Labels and Cross-References

In standard , the argument to the \label command cannot contain anything but characters (after expansion). In the geom style, this restriction is lifted, so labels contain just about anything. For example, \label{\TeX} and \label{Isometries of $\bold R^2$} are acceptable. Macros are no longer expanded.

One advantage of this flexibility is that subdivisions can be automatically labeled by their titles. In the geom format, by default, every sectioning command such as \section generates a label with its title. If this feature causes you to run out of memory, you can turn it off by saying \autolabelfalse in your document's preamble.

To refer to the place where a label occurs, you can still use the \ref command, as in standard . This gives a bald number—for example, \ref{Isometries of $\bold R^2$} will print ``8.3'' if Section 8.3 starts with \section{Isometries of $\bold R^2$} while automatic labeling is in effect.

A new feature is that you can replace \ref by \fullref or \Fullref, to get ``section 8.3'' or ``Section 8.3'', respectively. The way \fullref works is by storing in a certain register the type of environment in which a label was created. Thus, when a new section starts, this register gets the value ``section''. When a \begin command introducing one of the environments described in Theorems and Their Friends is encountered inside this section, the register gets updated to the type of that environment, say ``theorem''. It reverts to ``section'' when the matching \end command is read.

Even if you have nested environments, labels placed in each of them are assigned the correct type. Consider the input

\newtheorem[{\par\large}{\it}]{theorem}{Theorem}[section]
\equationwith{theorem}                  % These two lines are explained
...                                     % in later sections
\begin{theorem}[Erd\"os]
A theorem both deep and profound / Is that a circle is round:
\begin{equation}
\mathrm{curvature}\,S^1=\mathrm {const}.
\label{circle}
\end{equation}
In a paper by Erd\"os / Written in Kurdish / A counterexample is found.
\label{Kurdish}
\end{theorem}

to *If you disagree with 's idea of what the label type should be at a particular point, you can override is using \setlabeltype. For example, if we had \setlabeltype{formula} right before the \label command in the previous example, the result of \Fullref{circle} would be ``Formula '' instead of ``circle''.

Before \ref actually prints the number of a cross-reference, it applies the macro \preref to it. By default, this macro does nothing; but you can redefine it to do whatever you want with the number. If you want the number to appear upright even if the surrounding text is in italics, you can say

\def\preref#1{{\normalshape#1}}