lualatex and line breaks after em-dashes

Question

I’ve just installed TeX Live 2012. As mentioned elsewhere on this site, the release history explains that “the parameter \XeTeXdashbreakstate is set to 1 by default, for both xetex and xelatex. This allows line breaks after em-dashes and en-dashes, which has always been the behavior of plain TEX, LATEX, LuaTEX, etc.”

I rarely use xetex; I’ve worked almost exclusively with luatex since the release of TeX Live 2011. The only frequent problem I had was that I did not get line breaks after em-dashes, and now in 2012 I’m still not getting them, unless I refrain from using the Unicode character — which certainly seems against the spirit of luatex.

Consider this example, where the same paragraph is typed 3 times, with the em-dashes entered differently in each:

\documentclass[11pt,a5paper]{book}
\usepackage{fontspec,microtype}
\setmainfont[Ligatures=TeX,RawFeature={protrusion=default}]{TeX Gyre Termes}
\pdfprotrudechars=2
\usepackage[showframe]{geometry}

% Will Robertson’s macro from https://tex.stackexchange.com/questions/34608/
\DeclareRobustCommand\dash{%
 \unskip\nobreak\thinspace\textemdash\allowbreak\thinspace\ignorespaces}

\pdfpagewidth=\paperwidth
\pdfpageheight=\paperheight
\pdfinfo{/Title (Jeeves Takes Charge) /Author (P.G. Wodehouse)}
\begin{document}
% using ---
Now, touching this business of old Jeeves---my man, you know---how do we
stand? Lots of people think I’m much too dependent on him. My Aunt
Agatha, in fact, has even gone so far as to call him my keeper. Well,
what I say is: Why not? The man’s a genius. From the collar upward he
stands alone.

% typing the em-dash (easy in utf-8 locale with compose key)
Now, touching this business of old Jeeves—my man, you know—how do we
stand? Lots of people think I’m much too dependent on him. My Aunt
Agatha, in fact, has even gone so far as to call him my keeper. Well,
what I say is: Why not? The man’s a genius. From the collar upward he
stands alone.

% using Will Robertson’s macro
Now, touching this business of old Jeeves\dash my man, you know\dash how do we
stand? Lots of people think I’m much too dependent on him. My Aunt
Agatha, in fact, has even gone so far as to call him my keeper. Well,
what I say is: Why not? The man’s a genius. From the collar upward he
stands alone.
\end{document}

The result looks like this:

I’ve experimented with various free and commercial fonts, always with the same result.

Is there some luatex equivalent to \XeTeXdashbreakstate which I could set?

Answer 1

I am not sure what the correct answer is. The UTF-8 emdash is considered part of the word. You see this (when you leave out microtype) in the log file:

Overfull \hbox (19.42857pt too wide) in paragraph at lines 28--33
[][]\EU2/TeXGyreTermes(0)/m/n/10.95 Now, touch-ing this busi-ness of old Jeeves
—my man, you know—how

I am not sure if TeX is able at all to decompose the words when it encounters a character such as emdash. So my answer would be "it's not possible".

What you can do, it's slightly a hack:

\catcode`\—=13
\protected\def—{\unskip\nobreak\thinspace\textemdash\allowbreak\thinspace\ignorespaces}

This makes the emdash active and inserts a "breakable" emdash.

Another possibility would be to filter the emdash in a LuaTeX callback and replace it with something breakable.

Answer 2

Imho from the unicode view there shouldn't be automatically a break point after such dashes. With unicode you should probably insert a Zero Width Space (U+200B). It doesn't work out-of-the-box with lualatex, you need something like luaunicodespace (https://github.com/khaledhosny/luaunicodespace):

\documentclass[]{scrbook}
\usepackage{fontspec}
\directlua {
require('luaunicodespace')
luatexbase.add_to_callback("pre_linebreak_filter", luaunicodespace.handler, "luaotfload", 1)
luatexbase.add_to_callback("hpack_filter",         luaunicodespace.handler, "luaotfload", 1)
}
\textwidth=3cm
\begin{document}
Text—Text—Text—Text—Text—Text—Text—Text

Text—^^^^200bText—^^^^200bText—^^^^200bText—^^^^200bText—^^^^200bText—^^^^200bText—^^^^200bText

\end{document}

Answer 3

See the last paragraph here: http://www.heise.de/newsticker/meldung/TeX-Live-Jahrgang-2012-1634828.html, you have to set the parameter \XeTeXdashbreakstate to null explicitly starting TeX 2012.