path: root/Build/source/texk/web2c/vptovf.web

% This program by D. E. Knuth is not copyrighted and can be used freely.
% Version 1 was implemented in December 1989.
% Version 1.1 fixed some for-loop indices for stricter Pascal (April 1990).
% Version 1.2 fixed `nonexistent char 0' bug, and a bit more (September 1990).
% Version 1.3 has more robust `out_scaled' (March 1991).
% Version 1.4 (March 1995) initialized lk_step_ended (Armin K\"ollner).
% Version 1.5 (August 1998) corrected vf_fix(0) (Wayne Sullivan).
% Version 1.6 (January 2014) corrected possible end-of-line glitch (Ken Nakano),
%  and get_fix now treats -- as + (Peter Breitenlohner).

% 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=logo10 % for the METAFONT logo
\def\MF{{\logo METAFONT}}

\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{VP\lowercase{to}VF}
\def\contentspagenumber{201}
\def\topofcontents{\null
  \def\titlepage{F} % include headline on the contents page
  \def\rheader{\mainfont\hfil \contentspagenumber}
  \vfill
  \centerline{\titlefont The {\ttitlefont VPtoVF} processor}
  \vskip 15pt
  \centerline{(Version 1.6, January 2014)}
  \vfill}
\def\botofcontents{\vfill
  \centerline{\hsize 5in\baselineskip9pt
    \vbox{\ninerm\noindent
    The preparation of this program
    was supported in part by the National Science
    Foundation and by the System Development Foundation. `\TeX' is a
    trademark of the American Mathematical Society.}}}
\pageno=\contentspagenumber \advance\pageno by 1

@* Introduction.
The \.{VPtoVF} utility program converts virtual-property-list (``\.{VPL}'')
files into an equivalent pair of files called a virtual font (``\.{VF}'') file
and a \TeX\ font metric (``\.{TFM}'') file. It also makes a thorough check
of the given \.{VPL} file, so that the \.{VF} file should be acceptable to
device drivers and the \.{TFM} file should be acceptable to \TeX.

\indent\.{VPtoVF} is an extended version of the program \.{PLtoTF}, which
is part of the standard \TeX ware library.
The idea of a virtual font was inspired by the work of David R. Fuchs
@^Fuchs, David Raymond@>
who designed a similar set of conventions in 1984 while developing a
device driver for ArborText, Inc. He wrote a somewhat similar program
called \.{PLFONT}.

The |banner| string defined here should be changed whenever \.{VPtoVF}
gets modified.

@d banner=='This is VPtoVF, Version 1.6' {printed when the program starts}

@ This program is written entirely in standard \PASCAL, except that
it has to do some slightly system-dependent character code conversion
on input. Furthermore, lower case letters are used in error messages;
they could be converted to upper case if necessary. The input is read
from |vpl_file|, and the output is written on |vf_file| and |tfm_file|;
error messages and
other remarks are written on the |output| file, which the user may
choose to assign to the terminal if the system permits it.
@^system dependencies@>

The term |print| is used instead of |write| when this program writes on
the |output| file, so that all such output can be easily deflected.

@d print(#)==write(#)
@d print_ln(#)==write_ln(#)

@p program VPtoVF(@!vpl_file,@!vf_file,@!tfm_file,@!output);
const @<Constants in the outer block@>@/
type @<Types in the outer block@>@/
var @<Globals in the outer block@>@/
procedure initialize; {this procedure gets things started properly}
  var @<Local variables for initialization@>@/
  begin print_ln(banner);@/
  @<Set initial values@>@/
  end;

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

@<Constants...@>=
@!buf_size=60; {length of lines displayed in error messages}
@!max_header_bytes=100; {four times the maximum number of words allowed in
  the \.{TFM} file header block, must be 1024 or less}
@!vf_size=10000; {maximum length of |vf| data, in bytes}
@!max_stack=100; {maximum depth of simulated \.{DVI} stack}
@!max_param_words=30; {the maximum number of \.{fontdimen} parameters allowed}
@!max_lig_steps=5000;
  {maximum length of ligature program, must be at most $32767-257=32510$}
@!max_kerns=500; {the maximum number of distinct kern values}
@!hash_size=5003; {preferably a prime number, a bit larger than the number
  of character pairs in lig/kern steps}

@ 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 do_nothing == {empty statement}

@* Property list description of font metric data.
The idea behind \.{VPL} files is that precise details about fonts, i.e., the
facts that are needed by typesetting routines like \TeX, sometimes have to
be supplied by hand. The nested property-list format provides a reasonably
convenient way to do this.

A good deal of computation is necessary to parse and process a
\.{VPL} file, so it would be inappropriate for \TeX\ itself to do this
every time it loads a font. \TeX\ deals only with the compact descriptions
of font metric data that appear in \.{TFM} files. Such data is so compact,
however, it is almost impossible for anybody but a computer to read it.

Device drivers also need a compact way to describe mappings from \TeX's idea
of a font to the actual characters a device can produce. They can do this
conveniently when given a packed sequence of bytes called a \.{VF} file.

The purpose of \.{VPtoVF} is to convert from a human-oriented file of text
to computer-oriented files of binary numbers. There's a companion program,
\.{VFtoVP}, which goes the other way.

@<Glob...@>=
@!vpl_file:text;

@ @<Set init...@>=
reset(vpl_file);

@ A \.{VPL} file is like a \.{PL} file with a few extra features, so we
can begin to define it by reviewing the definition of \.{PL} files. The
material in the next few sections is copied from the program \.{PLtoTF}.

A \.{PL} file is a list of entries of the form
$$\.{(PROPERTYNAME VALUE)}$$
where the property name is one of a finite set of names understood by
this program, and the value may itself in turn be a property list.
The idea is best understood by looking at an example, so let's consider
a fragment of the \.{PL} file for a hypothetical font.
$$\vbox{\halign{\.{#}\hfil\cr
(FAMILY NOVA)\cr
(FACE F MIE)\cr
(CODINGSCHEME ASCII)\cr
(DESIGNSIZE D 10)\cr
(DESIGNUNITS D 18)\cr
(COMMENT A COMMENT IS IGNORED)\cr
(COMMENT (EXCEPT THIS ONE ISN'T))\cr
(COMMENT (ACTUALLY IT IS, EVEN THOUGH\cr
\qquad\qquad IT SAYS IT ISN'T))\cr
(FONTDIMEN\cr
\qquad   (SLANT R -.25)\cr
\qquad   (SPACE D 6)\cr
\qquad   (SHRINK D 2)\cr
\qquad   (STRETCH D 3)\cr
\qquad   (XHEIGHT R 10.55)\cr
\qquad   (QUAD D 18)\cr
\qquad   )\cr
(LIGTABLE\cr
\qquad   (LABEL C f)\cr
\qquad   (LIG C f O 200)\cr
\qquad   (SKIP D 1)\cr
\qquad   (LABEL O 200)\cr
\qquad   (LIG C i O 201)\cr
\qquad   (KRN O 51 R 1.5)\cr
\qquad   (/LIG C ? C f)\cr
\qquad   (STOP)\cr
\qquad   )\cr
(CHARACTER C f\cr
\qquad   (CHARWD D 6)\cr
\qquad   (CHARHT R 13.5)\cr
\qquad   (CHARIC R 1.5)\cr
\qquad   )\cr}}$$
This example says that the font whose metric information is being described
belongs to the hypothetical
\.{NOVA} family; its face code is medium italic extended;
and the characters appear in ASCII code positions. The design size is 10 points,
and all other sizes in this \.{PL} file are given in units such that 18 units
equals the design size. The font is slanted with a slope of $-.25$ (hence the
letters actually slant backward---perhaps that is why the family name is
\.{NOVA}). The normal space between words is 6 units (i.e., one third of
the 18-unit design size), with glue that shrinks by 2 units or stretches by 3.
The letters for which accents don't need to be raised or lowered are 10.55
units high, and one em equals 18 units.

The example ligature table is a bit trickier. It specifies that the
letter \.f followed by another \.f is changed to code @'200, while
code @'200 followed by \.i is changed to @'201; presumably codes @'200
and @'201 represent the ligatures `ff' and `ffi'.  Moreover, in both cases
\.f and @'200, if the following character is the code @'51 (which is a
right parenthesis), an additional 1.5 units of space should be inserted
before the @'51.  (The `\.{SKIP}~\.D~\.1' skips over one \.{LIG} or
\.{KRN} command, which in this case is the second \.{LIG}; in this way
two different ligature/kern programs can come together.)
Finally, if either \.f or @'200 is followed by a question mark,
the question mark is replaced by \.f and the ligature program is
started over. (Thus, the character pair `\.{f?}' would actually become
the ligature `ff', and `\.{ff?}' or `\.{f?f}' would become `fff'. To
avoid this restart procedure, the \.{/LIG} command could be replaced
by \.{/LIG>}; then `\.{f?}' would become `f\kern0ptf' and `\.{f?f}'
would become `f\kern0ptff'.)

Character \.f itself is 6 units wide and 13.5 units tall, in this example.
Its depth is zero (since \.{CHARDP} is not given), and its italic correction
is 1.5 units.

@ The example above illustrates most of the features found in \.{PL} files.
Note that some property names, like \.{FAMILY} or \.{COMMENT}, take a
string as their value; this string continues until the first unmatched
right parenthesis. But most property names, like \.{DESIGNSIZE} and \.{SLANT}
and \.{LABEL}, take a number as their value. This number can be expressed in
a variety of ways, indicated by a prefixed code; \.D stands for decimal,
\.H for hexadecimal, \.O for octal, \.R for real, \.C for character, and
\.F for ``face.''  Other property names, like \.{LIG}, take two numbers as
their value.  And still other names, like \.{FONTDIMEN} and \.{LIGTABLE} and
\.{CHARACTER}, have more complicated values that involve property lists.

A property name is supposed to be used only in an appropriate property
list.  For example, \.{CHARWD} shouldn't occur on the outer level or
within \.{FONTDIMEN}.

The individual property-and-value pairs in a property list can appear in
any order. For instance, `\.{SHRINK}' precedes `\.{STRETCH}' in the example
above, although the \.{TFM} file always puts the stretch parameter first.
One could even give the information about characters like `\.f' before
specifying the number of units in the design size, or before specifying the
ligature and kerning table. However, the \.{LIGTABLE} itself is an exception
to this rule; the individual elements of the \.{LIGTABLE} property list
can be reordered only to a certain extent without changing the meaning
of that table.

If property-and-value pairs are omitted, a default value is used. For example,
we have already noted that the default for \.{CHARDP} is zero. The default
for {\sl every\/} numeric value is, in fact, zero, unless otherwise stated
below.

If the same property name is used more than once, \.{VPtoVF} will not notice
the discrepancy; it simply uses the final value given. Once again, however, the
\.{LIGTABLE} is an exception to this rule; \.{VPtoVF} will complain if there
is more than one label for some character. And of course many of the
entries in the \.{LIGTABLE} property list have the same property name.

@ A \.{VPL} file also includes information about how to create each character,
by typesetting characters from other fonts and/or by drawing lines, etc.
Such information is the value of the `\.{MAP}' property, which can be
illustrated as follows:
$$\vbox{\halign{\.{#}\hfil\cr
(MAPFONT D 0 (FONTNAME Times-Roman))\cr
(MAPFONT D 1 (FONTNAME Symbol))\cr
(MAPFONT D 2 (FONTNAME cmr10)(FONTAT D 20))\cr
(CHARACTER O 0 (MAP (SELECTFONT D 1)(SETCHAR C G)))\cr
(CHARACTER O 76 (MAP (SETCHAR O 277)))\cr
(CHARACTER D 197 (MAP\cr
\qquad(PUSH)(SETCHAR C A)(POP)\cr
\qquad(MOVEUP R 0.937)(MOVERIGHT R 1.5)(SETCHAR O 312)))\cr
(CHARACTER O 200 (MAP (MOVEDOWN R 2.1)(SETRULE R 1 R 8)))\cr
(CHARACTER O 201 (MAP\cr
\qquad (SPECIAL ps: /SaveGray currentgray def .5 setgray)\cr
\qquad (SELECTFONT D 2)(SETCHAR C A)\cr
\qquad (SPECIAL ps: SaveGray setgray)))\cr
}}$$
(These specifications appear in addition to the conventional \.{PL}
information. The \.{MAP} attribute can be mixed in with other attributes
like \.{CHARWD} or it can be given separately.)

In this example, the virtual font is composed of characters that can be
fabricated from three actual fonts, `\.{Times-Roman}',
`\.{Symbol}', and `\.{cmr10} \.{at} \.{20\\u}' (where \.{\\u}
is the unit size in this \.{VPL} file). Character |@'0| is typeset as
a `G' from the symbol font. Character |@'76| is typeset as character |@'277|
from the ordinary Times font. (If no other font is selected, font
number~0 is the default. If no \.{MAP} attribute is given, the default map
is a character of the same number in the default font.)

Character 197 (decimal) is more interesting: First an A is typeset (in the
default font Times), and this is enclosed by \.{PUSH} and \.{POP} so that
the original position is restored. Then the accent character |@'312| is
typeset, after moving up .937 units and right 1.5 units.

To typeset character |@'200| in this virtual font, we move down 2.1 units,
then typeset a rule that is 1 unit high and 8 units wide.

Finally, to typeset character |@'201|, we do something that requires a
special ability to interpret PostScript commands; this example
sets the PostScript ``color'' to 50\char`\%\ gray and typesets an `A'
from \.{cmr10} \.{at} \.{20\\u} in that color.

In general, the \.{MAP} attribute of a virtual character can be any sequence
of typesetting commands that might appear in a page of a \.{DVI} file.
A single character might map into an entire page.

@ But instead of relying on a hypothetical example, let's consider a complete
grammar for \.{VPL} files, beginning with the (unchanged) grammatical rules
for \.{PL} files. At the outer level, the following property names
are valid in any \.{PL} file:

\yskip\hang\.{CHECKSUM} (four-byte value). The value, which should be a
nonnegative integer less than $2^{32}$, is used to identify a particular
version of a font; it should match the check sum value stored with the font
itself. An explicit check sum of zero is used to bypass
check sum testing. If no checksum is specified in the \.{VPL} file,
\.{VPtoVF} will compute the checksum that \MF\ would compute from the
same data.

\yskip\hang\.{DESIGNSIZE} (numeric value, default is 10). The value, which
should be a real number in the range |1.0<=x<2048|, represents the default
amount by which all quantities will be scaled if the font is not loaded
with an `\.{at}' specification. For example, if one says
`\.{\\font\\A=cmr10 at 15pt}' in \TeX\ language, the design size in the \.{TFM}
file is ignored and effectively replaced by 15 points; but if one simply
says `\.{\\font\\A=cmr10}' the stated design size is used. This quantity is
always in units of printer's points.

\yskip\hang\.{DESIGNUNITS} (numeric value, default is 1). The value
should be a positive real number; it says how many units equals the design
size (or the eventual `\.{at}' size, if the font is being scaled). For
example, suppose you have a font that has been digitized with 600 pixels per
em, and the design size is one em; then you could say `\.{(DESIGNUNITS R 600)}'
if you wanted to give all of your measurements in units of pixels.

\yskip\hang\.{CODINGSCHEME} (string value, default is `\.{UNSPECIFIED}').
The string should not contain parentheses, and its length must be less than 40.
It identifies the correspondence between the numeric codes and font characters.
(\TeX\ ignores this information, but other software programs make use of it.)

\yskip\hang\.{FAMILY} (string value, default is `\.{UNSPECIFIED}').
The string should not contain parentheses, and its length must be less than 20.
It identifies the name of the family to which this font belongs, e.g.,
`\.{HELVETICA}'.  (\TeX\ ignores this information; but it is needed, for
example, when converting \.{DVI} files to \.{PRESS} files for Xerox
equipment.)

\yskip\hang\.{FACE} (one-byte value). This number, which must lie between
0 and 255 inclusive, is a subsidiary ident\-ifi\-ca\-tion of the font within its
family. For example, bold italic condensed fonts might have the same family name
as light roman extended fonts, differing only in their face byte.  (\TeX\
ignores this information; but it is needed, for example, when converting
\.{DVI} files to \.{PRESS} files for Xerox equipment.)

\yskip\hang\.{SEVENBITSAFEFLAG} (string value, default is `\.{FALSE}'). The
value should start with either `\.T' (true) or `\.F' (false). If true, character
codes less than 128 cannot lead to codes of 128 or more via ligatures or
charlists or extensible characters. (\TeX82 ignores this flag, but older
versions of \TeX\ would only accept \.{TFM} files that were seven-bit safe.)
\.{VPtoVF} computes the correct value of this flag and gives an error message
only if a claimed ``true'' value is incorrect.

\yskip\hang\.{HEADER} (a one-byte value followed by a four-byte value).
The one-byte value should be between 18 and a maximum limit that can be
raised or lowered depending on the compile-time setting of |max_header_bytes|.
The four-byte value goes into the header word whose index is the one-byte
value; for example, to set |header[18]:=1|, one may write
`\.{(HEADER D 18 O 1)}'. This notation is used for header information that
is presently unnamed. (\TeX\ ignores it.)

\yskip\hang\.{FONTDIMEN} (property list value). See below for the names
allowed in this property list.

\yskip\hang\.{LIGTABLE} (property list value). See below for the rules
about this special kind of property list.

\yskip\hang\.{BOUNDARYCHAR} (one-byte value). If this character appears in
a \.{LIGTABLE} command, it matches ``end of word'' as well as itself.
If no boundary character is given and no \.{LABEL} \.{BOUNDARYCHAR} occurs
within \.{LIGTABLE}, word boundaries will not affect ligatures or kerning.

\yskip\hang\.{CHARACTER}. The value is a one-byte integer followed by
a property list. The integer represents the number of a character that is
present in the font; the property list of a character is defined below.
The default is an empty property list.

@ Numeric property list values can be given in various forms identified by
a prefixed letter.

\yskip\hang\.C denotes an ASCII character, which should be a standard visible
character that is not a parenthesis. The numeric value will therefore be
between @'41 and @'176 but not @'50 or @'51.

\yskip\hang\.D denotes an unsigned decimal integer, which must be
less than $2^{32}$, i.e., at most `\.{D 4294967295}'.

\yskip\hang\.F denotes a three-letter Xerox face code; the admissible codes
are \.{MRR}, \.{MIR}, \.{BRR}, \.{BIR}, \.{LRR}, \.{LIR}, \.{MRC}, \.{MIC},
\.{BRC}, \.{BIC}, \.{LRC}, \.{LIC}, \.{MRE}, \.{MIE}, \.{BRE}, \.{BIE},
\.{LRE}, and \.{LIE}, denoting the integers 0 to 17, respectively.

\yskip\hang\.O denotes an unsigned octal integer, which must be less than
$2^{32}$, i.e., at most `\.{O 37777777777}'.

\yskip\hang\.H denotes an unsigned hexadecimal integer, which must be less than
$2^{32}$, i.e., at most `\.{H FFFFFFFF}'.

\yskip\hang\.R denotes a real number in decimal notation, optionally preceded
by a `\.+' or `\.-' sign, and optionally including a decimal point. The
absolute value must be less than 2048.

@ The property names allowed in a \.{FONTDIMEN} property list correspond to
various \TeX\ parameters, each of which has a (real) numeric value. All
of the parameters except \.{SLANT} are in design units. The admissible
names are \.{SLANT}, \.{SPACE}, \.{STRETCH}, \.{SHRINK}, \.{XHEIGHT},
\.{QUAD}, \.{EXTRASPACE}, \.{NUM1}, \.{NUM2}, \.{NUM3}, \.{DENOM1},
\.{DENOM2}, \.{SUP1}, \.{SUP2}, \.{SUP3}, \.{SUB1}, \.{SUB2}, \.{SUPDROP},
\.{SUBDROP}, \.{DELIM1}, \.{DELIM2}, and \.{AXISHEIGHT}, for parameters
1~to~22. The alternate names \.{DEFAULTRULETHICKNESS},
\.{BIGOPSPACING1}, \.{BIGOPSPACING2}, \.{BIGOPSPACING3},
\.{BIGOPSPACING4}, and \.{BIGOPSPACING5}, may also be used for parameters
8 to 13.

The notation `\.{PARAMETER} $n$' provides another way to specify the
$n$th parameter; for example, `\.{(PARAMETER} \.{D 1 R -.25)}' is another way
to specify that the \.{SLANT} is $-0.25$. The value of $n$ must be positive
and less than |max_param_words|.

@ The elements of a \.{CHARACTER} property list can be of six different types.

\yskip\hang\.{CHARWD} (real value) denotes the character's width in
design units.

\yskip\hang\.{CHARHT} (real value) denotes the character's height in
design units.

\yskip\hang\.{CHARDP} (real value) denotes the character's depth in
design units.

\yskip\hang\.{CHARIC} (real value) denotes the character's italic correction in
design units.

\yskip\hang\.{NEXTLARGER} (one-byte value), specifies the character that
follows the present one in a ``charlist.'' The value must be the number of a
character in the font, and there must be no infinite cycles of supposedly
larger and larger characters.

\yskip\hang\.{VARCHAR} (property list value), specifies an extensible character.
This option and \.{NEXTLARGER} are mutually exclusive; i.e., they cannot
both be used within the same \.{CHARACTER} list.

\yskip\noindent
The elements of a \.{VARCHAR} property list are either \.{TOP}, \.{MID},
\.{BOT} or \.{REP}; the values are integers, which must be zero or the number
of a character in the font. A zero value for \.{TOP}, \.{MID}, or \.{BOT} means
that the corresponding piece of the extensible character is absent. A nonzero
value, or a \.{REP} value of zero, denotes the character code used to make
up the top, middle, bottom, or replicated piece of an extensible character.

@ A \.{LIGTABLE} property list contains elements of four kinds, specifying a
program in a simple command language that \TeX\ uses for ligatures and kerns.
If several \.{LIGTABLE} lists appear, they are effectively concatenated into
a single list.

\yskip\hang\.{LABEL} (one-byte value) means that the program for the
stated character value starts here. The integer must be the number of a
character in the font; its \.{CHARACTER} property list must not have a
\.{NEXTLARGER} or \.{VARCHAR} field. At least one \.{LIG} or \.{KRN} step
must follow.

\yskip\hang\.{LABEL} \.{BOUNDARYCHAR} means that the program for
beginning-of-word ligatures starts here.

\yskip\hang\.{LIG} (two one-byte values). The instruction `\.{(LIG} $c$ $r$\.)'
means, ``If the next character is $c$, then insert character~$r$ and
possibly delete the current character and/or~$c$;
otherwise go on to the next instruction.''
Characters $r$ and $c$ must be present in the font. \.{LIG} may be immediately
preceded or followed by a slash, and then immediately followed by \.>
characters not exceeding the number of slashes. Thus there are eight
possible forms:
$$\hbox to .8\hsize{\.{LIG}\hfil\.{/LIG}\hfil\.{/LIG>}\hfil
\.{LIG/}\hfil\.{LIG/>}\hfil\.{/LIG/}\hfil\.{/LIG/>}\hfil\.{/LIG/>>}}$$
The slashes specify retention of the left or right original character; the
\.> signs specify passing over the result without further ligature processing.

\yskip\hang\.{KRN} (a one-byte value and a real value). The instruction
`\.{(KRN} $c$ $r$\.)' means, ``If the next character is $c$, then insert
a blank space of width $r$ between the current character character and $c$;
otherwise go on to the next intruction.'' The value of $r$, which is in
design units, is often negative. Character code $c$ must exist
in the font.

\yskip\hang\.{STOP} (no value). This instruction ends a ligature/kern program.
It must follow either a \.{LIG} or \.{KRN} instruction, not a \.{LABEL}
or \.{STOP} or \.{SKIP}.

\yskip\hang\.{SKIP} (value in the range |0..127|). This instruction specifies
continuation of a ligature/kern program after the specified number of \.{LIG}
or \.{KRN} steps has been skipped over. The number of subsequent \.{LIG} and
\.{KRN} instructions must therefore exceed this specified amount.

@ In addition to all these possibilities, the property name \.{COMMENT} is
allowed in any property list. Such comments are ignored.

@ So that is what \.{PL} files hold. In a \.{VPL} file additional
properties are recognized; two of these are valid on the outermost level:

\yskip\hang\.{VTITLE} (string value, default is empty). The value will be
reproduced at the beginning of the \.{VF} file (and printed on the terminal
by \.{VFtoVP} when it examines that file).

\yskip\hang\.{MAPFONT}. The value is a nonnegative integer followed by
a property list. The integer represents an identifying number for fonts
used in \.{MAP} attributes. The property list, which identifies the font and
relative size, is defined below.

\yskip\noindent
And one additional ``virtual property'' is valid within a \.{CHARACTER}:

\yskip\hang\.{MAP}. The value is a property list consisting of typesetting
commands. Default is the single command \.{SETCHAR}~$c$, where $c$ is
the current character number.

@ The elements of a \.{MAPFONT} property list can be of the following types.

\yskip\hang\.{FONTNAME} (string value, default is \.{NULL}).
This is the font's identifying name.

\yskip\hang\.{FONTAREA} (string value, default is empty). If the font appears
in a nonstandard directory, according to local conventions, the directory
name is given here. (This is system dependent, just as in \.{DVI} files.)

\yskip\hang\.{FONTCHECKSUM} (four-byte value, default is zero). This value,
which should be a nonnegative integer less than $2^{32}$, can be used to
check that the font being referred to matches the intended font. If nonzero,
it should equal the \.{CHECKSUM} parameter in that font.

\yskip\hang\.{FONTAT} (numeric value, default is the \.{DESIGNUNITS} of the
present virtual font). This value is relative to the design units of
the present virtual font, hence it will be scaled when the virtual
font is magnified or reduced.  It represents the value that will
effectively replace the design size of the font being referred to,
so that all characters will be scaled appropriately.

\yskip\hang\.{FONTDSIZE} (numeric value, default is 10). This value is
absolute, in units of printer's points. It should equal the \.{DESIGNSIZE}
parameter in the font being referred to.

\yskip\noindent
If any of the
string values contain parentheses, the parentheses must be balanced. Leading
blanks are removed from the strings, but trailing blanks are not.

@ Finally, the elements of a \.{MAP} property list are an ordered sequence
of typesetting commands chosen from among the following:

\yskip\hang\.{SELECTFONT} (four-byte integer value). The value must be the
number of a previously defined \.{MAPFONT}. This font (or more precisely, the
final font that is mapped to that code number, if two \.{MAPFONT} properties
happen to specify the same code) will be used in subsequent \.{SETCHAR}
instructions until overridden by another \.{SELECTFONT}. The first-specified
\.{MAPFONT} is implicitly selected before the first \.{SELECTFONT} in every
character's map.

\yskip\hang\.{SETCHAR} (one-byte integer value). There must be a character of
this number in the currently selected font. (\.{VPtoVF} doesn't check that
the character is valid, but \.{VFtoVP} does.) That character is typeset at the
current position, and the typesetter moves right by the \.{CHARWD} in
that character's \.{TFM} file.

\yskip\hang\.{SETRULE} (two real values). The first value specifies height,
the second specifies width, in design units. If both height and width are
positive, a rule is typeset at the current position. Then the typesetter
moves right, by the specified width.

\yskip\hang\.{MOVERIGHT}, \.{MOVELEFT}, \.{MOVEUP}, \.{MOVEDOWN} (real
value). The typesetter moves its current position
by the number of design units specified.

\yskip\hang\.{PUSH}. The current typesetter position is remembered, to
be restored on a subsequent \.{POP}.

\yskip\hang\.{POP}. The current typesetter position is reset to where it
was on the most recent unmatched \.{PUSH}. The \.{PUSH} and \.{POP}
commands in any \.{MAP} must be properly nested like balanced parentheses.

\yskip\hang\.{SPECIAL} (string value). The subsequent characters, starting
with the first nonblank and ending just before the first `\.)' that has no
matching `\.(', are interpreted according to local conventions with the
same system-dependent meaning as a `special' (\\{xxx}) command
in a \.{DVI} file.

\yskip\hang\.{SPECIALHEX} (hexadecimal string value). The subsequent
nonblank characters before the next `\.)' must consist entirely of
hexadecimal digits, and they must contain an even number of such digits.
Each pair of hex digits specifies a byte, and this string of bytes is
treated just as the value of a \.{SPECIAL}. (This convention permits
arbitrary byte strings to be represented in an ordinary text file.)

@ Virtual font mapping is a recursive process, like macro expansion.
Thus, a \.{MAPFONT} might
specify another virtual font, whose characters are themselves mapped to
other fonts. As an example of this possibility, consider the
following curious file called \.{recurse.vpl}, which defines a
virtual font that is self-contained and self-referential:
$$\vbox{\halign{\.{#}\cr
(VTITLE Example of recursion)\cr
(MAPFONT D 0 (FONTNAME recurse)(FONTAT D 2))\cr
(CHARACTER C A (CHARWD D 1)(CHARHT D 1)(MAP (SETRULE D 1 D 1)))\cr
(CHARACTER C B (CHARWD D 2)(CHARHT D 2)(MAP (SETCHAR C A)))\cr
(CHARACTER C C (CHARWD D 4)(CHARHT D 4)(MAP (SETCHAR C B)))\cr
}}$$
The design size is 10 points (the default), hence the character \.A
in font \.{recurse} is a $10\times10$ point black square. Character \.B
is typeset as character \.A in \.{recurse} {scaled} {2000}, hence it
is a $20\times20$ point black square. And character \.C is typeset as
character \.{B} in \.{recurse} {scaled} {2000}, hence its size is
$40\times40$.

Users are responsible for making sure that infinite recursion doesn't happen.

@ So that is what \.{VPL} files hold. From these rules,
you can guess (correctly) that \.{VPtoVF} operates in four main stages.
First it assigns the default values to all properties; then it scans
through the \.{VPL} file, changing property values as new ones are seen; then
it checks the information and corrects any problems; and finally it outputs
the \.{VF} and \.{TFM} files.

@ The next question is, ``What are \.{VF} and
\.{TFM} files?'' A complete answer to that question appears in the
documentation of the companion programs, \.{VFtoVP} and
\.{TFtoPL}, so the details will not
be repeated here. Suffice it to say that a \.{VF} or
\.{TFM} file stores all of the
relevant font information in a sequence of 8-bit bytes. The number of
bytes is always a multiple of 4, so we could regard the files
as sequences of 32-bit words; but \TeX\ uses the byte interpretation,
and so does \.{VPtoVF}. Note that the bytes are considered to be unsigned
numbers.

@<Glob...@>=
@!vf_file:packed file of 0..255;
@!tfm_file:packed file of 0..255;

@ On some systems you may have to do something special to write a
packed file of bytes. For example, the following code didn't work
when it was first tried at Stanford, because packed files have to be
opened with a special switch setting on the \PASCAL\ that was used.
@^system dependencies@>

@<Set init...@>=
rewrite(vf_file); rewrite(tfm_file);

@* Basic input routines.
For the purposes of this program, a |byte| is an unsigned eight-bit quantity,
and an |ASCII_code| is an integer between @'40 and @'177. Such ASCII codes
correspond to one-character constants like \.{"A"} in \.{WEB} language.

@<Types...@>=
@!byte=0..255; {unsigned eight-bit quantity}
@!ASCII_code=@'40..@'177; {standard ASCII code numbers}

@ One of the things \.{VPtoVF} has to do is convert characters of strings
to ASCII form, since that is the code used for the family name and the
coding scheme in a \.{TFM} file. An array |xord| is used to do the
conversion from |char|; the method below should work with little or no change
on most \PASCAL\ systems.
@^system dependencies@>

@d first_ord=0 {ordinal number of the smallest element of |char|}
@d last_ord=127 {ordinal number of the largest element of |char|}

@<Global...@>=
@!xord:array[char] of ASCII_code; {conversion table}

@ @<Local variables for init...@>=
@!k:integer; {all-purpose initialization index}

@ Characters that should not appear in \.{VPL} files (except in comments)
are mapped into @'177.

@d invalid_code=@'177 {code deserving an error message}

@<Set init...@>=
for k:=first_ord to last_ord do xord[chr(k)]:=invalid_code;
xord[' ']:=" "; xord['!']:="!"; xord['"']:=""""; xord['#']:="#";
xord['$']:="$"; xord['%']:="%"; xord['&']:="&"; xord['''']:="'";
xord['(']:="("; xord[')']:=")"; xord['*']:="*"; xord['+']:="+"; xord[',']:=",";
xord['-']:="-"; xord['.']:="."; xord['/']:="/"; xord['0']:="0"; xord['1']:="1";
xord['2']:="2"; xord['3']:="3"; xord['4']:="4"; xord['5']:="5"; xord['6']:="6";
xord['7']:="7"; xord['8']:="8"; xord['9']:="9"; xord[':']:=":"; xord[';']:=";";
xord['<']:="<"; xord['=']:="="; xord['>']:=">"; xord['?']:="?";
xord['@@']:="@@"; xord['A']:="A"; xord['B']:="B"; xord['C']:="C";
xord['D']:="D"; xord['E']:="E"; xord['F']:="F"; xord['G']:="G"; xord['H']:="H";
xord['I']:="I"; xord['J']:="J"; xord['K']:="K"; xord['L']:="L"; xord['M']:="M";
xord['N']:="N"; xord['O']:="O"; xord['P']:="P"; xord['Q']:="Q"; xord['R']:="R";
xord['S']:="S"; xord['T']:="T"; xord['U']:="U"; xord['V']:="V"; xord['W']:="W";
xord['X']:="X"; xord['Y']:="Y"; xord['Z']:="Z"; xord['[']:="["; xord['\']:="\";
xord[']']:="]"; xord['^']:="^"; xord['_']:="_"; xord['`']:="`"; xord['a']:="a";
xord['b']:="b"; xord['c']:="c"; xord['d']:="d"; xord['e']:="e"; xord['f']:="f";
xord['g']:="g"; xord['h']:="h"; xord['i']:="i"; xord['j']:="j"; xord['k']:="k";
xord['l']:="l"; xord['m']:="m"; xord['n']:="n"; xord['o']:="o"; xord['p']:="p";
xord['q']:="q"; xord['r']:="r"; xord['s']:="s"; xord['t']:="t"; xord['u']:="u";
xord['v']:="v"; xord['w']:="w"; xord['x']:="x"; xord['y']:="y"; xord['z']:="z";
xord['{']:="{"; xord['|']:="|"; xord['}']:="}"; xord['~']:="~";

@ In order to help catch errors of badly nested parentheses, \.{VPtoVF}
assumes that the user will begin each line with a number of blank spaces equal
to some constant times the number of open parentheses at the beginning of
that line. However, the program doesn't know in advance what the constant
is, nor does it want to print an error message on every line for a user
who has followed no consistent pattern of indentation.

Therefore the following strategy is adopted: If the user has been consistent
with indentation for ten or more lines, an indentation error will be
reported. The constant of indentation is reset on every line that should
have nonzero indentation.

@<Glob...@>=
@!line:integer; {the number of the current line}
@!good_indent:integer; {the number of lines since the last bad indentation}
@!indent: integer; {the number of spaces per open parenthesis, zero if unknown}
@!level: integer; {the current number of open parentheses}

@ @<Set init...@>=
line:=0; good_indent:=0; indent:=0; level:=0;

@ The input need not really be broken into lines of any maximum length, and
we could read it character by character without any buffering. But we shall
place it into a small buffer so that offending lines can be displayed in error
messages.

@<Glob...@>=
@!left_ln,@!right_ln:boolean; {are the left and right ends of the buffer
  at end-of-line marks?}
@!limit:0..buf_size; {position of the last character present in the buffer}
@!loc:0..buf_size; {position of the last character read in the buffer}
@!buffer:array[1..buf_size] of char;
@!input_has_ended:boolean; {there is no more input to read}

@ @<Set init...@>=
limit:=0; loc:=0; left_ln:=true; right_ln:=true; input_has_ended:=false;

@ Just before each  \.{CHARACTER} property list is evaluated, the character
code is printed in octal notation. Up to eight such codes appear on a line;
so we have a variable to keep track of how many are currently there.

@<Glob...@>=
@!chars_on_line:0..8; {the number of characters printed on the current line}

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

@ The following routine prints an error message and an indication of
where the error was detected. The error message should not include any
final punctuation, since this procedure supplies its own.

@d err_print(#)==begin if chars_on_line>0 then print_ln(' ');
  print(#); show_error_context;
  end

@p procedure show_error_context; {prints the current scanner location}
var k:0..buf_size; {an index into |buffer|}
begin print_ln(' (line ',line:1,').');
if not left_ln then print('...');
for k:=1 to loc do print(buffer[k]); {print the characters already scanned}
print_ln(' ');
if not left_ln then print('   ');
for k:=1 to loc do print(' '); {space out the second line}
for k:=loc+1 to limit do print(buffer[k]); {print the characters yet unseen}
if right_ln then print_ln(' ')@+else print_ln('...');
chars_on_line:=0;
end;

@ Here is a procedure that does the right thing when we are done
reading the present contents of the buffer. It keeps |buffer[buf_size]|
empty, in order to avoid range errors on certain \PASCAL\ compilers.

An infinite sequence of right parentheses is placed at the end of the
file, so that the program is sure to get out of whatever level of nesting
it is in.

On some systems it is desirable to modify this code so that tab marks
in the buffer are replaced by blank spaces. (Simply setting
|xord[chr(@'11)]:=" "| would not work; for example, two-line
error messages would not come out properly aligned.)
@^system dependencies@>

@p procedure fill_buffer;
begin left_ln:=right_ln; limit:=0; loc:=0;
if left_ln then
  begin if line>0 then read_ln(vpl_file);
  incr(line);
  end;
if eof(vpl_file) then
  begin limit:=1; buffer[1]:=')'; right_ln:=false; input_has_ended:=true;
  end
else  begin while (limit<buf_size-2)and(not eoln(vpl_file)) do
    begin incr(limit); read(vpl_file,buffer[limit]);
    end;
  buffer[limit+1]:=' '; right_ln:=eoln(vpl_file);
  if right_ln then begin incr(limit); buffer[limit+1]:=' ';
    end;
  if left_ln then @<Set |loc| to the number of leading blanks in
    the buffer, and check the indentation@>;
  end;
end;

@ The interesting part about |fill_buffer| is the part that learns what
indentation conventions the user is following, if any.

@d bad_indent(#)==begin if good_indent>=10 then err_print(#);
  good_indent:=0; indent:=0;
  end

@<Set |loc|...@>=
begin while (loc<limit)and(buffer[loc+1]=' ') do incr(loc);
if loc<limit then
  begin if level=0 then
    if loc=0 then incr(good_indent)
    else bad_indent('Warning: Indented line occurred at level zero')
@.Warning: Indented line...@>
  else if indent=0 then
    if loc mod level=0 then
      begin indent:=loc div level; good_indent:=1;
      end
    else good_indent:=0
  else if indent*level=loc then incr(good_indent)
  else bad_indent('Warning: Inconsistent indentation; ',
@.Warning: Inconsistent indentation...@>
    'you are at parenthesis level ',level:1);
  end;
end

@* Basic scanning routines.
The global variable |cur_char| holds the ASCII code corresponding to the
character most recently read from the input buffer, or to a character that
has been substituted for the real one.

@<Global...@>=
@!cur_char:ASCII_code; {we have just read this}

@ Here is a procedure that sets |cur_char| to an ASCII code for the
next character of input, if that character is a letter or digit or slash
or \.>. Otherwise
it sets |cur_char:=" "|, and the input system will be poised to reread the
character that was rejected, whether or not it was a space.
Lower case letters are converted to upper case.

@p procedure get_keyword_char;
begin while (loc=limit)and(not right_ln) do fill_buffer;
if loc=limit then cur_char:=" " {end-of-line counts as a delimiter}
else  begin cur_char:=xord[buffer[loc+1]];
  if cur_char>="a" then cur_char:=cur_char-@'40;
  if ((cur_char>="0")and(cur_char<="9")) then incr(loc)
  else if ((cur_char>="A")and(cur_char<="Z")) then incr(loc)
  else if cur_char="/" then incr(loc)
  else if cur_char=">" then incr(loc)
  else cur_char:=" ";
  end;
end;

@ The following procedure sets |cur_char| to the next character code,
and converts lower case to upper case. If the character is a left or
right parenthesis, it will not be ``digested''; the character will
be read again and again, until the calling routine does something
like `|incr(loc)|' to get past it. Such special treatment of parentheses
insures that the structural information they contain won't be lost in
the midst of other error recovery operations.

@d backup==begin if (cur_char>")")or(cur_char<"(") then decr(loc);
  end {undoes the effect of |get_next|}

@p procedure get_next; {sets |cur_char| to next, balks at parentheses}
begin while loc=limit do fill_buffer;
incr(loc); cur_char:=xord[buffer[loc]];
if cur_char>="a" then
  if cur_char<="z" then cur_char:=cur_char-@'40 {uppercasify}
  else  begin if cur_char=invalid_code then
      begin err_print('Illegal character in the file');
@.Illegal character...@>
      cur_char:="?";
      end;
    end
else if (cur_char<=")")and(cur_char>="(") then decr(loc);
end;

@ Here's a procedure that scans a hexadecimal digit or a right parenthesis.

@p function get_hex:byte;
var @!a:integer; {partial result}
begin repeat get_next;
until cur_char<>" ";
a:=cur_char-")";
if a>0 then
  begin a:=cur_char-"0";
  if cur_char>"9" then
    if cur_char<"A" then a:=-1 else a:=cur_char-"A"+10;
  end;
if (a<0)or(a>15) then
  begin err_print('Illegal hexadecimal digit'); get_hex:=0;
@.Illegal hexadecimal digit@>
  end
else get_hex:=a;
end;

@ The next procedure is used to ignore the text of a comment, or to pass over
erroneous material. As such, it has the privilege of passing parentheses.
It stops after the first right parenthesis that drops the level below
the level in force when the procedure was called.

@p procedure skip_to_end_of_item;
var l:integer; {initial value of |level|}
begin l:=level;
while level>=l do
  begin while loc=limit do fill_buffer;
  incr(loc);
  if buffer[loc]=')' then decr(level)
  else if buffer[loc]='(' then incr(level);
  end;
if input_has_ended then err_print('File ended unexpectedly: No closing ")"');
@.File ended unexpectedly...@>
cur_char:=" "; {now the right parenthesis has been read and digested}
end;

@ A similar procedure copies the bytes remaining in an item. The copied bytes
go into an array |vf| that we'll declare later. Leading blanks are ignored.

@d vf_store(#)==
    begin vf[vf_ptr]:=#;
    if vf_ptr=vf_size then err_print('I''m out of memory---increase my vfsize!')
@.I'm out of memory...@>
    else incr(vf_ptr);
    end

@p procedure copy_to_end_of_item;
label 30;
var l:integer; {initial value of |level|}
@!nonblank_found:boolean; {have we seen a nonblank character yet?}
begin l:=level; nonblank_found:=false;
while true do
  begin while loc=limit do fill_buffer;
  if buffer[loc+1]=')' then
    if level=l then goto 30@+else decr(level);
  incr(loc);
  if buffer[loc]='(' then incr(level);
  if buffer[loc]<>' ' then nonblank_found:=true;
  if nonblank_found then
    if xord[buffer[loc]]=invalid_code then
      begin err_print('Illegal character in the file');
@.Illegal character...@>
      vf_store("?");
      end
    else vf_store(xord[buffer[loc]]);
  end;
30:end;

@ Sometimes we merely want to skip past characters in the input until we
reach a left or a right parenthesis. For example, we do this whenever we
have finished scanning a property value and we hope that a right parenthesis
is next (except for possible blank spaces).

@d skip_to_paren==repeat get_next@;@+ until (cur_char="(")or(cur_char=")")
@d skip_error(#)==begin err_print(#); skip_to_paren;
  end {this gets to the right parenthesis if something goes wrong}
@d flush_error(#)==begin err_print(#); skip_to_end_of_item;
  end {this gets past the right parenthesis if something goes wrong}

@ After a property value has been scanned, we want to move just past the
right parenthesis that should come next in the input (except for possible
blank spaces).

@p procedure finish_the_property; {do this when the value has been scanned}
begin while cur_char=" " do get_next;
if cur_char<>")" then err_print('Junk after property value will be ignored');
@.Junk after property value...@>
skip_to_end_of_item;
end;

@* Scanning property names.
We have to figure out the meaning of names that appear in the \.{VPL} file,
by looking them up in a dictionary of known keywords. Keyword number $n$
appears in locations |start[n]| through |start[n+1]-1| of an array called
|dictionary|.

@d max_name_index=100 {upper bound on the number of keywords}
@d max_letters=666 {upper bound on the total length of all keywords}

@<Global...@>=
@!start:array[1..max_name_index] of 0..max_letters;
@!dictionary:array[0..max_letters] of ASCII_code;
@!start_ptr:0..max_name_index; {the first available place in |start|}
@!dict_ptr:0..max_letters; {the first available place in |dictionary|}

@ @<Set init...@>=
start_ptr:=1; start[1]:=0; dict_ptr:=0;

@ When we are looking for a name, we put it into the |cur_name| array.
When we have found it, the corresponding |start| index will go into
the global variable |name_ptr|.

@d longest_name=20 {length of \.{DEFAULTRULETHICKNESS}}

@<Glob...@>=
@!cur_name:array[1..longest_name] of ASCII_code; {a name to look up}
@!name_length:0..longest_name; {its length}
@!name_ptr:0..max_name_index; {its ordinal number in the dictionary}

@ A conventional hash table with linear probing (cf.\ Algorithm 6.4L
in {\sl The Art of Computer Pro\-gram\-ming\/}) is used for the dictionary
operations. If |nhash[h]=0|, the table position is empty, otherwise |nhash[h]|
points into the |start| array.

@d hash_prime=141 {size of the hash table}

@<Glob...@>=
@!nhash:array[0..hash_prime-1] of 0..max_name_index;
@!cur_hash:0..hash_prime-1; {current position in the hash table}

@ @<Local...@>=
@!h:0..hash_prime-1; {runs through the hash table}

@ @<Set init...@>=
for h:=0 to hash_prime-1 do nhash[h]:=0;

@ Since there is no chance of the hash table overflowing, the procedure
is very simple. After |lookup| has done its work, |cur_hash| will point
to the place where the given name was found, or where it should be inserted.

@p procedure lookup; {finds |cur_name| in the dictionary}
var k:0..longest_name; {index into |cur_name|}
@!j:0..max_letters; {index into |dictionary|}
@!not_found:boolean; {clumsy thing necessary to avoid |goto| statement}
begin @<Compute the hash code, |cur_hash|, for |cur_name|@>;
not_found:=true;
while not_found do
  begin if cur_hash=0 then cur_hash:=hash_prime-1@+else decr(cur_hash);
  if nhash[cur_hash]=0 then not_found:=false
  else  begin j:=start[nhash[cur_hash]];
    if start[nhash[cur_hash]+1]=j+name_length then
      begin not_found:=false;
      for k:=1 to name_length do
      if dictionary[j+k-1]<>cur_name[k] then not_found:=true;
      end;
    end;
  end;
name_ptr:=nhash[cur_hash];
end;

@ @<Compute the hash...@>=
cur_hash:=cur_name[1];
for k:=2 to name_length do
  cur_hash:=(cur_hash+cur_hash+cur_name[k]) mod hash_prime

@ The ``meaning'' of the keyword that begins at |start[k]| in the
dictionary is kept in |equiv[k]|. The numeric |equiv| codes are given
symbolic meanings by the following definitions.

@d comment_code=0
@d check_sum_code=1
@d design_size_code=2
@d design_units_code=3
@d coding_scheme_code=4
@d family_code=5
@d face_code=6
@d seven_bit_safe_flag_code=7
@d header_code= 8
@d font_dimen_code=9
@d lig_table_code=10
@d boundary_char_code=11
@d virtual_title_code=12
@d map_font_code=13
@d character_code=14
@d font_name_code=20
@d font_area_code=21
@d font_checksum_code=22
@d font_at_code=23
@d font_dsize_code=24
@d parameter_code=30
@d char_info_code=60
@d width=1
@d height=2
@d depth=3
@d italic=4
@d char_wd_code=char_info_code+width
@d char_ht_code=char_info_code+height
@d char_dp_code=char_info_code+depth
@d char_ic_code=char_info_code+italic
@d next_larger_code=65
@d map_code=66
@d var_char_code=67
@d select_font_code=80
@d set_char_code=81
@d set_rule_code=82
@d move_right_code=83
@d move_down_code=85
@d push_code=87
@d pop_code=88
@d special_code=89
@d special_hex_code=90
@d label_code=100
@d stop_code=101
@d skip_code=102
@d krn_code=103
@d lig_code=104

@<Glo...@>=
@!equiv:array[0..max_name_index] of byte;
@!cur_code:byte; {equivalent most recently found in |equiv|}

@ We have to get the keywords into the hash table and into the dictionary in
the first place (sigh). The procedure that does this has the desired
|equiv| code as a parameter. In order to facilitate \.{WEB} macro writing
for the initialization, the keyword being initialized is placed into the
last positions of |cur_name|, instead of the first positions.

@p procedure enter_name(v:byte); {|cur_name| goes into the dictionary}
var k:0..longest_name;
begin for k:=1 to name_length do
  cur_name[k]:=cur_name[k+longest_name-name_length];
{now the name has been shifted into the correct position}
lookup; {this sets |cur_hash| to the proper insertion place}
nhash[cur_hash]:=start_ptr; equiv[start_ptr]:=v;
for k:=1 to name_length do
  begin dictionary[dict_ptr]:=cur_name[k]; incr(dict_ptr);
  end;
incr(start_ptr); start[start_ptr]:=dict_ptr;
end;

@ Here are the macros to load a name of up to 20 letters into the
dictionary. For example, the macro |load5| is used for five-letter keywords.

@d tail(#)==enter_name(#)
@d t20(#)==cur_name[20]:=#;tail
@d t19(#)==cur_name[19]:=#;t20
@d t18(#)==cur_name[18]:=#;t19
@d t17(#)==cur_name[17]:=#;t18
@d t16(#)==cur_name[16]:=#;t17
@d t15(#)==cur_name[15]:=#;t16
@d t14(#)==cur_name[14]:=#;t15
@d t13(#)==cur_name[13]:=#;t14
@d t12(#)==cur_name[12]:=#;t13
@d t11(#)==cur_name[11]:=#;t12
@d t10(#)==cur_name[10]:=#;t11
@d t9(#)==cur_name[9]:=#;t10
@d t8(#)==cur_name[8]:=#;t9
@d t7(#)==cur_name[7]:=#;t8
@d t6(#)==cur_name[6]:=#;t7
@d t5(#)==cur_name[5]:=#;t6
@d t4(#)==cur_name[4]:=#;t5
@d t3(#)==cur_name[3]:=#;t4
@d t2(#)==cur_name[2]:=#;t3
@d t1(#)==cur_name[1]:=#;t2
@d load3==name_length:=3;t18
@d load4==name_length:=4;t17
@d load5==name_length:=5;t16
@d load6==name_length:=6;t15
@d load7==name_length:=7;t14
@d load8==name_length:=8;t13
@d load9==name_length:=9;t12
@d load10==name_length:=10;t11
@d load11==name_length:=11;t10
@d load12==name_length:=12;t9
@d load13==name_length:=13;t8
@d load14==name_length:=14;t7
@d load15==name_length:=15;t6
@d load16==name_length:=16;t5
@d load17==name_length:=17;t4
@d load18==name_length:=18;t3
@d load19==name_length:=19;t2
@d load20==name_length:=20;t1

@ (Thank goodness for keyboard macros in the text editor used to create this
\.{WEB} file.)

@<Enter all the \.{PL} names and their equivalents,
 except the parameter names@>=
equiv[0]:=comment_code; {this is used after unknown keywords}
load8("C")("H")("E")("C")("K")("S")("U")("M")(check_sum_code);@/
load10("D")("E")("S")("I")("G")("N")("S")("I")("Z")("E")(design_size_code);@/
load11("D")("E")("S")("I")("G")("N")
  ("U")("N")("I")("T")("S")(design_units_code);@/
load12("C")("O")("D")("I")("N")("G")
  ("S")("C")("H")("E")("M")("E")(coding_scheme_code);@/
load6("F")("A")("M")("I")("L")("Y")(family_code);@/
load4("F")("A")("C")("E")(face_code);@/
load16("S")("E")("V")("E")("N")("B")("I")("T")@/@t\hskip2em@>
  ("S")("A")("F")("E")("F")("L")("A")("G")(seven_bit_safe_flag_code);@/
load6("H")("E")("A")("D")("E")("R")(header_code);@/
load9("F")("O")("N")("T")("D")("I")("M")("E")("N")(font_dimen_code);@/
load8("L")("I")("G")("T")("A")("B")("L")("E")(lig_table_code);@/
load12("B")("O")("U")("N")("D")("A")("R")("Y")("C")("H")("A")("R")
  (boundary_char_code);@/
load9("C")("H")("A")("R")("A")("C")("T")("E")("R")(character_code);@/
load9("P")("A")("R")("A")("M")("E")("T")("E")("R")(parameter_code);@/
load6("C")("H")("A")("R")("W")("D")(char_wd_code);@/
load6("C")("H")("A")("R")("H")("T")(char_ht_code);@/
load6("C")("H")("A")("R")("D")("P")(char_dp_code);@/
load6("C")("H")("A")("R")("I")("C")(char_ic_code);@/
load10("N")("E")("X")("T")("L")("A")("R")("G")("E")("R")(next_larger_code);@/
load7("V")("A")("R")("C")("H")("A")("R")(var_char_code);@/
load3("T")("O")("P")(var_char_code+1);@/
load3("M")("I")("D")(var_char_code+2);@/
load3("B")("O")("T")(var_char_code+3);@/
load3("R")("E")("P")(var_char_code+4);@/
load3("E")("X")("T")(var_char_code+4); {compatibility with older \.{PL} format}
load7("C")("O")("M")("M")("E")("N")("T")(comment_code);@/
load5("L")("A")("B")("E")("L")(label_code);@/
load4("S")("T")("O")("P")(stop_code);@/
load4("S")("K")("I")("P")(skip_code);@/
load3("K")("R")("N")(krn_code);@/
load3("L")("I")("G")(lig_code);@/
load4("/")("L")("I")("G")(lig_code+2);@/
load5("/")("L")("I")("G")(">")(lig_code+6);@/
load4("L")("I")("G")("/")(lig_code+1);@/
load5("L")("I")("G")("/")(">")(lig_code+5);@/
load5("/")("L")("I")("G")("/")(lig_code+3);@/
load6("/")("L")("I")("G")("/")(">")(lig_code+7);@/
load7("/")("L")("I")("G")("/")(">")(">")(lig_code+11);@/

@ \.{VPL} files may contain the following in addition to the \.{PL} names.

@<Enter all the \.{VPL} names@>=
load6("V")("T")("I")("T")("L")("E")(virtual_title_code);@/
load7("M")("A")("P")("F")("O")("N")("T")(map_font_code);@/
load3("M")("A")("P")(map_code);@/
load8("F")("O")("N")("T")("N")("A")("M")("E")(font_name_code);@/
load8("F")("O")("N")("T")("A")("R")("E")("A")(font_area_code);@/
load12("F")("O")("N")("T")
 ("C")("H")("E")("C")("K")("S")("U")("M")(font_checksum_code);@/
load6("F")("O")("N")("T")("A")("T")(font_at_code);@/
load9("F")("O")("N")("T")("D")("S")("I")("Z")("E")(font_dsize_code);@/
load10("S")("E")("L")("E")("C")("T")("F")("O")("N")("T")(select_font_code);@/
load7("S")("E")("T")("C")("H")("A")("R")(set_char_code);@/
load7("S")("E")("T")("R")("U")("L")("E")(set_rule_code);@/
load9("M")("O")("V")("E")("R")("I")("G")("H")("T")(move_right_code);@/
load8("M")("O")("V")("E")("L")("E")("F")("T")(move_right_code+1);@/
load8("M")("O")("V")("E")("D")("O")("W")("N")(move_down_code);@/
load6("M")("O")("V")("E")("U")("P")(move_down_code+1);@/
load4("P")("U")("S")("H")(push_code);@/
load3("P")("O")("P")(pop_code);@/
load7("S")("P")("E")("C")("I")("A")("L")(special_code);@/
load10("S")("P")("E")("C")("I")("A")("L")("H")("E")("X")(special_hex_code);@/

@ @<Enter the parameter names@>=
load5("S")("L")("A")("N")("T")(parameter_code+1);@/
load5("S")("P")("A")("C")("E")(parameter_code+2);@/
load7("S")("T")("R")("E")("T")("C")("H")(parameter_code+3);@/
load6("S")("H")("R")("I")("N")("K")(parameter_code+4);@/
load7("X")("H")("E")("I")("G")("H")("T")(parameter_code+5);@/
load4("Q")("U")("A")("D")(parameter_code+6);@/
load10("E")("X")("T")("R")("A")("S")("P")("A")("C")("E")(parameter_code+7);@/
load4("N")("U")("M")("1")(parameter_code+8);@/
load4("N")("U")("M")("2")(parameter_code+9);@/
load4("N")("U")("M")("3")(parameter_code+10);@/
load6("D")("E")("N")("O")("M")("1")(parameter_code+11);@/
load6("D")("E")("N")("O")("M")("2")(parameter_code+12);@/
load4("S")("U")("P")("1")(parameter_code+13);@/
load4("S")("U")("P")("2")(parameter_code+14);@/
load4("S")("U")("P")("3")(parameter_code+15);@/
load4("S")("U")("B")("1")(parameter_code+16);@/
load4("S")("U")("B")("2")(parameter_code+17);@/
load7("S")("U")("P")("D")("R")("O")("P")(parameter_code+18);@/
load7("S")("U")("B")("D")("R")("O")("P")(parameter_code+19);@/
load6("D")("E")("L")("I")("M")("1")(parameter_code+20);@/
load6("D")("E")("L")("I")("M")("2")(parameter_code+21);@/
load10("A")("X")("I")("S")("H")("E")("I")("G")("H")("T")(parameter_code+22);@/
load20("D")("E")("F")("A")("U")("L")("T")("R")("U")("L")("E")@/@t\hskip2em@>
  ("T")("H")("I")("C")("K")("N")("E")("S")("S")(parameter_code+8);@/
load13("B")("I")("G")("O")("P")
  ("S")("P")("A")("C")("I")("N")("G")("1")(parameter_code+9);@/
load13("B")("I")("G")("O")("P")
  ("S")("P")("A")("C")("I")("N")("G")("2")(parameter_code+10);@/
load13("B")("I")("G")("O")("P")
  ("S")("P")("A")("C")("I")("N")("G")("3")(parameter_code+11);@/
load13("B")("I")("G")("O")("P")
  ("S")("P")("A")("C")("I")("N")("G")("4")(parameter_code+12);@/
load13("B")("I")("G")("O")("P")
  ("S")("P")("A")("C")("I")("N")("G")("5")(parameter_code+13);@/

@ When a left parenthesis has been scanned, the following routine
is used to interpret the keyword that follows, and to store the
equivalent value in |cur_code|.

@p procedure get_name;
begin incr(loc); incr(level); {pass the left parenthesis}
cur_char:=" ";
while cur_char=" " do get_next;
if (cur_char>")")or(cur_char<"(") then decr(loc); {back up one character}
name_length:=0; get_keyword_char; {prepare to scan the name}
while cur_char<>" " do
  begin if name_length=longest_name then cur_name[1]:="X" {force error}
  else incr(name_length);
  cur_name[name_length]:=cur_char;
  get_keyword_char;
  end;
lookup;
if name_ptr=0 then err_print('Sorry, I don''t know that property name');
@.Sorry, I don't know...@>
cur_code:=equiv[name_ptr];
end;

@* Scanning numeric data.
The next thing we need is a trio of subroutines to read the one-byte,
four-byte, and real numbers that may appear as property values.
These subroutines are careful to stick to numbers between $-2^{31}$
and $2^{31}-1$, inclusive, so that a computer with two's complement
32-bit arithmetic will not be interrupted by overflow.

@ The first number scanner, which returns a one-byte value, surely has
no problems of arithmetic overflow.

@p function get_byte:byte; {scans a one-byte property value}
var acc:integer; {an accumulator}
@!t:ASCII_code; {the type of value to be scanned}
begin repeat get_next;
until cur_char<>" "; {skip the blanks before the type code}
t:=cur_char; acc:=0;
repeat get_next;
until cur_char<>" "; {skip the blanks after the type code}
if t="C" then @<Scan an ASCII character code@>
else if t="D" then @<Scan a small decimal number@>
else if t="O" then @<Scan a small octal number@>
else if t="H" then @<Scan a small hexadecimal number@>
else if t="F" then @<Scan a face code@>
else skip_error('You need "C" or "D" or "O" or "H" or "F" here');
@.You need "C" or "D" ...here@>
cur_char:=" "; get_byte:=acc;
end;

@ The |get_next| routine converts lower case to upper case, but it leaves
the character in the buffer, so we can unconvert it.

@<Scan an ASCII...@>=
if (cur_char>=@'41)and(cur_char<=@'176)and
 ((cur_char<"(")or(cur_char>")")) then
  acc:=xord[buffer[loc]]
else skip_error('"C" value must be standard ASCII and not a paren')
@:C value}\.{"C" value must be...@>

@ @<Scan a small dec...@>=
begin while (cur_char>="0")and(cur_char<="9") do
  begin acc:=acc*10+cur_char-"0";
  if acc>255 then
    begin skip_error('This value shouldn''t exceed 255');
@.This value shouldn't...@>
    acc:=0; cur_char:=" ";
    end
  else get_next;
  end;
backup;
end

@ @<Scan a small oct...@>=
begin while (cur_char>="0")and(cur_char<="7") do
  begin acc:=acc*8+cur_char-"0";
  if acc>255 then
    begin skip_error('This value shouldn''t exceed ''377');
@.This value shouldn't...@>
    acc:=0; cur_char:=" ";
    end
  else get_next;
  end;
backup;
end

@ @<Scan a small hex...@>=
begin while ((cur_char>="0")and(cur_char<="9"))or
   ((cur_char>="A")and(cur_char<="F")) do
  begin if cur_char>="A" then cur_char:=cur_char+"0"+10-"A";
  acc:=acc*16+cur_char-"0";
  if acc>255 then
    begin skip_error('This value shouldn''t exceed "FF');
@.This value shouldn't...@>
    acc:=0; cur_char:=" ";
    end
  else get_next;
  end;
backup;
end

@ @<Scan a face...@>=
begin if cur_char="B" then acc:=2
else if cur_char="L" then acc:=4
else if cur_char<>"M" then acc:=18;
get_next;
if cur_char="I" then incr(acc)
else if cur_char<>"R" then acc:=18;
get_next;
if cur_char="C" then acc:=acc+6
else if cur_char="E" then acc:=acc+12
else if cur_char<>"R" then acc:=18;
if acc>=18 then
  begin skip_error('Illegal face code, I changed it to MRR');
@.Illegal face code...@>
  acc:=0;
  end;
end

@ The routine that scans a four-byte value puts its output into |cur_bytes|,
which is a record containing (yes, you guessed it) four bytes.

@<Types...@>=
@!four_bytes=record @!b0:byte;@+@!b1:byte;@+@!b2:byte;@+@!b3:byte;@+end;

@ @d c0==cur_bytes.b0
@d c1==cur_bytes.b1
@d c2==cur_bytes.b2
@d c3==cur_bytes.b3

@<Glob...@>=
@!cur_bytes:four_bytes; {a four-byte accumulator}
@!zero_bytes:four_bytes; {four bytes all zero}

@ @<Set init...@>=
zero_bytes.b0:=0; zero_bytes.b1:=0; zero_bytes.b2:=0; zero_bytes.b3:=0;

@ Since the |get_four_bytes| routine is used very infrequently, no attempt
has been made to make it fast; we only want it to work.

@p procedure get_four_bytes; {scans an unsigned constant and sets |four_bytes|}
var c:integer; {local two-byte accumulator}
@!r:integer; {radix}
begin repeat get_next;
until cur_char<>" "; {skip the blanks before the type code}
r:=0; cur_bytes:=zero_bytes; {start with the accumulator zero}
if cur_char="H" then r:=16
else if cur_char="O" then r:=8
else if cur_char="D" then r:=10
else skip_error('Decimal ("D"), octal ("O"), or hex ("H") value needed here');
@.Decimal ("D"), octal ("O"), or hex...@>
if r>0 then
  begin repeat get_next;
  until cur_char<>" "; {skip the blanks after the type code}
  while ((cur_char>="0")and(cur_char<="9"))or@|
      ((cur_char>="A")and(cur_char<="F")) do
    @<Multiply by |r|, add |cur_char-"0"|, and |get_next|@>;
  end;
end;

@ @<Multiply by |r|...@>=
begin if cur_char>="A" then cur_char:=cur_char+"0"+10-"A";
if cur_char>="0"+r then skip_error('Illegal digit')
@.Illegal digit@>
else  begin c:=c3*r+cur_char-"0"; c3:=c mod 256;@/
  c:=c2*r+c div 256; c2:=c mod 256;@/
  c:=c1*r+c div 256; c1:=c mod 256;@/
  c:=c0*r+c div 256;
  if c<256 then c0:=c
  else begin cur_bytes:=zero_bytes;
    if r=8 then
      skip_error('Sorry, the maximum octal value is O 37777777777')
@.Sorry, the maximum...@>
    else if r=10 then
      skip_error('Sorry, the maximum decimal value is D 4294967295')
    else skip_error('Sorry, the maximum hex value is H FFFFFFFF');
    end;
  get_next;
  end;
end

@ The remaining scanning routine is the most interesting. It scans a real
constant and returns the nearest |fix_word| approximation to that constant.
A |fix_word| is a 32-bit integer that represents a real value that
has been multiplied by $2^{20}$. Since \.{VPtoVF} restricts the magnitude
of reals to 2048, the |fix_word| will have a magnitude less than $2^{31}$.

@d unity==@'4000000 {$2^{20}$, the |fix_word| 1.0}

@<Types...@>=
@!fix_word=integer; {a scaled real value with 20 bits of fraction}

@ When a real value is desired, we might as well treat `\.D' and `\.R'
formats as if they were identical.

@p function get_fix:fix_word; {scans a real property value}
var negative:boolean; {was there a minus sign?}
@!acc:integer; {an accumulator}
@!int_part:integer; {the integer part}
@!j:0..7; {the number of decimal places stored}
begin repeat get_next;
until cur_char<>" "; {skip the blanks before the type code}
negative:=false; acc:=0; {start with the accumulators zero}
if (cur_char<>"R")and(cur_char<>"D") then
  skip_error('An "R" or "D" value is needed here')
@.An "R" or "D" ... needed here@>
else  begin @<Scan the blanks and/or signs after the type code@>;
  while (cur_char>="0") and (cur_char<="9") do
    @<Multiply by 10, add |cur_char-"0"|, and |get_next|@>;
  int_part:=acc; acc:=0;
  if cur_char="." then @<Scan the fraction part and put it in |acc|@>;
  if (acc>=unity)and(int_part=2047) then
    skip_error('Real constants must be less than 2048')
@.Real constants must be...@>
  else acc:=int_part*unity+acc;
  end;
if negative then get_fix:=-acc@+else get_fix:=acc;
end;

@ @<Scan the blanks...@>=
repeat get_next;
if cur_char="-" then
  begin cur_char:=" "; negative:=not negative;
  end
else if cur_char="+" then cur_char:=" ";
until cur_char<>" "

@ @<Multiply by 10...@>=
begin acc:=acc*10+cur_char-"0";
if acc>=2048 then
  begin skip_error('Real constants must be less than 2048');
@.Real constants must be...@>
  acc:=0; cur_char:=" ";
  end
else get_next;
end

@ To scan the fraction $.d_1d_2\ldots\,$, we keep track of up to seven
of the digits $d_j$. A correct result is obtained if we first compute
$f^\prime=\lfloor 2^{21}(d_1\ldots d_j)/10^j\rfloor$, after which
$f=\lfloor(f^\prime+1)/2\rfloor$. It is possible to have $f=1.0$.

@<Glob...@>=
@!fraction_digits:array[1..7] of integer; {$2^{21}$ times $d_j$}

@ @<Scan the frac...@>=
begin j:=0; get_next;
while (cur_char>="0")and(cur_char<="9") do
  begin if j<7 then
    begin incr(j); fraction_digits[j]:=@'10000000*(cur_char-"0");
    end;
  get_next;
  end;
acc:=0;
while j>0 do
  begin acc:=fraction_digits[j]+(acc div 10); decr(j);
  end;
acc:=(acc+10) div 20;
end

@* Storing the property values.
When property values have been found, they are squirreled away in a bunch
of arrays. The header information is unpacked into bytes in an array
called |header_bytes|. The ligature/kerning program is stored in an array
of type |four_bytes|.
Another |four_bytes| array holds the specifications of extensible characters.
The kerns and parameters are stored in separate arrays of |fix_word| values.
Virtual font data goes into an array |vf| of single-byte values.

We maintain information about at most 256 local fonts. (If this is inadequate,
several arrays need to be made longer and we need to output font definitions
that go beyond |fnt1| and |fnt_def1| in the \.{VF} file.)

Instead of storing the design size in the header array, we will keep it
in a |fix_word| variable until the last minute. The number of units in the
design size is also kept in a |fix_word|.

@<Glob...@>=
@!header_bytes:array[header_index] of byte; {the header block}
@!header_ptr:header_index; {the number of header bytes in use}
@!design_size:fix_word; {the design size}
@!design_units:fix_word; {reciprocal of the scaling factor}
@!frozen_du:boolean; {have we used |design_units| irrevocably?}
@!seven_bit_safe_flag:boolean; {does the file claim to be seven-bit-safe?}
@!lig_kern:array[0..max_lig_steps] of four_bytes; {the ligature program}
@!nl:0..32767; {the number of ligature/kern instructions so far}
@!min_nl:0..32767; {the final value of |nl| must be at least this}
@!kern:array[0..max_kerns] of fix_word; {the distinct kerning amounts}
@!nk:0..max_kerns; {the number of entries of |kern|}
@!exten:array[0..255] of four_bytes; {extensible character specs}
@!ne:0..256; {the number of extensible characters}
@!param:array[1..max_param_words] of fix_word; {\.{FONTDIMEN} parameters}
@!np:0..max_param_words; {the largest parameter set nonzero}
@!check_sum_specified:boolean; {did the user name the check sum?}
@!bchar:0..256; {the right boundary character, or 256 if unspecified}
@!vf:array[0..vf_size] of byte; {stored bytes for \.{VF} file}
@!vf_ptr:0..vf_size; {first unused location in |vf|}
@!vtitle_start:0..vf_size; {starting location of \.{VTITLE} string}
@!vtitle_length:byte; {length of \.{VTITLE} string}
@!packet_start:array[byte] of 0..vf_size;
  {beginning location of character packet}
@!packet_length:array[byte] of integer; {length of character packet}
@!font_ptr:0..256; {number of distinct local fonts seen}
@!cur_font:0..256; {number of the current local font}
@!fname_start:array[byte] of 0..vf_size; {beginning of local font name}
@!fname_length:array[byte] of byte; {length of local font name}
@!farea_start:array[byte] of 0..vf_size; {beginning of local font area}
@!farea_length:array[byte] of byte; {length of local font area}
@!font_checksum:array[byte] of four_bytes; {local font checksum}
@!font_number:array[0..256] of four_bytes; {local font id number}
@!font_at:array[byte] of fix_word; {local font ``at size''}
@!font_dsize:array[byte] of fix_word; {local font design size}

@ @<Types...@>=
@!header_index=0..max_header_bytes;
@!indx=0..@'77777;

@ @<Local...@>=
@!d:header_index; {an index into |header_bytes|}

@ We start by setting up the default values.

@d check_sum_loc=0
@d design_size_loc=4
@d coding_scheme_loc=8
@d family_loc=coding_scheme_loc+40
@d seven_flag_loc=family_loc+20
@d face_loc=seven_flag_loc+3

@<Set init...@>=
for d:=0 to 18*4-1 do header_bytes[d]:=0;
header_bytes[8]:=11; header_bytes[9]:="U";
header_bytes[10]:="N";
header_bytes[11]:="S";
header_bytes[12]:="P";
header_bytes[13]:="E";
header_bytes[14]:="C";
header_bytes[15]:="I";
header_bytes[16]:="F";
header_bytes[17]:="I";
header_bytes[18]:="E";
header_bytes[19]:="D";
@.UNSPECIFIED@>
for d:=family_loc to family_loc+11 do header_bytes[d]:=header_bytes[d-40];
design_size:=10*unity; design_units:=unity; frozen_du:=false;
seven_bit_safe_flag:=false;@/
header_ptr:=18*4; nl:=0; min_nl:=0; nk:=0; ne:=0; np:=0;@/
check_sum_specified:=false; bchar:=256;@/
vf_ptr:=0; vtitle_start:=0; vtitle_length:=0; font_ptr:=0;
for k:=0 to 255 do packet_start[k]:=vf_size;
for k:=0 to 127 do packet_length[k]:=1;
for k:=128 to 255 do packet_length[k]:=2;

@ Most of the dimensions, however, go into the |memory| array. There are
at most 257 widths, 257 heights, 257 depths, and 257 italic corrections,
since the value 0 is required but it need not be used. So |memory| has room
for 1028 entries, each of which is a |fix_word|. An auxiliary table called
|link| is used to link these words together in linear lists, so that
sorting and other operations can be done conveniently.

We also add four ``list head'' words to the |memory| and |link| arrays;
these are in locations |width| through |italic|, i.e., 1 through 4.
For example, |link[height]| points to the smallest element in
the sorted list of distinct heights that have appeared so far, and
|memory[height]| is the number of distinct heights.

@d mem_size=1028+4 {number of nonzero memory addresses}

@<Types...@>=
@!pointer=0..mem_size; {an index into memory}

@ The arrays |char_wd|, |char_ht|, |char_dp|, and |char_ic| contain
pointers to the |memory| array entries where the corresponding dimensions
appear. Two other arrays, |char_tag| and |char_remainder|, hold
the other information that \.{TFM} files pack into a |char_info_word|.

@d no_tag=0 {vanilla character}
@d lig_tag=1 {character has a ligature/kerning program}
@d list_tag=2 {character has a successor in a charlist}
@d ext_tag=3 {character is extensible}
@d bchar_label==char_remainder[256]
  {beginning of ligature program for left boundary}

@<Glob...@>=
@!memory:array[pointer] of fix_word; {character dimensions and kerns}
@!mem_ptr:pointer; {largest |memory| word in use}
@!link:array[pointer] of pointer; {to make lists of |memory| items}
@!char_wd:array[byte] of pointer; {pointers to the widths}
@!char_ht:array[byte] of pointer; {pointers to the heights}
@!char_dp:array[byte] of pointer; {pointers to the depths}
@!char_ic:array[byte] of pointer; {pointers to italic corrections}
@!char_tag:array[byte] of no_tag..ext_tag; {character tags}
@!char_remainder:array[0..256] of 0..65535; {pointers to ligature labels,
  next larger characters, or extensible characters}

@ @<Local...@>=
@!c:byte; {runs through all character codes}

@ @<Set init...@>=
bchar_label:=@'77777;
for c:=0 to 255 do
  begin char_wd[c]:=0; char_ht[c]:=0; char_dp[c]:=0; char_ic[c]:=0;@/
  char_tag[c]:=no_tag; char_remainder[c]:=0;
  end;
memory[0]:=@'17777777777; {an ``infinite'' element at the end of the lists}
memory[width]:=0; link[width]:=0; {width list is empty}
memory[height]:=0; link[height]:=0; {height list is empty}
memory[depth]:=0; link[depth]:=0; {depth list is empty}
memory[italic]:=0; link[italic]:=0; {italic list is empty}
mem_ptr:=italic;

@ As an example of these data structures, let us consider the simple
routine that inserts a potentially new element into one of the dimension
lists. The first parameter indicates the list head (i.e., |h=width| for
the width list, etc.); the second parameter is the value that is to be
inserted into the list if it is not already present.  The procedure
returns the value of the location where the dimension appears in |memory|.
The fact that |memory[0]| is larger than any legal dimension makes the
algorithm particularly short.

We do have to handle two somewhat subtle situations. A width of zero must be
put into the list, so that a zero-width character in the font will not appear
to be nonexistent (i.e., so that its |char_wd| index will not be zero), but
this does not need to be done for heights, depths, or italic corrections.
Furthermore, it is necessary to test for memory overflow even though we
have provided room for the maximum number of different dimensions in any
legal font, since the \.{VPL} file might foolishly give any number of
different sizes to the same character.

@p function sort_in(@!h:pointer;@!d:fix_word):pointer; {inserts into list}
var p:pointer; {the current node of interest}
begin if (d=0)and(h<>width) then sort_in:=0
else begin p:=h;
  while d>=memory[link[p]] do p:=link[p];
  if (d=memory[p])and(p<>h) then sort_in:=p
  else if mem_ptr=mem_size then
    begin err_print('Memory overflow: more than 1028 widths, etc');
@.Memory overflow...@>
    print_ln('Congratulations! It''s hard to make this error.');
    sort_in:=p;
    end
  else  begin incr(mem_ptr); memory[mem_ptr]:=d;
    link[mem_ptr]:=link[p]; link[p]:=mem_ptr; incr(memory[h]);
    sort_in:=mem_ptr;
    end;
  end;
end;

@ When these lists of dimensions are eventually written to the \.{TFM}
file, we may have to do some rounding of values, because the \.{TFM} file
allows at most 256 widths, 16 heights, 16 depths, and 64 italic
corrections. The following procedure takes a given list head |h| and a
given dimension |d|, and returns the minimum $m$ such that the elements of
the list can be covered by $m$ intervals of width $d$.  It also sets
|next_d| to the smallest value $d^\prime>d$ such that the covering found
by this procedure would be different.  In particular, if $d=0$ it computes
the number of elements of the list, and sets |next_d| to the smallest
distance between two list elements. (The covering by intervals of width
|next_d| is not guaranteed to have fewer than $m$ elements, but in practice
this seems to happen most of the time.)

@<Glob...@>=
@!next_d:fix_word; {the next larger interval that is worth trying}

@ Once again we can make good use of the fact that |memory[0]| is ``infinite.''

@p function min_cover(@!h:pointer;@!d:fix_word):integer;
var p:pointer; {the current node of interest}
@!l:fix_word; {the least element covered by the current interval}
@!m:integer; {the current size of the cover being generated}
begin m:=0; p:=link[h]; next_d:=memory[0];
while p<>0 do
  begin incr(m); l:=memory[p];
  while memory[link[p]]<=l+d do p:=link[p];
  p:=link[p];
  if memory[p]-l<next_d then next_d:=memory[p]-l;
  end;
min_cover:=m;
end;

@ The following procedure uses |min_cover| to determine the smallest $d$
such that a given list can be covered with at most a given number of
intervals.

@p function shorten(@!h:pointer;m:integer):fix_word; {finds best way to round}
var d:fix_word; {the current trial interval length}
@!k:integer; {the size of a minimum cover}
begin if memory[h]>m then
  begin excess:=memory[h]-m;
  k:=min_cover(h,0); d:=next_d; {now the answer is at least |d|}
  repeat d:=d+d; k:=min_cover(h,d);
  until k<=m; {first we ascend rapidly until finding the range}
  d:=d div 2; k:=min_cover(h,d); {now we run through the feasible steps}
  while k>m do
    begin d:=next_d; k:=min_cover(h,d);
    end;
  shorten:=d;
  end
else shorten:=0;
end;

@ When we are nearly ready to output the \.{TFM} file, we will set
|index[p]:=k| if the dimension in |memory[p]| is being rounded to the
|k|th element of its list.

@<Glob...@>=
@!index:array[pointer] of byte;
@!excess:byte; {number of words to remove, if list is being shortened}

@ Here is the procedure that sets the |index| values. It also shortens
the list so that there is only one element per covering interval;
the remaining elements are the midpoints of their clusters.

@p procedure set_indices(@!h:pointer;@!d:fix_word); {reduces and indexes a list}
var p:pointer; {the current node of interest}
@!q:pointer; {trails one step behind |p|}
@!m:byte; {index number of nodes in the current interval}
@!l:fix_word; {least value in the current interval}
begin q:=h; p:=link[q]; m:=0;
while p<>0 do
  begin incr(m); l:=memory[p]; index[p]:=m;
  while memory[link[p]]<=l+d do
    begin p:=link[p]; index[p]:=m; decr(excess);
    if excess=0 then d:=0;
    end;
  link[q]:=p; memory[p]:=l+(memory[p]-l) div 2; q:=p; p:=link[p];
  end;
memory[h]:=m;
end;

@* The input phase.
We're ready now to read and parse the \.{VPL} file, storing property
values as we go.

@<Glob...@>=
@!c:byte; {the current character or byte being processed}
@!x:fix_word; {current dimension of interest}
@!k:integer; {general-purpose index}

@ @<Read all the input@>=
cur_char:=" ";
repeat while cur_char=" " do get_next;
if cur_char="(" then @<Read a font property value@>
else if (cur_char=")")and not input_has_ended then
  begin err_print('Extra right parenthesis');
  incr(loc); cur_char:=" ";
  end
@.Extra right parenthesis@>
else if not input_has_ended then junk_error;
until input_has_ended

@ The |junk_error| routine just referred to is called when something
appears in the forbidden area between properties of a property list.

@p procedure junk_error; {gets past no man's land}
begin err_print('There''s junk here that is not in parentheses');
@.There's junk here...@>
skip_to_paren;
end;

@ For each font property, we are supposed to read the data from the
left parenthesis that is the current value of |cur_char| to the right
parenthesis that matches it in the input. The main complication is
to recover with reasonable grace from various error conditions that might arise.

@<Read a font property value@>=
begin get_name;
if cur_code=comment_code then skip_to_end_of_item
else if cur_code>character_code then
  flush_error('This property name doesn''t belong on the outer level')
@.This property name doesn't belong...@>
else  begin @<Read the font property value specified by |cur_code|@>;
  finish_the_property;
  end;
end

@ @<Read the font property value spec...@>=
case cur_code of
check_sum_code: begin check_sum_specified:=true; read_four_bytes(check_sum_loc);
  end;
design_size_code: @<Read the design size@>;
design_units_code: @<Read the design units@>;
coding_scheme_code: read_BCPL(coding_scheme_loc,40);
family_code: read_BCPL(family_loc,20);
face_code:header_bytes[face_loc]:=get_byte;
seven_bit_safe_flag_code: @<Read the seven-bit-safe flag@>;
header_code: @<Read an indexed header word@>;
font_dimen_code: @<Read font parameter list@>;
lig_table_code: read_lig_kern;
boundary_char_code: bchar:=get_byte;
virtual_title_code: begin vtitle_start:=vf_ptr; copy_to_end_of_item;
  if vf_ptr>vtitle_start+255 then
    begin err_print('VTITLE clipped to 255 characters'); vtitle_length:=255;
@.VTITLE clipped...@>
    end
  else vtitle_length:=vf_ptr-vtitle_start;
  end;
map_font_code:@<Read a local font list@>;
character_code: read_char_info;
end

@ The |case| statement just given makes use of three subroutines that we
haven't defined yet. The first of these puts a 32-bit octal quantity
into four specified bytes of the header block.

@p procedure read_four_bytes(l:header_index);
begin get_four_bytes;
header_bytes[l]:=c0;
header_bytes[l+1]:=c1;
header_bytes[l+2]:=c2;
header_bytes[l+3]:=c3;
end;

@ The second little procedure is used to scan a string and to store it in
the ``{\mc BCPL} format'' required by \.{TFM} files. The string is supposed
to contain at most |n| bytes, including the first byte (which holds the
length of the rest of the string).

@p procedure read_BCPL(l:header_index;n:byte);
var k:header_index;
begin k:=l;
while cur_char=" " do get_next;
while (cur_char<>"(")and(cur_char<>")") do
  begin if k<l+n then incr(k);
  if k<l+n then header_bytes[k]:=cur_char;
  get_next;
  end;
if k=l+n then
  begin err_print('String is too long; its first ',n-1:1,
@.String is too long...@>
    ' characters will be kept'); decr(k);
  end;
header_bytes[l]:=k-l;
while k<l+n-1 do {tidy up the remaining bytes by setting them to nulls}
  begin incr(k); header_bytes[k]:=0;
  end;
end;

@ @<Read the design size@>=
begin next_d:=get_fix;
if next_d<unity then
  err_print('The design size must be at least 1')
@.The design size must...@>
else design_size:=next_d;
end

@ @<Read the design units@>=
begin next_d:=get_fix;
if next_d<=0 then
  err_print('The number of units per design size must be positive')
@.The number of units...@>
else if frozen_du then
  err_print('Sorry, it''s too late to change the design units')
@.Sorry, it's too late...@>
else design_units:=next_d;
end

@ @<Read the seven-bit-safe...@>=
begin while cur_char=" " do get_next;
if cur_char="T" then seven_bit_safe_flag:=true
else if cur_char="F" then seven_bit_safe_flag:=false
else err_print('The flag value should be "TRUE" or "FALSE"');
@.The flag value should be...@>
skip_to_paren;
end

@ @<Read an indexed header word@>=
begin c:=get_byte;
if c<18 then skip_error('HEADER indices should be 18 or more')
@.HEADER indices...@>
else if 4*c+4>max_header_bytes then
  skip_error('This HEADER index is too big for my present table size')
@.This HEADER index is too big...@>
else  begin while header_ptr<4*c+4 do
    begin header_bytes[header_ptr]:=0; incr(header_ptr);
    end;
  read_four_bytes(4*c);
  end;
end

@ The remaining kinds of font property values that need to be read are
those that involve property lists on higher levels. Each of these has a
loop similar to the one that was used at level zero. Then we put the
right parenthesis back so that `|finish_the_property|' will be happy;
there is probably a more elegant way to do this.

@d finish_inner_property_list==begin decr(loc); incr(level); cur_char:=")";
  end

@<Read font parameter list@>=
begin while level=1 do
  begin while cur_char=" " do get_next;
  if cur_char="(" then @<Read a parameter value@>
  else if cur_char=")" then skip_to_end_of_item
  else junk_error;
  end;
finish_inner_property_list;
end

@ @<Read a parameter value@>=
begin get_name;
if cur_code=comment_code then skip_to_end_of_item
else if (cur_code<parameter_code)or(cur_code>=char_wd_code) then
  flush_error('This property name doesn''t belong in a FONTDIMEN list')
@.This property name doesn't belong...@>
else  begin if cur_code=parameter_code then c:=get_byte
  else c:=cur_code-parameter_code;
  if c=0 then flush_error('PARAMETER index must not be zero')
@.PARAMETER index must not...@>
  else if c>max_param_words then
    flush_error('This PARAMETER index is too big for my present table size')
@.This PARAMETER index is too big...@>
  else  begin while np<c do
      begin incr(np); param[np]:=0;
      end;
    param[c]:=get_fix;
    finish_the_property;
    end;
  end;
end

@ @d numbers_differ==(font_number[cur_font].b3<>font_number[font_ptr].b3)or@|
(font_number[cur_font].b2<>font_number[font_ptr].b2)or@|
(font_number[cur_font].b1<>font_number[font_ptr].b1)or@|
(font_number[cur_font].b0<>font_number[font_ptr].b0)

@<Read a local font list@>=
begin get_four_bytes; font_number[font_ptr]:=cur_bytes; cur_font:=0;
while numbers_differ do incr(cur_font);
if cur_font=font_ptr then {it's a new font number}
  if font_ptr<256 then @<Initialize a new local font@>
  else err_print('I can handle only 256 different mapfonts');
@.I can handle only 256...@>
if cur_font=font_ptr then skip_to_end_of_item
else while level=1 do
  begin while cur_char=" " do get_next;
  if cur_char="(" then @<Read a local font property@>
  else if cur_char=")" then skip_to_end_of_item
  else junk_error;
  end;
finish_inner_property_list;
end

@ @<Initialize a new local font@>=
begin incr(font_ptr);
fname_start[cur_font]:=vf_size; fname_length[cur_font]:=4; {\.{NULL}}
farea_start[cur_font]:=vf_size; farea_length[cur_font]:=0;
font_checksum[cur_font]:=zero_bytes;
font_at[cur_font]:=@'4000000; {denotes design size of this virtual font}
font_dsize[cur_font]:=@'50000000; {the |fix_word| for 10}
end

@ @<Read a local font property@>=
begin get_name;
if cur_code=comment_code then skip_to_end_of_item
else if (cur_code<font_name_code)or(cur_code>font_dsize_code) then
  flush_error('This property name doesn''t belong in a MAPFONT list')
@.This property name doesn't belong...@>
else  begin case cur_code of
  font_name_code:@<Read a local font name@>;
  font_area_code:@<Read a local font area@>;
  font_checksum_code:begin get_four_bytes; font_checksum[cur_font]:=cur_bytes;
    end;
  font_at_code: begin frozen_du:=true;
    if design_units=unity then font_at[cur_font]:=get_fix
    else font_at[cur_font]:=round((get_fix/design_units)*1048576.0);
    end;
  font_dsize_code:font_dsize[cur_font]:=get_fix;
  end; {there are no other cases}
  finish_the_property;
  end;
end

@ @<Read a local font name@>=
begin fname_start[cur_font]:=vf_ptr; copy_to_end_of_item;
if vf_ptr>fname_start[cur_font]+255 then
  begin err_print('FONTNAME clipped to 255 characters');
@.FONTNAME clipped...@>
  fname_length[cur_font]:=255;
  end
else fname_length[cur_font]:=vf_ptr-fname_start[cur_font];
end

@ @<Read a local font area@>=
begin farea_start[cur_font]:=vf_ptr; copy_to_end_of_item;
if vf_ptr>farea_start[cur_font]+255 then
  begin err_print('FONTAREA clipped to 255 characters');
@.FONTAREA clipped...@>
  farea_length[cur_font]:=255;
  end
else farea_length[cur_font]:=vf_ptr-farea_start[cur_font];
end

@ @<Read ligature/kern list@>=
begin lk_step_ended:=false;
while level=1 do
  begin while cur_char=" " do get_next;
  if cur_char="(" then @<Read a ligature/kern command@>
  else if cur_char=")" then skip_to_end_of_item
  else junk_error;
  end;
finish_inner_property_list;
end

@ @<Read a ligature/kern command@>=
begin get_name;
if cur_code=comment_code then skip_to_end_of_item
else if cur_code<label_code then
  flush_error('This property name doesn''t belong in a LIGTABLE list')
@.This property name doesn't belong...@>
else  begin case cur_code of
  label_code:@<Read a label step@>;
  stop_code:@<Read a stop step@>;
  skip_code:@<Read a skip step@>;
  krn_code:@<Read a kerning step@>;
  lig_code,lig_code+1,lig_code+2,lig_code+3,lig_code+5,lig_code+6,lig_code+7,
    lig_code+11:@<Read a ligature step@>;
  end; {there are no other cases |>=label_code|}
  finish_the_property;
  end;
end

@ When a character is about to be tagged, we call the following
procedure so that an error message is given in case of multiple tags.

@p procedure check_tag(c:byte); {print error if |c| already tagged}
begin case char_tag[c] of
no_tag: do_nothing;
lig_tag: err_print('This character already appeared in a LIGTABLE LABEL');
@.This character already...@>
list_tag: err_print('This character already has a NEXTLARGER spec');
ext_tag: err_print('This character already has a VARCHAR spec');
end;
end;

@ @<Read a label step@>=
begin while cur_char=" " do get_next;
if cur_char="B" then
  begin bchar_label:=nl; skip_to_paren; {\.{LABEL BOUNDARYCHAR}}
  end
else begin backup; c:=get_byte;
  check_tag(c); char_tag[c]:=lig_tag; char_remainder[c]:=nl;
  end;
if min_nl<=nl then min_nl:=nl+1;
lk_step_ended:=false;
end

@ @d stop_flag=128 {value indicating `\.{STOP}' in a lig/kern program}
@d kern_flag=128 {op code for a kern step}

@<Globals...@>=
@!lk_step_ended:boolean;
  {was the last \.{LIGTABLE} property \.{LIG} or \.{KRN}?}
@!krn_ptr:0..max_kerns; {an index into |kern|}

@ @<Read a stop step@>=
if not lk_step_ended then
  err_print('STOP must follow LIG or KRN')
@.STOP must follow LIG or KRN@>
else begin lig_kern[nl-1].b0:=stop_flag; lk_step_ended:=false;
  end

@ @<Read a skip step@>=
if not lk_step_ended then
  err_print('SKIP must follow LIG or KRN')
@.SKIP must follow LIG or KRN@>
else begin c:=get_byte;
  if c>=128 then err_print('Maximum SKIP amount is 127')
@.Maximum SKIP amount...@>
  else if nl+c>=max_lig_steps then
    err_print('Sorry, LIGTABLE too long for me to handle')
@.Sorry, LIGTABLE too long...@>
  else begin lig_kern[nl-1].b0:=c;
    if min_nl<=nl+c then min_nl:=nl+c+1;
    end;
  lk_step_ended:=false;
  end

@ @<Read a ligature step@>=
begin lig_kern[nl].b0:=0;
lig_kern[nl].b2:=cur_code-lig_code;
lig_kern[nl].b1:=get_byte;
lig_kern[nl].b3:=get_byte;
if nl>=max_lig_steps-1 then
  err_print('Sorry, LIGTABLE too long for me to handle')
@.Sorry, LIGTABLE too long...@>
else incr(nl);
lk_step_ended:=true;
end

@ @<Read a kerning step@>=
begin lig_kern[nl].b0:=0; lig_kern[nl].b1:=get_byte;
kern[nk]:=get_fix; krn_ptr:=0;
while kern[krn_ptr]<>kern[nk] do incr(krn_ptr);
if krn_ptr=nk then
  begin if nk<max_kerns then incr(nk)
  else  begin err_print('Sorry, too many different kerns for me to handle');
@.Sorry, too many different kerns...@>
    decr(krn_ptr);
    end;
  end;
lig_kern[nl].b2:=kern_flag+(krn_ptr div 256);
lig_kern[nl].b3:=krn_ptr mod 256;
if nl>=max_lig_steps-1 then
  err_print('Sorry, LIGTABLE too long for me to handle')
@.Sorry, LIGTABLE too long...@>
else incr(nl);
lk_step_ended:=true;
end

@ Finally we come to the part of \.{VPtoVF}'s input mechanism
that is used most, the processing of individual character data.

@<Read character info list@>=
begin c:=get_byte; {read the character code that is being specified}
@<Print |c| in octal notation@>;
while level=1 do
  begin while cur_char=" " do get_next;
  if cur_char="(" then @<Read a character property@>
  else if cur_char=")" then skip_to_end_of_item
  else junk_error;
  end;
if char_wd[c]=0 then char_wd[c]:=sort_in(width,0); {legitimatize |c|}
finish_inner_property_list;
end

@ @<Read a character prop...@>=
begin get_name;
if cur_code=comment_code then skip_to_end_of_item
else if (cur_code<char_wd_code)or(cur_code>var_char_code) then
  flush_error('This property name doesn''t belong in a CHARACTER list')
@.This property name doesn't belong...@>
else  begin case cur_code of
  char_wd_code:char_wd[c]:=sort_in(width,get_fix);
  char_ht_code:char_ht[c]:=sort_in(height,get_fix);
  char_dp_code:char_dp[c]:=sort_in(depth,get_fix);
  char_ic_code:char_ic[c]:=sort_in(italic,get_fix);
  next_larger_code:begin check_tag(c); char_tag[c]:=list_tag;
    char_remainder[c]:=get_byte;
    end;
  map_code:read_packet(c);
  var_char_code:@<Read an extensible recipe for |c|@>;
  end;@/
  finish_the_property;
  end;
end

@ @<Read an extensible r...@>=
begin if ne=256 then
  err_print('At most 256 VARCHAR specs are allowed')
@.At most 256 VARCHAR specs...@>
else  begin check_tag(c); char_tag[c]:=ext_tag; char_remainder[c]:=ne;@/
  exten[ne]:=zero_bytes;
  while level=2 do
    begin while cur_char=" " do get_next;
    if cur_char="(" then @<Read an extensible piece@>
    else if cur_char=")" then skip_to_end_of_item
    else junk_error;
    end;
  incr(ne);
  finish_inner_property_list;
  end;
end

@ @<Read an extensible p...@>=
begin get_name;
if cur_code=comment_code then skip_to_end_of_item
else if (cur_code<var_char_code+1)or(cur_code>var_char_code+4) then
  flush_error('This property name doesn''t belong in a VARCHAR list')
@.This property name doesn't belong...@>
else  begin case cur_code-(var_char_code+1) of
  0:exten[ne].b0:=get_byte;
  1:exten[ne].b1:=get_byte;
  2:exten[ne].b2:=get_byte;
  3:exten[ne].b3:=get_byte;
  end;@/
  finish_the_property;
  end;
end

@* Assembling the mappings.
Each \.{MAP} property is a sequence of \.{DVI} instructions, for which
we need to know some of the opcodes.

@d set_char_0=0 {\.{DVI} command to 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 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}

@ We keep stacks of movement values, in order to optimize the \.{DVI} code
in simple cases.

@<Glob...@>=
@!hstack:array[0..max_stack] of 0..2; {number of known horizontal movements}
@!vstack:array[0..max_stack] of 0..2; {number of known vertical movements}
@!wstack,@!xstack,@!ystack,@!zstack:array[0..max_stack] of fix_word;
@!stack_ptr:0..max_stack;

@ The packet is built by straightforward assembly of \.{DVI} instructions.

@p @<Declare the |vf_fix| procedure@>@;@/
procedure read_packet(@!c:byte);
var @!cc:byte; {character being typeset}
@!x:fix_word; {movement}
@!h,@!v:0..2; {top of |hstack| and |vstack|}
@!special_start:0..vf_size; {location of |xxx1| command}
@!k:0..vf_size; {loop index}
begin packet_start[c]:=vf_ptr; stack_ptr:=0; h:=0; v:=0;
cur_font:=0;
while level=2 do
  begin while cur_char=" " do get_next;
  if cur_char="(" then @<Read and assemble a list of \.{DVI} commands@>
  else if cur_char=")" then skip_to_end_of_item
  else junk_error;
  end;
while stack_ptr>0 do
  begin err_print('Missing POP supplied');
@.Missing POP supplied@>
  vf_store(pop); decr(stack_ptr);
  end;
packet_length[c]:=vf_ptr-packet_start[c];
finish_inner_property_list;
end;

@ @<Read and assemble a list of \.{DVI}...@>=
begin get_name;
if cur_code=comment_code then skip_to_end_of_item
else if (cur_code<select_font_code)or(cur_code>special_hex_code) then
  flush_error('This property name doesn''t belong in a MAP list')
@.This property name doesn't belong...@>
else  begin case cur_code of
  select_font_code:@<Assemble a font selection@>;
  set_char_code:@<Assemble a typesetting instruction@>;
  set_rule_code:@<Assemble a rulesetting instruction@>;
  move_right_code,move_right_code+1:@<Assemble a horizontal movement@>;
  move_down_code,move_down_code+1:@<Assemble a vertical movement@>;
  push_code:@<Assemble a stack push@>;
  pop_code:@<Assemble a stack pop@>;
  special_code,special_hex_code:@<Assemble a special command@>;
  end;@/
  finish_the_property;
  end;
end

@ @<Assemble a font selection@>=
begin get_four_bytes; font_number[font_ptr]:=cur_bytes;
cur_font:=0;
while numbers_differ do incr(cur_font);
if cur_font=font_ptr then err_print('Undefined MAPFONT cannot be selected')
@.Undefined MAPFONT...@>
else if cur_font<64 then vf_store(fnt_num_0+cur_font)
else begin vf_store(fnt1); vf_store(cur_font);
  end;
end

@ @<Assemble a typesetting instruction@>=
if cur_font=font_ptr then
 err_print('Character cannot be typeset in undefined font')
@.Character cannot be typeset...@>
else begin cc:=get_byte;
  if cc>=128 then vf_store(set1);
  vf_store(cc);
  end

@ Here's a procedure that converts a |fix_word| to a sequence of
\.{DVI} bytes.

@<Declare the |vf_fix|...@>=
procedure vf_fix(@!opcode:byte;@!x:fix_word);
var negative:boolean;
@!k:0..4; {number of bytes to typeset}
@!t:integer; {threshold}
begin frozen_du:=true;
if design_units<>unity then x:=round((x/design_units)*1048576.0);
if x>=0 then negative:=false
else begin negative:=true; x:=-1-x;@+end;
if opcode=0 then
  begin k:=4; t:=@'100000000;@+end
else begin t:=127; k:=1;
  while x>t do
    begin t:=256*t+255; incr(k);
    end;
  vf_store(opcode+k-1); t:=t div 128 +1;
  end;
repeat if negative then
  begin vf_store(255-(x div t)); negative:=false;
  x:=(x div t)*t+t-1-x;
  end
else vf_store((x div t) mod 256);
decr(k); t:=t div 256;
until k=0;
end;

@ @<Assemble a rulesetting instruction@>=
begin vf_store(set_rule); vf_fix(0,get_fix); vf_fix(0,get_fix);
end

@ @<Assemble a horizontal movement@>=
begin if cur_code=move_right_code then x:=get_fix@+else x:=-get_fix;
if h=0 then
  begin wstack[stack_ptr]:=x; h:=1; vf_fix(w1,x);@+end
else if x=wstack[stack_ptr] then vf_store(w0)
else if h=1 then
  begin xstack[stack_ptr]:=x; h:=2; vf_fix(x1,x);@+end
else if x=xstack[stack_ptr] then vf_store(x0)
else vf_fix(right1,x);
end

@ @<Assemble a vertical movement@>=
begin if cur_code=move_down_code then x:=get_fix@+else x:=-get_fix;
if v=0 then
  begin ystack[stack_ptr]:=x; v:=1; vf_fix(y1,x);@+end
else if x=ystack[stack_ptr] then vf_store(y0)
else if v=1 then
  begin zstack[stack_ptr]:=x; v:=2; vf_fix(z1,x);@+end
else if x=zstack[stack_ptr] then vf_store(z0)
else vf_fix(down1,x);
end

@ @<Assemble a stack push@>=
if stack_ptr=max_stack then {too pushy}
  err_print('Don''t push so much---stack is full!')
@.Don't push so much...@>
else begin vf_store(push); hstack[stack_ptr]:=h; vstack[stack_ptr]:=v;
  incr(stack_ptr); h:=0; v:=0;
  end

@ @<Assemble a stack pop@>=
if stack_ptr=0 then
  err_print('Empty stack cannot be popped')
@.Empty stack...@>
else begin vf_store(pop); decr(stack_ptr);
  h:=hstack[stack_ptr]; v:=vstack[stack_ptr];
  end

@ @<Assemble a special command@>=
begin vf_store(xxx1); vf_store(0); {dummy length}
special_start:=vf_ptr;
if cur_code=special_code then copy_to_end_of_item
else begin repeat x:=get_hex;
   if cur_char>")" then vf_store(x*16+get_hex);
  until cur_char<=")";
  end;
if vf_ptr-special_start>255 then @<Convert |xxx1| command to |xxx4|@>
else vf[special_start-1]:=vf_ptr-special_start;
end

@ @<Convert |xxx1|...@>=
if vf_ptr+3>vf_size then
  begin err_print('Special command being clipped---no room left!');
@.Special command being clipped...@>
  vf_ptr:=special_start+255; vf[special_start-1]:=255;
  end
else begin for k:=vf_ptr downto special_start do vf[k+3]:=vf[k];
  x:=vf_ptr-special_start; vf_ptr:=vf_ptr+3;
  vf[special_start-2]:=xxx4;
  vf[special_start-1]:=x div @'100000000;
  vf[special_start]:=(x div @'200000) mod 256;
  vf[special_start+1]:=(x div @'400) mod 256;
  vf[special_start+2]:=x mod 256;
  end

@ The input routine is now complete except for the following code,
which prints a progress report as the file is being read.

@p procedure print_octal(c:byte); {prints three octal digits}
begin print('''',(c div 64):1,((c div 8) mod 8):1,(c mod 8):1);
end;

@ @<Print |c| in octal...@>=
begin if chars_on_line=8 then
  begin print_ln(' '); chars_on_line:=1;
  end
else  begin if chars_on_line>0 then print(' ');
  incr(chars_on_line);
  end;
print_octal(c); {progress report}
end

@* The checking and massaging phase.
Once the whole \.{VPL} file has been read in, we must check it for consistency
and correct any errors. This process consists mainly of running through
the characters that exist and seeing if they refer to characters that
don't exist. We also compute the true value of |seven_unsafe|; we make sure
that the charlists and ligature programs contain no loops; and we
shorten the lists of widths, heights, depths, and italic corrections,
if necessary, to keep from exceeding the required maximum sizes.

@<Glob...@>=
@!seven_unsafe:boolean; {do seven-bit characters generate eight-bit ones?}

@ @<Correct and check the information@>=
if nl>0 then @<Make sure the ligature/kerning program ends appropriately@>;
seven_unsafe:=false;
for c:=0 to 255 do if char_wd[c]<>0 then
    @<For all characters |g| generated by |c|,
    make sure that |char_wd[g]| is nonzero, and
    set |seven_unsafe| if |c<128<=g|@>;
if bchar_label<@'77777 then
  begin c:=256; @<Check ligature program of |c|@>;
  end;
if seven_bit_safe_flag and seven_unsafe then
  print_ln('The font is not really seven-bit-safe!');
@.The font is not...safe@>
@<Check for infinite ligature loops@>;
@<Doublecheck the lig/kern commands and the extensible recipes@>;
for c:=0 to 255 do
  @<Make sure that |c| is not the largest element of a charlist cycle@>;
@<Put the width, height, depth, and italic lists into final form@>

@ The checking that we need in several places is accomplished by three
macros that are only slightly tricky.

@d existence_tail(#)==begin char_wd[g]:=sort_in(width,0);
    print(#,' '); print_octal(c);
    print_ln(' had no CHARACTER spec.');
    end;
  end
@d check_existence_and_safety(#)==begin g:=#;
  if (g>=128)and(c<128) then seven_unsafe:=true;
  if char_wd[g]=0 then existence_tail
@d check_existence(#)==begin g:=#;
  if char_wd[g]=0 then existence_tail

@<For all characters |g| generated by |c|...@>=
case char_tag[c] of
no_tag: do_nothing;
lig_tag: @<Check ligature program of |c|@>;
list_tag: check_existence_and_safety(char_remainder[c])
  ('The character NEXTLARGER than');
@.The character NEXTLARGER...@>
ext_tag:@<Check the pieces of |exten[c]|@>;
end

@ @<Check the pieces...@>=
begin if exten[char_remainder[c]].b0>0 then
  check_existence_and_safety(exten[char_remainder[c]].b0)
    ('TOP piece of character');
@.TOP piece of character...@>
if exten[char_remainder[c]].b1>0 then
  check_existence_and_safety(exten[char_remainder[c]].b1)
    ('MID piece of character');
@.MID piece of character...@>
if exten[char_remainder[c]].b2>0 then
  check_existence_and_safety(exten[char_remainder[c]].b2)
    ('BOT piece of character');
@.BOT piece of character...@>
check_existence_and_safety(exten[char_remainder[c]].b3)
    ('REP piece of character');
@.REP piece of character...@>
end

@ @<Make sure that |c| is not the largest element of a charlist cycle@>=
if char_tag[c]=list_tag then
  begin g:=char_remainder[c];
  while (g<c)and(char_tag[g]=list_tag) do g:=char_remainder[g];
  if g=c then
    begin char_tag[c]:=no_tag;
    print('A cycle of NEXTLARGER characters has been broken at ');
@.A cycle of NEXTLARGER...@>
    print_octal(c); print_ln('.');
    end;
  end

@ @<Glob...@>=
@!delta:fix_word; {size of the intervals needed for rounding}

@ @d round_message(#)==if delta>0 then print_ln('I had to round some ',
@.I had to round...@>
  #,'s by ',(((delta+1) div 2)/@'4000000):1:7,' units.')

@<Put the width, height, depth, and italic lists into final form@>=
delta:=shorten(width,255); set_indices(width,delta); round_message('width');@/
delta:=shorten(height,15); set_indices(height,delta); round_message('height');@/
delta:=shorten(depth,15); set_indices(depth,delta); round_message('depth');@/
delta:=shorten(italic,63); set_indices(italic,delta);
  round_message('italic correction');

@ @d clear_lig_kern_entry== {make an unconditional \.{STOP}}
  lig_kern[nl].b0:=255; lig_kern[nl].b1:=0;
  lig_kern[nl].b2:=0; lig_kern[nl].b3:=0

@<Make sure the ligature/kerning program ends...@>=
begin if bchar_label<@'77777 then {make room for it}
  begin clear_lig_kern_entry; incr(nl);
  end; {|bchar_label| will be stored later}
while min_nl>nl do
  begin clear_lig_kern_entry; incr(nl);
  end;
if lig_kern[nl-1].b0=0 then lig_kern[nl-1].b0:=stop_flag;
end

@ It's not trivial to check for infinite loops generated by repeated
insertion of ligature characters. But fortunately there is a nice
algorithm for such testing, copied here from the program \.{TFtoPL}
where it is explained further.

@d simple=0 {$f(x,y)=z$}
@d left_z=1 {$f(x,y)=f(z,y)$}
@d right_z=2 {$f(x,y)=f(x,z)$}
@d both_z=3 {$f(x,y)=f(f(x,z),y)$}
@d pending=4 {$f(x,y)$ is being evaluated}


@ @<Glo...@>=
@!lig_ptr:0..max_lig_steps; {an index into |lig_kern|}
@!hash:array[0..hash_size] of 0..66048; {$256x+y+1$ for $x\le257$ and $y\le255$}
@!class:array[0..hash_size] of simple..pending;
@!lig_z:array[0..hash_size] of 0..257;
@!hash_ptr:0..hash_size; {the number of nonzero entries in |hash|}
@!hash_list:array[0..hash_size] of 0..hash_size; {list of those nonzero entries}
@!h,@!hh:0..hash_size; {indices into the hash table}
@!tt:indx; {temporary register}
@!x_lig_cycle,@!y_lig_cycle:0..256; {problematic ligature pair}

@ @<Set init...@>=
hash_ptr:=0; y_lig_cycle:=256;
for k:=0 to hash_size do hash[k]:=0;

@ @d lig_exam==lig_kern[lig_ptr].b1
@d lig_gen==lig_kern[lig_ptr].b3

@<Check lig...@>=
begin lig_ptr:=char_remainder[c];
repeat if hash_input(lig_ptr,c) then
  begin if lig_kern[lig_ptr].b2<kern_flag then
    begin if lig_exam<>bchar then
      check_existence(lig_exam)('LIG character examined by');
@.LIG character examined...@>
    check_existence(lig_gen)('LIG character generated by');
@.LIG character generated...@>
    if lig_gen>=128 then if(c<128)or(c=256) then
      if(lig_exam<128)or(lig_exam=bchar) then seven_unsafe:=true;
    end
  else if lig_exam<>bchar then
    check_existence(lig_exam)('KRN character examined by');
@.KRN character examined...@>
  end;
if lig_kern[lig_ptr].b0>=stop_flag then lig_ptr:=nl
else lig_ptr:=lig_ptr+1+lig_kern[lig_ptr].b0;
until lig_ptr>=nl;
end

@ The |hash_input| procedure is copied from \.{TFtoPL}, but it is made
into a boolean function that returns |false| if the ligature command
was masked by a previous one.

@p function hash_input(@!p,@!c:indx):boolean;
 {enter data for character |c| and command in location |p|, unless it isn't new}
label 30; {go here for a quick exit}
var @!cc:simple..both_z; {class of data being entered}
@!zz:0..255; {function value or ligature character being entered}
@!y:0..255; {the character after the cursor}
@!key:integer; {value to be stored in |hash|}
@!t:integer; {temporary register for swapping}
begin if hash_ptr=hash_size then
  begin hash_input:=false; goto 30;@+end;
@<Compute the command parameters |y|, |cc|, and |zz|@>;
key:=256*c+y+1; h:=(1009*key) mod hash_size;
while hash[h]>0 do
  begin if hash[h]<=key then
    begin if hash[h]=key then
      begin hash_input:=false; goto 30; {unused ligature command}
      end;
    t:=hash[h]; hash[h]:=key; key:=t; {do ordered-hash-table insertion}
    t:=class[h]; class[h]:=cc; cc:=t; {namely, do a swap}
    t:=lig_z[h]; lig_z[h]:=zz; zz:=t;
    end;
  if h>0 then decr(h)@+else h:=hash_size;
  end;
hash[h]:=key; class[h]:=cc; lig_z[h]:=zz;
incr(hash_ptr); hash_list[hash_ptr]:=h;
hash_input:=true;
30:end;

@ @<Compute the command param...@>=
y:=lig_kern[p].b1; t:=lig_kern[p].b2; cc:=simple;
zz:=lig_kern[p].b3;
if t>=kern_flag then zz:=y
else begin case t of
  0,6:do_nothing; {\.{LIG},\.{/LIG>}}
  5,11:zz:=y; {\.{LIG/>}, \.{/LIG/>>}}
  1,7:cc:=left_z; {\.{LIG/}, \.{/LIG/>}}
  2:cc:=right_z; {\.{/LIG}}
  3:cc:=both_z; {\.{/LIG/}}
  end; {there are no other cases}
  end

@ (More good stuff from \.{TFtoPL}.)

@p function f(@!h,@!x,@!y:indx):indx; forward;@t\2@>
  {compute $f$ for arguments known to be in |hash[h]|}
function eval(@!x,@!y:indx):indx; {compute $f(x,y)$ with hashtable lookup}
var @!key:integer; {value sought in hash table}
begin key:=256*x+y+1; h:=(1009*key) mod hash_size;
while hash[h]>key do
  if h>0 then decr(h)@+else h:=hash_size;
if hash[h]<key then eval:=y {not in ordered hash table}
else eval:=f(h,x,y);
end;

@ Pascal's beastly convention for |forward| declarations prevents us from
saying |function f(h,x,y:indx):indx| here.

@p function f;
begin case class[h] of
simple: do_nothing;
left_z: begin class[h]:=pending; lig_z[h]:=eval(lig_z[h],y); class[h]:=simple;
  end;
right_z: begin class[h]:=pending; lig_z[h]:=eval(x,lig_z[h]); class[h]:=simple;
  end;
both_z: begin class[h]:=pending; lig_z[h]:=eval(eval(x,lig_z[h]),y);
  class[h]:=simple;
  end;
pending: begin x_lig_cycle:=x; y_lig_cycle:=y; lig_z[h]:=257; class[h]:=simple;
  end; {the value 257 will break all cycles, since it's not in |hash|}
end; {there are no other cases}
f:=lig_z[h];
end;

@ @<Check for infinite...@>=
if hash_ptr<hash_size then for hh:=1 to hash_ptr do
  begin tt:=hash_list[hh];
  if class[tt]>simple then {make sure $f$ is well defined}
  tt:=f(tt,(hash[tt]-1)div 256,(hash[tt]-1)mod 256);
  end;
if(hash_ptr=hash_size)or(y_lig_cycle<256) then
  begin if hash_ptr<hash_size then
    begin print('Infinite ligature loop starting with ');
@.Infinite ligature loop...@>
    if x_lig_cycle=256 then print('boundary')@+else print_octal(x_lig_cycle);
    print(' and '); print_octal(y_lig_cycle); print_ln('!');
    end
  else print_ln('Sorry, I haven''t room for so many ligature/kern pairs!');
@.Sorry, I haven't room...@>
  print_ln('All ligatures will be cleared.');
  for c:=0 to 255 do if char_tag[c]=lig_tag then
    begin char_tag[c]:=no_tag; char_remainder[c]:=0;
    end;
  nl:=0; bchar:=256; bchar_label:=@'77777;
  end

@ The lig/kern program may still contain references to nonexistent characters,
if parts of that program are never used. Similarly, there may be extensible
characters that are never used, because they were overridden by
\.{NEXTLARGER}, say. This would produce an invalid \.{TFM} file; so we
must fix such errors.

@d double_check_tail(#)==@t\1@>if char_wd[0]=0
      then char_wd[0]:=sort_in(width,0);
    print('Unused ',#,' refers to nonexistent character ');
    print_octal(c); print_ln('!');
    end;
  end
@d double_check_lig(#)==begin c:=lig_kern[lig_ptr].#;
  if char_wd[c]=0 then if c<>bchar then
    begin lig_kern[lig_ptr].#:=0; double_check_tail
@d double_check_ext(#)==begin c:=exten[g].#;
  if c>0 then if char_wd[c]=0 then
    begin exten[g].#:=0; double_check_tail
@d double_check_rep(#)==begin c:=exten[g].#;
  if char_wd[c]=0 then
    begin exten[g].#:=0; double_check_tail

@<Doublecheck...@>=
if nl>0 then for lig_ptr:=0 to nl-1 do
  if lig_kern[lig_ptr].b2<kern_flag then
    begin if lig_kern[lig_ptr].b0<255 then
      begin double_check_lig(b1)('LIG step'); double_check_lig(b3)('LIG step');
      end;
    end
  else double_check_lig(b1)('KRN step');
@.Unused LIG step...@>
@.Unused KRN step...@>
if ne>0 then for g:=0 to ne-1 do
  begin double_check_ext(b0)('VARCHAR TOP');
  double_check_ext(b1)('VARCHAR MID');
  double_check_ext(b2)('VARCHAR BOT');
  double_check_rep(b3)('VARCHAR REP');
@.Unused VARCHAR...@>
  end

@* The TFM output phase.
Now that we know how to get all of the font data correctly stored in
\.{VPtoVF}'s memory, it only remains to write the answers out.

First of all, it is convenient to have an abbreviation for output to the
\.{TFM} file:

@d out(#)==write(tfm_file,#)

@ The general plan for producing \.{TFM} files is long but simple:

@<Do the \.{TFM} output@>=
@<Compute the twelve subfile sizes@>;
@<Output the twelve subfile sizes@>;
@<Output the header block@>;
@<Output the character info@>;
@<Output the dimensions themselves@>;
@<Output the ligature/kern program@>;
@<Output the extensible character recipes@>;
@<Output the parameters@>

@ A \.{TFM} file begins with 12 numbers that tell how big its subfiles are.
We already know most of these numbers; for example, the number of distinct
widths is |memory[width]+1|, where the $+1$ accounts for the zero width that
is always supposed to be present. But we still should compute the beginning
and ending character codes (|bc| and |ec|), the number of header words (|lh|),
and the total number of words in the \.{TFM} file (|lf|).

@<Gl...@>=
@!bc:byte; {the smallest character code in the font}
@!ec:byte; {the largest character code in the font}
@!lh:byte; {the number of words in the header block}
@!lf:0..32767; {the number of words in the entire \.{TFM} file}
@!not_found:boolean; {has a font character been found?}
@!temp_width:fix_word; {width being used to compute a check sum}

@ It might turn out that no characters exist at all. But \.{VPtoVF} keeps
going and writes the \.{TFM} anyway. In this case |ec| will be~0 and |bc|
will be~1.

@<Compute the twelve...@>=
lh:=header_ptr div 4;@/
not_found:=true; bc:=0;
while not_found do
  if (char_wd[bc]>0)or(bc=255) then not_found:=false
  else incr(bc);
not_found:=true; ec:=255;
while not_found do
  if (char_wd[ec]>0)or(ec=0) then not_found:=false
  else decr(ec);
if bc>ec then bc:=1;
incr(memory[width]); incr(memory[height]); incr(memory[depth]);
incr(memory[italic]);@/
@<Compute the ligature/kern program offset@>;
lf:=6+lh+(ec-bc+1)+memory[width]+memory[height]+memory[depth]+
memory[italic]+nl+lk_offset+nk+ne+np;

@ @d out_size(#)==out((#) div 256); out((#) mod 256)

@<Output the twelve subfile sizes@>=
out_size(lf); out_size(lh); out_size(bc); out_size(ec);
out_size(memory[width]); out_size(memory[height]);
out_size(memory[depth]); out_size(memory[italic]);
out_size(nl+lk_offset); out_size(nk); out_size(ne); out_size(np);

@ The routines that follow need a few temporary variables of different types.

@<Gl...@>=
@!j:0..max_header_bytes; {index into |header_bytes|}
@!p:pointer; {index into |memory|}
@!q:width..italic; {runs through the list heads for dimensions}
@!par_ptr:0..max_param_words; {runs through the parameters}

@ The header block follows the subfile sizes. The necessary information all
appears in |header_bytes|, except that the design size and the seven-bit-safe
flag must still be set.

@<Output the header block@>=
if not check_sum_specified then @<Compute the check sum@>;
header_bytes[design_size_loc]:=design_size div @'100000000;
  {this works since |design_size>0|}
header_bytes[design_size_loc+1]:=(design_size div @'200000) mod 256;
header_bytes[design_size_loc+2]:=(design_size div 256) mod 256;
header_bytes[design_size_loc+3]:=design_size  mod 256;
if not seven_unsafe then header_bytes[seven_flag_loc]:=128;
for j:=0 to header_ptr-1 do out(header_bytes[j]);

@ @<Compute the check sum@>=
begin c0:=bc; c1:=ec; c2:=bc; c3:=ec;
for c:=bc to ec do if char_wd[c]>0 then
  begin temp_width:=memory[char_wd[c]];
  if design_units<>unity then
    temp_width:=round((temp_width/design_units)*1048576.0);
  temp_width:=temp_width + (c+4)*@'20000000; {this should be positive}
  c0:=(c0+c0+temp_width) mod 255;
  c1:=(c1+c1+temp_width) mod 253;
  c2:=(c2+c2+temp_width) mod 251;
  c3:=(c3+c3+temp_width) mod 247;
  end;
header_bytes[check_sum_loc]:=c0;
header_bytes[check_sum_loc+1]:=c1;
header_bytes[check_sum_loc+2]:=c2;
header_bytes[check_sum_loc+3]:=c3;
end

@ The next block contains packed |char_info|.

@<Output the character info@>=
index[0]:=0;
for c:=bc to ec do
  begin out(index[char_wd[c]]);
  out(index[char_ht[c]]*16+index[char_dp[c]]);
  out(index[char_ic[c]]*4+char_tag[c]);
  out(char_remainder[c]);
  end

@ When a scaled quantity is output, we may need to divide it by |design_units|.
The following subroutine takes care of this, using floating point arithmetic
only if |design_units<>1.0|.

@p procedure out_scaled(x:fix_word); {outputs a scaled |fix_word|}
var @!n:byte; {the first byte after the sign}
@!m:0..65535; {the two least significant bytes}
begin if abs(x/design_units)>=16.0 then
  begin print_ln('The relative dimension ',x/@'4000000:1:3,
    ' is too large.');
@.The relative dimension...@>
  print('  (Must be less than 16*designsize');
  if design_units<>unity then print(' =',design_units/@'200000:1:3,
      ' designunits');
  print_ln(')'); x:=0;
  end;
if design_units<>unity then x:=round((x/design_units)*1048576.0);
if x<0 then
  begin out(255); x:=x+@'100000000;
  if x<=0 then x:=1;
  end
else begin out(0);
  if x>=@'100000000 then x:=@'77777777;
  end;
n:=x div @'200000; m:=x mod @'200000;
out(n); out(m div 256); out(m mod 256);
end;

@ We have output the packed indices for individual characters.
The scaled widths, heights, depths, and italic corrections are next.

@<Output the dimensions themselves@>=
for q:=width to italic do
  begin out(0); out(0); out(0); out(0); {output the zero word}
  p:=link[q]; {head of list}
  while p>0 do
    begin out_scaled(memory[p]);
    p:=link[p];
    end;
  end;

@ One embarrassing problem remains: The ligature/kern program might be very
long, but the starting addresses in |char_remainder| can be at most~255.
Therefore we need to output some indirect address information; we want to
compute |lk_offset| so that addition of |lk_offset| to all remainders makes
all but |lk_offset| distinct remainders less than~256.

For this we need a sorted table of all relevant remainders.

@<Glob...@>=
@!label_table:array[0..256] of record
  @!rr: -1..@'77777; {sorted label values}
  @!cc: byte; {associated characters}
  end;
@!label_ptr:0..256; {index of highest entry in |label_table|}
@!sort_ptr:0..256; {index into |label_table|}
@!lk_offset:0..256; {smallest offset value that might work}
@!t:0..@'77777; {label value that is being redirected}
@!extra_loc_needed:boolean; {do we need a special word for |bchar|?}

@ @<Compute the ligature/kern program offset@>=
@<Insert all labels into |label_table|@>;
if bchar<256 then
  begin extra_loc_needed:=true; lk_offset:=1;
  end
else begin extra_loc_needed:=false; lk_offset:=0;
  end;
@<Find the minimum |lk_offset| and adjust all remainders@>;
if bchar_label<@'77777 then
  begin lig_kern[nl-1].b2:=(bchar_label+lk_offset)div 256;
  lig_kern[nl-1].b3:=(bchar_label+lk_offset)mod 256;
  end

@ @<Insert all labels...@>=
label_ptr:=0; label_table[0].rr:=-1; {sentinel}
for c:=bc to ec do if char_tag[c]=lig_tag then
  begin sort_ptr:=label_ptr; {there's a hole at position |sort_ptr+1|}
  while label_table[sort_ptr].rr>char_remainder[c] do
    begin label_table[sort_ptr+1]:=label_table[sort_ptr];
    decr(sort_ptr); {move the hole}
    end;
  label_table[sort_ptr+1].cc:=c;
  label_table[sort_ptr+1].rr:=char_remainder[c];
  incr(label_ptr);
  end

@ @<Find the minimum |lk_offset| and adjust all remainders@>=
begin sort_ptr:=label_ptr; {the largest unallocated label}
if label_table[sort_ptr].rr+lk_offset > 255 then
  begin lk_offset:=0; extra_loc_needed:=false; {location 0 can do double duty}
  repeat char_remainder[label_table[sort_ptr].cc]:=lk_offset;
  while label_table[sort_ptr-1].rr=label_table[sort_ptr].rr do
    begin decr(sort_ptr); char_remainder[label_table[sort_ptr].cc]:=lk_offset;
    end;
  incr(lk_offset); decr(sort_ptr);
  until lk_offset+label_table[sort_ptr].rr<256;
    {N.B.: |lk_offset=256| satisfies this when |sort_ptr=0|}
  end;
if lk_offset>0 then while sort_ptr>0 do
  begin char_remainder[label_table[sort_ptr].cc]:=
    char_remainder[label_table[sort_ptr].cc]+lk_offset;
  decr(sort_ptr);
  end;
end

@ @<Output the ligature/kern program@>=
if extra_loc_needed then {|lk_offset=1|}
  begin out(255); out(bchar); out(0); out(0);
  end
else for sort_ptr:=1 to lk_offset do {output the redirection specs}
  begin t:=label_table[label_ptr].rr;
  if bchar<256 then
    begin out(255); out(bchar);
    end
  else begin out(254); out(0);
    end;
  out_size(t+lk_offset);
  repeat decr(label_ptr); until label_table[label_ptr].rr<t;
  end;
if nl>0 then for lig_ptr:=0 to nl-1 do
  begin out(lig_kern[lig_ptr].b0);
  out(lig_kern[lig_ptr].b1);
  out(lig_kern[lig_ptr].b2);
  out(lig_kern[lig_ptr].b3);
  end;
if nk>0 then for krn_ptr:=0 to nk-1 do out_scaled(kern[krn_ptr])

@ @<Output the extensible character recipes@>=
if ne>0 then for c:=0 to ne-1 do
  begin out(exten[c].b0);
  out(exten[c].b1);
  out(exten[c].b2);
  out(exten[c].b3);
  end;

@ For our grand finale, we wind everything up by outputting the parameters.

@<Output the parameters@>=
for par_ptr:=1 to np do
  begin if par_ptr=1 then
    @<Output the slant (|param[1]|) without scaling@>
  else out_scaled(param[par_ptr]);
  end

@ @<Output the slant...@>=
begin if param[1]<0 then
  begin param[1]:=param[1]+@'10000000000;
  out((param[1] div @'100000000)+256-64);
  end
else out(param[1] div @'100000000);
out((param[1] div @'200000) mod 256);
out((param[1] div 256) mod 256);
out(param[1] mod 256);
end

@* The VF output phase.
Output to |vf_file| is considerably simpler.

@d id_byte=202 {current version of \.{VF} format}
@d vout(#)==write(vf_file,#)

@<Glob...@>=
@!vcount:integer; {number of bytes written to |vf_file|}

@ We need a routine to output integers as four bytes. Negative values
will never be less than $-2^{24}$.

@p procedure vout_int(@!x:integer);
begin if x>=0 then vout(x div @'100000000)
else begin vout(255); x:=x+@'100000000;
  end;
vout((x div @'200000) mod 256);
vout((x div @'400) mod 256); vout(x mod 256);
end;

@ @<Do the \.{VF} output@>=
vout(pre); vout(id_byte); vout(vtitle_length);
for k:=0 to vtitle_length-1 do vout(vf[vtitle_start+k]);
for k:=check_sum_loc to design_size_loc+3 do vout(header_bytes[k]);
vcount:=vtitle_length+11;
for cur_font:=0 to font_ptr-1 do @<Output a local font definition@>;
for c:=bc to ec do if char_wd[c]>0 then
  @<Output a packet for character |c|@>;
repeat vout(post); incr(vcount);
until vcount mod 4 = 0

@ @<Output a local font definition@>=
begin vout(fnt_def1); vout(cur_font);@/
vout(font_checksum[cur_font].b0);
vout(font_checksum[cur_font].b1);
vout(font_checksum[cur_font].b2);
vout(font_checksum[cur_font].b3);
vout_int(font_at[cur_font]);
vout_int(font_dsize[cur_font]);
vout(farea_length[cur_font]);
vout(fname_length[cur_font]);
for k:=0 to farea_length[cur_font]-1 do vout(vf[farea_start[cur_font]+k]);
if fname_start[cur_font]=vf_size then
  begin vout("N"); vout("U"); vout("L"); vout("L");
  end
else for k:=0 to fname_length[cur_font]-1 do vout(vf[fname_start[cur_font]+k]);
vcount:=vcount+12+farea_length[cur_font]+fname_length[cur_font];
end

@ @<Output a packet for character |c|@>=
begin x:=memory[char_wd[c]];
if design_units<>unity then x:=round((x/design_units)*1048576.0);
if (packet_length[c]>241)or(x<0)or(x>=@'100000000) then
  begin vout(242); vout_int(packet_length[c]); vout_int(c); vout_int(x);
  vcount:=vcount+13+packet_length[c];
  end
else begin vout(packet_length[c]); vout(c); vout(x div @'200000);
  vout((x div @'400) mod 256); vout(x mod 256);
  vcount:=vcount+5+packet_length[c];
  end;
if packet_start[c]=vf_size then
  begin if c>=128 then vout(set1);
  vout(c);
  end
else for k:=0 to packet_length[c]-1 do vout(vf[packet_start[c]+k]);
end

@* The main program.
The routines sketched out so far need to be packaged into separate procedures,
on some systems, since some \PASCAL\ compilers place a strict limit on the
size of a routine. The packaging is done here in an attempt to avoid some
system-dependent changes.

@p procedure param_enter;
begin @<Enter the parameter names@>;
end;
@#
procedure vpl_enter;
begin @<Enter all the \.{VPL} names@>;
end;
@#
procedure name_enter; {enter all names and their equivalents}
begin @<Enter all the \.{PL} names...@>;
vpl_enter; param_enter;
end;
@#
procedure read_lig_kern;
var @!krn_ptr:0..max_kerns; {an index into |kern|}
@!c:byte; {runs through all character codes}
begin @<Read ligature/kern list@>;
end;
@#
procedure read_char_info;
var @!c:byte; {the char}
begin @<Read character info list@>;
end;
@#
procedure read_input;
var @!c:byte; {header or parameter index}
begin @<Read all the input@>;
end;
@#
procedure corr_and_check;
var @!c:0..256; {runs through all character codes}
@!hh:0..hash_size; {an index into |hash_list|}
@!lig_ptr:0..max_lig_steps; {an index into |lig_kern|}
@!g:byte; {a character generated by the current character |c|}
begin @<Correct and check the information@>
end;
@#
procedure vf_output;
var @!c:byte; {runs through all character codes}
@!cur_font:0..256; {runs through all local fonts}
@!k:integer; {loop index}
begin @<Do the \.{VF} output@>;
end;

@ Here is where \.{VPtoVF} begins and ends.

@p begin initialize;@/
name_enter;@/
read_input; print_ln('.');@/
corr_and_check;@/
@<Do the \.{TFM} output@>;
vf_output;
end.

@* System-dependent changes.
This section should be replaced, if necessary, by changes to the program
that are necessary to make \.{VPtoVF} 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.