Bad vertical alignment with double-line inference rule

Question

I'm trying to write a bi-directional inference rule with a double horizontal line, using mathpartir. The documentation says I should be able to use \mprset{fraction={===}} as follows:

\documentclass{article}

\usepackage{mathpartir}

\begin{document}
\begin{mathpar}
\mprset{fraction={===}}
\inferrule*
{
}
{
  C \rightarrow C'
}
\and
\inferrule*
{
  C \rightarrow C
}
{
  C \rightarrow C'
}
\and
\inferrule*
{
  C \rightarrow C_{\beta}
}
{
  C \rightarrow C'
}
\end{mathpar}
\end{document}

This does indeed format the rule with a double horizontal line, but it also breaks the vertical alignment, so that, in the last example above, the C \rightarrow C' now overlaps with the double line.

See also here, Frac or inference rule with dotted line, but I would like a solution based on inferrule if possible.

Answer 1

Three parameters to tune:

\eqgap distance between double rule lines;

\overgap separation between the text and the rules; and

\inferrulerule thickness of two rule lines.

This MWE shows it used in math mode, but it works outside of math mode, too. I don't know enough about inference rules to know, but if the math axis is aligned with the "denominator", then this MWE would work.

\documentclass{article}
\usepackage{stackengine}
\def\eqgap{.2ex}
\def\overgap{.4ex}
\def\inferrulerule{.2pt}

\newlength\rulelength
\newlength\toplength
\newlength\bottomlength

\newcommand\myinferrule[2]{%
  \stackMath%
  \setlength\bottomlength{\widthof{$#1$}}%
  \setlength\toplength{\widthof{$#2$}}%
  \ifdim\toplength>\bottomlength%
    \setlength\rulelength{\the\toplength}%
  \else%
    \setlength\rulelength{\the\bottomlength}%
  \fi%
  \mathrel{%
  \stackon[\overgap]{\stackon[\eqgap]{\stackon[\overgap]{#1}%
    {\rule{\the\rulelength}{\inferrulerule}}}%
    {\rule{\the\rulelength}{\inferrulerule}}}{#2}%
  }%
}
\begin{document}
\[
\mathrm{Math~Axis} \myinferrule{C \rightarrow C'}{C \rightarrow C_{\beta}}
\]
\end{document}

If the math axis is supposed to align with the rule, then this MWE would apply:

\documentclass{article}
\usepackage{stackengine}
\def\eqgap{.2ex}
\def\overgap{.4ex}
\def\inferrulerule{.2pt}

\newlength\rulelength
\newlength\toplength
\newlength\bottomlength

\newcommand\myinferrule[2]{%
  \stackMath%
  \setlength\bottomlength{\widthof{$#1$}}%
  \setlength\toplength{\widthof{$#2$}}%
  \ifdim\toplength>\bottomlength%
    \setlength\rulelength{\the\toplength}%
  \else%
    \setlength\rulelength{\the\bottomlength}%
  \fi%
  \mathrel{%
    \stackunder[\overgap]{%
      \stackon[\overgap]{%
        \stackanchor[\eqgap]%
          {\rule{\the\rulelength}{\inferrulerule}}%
        {\rule{\the\rulelength}{\inferrulerule}}%
      }{#2}%
    }{#1}%
  }%
}
\begin{document}
\[
\mathrm{Math~Axis} \myinferrule{C \rightarrow C'}{C \rightarrow C_{\beta}}
\]
\end{document}

With these parameters, the following

\[
\myinferrule{C \rightarrow C'}{C \rightarrow C_{\beta}} ~
\myinferrule{\scriptstyle C \rightarrow C'}{\scriptstyle C \rightarrow C_{\beta}} ~
\myinferrule{\scriptscriptstyle C \rightarrow C'}{\scriptscriptstyle C \rightarrow C_{\beta}}
\]

typesets to different math styles:

If you need to regularly set this to different math styles, I can make it (at the expense of extra coding) so that the style need not be set within the arguments themselves, as it is here. So let me know if that's the case.

Answer 2

The best thing to do is pick up the new version of mathpartir, in which this problem as been fixed. In reponse to my bug report, the author of mathpartir responded as follows:

I pushed a new version 1.2.1 where the fraction macro is handled a bit differently and I hope behaves much better with respect to vertical spacing.

http://cristal.inria.fr/~remy/latex/mathpartir.sty
http://cristal.inria.fr/~remy/latex/mathpartir.tex
http://cristal.inria.fr/~remy/latex/mathpartir.pdf

I also added options to adjust vertical spacing in case it does not work properly.

This seems to work as I would expect, and improves on either solution proposed above. The author does not have a Stack Exchange account.