eqn

Anatomy of an equation

eqn input consists of tokens. Consider a form of Newton's second law of motion for constant mass. The input

.EQ
F =
m a
.EN

becomes $F=ma.$ Each of F, =, m, and a is a token. Spaces and newlines are interchangeable; they separate tokens but do not break lines or produce space in the output.

Beyond their primary functions, the following input characters separate tokens as well.

{ }

Braces perform grouping. Whereas “e sup a b” expresses “(e to the a) times b”, “e sup { a b }” means “e to the (a times b)”. When immediately preceded by a “left” or “right” primitive, a brace loses its special meaning.

^ ~

are the half space and full space, respectively. Use them to tune the appearance of the output.

Tab and leader characters separate tokens as well as advancing the drawing position to the next tab stop, but are seldom used in eqn input. When they occur, they must appear at the outermost lexical scope. This roughly means that they can't appear within braces that are necessary to disambiguate the input; eqn will diagnose an error in this event. (See subsection “Macros” below for additional token separation rules.)

Other tokens are primitives, macros, an argument to either of the foregoing, or components of an equation.

Primitives are fundamental keywords of the eqn language. They can configure an aspect of the preprocessor's state, as when setting a “global” font selection or type size (gifont and gsize), or declaring or deleting macros (“define” and undef); these are termed commands. Other primitives perform formatting operations on the tokens around them (as with fat, over, sqrt, or up).

Equation components include mathematical variables, constants, numeric literals, and operators. eqn remaps some input character sequences to groff special character escape sequences for economy in equation entry and to ensure that glyphs from an unstyled font are used; see groff_char(7).

Macros permit primitives, components, and other macros to be collected and used together as a single token. Predefined macros make convenient the preparation of eqn input in a form resembling its spoken expression; for example, consider cos, hat, inf, and lim.

Spacing and typeface

GNU eqn imputes a type to each equation component, adjusting the space around it accordingly. Recognized types follow; most affect spacing only, whereas the “letter” subtype of “ordinary” also assigns a style.

Two primitives apply types to equation components.

type t e

Apply type t to expression e.

chartype t text

Assign each character in (unquoted) text type t, persistently.

eqn sets up spacings and styles as if by the following commands.

eqn assigns all other ordinary and special roff characters, including numerals 0–9, the “ordinary” type. (The “digit” type is not used, but is available for customization.) In keeping with common practice in mathematical typesetting, lowercase, but not uppercase, Greek letters are assigned the “letter” type to style them in italics. The macros for producing ellipses, “...”, cdots, and ldots, use the “inner” type.

Primitives

eqn supports without alteration the AT&T eqn primitives above, back, bar, bold, define, down, fat, font, from, fwd, gfont, gsize, italic, left, lineup, mark, matrix, ndefine, over, right, roman, size, sqrt, sub, sup, tdefine, to, under, and up.

New primitives

We describe the GNU extension primitives “type” and chartype in subsection “Spacing and typeface” above; “set” and “reset” in subsection “Customization” below; and gbfont, gifont, and grfont in subsection “Fonts” below. In the following synopses, X can be any character not appearing in the parameter thus bracketed.

e1 accent e2

Set e2 as an accent over e1. eqn assumes that e2 is at the appropriate height for a lowercase letter without an ascender, and shifts it vertically based on e1's height. For example, eqn defines hat as follows.

dotdot, dot, tilde, vec, and dyad are also defined using the accent primitive.

big e

Enlarge the expression e; semantics like those of CSS “large” are intended. In troff output, the type size is increased by 5 scaled points. MathML output emits the following.

copy file

include file

Interpolate the contents of file, omitting lines beginning with .EQ or .EN. file is sought relative to the current working directory.

ifdef name X anything X

If name is defined as a primitive or macro, interpret anything.

nosplit text

As "text", but since text is not quoted it is subject to macro expansion; it is not split up and the spacing between characters not adjusted per subsection “Spacing and typeface” above.

e opprime

As prime, but set the prime symbol as an operator on e. In the input “A opprime sub 1”, the “1” is tucked under the prime as a subscript to the “A” (as is conventional in mathematical typesetting), whereas when prime is used, the “1” is a subscript to the prime character. The precedence of opprime is the same as that of bar and “under”, and higher than that of other primitives except accent and uaccent. In unquoted text, a neutral apostrophe (') that is not the first character on the input line is treated like opprime.

sdefine name X anything X

As “define”, but name is not recognized as a macro if called with arguments.

e1 smallover e2

As over, but reduce the type size of e1 and e2, and put less vertical space between e1 and e2 and the fraction bar. The over primitive corresponds to the TeX \over primitive in displayed equation styles; smallover corresponds to \over in non-display (“inline”) styles.

space n

Set extra vertical spacing around the equation, replacing the default values, where n is an integer in hundredths of an em. If positive, n increases vertical spacing before the equation; if negative, it does so after the equation. This primitive provides an interface to groff's \x escape sequence, but with the opposite sign convention. It has no effect if the equation is part of a pic(1) picture.

special troff-macro e

Construct an object by calling troff-macro on e. The troff string 0s contains the eqn output for e, and the registers 0w, 0h, 0d, 0skern, and 0skew the width, height, depth, subscript kern, and skew of e, respectively. (The subscript kern of an object indicates how much a subscript on that object should be “tucked in”, or placed to the left relative to a non-subscripted glyph of the same size. The skew of an object is how far to the right of the center of the object an accent over it should be placed.) The macro must modify 0s so that it outputs the desired result, returns the drawing position to the text baseline at the beginning of e, and updates the foregoing registers to correspond to the new dimensions of the result.

Suppose you want a construct that “cancels” an expression by drawing a diagonal line through it.

We use the \[mi] and \[pl] special characters instead of + and - because they are part of the argument to a troff macro, so eqn does not transform them to mathematical glyphs for us. Here's a more complicated construct that draws a box around an expression; the bottom of the box rests on the text baseline. We define the eqn macro box to wrap the call of the troff macro Bx.

split "text"

As text, but since text is quoted, it is not subject to macro expansion; it is split up and the spacing between characters adjusted per subsection “Spacing and typeface” above.

e1 uaccent e2

Set e2 as an accent under e1. e2 is assumed to be at the appropriate height for a letter without a descender; eqn vertically shifts it depending on whether e1 has a descender. utilde is predefined using uaccent as a tilde accent below the baseline.

undef name

Remove definition of macro or primitive name, making it undefined.

vcenter e

Vertically center e about the math axis, a horizontal line upon which fraction bars and characters such as “+” and “−” are aligned. MathML already behaves this way, so eqn ignores this primitive when producing that output format. The built-in sum macro is defined as if by the following.

Extended primitives

GNU eqn extends the syntax of some AT&T eqn primitives, introducing one deliberate incompatibility.

delim on

GNU eqn recognizes an “on” argument to the delim primitive specially, restoring any delimiters previously disabled with “delim off”. If delimiters haven't been specified, neither command has effect. Few eqn documents are expected to use “o” and “n” as left and right delimiters, respectively. If yours does, swap them, or select others.

col n { ... }

ccol n { ... }

lcol n { ... }

rcol n { ... }

pile n { ... }

cpile n { ... }

lpile n { ... }

rpile n { ... }

The integer value n, in hundredths of an em, uses the formatter's \x escape sequence to increase the vertical spacing between rows; eqn ignores it when producing MathML. Negative values are accepted but have no effect. If more than one n occurs in a matrix or pile, the largest is used.

Customization

When eqn generates troff input, the appearance of equations is controlled by a large number of parameters. They have no effect when generating MathML, which delegates typesetting to a MathML rendering engine. Configure these parameters with the “set” and “reset” primitives.

set p n

assigns parameter p the integer value n; n is interpreted in units of hundredths of an em unless otherwise stated. For example,

set x_height 45

reset p

restores the default value of parameter p.

Available parameters p are as follows; defaults are shown in parentheses. We intend these descriptions to be expository rather than rigorous.

minimum_size

sets a floor for the type size (in scaled points) at which equations are set (5).

fat_offset

The fat primitive emboldens an equation by overprinting two copies of the equation horizontally offset by this amount (4). When producing MathML output, components to which fat_offset applies instead use the following.

<mstyle mathvariant='double-struck'>

over_hang

A fraction bar is longer by twice this amount than the maximum of the widths of the numerator and denominator; in other words, it overhangs the numerator and denominator by at least this amount (0).

accent_width

When bar or under is applied to a single character, the line is this long (31). Normally, bar or under produces a line whose length is the width of the object to which it applies; in the case of a single character, this tends to produce a line that looks too long.

delimiter_factor

Extensible delimiters produced with the left and right primitives have a combined height and depth of at least this many thousandths of twice the maximum amount by which the sub-equation that the delimiters enclose extends away from the axis (900).

delimiter_shortfall

Extensible delimiters produced with the left and right primitives have a combined height and depth not less than the difference of twice the maximum amount by which the sub-equation that the delimiters enclose extends away from the axis and this amount (50).

null_delimiter_space

This much horizontal space is inserted on each side of a fraction (12).

script_space

The width of subscripts and superscripts is increased by this amount (5).

thin_space

This amount of space is automatically inserted after punctuation characters (17).

medium_space

This amount of space is automatically inserted on either side of binary operators (22).

thick_space

This amount of space is automatically inserted on either side of relations (28).

half_space

configures the width of the space produced by the ^ token (17).

full_space

configures the width of the space produced by the ~ token (28).

x_height

The height of lowercase letters without ascenders such as “x” (45).

axis_height

The height above the baseline of the center of characters such as “+” and “−” (26). It is important that this value is correct for the font you are using.

default_rule_thickness

This should be set to the thickness of the \[ru] character, or the thickness of horizontal lines produced with the \D escape sequence (4).

num1

The over primitive shifts up the numerator by at least this amount (70).

num2

The smallover primitive shifts up the numerator by at least this amount (36).

denom1

The over primitive shifts down the denominator by at least this amount (70).

denom2

The smallover primitive shifts down the denominator by at least this amount (36).

sup1

Normally superscripts are shifted up by at least this amount (42).

sup2

Superscripts within superscripts or upper limits or numerators of smallover fractions are shifted up by at least this amount (37). Conventionally, this is less than sup1.

sup3

Superscripts within denominators or square roots or subscripts or lower limits are shifted up by at least this amount (28). Conventionally, this is less than sup2.

sub1

Subscripts are normally shifted down by at least this amount (20).

sub2

When there is both a subscript and a superscript, the subscript is shifted down by at least this amount (23).

sup_drop

The baseline of a superscript is no more than this much below the top of the object on which the superscript is set (38).

sub_drop

The baseline of a subscript is at least this much below the bottom of the object on which the subscript is set (5).

big_op_spacing1

The baseline of an upper limit is at least this much above the top of the object on which the limit is set (11).

big_op_spacing2

The baseline of a lower limit is at least this much below the bottom of the object on which the limit is set (17).

big_op_spacing3

The bottom of an upper limit is at least this much above the top of the object on which the limit is set (20).

big_op_spacing4

The top of a lower limit is at least this much below the bottom of the object on which the limit is set (60).

big_op_spacing5

This much vertical space is added above and below limits (10).

baseline_sep

The baselines of the rows in a pile or matrix are normally this far apart (140). Usually equal to the sum of num1 and denom1.

shift_down

The midpoint between the top baseline and the bottom baseline in a matrix or pile is shifted down by this much from the axis (26). Usually equal to axis_height.

column_sep

This much space is added between columns in a matrix (100).

matrix_side_sep

This much space is added at each side of a matrix (17).

draw_lines

If non-zero, eqn draws lines using the troff \D escape sequence, rather than the \l escape sequence and the \[ru] special character. The eqnrc file sets the default: 1 on ps, html, and the X11 devices, otherwise 0.

body_height

is the presumed height of an equation above the text baseline; eqn adds any excess as extra pre-vertical line spacing with troff's \x escape sequence (85).

body_depth

is the presumed depth of an equation below the text baseline; eqn adds any excess as extra post-vertical line spacing with troff's \x escape sequence (35).

nroff

If non-zero, then ndefine behaves like define and tdefine is ignored, otherwise tdefine behaves like define and ndefine is ignored. The eqnrc file sets the default: 1 on ascii, latin1, and utf8 devices, otherwise 0.

Macros

In GNU eqn, macros can take arguments. A word defined by any of the define, ndefine, or tdefine primitives followed immediately by a left parenthesis is treated as a parameterized macro call: subsequent tokens up to a matching right parenthesis are treated as comma-separated arguments. In this context only, commas and parentheses also serve as token separators. A macro argument is not terminated by a comma inside parentheses nested within it. In a macro definition, $n, where n is between 1 and 9 inclusive, is replaced by the nth argument; if there are fewer than n arguments, it is replaced by nothing.

Predefined macros

GNU eqn supports the predefined macros offered by AT&T eqn: and, approx, arc, cos, cosh, del, det, dot, dotdot, dyad, exp, for, grad, half, hat, if, inter, Im, inf, int, lim, ln, log, max, min, nothing, partial, prime, prod, Re, sin, sinh, sum, tan, tanh, tilde, times, union, vec, ==, !=, +=, ->, <-, <<, >>, and “...”. The lowercase classical Greek letters are available as alpha, beta, chi, delta, epsilon, eta, gamma, iota, kappa, lambda, mu, nu, omega, omicron, phi, pi, psi, rho, sigma, tau, theta, upsilon, xi, and zeta. Spell them with an initial capital letter (Alpha) or in full capitals (ALPHA) to obtain uppercase forms.

GNU eqn further defines the macros cdot, cdots, and utilde (all discussed above), dollar, which sets a dollar sign, and ldots, which sets an ellipsis on the text baseline.

Fonts

eqn uses up to three typefaces to set an equation: italic (oblique), roman (upright), and bold (heavy). Assign each a groff typeface with the GNU extension primitives grfont, gifont, and gbfont. The defaults are the styles R, I, and B (applied to the current font family). The chartype primitive (see above) sets a character's type, which determines the face used to set it. The “letter” type is set in italics; others are set in roman. Use the bold primitive to select an (upright) bold style.

gbfont f

Select f as the bold font.

gifont f

Select f as the italic font. GNU eqn recognizes gfont as a synonym for AT&T compatibility.

grfont f

Select f as the roman font.

roff interface

The string 10 stores the most recently formatted equation. The register MK stores the horizontal drawing position of the most recently assigned mark. Avoid using these names for any other purpose.

GNU eqn internally employs register, string, macro, and diversion names beginning with the numeral 0. (AT&T eqn used identifiers starting with the numeral 1.) A document to be preprocessed with GNU eqn should not use any such identifiers.