b3| as non-negative integer}
if tfm_b2>127 then bad_font
else #:=tfm_b2*256+tfm_b3
@d tfm_squad(#)== {|tfm_b0..tfm_b3| as signed integer}
if tfm_b0<128 then #:=((tfm_b0*256+tfm_b1)*256+tfm_b2)*256+tfm_b3
else #:=(((tfm_b0-256)*256+tfm_b1)*256+tfm_b2)*256+tfm_b3
@d tfm_uquad== {|tfm_b0..tfm_b3| as unsigned integer}
(((tfm_b0*256+tfm_b1)*256+tfm_b2)*256+tfm_b3)

@<TFM: Read past the header data@>=
read_tfm_word; tfm_b23(lh);
read_tfm_word; tfm_b01(bc); tfm_b23(ec);
if ec<bc then
  begin bc:=1; ec:=0;
  end
else if ec>255 then bad_font;
read_tfm_word; tfm_b01(nw);
if (nw=0)or(nw>256) then bad_font;
for l:=-2 to lh do
  begin read_tfm_word;
  if l=1 then
    begin tfm_squad(w); check_check_sum(w,true);
    end
  else if l=2 then
    begin if tfm_b0>127 then bad_font;
    check_design_size(round(tfm_conv*tfm_uquad));
    end;
  end

@ The width indices for the characters are stored in positions |n_chars|
through |n_chars-bc+ec| of the |char_widths| array; if characters on
either end of the range |bc..ec| do not exist, they are ignored and the
range is adjusted accordingly.

@<TFM: Store character-width indices@>=
read_tfm_word;
while (tfm_b0=0)and(bc<=ec) do
  begin incr(bc); read_tfm_word;
  end;
font_bc(cur_fnt):=bc; font_chars(cur_fnt):=n_chars-bc;
if ec>=max_chars-font_chars(cur_fnt) then overflow(str_chars,max_chars);
for l:=bc to ec do
  begin char_widths[n_chars]:=tfm_b0; incr(n_chars); read_tfm_word;
  end;
while (char_widths[n_chars-1]=0)and(ec>=bc) do
  begin decr(n_chars); decr(ec);
  end;
font_ec(cur_fnt):=ec

@ The most important part of |load_font| is the width computation, which
involves multiplying the relative widths in the \.{TFM} file by the
scaling factor in the \.{DVI} file. A similar computation is used for
dimensions read from \.{VF} files. This fixed-point multiplication must
be done with precisely the same accuracy by all \.{DVI}-reading programs,
in order to validate the assumptions made by \.{DVI}-writing programs
like \TeX82.

Let us therefore summarize what needs to be done. Each width in a \.{TFM}
file appears as a four-byte quantity called a |fix_word|.  A |fix_word|
whose respective bytes are $(a,b,c,d)$ represents the number
$$x=\left\{\vcenter{\halign{$#$,\hfil\qquad&if $#$\hfil\cr
b\cdot2^{-4}+c\cdot2^{-12}+d\cdot2^{-20}&a=0;\cr
-16+b\cdot2^{-4}+c\cdot2^{-12}+d\cdot2^{-20}&a=255.\cr}}\right.$$
(No other choices of $a$ are allowed, since the magnitude of a \.{TFM}
dimension must be less than 16.)  We want to multiply this quantity by the
integer~|z|, which is known to be less than $2^{27}$.
If $|z|<2^{23}$, the individual multiplications $b\cdot z$, $c\cdot z$,
$d\cdot z$ cannot overflow; otherwise we will divide |z| by 2, 4, 8, or
16, to obtain a multiplier less than $2^{23}$, and we can compensate for
this later. If |z| has thereby been replaced by $|z|^\prime=|z|/2^e$, let
$\beta=2^{4-e}$; we shall compute
$$\lfloor(b+c\cdot2^{-8}+d\cdot2^{-16})\,z^\prime/\beta\rfloor$$ if $a=0$,
or the same quantity minus $\alpha=2^{4+e}z^\prime$ if $a=255$.
This calculation must be done exactly, for the reasons stated above; the
following program does the job in a system-independent way, assuming
that arithmetic is exact on numbers less than $2^{31}$ in magnitude. We
use \.{WEB} macros for various versions of this computation.
@^system dependencies@>
@^optimization@>

@d tfm_fix3u== {convert |tfm_b1..tfm_b3| to an unsigned scaled dimension}
(((((tfm_b3*z)div@'400)+(tfm_b2*z))div@'400)+(tfm_b1*z))div beta
@#
@d tfm_fix4(#)== {convert |tfm_b0..tfm_b3| to a scaled dimension}
  #:=tfm_fix3u;
  if tfm_b0>0 then if tfm_b0=255 then Decr(#)(alpha) else bad_font
@d tfm_fix3(#)== {convert |tfm_b1..tfm_b3| to a scaled dimension}
  #:=tfm_fix3u; @+ if tfm_b1>127 then Decr(#)(alpha)
@d tfm_fix2== {convert |tfm_b2..tfm_b3| to a scaled dimension}
  if tfm_b2>127 then tfm_b1:=255 else tfm_b1:=0;
  tfm_fix3
@d tfm_fix1== {convert |tfm_b3| to a scaled dimension}
  if tfm_b3>127 then tfm_b1:=255 else tfm_b1:=0;
  tfm_b2:=tfm_b1; tfm_fix3

@<Variables for scaling computation@>=
@!z:int_32; {multiplier}
@!alpha:int_32; {correction for negative values}
@!beta:int_15; {divisor}

@ @<Replace |z| by $|z|^\prime$ and compute $\alpha,\beta$@>=
alpha:=16;
while z>=@'40000000 do
  begin z:=z div 2; alpha:=alpha+alpha;
  end;
beta:=256 div alpha; alpha:=alpha*z

@ The first width value, which indicates that a character does not exist
and which must vanish, is converted to |invalid_width|; the other width
values are scaled by |font_scaled(cur_fnt)| and converted to width
pointers by |make_width|. The resulting width pointers are stored
temporarily in the |char_widths| array, following the with indices.

@<TFM: Read and convert the width values@>=
if nw-1>max_chars-n_chars then overflow(str_chars,max_chars);
if (tfm_b0<>0)or(tfm_b1<>0)or(tfm_b2<>0)or(tfm_b3<>0) then bad_font
  else char_widths[n_chars]:=invalid_width;
z:=font_scaled(cur_fnt);
@<Replace |z|...@>;
for p:=n_chars+1 to n_chars+nw-1 do
  begin read_tfm_word; tfm_fix4(w);
  char_widths[p]:=make_width(w);
  end

@ We simply translate the width indices into width pointers. In addition
we initialize the character packets with the invalid packet.

@<TFM: Convert character-width indices to character-width pointers@>=
for p:=font_chars(cur_fnt)+bc to n_chars-1 do
  begin q:=char_widths[n_chars+char_widths[p]]; char_widths[p]:=q;
  char_packets[p]:=invalid_packet;
  end

@ When processing a font definition we put the data extracted from the
\.{DVI} or \.{VF} file into position |nf| of the font data arrays and
call |define_font| to obtain the internal font number for this font.
The parameter |load| is true if the \.{TFM} file should be loaded.

@p function define_font(@!load:boolean):font_number;
var save_fnt:font_number; {used to save |cur_fnt|}
begin save_fnt:=cur_fnt; {save}
cur_fnt:=0;
while (font_name(cur_fnt)<>font_name(nf))or@|
  (font_scaled(cur_fnt)<>font_scaled(nf)) do incr(cur_fnt);
d_print(' => ',cur_fnt:1); print_font(cur_fnt);
if cur_fnt<nf then
  begin check_check_sum(font_check(nf),true);
  check_design_size(font_design(nf));
  @!debug if font_type(cur_fnt)=defined_font then print(' defined')
  else print(' loaded');
  print(' previously');
  gubed@;
  end
else  begin if nf=max_fonts then overflow(str_fonts,max_fonts);
  incr(nf); font_font(cur_fnt):=invalid_font;
  font_type(cur_fnt):=defined_font;
  d_print(' defined');
  end;
print_ln('.');
if load and(font_type(cur_fnt)=defined_font) then load_font;
define_font:=cur_fnt;
cur_fnt:=save_fnt; {restore}
end;

@* Low-level DVI input routines.
The program uses the binary file variable |dvi_file| for its main input
file; |dvi_loc| is the number of the byte about to be read next from
|dvi_file|.

@<Glob...@>=
@!dvi_file:byte_file; {the stuff we are \.{\title}ing}
@!dvi_loc:int_32; {where we are about to look, in |dvi_file|}

@ If the \.{DVI} file is badly malformed, we say |bad_dvi|; this
procedure gives an error message which refers the user to \.{DVItype},
and terminates \.{\title}.

@<Error handling...@>=
procedure bad_dvi;
begin new_line; print_ln('Bad DVI file: loc=',dvi_loc:1,'!');
@.Bad DVI file@>
print(' Use DVItype with output level');
@.Use DVItype@>
if random_reading then print('=4') @+ else print('<4');
abort('to diagnose the problem');
end;

@ To prepare |dvi_file| for input, we |reset| it.

@<Open input file(s)@>=
reset(dvi_file); {prepares to read packed bytes from |dvi_file|}
dvi_loc:=0;

@ Reading the \.{DVI} file should be done as efficient as possible for a
particular system; on many systems this means that a large number of
bytes from |dvi_file| is read into a buffer and will then be extracted
from that buffer. In order to simplify such system dependent changes
we use a pair of \.{WEB} macros: |dvi_byte| extracts the next \.{DVI}
byte and |dvi_eof| is |true| if we have reached the end of the \.{DVI}
file. Here we give simple minded definitions for these macros in terms
of standard \PASCAL.
@^system dependencies@>
@^optimization@>

@d dvi_eof == eof(dvi_file) {has the \.{DVI} file been exhausted?}
@d dvi_byte(#) ==
  if dvi_eof then bad_dvi
  else read(dvi_file,#) {obtain next \.{DVI} byte}

@ Next we come to the routines that are used only if |random_reading|    is
|true|. The driver program below needs two such routines: |dvi_length| should
compute the total number of bytes in |dvi_file|, possibly also
causing |eof(dvi_file)| to be true; and |dvi_move(n)| should position
|dvi_file| so that the next |dvi_byte| will read byte |n|, starting with
|n=0| for the first byte in the file.
@^system dependencies@>

Such routines are, of course, highly system dependent. They are implemented
here in terms of two assumed system routines called |set_pos| and |cur_pos|.
The call |set_pos(f,n)| moves to item |n| in file |f|, unless |n| is
negative or larger than the total number of items in |f|; in the latter
case, |set_pos(f,n)| moves to the end of file |f|.
The call |cur_pos(f)| gives the total number of items in |f|, if
|eof(f)| is true; we use |cur_pos| only in such a situation.

@p function dvi_length:int_32;
begin set_pos(dvi_file,-1); dvi_length:=cur_pos(dvi_file);
end;
@#
procedure dvi_move(@!n:int_32);
begin set_pos(dvi_file,n); dvi_loc:=n;
end;

@ We need seven simple functions to read the next byte or bytes
from |dvi_file|.

@p function dvi_sbyte:int_8; {returns the next byte, signed}
@!begin_byte(dvi_byte); incr(dvi_loc); comp_sbyte(dvi_sbyte);
end;
@#
function dvi_ubyte:int_8u; {returns the next byte, unsigned}
@!begin_byte(dvi_byte); incr(dvi_loc); comp_ubyte(dvi_ubyte);
end;
@#
function dvi_spair:int_16; {returns the next two bytes, signed}
@!begin_pair(dvi_byte); Incr(dvi_loc)(2); comp_spair(dvi_spair);
end;
@#
function dvi_upair:int_16u; {returns the next two bytes, unsigned}
@!begin_pair(dvi_byte); Incr(dvi_loc)(2); comp_upair(dvi_upair);
end;
@#
function dvi_strio:int_24; {returns the next three bytes, signed}
@!begin_trio(dvi_byte); Incr(dvi_loc)(3); comp_strio(dvi_strio);
end;
@#
function dvi_utrio:int_24u; {returns the next three bytes, unsigned}
@!begin_trio(dvi_byte); Incr(dvi_loc)(3); comp_utrio(dvi_utrio);
end;
@#
function dvi_squad:int_32; {returns the next four bytes, signed}
@!begin_quad(dvi_byte); Incr(dvi_loc)(4); comp_squad(dvi_squad);
end;

@ Three other functions are used in cases where a four byte integer
(which is always signed) must have a non-negative value, a positive
value, or is a pointer which must be either positive or |=-1|.

@p function dvi_uquad:int_31; {result must be non-negative}
var x:int_32;
begin x:=dvi_squad; if x<0 then bad_dvi
else dvi_uquad:=x;
end;
@#
function dvi_pquad:int_31; {result must be positive}
var x:int_32;
begin x:=dvi_squad; if x<=0 then bad_dvi
else dvi_pquad:=x;
end;
@#
function dvi_pointer:int_32; {result must be positive or |=-1|}
var x:int_32;
begin x:=dvi_squad; if (x<=0)and(x<>-1) then bad_dvi
else dvi_pointer:=x;
end;

@ Given the structure of the \.{DVI} commands it is fairly obvious
that their interpretation consists of two steps: First zero to four
bytes are read in order to obtain the value of the first parameter
(e.g., zero bytes for |set_char_0|, four bytes for |set4|); then,
depending on the command class, a specific action is performed (e.g.,
typeset a character but don't move the reference point for |put1..put4|).

The \.{DVItype} program uses large case statements for both steps;
unfortunately some \PASCAL\ compilers fail to implement large case
statements efficiently -- in particular those as the one used in the
|first_par| function of \.{DVItype}. Here we use a pair of look up tables:
|dvi_par| determines how to obtain the value of the first parameter, and
|dvi_cl| determines the command class.

A slight complication arises from the fact that we want to decompose the
character code of each character to be typeset into a residue
|0<=char_res<256| and extension: |char_code=char_res+256*char_ext|;
the \.{TFM} widths as well as the pixel widths for a given resolution
are the same for all characters in a font with the same residue.

@d two_cases(#)==#,#+1
@d three_cases(#)==#,#+1,#+2
@d five_cases(#)==#,#+1,#+2,#+3,#+4

@ First we define the values used as array elements of |dvi_par|; we
distinguish between pure numbers and dimensions because dimensions read
from a \.{VF} file must be scaled.

@d char_par=0 {character for \\{set} and |put|}
@d no_par=1 {no parameter}
@d dim1_par=2 {one-byte signed dimension}
@d num1_par=3 {one-byte unsigned number}
@d dim2_par=4 {two-byte signed dimension}
@d num2_par=5 {two-byte unsigned number}
@d dim3_par=6 {three-byte signed dimension}
@d num3_par=7 {three-byte unsigned number}
@d dim4_par=8 {four-byte signed dimension}
@d num4_par=9 {four-byte signed number}
@d numu_par=10 {four-byte non-negative number}
@d rule_par=11 {dimensions for |set_rule| and |put_rule|}
@d fnt_par=12 {font for |fnt_num| commands}
@d max_par=12 {largest possible value}

@<Types...@>=
@!cmd_par=char_par..max_par;

@ Here we declare the array |dvi_par|.

@<Globals...@>=
@!dvi_par:packed array [eight_bits] of cmd_par;

@ And here we initialize it.

@<Set init...@>=
for i:=0 to put1+3 do dvi_par[i]:=char_par;@/
for i:=nop to 255 do dvi_par[i]:=no_par;@/
dvi_par[set_rule]:=rule_par; dvi_par[put_rule]:=rule_par;@/
dvi_par[right1]:=dim1_par; dvi_par[right1+1]:=dim2_par;
dvi_par[right1+2]:=dim3_par; dvi_par[right1+3]:=dim4_par;@/
for i:=fnt_num_0 to fnt_num_0+63 do dvi_par[i]:=fnt_par;@/
dvi_par[fnt1]:=num1_par; dvi_par[fnt1+1]:=num2_par;
dvi_par[fnt1+2]:=num3_par; dvi_par[fnt1+3]:=num4_par;@/
dvi_par[xxx1]:=num1_par; dvi_par[xxx1+1]:=num2_par;
dvi_par[xxx1+2]:=num3_par; dvi_par[xxx1+3]:=numu_par;@/
for i:=0 to 3 do
  begin dvi_par[i+w1]:=dvi_par[i+right1];
  dvi_par[i+x1]:=dvi_par[i+right1];
  dvi_par[i+down1]:=dvi_par[i+right1];
  dvi_par[i+y1]:=dvi_par[i+right1];
  dvi_par[i+z1]:=dvi_par[i+right1];
  dvi_par[i+fnt_def1]:=dvi_par[i+fnt1];
  end;

@ Next we define the values used as array elements of |dvi_cl|;
several \.{DVI} commands (e.g., |nop|, |bop|, |eop|, |pre|, |post|) will
always be treated separately and are therefore assigned to the invalid
class here.

@d char_cl=0
@d rule_cl=char_cl+1
@d xxx_cl=char_cl+2
@d push_cl=3
@d pop_cl=4
@d w0_cl=5
@d x0_cl=w0_cl+1
@d right_cl=w0_cl+2
@d w_cl=w0_cl+3
@d x_cl=w0_cl+4
@d y0_cl=10
@d z0_cl=y0_cl+1
@d down_cl=y0_cl+2
@d y_cl=y0_cl+3
@d z_cl=y0_cl+4
@d fnt_cl=15
@d fnt_def_cl=16
@d invalid_cl=17
@d max_cl=invalid_cl {largest possible value}

@<Types...@>=
@!cmd_cl=char_cl..max_cl;

@ Here we declare the array |dvi_cl|.

@<Globals...@>=
@!dvi_cl:packed array [eight_bits] of cmd_cl;

@ And here we initialize it.

@<Set init...@>=
for i:=set_char_0 to put1+3 do dvi_cl[i]:=char_cl;
dvi_cl[set_rule]:=rule_cl; dvi_cl[put_rule]:=rule_cl;@/
dvi_cl[nop]:=invalid_cl;
dvi_cl[bop]:=invalid_cl; dvi_cl[eop]:=invalid_cl;@/
dvi_cl[push]:=push_cl; dvi_cl[pop]:=pop_cl;@/
dvi_cl[w0]:=w0_cl; dvi_cl[x0]:=x0_cl;@/
dvi_cl[y0]:=y0_cl; dvi_cl[z0]:=z0_cl;@/
for i:=0 to 3 do
  begin dvi_cl[i+right1]:=right_cl;
  dvi_cl[i+w1]:=w_cl;
  dvi_cl[i+x1]:=x_cl;@/
  dvi_cl[i+down1]:=down_cl;
  dvi_cl[i+y1]:=y_cl;
  dvi_cl[i+z1]:=z_cl;@/
  dvi_cl[i+xxx1]:=xxx_cl;
  dvi_cl[i+fnt_def1]:=fnt_def_cl;
  end;
for i:=fnt_num_0 to fnt1+3 do dvi_cl[i]:=fnt_cl;
for i:=pre to 255 do dvi_cl[i]:=invalid_cl;

@ A few small arrays are used to generate \.{DVI} commands.

@<Glob...@>=
@!dvi_char_cmd:array[boolean] of eight_bits; {|put1| and |set1|}
@!dvi_rule_cmd:array[boolean] of eight_bits; {|put_rule| and |set_rule|}
@!dvi_right_cmd:array[right_cl..x_cl] of eight_bits; {|right1|, |w1|, and |x1|}
@!dvi_down_cmd:array[down_cl..z_cl] of eight_bits; {|down1|, |y1|, and |z1|}

@ @<Set init...@>=
dvi_char_cmd[false]:=put1;
dvi_char_cmd[true]:=set1;@/
dvi_rule_cmd[false]:=put_rule;
dvi_rule_cmd[true]:=set_rule;@/
dvi_right_cmd[right_cl]:=right1;
dvi_right_cmd[w_cl]:=w1;
dvi_right_cmd[x_cl]:=x1;@/
dvi_down_cmd[down_cl]:=down1;
dvi_down_cmd[y_cl]:=y1;
dvi_down_cmd[z_cl]:=z1;

@ The global variables |cur_cmd|, |cur_parm|, and |cur_class| are used
for the current \.{DVI} command, its first parameter (if any), and its
command class respectively.

@<Glob...@>=
@!cur_cmd:eight_bits; {current \.{DVI} command byte}
@!cur_parm:int_32; {its first parameter (if any)}
@!cur_class:cmd_cl; {its class}

@ When typesetting a character or rule, the boolean variable |cur_upd|
is |true| for \\{set} commands, |false| for |put| commands.

@<Glob...@>=
@!cur_cp:char_pointer; {|char_widths| index for the current character}
@!cur_wp:width_pointer; {width pointer of the current character}
@!cur_upd:boolean; {is this a \\{set} or |set_rule| command ?}
@!cur_v_dimen:int_32; {a vertical dimension}
@!cur_h_dimen:int_32; {a horizontal dimension}

@ @<Set init...@>=
cur_cp:=0; cur_wp:=invalid_width; {so they can be saved and restored!}

@ The |dvi_first_par| procedure first reads \.{DVI} command bytes into
|cur_cmd| until |cur_cmd<>nop|; then |cur_parm| is set to the value of
the first parameter (if any) and |cur_class| to the command class.

@d set_cur_char(#)== {set up |cur_res|, |cur_ext|, and |cur_upd|}
begin cur_ext:=0;
if cur_cmd<set1 then
  begin cur_res:=cur_cmd; cur_upd:=true
  end
else  begin cur_res:=#; cur_upd:=(cur_cmd<put1);
  Decr(cur_cmd)(dvi_char_cmd[cur_upd]);
  while cur_cmd>0 do
    begin if cur_cmd=3 then if cur_res>127 then cur_ext:=-1;
    cur_ext:=cur_ext*256+cur_res; cur_res:=#; decr(cur_cmd);
    end;
  end;
end

@p procedure dvi_first_par;
begin repeat cur_cmd:=dvi_ubyte;
until cur_cmd<>nop; {skip over |nop|s}
case dvi_par[cur_cmd] of
char_par: set_cur_char(dvi_ubyte);
no_par: do_nothing;
dim1_par: cur_parm:=dvi_sbyte;
num1_par: cur_parm:=dvi_ubyte;
dim2_par: cur_parm:=dvi_spair;
num2_par: cur_parm:=dvi_upair;
dim3_par: cur_parm:=dvi_strio;
num3_par: cur_parm:=dvi_utrio;
two_cases(dim4_par): cur_parm:=dvi_squad; {|dim4_par| and |num4_par|}
numu_par: cur_parm:=dvi_uquad;
rule_par:
  begin cur_v_dimen:=dvi_squad; cur_h_dimen:=dvi_squad;
  cur_upd:=(cur_cmd=set_rule);
  end;
fnt_par:cur_parm:=cur_cmd-fnt_num_0;
othercases abort('internal error');
endcases;
cur_class:=dvi_cl[cur_cmd];
end;

@ The global variable |dvi_nf| is used for the number of different
\.{DVI} fonts defined so far; their external font numbers (as extracted
from the \.{DVI} file) are stored in the array |dvi_e_fnts|, the
corresponding internal font numbers used internally by \.{\title} are
stored in the array |dvi_i_fnts|.

@<Glob...@>=
@!dvi_e_fnts:array[font_number] of int_32; {external font numbers}
@!dvi_i_fnts:array[font_number] of font_number; {corresponding
  internal font numbers}
@!dvi_nf:font_number; {number of \.{DVI} fonts defined so far}

@ @<Set ini...@>=
dvi_nf:=0;

@ The |dvi_font| procedure sets |cur_fnt| to the internal font number
corresponding to the external font number |cur_parm| (or aborts the
program if such a font was never defined).

@p procedure dvi_font; {computes |cur_fnt| corresponding to |cur_parm|}
var f:font_number; {where the font is sought}
begin @<DVI: Locate font |cur_parm|@>;
if f=dvi_nf then bad_dvi;
cur_fnt:=dvi_i_fnts[f];
if font_type(cur_fnt)=defined_font then load_font;
end;

@ @<DVI: Locate font |cur_parm|@>=
f:=0; dvi_e_fnts[dvi_nf]:=cur_parm;
while cur_parm<>dvi_e_fnts[f] do incr(f)

@ Finally the |dvi_do_font| procedure is called when one of the commands
|fnt_def1..fnt_def4| and its first parameter have been read from the
\.{DVI} file; the argument indicates whether this should be the second
definition of the font (|true|) or not (|false|).

@p procedure dvi_do_font(@!second:boolean);
var f:font_number; {where the font is sought}
@!k:int_15; {general purpose variable}
begin print('DVI: font ',cur_parm:1);
@<DVI: Locate font |cur_parm|@>;
if (f=dvi_nf)=second then bad_dvi;
font_check(nf):=dvi_squad;
font_scaled(nf):=dvi_pquad;
font_design(nf):=dvi_pquad;
k:=dvi_ubyte; pckt_room(1); append_byte(k);
Incr(k)(dvi_ubyte); pckt_room(k);
while k>0 do  begin append_byte(dvi_ubyte); decr(k);
  end;
font_name(nf):=make_packet; {the font area plus name}
dvi_i_fnts[dvi_nf]:=define_font(false);
if not second then
  begin if dvi_nf=max_fonts then overflow(str_fonts,max_fonts);
  incr(dvi_nf);
  end
else if dvi_i_fnts[f]<>dvi_i_fnts[dvi_nf] then bad_dvi;
end;

@* Low-level VF input routines.
A detailed description of the \.{VF} file format can be found in the
documentation of \.{VFtoVP}; here we just define symbolic names for
some of the \.{VF} command bytes.

@d long_char=242 {\.{VF} command for general character packet}
@#
@d vf_id=202 {identifies \.{VF} files}

@ The program uses the binary file variable |vf_file| for input from
\.{VF} files; |vf_loc| is the number of the byte about to be read next
from |vf_file|.

@<Glob...@>=
@!vf_file:byte_file; {a \.{VF} file}
@!vf_loc:int_32; {where we are about to look, in |vf_file|}
@!vf_limit:int_32; {value of |vf_loc| at end of a character packet}
@!vf_ext:pckt_pointer; {extension for \.{VF} files}
@!vf_cur_fnt:font_number; {current font number in a \.{VF} file}

@ @<Initialize predefined strings@>=
id3(".")("V")("F")(vf_ext); {file name extension for \.{VF} files}

@ If a \.{VF} file is badly malformed, we say |bad_font|; this procedure
gives an error message which refers the user to \.{VFtoVP} and \.{VPtoVF},
and terminates \.{\title}.

@<Cases for |bad_font|@>=
vf_font_type: begin print('Bad VF file'); print_font(cur_fnt);
@.Bad VF file@>
  print_ln(' loc=',vf_loc:1);
  abort('Use VFtoVP/VPtoVF to diagnose and correct the problem');
@.Use VFtoVP/VPtoVF@>
  end;

@ If no font directory has been specified, \.{\title} is supposed to use
the default \.{VF} directory, which is a system-dependent place where
the \.{VF} files for standard fonts are kept.
The string variable |VF_default_area| contains the name of this area.
@^system dependencies@>

@d VF_default_area_name=='TeXvfonts:' {change this to the correct name}
@d VF_default_area_name_length=10 {change this to the correct length}

@<Glob...@>=
@!VF_default_area:packed array[1..VF_default_area_name_length] of char;

@ @<Set init...@>=
VF_default_area:=VF_default_area_name;

@ To prepare |vf_file| for input we |reset| it.

@<VF: Open |vf_file| or |goto not_found|@>=
make_font_name(VF_default_area_name_length)(VF_default_area)(vf_ext);
reset(vf_file,cur_name);
if eof(vf_file) then
@^system dependencies@>
  goto not_found;
vf_loc:=0

@ Reading a \.{VF} file should be done as efficient as possible for a
particular system; on many systems this means that a large number of
bytes from |vf_file| is read into a buffer and will then be extracted
from that buffer. In order to simplify such system dependent changes
we use a pair of \.{WEB} macros: |vf_byte| extracts the next \.{VF}
byte and |vf_eof| is |true| if we have reached the end of the \.{VF}
file. Here we give simple minded definitions for these macros in terms
of standard \PASCAL.
@^system dependencies@>
@^optimization@>

@d vf_eof == eof(vf_file) {has the \.{VF} file been exhausted?}
@d vf_byte(#) ==
  if vf_eof then bad_font
  else read(vf_file,#) {obtain next \.{VF} byte}

@ We need several simple functions to read the next byte or bytes
from |vf_file|.

@p function vf_ubyte:int_8u; {returns the next byte, unsigned}
@!begin_byte(vf_byte); incr(vf_loc); comp_ubyte(vf_ubyte);
end;
@#
function vf_upair:int_16u; {returns the next two bytes, unsigned}
@!begin_pair(vf_byte); Incr(vf_loc)(2); comp_upair(vf_upair);
end;
@#
function vf_strio:int_24; {returns the next three bytes, signed}
@!begin_trio(vf_byte); Incr(vf_loc)(3); comp_strio(vf_strio);
end;
@#
function vf_utrio:int_24u; {returns the next three bytes, unsigned}
@!begin_trio(vf_byte); Incr(vf_loc)(3); comp_utrio(vf_utrio);
end;
@#
function vf_squad:int_32; {returns the next four bytes, signed}
@!begin_quad(vf_byte); Incr(vf_loc)(4); comp_squad(vf_squad);
end;

@ All dimensions in a \.{VF} file, except the design sizes of a virtual
font and its local fonts, are |fix_word|s that must be scaled in exactly
the same way as the character widths from a \.{TFM} file; we can use the
same code, but this time |z|, |alpha|, and |beta| are global variables.

@<Glob...@>=
@<Variables for scaling computation@>@;

@ We need five functions to read the next byte or bytes and convert a
|fix_word| to a scaled dimension.

@p function vf_fix1:int_32; {returns the next byte as scaled value}
var x:int_32; {accumulator}
begin vf_byte(tfm_b3); incr(vf_loc);
tfm_fix1(x); vf_fix1:=x;
end;
@#
function vf_fix2:int_32; {returns the next two bytes as scaled value}
var x:int_32; {accumulator}
begin vf_byte(tfm_b2); vf_byte(tfm_b3); Incr(vf_loc)(2);
tfm_fix2(x); vf_fix2:=x;
end;
@#
function vf_fix3:int_32; {returns the next three bytes as scaled value}
var x:int_32; {accumulator}
begin vf_byte(tfm_b1); vf_byte(tfm_b2); vf_byte(tfm_b3);
Incr(vf_loc)(3);@/
tfm_fix3(x); vf_fix3:=x;
end;
@#
function vf_fix3u:int_32; {returns the next three bytes as scaled value}
begin vf_byte(tfm_b1); vf_byte(tfm_b2); vf_byte(tfm_b3);
Incr(vf_loc)(3);@/
vf_fix3u:=tfm_fix3u;
end;
@#
function vf_fix4:int_32; {returns the next four bytes as scaled value}
var x:int_32; {accumulator}
begin vf_byte(tfm_b0); vf_byte(tfm_b1); vf_byte(tfm_b2); vf_byte(tfm_b3);
Incr(vf_loc)(4);@/
tfm_fix4(x); vf_fix4:=x;
end;

@ Three other functions are used in cases where the result must have a
non-negative value or a positive value.

@p function vf_uquad:int_31; {result must be non-negative}
var x:int_32;
begin x:=vf_squad; if x<0 then bad_font @+ else vf_uquad:=x;
end;
@#
function vf_pquad:int_31; {result must be positive}
var x:int_32;
begin x:=vf_squad; if x<=0 then bad_font @+ else vf_pquad:=x;
end;
@#
function vf_fixp:int_31; {result must be positive}
begin vf_byte(tfm_b0); vf_byte(tfm_b1); vf_byte(tfm_b2); vf_byte(tfm_b3);
Incr(vf_loc)(4);@/
if tfm_b0>0 then bad_font;
vf_fixp:=tfm_fix3u;
end;

@ The |vf_first_par| procedure first reads a \.{VF} command byte into
|cur_cmd|; then |cur_parm| is set to the value of the first parameter
(if any) and |cur_class| to the command class.

@d set_cur_wp_end(#)== if cur_wp=invalid_width then #
@d set_cur_wp(#)== {set |cur_wp| to the char's width pointer}
cur_wp:=invalid_width;
if #<>invalid_font then
  if (cur_res>=font_bc(#))and(cur_res<=font_ec(#)) then
    begin cur_cp:=font_chars(#)+cur_res; cur_wp:=char_widths[cur_cp];
    end;
set_cur_wp_end

@p procedure vf_first_par;
begin cur_cmd:=vf_ubyte;
case dvi_par[cur_cmd] of
char_par:
  begin set_cur_char(vf_ubyte); set_cur_wp(vf_cur_fnt)(bad_font);
  end;
no_par: do_nothing;
dim1_par: cur_parm:=vf_fix1;
num1_par: cur_parm:=vf_ubyte;
dim2_par: cur_parm:=vf_fix2;
num2_par: cur_parm:=vf_upair;
dim3_par: cur_parm:=vf_fix3;
num3_par: cur_parm:=vf_utrio;
dim4_par: cur_parm:=vf_fix4;
num4_par: cur_parm:=vf_squad;
numu_par: cur_parm:=vf_uquad;
rule_par:
  begin cur_v_dimen:=vf_fix4; cur_h_dimen:=vf_fix4;
  cur_upd:=(cur_cmd=set_rule);
  end;
fnt_par:cur_parm:=cur_cmd-fnt_num_0;
othercases abort('internal error');
endcases;
cur_class:=dvi_cl[cur_cmd];
end;

@ For a virtual font we set |font_type(f):=vf_font_type|; in this case
|font_font(f)| is the default font for character packets from virtual
font~|f|.
@^font types@>

The global variable |vf_nf| is used for the number of different local
fonts defined in a \.{VF} file so far; their external font numbers (as
extracted from the \.{VF} file) are stored in the array |vf_e_fnts|, the
corresponding internal font numbers used internally by \.{\title} are
stored in the array |vf_i_fnts|.

@<Glob...@>=
@!vf_e_fnts:array[font_number] of int_32; {external font numbers}
@!vf_i_fnts:array[font_number] of font_number; {corresponding
  internal font numbers}
@!vf_nf:font_number; {number of local fonts defined so far}
@!lcl_nf:font_number; {largest |vf_nf| value for any \.{VF} file}

@ @<Set init...@>=
lcl_nf:=0;

@ The |vf_font| procedure sets |vf_cur_fnt| to the internal font number
corresponding to the external font number |cur_parm| (or aborts the
program if such a font was never defined).

@p procedure vf_font; {computes |vf_cur_fnt| corresponding to |cur_parm|}
var f:font_number; {where the font is sought}
begin @<VF: Locate font |cur_parm|@>;
if f=vf_nf then bad_font;
vf_cur_fnt:=vf_i_fnts[f];
end;

@ @<VF: Locate font |cur_parm|@>=
f:=0; vf_e_fnts[vf_nf]:=cur_parm;
while cur_parm<>vf_e_fnts[f] do incr(f)

@ Finally the |vf_do_font| procedure is called when one of the commands
|fnt_def1..fnt_def4| and its first parameter have been read from the
\.{VF} file.

@p procedure vf_do_font;
var f:font_number; {where the font is sought}
@!k:int_15; {general purpose variable}
begin print('VF: font ',cur_parm:1);@/
@<VF: Locate font |cur_parm|@>;
if f<>vf_nf then bad_font;
font_check(nf):=vf_squad;
font_scaled(nf):=vf_fixp;
font_design(nf):=round(tfm_conv*vf_pquad);
k:=vf_ubyte; pckt_room(1); append_byte(k);
Incr(k)(vf_ubyte); pckt_room(k);
while k>0 do  begin append_byte(vf_ubyte); decr(k);
  end;
font_name(nf):=make_packet; {the font area plus name}
vf_i_fnts[vf_nf]:=define_font(true);
if vf_nf=lcl_nf then
  if lcl_nf=max_fonts then overflow(str_fonts,max_fonts)
  else incr(lcl_nf);
incr(vf_nf);
end;

@* Reading VF files.
The |do_vf| function attempts to read the \.{VF} file for a font and
returns |false| if the \.{VF} file could not be found; otherwise the
font type is changed to |vf_font_type|.

@p function do_vf:boolean; {read a \.{VF} file}
label reswitch,done,not_found,exit;
var temp_byte:int_8u; {byte for temporary variables}
@!k:byte_pointer; {index into |byte_mem|}
@!l:int_15; {general purpose variable}
@!save_ext:int_24; {used to save |cur_ext|}
@!save_res:int_8u; {used to save |cur_res|}
@!save_cp:width_pointer; {used to save |cur_cp|}
@!save_wp:width_pointer; {used to save |cur_wp|}
@!save_upd:boolean; {used to save |cur_upd|}
@!vf_wp:width_pointer; {width pointer for the current character packet}
@!vf_fnt:font_number; {current font in the current character packet}
@!move_zero:boolean; {|true| if rule 1 is used}
@!last_pop:boolean; {|true| if final |pop| has been manufactured}
begin @<VF: Open |vf_file| or |goto not_found|@>;
save_ext:=cur_ext; save_res:=cur_res; save_cp:=cur_cp; save_wp:=cur_wp;
save_upd:=cur_upd; {save}
font_type(cur_fnt):=vf_font_type;@/
@<VF: Process the preamble@>;@/
@<VF: Process the font definitions@>;@/
while cur_cmd<=long_char do @<VF: Build a character packet@>;
if cur_cmd<>post then bad_font;
@!debug print('VF file for font ',cur_fnt:1); print_font(cur_fnt);
print_ln(' loaded.');
gubed @;@/
close_in(vf_file);
cur_ext:=save_ext; cur_res:=save_res; cur_cp:=save_cp; cur_wp:=save_wp;
cur_upd:=save_upd; {restore}
do_vf:=true; return;
not_found:do_vf:=false;
exit:end;

@ @<VF: Process the preamble@>=
if vf_ubyte<>pre then bad_font;
if vf_ubyte<>vf_id then bad_font;
temp_byte:=vf_ubyte; pckt_room(temp_byte);
for l:=1 to temp_byte do append_byte(vf_ubyte);
print('VF file: '''); print_packet(new_packet); print(''',');
flush_packet;@/
check_check_sum(vf_squad,false);
check_design_size(round(tfm_conv*vf_pquad));@/
z:=font_scaled(cur_fnt);
@<Replace |z|...@>;@/
print_nl('   for font ',cur_fnt:1); print_font(cur_fnt); print_ln('.')

@ @<VF: Process the font definitions@>=
vf_i_fnts[0]:=invalid_font; vf_nf:=0;@/
cur_cmd:=vf_ubyte;
while (cur_cmd>=fnt_def1)and(cur_cmd<=fnt_def1+3) do
  begin case cur_cmd-fnt_def1 of
  0: cur_parm:=vf_ubyte;
  1: cur_parm:=vf_upair;
  2: cur_parm:=vf_utrio;
  3: cur_parm:=vf_squad;
  end; {there are no other cases}
  vf_do_font;
  cur_cmd:=vf_ubyte;
  end;
font_font(cur_fnt):=vf_i_fnts[0]

@ The \.{VF} format specifies that the interpretation of each packet
begins with |w=x=y=z=0|; any |w0|, |x0|, |y0|, or |z0| command using
these initial values will be ignored.

@<Types...@>=
@!vf_state=array[0..1,0..1] of boolean; {state of |w|, |x|, |y|, and |z|}

@ As implied by the \.{VF} format the \.{DVI} commands read from the \.{VF}
file are enclosed by |push| and |pop|; as we read \.{DVI}
commands and append them to |byte_mem|, we perform a set of
transformations in order to simplify the resulting packet: Let |zero| be
any of the commands |put|, |put_rule|, |fnt_num|, |fnt|, or |xxx| which
all leave the current position on the page unchanged, let |move| be any
of the horizontal or vertical movement commands |right1..z4|, and let
|any| be any sequence of commands containing |push| and |pop| in
properly nested pairs; whenever possible we apply one of the following
transformation rules: $$\def\n#1:{\hbox to 3cm{\hfil#1:}}
\leqalignno{
\hbox{|push| |zero|}&\RA\hbox{|zero| |push|}&\n1:\cr
\hbox{|move| |pop|}&\RA\hbox{|pop|}&\n2:\cr
\hbox{|push| |pop|}&\RA{}&\n3:\cr
\hbox{|push| |set_char| |pop|}&\RA\hbox{|put|}&\n4a:\cr
\hbox{|push| \\{set} |pop|}&\RA\hbox{|put|}&\n4b:\cr
\hbox{|push| |set_rule| |pop|}&\RA\hbox{|put_rule|}&\n4c:\cr
\hbox{|push| |push| |any| |pop|}&\RA\hbox{|push| |any| |pop| |push|}&\n5:\cr
\hbox{|push| |any| |pop| |pop|}&\RA\hbox{|any| |pop|}&\n6:\cr
}$$

@ In order to perform these transformations we need a stack which is
indexed by |vf_ptr|, the number of |push| commands without corresponding
|pop| in the packet we are building; the |vf_push_loc| array contains
the locations in |byte_mem| following such |push| commands.
In view of rule~5 consecutive |push| commands are never stored, the
|vf_push_num| array is used to count them.
The |vf_last| array indicates the type of the last non-discardable item:
a character, a rule, or a group enclosed by |push| and |pop|;
the |vf_last_end| array points to the ending locations and, if
|vf_last<>vf_other|, the |vf_last_loc| array points to the starting
locations of these items.

@d vf_set=0 {|vf_set=char_cl|, last item is a |set_char| or \\{set}}
@d vf_rule=1 {|vf_rule=rule_cl|, last item is a |set_rule|}
@d vf_group=2 {last item is a group enclosed by |push| and |pop|}
@d vf_put=3 {last item is a |put|}
@d vf_other=4 {last item (if any) is none of the above}

@<Types...@>=
@!vf_type=vf_set..vf_other;

@ @<Glob...@>=
@!vf_move: array[stack_pointer] of vf_state; {state of |w|, |x|, |y|, and |z|}
@!vf_push_loc: array[stack_pointer] of byte_pointer; {end of a |push|}
@!vf_last_loc: array[stack_pointer] of byte_pointer; {start of an item}
@!vf_last_end: array[stack_pointer] of byte_pointer; {end of an item}
@!vf_push_num: array[stack_pointer] of eight_bits; {|push| count}
@!vf_last: array[stack_pointer] of vf_type; {type of last item}
@!vf_ptr:stack_pointer; {current number of unfinished groups}
@!stack_used:stack_pointer; {largest |vf_ptr| or |stack_ptr| value}

@ We use two small arrays to determine the item type of a character or a
rule.

@<Glob...@>=
@!vf_char_type:array[boolean] of vf_type;
@!vf_rule_type:array[boolean] of vf_type;

@ @<Set init...@>=
vf_move[0][0][0]:=false; vf_move[0][0][1]:=false;
vf_move[0][1][0]:=false; vf_move[0][1][1]:=false;@/
stack_used:=0;@/
vf_char_type[false]:=vf_put; vf_char_type[true]:=vf_set;@/
vf_rule_type[false]:=vf_other; vf_rule_type[true]:=vf_rule;

@ Here we read the first bytes of a character packet from the \.{VF}
file and initialize the packet being built in |byte_mem|; the start of
the whole packet is stored in |vf_push_loc[0]|. When the character
packet is finished, a type is assigned to it: |vf_simple| if the
packet ends with a character of the correct width, or |vf_complex|
otherwise. Moreover, if such a packet for a character with
extension zero consists of just one character with extension zero and
the same residue, and if there is no previous packet, the whole packet
is replaced by the empty packet.

@d vf_simple=0 {the packet ends with a character of the correct width}
@d vf_complex=vf_simple+1 {otherwise}

@<VF: Build a character packet@>=
begin if cur_cmd<long_char then
  begin vf_limit:=cur_cmd;
  cur_ext:=0; cur_res:=vf_ubyte; vf_wp:=check_width(vf_fix3u);
  end
else  begin vf_limit:=vf_uquad;
  cur_ext:=vf_strio; cur_res:=vf_ubyte; vf_wp:=check_width(vf_fix4);
  end;
Incr(vf_limit)(vf_loc);
vf_push_loc[0]:=byte_ptr; vf_last_end[0]:=byte_ptr;
vf_last[0]:=vf_other; vf_ptr:=0;@/
start_packet(vf_complex);
@<VF: Append \.{DVI} commands to the character packet@>;@/
k:=pckt_start[pckt_ptr];
if vf_last[0]=vf_put then if cur_wp=vf_wp then
  begin decr(byte_mem[k]); {change |vf_complex| into |vf_simple|}
  if (byte_mem[k]=bi(0))and@|(vf_push_loc[0]=vf_last_loc[0])and@|
    (cur_ext=0)and@|(cur_res=pckt_res) then byte_ptr:=k;
  end;
build_packet;
cur_cmd:=vf_ubyte;
end

@ For every \.{DVI} command read from the \.{VF} file some action is
performed; in addition the initial |push| and the final |pop| are
manufactured here.

@<VF: Append \.{DVI} commands to the character packet@>=
vf_cur_fnt:=font_font(cur_fnt); vf_fnt:=vf_cur_fnt;@/
last_pop:=false; cur_class:=push_cl; {initial |push|}
loop  begin
reswitch:case cur_class of
  three_cases(char_cl): @<VF: Do a |char|, |rule|, or |xxx|@>;
  push_cl: @<VF: Do a |push|@>;
  pop_cl: @<VF: Do a |pop|@>;
  two_cases(w0_cl):
    if vf_move[vf_ptr][0][cur_class-w0_cl] then append_one(cur_cmd);
  three_cases(right_cl):
    begin pckt_signed(dvi_right_cmd[cur_class],cur_parm);
    if cur_class>=w_cl then vf_move[vf_ptr][0][cur_class-w_cl]:=true;
    end;
  two_cases(y0_cl):
    if vf_move[vf_ptr][1][cur_class-y0_cl] then append_one(cur_cmd);
  three_cases(down_cl):
    begin pckt_signed(dvi_down_cmd[cur_class],cur_parm);
    if cur_class>=y_cl then vf_move[vf_ptr][1][cur_class-y_cl]:=true;
    end;
  fnt_cl: vf_font;
  fnt_def_cl: bad_font;
  invalid_cl: if cur_cmd<>nop then bad_font;
  othercases abort('internal error');
  endcases;
  if vf_loc<vf_limit then vf_first_par
  else if last_pop then goto done
  else  begin cur_class:=pop_cl; last_pop:=true; {final |pop|}
    end;
  end;
done:if (vf_ptr<>0)or(vf_loc<>vf_limit) then bad_font

@ For a |push| we either increase |vf_push_num| or start a new level and
append a |push|.

@d incr_stack(#)==
if #=stack_used then
  if stack_used=stack_size then overflow(str_stack,stack_size)
  else incr(stack_used);
incr(#)

@<VF: Do a |push|@>=
if (vf_ptr>0)and(vf_push_loc[vf_ptr]=byte_ptr) then
  begin if vf_push_num[vf_ptr]=255 then overflow(str_stack,255);
  incr(vf_push_num[vf_ptr]);
  end
else  begin incr_stack(vf_ptr);
  @<VF: Start a new level@>;
  vf_push_num[vf_ptr]:=0;
  end

@ @<VF: Start a new level@>=
append_one(push);
vf_move[vf_ptr]:=vf_move[vf_ptr-1];
vf_push_loc[vf_ptr]:=byte_ptr;
vf_last_end[vf_ptr]:=byte_ptr;
vf_last[vf_ptr]:=vf_other

@ When a character, a rule, or an |xxx| is appended, transformation
rule~1 might be applicable.

@<VF: Do a |char|, |rule|, or |xxx|@>=
begin if (vf_ptr=0)or(byte_ptr>vf_push_loc[vf_ptr]) then move_zero:=false
else case cur_class of
char_cl: move_zero:=(not cur_upd)or(vf_cur_fnt<>vf_fnt);
rule_cl: move_zero:=not cur_upd;
xxx_cl: move_zero:=true;
othercases abort('internal error');
endcases;
if move_zero then
  begin decr(byte_ptr); decr(vf_ptr);
  end;
case cur_class of
char_cl: @<VF: Do a |fnt|, a |char|, or both@>;
rule_cl: @<VF: Do a |rule|@>;
xxx_cl: @<VF: Do an |xxx|@>;
end; {there are no other cases}
vf_last_end[vf_ptr]:=byte_ptr;
if move_zero then
  begin incr(vf_ptr); append_one(push); vf_push_loc[vf_ptr]:=byte_ptr;
  vf_last_end[vf_ptr]:=byte_ptr;
  if cur_class=char_cl then if cur_upd then goto reswitch;
  end;
end

@ A special situation arises if transformation rule~1 is applied to a
|fnt_num| of |fnt| command, but not to the |set_char| or \\{set} command
following it; in this case |cur_upd| and |move_zero| are both |true| and
the |set_char| or \\{set} command will be appended later.

@<VF: Do a |fnt|, a |char|, or both@>=
begin if vf_cur_fnt<>vf_fnt then
  begin vf_last[vf_ptr]:=vf_other;
  pckt_unsigned(fnt1,vf_cur_fnt); vf_fnt:=vf_cur_fnt;
  end;
if (not move_zero)or(not cur_upd) then
  begin vf_last[vf_ptr]:=vf_char_type[cur_upd];
  vf_last_loc[vf_ptr]:=byte_ptr;
  pckt_char(cur_upd,cur_ext,cur_res);
  end;
end

@ @<VF: Do a |rule|@>=
begin vf_last[vf_ptr]:=vf_rule_type[cur_upd];
vf_last_loc[vf_ptr]:=byte_ptr;
append_one(dvi_rule_cmd[cur_upd]);
pckt_four(cur_v_dimen); pckt_four(cur_h_dimen);
end

@ @<VF: Do an |xxx|@>=
begin vf_last[vf_ptr]:=vf_other;
pckt_unsigned(xxx1,cur_parm); pckt_room(cur_parm);
while cur_parm>0 do
  begin append_byte(vf_ubyte); decr(cur_parm);
  end;
end

@ Transformation rules 2--6 are triggered by a |pop|, either read from
the \.{VF} file or manufactured at the end of the packet.

@<VF: Do a |pop|@>=
begin if vf_ptr<1 then bad_font;
byte_ptr:=vf_last_end[vf_ptr]; {this is rule 2}
if vf_last[vf_ptr]<=vf_rule then
 if vf_last_loc[vf_ptr]=vf_push_loc[vf_ptr] then
  @<VF: Prepare for rule 4@>;
if byte_ptr=vf_push_loc[vf_ptr] then @<VF: Apply rule 3 or 4@>
else  begin if vf_last[vf_ptr]=vf_group then @<VF: Apply rule 6@>;
  append_one(pop); decr(vf_ptr); vf_last[vf_ptr]:=vf_group;
  vf_last_loc[vf_ptr]:=vf_push_loc[vf_ptr+1]-1;
  vf_last_end[vf_ptr]:=byte_ptr;
  if vf_push_num[vf_ptr+1]>0 then @<VF: Apply rule 5@>;
  end;
end

@ In order to implement transformation rule~4, we cancel the |set_char|,
\\{set}, or |set_rule|, append a |pop|, and insert a |put| or |put_rule|
with the old parameters.

@<VF: Prepare for rule 4@>=
begin cur_class:=vf_last[vf_ptr]; cur_upd:=false;
byte_ptr:=vf_push_loc[vf_ptr];
end

@ @<VF: Apply rule 3 or 4@>=
begin if vf_push_num[vf_ptr]>0 then
  begin decr(vf_push_num[vf_ptr]);
  vf_move[vf_ptr]:=vf_move[vf_ptr-1];
  end
else  begin decr(byte_ptr); decr(vf_ptr);
  end;
if cur_class<>pop_cl then goto reswitch; {this is rule 4}
end

@ @<VF: Apply rule 6@>=
begin Decr(byte_ptr)(2);
for k:=vf_last_loc[vf_ptr]+1 to byte_ptr do byte_mem[k-1]:=byte_mem[k];
vf_last[vf_ptr]:=vf_other; vf_last_end[vf_ptr]:=byte_ptr;
end

@ @<VF: Apply rule 5@>=
begin incr(vf_ptr);
@<VF: Start a new level@>;
decr(vf_push_num[vf_ptr]);
end

@ The \.{VF} format specifies that after a character packet invoked by a
|set_char| or \\{set} command, ``|h|~is increased by the \.{TFM} width
(properly scaled)---just as if a simple character had been typeset'';
for |vf_simple| packets this is achieved by changing the final |put|
command into |set_char| or \\{set}, but for |vf_complex| packets an
explicit movement must be done. This poses a problem for programs,
such as \.{DVIcopy}, which write a new \.{DVI} file with all references
to characters from virtual fonts replaced by their character packets:
The \.{DVItype} program specifies that the horizontal movements after a
|set_char| or \\{set} command, after a |set_rule| command, and after one
of the commands |right1..x4|, are all treated differently when \.{DVI}
units are converted to pixels.

Thus we introduce a slight extension of \.{DVItype}'s pixel rounding
algorithm and hope that this extension will become part of the standard
\.{DVItype} program in the near future: If a \.{DVI} file contains a
|set_rule| command for a rule with the negative height |width_dimen|,
then this rule shall be treated in exactly the same way as a ficticious
character whose width is the width of that rule; as value of |width_dimen|
we choose $-2^{31}$, the smallest signed 32-bit integer.

@<Glob...@>=
@!width_dimen:int_32; {vertical dimension of special rules}

@ When initializing |width_dimen| we are careful to avoid arithmetic
overflow.

@<Set init...@>=
width_dimen:=-@"40000000; Decr(width_dimen)(@"40000000);

@* Terminal communication.
When \.{\title} begins, it engages the user in a brief dialog so that
various options may be specified. This part of \.{\title} requires
nonstandard \PASCAL\ constructions to handle the online interaction; so
it may be preferable in some cases to omit the dialog and simply to
stick to the default options. On other hand, the system-dependent
routines that are needed are not complicated, so it will not be terribly
difficult to introduce them; furthermore they are similar to those in
\.{DVItype}.

It may be desirable to (optionally) specify all the options in the
command line and skip the dialog with the user, provided the operating
system permits this. Here we just define the system-indepent part of the
code required for this possibility. Since a complete option (a keyword
possibly followed by one or several parameters) may have embedded blanks
it might be necessary to replace these blanks by some other separator,
e.g., by a '/'. Using, e.g., \.{UNIX} style options one might then say
$$\.{\title\space-mag/2000 -sel/17.3/5 -sel/47 ...}$$
to override the magnification factor that is stated in the \.{DVI} file,
and to select five pages starting with the page numbered~17.3 as well as
all remaining pages starting with the one numbered~47; alternatively one
might simply say
$$\.{\title\space- ...}$$
to skip the dialog and use the default options.

The system-dependent initialization code should set the |n_opt| variable
to the number of options found in the command line.  If |n_opt=0| the
|input_ln| procedure defined below will prompt the user for options.  If
|n_opt>0| the |k_opt| variable will be incremented and another piece of
system-dependent code is invoked instead of the dialog; that code should
place the value of command line option number |k_opt| as temporary
string into the |byte-mem| array.  This process will be repeated until
|k_opt=n_opt|, indicating that all command line options have been
processed.
@^system dependencies@>

@d opt_separator="/" {acts as blank when scanning (command line) options}

@<Set init...@>=
n_opt:=0; {change this to indicate the presence of command line options}
k_opt:=0; {just in case}

@ The |input_ln| routine waits for the user to type a line at his or her
terminal; then it puts ASCII-code equivalents for the characters on that
line into the |byte_mem| array as a temporary string. \PASCAL's
standard |input| file is used for terminal input, as |output| is used
for terminal output.

Since the terminal is being used for both input and output, some systems
need a special routine to make sure that the user can see a prompt message
before waiting for input based on that message. (Otherwise the message
may just be sitting in a hidden buffer somewhere, and the user will have
no idea what the program is waiting for.) We shall invoke a system-dependent
subroutine |update_terminal| in order to avoid this problem.
@^system dependencies@>

@d update_terminal == break(output) {empty the terminal output buffer}
@#
@d scan_blank(#)== {tests for `blank' when scanning (command line) options}
  ((byte_mem[#]=bi(" "))or(byte_mem[#]=bi(opt_separator)))
@d scan_skip== {skip `blanks'}
  while scan_blank(scan_ptr)and(scan_ptr<byte_ptr) do incr(scan_ptr)
@d scan_init== {initialize |scan_ptr|}
  byte_mem[byte_ptr]:=bi(" "); scan_ptr:=pckt_start[pckt_ptr-1]; scan_skip

@<Action procedures for |dialog|@>=
procedure input_ln; {inputs a line from the terminal}
var k:0..terminal_line_length;
begin if n_opt=0 then
  begin print('Enter option: '); update_terminal; reset(input);
  if eoln(input) then read_ln(input);
  k:=0; pckt_room(terminal_line_length);
  while (k<terminal_line_length)and not eoln(input) do
    begin append_byte(xord[input^]); incr(k); get(input);
    end;
  end
else if k_opt<n_opt then
  begin incr(k_opt);
  {Copy command line option number |k_opt| into |byte_mem| array!}
  end;
end;

@ The global variable |scan_ptr| is used while scanning the temporary
packet; it points to the next byte in |byte_mem| to be examined.

@<Glob...@>=
@!n_opt:int_16; {number of options found in command line}
@!k_opt:int_16; {number of command line options processed}
@!scan_ptr:byte_pointer; {pointer to next byte to be examined}
@!sep_char:text_char; {|' '| or |xchr[opt_separator]|}

@ The |scan_keyword| function is used to test for keywords in a character
string stored as temporary packet in |byte_mem|; the result is |true|
(and |scan_ptr| is updated) if the characters starting at position
|scan_ptr| are an abbreviation of a given keyword followed by at least
one blank.

@<Action procedures for |dialog|@>=
function scan_keyword(@!p:pckt_pointer;@!l:int_7):boolean;
var i,@!j,@!k:byte_pointer; {indices into |byte_mem|}
begin i:=pckt_start[p]; j:=pckt_start[p+1]; k:=scan_ptr;
while (i<j)and((byte_mem[k]=byte_mem[i])or(byte_mem[k]=byte_mem[i]-"a"+"A")) do
  begin incr(i); incr(k);
  end;
if scan_blank(k)and(i-pckt_start[p]>=l) then
  begin scan_ptr:=k; scan_skip; scan_keyword:=true;
  end
else scan_keyword:=false;
end;

@ Here is a routine that scans a (possibly signed) integer and computes
the decimal value. If no decimal integer starts at |scan_ptr|, the
value~0 is returned. The integer should be less than $2^{31}$ in
absolute value.

@<Action procedures for |dialog|@>=
function scan_int:int_32;
var x:int_32; {accumulates the value}
@!negative:boolean; {should the value be negated?}
begin if byte_mem[scan_ptr]="-" then
  begin negative:=true; incr(scan_ptr);
  end
else negative:=false;
x:=0;
while (byte_mem[scan_ptr]>="0")and(byte_mem[scan_ptr]<="9") do
  begin x:=10*x+byte_mem[scan_ptr]-"0"; incr(scan_ptr);
  end;
scan_skip;
if negative then scan_int:=-x @+ else scan_int:=x;
end;

@ The selected options are put into global variables by the |dialog|
procedure, which is called just as \.{\title} begins.
@^system dependencies@>

@p @<Action procedures for |dialog|@>@;
procedure dialog;
label exit;
var p:pckt_pointer; {packet being created}
begin @<Initialize options@>@;
loop  begin input_ln; p:=new_packet; scan_init;
  if scan_ptr=byte_ptr then
    begin flush_packet; return;
    end@;@/
  @<Cases for options@>@;@/
  else  begin if n_opt=0 then sep_char:=' '
    else sep_char:=xchr[opt_separator];
    print_options;
    if n_opt>0 then
      begin print('Bad command line option: ');
      print_packet(p); abort('---run terminated');
      end;
    end;
  flush_packet;
  end;
exit:end;

@ The |print_options| procedure might be used in a `Usage message'
displaying the command line syntax.

@<Basic printing...@>=
procedure print_options;
begin print_ln('Valid options are:');
@<Print valid options@>@;
end;

@* Subroutines for typesetting commands.
This is the central part of the whole \.{\title} program:
When a typesetting command from the \.{DVI} file or from a \.{VF} packet
has been decoded, one of the typesetting routines defined below is
invoked to execute the command; apart from the necessary book keeping,
these routines invoke device dependent code defined later.

@p @<Declare typesetting procedures@>

@ These typesetting routines communicate with the rest of the program
through global variables.

@<Glob...@>=
@!type_setting:boolean; {|true| while typesetting a page}

@ @<Set init...@>=
type_setting:=false;

@ The user may select up to |max_select| ranges of consecutive pages to
be processed. Each starting page specification is recorded in two global
arrays called |start_count| and |start_there|. For example, `\.{1.*.-5}'
is represented by |start_there[0]=true|, |start_count[0]=1|,
|start_there[1]=false|, |start_there[2]=true|, |start_count[2]=-5|. We
also set |start_vals=2|, to indicate that count 2 was the last one
mentioned. The other values of |start_count| and |start_there| are not
important, in this example. The number of pages is recorded in
|max_pages|; a non positive value indicates that there is no limit.

@d start_count==select_count[cur_select] {count values to select
  starting page}
@d start_there==select_there[cur_select] {is the |start_count| value
  relevant?}
@d start_vals==select_vals[cur_select] {the last count considered
  significant}
@d max_pages==select_max[cur_select] {at most this many |bop..eop| pages
  will be printed}

@<Glob...@>=
@!select_count:array[0..max_select-1,0..9] of int_32;
@!select_there:array[0..max_select-1,0..9] of boolean;
@!select_vals:array[0..max_select-1] of 0..9;
@!select_max:array[0..max_select-1] of int_32;
@!out_mag:int_32; {output maginfication}
@!count:array[0..9] of int_32; {the count values on the current page}
@!num_select:0..max_select; {number of page selection ranges specified}
@!cur_select:0..max_select; {current page selection range}
@!selected:boolean; {has starting page been found?}
@!all_done:boolean; {have all selected pages been processed?}
@!str_mag,@!str_select:pckt_pointer;

@ Here is a simple subroutine that tests if the current page might be the
starting page.

@p function start_match:boolean; {does |count| match the starting spec?}
var k:0..9;  {loop index}
@!match:boolean; {does everything match so far?}
begin match:=true;
for k:=0 to start_vals do
  if start_there[k]and(start_count[k]<>count[k]) then match:=false;
start_match:=match;
end;

@ @<Initialize options@>=
out_mag:=0; cur_select:=0; max_pages:=0; selected:=true;

@ @<Print valid options@>=
print_ln('  mag',sep_char,'<new_mag>');
print_ln('  select',sep_char,'<start_count>',sep_char,
  '[<max_pages>]  (up to ',max_select:1,' ranges)');

@ @<Action procedures for |dialog|@>=
procedure scan_count; {scan a |start_count| value}
begin if byte_mem[scan_ptr]=bi("*") then
  begin start_there[start_vals]:=false; incr(scan_ptr); scan_skip;
  end
else  begin start_there[start_vals]:=true;
  start_count[start_vals]:=scan_int;
  if cur_select=0 then selected:=false; {don't start at first page}
  end;
end;

@ @<Cases for options@>=
else if scan_keyword(str_mag,3) then out_mag:=scan_int
else if scan_keyword(str_select,3) then
  if cur_select=max_select then print_ln('Too many page selections')
  else  begin start_vals:=0; scan_count;
    while (start_vals<9)and(byte_mem[scan_ptr]=bi(".")) do
      begin incr(start_vals); incr(scan_ptr); scan_count;
      end;
    max_pages:=scan_int; incr(cur_select);
    end

@ @<Initialize predefined strings@>=
id3("m")("a")("g")(str_mag);
id6("s")("e")("l")("e")("c")("t")(str_select);

@ A stack is used to keep track of the current horizonal and vertical
position, |h| and |v|, and the four registers |w|, |x|, |y|, and |z|;
the register pairs |(w,x)| and |(y,z)| are maintained as arrays.

@<Types...@>=
@!device @<Declare device dependend types@>@; @+ ecived @; @/
@!stack_pointer=0..stack_size;@/
@!stack_index=1..stack_size;@/
@!pair_32=array[0..1] of int_32; {a pair of |int_32| variables}
@!stack_record=record@;@/
  @!h_field:int_32; {horizontal position |h|}
  @!v_field:int_32; {vertical position |v|}
  @!w_x_field:pair_32; {|w| and |x| register for horizontal movements}
  @!y_z_field:pair_32; {|y| and |z| register for vertical movements}
  @!device @<Device dependent stack record fields@>@; @+ ecived @; @/
  end;

@ The current values are kept in |cur_stack|; they are pushed onto and
popped from |stack|. We use \.{WEB} macros to access the current values.

@d cur_h==cur_stack.h_field {the current |@!h| value}
@d cur_v==cur_stack.v_field {the current |@!v| value}
@d cur_w_x==cur_stack.w_x_field {the current |@!w| and |@!x| value}
@d cur_y_z==cur_stack.y_z_field {the current |@!y| and |@!z| value}

@<Glob...@>=
@!stack:array[stack_index] of stack_record; {the pushed values}
@!cur_stack:stack_record; {the current values}
@!zero_stack:stack_record; {initial values}
@!stack_ptr:stack_pointer; {last used position in |stack|}

@ @<Set init...@>=
zero_stack.h_field:=0; zero_stack.v_field:=0;
for i:=0 to 1 do
  begin zero_stack.w_x_field[i]:=0; zero_stack.y_z_field[i]:=0;
  end;
@!device @<Initialize device dependent stack record fields@>@; @+ ecived @; @/

@ When typesetting for a real device we must convert the current
position from \.{DVI} units to pixels, i.e., |cur_h| and |cur_v| into
|cur_hh| and |cur_vv|.  This might be a good place to collect everything
related to the conversion from \.{DVI} units to pixels and in particular
all the pixel rounding algorithms.

@d font_space(#)==fnt_space[#] {boundary between ``small'' and ``large''
  spaces}

@<Declare device dependent font data arrays@>=
@!fnt_space:array [font_number] of int_32; {boundary between ``small''
  and ``large'' spaces}

@ @<Initialize device dependent font data@>=
font_space(invalid_font):=0;

@ @<Initialize device dependent data for a font@>=
font_space(cur_fnt):=font_scaled(cur_fnt) div 6;
  {this is a 3-unit ``thin space''}

@ The |char_pixels| array is used to store the horizontal character
escapements:  for \.{PK} or \.{GF} files we use the values given there,
otherwise we must convert the character widths to (horizontal) pixels.
The horizontal escapement of character~|c| in font~|f| is given by
|font_pixel(f)(c)|.

@d font_pixel(#)==char_pixels[font_chars(#)+font_width_end
@#
@d max_pix_value==@"7FFF {largest allowed pixel value; this range may not
  suffice for high resolution output devices}

@<Declare device dependend types@>=
@!pix_value=-max_pix_value..max_pix_value; {a pixel coordinate or displacement}

@ @<Glob...@>=
@!device
@!char_pixels:array[char_pointer] of pix_value; {character escapements}
@!h_pixels:pix_value; {a horizontal dimension in pixels}
@!v_pixels:pix_value; {a vertical dimension in pixels}
@!temp_pix:pix_value; {temporary value for pixel rounding}
ecived

@ @d cur_hh==cur_stack.hh_field {the current |@!hh| value}
@d cur_vv==cur_stack.vv_field {the current |@!vv| value}

@<Device dependent stack record fields@>=
@!hh_field:pix_value; {horizontal pixel position |hh|}
@!vv_field:pix_value; {vertical pixel position |vv|}

@ @<Initialize device dependent stack record fields@>=
zero_stack.hh_field:=0; zero_stack.vv_field:=0;

@ For small movements we round the increment in position, for large
movements we round the incremented position.  The same applies to rule
dimensions with the only difference that they will always be rounded
towards larger values.  For characters we increment the horizontal
position by the escapement values obtained, e.g., from a \.{PK} file or
by the \.{TFM} width converted to pixels.

@d h_pixel_round(#)==round(h_conv*(#))
@d v_pixel_round(#)==round(v_conv*(#))
@^system dependencies@>
@#
@d large_h_space(#)==(#>=font_space(cur_fnt))or(#<=-4*font_space(cur_fnt))
  {is this a ``large'' horizontal distance?}
@d large_v_space(#)==(abs(#)>=5*font_space(cur_fnt))
  {is this a ``large'' vertical distance?}
@#
@d h_rule_pixels== {converts the rule width |cur_h_dimen| to pixels}
@!device if large_h_space(cur_h_dimen) then
  begin h_pixels:=h_pixel_round(cur_h+cur_h_dimen)-cur_hh;
  if h_pixels<=0 then if cur_h_dimen>0 then h_pixels:=1;
  end
else  begin h_pixels:=trunc(h_conv*cur_h_dimen);
  if h_pixels<h_conv*cur_h_dimen then incr(h_pixels);
  end;
ecived
@#
@d v_rule_pixels== {converts the rule height |cur_v_dimen| to pixels}
@!device if large_v_space(cur_v_dimen) then
  begin v_pixels:=cur_vv-v_pixel_round(cur_v-cur_v_dimen);
  if v_pixels<=0 then v_pixels:=1; {used only for |cur_v_dimen>0|}
  end
else  begin v_pixels:=trunc(v_conv*cur_v_dimen);
  if v_pixels<v_conv*cur_v_dimen then incr(v_pixels);
  end;
ecived

@ A sequence of consecutive rules, or consecutive characters in a
fixed-width font whose width is not an integer number of pixels, can
cause |hh| to drift far away from a correctly rounded value.  \.{\title}
ensures that the amount of drift will never exceed |max_h_drift| pixels;
similarly |vv| shall never drift away from the correctly rounded value
by more than |max_v_drift| pixels.

@d h_upd_end(#)== {check for proper horizontal pixel rounding}
begin Incr(cur_hh)(#); temp_pix:=h_pixel_round(cur_h);
if abs(temp_pix-cur_hh)>max_h_drift then
  if temp_pix>cur_hh then cur_hh:=temp_pix-max_h_drift
  else cur_hh:=temp_pix+max_h_drift;
end @+ ecived
@d h_upd_char(#)==Incr(cur_h)(#)@;
  @!device; h_upd_end
@d h_upd_move(#)==Incr(cur_h)(#)@;
  @!device; if large_h_space(#) then cur_hh:=h_pixel_round(cur_h)
  else h_upd_end
@#
@d v_upd_end(#)== {check for proper vertical pixel rounding}
begin Incr(cur_vv)(#); temp_pix:=v_pixel_round(cur_v);
if abs(temp_pix-cur_vv)>max_v_drift then
  if temp_pix>cur_vv then cur_vv:=temp_pix-max_v_drift
  else cur_vv:=temp_pix+max_v_drift;
end @+ ecived
@d v_upd_move(#)==Incr(cur_v)(#)@;
  @!device; if large_v_space(#) then cur_vv:=v_pixel_round(cur_v)
  else v_upd_end

@ The routines defined below use sections named `Declare local variables
(if any) for \dots' or `Declare additional local variables for \dots';
the former may declare variables (including the keyword \&{var}), whereas
the later must at least contain the keyword \&{var}. In general, both may
start with the declaration of labels, constants, and\slash or types.

Let us start with the simple cases:
The |do_pre| procedure is called when the preamble has been read from
the \.{DVI} file; the preamble comment has just been converted into a
temporary packet with the |new_packet| procedure.

@p procedure do_pre;@/
@<OUT: Declare local variables (if any) for |do_pre|@>@;
begin all_done:=false; num_select:=cur_select; cur_select:=0;
if num_select=0 then max_pages:=0;
@!device
h_conv:=(dvi_num/254000.0)*(h_resolution/dvi_den)*(out_mag/1000.0);
v_conv:=(dvi_num/254000.0)*(v_resolution/dvi_den)*(out_mag/1000.0);
ecived @; @/
@<OUT: Process the |pre|@>@;@/
end;

@ The |do_bop| procedure is called when a |bop| has been read. This
routine determines whether a page shall be processed or skipped and sets
the variable |type_setting| accordingly.

@p procedure do_bop;@/
@<OUT: Declare additional local variables |do_bop|@>@;
@!i,@!j:0..9; {indices into |count|}
begin @<Determine whether this page should be processed or skipped@>;
print('DVI: ');
if type_setting then print('process') @+ else print('skipp');
print('ing page ',count[0]:1); j:=9;
while (j>0)and(count[j]=0) do decr(j);
for i:=1 to j do print('.',count[i]:1);
d_print(' at ',dvi_loc-45:1);
print_ln('.');
if type_setting then
  begin stack_ptr:=0; cur_stack:=zero_stack; cur_fnt:=invalid_font;@/
  @<OUT: Process a |bop|@>@;@/
  end;
end;

@ Note that the device dependent code `OUT: Process a |bop|' may choose
to set |type_setting| to false even if |selected| is true.

@<Determine whether this page...@>=
if not selected then selected:=start_match;
type_setting:=selected

@ The |do_eop| procedure is called in order to process an |eop|; the
stack should be empty.

@p procedure do_eop;@/
@<OUT: Declare local variables (if any) for |do_eop|@>@;
begin if stack_ptr<>0 then bad_dvi;
@<OUT: Process an |eop|@>@;
if max_pages>0 then
  begin decr(max_pages);
  if max_pages=0 then
    begin selected:=false; incr(cur_select);
   if cur_select=num_select then all_done:=true;
    end;
  end;
type_setting:=false;
end;

@ The procedures |do_push| and |do_pop| are called in order to process
|push| and |pop| commands; |do_push| must check for stack overflow,
|do_pop| should never be called when the stack is empty.

@p procedure do_push; {push onto stack}
@<OUT: Declare local variables (if any) for |do_push|@>@;
begin incr_stack(stack_ptr); stack[stack_ptr]:=cur_stack;@/
@<OUT: Process a |push|@>@;
end;
@#
procedure do_pop; {pop from stack}
@<OUT: Declare local variables (if any) for |do_pop|@>@;
begin if stack_ptr=0 then bad_dvi;
cur_stack:=stack[stack_ptr]; decr(stack_ptr);
@<OUT: Process a |pop|@>@;@/
end;

@ The |do_xxx| procedure is called in order to process a special command.
The bytes of the special string have been put into |byte_mem| as the
current string. They are converted to a temporary packet and discarded
again.

@p procedure do_xxx;@/
@<OUT: Declare additional local variables for |do_xxx|@>@;
@!p:pckt_pointer; {temporary packet}
begin p:=new_packet;@/
@<OUT: Process an |xxx|@>@;@/
flush_packet;
end;

@ Next are the movement commands:
The |do_right| procedure is called in order to process the horizontal
movement commands |right|, |w|, and |x|.


@p procedure do_right;@/
@<OUT: Declare local variables (if any) for |do_right|@>@;
begin if cur_class>=w_cl then cur_w_x[cur_class-w_cl]:=cur_parm
else if cur_class<right_cl then cur_parm:=cur_w_x[cur_class-w0_cl];
@<OUT: Process a |right| or |w| or |x|@>@;@/
h_upd_move(cur_parm)(h_pixel_round(cur_parm));
@<OUT: Move right@>@;
end;

@ The |do_down| procedure is called in order to process the vertical
movement commands |down|, |y|, and |z|.

@p procedure do_down;@/
@<OUT: Declare local variables (if any) for |do_down|@>@;
begin if cur_class>=y_cl then cur_y_z[cur_class-y_cl]:=cur_parm
else if cur_class<down_cl then cur_parm:=cur_y_z[cur_class-y0_cl];
@<OUT: Process a |down| or |y| or |z|@>@;@/
v_upd_move(cur_parm)(v_pixel_round(cur_parm));
@<OUT: Move down@>@;
end;

@ The |do_width| procedure, or actually the |do_a_width| macro, is
called in order to increase the current horizontal position |cur_h| by
|cur_h_dimen| in exactly the same way as if a character of width
|cur_h_dimen| had been typeset.

@d do_a_width(#)==
  begin @!device h_pixels:=#; @+ ecived @; @+ do_width;
  end

@p procedure do_width;@/
@<OUT: Declare local variables (if any) for |do_width|@>@;
begin @<OUT: Typeset a |width|@>@;@/
h_upd_char(cur_h_dimen)(h_pixels);
@<OUT: Move right@>@;
end;

@ Finally we have the commands for the typesetting of rules and characters;
the global variable |cur_upd| is |true| if the horizontal position shall
be updated (\\{set} commands).

The |do_rule| procedure is called in order to typeset a rule.

@p procedure do_rule;@/
@<OUT: Declare additional local variables |do_rule|@>@;
@!visible:boolean;
begin h_rule_pixels@;
if (cur_h_dimen>0)and(cur_v_dimen>0) then
  begin visible:=true; v_rule_pixels@;
  @<OUT: Typeset a visible |rule|@>@;
  end
else  begin visible:=false;
  @<OUT: Typeset an invisible |rule|@>@;
  end;
if cur_upd then
  begin h_upd_move(cur_h_dimen)(h_pixels);
  @<OUT: Move right@>@;
  end;
end;

@ Last not least the |do_char| procedure is called in order to typeset
character~|cur_res| with extension~|cur_ext| from the real font~|cur_fnt|.

@p procedure do_char;@/
@<OUT: Declare local variables (if any) for |do_char|@>@;
begin @<OUT: Typeset a |char|@>@;
if cur_upd then
  begin h_upd_char(widths[cur_wp])(char_pixels[cur_cp]);
  @<OUT: Move right@>@;
  end;
end;

@ If the program terminates abnormally, the following code may be
invoked in the middle of a page.

@<Finish output file(s)@>=
begin if type_setting then @<OUT: Finish incomplete page@>;
@<OUT: Finish output file(s)@>@;
end

@ When the first character of font~|cur_fnt| is about to be typeset, the
|do_font| procedure is called in order to decide whether this is a
virtual font or a real font.

One step in this decision is the attempt to find and read the \.{VF}
file for this font; other attempts to locate a font file may be
performed before and after that, depending on the nature of the output
device and on the structure of the file system at a particular
installation.  For a real device we convert the character widths to
(horizontal) pixels.

In any case |do_font| must change |font_type(cur_fnt)| to a value
|>defined_font|; as a last resort one might use the \.{TFM} width data
and draw boxes or leave blank spaces in the output.

@p procedure do_font;@/
label done;@/
@<OUT: Declare additional local variables for |do_font|@>@;
@!p:char_pointer; {index into |char_widths| and |char_pixels|}
begin @!debug if font_type(cur_fnt)=defined_font then confusion(str_fonts);
gubed@;
p:=0; {such that |p| is used}
@!device for p:=font_chars(cur_fnt)+font_bc(cur_fnt)
  to font_chars(cur_fnt)+font_ec(cur_fnt) do
    char_pixels[p]:=h_pixel_round(widths[char_widths[p]]);
ecived@;
@<OUT: Look for a font file before trying to read the \.{VF} file;
  if found |goto done|@>@;@/
if do_vf then goto done; {try to read the \.{VF} file}
@<OUT: Look for a font file after trying to read the \.{VF} file@>@;@/
done:
@!debug if font_type(cur_fnt)<=loaded_font then confusion(str_fonts);
gubed@;
end;

@ Before a character of font~|cur_fnt| is typeset the following piece of
code ensures that the font is ready to be used.

@<Prepare to use font |cur_fnt|@>=
@<OUT: Prepare to use font |cur_fnt|@>@;
if font_type(cur_fnt)<=loaded_font then do_font {|cur_fnt| was not yet used}

@* Interpreting VF packets.
The |pckt_first_par| procedure first reads a \.{DVI} command byte from
the packet into |cur_cmd|; then |cur_parm| is set to the value of the
first parameter (if any) and |cur_class| to the command class.

@p procedure pckt_first_par;
begin cur_cmd:=pckt_ubyte;
case dvi_par[cur_cmd] of
char_par: set_cur_char(pckt_ubyte);
no_par: do_nothing;
dim1_par: cur_parm:=pckt_sbyte;
num1_par: cur_parm:=pckt_ubyte;
dim2_par: cur_parm:=pckt_spair;
num2_par: cur_parm:=pckt_upair;
dim3_par: cur_parm:=pckt_strio;
num3_par: cur_parm:=pckt_utrio;
three_cases(dim4_par): cur_parm:=pckt_squad; {|dim4|, |num4|, or |numu|}
rule_par:
  begin cur_v_dimen:=pckt_squad; cur_h_dimen:=pckt_squad;
  cur_upd:=(cur_cmd=set_rule);
  end;
fnt_par:cur_parm:=cur_cmd-fnt_num_0;
othercases abort('internal error');
endcases;
cur_class:=dvi_cl[cur_cmd];
end;

@ The |do_vf_packet| procedure is called in order to interpret the
character packet for a virtual character. Such a packet may contain the
instruction to typeset a character from the same or an other virtual
font; in such cases |do_vf_packet| calls itself recursively. The
recursion level, i.e., the number of times this has happened, is kept
in the global variable |n_recur| and should not exceed |max_recursion|.
@^recursion@>

@<Types...@>=
@!recur_pointer=0..max_recursion;

@ The \.{\title} processor should detect an infinite recursion caused by
bad \.{VF} files; thus a new recursion level is entered even in cases
where this could be avoided without difficulty.

If the recursion level exceeds the allowed maximum, we want to give
a traceback how this has happened; thus some of the global variables
used in different invocations of |do_vf_packet| are saved in a stack,
others are saved as local variables of |do_vf_packet|.

@<Glob...@>=
@!recur_fnt:array[recur_pointer] of font_number; {this packet's font}
@!recur_ext:array[recur_pointer] of int_24; {this packet's extension}
@!recur_res:array[recur_pointer] of eight_bits; {this packet's residue}
@!recur_pckt:array[recur_pointer] of pckt_pointer; {the packet}
@!recur_loc:array[recur_pointer] of byte_pointer; {next byte of packet}
@!n_recur:recur_pointer; {current recursion level}
@!recur_used:recur_pointer; {highest recursion level used so far}

@ @<Set init...@>=
n_recur:=0; recur_used:=0;

@ Here now is the |do_vf_packet| procedure.

@p procedure do_vf_packet;
label continue,found,done;
var k:recur_pointer; {loop index}
@!f:int_8u; {packet type flag}
@!save_upd:boolean; {used to save |cur_upd|}
@!save_cp:width_pointer; {used to save |cur_cp|}
@!save_wp:width_pointer; {used to save |cur_wp|}
@!save_limit:byte_pointer; {used to save |cur_limit|}
begin @<VF: Save values on entry to |do_vf_packet|@>;@/
@<VF: Interpret the \.{DVI} commands in the packet@>@;@/
if save_upd then
  begin cur_h_dimen:=widths[save_wp]; do_a_width(char_pixels[save_cp]);
  end;
@<VF: Restore values on exit from |do_vf_packet|@>;@/
end;

@ On entry to |do_vf_packet| several values must be saved.

@<VF: Save values on entry to |do_vf_packet|@>=
save_upd:=cur_upd; save_cp:=cur_cp; save_wp:=cur_wp;@/
recur_fnt[n_recur]:=cur_fnt;
recur_ext[n_recur]:=cur_ext;
recur_res[n_recur]:=cur_res

@ Some of these values must be restored on exit from |do_vf_packet|.

@<VF: Restore values on exit from |do_vf_packet|@>=
cur_fnt:=recur_fnt[n_recur]

@ If |cur_pckt| is the empty packet, we manufacture a |put| command;
otherwise we read and interpret \.{DVI} commands from the packet.

@<VF: Interpret the \.{DVI} commands in the packet@>=
if find_packet then f:=cur_type @+ else goto done;
recur_pckt[n_recur]:=cur_pckt;
save_limit:=cur_limit;
cur_fnt:=font_font(cur_fnt);
if cur_pckt=empty_packet then
  begin cur_class:=char_cl; goto found;
  end;
if cur_loc>=cur_limit then goto done;
continue: pckt_first_par;
found: case cur_class of
char_cl: @<VF: Typeset a |char|@>;
rule_cl: do_rule;
xxx_cl:
  begin pckt_room(cur_parm);
  while cur_parm>0 do
    begin append_byte(pckt_ubyte); decr(cur_parm);
    end;
  do_xxx;
  end;
push_cl: do_push;
pop_cl: do_pop;
five_cases(w0_cl): do_right; {|right|, |w|, or |x|}
five_cases(y0_cl): do_down; {|down|, |y|, or |z|}
fnt_cl: cur_fnt:=cur_parm;
othercases confusion(str_packets); {font definition or invalid}
endcases;
if cur_loc<cur_limit then goto continue;
done:

@ The final |put| of a simple packet may be changed into |set_char| or
\\{set}.

@<VF: Typeset a |char|@>=
begin @<Prepare to use font |cur_fnt|@>;
cur_cp:=font_chars(cur_fnt)+cur_res; cur_wp:=char_widths[cur_cp];
if (cur_loc=cur_limit)and(f=vf_simple) and save_upd then
  begin save_upd:=false; cur_upd:=true;
  end;
if font_type(cur_fnt)=vf_font_type then
  @<VF: Enter a new recursion level@>
else do_char;
end

@ Before entering a new recursion level we must test for overflow; in
addition a few variables must be saved and restored.
A |set_char| or \\{set} followed by |pop| is changed into |put|.

@<VF: Enter a new recursion level@>=
begin recur_loc[n_recur]:=cur_loc; {save}
if cur_loc<cur_limit then
  if byte_mem[cur_loc]=bi(pop) then cur_upd:=false;
if n_recur=recur_used then
  if recur_used=max_recursion then
    @<VF: Display the recursion traceback and terminate@>
  else incr(recur_used);@/
incr(n_recur); do_vf_packet; decr(n_recur); {recurse}
cur_loc:=recur_loc[n_recur]; cur_limit:=save_limit; {restore}
end

@ @<VF: Display the recursion traceback and terminate@>=
begin print_ln(' !Infinite VF recursion?');
@.Infinite VF recursion?@>
for k:=max_recursion downto 0 do
  begin print('level=',k:1,' font');
  d_print('=',recur_fnt[k]:1);
  print_font(recur_fnt[k]);
  print(' char=',recur_res[k]:1);
  if recur_ext[k]<>0 then print('.',recur_ext[k]:1);
  new_line;
  @!debug hex_packet(recur_pckt[k]); print_ln('loc=',recur_loc[k]:1);
  gubed@;
  end;
overflow(str_recursion,max_recursion);
end

@* Interpreting the DVI file.
The |do_dvi| procedure reads the entire \.{DVI} file and initiates
whatever actions may be necessary.

@p procedure do_dvi;
label done,exit;
var temp_byte:int_8u; {byte for temporary variables}
@!temp_int:int_32; {integer for temporary variables}
@!dvi_start:int_32; {starting location}
@!dvi_bop_post:int_32; {location of |bop| or |post|}
@!dvi_back:int_32; {a back pointer}
@!k:int_15; {general purpose variable}
begin @<DVI: Process the preamble@>;
if random_reading then @<DVI: Process the postamble@>;
repeat dvi_first_par;
  while cur_class=fnt_def_cl do
    begin dvi_do_font(random_reading); dvi_first_par;
    end;
  if cur_cmd=bop then @<DVI: Process one page@>;
until cur_cmd<>eop;
if cur_cmd<>post then bad_dvi;
exit:end;

@ @<DVI: Process the preamble@>=
if dvi_ubyte<>pre then bad_dvi;
if dvi_ubyte<>dvi_id then bad_dvi;
dvi_num:=dvi_pquad; dvi_den:=dvi_pquad; dvi_mag:=dvi_pquad;
tfm_conv:=(25400000.0/dvi_num)*(dvi_den/473628672)/16.0;
temp_byte:=dvi_ubyte; pckt_room(temp_byte);
for k:=1 to temp_byte do append_byte(dvi_ubyte);
print('DVI file: '''); print_packet(new_packet); print_ln(''',');
print('   num=',dvi_num:1,', den=',dvi_den:1,', mag=',dvi_mag:1);
if out_mag<=0 then out_mag:=dvi_mag @+ else print(' => ',out_mag:1);
print_ln('.');
do_pre; flush_packet

@ @<Glob...@>=
@!dvi_num:int_31; {numerator}
@!dvi_den:int_31; {denominator}
@!dvi_mag:int_31; {magnification}

@ @<DVI: Process the postamble@>=
begin dvi_start:=dvi_loc; {remember start of first page}
@<DVI: Find the postamble@>;
d_print_ln('DVI: postamble at ',dvi_bop_post:1);
dvi_back:=dvi_pointer;
if dvi_num<>dvi_pquad then bad_dvi;
if dvi_den<>dvi_pquad then bad_dvi;
if dvi_mag<>dvi_pquad then bad_dvi;
temp_int:=dvi_squad; temp_int:=dvi_squad;
if stack_size<dvi_upair then overflow(str_stack,stack_size);
temp_int:=dvi_upair;
dvi_first_par;
while cur_class=fnt_def_cl do
  begin dvi_do_font(false); dvi_first_par;
  end;
if cur_cmd<>post_post then bad_dvi;
if not selected then @<DVI: Find the starting page@>;
dvi_move(dvi_start); {go to first or starting page}
end

@ @<DVI: Find the postamble@>=
temp_int:=dvi_length-5;
repeat if temp_int<49 then bad_dvi;
dvi_move(temp_int); temp_byte:=dvi_ubyte; decr(temp_int);
until temp_byte<>dvi_pad;
if temp_byte<>dvi_id then bad_dvi;
dvi_move(temp_int-4); if dvi_ubyte<>post_post then bad_dvi;
dvi_bop_post:=dvi_pointer;
if (dvi_bop_post<15)or(dvi_bop_post>dvi_loc-34) then bad_dvi;
dvi_move(dvi_bop_post); if dvi_ubyte<>post then bad_dvi

@ @<DVI: Find the starting page@>=
begin dvi_start:=dvi_bop_post; {just in case}
while dvi_back<>-1 do
  begin if (dvi_back<15)or(dvi_back>dvi_bop_post-46) then bad_dvi;
  dvi_bop_post:=dvi_back; dvi_move(dvi_back);
  if dvi_ubyte<>bop then bad_dvi;
  for k:=0 to 9 do count[k]:=dvi_squad;
  if start_match then dvi_start:=dvi_bop_post;
  dvi_back:=dvi_pointer;
  end;
end

@ When a |bop| has been read, the \.{DVI} commands for one page are
interpreted until an |eop| is found.

@<DVI: Process one page@>=
begin for k:=0 to 9 do count[k]:=dvi_squad;
temp_int:=dvi_pointer; do_bop;
dvi_first_par;
if type_setting then @<DVI: Process a page; then |goto done|@>
else @<DVI: Skip a page; then |goto done|@>;
done:if cur_cmd<>eop then bad_dvi;
if selected then
  begin do_eop;
  if all_done then return;
  end;
end

@ All \.{DVI} commands are processed, as long as |cur_class<>invalid_cl|;
then we should have found an |eop|.

@<DVI: Process a page; then |goto done|@>=
loop begin
  case cur_class of
  char_cl: @<DVI: Typeset a |char|@>;
  rule_cl:
    if cur_upd and(cur_v_dimen=width_dimen) then
      do_a_width(h_pixel_round(cur_h_dimen))
    else do_rule;
  xxx_cl:
    begin pckt_room(cur_parm);
    while cur_parm>0 do
      begin append_byte(dvi_ubyte); decr(cur_parm);
      end;
    do_xxx;
    end;
  push_cl: do_push;
  pop_cl: do_pop;
  five_cases(w0_cl): do_right; {|right|, |w|, or |x|}
  five_cases(y0_cl): do_down; {|down|, |y|, or |z|}
  fnt_cl: dvi_font;
  fnt_def_cl: dvi_do_font(random_reading);
  invalid_cl: goto done;
  othercases abort('internal error');
  endcases;
dvi_first_par; {get the next command}
end

@ While skipping a page all commands other than font definitions are
ignored.

@<DVI: Skip a page; then |goto done|@>=
loop begin
  case cur_class of
  xxx_cl: while cur_parm>0 do
    begin temp_byte:=dvi_ubyte; decr(cur_parm);
    end;
  fnt_def_cl: dvi_do_font(random_reading);
  invalid_cl: goto done;
  othercases do_nothing;
  endcases;
dvi_first_par; {get the next command}
end

@ @<DVI: Typeset a |char|@>=
begin @<Prepare to use font |cur_fnt|@>;
set_cur_wp(cur_fnt)(bad_dvi);
if font_type(cur_fnt)=vf_font_type then do_vf_packet @+ else do_char;
end

@* The main program.
The code for real devices is still rather incomplete.
Moreover several branches of the program have not been tested because
they are never used with \.{DVI} files made by \TeX\ and \.{VF} files
made by \.{VPtoVF}.

@ At the end of the program the output file(s) have to be finished and
on some systems it may be necessary to close input and\slash or output
files.
@^system dependencies@>

@p procedure close_files_and_terminate;
var k:@!int_15; {general purpose index}
begin close_in(dvi_file);
if history<fatal_message then @<Finish output file(s)@>;
stat @<Print memory usage statistics@>;@+tats@;@/
@<Close output file(s)@>@;
@<Print the job |history|@>;
end;

@ Now we are ready to put it all together.
Here is where \.{\title} starts, and where it ends.
@^system dependencies@>

@p begin initialize; {get all variables initialized}
@<Initialize predefined strings@>@;
dialog; {get options}
@<Open input file(s)@>@;
@<Open output file(s)@>@;
do_dvi; {process the entire \.{DVI} file}
close_files_and_terminate;
final_end:end.

@ @<Print memory usage statistics@>=
print_ln('Memory usage statistics:');
print(dvi_nf:1,' dvi, ',lcl_nf:1,' local, ');
@<Print more font usage statistics@>@;@/
print_ln('and ',nf:1,' internal fonts of ',max_fonts:1);
print_ln(n_widths:1,' widths of ',max_widths:1,' for ',
  n_chars:1,' characters of ',max_chars:1);
print_ln(pckt_ptr:1,' byte packets of ',max_packets:1,' with ',
  byte_ptr:1,' bytes of ',max_bytes:1);
@<Print more memory usage statistics@>@;@/
print_ln(stack_used:1,' of ',stack_size:1,' stack and ',
  recur_used:1,' of ',max_recursion:1,' recursion levels.')

@ Some implementations may wish to pass the |history| value to the
operating system so that it can be used to govern whether or not other
programs are started. Here we simply report the history to the user.
@^system dependencies@>

@<Print the job |history|@>=
case history of
spotless: print_ln('(No errors were found.)');
harmless_message: print_ln('(Did you see the warning message above?)');
error_message: print_ln('(Pardon me, but I think I spotted something wrong.)');
fatal_message: print_ln('(That was a fatal error, my friend.)');
end {there are no other cases}

@* Low-level output routines.
The program uses the binary file variable |out_file| for its main output
file; |out_loc| is the number of the byte about to be written next on
|out_file|.

@<Glob...@>=
@!out_file:byte_file; {the \.{DVI} file we are writing}
@!out_loc:int_32; {where we are about to write, in |out_file|}
@!out_back:int_32; {a back pointer}
@!out_max_v:int_31; {maximum |v| value so far}
@!out_max_h:int_31; {maximum |h| value so far}
@!out_stack:int_16u; {maximum stack depth}
@!out_pages:int_16u; {total number of pages}

@ @<Set ini...@>=
out_loc:=0; out_back:=-1;
out_max_v:=0; out_max_h:=0;
out_stack:=0; out_pages:=0;

@ To prepare |out_file| for output, we |rewrite| it.

@<Open output file(s)@>=
rewrite(out_file); {prepares to write packed bytes to |out_file|}

@ For some operating systems it may be necessary to close |out_file|.

@<Close output file(s)@>=

@ Writing the |out_file| should be done as efficient as possible for a
particular system; on many systems this means that a large number of
bytes will be accumulated in a buffer and is then written from that
buffer to |out_file|. In order to simplify such system dependent changes
we use the \.{WEB} macro |out_byte| to write the next \.{DVI} byte. Here
we give a simple minded definition for this macro in terms of standard
\PASCAL.
@^system dependencies@>
@^optimization@>

@d out_byte(#) == write(out_file,#) {write next \.{DVI} byte}

@ The \.{WEB} macro |out_one| is used to write one byte and to update
|out_loc|.

@d out_one(#) == begin out_byte(#); incr(out_loc); @+ end

@ First the |out_packet| procedure copies a packet to |out_file|.

@<Declare typesetting procedures@>=
procedure out_packet(@!p:pckt_pointer);
var k:byte_pointer; {index into |byte_mem|}
begin Incr(out_loc)(pckt_length(p));
for k:=pckt_start[p] to pckt_start[p+1]-1 do out_byte(bo(byte_mem[k]));
end;

@ Next are the procedures used to write integer numbers or even complete
\.{DVI} commands to |out_file|; they all keep |out_loc| up to date.

The |out_four| procedure outputs four bytes in two's complement notation,
without risking arithmetic overflow.

@<Declare typesetting procedures@>=
procedure out_four(@!x:int_32); {output four bytes}
@!begin_four; comp_four(out_byte); Incr(out_loc)(4);
end;

@ The |out_char| procedure outputs a |set_char| or \\{set} command or, if
|upd=false|, a |put| command.

@<Declare typesetting procedures@>=
procedure out_char(@!upd:boolean;@!ext:int_32;@!res:eight_bits);
  {output \\{set} or |put|}
@!begin_char; comp_char(out_one);
end;

@ The |out_unsigned| procedure outputs a |fnt|, |xxx|, or |fnt_def|
command with its first parameter (normally unsigned); a |fnt| command
is converted into |fnt_num| whenever this is possible.

@<Declare typesetting procedures@>=
procedure out_unsigned(@!o:eight_bits;@!x:int_32);
  {output |fnt_num|, |fnt|, |xxx|, or |fnt_def|}
@!begin_unsigned; comp_unsigned(out_one);
end;

@ The |out_signed| procedure outputs a movement (|right|, |w|,
|x|, |down|, |y|, or |z|) command with its (signed) parameter.

@<Declare typesetting procedures@>=
procedure out_signed(@!o:eight_bits;@!x:int_32);
  {output |right|, |w|, |x|, |down|, |y|, or |z|}
@!begin_signed; comp_signed(out_one);
end;

@ For an output font we set |font_type(f):=out_font_type|; in this case
|font_font(f)| is the font number used for font~|f| in |out_file|.
@^font types@>

The global variable |out_nf| is the number of fonts already used in
|out_file| and the array |out_fnts| contains their internal font numbers;
the current font in |out_file| is called |out_fnt|.

@<Glob...@>=
@!out_fnts:array[font_number] of font_number; {internal font numbers}
@!out_nf:font_number; {number of fonts used in |out_file|}
@!out_fnt:font_number; {internal font number of current output font}

@ @<Set init...@>=
out_nf:=0;

@ @<Print more font usage statistics@>=
print(out_nf:1,' out, ');

@ The |out_fnt_def| procedure outputs a complete font definition
command.

@<Declare typesetting procedures@>=
procedure out_fnt_def(@!f:font_number);
var p:pckt_pointer; {the font name packet}
@!k,@!l:byte_pointer; {indices into |byte_mem|}
@!a:eight_bits; {length of area part}
begin out_unsigned(fnt_def1,font_font(f)); out_four(font_check(f));
out_four(font_scaled(f)); out_four(font_design(f));@/
p:=font_name(f); k:=pckt_start[p]; l:=pckt_start[p+1]-1;
a:=bo(byte_mem[k]);@/
Incr(out_loc)(l-k+2); out_byte(a); out_byte(l-k-a);
while k<l do
  begin incr(k); out_byte(bo(byte_mem[k]));
  end;
end;

@* Writing the output file.
Here we define the device dependent parts of the typesetting routines
described earlier in this program.

First we define a few quantities required by the device dependent code
for a real output device in order to demonstrate how they might be
defined and in order to be able to compile \.{DVIcopy} with the device
dependent code included.

@d h_resolution==300 {horizontal resolution in pixels per inch (dpi)}
@d v_resolution==300 {vertical resolution in pixels per inch (dpi)}

@d max_h_drift==2 {we insist that |abs(hh-h_pixel_round(h))<=max_h_drift|}
@d max_v_drift==2 {we insist that |abs(vv-v_pixel_round(v))<=max_v_drift|}

@<Glob...@>=
@!device
@!h_conv:real; {converts \.{DVI} units to horizontal pixels}
@!v_conv:real; {converts \.{DVI} units to vertical pixels}
ecived

@ These are the local variables (if any) needed for |do_pre|.

@<OUT: Declare local variables (if any) for |do_pre|@>=
var k:int_15; {general purpose variable}
@!p,@!q,@!r:byte_pointer; {indices into |byte_mem|}
@!comment:packed array[1..comm_length] of char; {preamble comment prefix}

@ And here is the device dependent code for |do_pre|; the \.{DVI} preamble
comment written to |out_file| is similar to the one produced by \.{GFtoPK},
but we want to apply our preamble comment prefix only once.

@<OUT: Process the |pre|@>=
out_one(pre); out_one(dvi_id);
out_four(dvi_num); out_four(dvi_den); out_four(out_mag);@/
p:=pckt_start[pckt_ptr-1]; q:=byte_ptr; {location of old \.{DVI} comment}
comment:=preamble_comment; pckt_room(comm_length);
for k:=1 to comm_length do append_byte(xord[comment[k]]);
while byte_mem[p]=bi(" ") do incr(p); {remove leading blanks}
if p=q then Decr(byte_ptr)(from_length)
else begin k:=0;
  while (k<comm_length)and(byte_mem[p+k]=byte_mem[q+k]) do incr(k);
  if k=comm_length then Incr(p)(comm_length);
  end;
k:=byte_ptr-p; {total length}
if k>255 then
  begin k:=255; q:=p+255-comm_length; {at most 255 bytes}
  end;
out_one(k); out_packet(new_packet); flush_packet;
for r:=p to q-1 do out_one(bo(byte_mem[r]));

@ These are the additional local variables (if any) needed for |do_bop|;
the variables |@!i| and |@!j| are already declared.

@<OUT: Declare additional local variables |do_bop|@>=
var

@ And here is the device dependent code for |do_bop|.

@<OUT: Process a |bop|@>=
out_one(bop); incr(out_pages);
for i:=0 to 9 do out_four(count[i]);
out_four(out_back); out_back:=out_loc-45;
out_fnt:=invalid_font;

@ These are the local variables (if any) needed for |do_eop|.

@<OUT: Declare local variables (if any) for |do_eop|@>=

@ And here is the device dependent code for |do_eop|.

@<OUT: Process an |eop|@>=
out_one(eop);

@ These are the local variables (if any) needed for |do_push|.

@<OUT: Declare local variables (if any) for |do_push|@>=

@ And here is the device dependent code for |do_push|.

@<OUT: Process a |push|@>=
if stack_ptr>out_stack then out_stack:=stack_ptr;
out_one(push);

@ These are the local variables (if any) needed for |do_pop|.

@<OUT: Declare local variables (if any) for |do_pop|@>=

@ And here is the device dependent code for |do_pop|.

@<OUT: Process a |pop|@>=
out_one(pop);

@ These are the additional local variables (if any) needed for |do_xxx|;
the variable |@!p|, the pointer to the packet containing the special
string, is already declared.

@<OUT: Declare additional local variables for |do_xxx|@>=
var

@ And here is the device dependent code for |do_xxx|.

@<OUT: Process an |xxx|@>=
out_unsigned(xxx1,pckt_length(p)); out_packet(p);

@ These are the local variables (if any) needed for |do_right|.

@<OUT: Declare local variables (if any) for |do_right|@>=

@ And here is the device dependent code for |do_right|.

@<OUT: Process a |right| or |w| or |x|@>=
if cur_class<right_cl then out_one(cur_cmd) {|w0| or |x0|}
else out_signed(dvi_right_cmd[cur_class],cur_parm); {|right|, |w|, or |x|}

@ Here we update the |out_max_h| value.

@<OUT: Move right@>=
if abs(cur_h)>out_max_h then out_max_h:=abs(cur_h);

@ These are the local variables (if any) needed for |do_down|.

@<OUT: Declare local variables (if any) for |do_down|@>=

@ And here is the device dependent code for |do_down|.

@<OUT: Process a |down| or |y| or |z|@>=
if cur_class<down_cl then out_one(cur_cmd) {|y0| or |z0|}
else out_signed(dvi_down_cmd[cur_class],cur_parm); {|down|, |y|, or |z|}

@ Here we update the |out_max_v| value.

@<OUT: Move down@>=
if abs(cur_v)>out_max_v then out_max_v:=abs(cur_v);

@ These are the local variables (if any) needed for |do_width|.

@<OUT: Declare local variables (if any) for |do_width|@>=

@ And here is the device dependent code for |do_width|.

@<OUT: Typeset a |width|@>=
out_one(set_rule);
out_four(width_dimen); out_four(cur_h_dimen);

@ These are the additional local variables (if any) needed for |do_rule|;
the variable |@!visible| is already declared.

@<OUT: Declare additional local variables |do_rule|@>=
var

@ And here is the device dependent code for |do_rule|.

@<OUT: Typeset a visible |rule|@>=
out_one(dvi_rule_cmd[cur_upd]);
out_four(cur_v_dimen); out_four(cur_h_dimen);

@ @<OUT: Typeset an invisible |rule|@>=
@<OUT: Typeset a visible |rule|@>

@ These are the additional local variables (if any) needed for |do_font|;
the variable |@!p| is already declared.

@<OUT: Declare additional local variables for |do_font|@>=
var

@ And here is the device dependent code for |do_font|; if the \.{VF} file
for a font could not be found, we simply assume this must be a real font.

@<OUT: Look for a font file before trying to read the \.{VF} file;
  if found |goto done|@>=

@ @<OUT: Look for a font file after trying to read the \.{VF} file@>=
if(out_nf>=max_fonts) then overflow(str_fonts,max_fonts);
print('OUT: font ',cur_fnt:1); d_print(' => ',out_nf:1);
print_font(cur_fnt);
d_print(' at ',font_scaled(cur_fnt):1,' DVI units'); print_ln('.');
font_type(cur_fnt):=out_font_type; font_font(cur_fnt):=out_nf;
out_fnts[out_nf]:=cur_fnt; incr(out_nf);
out_fnt_def(cur_fnt);

@ And here is some device dependent code used before each character.

@<OUT: Prepare to use font |cur_fnt|@>=

@ These are the local variables (if any) needed for |do_char|.

@<OUT: Declare local variables (if any) for |do_char|@>=

@ And here is the device dependent code for |do_char|.

@<OUT: Typeset a |char|@>=
@!debug if font_type(cur_fnt)<>out_font_type then confusion(str_fonts);
gubed @;
if cur_fnt<>out_fnt then
  begin out_unsigned(fnt1,font_font(cur_fnt)); out_fnt:=cur_fnt;
  end;
out_char(cur_upd,cur_ext,cur_res);

@ If the program terminates in the middle of a page, we write as many
|pop|s as necessary and one |eop|.

@<OUT: Finish incomplete page@>=
begin while stack_ptr>0 do
  begin out_one(pop); decr(stack_ptr);
  end;
  out_one(eop);
end

@ If the output file has been started, we write the postamble; in
addition we print the number of bytes and pages written to |out_file|.

@<OUT: Finish output file(s)@>=
if out_loc>0 then
  begin @<OUT: Write the postamble@>;
  k:=7-((out_loc-1) mod 4); {the number of |dvi_pad| bytes}
  while k>0 do
    begin out_one(dvi_pad); decr(k);
    end;
  print('OUT file: ',out_loc:1,' bytes, ',out_pages:1,' page');
  if out_pages<>1 then print('s');
  end
else print('OUT file: no output');
print_ln(' written.');
if out_pages=0 then mark_harmless;

@ Here we simply write the values accumulated during the \.{DVI} output.

@<OUT: Write the postamble@>=
out_one(post); out_four(out_back); out_back:=out_loc-5;@/
out_four(dvi_num); out_four(dvi_den); out_four(out_mag);@/
out_four(out_max_v); out_four(out_max_h);@/
out_one(out_stack div @"100); out_one(out_stack mod @"100);@/
out_one(out_pages div @"100); out_one(out_pages mod @"100);@/
k:=out_nf;
while k>0 do
  begin decr(k); out_fnt_def(out_fnts[k]);
  end;
out_one(post_post); out_four(out_back);@/
out_one(dvi_id)

@ Here we could print more memory usage statistics; this possibility is,
however, not used for \.{DVIcopy}.

@<Print more memory usage statistics@>=

@* System-dependent changes.
This section should be replaced, if necessary, by changes to the program
that are necessary to make \.{DVIcopy} work at a particular installation.
It is usually best to design your change file so that all changes to
previous sections preserve the section numbering; then everybody's version
will be consistent with the printed program. More extensive changes,
which introduce new sections, can be inserted here; then only the index
itself will get a new section number.
@^system dependencies@>

@* Index.
Pointers to error messages appear here together with the section numbers
where each ident\-i\-fier is used.