53

I'd like to typeset two compound symbols for making definitions. The following

  • \newcommand{\defeq}{\mathrel{\mathop:}=}
  • \newcommand{\eqdef}{\mathrel{\mathop=}:}

are suggested by the CTAN symbol list in the footnote on page 35. (Well, the first is suggested, and I made up the second.) Is there a better way to do this, short of including txfonts? Basically, it always looks like \defeq has more space on the left than on the right, and conversely for \eqdef.

EDIT - Most of the answers are taking about packages that fix the vertical placement of the colon. Ok, that is great and I'm interested to hear it. But let me restate my question. I want to fix the apparent difference in space on the left and right sides of \defeq. To see what I mean, look at typeset equation $a \defeq b$ in Stefan's answer. The $a$ and the $b$ are equal distances from the center. But because the colon is a lighter symbol _it_looks_ the $a$ is further away. Does that make sense?

bodo
  • 6,228
Sam Nead
  • 1,051

4 Answers4

42

I recommend using a vertically centered colon which matches the equal sign better, such as \vcentcolon from the mathtools package. You can also change it generally to vertical by \mathtoolsset{centercolon}.

Compare standard :=

def equal

to the vertically centered:

def equal

So your definition could become

\usepackage{mathtools}
\newcommand{\defeq}{\vcentcolon=}
\newcommand{\eqdef}{=\vcentcolon}

with tighter spacing. And since both symbols are already relation symbols, as egreg said, we even could omit \mathrel.

And you could do that even easier by mathtools provided commands:

\documentclass{article}
\usepackage{mathtools}
\begin{document}
$a\coloneqq b \qquad b\eqqcolon a$
\end{document}
Stefan Kottwitz
  • 231,401
  • 6
    There's no need to add \mathrel with the second set of definitions, as both = and \vcentcolon are already relation symbols. The \mathop in the first definition in the first set does nothing; in the second it's wrong. – egreg Sep 18 '11 at 20:10
  • In the latter group the \mathrel is useless. The former group gives obviously wrong spacing. – egreg Sep 18 '11 at 20:24
  • The use of \mathop in the question is correct in the first definition and incorrect in the second one, as I remarked in my answer. The fact is that \vcentcolon is defined exactly as \mathrel{\mathop:}; TeX doesn't space consecutive Rel atoms. – egreg Sep 18 '11 at 20:39
12

The mathtools package offers you the centercolon = true|false option, as well as a wide range of predefined symbols that include a vertically centered colon:

\documentclass{article}
\usepackage{mathtools}

\begin{document}

\[
a \coloneqq b \quad c \Colonapprox d \quad e \dblcolon f
\]

\mathtoolsset{centercolon}
\[
a := b
\]

\end{document}
Stefan Pinnow
  • 29,535
Gonzalo Medina
  • 505,128
  • 1
    Be aware that centercolon cannot be used with french babel, og other setups that leave : active. The full names are recommended. – daleif Sep 20 '11 at 09:40
11

This, as far as I know, is the best kludge around (it's the same used by mathtools for its \vcentcolon command). However the second definition should be

\newcommand{\eqdef}{=\mathrel{\mathop:}}

otherwise both symbols would be misplaced with respect to the other, and the result would be bad anyway.

Note
As a mathematician I've never felt any need of an equals-by-definition symbol as did many others for centuries. But this risks to be the source for another religion war. :)

egreg
  • 1,121,712
  • 2
    I completely agree with you. I don't see the need for the equals-by-definition symbol. – Gonzalo Medina Sep 18 '11 at 19:56
  • 1
    I've seen eqdef used in applied maths (control theory) and I quite like it there; sort of like a capital letter at the start of a sentence for maths. But I'm not a mathematician, only a lowly engineer :-) – Will Robertson Sep 18 '11 at 22:15
  • 2
    @WillRobertson It makes sense in assignments, where it doesn't represent equality; but in definitions it's not necessary. Clarity is achieved through words, not symbols. – egreg Sep 18 '11 at 22:31
  • 2
    A colleague of mine refers to that urge that some mathematicians feel to introduce unnecessary new symbols as "conjuntivitis" (playing with "misusing the set theory object language" (set is Spanish is "conjunto") and the well known inflammation of the conjunctiva). – Gonzalo Medina Sep 19 '11 at 01:22
  • The whole definition must to enclosed in \mathrel, otherwise the symbol can break between the = and : (The mathtools package had this bug for some time). – Aditya Sep 19 '11 at 05:23
  • 1
    @Aditya: can you make an example? As far as I know, there's no legal break point between two consecutive Rel atoms; compare, for example, the Plain TeX definition of \longrightarrow, where three consecutive Rel atoms are used, not enclosed in \mathrel. It would be different if a \mkern were inserted between the two symbols, of course; mathtools actually does this, which was the real cause of the bug. – egreg Sep 19 '11 at 08:49
  • @egreg: You are right. What I had in mind was the bug that I had reported on comp.text.tex, but that was because mathtools was using a \mkern. – Aditya Sep 20 '11 at 20:01
  • The best code is indeed \vcentcolon\mathrel{\mkern-1.2mu}= that avoids the bug and is probably slightly more efficient, or at least more elegant. – egreg Sep 20 '11 at 20:04
5

Looking at your code for the \defeq command, I noticed an asymmetry w.r.t. the scope of the \mathrel instruction. The following commands avoid this asymmetry in the spacing around the symbols:

\newcommand{\defeq}{\mathrel{:\mkern-0.25mu=}}
\newcommand{\eqdef}{\mathrel{=\mkern-0.25mu:}}

where I've made use of the \mkern ("math-kern") command to tighten up the space (by a very small amount) between the two constituent symbols. The value of the mkern instruction was determined by the eyeball method.

A downside of this approach is that the colon isn't centered vertically relative to the equals sign. This could be fixed by fiddling some more, but I think you're better off using a predefined macro. One actually almost has an embarrassment of riches for the "colon-equals" symbol:

  • packages mathabx, MnSymbol, and mtpro2: \coloneq
  • packages txfonts, pxfonts, and mathtools: \coloneqq
  • package colonequals: \colonequals (what else?!)

Note that the mtpro2 package (short for "MathTime Pro 2") is available in a "lite" version (free) as well as a "professional" (not free) version. The mtpro2 and txfonts will use a Times-Roman font style, and pxfonts will use a Palatino-type font; this may or may not be acceptable to you, depending on what text font you're using. The version from the \mathabx package has the interesting feature that the length of its "equal" symbol is shorter than normal, and the resulting width of its \coloneq command is the same as that of the normal equals sign.

For a predefined version of the reverse symbol, i.e., "equals-colon", you could use the \eqcolon command of the mtpro2 package (may be easiest if you already have that package) or the \equalscolon command of the colonequals package. In addition, you can always rotate your preferred form of the colon-equals symbol by 180 degrees to generate the reverse symbol using, say, the \rotatebox command of the graphicx package. Just remember that you need to tell TeX about the "type" of the new symbol, which should be "mathrel". The following MWE explains how to achieve this effect. 

\documentclass{article}
\usepackage{MnSymbol, graphicx}
\newcommand{\eqcolon}{\mathrel{\rotatebox[units=360]{180}{\ensuremath{\coloneq}}}}
\begin{document}
$a \coloneq b$ and $c \eqcolon d$.
\end{document}

Happy TeXing!

Mico
  • 506,678