groff_font − format of groff device and font description files


The groff font format is roughly a superset of the ditroff font format. The font files for device name are stored in a directory devname. There are two types of file: a device description file called DESC and for each font F a font file called F. These are text files; unlike the ditroff font format, there is no associated binary format.

DESC file format
The DESC file can contain the following types of line as shown below. Later entries in the file override previous values.

This line and everything following in the file are ignored. It is allowed for the sake of backwards compatibility.

family fam

The default font family is fam.

fonts n F1 F2 F3...Fn

Fonts F1...Fn will be mounted in the font positions m+1,...,m+n where m is the number of styles. This command may extend over more than one line. A font name of 0 will cause no font to be mounted on the corresponding font position.

hor n

The horizontal resolution is n machine units.

image_generator string

Needed for grohtml only. It specifies the program to generate PNG images from PostScript input. Under GNU/Linux this is usually gs but under other systems (notably cygwin) it might be set to another name.

paperlength n

The physical vertical dimension of the output medium in machine units. This isn’t used by troff itself but by output devices. Deprecated. Use papersize instead.

papersize string

Select a paper size. Valid values for string are the ISO paper types A0-A7, B0-B7, C0-C7, D0-D7, DL, and the US paper types letter, legal, tabloid, ledger, statement, executive, com10, and monarch. Case is not significant for string if it holds predefined paper types. Alternatively, string can be a file name (e.g. ‘/etc/papersize’); if the file can be opened, groff reads the first line and tests for the above paper sizes. Finally, string can be a custom paper size in the format length,width (no spaces before and after the comma). Both length and width must have a unit appended; valid values are ‘i’ for inches, ‘c’ for centimeters, ‘p’ for points, and ‘P’ for picas. Example: 12c,235p. An argument which starts with a digit is always treated as a custom paper format. papersize sets both the vertical and horizontal dimension of the output medium.

More than one argument can be specified; groff scans from left to right and uses the first valid paper specification.

paperwidth n

The physical horizontal dimension of the output medium in machine units. Deprecated. Use papersize instead. This isn’t used by troff itself but by output devices.


Make troff tell the driver the source file name being processed. This is achieved by another tcommand: F filename.

postpro program

Use program as the postprocessor.

prepro program

Call program as a preprocessor.

print program

Use program as the spooler program for printing. If omitted, the −l and −L options of groff are ignored.

res n

There are n machine units per inch.

sizes s1 0

This means that the device has fonts at s1, s2, scaled points. The list of sizes must be terminated by a 0. Each si can also be a range of sizes mn. The list can extend over more than one line.

sizescale n

The scale factor for pointsizes. By default this has a value of 1. One scaled point is equal to one point/n. The arguments to the unitwidth and sizes commands are given in scaled points.

styles S1 S2...Sm

The first m font positions will be associated with styles S1...Sm.


This means that the postprocessor can handle the t and u output commands.

unitwidth n

Quantities in the font files are given in machine units for fonts whose point size is n scaled points.


Make the font handling module always return unscaled character widths. Needed for the grohtml device.


This command indicates that troff should encode named characters inside special commands.

vert n

The vertical resolution is n machine units.

The res, unitwidth, fonts, and sizes lines are compulsory. Not all commands in the DESC file are used by troff itself; some of the keywords (or even additional ones) are used by postprocessors to store arbitrary information about the device.

Here a list of obsolete keywords which are recognized by groff but completely ignored: spare1, spare2, biggestfont.

Font file format
A font file has two sections. The first section is a sequence of lines each containing a sequence of blank delimited words; the first word in the line is a key, and subsequent words give a value for that key.
lig1 lig2...lign [0]

Characters lig1, lig2, ..., lign are ligatures; possible ligatures are ff, fi, fl, ffi and ffl. For backwards compatibility, the list of ligatures may be terminated with a 0. The list of ligatures may not extend over more than one line.

name F

The name of the font is F.

slant n

The characters of the font have a slant of n degrees. (Positive means forward.)

spacewidth n

The normal width of a space is n.


The font is special; this means that when a character is requested that is not present in the current font, it will be searched for in any special fonts that are mounted.

Other commands are ignored by troff but may be used by postprocessors to store arbitrary information about the font in the font file.

The first section can contain comments which start with the # character and extend to the end of a line.

The second section contains one or two subsections. It must contain a charset subsection and it may also contain a kernpairs subsection. These subsections can appear in any order. Each subsection starts with a word on a line by itself.

The word charset starts the charset subsection. The charset line is followed by a sequence of lines. Each line gives information for one character. A line comprises a number of fields separated by blanks or tabs. The format is

name metrics type code [entity_name] [-- comment]

name identifies the character: if name is a single character c then it corresponds to the groff input character c; if it is of the form \c where c is a single character, then it corresponds to the special character \[c]; otherwise it corresponds to the groff input character \[name]. If it is exactly two characters xx it can be entered as \(xx. Note that single-letter special characters can’t be accessed as \c; the only exception is ‘\-’ which is identical to ‘\[-]’. The name −−− is special and indicates that the character is unnamed; such characters can only be used by means of the \N escape sequence in troff.

Groff supports eight-bit characters; however some utilities have difficulties with eight-bit characters. For this reason, there is a convention that the name charn is equivalent to the single character whose code is n. For example, char163 would be equivalent to the character with code 163 which is the pounds sterling sign in ISO Latin-1.

The type field gives the character type:


means the character has a descender, for example, p;


means the character has an ascender, for example, b;


means the character has both an ascender and a descender, for example, (.

The code field gives the code which the postprocessor uses to print the character. The character can also be input to groff using this code by means of the \N escape sequence. The code can be any integer. If it starts with a 0 it will be interpreted as octal; if it starts with 0x or 0X it will be intepreted as hexadecimal. Note, however, that the \N escape sequence only accepts a decimal integer.

The entity_name field gives an ascii string identifying the glyph which the postprocessor uses to print the character. This field is optional and has been introduced so that the html device driver can encode its character set. For example, the character ‘\[Po]’ is represented as ‘£’ in html 4.0.

Anything on the line after the encoding field resp. after ‘--’ will be ignored.

The metrics field has the form (in one line; it is broken here for the sake of readability):


There must not be any spaces between these subfields. Missing subfields are assumed to be 0. The subfields are all decimal integers. Since there is no associated binary format, these values are not required to fit into a variable of type char as they are in ditroff. The width subfields gives the width of the character. The height subfield gives the height of the character (upwards is positive); if a character does not extend above the baseline, it should be given a zero height, rather than a negative height. The depth subfield gives the depth of the character, that is, the distance below the lowest point below the baseline to which the character extends (downwards is positive); if a character does not extend below above the baseline, it should be given a zero depth, rather than a negative depth. The italic-correction subfield gives the amount of space that should be added after the character when it is immediately to be followed by a character from a roman font. The left-italic-correction subfield gives the amount of space that should be added before the character when it is immediately to be preceded by a character from a roman font. The subscript-correction gives the amount of space that should be added after a character before adding a subscript. This should be less than the italic correction.

A line in the charset section can also have the format

name "

This indicates that name is just another name for the character mentioned in the preceding line.

The word kernpairs starts the kernpairs section. This contains a sequence of lines of the form:

c1 c2 n

This means that when character c1 appears next to character c2 the space between them should be increased by n. Most entries in kernpairs section will have a negative value for n.



Device description file for device name.


Font file for font F of device name.


groff_out(5), troff(1).