Spotting strings that are typos in “text mode” but that are right in “math mode”

Question

My copy editor work is largely dedicate do spot typos. I have a kind of typos formed by strings like, e.g., ".", ",", ":"... (a space before a punctuation sign) that we consider typos in text-mode (e.g. "as follow :") but aren't a problem inside in-line math or math environments (e.g. "$\mu .$").

Usually we use custom emacs' query-replace regexp to find and correct these typos but, as emacs doesn't distinguish math mode from text mode (I know, I can do some tricks to do that but it is not a robust solution), I typically have a lot of false positives (and this is very boring).

I was wondering if there is a native LaTeX solution to do that. I mean not the query-replace but to highlight some matched strings.

I found, e.g., the xesearch package but it doesn't handle punctuation search.

Here is a MWE:

\documentclass[11pt]{article}
\pagestyle{empty}
\begin{document}

Punctuation , typos .

No problem with this string
\begin{equation}
(\mu + \nu) ,
\end{equation}
\end{document}

I also found l3regex but I think it performs regexp works on the LaTeX code and not on the pre-formatted code as xesearch does.

Any hints, suggestions or helpful ideas are welcome.

Edit. As I answered in comments: "I'm asking for a technique to distinguish strings that are in math mode from strings that are in text mode and highlight only the second ones".

Edit 2. I moved and posted a related question here: XeTeXinterchartoks: Why the “)” in the layout of “\eqref” and (\ref{…}) are interpreted differently?

Answer 1

You could use xelatex and \XeTeXinterchartoks:

\documentclass[11pt]{article}
%
\usepackage{xcolor}

\newXeTeXintercharclass\mypunctclass
\XeTeXcharclass `\. \mypunctclass
\XeTeXcharclass `\, \mypunctclass
\XeTeXinterchartokenstate=1
\XeTeXinterchartoks 4095 \mypunctclass{\textcolor{red}{TYPO}}



\pagestyle{empty}
\begin{document}

Punctuation , typos .

Punctuation, typos.

No problem with this string
\begin{equation}
(\mu + \nu) ,
\end{equation}
\end{document}

Answer 2

In keeping with the Carlisle tradition, I am naming this "gem" \z. Done here for commas, periods and colons, but added loops could be set up for other tests.

I would be remiss if I didn't draw your attention to Ulrike's comments, noting that there are many pitfalls for \z. For example verbatim material, as well as the occurrence of the offending syntax inside a valid argument list are likely to cause compilation errors.

EDITED so that it can now work when the offending bit falls, for example, inside an align environment. Rather than outputting the parts on the fly (which causes problems if the offending bit falls inside an argument), I instead create a concatenated \def in which every ., etc. is replaced with an \ifmmode{} .\else\colorbox{cyan}{ .}\fi inside the \def. Only at the end, is the defed macro named \z@save regurgitated. [Just for completeness, the original behavior can be obtained by defining \newcommand\zz@[1]{#1} instead of the revised \newcommand\zz@[1]{\g@addto@macro\z@save{#1}} ]

\documentclass[11pt]{article}
\usepackage{xcolor,amsmath}
\makeatletter
\newcommand\z[1]{\def\z@save{}\z@A#1 ,\z@end\z@save}
\long\def\z@A#1 ,#2\z@end{%
  \z@B#1 .\z@end%
  \ifx\z@end#2\else\zz@{\ifmmode{} ,\else\colorbox{cyan}{ ,}\fi}\z@A#2\z@end\fi}
\long\def\z@B#1 .#2\z@end{%
  \z@C#1 :\z@end%
  \ifx\z@end#2\else\zz@{\ifmmode{} .\else\colorbox{orange}{ .}\fi}\z@B#2\z@end\fi}
\long\def\z@C#1 :#2\z@end{%
  \zz@{#1}%
  \ifx\z@end#2\else\zz@{\ifmmode{} :\else\colorbox{yellow}{ :}\fi}\z@C#2\z@end\fi}
\newcommand\zz@[1]{\g@addto@macro\z@save{#1}}
\makeatother
\pagestyle{empty}
\begin{document}
\z{
Punctuation , typos : such as 3 .75 and such .

No problem with this string
\begin{equation}
(3 .7\mu + \nu) ,
\end{equation}

\begin{align} 
\mathcal{A}_{\text{reg}}:\qquad ( - l_u/2+\epsilon_u , - l_\phi/2+\epsilon_\phi ) 
  \rightarrow ( l_u/2-\epsilon_u , l_\phi/2-\epsilon_\phi )\,. \end{align}
And again , are there these problems : a nd b .
}
\end{document}