man groff_diff (Conventions) - differences between GNU troff and classical troff

NAME

groff_diff - differences between GNU troff and classical troff

DESCRIPTION

This manual page describes the language differences between groff, the GNU roff text processing system and the classical roff formatter of the freely available Unix~7 of the 1970s, documented in the Troff User's Manual by Osanna and Kernighan. This inludes the roff language as well as the intermediate output format (troff output).

The section SEE ALSO gives pointers to both the classical roff and the modern groff documentation.

At the moment, this document is the place of the most actual documentation within the groff system. This might change in the future. Actually, all novelties of the groff language are first described here and will pervade into the other documents only at a later stage.

GROFF LANGUAGE

In this section, all additional features of groff compared to the classical Unix~7 troff are described in detail.

Long names

The names of number registers, fonts, strings/:macros/:diversions, special characters, and colors can be of any length. In escape sequences, additionally to the classical (xx construction for a two character name, you can use [xxx] for a name of arbitrary length, for example in

[rs][xxx]
Print the special character called xxx.
[rs]f[xxx]
Set font xxx. Additionally, [rs]f[] is a new syntax equal to [rs]fP, i.e., to return to the previous font.
[rs]*[xxx arg1 arg2 ...]
Interpolate string xxx, taking arg1, arg2, ... as arguments.
[rs]n[xxx]
Interpolate number register xxx.

Fractional pointsizes

A scaled point is equal to 1/sizescale points, where sizescale is specified in the DESC file (1 by default). There is a new scale indicator z that has the effect of multiplying by sizescale. Requests and escape sequences in troff interpret arguments that represent a pointsize as being in units of scaled points, but they evaluate each such argument using a default scale indicator of z. Arguments treated in this way are the argument to the ps request, the third argument to the cs request, the second and fourth arguments to the tkf request, the argument to the [rs]H escape sequence, and those variants of the [rs]s escape sequence that take a numeric expression as their argument.

For example, suppose sizescale is 1000; then a scaled point will be equivalent to a millipoint; the call .ps 10.25 is equivalent to .ps 10.25z and so sets the pointsize to 10250 scaled points, which is equal to 10.25 points.

The number register [rs]n[.s] returns the pointsize in points as decimal fraction. There is also a new number register [rs]n[.ps] that returns the pointsize in scaled points.

It would make no sense to use the z scale indicator in a numeric expression whose default scale indicator was neither u nor z, and so troff disallows this. Similarly it would make no sense to use a scaling indicator other than z or u in a numeric expression whose default scale indicator was z, and so troff disallows this as well.

There is also new scale indicator~c s which multiplies by the number of units in a scaled point. So, for example, [rs]n[.ps]s is equal to 1m. Be sure not to confuse the s and z scale indicators.

Numeric expressions

Spaces are permitted in a number expression within parentheses.

M indicates a scale of 100ths of an em. f indicates a scale of 65536 units, providing fractions for color definitions with the defcolor request. For example, 0.5f = 32768u.

e1>?e2
The maximum of e1 and e2.
e1<?e2
The minimum of e1 and e2.
(c;e)
Evaluate e using c as the default scaling indicator. If c is missing, ignore scaling indicators in the evaluation of e.

New escape sequences

[rs]A'anything'
This expands to 1 or 0 resp., depending on whether anything is or is not acceptable as the name of a string, macro, diversion, number register, environment, font, or color. It will return~c 0 if anything is empty. This is useful if you want to lookup user input in some sort of associative table.
[rs]B'anything'
This expands to 1 or 0 resp., depending on whether anything is or is not a valid numeric expression. It will return~c 0 if anything is empty.
[rs]C'xxx'
Typeset character named xxx. Normally it is more convenient to use [rs][xxx]R]. But [rs]C has the advantage that it is compatible with recent versions of UNIX and is available in compatibility mode.
[rs]E
This is equivalent to an escape character, but it is not interpreted in copy-mode. For example, strings to start and end superscripting could be defined like this


The use of [rs]E ensures that these definitions will work even if [rs]*{ gets interpreted in copy-mode (for example, by being used in a macro argument).

[rs]Ff
[rs]F(fm [rs]F[fam] Change font family. This is the same as the fam request. [rs]F[] switches back to the previous color (note that [rs]FP won't work; it selects font family `P' instead).
[rs]mx
[rs]m(xx [rs]m[xxx] Set drawing color. [rs]m[] switches back to the previous color.
[rs]Mx
[rs]M(xx [rs]M[xxx] Set background color for filled objects drawn with the [rs]D'...' commands. [rs]M[] switches back to the previous color.
[rs]N'n'
Typeset the character with code n in the current font. n can be any integer. Most devices only have characters with codes between 0 and 255. If the current font does not contain a character with that code, special fonts will not be searched. The [rs]N escape sequence can be conveniently used in conjunction with the char request, for example
The code of each character is given in the fourth column in the font description file after the charset command. It is possible to include unnamed characters in the font description file by using a name of ---; the [rs]N escape sequence is the only way to use these.
[rs]On
[rs]O[n] Suppressing troff output. The escapes [rs]02, [rs]O3, [rs]O4, and [rs]O5 are intended for internal use by %grohtml.
[rs]O0
Disable any ditroff glyphs from being emitted to the device driver, provided that the escape occurs at the outer level (see [rs]O3 and [rs]O4).
[rs]O1
Enable output of glyphs, provided that the escape occurs at the outer level.
[rs]O0 and [rs]O1 also reset the registers [rs]n[opminx], [rs]n[opminy], [rs]n[opmaxx], and [rs]n[opmaxy] to~-1. These four registers mark the top left and bottom right hand corners of a box which encompasses all written glyphs.
[rs]O2
Provided that the escape occurs at the outer level, enable output of glyphs and also write out to stderr the page number and four registers encompassing the glyphs previously written since the last call to [rs]O.
[rs]O3
Begin a nesting level. At start-up, troff is at outer level. This is really an internal mechanism for %grohtml while producing images. They are generated by running the troff source through troff to the postscript device and ghostscript to produce images in PNG format. The [rs]O3 escape will start a new page if the device is not html (to reduce the possibility of images crossing a page boundary).
[rs]O4
End a nesting level.
[rs]O5[Pfilename]
This escape is %grohtml specific. Provided that this escape occurs at the outer nesting level, write filename to stderr. The position of the image, P, must be specified and must be one of l, r, c, or i (left, right, centered, inline). filename will be associated with the production of the next inline image.
[rs]R'name [+-]n'
This has the same effect as
.nr name [+-]n
[rs]s(nn
[rs]s[+-](nn Set the point size to nn points; nn must be exactly two digits.
[rs]s[[+-]n]
[rs]s[+-][n] [rs]s'[+-]n' [rs]s[+-]'n' Set the point size to n scaled points; n is a numeric expression with a default scale indicator of~c z.
[rs]Vx
[rs]V(xx [rs]V[xxx] Interpolate the contents of the environment variable xxx, as returned by getenv(3). [rs]V is interpreted in copy-mode.
[rs]Yx
[rs]Y(xx [rs]Y[xxx] This is approximately equivalent to [rs]X'[rs]*[xxx]'R]. However the contents of the string or macro xxx are not interpreted; also it is permitted for xxx to have been defined as a macro and thus contain newlines (it is not permitted for the argument to [rs]X to contain newlines). The inclusion of newlines requires an extension to the UNIX troff output format, and will confuse drivers that do not know about this extension.
[rs]Z'anything'
Print anything and then restore the horizontal and vertical position; anything may not contain tabs or leaders.
[rs]$0
The name by which the current macro was invoked. The als request can make a macro have more than one name.
[rs]$*
In a macro or string, the concatenation of all the arguments separated by spaces.
[rs]$@
In a macro or string, the concatenation of all the arguments with each surrounded by double quotes, and separated by spaces.
[rs]$(nn
[rs]$[nnn] In a macro or string, this gives the nn-th or nnn-th argument. Macros and strings can have an unlimited number of arguments.
[rs]?anything[rs]?
When used in a diversion, this will transparently embed anything in the diversion. anything is read in copy mode. When the diversion is reread, anything will be interpreted. anything may not contain newlines; use [rs]! if you want to embed newlines in a diversion. The escape sequence [rs]? is also recognised in copy mode and turned into a single internal code; it is this code that terminates anything. Thus

will print~c 4.
[rs]/
This increases the width of the preceding character so that the spacing between that character and the following character will be correct if the following character is a roman character. It is a good idea to use this escape sequence whenever an italic character is immediately followed by a roman character without any intervening space.
[rs],
This modifies the spacing of the following character so that the spacing between that character and the preceding character will correct if the preceding character is a roman character. It is a good idea to use this escape sequence whenever a roman character is immediately followed by an italic character without any intervening space.
[rs])
Like [rs]& except that it behaves like a character declared with the cflags request to be transparent for the purposes of end-of-sentence recognition.
[rs]~
This produces an unbreakable space that stretches like a normal inter-word space when a line is adjusted.
[rs]:
This causes the insertion of a zero-width break point. It is equal to [rs]% within a word but without insertion of a soft hyphen character.
[rs]#
Everything up to and including the next newline is ignored. This is interpreted in copy mode. It is like [rs]" except that [rs]" does not ignore the terminating newline.

New requests

.aln xx yy
Create an alias xx for number register object named yy. The new name and the old name will be exactly equivalent. If yy is undefined, a warning of type reg will be generated, and the request will be ignored.
.als xx yy
Create an alias xx for request, string, macro, or diversion object named yy. The new name and the old name will be exactly equivalent (it is similar to a hard rather than a soft link). If yy is undefined, a warning of type mac will be generated, and the request will be ignored. The de, am, di, da, ds, and as requests only create a new object if the name of the macro, diversion or string diversion is currently undefined or if it is defined to be a request; normally they modify the value of an existing object.
.ami xx yy
Append to macro indirectly. See the dei request below for more information.
.am1 xx yy
Similar to .am, but compatibility mode is switched off during execution. To be more precise, a `compatibility save' token is inserted at the beginning of the macro addition, and a `compatibility restore' token at the end. As a consequence, the requests am, am1, de, and de1 can be intermixed freely since the compatibility save/:restore tokens only affect the macro parts defined by .am1 and .ds1.
.asciify xx
This request `unformats' the diversion xx in such a way that ASCII and space characters (and some escape sequences) that were formatted and diverted into xx will be treated like ordinary input characters when xx is reread. Useful for diversions in conjunction with the .writem request. It can be also used for gross hacks; for example, this

will set register~c n to~1. Note that glyph information (font, font size, etc.) is not preserved; use .unformat instead.
.as1 xx yy
Similar to .as, but compatibility mode is switched off during expansion. To be more precise, a `compatibility save' token is inserted at the beginning of the string, and a `compatibility restore' token at the end. As a consequence, the requests as, as1, ds, and ds1 can be intermixed freely since the compatibility save/:restore tokens only affect the (sub)strings defined by as1 and ds1.
.backtrace
Print a backtrace of the input stack on stderr.
.blm xx
Set the blank line macro to xx. If there is a blank line macro, it will be invoked when a blank line is encountered instead of the usual troff behaviour.
.box xx
.boxa xx These requests are similar to the di and da requests with the exception that a partially filled line will not become part of the diversion (i.e., the diversion always starts with a new line) but restored after ending the diversion, discarding the partially filled line which possibly comes from the diversion.
.break
Break out of a while loop. See also the while and continue requests. Be sure not to confuse this with the br request.
.brp
This is the same as [rs]p.
.cflags n c1 c2...
Characters c1, c2,... have properties determined by n, which is ORed from the following:
1
The character ends sentences (initially characters .?! have this property).
2
Lines can be broken before the character (initially no characters have this property); a line will not be broken at a character with this property unless the characters on each side both have non-zero hyphenation codes.
4
Lines can be broken after the character (initially characters -[rs](hy[rs](em have this property); a line will not be broken at a character with this property unless the characters on each side both have non-zero hyphenation codes.
8
The character overlaps horizontally (initially characters [rs](ul[rs](rn[rs](ru have this property).
16
The character overlaps vertically (initially character [rs](br has this property).
32
An end-of-sentence character followed by any number of characters with this property will be treated as the end of a sentence if followed by a newline or two spaces; in other words the character is transparent for the purposes of end-of-sentence recognition; this is the same as having a zero space factor in [tx] (initially characters [dq]')]*[rs](dg[rs](rq have this property).
.char c string
Define character c to be string. Every time character c needs to be printed, string will be processed in a temporary environment and the result will be wrapped up into a single object. Compatibility mode will be turned off and the escape character will be set to [rs] while string is being processed. Any emboldening, constant spacing or track kerning will be applied to this object rather than to individual characters in string.
A character defined by this request can be used just like a normal character provided by the output device. In particular other characters can be translated to it with the tr request; it can be made the leader character by the lc request; repeated patterns can be drawn with the character using the [rs]l and [rs]L escape sequences; words containing the character can be hyphenated correctly, if the hcode request is used to give the character a hyphenation code.
There is a special anti-recursion feature: use of character within the character's definition will be handled like normal characters not defined with char.
A character definition can be removed with the rchar request.
.chop xx
Chop the last character off macro, string, or diversion xx. This is useful for removing the newline from the end of diversions that are to be interpolated as strings.
.close stream
Close the stream named stream; stream will no longer be an acceptable argument to the write request. See the open request.
.continue
Finish the current iteration of a while loop. See also the while and break requests.
.color n
If n is non-zero or missing, enable colors (this is the default), otherwise disable them.
.cp n
If n is non-zero or missing, enable compatibility mode, otherwise disable it. In compatibility mode, long names are not recognised, and the incompatibilities caused by long names do not arise.
.defcolor xxx scheme color_components
Define color. scheme can be one of the following values: rgb (three components), cym (three components), cmyk (four components), and gray or grey (one component). Color components can be given either as a hexadecimal string or as positive decimal integers in the range 0-65535. A hexadecimal string contains all color components concatenated; it must start with either # or ##. The former specifies hex values in the range 0-255 (which are internally multiplied by~257), the latter in the range 0-65535. Examples: #FFC0CB (pink), ##ffff0000ffff (magenta). A new scaling indicator~c f has been introduced which multiplies its value by~65536; this makes it convenient to specify color components as fractions in the range 0 to~1. Example:


Note that f is the default scaling indicator for the defcolor request, thus the above statement is equivalent to


The color named default (which is device-specific) can't be redefined. It is possible that the default color for and is not the same.
.dei xx yy
Define macro indirectly. The following example

is equivalent to


.de1 xx yy
Similar to .de, but compatibility mode is switched off during execution. On entry, the current compatibility mode is saved and restored at exit.
.do xxx
Interpret .xxx with compatibility mode disabled. For example,


would have the same effect as



except that it would work even if compatibility mode had been enabled. Note that the previous compatibility mode is restored before any files sourced by xxx are interpreted.

.ds1 xx yy
Similar to .ds, but compatibility mode is switched off during expansion. To be more precise, a `compatibility save' token is inserted at the beginning of the string, and a `compatibility restore' token at the end.
.ecs
Save current escape character.
.ecr
Restore escape character saved with ecs. Without a previous call to ecs, `[rs]' will be the new escape character.
.evc xx
Copy the contents of environment xx to the current environment. No pushing or popping of environments will be done.
.fam xx
Set the current font family to xx. The current font family is part of the current environment. If xx is missing, switch back to previous font family. The value at start-up is `T'. See the description of the sty request for more information on font families.
.fchar c string
Define fallback character c to be string. The syntax of this request is the same as the char request; the only difference is that a character defined with char hides the glyph with the same name in the current font, whereas a character defined with fchar is checked only if the particular glyph isn't found in the current font. This test happens before checking special fonts.
.fspecial f s1 s2...
When the current font is f, fonts s1, s2,... will be special, that is, they will searched for characters not in the current font. Any fonts specified in the special request will be searched after fonts specified in the fspecial request.
.ftr f g
Translate font f to g. Whenever a font named f is referred to in an [rs]f escape sequence, or in the ft, ul, bd, cs, tkf, special, fspecial, fp, or sty requests, font g will be used. If g is missing, or equal to f then font f will not be translated.
.hcode c1 code1 c2 code2...
Set the hyphenation code of character c1 to code1 and that of c2 to code2. A hyphenation code must be a single input character (not a special character) other than a digit or a space. Initially each lower-case letter %a-z has a hyphenation code, which is itself, and each upper-case letter %A-Z has a hyphenation code which is the lower-case version of itself. See also the hpf request.
.hla lang
Set the current hyphenation language to lang. Hyphenation exceptions specified with the hw request and hyphenation patterns specified with the hpf request are both associated with the current hyphenation language. The hla request is usually invoked by the troffrc file.
.hlm n
Set the maximum number of consecutive hyphenated lines to~c n. If n is negative, there is no maximum. The default value is~-1. This value is associated with the current environment. Only lines output from an environment count towards the maximum associated with that environment. Hyphens resulting from [rs]% are counted; explicit hyphens are not.
.hpf file
Read hyphenation patterns from file; this will be searched for in the same way that name.tmac is searched for when the -mname option is specified. It should have the same format as (simple) [tx] patterns files. More specifically, the following scanning rules are implemented.
[bu]
A percent sign starts a comment (up to the end of the line) even if preceded by a backslash.
[bu]
No support for `digraphs' like [rs]$.
[bu]
^^xx (x is 0-9 or a-f) and ^^x (character code of~c x in the range 0-127) are recognized; other use of ^ causes an error.
[bu]
No macro expansion.
[bu]
hpf checks for the expression [rs]patterns{...} (possibly with whitespace before and after the braces). Everything between the braces is taken as hyphenation patterns. Consequently, { and } are not allowed in patterns.
[bu]
Similarly, [rs]hyphenation{...} gives a list of hyphenation exceptions.
[bu]
[rs]endinput is recognized also.
[bu]
For backwards compatibility, if [rs]patterns is missing, the whole file is treated as a list of hyphenation patterns (only recognizing the %~c character as the start of a comment).
Use the hpfcode request to map the encoding used in hyphenation patterns files to groff's input encoding.
The set of hyphenation patterns is associated with the current language set by the hla request. The hpf request is usually invoked by the troffrc file; a second call replaces the old patterns with the new ones.
.hpfa file
The same as hpf except that the hyphenation patterns from file are appended to the patterns already loaded in the current language.
.hpfcode a b c d ...
After reading a hyphenation patterns file with the hpf or hpfa request, convert all characters with character code~c a in the recently read patterns to character code~c b, character code~c c to~c d, etc. Initially, all character codes map to themselves. The arguments of hpfcode must be integers in the range 0 to~255. Note that it is even possible to use character codes which are invalid in groff otherwise.
.hym n
Set the hyphenation margin to~c n: when the current adjustment mode is not~c b, the line will not be hyphenated if the line is no more than n short. The default hyphenation margin is~0. The default scaling indicator for this request is~c m. The hyphenation margin is associated with the current environment. The current hyphenation margin is available in the [rs]n[.hym] register.
.hys n
Set the hyphenation space to~c n: when the current adjustment mode is~c b don't hyphenate the line if the line can be justified by adding no more than n extra space to each word space. The default hyphenation space is~0. The default scaling indicator for this request is~c m. The hyphenation space is associated with the current environment. The current hyphenation space is available in the [rs]n[.hys] register.
.itc n macro
Variant of .it for which a line interrupted with [rs]c counts as one input line.
.kern n
If n is non-zero or missing, enable pairwise kerning, otherwise disable it.
.length xx string
Compute the length of string and return it in the number register xx (which is not necessarily defined before).
.linetabs n
If n is non-zero or missing, enable line-tabs mode, otherwise disable it (which is the default). In line-tabs mode, tab distances are computed relative to the (current) output line. Otherwise they are taken relative to the input line. For example, the following

yields
a b c
In line-tabs mode, the same code gives
a b c
Line-tabs mode is associated with the current environment; the read-only number register \[rs]n[.linetabs] is set to~1 if in line-tabs mode, and 0 otherwise.
.mso file
The same as the so request except that file is searched for in the same directories as macro files for the the -m command line option. If the file name to be included has the form name.tmac and it isn't found, mso tries to include tmac.name instead and vice versa.
.nop anything
Execute anything. This is similar to `.if 1'.
.nroff
Make the n built-in condition true and the t built-in condition false. This can be reversed using the troff request.
.open stream filename
Open filename for writing and associate the stream named stream with it. See also the close and write requests.
.opena stream filename
Like open, but if filename exists, append to it instead of truncating it.
.output string
Emit string directly to the intermediate output (subject to copy-mode interpretation); this is similar to [rs]! used at the top level. An initial double quote in string is stripped off to allow initial blanks.
.pnr
Print the names and contents of all currently defined number registers on stderr.
.psbb filename
Get the bounding box of a PostScript image filename. This file must conform to Adobe's Document Structuring Conventions; the command looks for a %%%BoundingBox comment to extract the bounding box values. After a successful call, the coordinates (in PostScript units) of the lower left and upper right corner can be found in the registers [rs]n[llx], [rs]n[lly], [rs]n[urx], and [rs]n[ury], respectively. If some error has occurred, the four registers are set to zero.
.pso command
This behaves like the so request except that input comes from the standard output of command.
.ptr
Print the names and positions of all traps (not including input line traps and diversion traps) on stderr. Empty slots in the page trap list are printed as well, because they can affect the priority of subsequently planted traps.
.pvs [+-]n
Set the post-vertical line space to n; default scale indicator is~c p. This value will be added to each line after it has been output. With no argument, the post-vertical line space is set to its previous value.
The total vertical line spacing consists of four components: .vs and [rs]x with a negative value which are applied before the line is output, and .pvs and [rs]x with a positive value which are applied after the line is output.
.rchar c1 c2...
Remove the definitions of characters c1, c2,... This undoes the effect of a char request.
.return
Within a macro, return immediately. No effect otherwise.
.rj
.rj~n Right justify the next n~c input lines. Without an argument right justify the next input line. The number of lines to be right justified is available in the [rs]n[.rj] register. This implicitly does .ce~0. The ce request implicitly does .rj~0.
.rnn xx yy
Rename number register xx to yy.
.shc c
Set the soft hyphen character to c. If c is omitted, the soft hyphen character will be set to the default [rs](hy. The soft hyphen character is the character which will be inserted when a word is hyphenated at a line break. If the soft hyphen character does not exist in the font of the character immediately preceding a potential break point, then the line will not be broken at that point. Neither definitions (specified with the char request) nor translations (specified with the tr request) are considered when finding the soft hyphen character.
.shift n
In a macro, shift the arguments by n positions: argument~c i becomes argument i-n; arguments 1 to~c n will no longer be available. If n is missing, arguments will be shifted by~1. Shifting by negative amounts is currently undefined.
.sizes s1 s2...sn [0]
This command is similar to the sizes command of a DESC file. It sets the available font sizes for the current font to s1, s2,...,~sn scaled points. The list of sizes can be terminated by an optional~c 0. Each si can also be a range of sizes m-n. Contrary to the font file command, the list can't extend over more than a single line.
.special s1 s2...
Fonts s1, s2, are special and will be searched for characters not in the current font.
.spreadwarn limit
Make troff emit a warning if the additional space inserted for each space between words in an output line is larger or equal to limit. A negative value is changed to zero; no argument toggles the warning on and off without changing limit. The default scaling indicator is~c m. At startup, spreadwarn is deactivated, and limit is set to 3m. For example, .spreadwarn 0.2m will cause a warning if troff must add 0.2m or more for each interword space in a line. This request is active only if text is justified to both margins (using .ad b).
.sty n f
Associate style~c f with font position~c n. A font position can be associated either with a font or with a style. The current font is the index of a font position and so is also either a font or a style. When it is a style, the font that is actually used is the font the name of which is the concatenation of the name of the current family and the name of the current style. For example, if the current font is~1 and font position~1 is associated with style~c R and the current font family is~c T, then font TR will be used. If the current font is not a style, then the current family is ignored. When the requests cs, bd, tkf, uf, or fspecial are applied to a style, then they will instead be applied to the member of the current family corresponding to that style. The default family can be set with the -f option. The styles command in the DESC file controls which font positions (if any) are initially associated with styles rather than fonts.
.substring xx n1 [n2]
Replace the string named xx with the substring defined by the indices n1 and n2. The first character in the string has index~0. If n2 is omitted, it is taken to be equal to the string's length. If the index value n1 or n2 is negative, it will be counted from the end of the string, going backwards: The last character has index~-1, the character before the last character has index~-2, etc.
.tkf f s1 n1 s2 n2
Enable track kerning for font f. When the current font is f the width of every character will be increased by an amount between n1 and n2; when the current point size is less than or equal to s1 the width will be increased by n1; when it is greater than or equal to s2 the width will be increased by n2; when the point size is greater than or equal to s1 and less than or equal to s2 the increase in width is a linear function of the point size.
.tm1 string
Similar to the tm request, string is read in copy mode and written on the standard error, but an initial double quote in string is stripped off to allow initial blanks.
.tmc string
Similar to tm1 but without writing a final newline.
.trf filename
Transparently output the contents of file filename. Each line is output as if preceded by [rs]!; however, the lines are not subject to copy-mode interpretation. If the file does not end with a newline, then a newline will be added. For example, you can define a macro~c x containing the contents of file~c f, using

Unlike with the cf request, the file cannot contain characters such as NUL that are not legal troff input characters.
.trin abcd
This is the same as the tr request except that the asciify request will use the character code (if any) before the character translation. Example:

The result is x a. Using tr, the result would be x x.
.trnt abcd
This is the same as the tr request except that the translations do not apply to text that is transparently throughput into a diversion with [rs]!. For example,

will print~c b; if trnt is used instead of tr it will print~c a.
.troff
Make the n built-in condition false, and the t built-in condition true. This undoes the effect of the nroff request.
.unformat xx
This request `unformats' the diversion xx. Contrary to the .asciify request, which tries to convert formatted elements of the diversion back to input tokens as much as possible, .unformat will only handle tabs and spaces between words (usually caused by spaces or newlines in the input) specially. The former are treated as if they were input tokens, and the latter are stretchable again. Note that the vertical size of lines is not preserved. Glyph information (font, font size, space width, etc.) is retained. Useful in conjunction with the .box and .boxa requests.
.vpt n
Enable vertical position traps if n is non-zero, disable them otherwise. Vertical position traps are traps set by the wh or dt requests. Traps set by the it request are not vertical position traps. The parameter that controls whether vertical position traps are enabled is global. Initially vertical position traps are enabled.
.warn n
Control warnings. n is the sum of the numbers associated with each warning that is to be enabled; all other warnings will be disabled. The number associated with each warning is listed in troff(1). For example, .warn~0 will disable all warnings, and .warn~1 will disable all warnings except that about missing characters. If n is not given, all warnings will be enabled.
.warnscale si
Set the scaling indicator used in warnings to si. Valid values for si are u, i, c, p, and P. At startup, it is set to~c i.
.while c anything
While condition~c c is true, accept anything as input; c~c can be any condition acceptable to an if request; anything can comprise multiple lines if the first line starts with [rs]{ and the last line ends with [rs]}. See also the break and continue requests.
.write stream anything
Write anything to the stream named stream. stream must previously have been the subject of an open request. anything is read in copy mode; a leading~c [dq] will be stripped.
.writec stream anything
Similar to write but without writing a final newline.
.writem stream xx
Write the contents of the macro or string xx to the stream named stream. stream must previously have been the subject of an open request. xx is read in copy mode.

Extended requests

.cf filename
When used in a diversion, this will embed in the diversion an object which, when reread, will cause the contents of filename to be transparently copied through to the output. In UNIX troff, the contents of filename is immediately copied through to the output regardless of whether there is a current diversion; this behaviour is so anomalous that it must be considered a bug.
.ev xx
If xx is not a number, this will switch to a named environment called xx. The environment should be popped with a matching ev request without any arguments, just as for numbered environments. There is no limit on the number of named environments; they will be created the first time that they are referenced.
.ss m n
When two arguments are given to the ss request, the second argument gives the sentence space size. If the second argument is not given, the sentence space size will be the same as the word space size. Like the word space size, the sentence space is in units of one twelfth of the spacewidth parameter for the current font. Initially both the word space size and the sentence space size are~12. Contrary to UNIX troff, GNU troff handles this request in nroff mode also; a given value is then rounded down to the nearest multiple of~12. The sentence space size is used in two circumstances. If the end of a sentence occurs at the end of a line in fill mode, then both an inter-word space and a sentence space will be added; if two spaces follow the end of a sentence in the middle of a line, then the second space will be a sentence space. Note that the behaviour of UNIX troff will be exactly that exhibited by GNU troff if a second argument is never given to the ss request. In GNU troff, as in UNIX troff, you should always follow a sentence with either a newline or two spaces.
.ta n1 n2...nn T r1 r2...rn
Set tabs at positions n1, n2,..., nn and then set tabs at nn+r1, nn+r2,..., nn+rn and then at nn+rn+r1, nn+rn+r2,..., nn+rn+rn, and so on. For example,


will set tabs every half an inch.

New number registers

The following read-only registers are available:

[rs]n[.C]
1~if compatibility mode is in effect, 0~otherwise.
[rs]n[.cdp]
The depth of the last character added to the current environment. It is positive if the character extends below the baseline.
[rs]n[.ce]
The number of lines remaining to be centered, as set by the ce request.
[rs]n[.cht]
The height of the last character added to the current environment. It is positive if the character extends above the baseline.
[rs]n[.color]
1~if colors are enabled, 0~otherwise.
[rs]n[.csk]
The skew of the last character added to the current environment. The skew of a character is how far to the right of the center of a character the center of an accent over that character should be placed.
[rs]n[.ev]
The name or number of the current environment. This is a string-valued register.
[rs]n[.fam]
The current font family. This is a string-valued register.
[rs]n[.fn]
The current (internal) real font name. This is a string-valued register. If the current font is a style, the value of [rs]n[.fn] is the proper concatenation of family and style name.
[rs]n[.fp]
The number of the next free font position.
[rs]n[.g]
Always~1. Macros should use this to determine whether they are running under GNU troff.
[rs]n[.hla]
The current hyphenation language as set by the hla request.
[rs]n[.hlc]
The number of immediately preceding consecutive hyphenated lines.
[rs]n[.hlm]
The maximum allowed number of consecutive hyphenated lines, as set by the hlm request.
[rs]n[.hy]
The current hyphenation flags (as set by the hy request).
[rs]n[.hym]
The current hyphenation margin (as set by the hym request).
[rs]n[.hys]
The current hyphenation space (as set by the hys request).
[rs]n[.in]
The indent that applies to the current output line.
[rs]n[.int]
Set to a positive value if last output line is interrupted (i.e., if it contains [rs]c).
[rs]n[.kern]
1~if pairwise kerning is enabled, 0~otherwise.
[rs]n[.lg]
The current ligature mode (as set by the lg request).
[rs]n[.linetabs]
The current line-tabs mode (as set by the linetabs request).
[rs]n[.ll]
The line length that applies to the current output line.
[rs]n[.lt]
The title length as set by the lt request.
[rs]n[.ne]
The amount of space that was needed in the last ne request that caused a trap to be sprung. Useful in conjunction with the [rs]n[.trunc] register.
[rs]n[.ns]
1~if no-space mode is active, 0~otherwise.
[rs]n[.pn]
The number of the next page, either the value set by a pn request, or the number of the current page plus~1.
[rs]n[.ps]
The current pointsize in scaled points.
[rs]n[.psr]
The last-requested pointsize in scaled points.
[rs]n[.pvs]
The current post-vertical line space as set with the pvs request.
[rs]n[.rj]
The number of lines to be right-justified as set by the rj request.
[rs]n[.sr]
The last requested pointsize in points as a decimal fraction. This is a string-valued register.
[rs]n[.ss]
[rs]n[.sss] These give the values of the parameters set by the first and second arguments of the ss request.
[rs]n[.tabs]
A string representation of the current tab settings suitable for use as an argument to the ta request.
[rs]n[.trunc]
The amount of vertical space truncated by the most recently sprung vertical position trap, or, if the trap was sprung by a ne request, minus the amount of vertical motion produced by the ne request. In other words, at the point a trap is sprung, it represents the difference of what the vertical position would have been but for the trap, and what the vertical position actually is. Useful in conjunction with the [rs]n[.ne] register.
[rs]n[.vpt]
1~if vertical position traps are enabled, 0~otherwise.
[rs]n[.warn]
The sum of the numbers associated with each of the currently enabled warnings. The number associated with each warning is listed in troff(1).
[rs]n[.x]
The major version number. For example, if the version number is 1.03, then [rs]n[.x] will contain~1.
[rs]n[.y]
The minor version number. For example, if the version number is 1.03, then [rs]n[.y] will contain~03.
[rs]n[.Y]
The revision number of groff.
[rs]n[llx]
[rs]n[lly] [rs]n[urx] [rs]n[ury] These four registers are set by the .psbb request and contain the bounding box values (in PostScript units) of a given PostScript image.

The following read/write registers are set by the [rs]w escape sequence:

[rs]n[rst]
[rs]n[rsb] Like the st and sb registers, but take account of the heights and depths of characters.
[rs]n[ssc]
The amount of horizontal space (possibly negative) that should be added to the last character before a subscript.
[rs]n[skw]
How far to right of the center of the last character in the [rs]w argument, the center of an accent from a roman font should be placed over that character.

Other available read/write number registers are:

[rs]n[c.]
The current input line number. [rs]n[.c] is a read-only alias to this register.
[rs]n[hours]
The number of hours past midnight. Initialized at start-up.
[rs]n[hp]
The current horizontal position at input line.
[rs]n[minutes]
The number of minutes after the hour. Initialized at start-up.
[rs]n[seconds]
The number of seconds after the minute. Initialized at start-up.
[rs]n[systat]
The return value of the system() function executed by the last sy request.
[rs]n[slimit]
If greater than~0, the maximum number of objects on the input stack. If less than or equal to~0, there is no limit on the number of objects on the input stack. With no limit, recursion can continue until virtual memory is exhausted.
[rs]n[year]
The current year. Note that the traditional troff number register [rs]n[yr] is the current year minus 1900.

Miscellaneous

troff predefines a single (read/write) string-based register, [rs]*(.T, which contains the argument given to the -T command line option, namely the current output device (for example, latin1 or ascii). Note that this is not the same as the (read-only) number register [rs]n[.T] which is defined to be~1 if troff is called with the -T command line option, and zero otherwise. This behaviour is different to UNIX troff.

Fonts not listed in the DESC file are automatically mounted on the next available font position when they are referenced. If a font is to be mounted explicitly with the fp request on an unused font position, it should be mounted on the first unused font position, which can be found in the [rs]n[.fp] register; although troff does not enforce this strictly, it will not allow a font to be mounted at a position whose number is much greater than that of any currently used position.

Interpolating a string does not hide existing macro arguments. Thus in a macro, a more efficient way of doing

.xx [rs][rs]$@

is

[rs][rs]*[xx][rs][rs]

If the font description file contains pairwise kerning information, characters from that font will be kerned. Kerning between two characters can be inhibited by placing a [rs]& between them.

In a string comparison in a condition, characters that appear at different input levels to the first delimiter character will not be recognised as the second or third delimiters. This applies also to the tl request. In a [rs]w escape sequence, a character that appears at a different input level to the starting delimiter character will not be recognised as the closing delimiter character. The same is true for [rs]A, [rs]b, [rs]B, [rs]C, [rs]l, [rs]L, [rs]o, [rs]X, and [rs]Z. When decoding a macro or string argument that is delimited by double quotes, a character that appears at a different input level to the starting delimiter character will not be recognised as the closing delimiter character. The implementation of [rs]$@ ensures that the double quotes surrounding an argument will appear the same input level, which will be different to the input level of the argument itself. In a long escape name ] will not be recognized as a closing delimiter except when it occurs at the same input level as the opening ]. In compatibility mode, no attention is paid to the input-level.

There are some new types of condition:

.if rxxx
True if there is a number register named xxx.
.if dxxx
True if there is a string, macro, diversion, or request named xxx.
.if mxxx
True if there is a color named xxx.
.if cch
True if there is a character ch available; ch is either an ASCII character or a special character [rs](xx or [rs][xxx]R]; the condition will also be true if ch has been defined by the char request.

The tr request can now map characters onto [rs]~.

It is now possible to have whitespace between the first and second dot (or the name of the ending macro) to end a macro definition. Example:


INTERMEDIATE OUTPUT FORMAT

This section describes the format output by GNU troff. The output format used by GNU troff is very similar to that used by Unix device-independent troff. Only the differences are documented here.

Units

The argument to the s command is in scaled points (units of points/n, where n is the argument to the sizescale command in the DESC file). The argument to the x Height command is also in scaled points.

Text Commands

Nn
Print character with index~c n (a non-negative integer) of the current font.

If the tcommand line is present in the DESC file, troff will use the following two commands.

txxx
xxx is any sequence of characters terminated by a space or a newline; the first character should be printed at the current position, the current horizontal position should be increased by the width of the first character, and so on for each character. The width of the character is that given in the font file, appropriately scaled for the current point size, and rounded so that it is a multiple of the horizontal resolution. Special characters cannot be printed using this command.
un xxx
This is same as the t command except that after printing each character, the current horizontal position is increased by the sum of the width of that character and n.

Note that single characters can have the eighth bit set, as can the names of fonts and special characters.

The names of characters and fonts can be of arbitrary length; drivers should not assume that they will be only two characters long.

When a character is to be printed, that character will always be in the current font. Unlike device-independent troff, it is not necessary for drivers to search special fonts to find a character.

For color support, some new commands have been added:

Set the color components of the current drawing color, using various color schemes. md resets the drawing color to the default value. The arguments are integers in the range 0 to 65536.

The x device control command has been extended.

If n is~1, start underlining of spaces. If n is~0, stop underlining of spaces. This is needed for the cu request in nroff mode and is ignored otherwise.

Drawing Commands

The D drawing command has been extended. These extensions will not be used by GNU pic if the -n option is given.

Set the shade of gray to be used for filling solid objects to n; n must be an integer between 0 and 1000, where 0 corresponds solid white and 1000 to solid black, and values in between correspond to intermediate shades of gray. This applies only to solid circles, solid ellipses and solid polygons. By default, a level of 1000 will be used. Whatever color a solid object has, it should completely obscure everything beneath it. A value greater than 1000 or less than 0 can also be used: this means fill with the shade of gray that is currently being used for lines and text. Normally this will be black, but some drivers may provide a way of changing this.
Draw a solid circle with a diameter of d with the leftmost point at the current position.
Draw a solid ellipse with a horizontal diameter of dx and a vertical diameter of dy with the leftmost point at the current position. delim $$
Draw a polygon with, for $i = 1 ,..., n+1$, the i-th vertex at the current position $+ sum from j=1 to i-1 ( dx sub j , dy sub j )$. At the moment, GNU pic only uses this command to generate triangles and rectangles.
Like Dp but draw a solid rather than outlined polygon.
Set the current line thickness to n machine units. Traditionally Unix troff drivers use a line thickness proportional to the current point size; drivers should continue to do this if no Dt command has been given, or if a Dt command has been given with a negative value of n. A zero value of n selects the smallest available line thickness.

A difficulty arises in how the current position should be changed after the execution of these commands. This is not of great importance since the code generated by GNU pic does not depend on this. Given a drawing command of the form

B][rs]D[fm]I]cR] $x sub 1$ $y sub 1$ $x sub 2$ $y sub 2$ $...$ $x sub n$ $y sub n$[fm]

where c is not one of c, e, l, a, or ~, Unix troff will treat each of the $x sub i$ as a horizontal quantity, and each of the $y sub i$ as a vertical quantity and will assume that the width of the drawn object is $sum from i=1 to n x sub i$, and that the height is $sum from i=1 to n y sub i$. (The assumption about the height can be seen by examining the st and sb registers after using such a D command in a [rs]w escape sequence). This rule also holds for all the original drawing commands with the exception of De. For the sake of compatibility GNU troff also follows this rule, even though it produces an ugly result in the case of the Dt, and, to a lesser extent, DE commands. Thus after executing a D command of the form

B]DI]cR] $x sub 1$ $y sub 1$ $x sub 2$ $y sub 2$ $...$ c $x sub n$ $y sub n$[rs]n

the current position should be increased by $( sum from i=1 to n x sub i , sum from i=1 to n y sub i )$.

Another set of extensions is

Set the color components of the filling color similar to the m commands above.

Note that Df is now mapped onto DFg. The current position isn't changed by those colour commands.

Device Control Commands

There is a continuation convention which permits the argument to the x X command to contain newlines: when outputting the argument to the x X command, GNU troff will follow each newline in the argument with a + character (as usual, it will terminate the entire argument with a newline); thus if the line after the line containing the x X command starts with +, then the newline ending the line containing the x X command should be treated as part of the argument to the x X command, the + should be ignored, and the part of the line following the + should be treated like the part of the line following the x X command.

The first three output commands are guaranteed to be:

x T device

x res n h v

x init

INCOMPATIBILITIES

In spite of the many extensions, groff has retained compatibility to classical troff to a large degree. For the cases where the extensions lead to collisions, a special compatibility mode with the restricted, old functionality was created for groff.

Groff Language

groff provides a compatibility mode that allows to process roff code written for classical or for other implementations of roff in a consistent way.

Compatibility mode can be turned on with the command line option, and turned on or off with the request. The number register is~1 if compatibility mode is on, 0~otherwise.

This became necessary because the GNU concept for long names causes some incompatibilities. Classical troff interprets

as defining a string ab with contents cd. In groff mode, this will be considered as a call of a macro named

Also classical troff interprets or as references to a string or number register called while groff takes this as the start of a long name.

In compatibility mode, groff interprets these things in the traditional way; so long names are not recognized.

On the other hand, groff in GNU native mode does not allow to use the single-character escapes (backslash), (vertical bar), (caret), (ampersand), (opening brace), (closing brace), (space), (single quote), (backquote), (minus), (underline), (bang), (percent), and (character c) in names of strings, macros, diversions, number registers, fonts or environments, whereas classical troff does.

The escape sequence can be helpful in avoiding these escape sequences in names.

Fractional pointsizes cause one noteworthy incompatibility. In classical troff, the request ignores scale indicators and so

.ps~10u

will set the pointsize to 10~points, whereas in groff native mode the pointsize will be set to 10~scaled points.

In groff mode, there is a fundamental difference between unformatted input characters, and formatted output characters. Everything that affects how an output character will be output is stored with the character; once an output character has been constructed it is unaffected by any subsequent requests that are executed, including the or requests.

Normally output characters are constructed from input characters at the moment immediately before the character is added to the current output line. Macros, diversions and strings are all, in fact, the same type of object; they contain lists of input characters and output characters in any combination.

An output character does not behave like an input character for the purposes of macro processing; it does not inherit any of the special properties that the input character from which it was constructed might have had. The following example will make things clearer.


In GNU mode this will be printed as So each pair of input backslashes is turned into a single output backslash and the resulting output backslashes are not interpreted as escape characters when they are reread.

Classical troff would interpret them as escape characters when they were reread and would end up printing a single backslash

In GNU, the correct way to get a printable version of the backslash character is the escape sequence, but classical troff does not provide a clean feature for getting a non-syntactical backslash. A close method is the printable version of the current escape character using the escape sequence; this works if the current escape character is not redefined. It works in both GNU mode and compatibility mode, while dirty tricks like specifying a sequence of multiple backslashes do not work reliably; for the different handling in diversions, macro definitions, or text mode quickly leads to a confusion about the necessary number of backslashes.

To store an escape sequence in a diversion that will be interpreted when the diversion is reread, either the traditional transparent output facility or the new escape sequence can be used.

Intermediate Output

The groff intermediate output format is in a state of evolution. So far it has some incompatibilities, but it is intended to establish a full compatibility to the classical troff output format. Actually the following incompatibilities exist: The positioning after the drawing of the polygons conflicts with the classical definition. The intermediate output cannot be rescaled to other devices as classical "device-independent" troff did.

AUTHORS

Copyright (C) 1989, 2001, 2002 Free Software Foundation, Inc.

This document is distributed under the terms of the FDL (GNU Free Documentation License) version 1.1 or later. You should have received a copy of the FDL on your system, it is also available on-line at the This document was written by James Clark, with modifications by and

This document is part of groff, the GNU roff distribution. Formerly, the contents of this document was kept in the manual page troff(1). Only the parts dealing with the language aspects of the different roff systems were carried over into this document. The troff command line options and warnings are still documented in troff(1).

SEE ALSO

The groff info file, cf. info(1) presents all groff documentation within a single document.

groff(1)
A list of all documentation around groff.
groff(7)
A description of the groff language, including a short, but complete reference of all predefined requests, registers, and escapes of plain groff. From the command line, this is called using
roff(7)
A survey of roff systems, including pointers to further historical documentation.
[CSTR~#54/]
The Nroff/:Troff User's Manual by J. F. Osanna of 1976 in the revision of Brian Kernighan of 1992, being the "classical troff documentation" .