Semantic line breaks and spaces around em-dashes

Question

I would like to use semantic line breaks in my document's source code, but I'm not sure what to do with em dashes: ---.

sentence-per-line, no em-dash spaces:

The food---which was delicious---reminded me of home.
The food, which was delicious, reminded me of home.

clause-per-line, inconsistent em-dash spaces:

The food
---which was delicious---
reminded me of home.
The food,
which was delicious,
reminded me of home.

I would like both sources to render the same, but the clause-per-line version introduces a space before or after an em-dash.

I could instead use

sentence-per-line, em-dash spaces:

The food --- which was delicious --- reminded me of home.

and

clause-per-line, em-dash spaces:

The food
--- which was delicious ---
reminded me of home.

which render the same as each other and make the em-dash spacing situation consistent with the comma situation.

I would like to defer the style choice of how much space is around an em-dash, and be able to try different choices without modifying the source throughout.

I would prefer to continue using triple ASCII dash ---, and could also consider using a Unicode character, but I do not want to use any backslash \ to invoke a command.

Related: Semantic linefeeds and problem with space before footnote mark

Answer 1

This answer, which drew directly from my answer at Automatic text highlighting based on a dictionary, uses the same technique: running the tokens of the environment through a scanner in search of, in this case - tokens. If they are found to occur in a group of three, the tokens are prepended with \unskip and appended with \ignorespaces, before the token list is executed.

To start the environment, invoke \def\currentword{}\tokencyclexpress and to close it out, invoke \endtokencyclexpress. This can be done once, at the beginning and end of the document.

\documentclass{article}
\usepackage{tokcycle,listofitems,xcolor}
\setsepchar{-}
\newcommand\testdict{%
  \if\relax\detokenize\expandafter{\currentword}\relax\else
    {\ignoreemptyitems
      \greadlist\dictcompA{\currentword}}%
    \ifnum\listlen\dictcompA[]=0\relax
      \addcytoks[1]{\autohighlightStyleA}%
      \addcytoks[1]{\expandafter{\currentword}}
    \else
      \addcytoks[1]{\currentword}%
    \fi
  \fi
  \gdef\currentword{}%
}
\makeatletter
\Characterdirective{\tctestifx 
  {-#1}{\g@addto@macro\currentword{#1}}
  {\testdict\addcytoks{#1}}}
\stripgroupingtrue
\Groupdirective{\testdict\groupedcytoks{\processtoks{#1}\testdict}}
\Macrodirective{\g@addto@macro\currentword{#1}}
\Spacedirective{\testdict\addcytoks{#1}}
\makeatother
\newcommand\autohighlightStyleA{\zz}
\def\emdashref{---}
\newcommand\zz[1]{\def\tmp{#1}\ifx\emdashref\tmp
  \unskip#1\ignorespaces\else#1\fi}
\begin{document}
\def\currentword{}\tokencyclexpress
The food---which was delicious---reminded me of home.
The food, which was delicious, reminded me of home. - . -- .

The food
---which was delicious---
reminded me of home.
The food,
which was delicious,
reminded me of home. - . -- .
\endtokencyclexpress
\end{document}

Answer 2

You can add % at the end of a line to suppress the space:

The food%
---which was delicious---%
reminded me of home.
The food,
which was delicious,
reminded me of home.