mhchem vs expl3 bug

Question

Does anybody know how to fix this problem with mhchem package?

! Undefined control sequence.
<argument> \l__mhchem_cg_isScript_bool

Recently I've found this part of mhchem.sty code:

\cs_new_protected:Npn \__mhchem_output_escapeFromTextToItalicMath:n #1
  {
    \ensuremath
      {
        \mhchem@hook@beforeItalicMath
        \bool_if:NT \l__mhchem_cg_isScript_bool
          {
            \medmuskip=0mu
            \thinmuskip=0mu
            \thickmuskip=0mu
          }
        #1
      }
  }

It seems to me problem is connected with \bool_if:NT \l__mhchem_cg_isScript_bool condition which can be wrong with newest expl3 package. The same bug has successfully been solved in siunitx package.

Win 10, MikTeX 2.9.6236, up-to-date

https://listserv.uni-heidelberg.de/cgi-bin/wa?A2=latex-l;3fb6e8f0.1707 — Henri Menke, Jul 24 '17 at 10:47
The siunitx problem has been solved because I notified the author: https://github.com/josephwright/siunitx/issues/311. So if you want to get a fix for mhchem: write to the author. — Ulrike Fischer, Jul 24 '17 at 11:10
Thank you. I sent him an e-mail with link to forum's question. Seems simple replacement bool<-bool_lazy can't solve this :) — Alexey, Jul 24 '17 at 11:33
Temporary, I make this: code \bool_new:N \l__mhchem_cg_isScript \cs_new_protected:Npn __mhchem_output_escapeFromTextToItalicMath:n #1 { \ensuremath { \mhchem@hook@beforeItalicMath \bool_lazy_any:nT \l__mhchem_cg_isScript { \medmuskip=0mu \thinmuskip=0mu \thickmuskip=0mu } #1 } } — Alexey, Jul 24 '17 at 11:44
The problem is that the boolean is not defined. Probably in the past an undefined boolean was considered false. This is quite likely a left-over from some older version. By the way, setting \thinmuskip to zero is wrong. — egreg, Jul 24 '17 at 11:48
The \bool_if:NT code was added in version 4.04, but not accompanied by a definition of the boolean variable, nor with any way to toggle it. Hence it is completely useless. — egreg, Jul 24 '17 at 11:57
@egreg, the code \thinmuskip=0mu is not mine, but your suggestion about declaration is right. I've made it like \bool_new:N \l__mhchem_cg_isScript — Alexey, Jul 24 '17 at 11:58
@Alexey: your last comment should have \bool_new:N \l__mhchem_cg_isScript_bool. — Bruno Le Floch, Jul 24 '17 at 13:26

score 10 · Accepted Answer · answered Jul 24 '17 at 12:16

The code \bool_if:NT \l__mhchem_cg_isScript_bool {...} was added in version 4.04 of mhchem.sty. However, the addressed boolean variable is nowhere defined.

With previous versions of expl3, such a case was dealt with by considering the boolean as false, now it is not so any longer.

By the way, as there obviously is no way to toggle the boolean's value, the code is completely useless (and it would be so even if the boolean was defined).

Here's a temporary fix, which wouldn't be invalid even after the package author defines the boolean variable.

\documentclass{article}
\usepackage[version=4]{mhchem}

\ExplSyntaxOn
\bool_if_exist:NF \l__mhchem_cg_isScript_bool
 {
  \bool_new:N \l__mhchem_cg_isScript_bool
 }
\ExplSyntaxOff

\begin{document}

$\ce{NO_$x$}$

$\ce{NO_$x+y$}$

\end{document}

score 7 · Answer 2 · edited Jul 25 '17 at 07:28

7

mhchem v4.07 fixes this bug.

Thanks a lot for reporting this and all your research into that matter. Thanks @egreg! This seems indeed to be a leftover of an experiment I did once and did not completely reverse. And my (wrong) code stopped working with an expl3 update. I will publish an update of mhchem during the next couple of days.

edited Jul 25 '17 at 07:28

Schweinebacke

26,336

answered Jul 24 '17 at 13:59

mhchem

3,485
12
37

5

Please do not add thank you as answer. Once the update is published you can post an answer showing how it works with the updated version. – TeXnician Jul 24 '17 at 16:05
@TeXnician This is no thank you answer. It is an "I noticed this and I am working on an update". Am I not allowed to say thank you on the way? --> meta – mhchem Jul 24 '17 at 18:44
4

Well, it is. The appropriate answer would be: "It's fixed in version X (problem was...) and this is how it works..." The rest is commentary. – TeXnician Jul 24 '17 at 18:49
You can use comments if you need to say thank you not only using the up-vote button. – Schweinebacke Jul 25 '17 at 07:26

Alexey · Answer 3 · 2017-07-24T14:49:56.653

0

I've fixed it directly in mhchem.sty:

\bool_new:nT \l__mhchem_cg_isScript_bool
\cs_new_protected:Npn \__mhchem_output_escapeFromTextToItalicMath:n #1
  {
    \ensuremath
      {
        \mhchem@hook@beforeItalicMath
        \bool_lazy_any:nT \l__mhchem_cg_isScript_bool
          {
            \medmuskip=0mu
            \thinmuskip=0mu
            \thickmuskip=0mu
          }
        #1
      }
  }

edited Jul 24 '17 at 14:49

answered Jul 24 '17 at 12:39

Alexey

415

4

The name of a boolean variable should end with _bool. Besides, changing package code is not the best way to solve such issues. On another machine the document would choke again. – egreg Jul 24 '17 at 12:40
@egreg, Ok. Also, I'm glad to see the author of this package here. – Alexey Jul 24 '17 at 14:52

mhchem vs expl3 bug

3 Answers3

Linked