changed response to \fam\z@ command (from dchem) in XeLatex/mathspec

Question

I am a dedicated user of Donald Arsenau's dchem package, which has been distributed privately for a long time but which he has never felt was sufficiently developed to be released publicly (or, if that is no longer true, a public release has apparently not floated high enough on his priority list to get done.)

I am now trying to switch to xelatex/mathspec. Unfortunately I find that this change has broken the main \ch command of the dchem package. I wrote to D. Arsenau privately, and have asked on c.t.t., but I also thought I might try here to find advice.

I have tried to reduce the problem to an absolute minimum. I therefore extracted a single line of dchem.sty, and renamed it gchem.sty:

\newcommand\ch{\fam\z@\everymath{\fam\z@}}
% like \rm but don't have to repeat 

This command is typically used like $\ch H_2O$ to produce chemical formulas set in roman (more accurately, mathrm) type.

This line falls within some complex conditionals in dchem.sty that try to determine something about the current latex environment and its \everymath, and the implementation is somewhat different depending on the conditional branches, but I believe that this single line is the essence of all versions of the \ch command.

Now, I write a simple test file using the Fourier package:

\documentclass{article}
\usepackage{fourier}
\usepackage{gchem}
\begin{document}
$\rm CH_2O$, $\ch CH_2O$
\end{document}

When I process this package with pdflatex, both the formula produced with \rm and the formula produced with \ch come out in Utopia as expected.

I then modify the test file to use xelatex/mathspec:

\documentclass{article}
\usepackage{mathspec}
\setmainfont{Minion Pro}
\setmathsfont(Digits,Latin,Greek){Minion Pro}
\setmathrm{Minion Pro}
\usepackage{gchem}
\begin{document}
$\rm CH_2O$, $\ch CH_2O$
\end{document}

When I process this file with xelatex (This is XeTeX, Version 3.1415926-2.3-0.9997.5 (MiKTeX 2.9), the \rm formula comes out in Minion Pro as expected but the \ch version comes out in Computer Modern except for the subscript 2 which is Minion Pro.

I have log files, including ones with tracing turned on, that show distinct differences between the behavior of \rm and \ch, and I can post or forward them if needed.

But I think these examples get the point across: it seems that the short \fam\z@ command used by dchem does effect a switch to \mathrm in pdflatex but does not do so in xelatex. Is this expected? If not, where is the problem? If so, how should the \ch command be modified to be compatible with xetex?

Answer 1

mathspec doesn't change the basic math group 0, but assigns new math alphabets or math symbol fonts. One that should correspond to upright characters is the one used for digits, so

\makeatletter
\newcommand\ch{\fam\csname sym\eu@DigitsArabic@symfont\endcsname
  \everymath{\fam\csname sym\eu@DigitsArabic@symfont\endcsname }}
\makeatother

might be what you need.

Let's see what happens. At a certain point, mathspec declares

\DeclareSymbolFont{Digits:m:n}{EU1}{\eu@Digitsmathsfont}{m}{n}

and \eu@Digitsmathsfont is translated into the correct family name. This defines a new math group (called a family in the TeXbook), that LaTeX accesses by adding the string sym to Digits:m:n. But it's safer, in my opinion, to use \eu@DigitsArabic@symfont, that will be the right string in any case, independently of the options given at package loading.

So what we do is

\fam\csname sym\eu@DigitsArabic@symfont\endcsname

Since \fam expects a number after it, it will expand \csname...\endcsname; this causes the expansion of \eu@DigitsArabic@symfont (in your case to Digits:m:n) and the formation of the token

\symDigits:m:n

that is equivalent to the number assigned by \DeclareSymbolFont to the desired math group.