"Automatic Labelling" of (sub)(sub)sections does not work with math

Question

I followed the code in this answer. (sub)(sub)sections. However, if there is any math in the (sub)(sub)section title it fails to work. This is probably expected, but is there a way to overcome this? Or what label should I reference using \ref instead? Thanks in advance.

MWE:

\documentclass{article}
\usepackage{amsmath}
\usepackage{amsfonts}
\usepackage{etoolbox}
\makeatletter
\patchcmd{@sect}% <cmd>
  {@xsect}% <search>
  {\label{sec:#8}@xsect}% <replace>
  {}{}% <successs><failure>
\makeatother
\begin{document}
\section{The ($\epsilon$, $\delta$)-definition --- the rigorous approach}
Symbolically, $\displaystyle (\forall\ \epsilon > 0 ) , (\exists\ \delta > 0) , (\forall x \in \mathbb{R}) , (0 < |x - p| < \delta \implies |f(x) - L| < \varepsilon)$.
\section{Continuity and Differentiability}
... This can be trivially proved using any of the methods stated in section \ref{sec:The ($\epsilon$, $\delta$)-definition --- the rigorous approach}.
\end{document}

Answer 1

Automatic label generation from the title of a section may sound attractive, but mostly it is not, because it leads to lots of problems. This affects not only math mode, but macro expansion in general. In addition, all cross-references must be changed if a title is modified even minimally. This is almost as impractical as resolving all cross-references manually instead of automatically by \label/\ref.

Nevertheless, you can define substitutions for all sorts of macros to try to achieve it anyway. Here is a small example using LaTeX3:

\documentclass{article}
\usepackage{amsmath}
\usepackage{amsfonts}
\usepackage{etoolbox}
\makeatletter
\patchcmd{@sect}% <cmd>
  {@xsect}% <search>
  {\SecLabel{#8}@xsect}% <replace>
  {}{}% <successs><failure>
\makeatother
\ExplSyntaxOn
\cs_new:Nn \doc_label_set_sanitized:Nn
  {
    \str_set:Nn #1 { #2 }
    \str_remove_all:Nn #1 { $ }
    \str_replace_all:Nnn #1 { \epsilon } { epsilon }
    \str_replace_all:Nnn #1 { \delta } { delta }
    % Extend this replacement list according to your own needs.
  }
\NewDocumentCommand { \SecLabel } { m }
  {
    \doc_label_set_sanitized:Nn \l_tmpa_str { #1 }
    \label { sec: \l_tmpa_str }
  }
\NewDocumentCommand { \SecRef } { m }
  {
    \doc_label_set_sanitized:Nn \l_tmpa_str { #1 }
    \ref { sec: \l_tmpa_str }
  }
\ExplSyntaxOff
\begin{document}
\section{The ($\epsilon$, $\delta$)-definition --- the rigorous approach}
Symbolically, $\displaystyle (\forall\ \epsilon > 0 ) , (\exists\ \delta > 0) , (\forall x \in \mathbb{R}) , (0 < |x - p| < \delta \implies |f(x) - L| < \varepsilon)$.
\section{Continuity and Differentiability}
... This can be trivially proved using any of the methods stated in section \SecRef{The ($\epsilon$, $\delta$)-definition --- the rigorous approach}.
\end{document}

However, if you want to do this for the complete possible mathematics, this could become more than time-consuming.

As an alternative, you also could replace all macros by their names and additionally remove, e.g., subscripts and superscripts:

\documentclass{article}
\usepackage{amsmath}
\usepackage{amsfonts}
\usepackage{etoolbox}
\makeatletter
\patchcmd{@sect}% <cmd>
  {@xsect}% <search>
  {\SecLabel{#8}@xsect}% <replace>
  {}{}% <successs><failure>
\makeatother
\ExplSyntaxOn
\cs_new:Nn \doc_label_set_sanitized:Nn
  {
    \str_set:Nn #1 { #2 }
    \str_remove_all:Nn #1 { $ }
    \regex_replace_all:nnN { [\_^] } { } #1 % replace each macro by its name
                                % and also remove subscripts and superscripts
  }
\NewDocumentCommand { \SecLabel } { m }
  {
    \doc_label_set_sanitized:Nn \l_tmpa_str { #1 }
    \label { sec: \l_tmpa_str }
  }
\NewDocumentCommand { \SecRef } { m }
  {
    \doc_label_set_sanitized:Nn \l_tmpa_str { #1 }
    \ref { sec: \l_tmpa_str }
  }
\ExplSyntaxOff
\begin{document}
\section{The ($\epsilon^1$, $\delta$)-\emph{definition} --- the rigorous approach}
Symbolically, $\displaystyle (\forall\ \epsilon > 0 ) , (\exists\ \delta > 0) , (\forall x \in \mathbb{R}) , (0 < |x - p| < \delta \implies |f(x) - L| < \varepsilon)$.
\section{Continuity and Differentiability}
... This can be trivially proved using any of the methods stated in section \SecRef{The ($\epsilon^1$, $\delta$)-\emph{definition} --- the rigorous approach}.
\end{document}

See “The LaTeX3 Interface” for more information.