Autres langues

Langue: en

Version: 12 March 2009 (debian - 07/07/09)

Autres sections - même nom

Section: 7 (Divers)


groff - a short reference for the GNU roff language


The name groff stands for GNU roff and is the free implementation of the roff type-setting system. See roff(7) for a survey and the background of the groff system.

This document gives only short descriptions of the predefined roff language elements as used in groff. Both the classical features and the groff extensions are provided.

Historically, the roff language was called troff. groff is compatible with the classical system and provides proper extensions. So in GNU, the terms roff, troff, and groff language could be used as synonyms. However troff slightly tends to refer more to the classical aspects, whereas groff emphasizes the GNU extensions, and roff is the general term for the language.

This file is only a short version of the complete documentation that is found in the groff info(1) file, which contains more detailed, actual, and concise information.

The general syntax for writing groff documents is relatively easy, but writing extensions to the roff language can be a bit harder.

The roff language is line-oriented. There are only two kinds of lines, control lines and text lines. The control lines start with a control character, by default a period Text " " or a single quote Text " " all other lines are text lines.

Control lines represent commands, optionally with arguments. They have the following syntax. The leading control character can be followed by a command name; arguments, if any, are separated by blanks from the command name and among themselves, for example,

For indentation, any number of space or tab characters can be inserted between the leading control character and the command name, but the control character must be on the first position of the line.

Text lines represent the parts that will be printed. They can be modified by escape sequences, which are recognized by a leading backslash Text " " These are in-line or even in-word formatting elements or functions. Some of these take arguments separated by single quotes Text " " others are regulated by a length encoding introduced by an open parenthesis Text " " or enclosed in brackets Text " " and Text " "

The roff language provides flexible instruments for writing language extension, such as macros. When interpreting macro definitions, the roff system enters a special operating mode, called the copy mode.

The copy mode behavior can be quite tricky, but there are some rules that ensure a safe usage.

Printable backslashes must be denoted as \e. To be more precise, \e represents the current escape character. To get a backslash glyph, use \(rs or \[rs].
Double all backslashes.
Begin all text lines with the special non-spacing character \&.

This does not produce the most efficient code, but it should work as a first measure. For better strategies, see the groff info file and groff_tmac(5).

Reading roff source files is easier, just reduce all double backslashes to a single one in all macro definitions.


The roff language elements add formatting information to a text file. The fundamental elements are predefined commands and variables that make roff a full-blown programming language.

There are two kinds of roff commands, possibly with arguments. Requests are written on a line of their own starting with a dot Text " " or a Text " " whereas Escape sequences are in-line functions and in-word formatting elements starting with a backslash Text " "

The user can define her own formatting commands using the de request. These commands are called macros, but they are used exactly like requests. Macro packages are pre-defined sets of macros written in the groff language. A user's possibilities to create escape sequences herself is very limited, only special characters can be mapped.

The groff language provides several kinds of variables with different interfaces. There are pre-defined variables, but the user can define her own variables as well.

String variables store character sequences. They are set with the ds request and retrieved by the \* escape sequences. Strings can have variables.

Register variables can store numerical values, numbers with a scale unit, and occasionally string-like objects. They are set with the nr request and retrieved by the \n escape sequences.

Environments allow the user to temporarily store global formatting parameters like line length, font size, etc. for later reuse. This is done by the ev request.

Fonts are identified either by a name or by an internal number. The current font is chosen by the ft request or by the \f escape sequences. Each device has special fonts, but the following fonts are available for all devices. R is the standard font Roman. B is its bold counterpart. The italic font is called I and is available everywhere, but on text devices it is displayed as an underlined Roman font. For the graphical output devices, there exist constant-width pendants of these fonts, CR, CI, and CB. On text devices, all characters have a constant width anyway.

Moreover, there are some advanced roff elements. A diversion stores information into a macro for later usage. A trap is a positional condition like a certain number of lines from page top or in a diversion or in the input. Some action can be prescribed to be run automatically when the condition is met.

More detailed information and examples can be found in the groff info file.


There is a small set of characters that have a special controlling task in certain conditions.
Text " "
A dot is only special at the beginning of a line or after the condition in the requests if , ie , el , and while . There it is the control character that introduces a request (or macro). The special behavior can be delayed by using the \. escape. By using the cc request, the control character can be set to a different character, making the dot Text " " a non-special character.
In all other positions, it just means a dot character. In text paragraphs, it is advantageous to start each sentence at a line of its own.
Text " "
The single quote has two controlling tasks. At the beginning of a line and in the conditional requests it is the non-breaking control character. That means that it introduces a request like the dot, but with the additional property that this request doesn't cause a linebreak. By using the c2 request, the non-break control character can be set to a different character.
As a second task, it is the most commonly used argument separator in some functional escape sequences (but any pair of characters not part of the argument will work). In all other positions, it denotes the single quote or apostrophe character. Groff provides a printable representation with the \(cq escape sequence.
Text " "
The double quote is used to enclose arguments in requests, macros, and strings. In the ds and as requests, a leading double quote in the argument will be stripped off, making everything else afterwards the string to be defined (enabling leading whitespace). The escaped double quote \" introduces a comment. Otherwise, it is not special. Groff provides a printable representation with the \(dq escape sequence.
Text " "
The backslash usually introduces an escape sequence (this can be changed with the ec request). A printed version of the escape character is the \e escape; a backslash glyph can be obtained by \(rs.
Text " "
The open parenthesis is only special in escape sequences when introducing an escape name or argument consisting of exactly two characters. In groff, this behavior can be replaced by the [] construct.
Text " "
The opening bracket is only special in groff escape sequences; there it is used to introduce a long escape name or long escape argument. Otherwise, it is non-special, e.g. in macro calls.
Text " "
The closing bracket is only special in groff escape sequences; there it terminates a long escape name or long escape argument. Otherwise, it is non-special.
Space characters are only functional characters. They separate the arguments in requests, macros, and strings, and the words in text lines. They are subject to groff's horizontal spacing calculations. To get a defined space width, escape sequences like Text " " (this is the escape character followed by a space), \|, \^, or \h should be used.
In text paragraphs, newlines mostly behave like space characters. Continuation lines can be specified by an escaped newline, i.e., by specifying a backslash Text " " as the last character of a line.
If a tab character occurs during text the interpreter makes a horizontal jump to the next pre-defined tab position. There is a sophisticated interface for handling tab positions.


A numerical value is a signed or unsigned integer or float with or without an appended scaling indicator. A scaling indicator is a one-character abbreviation for a unit of measurement. A number followed by a scaling indicator signifies a size value. By default, numerical values do not have a scaling indicator, i.e., they are normal numbers.

The roff language defines the following scaling indicators.

Pica = 1/6 inch
Point = 1/72 inch
Em = the font size in points (width of letter `m')
100th of an Em
En = Em/2
Basic unit for actual output device
Vertical line space in basic units scaled point = 1/sizescale of a point (defined in font DESC file)
Scale by 65536.

Numerical expressions are combinations of the numerical values defined above with the following arithmetical operators already defined in classical troff.

Less than
Greater than
Less or equal
Greater or equal
Logical and
Logical or
Logical not
Grouping of expressions
Close current grouping

Moreover, groff added the following operators for numerical expressions:

Text " " The maximum of e1 and e2.
Text " " The minimum of e1 and e2.
Text " " Evaluate e using c as the default scaling indicator.

For details see the groff info file.


Conditions occur in tests raised by the if , ie , and the while requests. The following table characterizes the different types of conditions.
A numerical expression N yields true if its value is greater than~0.
True if the value of I is~0.
True if string~s1 is identical to string~s2.
True if string~s1 is not identical to string~s2.
True if there is a character~ch available.
True if there is a string, macro, diversion, or request called name.
Current page number is even.
Current page number is odd.
True if there is a color called name.
Formatter is nroff.
True if there is a register named reg.
Formatter is troff.


This section provides a short reference for the predefined requests. In groff, request and macro names can be arbitrarily long. No bracketing or marking of long names is needed.

Most requests take one or more arguments. The arguments are separated by space characters (no tabs!); there is no inherent limit for their length or number. An argument can be enclosed by a pair of double quotes. This is very handy if an argument contains space characters, e.g., "arg with space" denotes a single argument.

Some requests have optional arguments with a different behaviour. Not all of these details are outlined here. Refer to the groff info file and groff_diff(7) for all details.

In the following request specifications, most argument names were chosen to be descriptive. Only the following denotations need clarification.

denotes a single character.
a font either specified as a font name or a font number.
all characters up to the end of the line or within \{ and \}.
is a numerical expression that evaluates to an integer value.
is an arbitrary numerical expression, signed or unsigned.
has three meanings depending on its sign, described below.

If an expression defined as ±N starts with a Text " " sign the resulting value of the expression will be added to an already existing value inherent to the related request, e.g. adding to a number register. If the expression starts with a Text " " the value of the expression will be subtracted from the request value.

Without a sign, N replaces the existing value directly. To assign a negative number either prepend~0 or enclose the negative number in parentheses.

Request Short Reference

Empty line, ignored. Useful for structuring documents.
Complete line is a comment.
Print string on standard error, exit program.
Begin line adjustment for output lines in current adjust mode.
Start line adjustment in mode c (c=l,r,b,n).
Assign format c to register (c=l,i,I,a,A).
Create alias name for register.
Create alias name for request, string, macro, or diversion object.
Append to macro until .. is encountered.
Append to macro until .end is called.
Append to a macro whose name is contained in the string register macro until .. is encountered.
Append to a macro indirectly. macro and end are string registers whose contents are interpolated for the macro name and the end macro, respectively.
Same as .am but with compatibility mode switched off during macro expansion.
Same as .am but with compatibility mode switched off during macro expansion.
Append anything to stringvar.
Unformat ASCII characters, spaces, and some escape sequences in diversion.
Same as .as but with compatibility mode switched off during string expansion.
Print a backtrace of the input on stderr.
Embolden font by N-1 units.
Embolden Special Font S when current font is font.
Unset the blank line macro.
Set the blank line macro to macro.
End current diversion.
Divert to macro, omitting a partially filled line.
End current diversion.
Divert and append to macro, omitting a partially filled line.
Eject current page and begin new page.
Eject current page; next page number ±N.
Line break.
Break and spread output line. Same as \p.
Break out of a while loop.
Reset no-break control character to Text " "
Set no-break control character to c.
Reset control character to Text " "
Set control character to c.
Center the next input line.
Center following N input lines.
Copy contents of file filename unprocessed to stdout or to the diversion.
Treat characters c1, c2, ... according to mode number.
Change trap location to N .
Define character c as string anything.
Chop the last character off macro, string, or diversion object.
Close the stream.
Enable colors.
If N is zero disable colors, otherwise enable them.
Finish the current iteration of a while loop.
Enable compatibility mode.
If N is zero disable compatibility mode, otherwise enable it.
Set constant character width mode for font to N/36 ems with em M.
Continuous underline in nroff, like .ul in troff.
End current diversion.
Divert and append to macro.
Define or redefine macro until .. is encountered.
Define or redefine macro until .end is called.
Same as .de but with compatibility mode switched off during macro expansion.
Same as .de but with compatibility mode switched off during macro expansion.
Define or redefine a color with name color. scheme can be rgb, cym, cymk, gray, or grey. component can be single components specified as fractions in the range 0 to 1 (default scaling indicator~ as a string of two-digit hexadecimal color components with a leading #, or as a string of four-digit hexadecimal components with two leading #. The color default can't be redefined.
Define or redefine a macro whose name is contained in the string register macro until .. is encountered.
Define or redefine a macro indirectly. macro and end are string registers whose contents are interpolated for the macro name and the end macro, respectively.
End current diversion.
Divert to macro .
Interpret .name with compatibility mode disabled.
Set stringvar to anything.
Same as .ds but with compatibility mode switched off during string expansion.
Set diversion trap to position N (default scaling indicator~
Reset escape character to Text " "
Set escape character to c.
Restore escape character saved with .ecs .
Save current escape character.
Else part for if-else ( ie ) request.
The macro will be run after the end of input.
Turn off escape character mechanism.
Switch to previous environment.
Push down environment number or name env and switch to it.
Copy the contents of environment env to the current environment. No pushing or popping.
Exit from roff processing.
Return to previous font family.
Set the current font family to name.
Disable field mechanism.
Set field delimiter to a and pad character to space.
Set field delimiter to a and pad character to b.
Define fallback character c as string anything.
Fill output lines.
Flush output buffer.
Mount font on position n.
Mount font with long external name to short internal name on position n.
When the current font is font, then the fonts s1, s2, ... will be special.
Return to previous font. Same as \f[] or \fP .
Change to font name or number font; same as escape sequence.
Translate font1 to font2.
Remove additional hyphenation indicator character.
Set up additional hyphenation indicator character~c.
Set the hyphenation code of character c1 to code1, that of c2 to code2, etc.
Set the current hyphenation language to lang.
Set the maximum number of consecutive hyphenated lines to n.
Read hyphenation patterns from file.
Append hyphenation patterns from file.
Set input mapping for .hpf .
List of words with exceptional hyphenation.
Switch to hyphenation mode N.
Set the hyphenation margin to n (default scaling indicator~
Set the hyphenation space to n.
If cond then anything else goto .el .
If cond then anything; otherwise do nothing.
Ignore text until .. is encountered.
Ignore text until .end .
Change to previous indent value.
Change indent according to ±N (default scaling indicator~
Set an input-line count trap for the next N lines.
Same as .it but count lines interrupted with \c as one line.
Enable pairwise kerning.
If n is zero, disable pairwise kerning, otherwise enable it.
Remove leader repetition character.
Set leader repetition character to~c.
Write the length of the string anything in register.
Enable line-tabs mode (i.e., calculate tab positions relative to output line).
If n is zero, disable line-tabs mode, otherwise enable it.
Set input line number to N and filename to file.
Ligature mode on if N>0.
Change to previous line length.
Set line length according to ±N (default size default scaling indicator~
Change to the previous value of additional intra-line skip.
Set additional intra-line skip value to N, i.e., N-1 blank lines are inserted after each text output line.
Length of title (default scaling indicator~
Margin character off.
Print character c after each text line at actual distance from right margin.
Set margin character to c and distance to N from right margin (default scaling indicator~
Mark current vertical position in register.
The same as the .so request except that file is searched in the tmac directories.
No output-line adjusting.
Need a one-line vertical space.
Need N vertical space (default scaling indicator~
No filling or adjusting of output-lines.
No hyphenation.
Number mode off.
In line number mode, set number, multiple, spacing, and indent.
Do not number next line.
Do not number next N lines.
Always execute anything.
Define or modify register using ±N with auto-increment M.
Make the built-in condition n true and t false.
Turn no-space mode on.
Immediately jump to end of current file.
Next file.
Open filename for writing and associate the stream named stream with it.
Like .open but append to it.
Output vertical distance that was saved by the sv request.
Emit string directly to intermediate output, allowing leading whitespace if string starts with Text " " (which will be stripped off).
Reset page number character to~ Text " "
Page number character.
Pipe output to program (nroff only).
Set page length to default The current page length is stored in .p .
Change page length to ±N (default scaling indicator~
Print macro names and sizes (number of blocks of 128 bytes).
Print only total of sizes of macros (number of 128 bytes blocks).
Next page number N.
Print the names and contents of all currently defined number registers on stderr.
Change to previous page offset. The current page offset is available in .o .
Page offset N.
Return to previous point-size.
Point size; same as
Get the bounding box of a PostScript image filename.
This behaves like the so request except that input comes from the standard output of command.
Print the names and positions of all traps (not including input line traps and diversion traps) on stderr.
Change to previous post-vertical line spacing.
Change post-vertical line spacing according to ±N (default scaling indicator~
Remove the definitions of characters c1, c2, ...
Read insertion.
Return from a macro.
Right justify the next n input lines.
Remove request, macro, or string name.
Rename request, macro, or string old to new.
Rename register reg1 to reg2.
Remove register.
Restore spacing; turn no-space mode off.
Return (upward only) to marked vertical place (default scaling indicator~
Reset soft hyphen character to \(hy.
Set the soft hyphen character to c.
In a macro, shift the arguments by n~positions.
Set available font sizes similar to the sizes command in a DESC file.
Include source file.
Skip one line vertically.
Space vertical distance N up or down according to sign of N (default scaling indicator~
Fonts s1, s2, etc. are special and will be searched for characters not in the current font.
Toggle the spread warning on and off without changing its value.
Emit a warning if each space in an output line is widened by limit or more (default scaling indicator~
Space-character size set to N/12 of the spacewidth in the current font.
Space-character size set to N/12 and sentence space size set to M/12 of the spacewidth in the current font (=1/3 em).
Associate style with font position n.
Replace the string named xx with the substring defined by the indices n1 and n2.
Save of vertical space.
Save the vertical distance N for later output with os request.
Execute program command-line.
Set tabs after every position that is a multiple of N (default scaling indicator~
Set tabs at positions n1, n2, nn, then set tabs at nn+r1, nn+r2, nn+rn, then at nn+rn+r1, nn+rn+r2, nn+rn+rn, and so on.
Remove tab repition character.
Set tab repetition character to~c.
Temporary indent next line (default scaling indicator~
Enable track kerning for font.
Three-part title.
Print anything on terminal (UNIX standard message output).
Print anything on terminal (UNIX standard message output), allowing leading whitespace if anything starts with Text " " (which will be stripped off).
Similar to .tm1 without emitting a final newline.
Translate a to b, c to d, etc. on output.
Transparently output the contents of file filename.
This is the same as the tr request except that the asciify request will use the character code (if any) before the character translation.
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 \!.
Make the built-in condition t true and n false.
Underline font set to font (to be switched to by .ul ).
Underline (italicize in troff) N input lines.
Unformat space characters and tabs, preserving font information in diversion.
Enable vertical position traps if n is non-zero, disable them otherwise.
Change to previous vertical base line spacing.
Set vertical base line spacing according to ±N (default scaling indicator~ Default value is
Set warnings code to n.
Set scaling indicator used in warnings to si.
Remove (first) trap at position N.
Set location trap; negative means from page bottom.
While condition cond is true, accept anything as input.
Write anything to the stream named stream.
Similar to .write without emitting a final newline.
Write contents of macro or string xx to the stream named stream.

Besides these standard groff requests, there might be further macro calls. They can originate from a macro package (see roff(7) for an overview) or from a preprocessor.

Preprocessor macros are easy to be recognized. They enclose their code into a pair of characteristic macros.

preprocessor start macro end macro