tmac.ps: do not rename the EPS macro

[neatroff_make.git] / demo / neatroff.ms

.\" PSTITLE: Neatroff Introduction
.so neat__.ms
.ds en.cl "#237
.ps.info "Neatroff Introduction" "Ali Gholami Rudi"
.de MH
.       sp
.       LP
.       ne 2
\m[#237]\fI\\$1\fP\m[]
.       IP
..
.de NR
.       sp
.       LP
.       ne 2
.       ta 4.5i 6iR
.       nf
.       lg 0
\f(CB\m[#237]\\$1\m0\fP \s-3\m[#237]\fB\\$2\fP  \fB\\$4\m0\fP
.       lg 1
.       IP
..
.HD
.TL
\s+8Neatroff\m0\s-8
.AU
\fIAli Gholami Rudi\fP
.ce
\s-3\fIUpdated in October 2017\s+3\fP\&
.sp 3
Neatroff is a new implementation of Troff typesetting system in C
programming language, which tries to address, as neatly as possible,
some of the shortcomings of the original Troff based on the ideas and
features available in Plan 9 Troff, Heirloom Troff, and Groff.
Neatroff, its postscript postprocessor, neatpost, and its eqn
preprocessor, neateqn, are available at \*[ps.url http://litcave.rudi.ir/].
This document enumerates Neatroff's features, its new requests, and
its differences compared to other Troff implementations.  On the other
hand, the document \(lqGetting Started with Neatroff,\(rq available at
http://litcave.rudi.ir/\:neatstart.pdf, explains how to set up and
use Neatroff.

.bp
.SH "Noteworthy Features"
The following list describes the main extensions of Neatroff compared
to the original Troff (many of these extensions are available in
Groff and Heirloom Troff as well).  The number register \&.neat,
which is always one in Neatroff, can be used to distinguish Neatroff
from other Troff implementations.

.MH "UTF-8 encoding
In Neatroff, input files and characters, glyph names, ligatures,
hyphenation patterns and dictionary, as well as quoted escape sequence
delimiters and arguments of commands like .tr, .hc, .tc, .lc, .mc, and
\&.fc are in UTF-8 encoding.

.MH "Long macro, register, and environment names
When not in compatibility mode (activated with -C command line option or the .cp
request, as in Groff), Neatroff supports long macro, register, and environment
names.  It also supports Groff-style escape sequences with long arguments
(for \\[], \\*[], \\$[], \\f[], \\g[], \\k[], \\m[], \\n[], and \\s[])
and interpolating string registers with arguments (\\*[xyz arg1 arg2 ...]).
Note that like Groff, Neatroff supports named environments and
is not limited to original Troff's three fixed environments.

.MH "Advanced font features, ligature, and pairwise kerning
Neatroff and neatmkfn (which generates Neatroff's font descriptions)
support many of the advanced font features available in modern fonts.
In a font, a set of substitution and positioning rules may be
specified, which are grouped into several features and scripts.  In
Neatroff, features can be enabled with \&.ff and the active script can
be selected with \&.ffsc.  Old-style ligatures (\&.lg) and pairwise
kerning (\&.kn) are still supported, as described in a later section.

.MH "Whole paragraph text formatting
Neatroff supports filling whole paragraphs at once, to achieve
more uniform word spacing.  Like Heirloom Troff, the \&.ad request
accepts arguments p or pb, pl, and pr, which are equivalent to b, l,
and r, except that the filling is done for whole paragraphs, i.e.,
words are collected until a line break is issued.
This inevitably changes the behaviour of some requests and traps: several
lines may be collected and ready to be output while executing them.
For the end macro, Troff invokes the macro specified
with \&.em request without flushing the last incomplete line.
Neatroff follows the same behaviour even when formatting whole
paragraphs and does not write any of the collected lines to the output.
Since after the end macro no new page is started, collected lines may
be unexpectedly written to the end of the last page of the document.
To change that, the end macro can invoke the \&.br request.
For requests that cause break, using \(aq as the control character
prevents writing any line of the collected paragraph to the output,
as expected.
The exception to this rule is \(aqbr, which formats the words
collected so far and outputs all resulting lines except the final
incomplete line (this is useful, for instance, for footnotes, which
should be inserted in the same page).

.MH "Paragraph formatting algorithm
For deciding at what points to break a paragraph into lines, Neatroff
assigns a cost to each possible outcome: a cost of 100 is assigned to
each stretchable space that has to be stretched 100 percent.  The cost
grows quadratically and the cost of stretching a space 200 percent is 400.
There are requests that adjust the algorithm Neatroff uses for
performing paragraph formatting.
The \&.hycost request changes the cost of hyphenating words.  The
default value is zero.
The \\j escape sequence, as in Heirloom Troff, specifies the extra
cost of line break after a word; for instance, in \(lqHello\\j'10000'
world\(rq, the words are not split by the line breaking algorithm,
unless absolutely necessary (i.e., if other options are more costly).
The escape sequence \\\(ti introduces non-breakable stretchable space.
Also, to prevent paragraphs with
very short last lines, the \&.pmll (paragraph minimum line length)
sets the minimum length of a formatted line, specified as a
percentage of \\n(.l; \(lq.pmll 15\(rq, for instance, makes sure that
the length of last line of each paragraph is at least 15% of its other
lines; otherwise, a cost proportional to the value specified
as its second argument is added.

.MH "Controlling word spaces
The \&.ssh request sets the amount (in percentage) by which the
stretchable spaces in a line may be shrunk while formatting lines.
The default value is zero.
Also, the second
argument of \&.ss request specifies sentence space, as in Groff or
Heirloom Troff.

.MH "Text direction
The dir branch of Neatroff supports text direction to render
right-to-left languages.  The \&.<< and \&.>> requests specify text
direction and the \\< and \\> escape sequences change it temporarily
for including words in the reverse direction.  The value of number
registers \&.td and \&.cd indicate the current text and temporary
directions respectively; zero means left-to-right and one means
right-to-left.

.MH "Keshideh justification and cursive scripts
In the dir branch of Neatroff, a new adjustment type (.ad k) allows
inserting Keshideh characters before justifying text with hyphenation
and spaces.  The dir branch also supports cursive scripts, which
require connecting glyphs at their cursive attachment positions,
as defined in the fonts.

.MH "Font manipulation
In Neatroff, the mapping between Troff character names and glyphs in a
font can be modified with \&.fmap request: \(lq.fmap F C G\(rq
maps Troff character C to the glyph with device-specific name G for
font F.  When this glyph does not exist in F, Neatroff assumes that
the character C is not present in the font.  When G is missing, the
effect of \&.fmap for character C is cancelled.  Neatroff also implements
Groff's \&.fspecial and \&.fzoom requests: after \(lq.fspecial FN S1
S2 ...\(rq, when the current font is FN, the fonts S1, S2, ... are
assumed to be special.  Also, \(lq.fzoom FN zoom\(rq scales font FN by
the second argument after dividing it by 1000.

.MH "Colour support
Neatroff supports colours with .cl request and \\m[] escape sequence.
Unlike Groff, colours need not be defined beforehand and can be
specified directly.  The argument of \\m can be predefined colour
names (e.g. blue), predefined colour numbers (0 for black, 1 for red,
2 for green, 3 for yellow, 4 for blue, 5 for magenta, 6 for cyan, and
7 for white), #rgb and #rrggbb for specifying colours in hexadecimal
RGB format, #g and #gg for specifying grey with the given hexadecimal
level, and empty (\\m[]) for the previous colour.  The current colour
is available in \&.m number register.

.MH "Hyphenation language
The \&.hpf request loads hyphenation patterns, exceptions, and
character mappings from the addresses specified via its arguments.
The specified files should contain nothing but utf-8 patterns,
exceptions and mappings respectively (i.e. no TeX code), just like the
files whose names end with \&.pat.txt, \&.hyp.txt and \&.chr.txt in
ftp:/\h'-.3n'/ftp.ctan.org/\:tex-archive/\:language/\:hyph-utf8/\:tex/\:generic/\:hyph-utf8/\:patterns/\:txt/.
The \&.hpfa request is like \&.hpf, except that it does not clear the
previous hyphenation patterns and exceptions.  The second and third
arguments of these requests are optional.  With no arguments, these
requests load English hyphenation patterns and exceptions.  Also the
\&\(lq.hcode abcd...\(rq request, assigns the hyphenation code of b to
a and the hyphenation code of d to c; initially all upper-case ASCII
letters are mapped to their lower-case forms.

.MH "Filled drawing objects
Neatroff supports Groff-style polygons and filled drawing objects (p,
C, E and P commands for \\D escape sequence).  In Neatroff, however,
there is no specific background colour; objects are filled with the
current colour (.m number register).  Furthermore, in Neatroff
the edges of polygons can be lines, arcs, or splines; a letter among the
arguments of \\D'p ..' specifies the type of the subsequent
edges: \(oql\(cq, \(oqa\(cq, and \(oq\(ti\(cq for lines, arcs,
and splines respectively.

.MH "Conditional escape sequence
Neatroff supports a new escape sequence for conditional interpolation:
the escape sequence \\?'cond@expr1@expr2@', evaluates cond (exactly as
if it is a \&.if condition) and interpolates expr1, if the condition
is true, and expr2, otherwise.  The delimiter (@ in this example) can
be any character that cannot be part of the condition; for numerical
expressions, for instance, it cannot be a digit, an operator sign, or
a scale indicator, unless separated from the condition with \\&.  The
final delimiter, and even expr2, may be omitted, thus \\?'cond@expr'
is valid; Neatroff interpolates expr if cond is true.  Note that this
escape sequence is not interpolated in copy mode.

.MH "Neatpost-specific device commands
Neatroff supports \\X'rotate deg' for rotating the current page
around the current point.  Also \\X'eps file [width [height]]'
includes the given EPS file with its lower left corner at the current
point.  If width and height are given (in basic units), the file is
scaled appropriately.

.bp
.SH "Summary of New Requests"
This is the list of new request available in Neatroff compared
to those documented in \(lqTroff User's Manual\(rq by Ossanna
and Kernighan.

.NR "\&.cl C" "0" "previous" "E"
Change text colour.  The current colour is available in the number
register \\n(.m.  With no arguments, the previous colour is selected.
The format of the argument and the \\m escape sequence are described
in the previous section.

.NR "\&.ssh N" "0" "0" "E"
Set the amount stretchable spaces in formatted lines may be
shrunk in percentage (available through \\n[.ssh]).

.NR "\&.ad p*" "b" "adjust" "E"
With values pl, pr, pb, and p, this request instructs Neatroff to perform
whole-paragraph line formatting.  Also in dir branch of Neatroff, the
value k enables Keshideh justification (kp is the equivalent for
whole-paragraph formatting).

.NR "\&.eos S T" "S=\&.?!  T='"")]*" "none" "\-"
Specify sentence characters.  The first argument specifies
the characters that end a sentence and the second argument
specifies the characters ignored after them.

.NR "\&.ss M N" "M=12 N=12" "none" "E"
The second argument sets sentence space size (available in \\n[.sss]).

.NR "\&.fzoom F N" "1000" "none" "\-"
Magnify the given font by N/1000.

.NR "\&.fp N F L" "\-" "none" "\-"
In Neatroff, if instead of the position of the font to be mounted, N
is a dash, the position of the font is decided automatically: if a
font with the same name is already mounted, the same position is
reused.  Otherwise the font is mounted on the next available
position.

.NR "\&.ff F +F1 -F2" "\-" "none" "\-"
Enable or disable font features; the first argument specifies the
font and the rest of the arguments specify feature names,
prefixed with a plus to enable or a minus to disable.
When a feature is enabled, all substitution and positioning rules
of a font with that feature name are applied when laying out the
glyphs.

.NR "\&.ffsc F SC" "\-" "none" "\-"
Specify font's script.  A font description specifies a set of
rules for each script, grouped into several features.  With this
request, only the rules for the specified script are enabled.
By default, or when SC is missing, all scripts are selected.

.NR "\&.kn N" "1" "none" "E"
Enable or disable pairwise kerning (current value available through \\n[.kn]).

.NR "\&.fspecial F S1 S2" "\-" "none" "\-"
Set special fonts when the current font is F.

.NR "\&.fmap FN CH GID" "\-" "none" "\-"
Map Troff character CH to glyph with device dependent name GID for
font FN.  When gid is missing, the effect of mapping CH is cancelled.
Neatroff implicitly calls \&.fmap for all aliases in font descriptions
(character definitions whose second column is ").

.NR "\&.tkf FN S1 N1 S2 N2" "\-" "none" "\-"
Enable track kerning for font FN.  If the point size is at most
S1, the width of each character is increased by N1 points, if it
is at least S2, the width of each character is increased by N2
points, and if it is between S1 and S2, the width of each character
is increased by a value between N1 and N2, relative to the
difference between the current point size and S1.

.NR "\&.pmll N C" "0" "0" "E"
Set paragraph minimum line length in percentage.  To shorter lines,
Neatroff assigns a cost proportional to the value specified as the
second argument (or 100, if missing) when formatting paragraphs.
Number registers \\n[.pmll] and \\n[.pmllcost] store the
values passed to \&.pmll.

.NR "\&.hycost N N2 N3" "0" "none" "E"
Change the cost of hyphenating words when adjusting lines.
An argument of 100 assigns to each hyphenation the cost of stretching
a space one hundred percent while formatting.  The second and third
arguments specify additional costs for two and three consecutive
hyphenated lines (only when formatting whole paragraphs).

.NR "\&.hlm n" "0" "0" "E"
Set the maximum number of consecutive hyphenated lines (only when
formatting whole paragraphs).  The current value is available via
\\n[.hlm].  An argument of zero or a negative number implies no
limitation.

.NR "\&.hydash C" "\\\\:\\\\(hy\\\\(en\\\\(em-\\\\-\\\\(--" "none" "\-"
Specify the list of characters after which words may be broken (even
when hyphenation is disabled) without inserting hyphens.

.NR "\&.hystop C" "\\\\%" "none" "\-"
Specify hyphenation inhibiting characters.  Words containing any of
the given characters are not hyphenated, unless after dashes
(characters specified via \&.hydash) or hyphenation indicators (\\%).

.NR "\&.hpf H P C" "\-" "English" "\-"
Set hyphenation files for patterns, exceptions, and mappings.
With no arguments, loads English hyphenation patterns and exceptions.

.NR "\&.hpfa H P C" "\-" "English" "\-"
Like, \&.hpf, but do not clear the previous hyphenation patterns.

.NR "\&.hcode abcd..." "\-" "none" "\-"
Assign the hyphenation code of b to a and the hyphenation code of d to c.

.NR "\&.chop xx" "\-" "none" "\-"
Remove the last character of a string register.

.NR "\&.>>  \&.<<" "left-to-right" "\-" "E"
Render text in left-to-right or right-to-left direction (available
only in dir branch of Neatroff).  See the first section for
an explanation of escape sequences \\> and \\<.

.NR "\&.in2" "0" "previous" "E"
Right-side indentation (available only in dir branch of Neatroff).
The current right-side indentation is available in register \\n(.I.

.NR "\&.ti2" "0" "\-" "E"
Right-side temporary indentation (available only in dir branch of Neatroff).

.NR "\&.char C DEF" "\-" "\-" "\-"
Define Troff character C.
If DEF is missing, previous definitions for character C are removed.

.NR "\&.ochar FN C DEF" "\-" "\-" "\-"
Define Troff character C only for font FN.
If DEF is missing, previous definitions
for character C are removed.

.NR "\&.blm" "\-" "\-" "\-"
Specify the blank line macro.  If specified, each blank line is
treated as an invocation of this macro.

.bp
.SH "Notes"
.sp -1
.MH "Generating the output device
The neatmkfn program (available at \*[ps.url http://litcave.rudi.ir/]) generates
Neatroff font descriptions for AFM or TrueType fonts (OpenType fonts
are converted to TrueType first).  It includes a script to create
a complete output device for Neatroff.

.MH "Formatting equations with neateqn
Neateqn is an eqn preprocessor for Neatroff.  It implements many
of the extensions introduced in Groff's eqn preprocessor.  It
can use TeX's Computer Modern-style bracket-building symbols, if
available.

.MH "The standard macro packages
The standard Troff macro packages and a top-level build script to
obtain and install Neatroff are available in neatroff_make git
repository (available at \*[ps.url http://litcave.rudi.ir/]).

.MH "Missing requests
A few requests of the original Troff are not implemented:
\&.pi, .cf, .rd, .pm, .ul, .cu, .uf, \\H, and \\S.

.MH "Porting and distribution
Given that Neatroff, neatpost and neateqn can be compiled with Neatcc,
porting them to other Unix variants besides Linux should not be
difficult.  Note that Neatroff is released under the ISC licence.

.MH "List of OpenType font features
As mentioned in previous sections, font features can be enabled and
disabled with \&.ff request.  For the list of OpenType features in
general and their descriptions, see
https://en.wikipedia.org/\:wiki/\:List_of_typographic_features or
http://www.microsoft.com/\:typography/\:\s-1OTSPEC\s+1/\:featurelist.htm.

.bp
.SH "Font Description Files"
The format of font description files in Neatroff, although still mostly
backward compatible, has been slightly changed.  The value of special,
spacewidth, and ligatures parameters retain their old meanings; sizes
and name parameters are ignored, however.  The value of the fontname
parameter in Neatroff specifies the device name of the font
(e.g. Times-Roman); neatpost uses it to map Troff fonts to postscript fonts.
In the charset section, the forth field is always the
device-specific name of the glyph (accessible with \\N escape sequence)
and the optional fifth field specifies glyph's code (the fourth field
of the original Troff).

In addition to the old charset section of the original Troff, Neatroff
supports a new syntax for defining characters and kerning pairs.  Lines
starting with the word \(lqchar\(rq define characters (similar to lines in
the charset section) and lines starting with \(lqkern\(rq specify kerning pairs.
For the latter, \(lqkern\(rq is followed by three tokens: the name of the
first glyph, the name of the second glyph, and the amount of kerning
between them.  Specifying the name of glyphs (the fourth field after \(lqchar\(rq)
instead of character names allows specifying kerning pairs for glyphs
not mapped to any characters (may be later with \&.fmap request) or
specifying kerning pairs only once for all aliases of a character.
Here are a few lines of a font description file for Neatroff, created
with neatmkfn.

.cc.beg
.ta 5 10 15 20 25
name R
fontname Times-Roman
spacewidth 25
ligatures fi fl 0
# the list of characters
char    !       33      2       exclam  33
char    .       25      0       period  46
char    A       72      2       A       65
char    B       67      2       B       66
char    C       67      2       C       67
# the kerning pairs
kern    A       C       -5
kern    A       period  -1
.cc.end
.LP
The width column of the character definition lines can optionally
include four more numbers, separated with commas, that describe the
bounding boxes of the glyphs.  The bounding boxes are used in the \\w
escape sequence; after this escape sequence, the value of the bbllx,
bblly, bburx and bbury number registers are modified to represent the
bounding box of the argument of \\w.

To use the advanced features present in TrueType and OpenType fonts,
Neatroff supports lines that define substitution and positioning
rules (lines starting with \(lqgsub\(rq and \(lqgpos\(rq respectively).
Note that unlike Heirloom Troff, which implements non-contextual
single-character substitutions, Neatroff implements many of the more
complex OpenType substitution and positioning features.  The following
example shows how such features are defined in Neatroff font
descriptions:

.cc.beg
.ta 5 10 15 20 25
gsub liga:latn 4 -gl1 -gl2 -gl3 +gl123
gpos kern:latn 2 gl1:+0+0-5+0 gl2
.cc.end
.LP
In this example, the first line defines a 3-character ligature (with
feature name \(lqliga\(rq and script name \(lqlatn\(rq) and the second
defines pairwise kerning for the pair of glyphs gl1 and gl2
(decreasing the horizontal advance of gl1 by 5 basic units; with
feature name \(lqkern\(rq and script name \(lqlatn\(rq).  The patterns
can be longer and more detailed, defining context or glyph groups, to
support OpenType features that require them; for examples, see the
files generated by neatmkfn.

.bp
.SH "Source Code Organization
The following figure shows where Neatroff's major
layers and features are implemented in its source tree.
.sp -1

.KS
.RT
.PS
# part(title, description)
define part { [
M:      box ht 0.5 wid 2.3
.ps 15
        $1 at 1/2 <M.w, M.nw> + (.2, 0) ljust
.ps 9
        $2 at 1/2 <M.w, M.sw> + (.2, 0) ljust
.ps 20
] }
# part2(title, description)
define part2 { [
M:      box ht 0.3 wid 2.6
.ps 13
        $1 ljust at M.w + (.2, 0)
.ps 9
        $2 ljust at M + (-.4, 0)
.ps 20
] }
.ps 20
lineht = .3
HEAD:   "\s(17\fI\fP\s0"
        down
        move .3
IN:     part("in.c", "Input handling")
        arrow
CP:     part("cp.c", "Copy-mode interpretation")
        arrow
TR:     part("tr.c", "Troff request/macro execution")
        arrow
REN:    part("ren.c", "Rendering, traps, and diversions")
        arrow
OUT:    part("out.c", "Generating Troff output")
        "\s(17\fI\fP\s0" at HEAD + (3, 0)
        move .3
REG:    part2("reg.c", "Registers and environments")
        move down 0.2
FMT:    part2("wb.c", "Word buffer")
        move same
        part2("eval.c", "Integer expression evaluation")
        move same
        part2("fmt.c", "Line formatting")
        move same
DEV:    part2("dev.c", "Output device")
        move same
FONT:   part2("font.c", "Fonts")
        move same
        part2("hyph.c", "Tex hyphenation")
        move same
        part2("dir.c", "Text direction")
.PE
.KE