Remap characters for \setminus and \smallsetminus

[jakubkaczor]

Status

READY

Description

STIX Two has been updated. (stipub/stixfonts#179) Therefore, this change achieves Unicode compliance without producing inconsistencies with STIX. This is regarding issue #181.

Todo

• [X] Tests added to cover new/fixed functionality
• [X] Documentation added if necessary
• [X] Code follows expl3 style guidelines

Minimal example demonstrating the new/fixed functionality

\documentclass{article}
\usepackage{unicode-math}
\setmathfont{texgyrepagella-math.otf}% filename only please!
\begin{document}
\[
  \setminus, \smallsetminus
\]
\end{document}

[wspr]

Sorry for the slow reply. To do this taking into account variations between fonts, wouldn't we need to have a package option or font-loading option to choose whether or not to switch these?

[jakubkaczor]

I am sorry, but after this time I can't remember the entire picture. I believe I based this PR on the following. You once said: “The unicode-math definitions come from the STIX table and match up with the STIX fonts (see below), which are the reference implementation for unicode math fonts”^1. The STIX fonts has been updated, therefore unicode-math has to be updated if STIX is supposed to be a “reference implementation”. I only considered defaults, not a way of handling different fonts.

[davidcarlisle]

\documentclass{article}
\usepackage{fontspec}
\newfontface\lm{lmroman10-regular.otf}
\newfontface\lmm{latinmodern-math.otf}
\newfontface\tgp{texgyrepagella-regular.otf}
\newfontface\tgpm{texgyrepagella-math.otf}
\newfontface\cbia{Cambria}
\newfontface\cbiam{Cambria Math}
\newfontface\sx{Stix Two Text}
\newfontface\sxm{Stix Two Math}
\newfontface\ncm{NewCM10-Book.otf}
\newfontface\ncmm{NewCMMath-Book.otf}
\newfontface\fira{FiraSans-Regular.otf}
\newfontface\firam{FiraMath-Regular.otf}

\usepackage{amssymb}
\begin{document}

{\tiny
U+005C REVERSE SOLIDUS

U+2216 SET MINUS

U+27CD MATHEMATICAL FALLING DIAGONAL

U+29F5 REVERSE SOLIDUS OPERATOR

U+FE68 SMALL REVERSE SOLIDUS

}
\def\x#1 #2 #3:#4\relax{#3}
\newcommand\test[2]{
#1 \expandafter\expandafter\expandafter\x\expandafter\meaning\the\font\relax
&#1  \char"005C   &#1  ^^^^29f5 &#1  ^^^^2216  &#1  ^^^^27cd &#1  ^^^^fe68  \\
#2 \expandafter\expandafter\expandafter\x\expandafter\meaning\the\font\relax
&#2  \char"005C   &#2  ^^^^29f5 &#2  ^^^^2216  &#2  ^^^^27cd &#2  ^^^^fe68  \\[10pt]
}

\begin{tabular}{llllll}
    &005C             &2216         &27CD          &29F5         &FE68        \\
amssymb &$\backslash$&$\setminus$& ? &$\smallsetminus$&?\\[10pt]
\test\lm\lmm
\test\tgp\tgpm
\test\cbia\cbiam
\test\sx\sxm
\test\ncm\ncmm
\test\fira\firam
\end{tabular}

\end{document}

Which is fairly depressing, none of the OTF really has a small set minus apart from NCM but at least U+29F5 is more slanted so that's a better map for smallsetminus in the fonts where it exists. So I think that should be the used were possible but neither is usable as a straight default setting I think the settings need to be guarded by tests to see if the characters exist in the font and set them to something that exists in all cases.

[davidcarlisle]

@wspr perhaps something like this, modify the table as in this PR but at the end make sure both commands point to something

\documentclass{article}

\usepackage{unicode-math}
\AtBeginDocument{%
% from (new) unicode-math-table
%\UnicodeMathSymbol{"02216}{\setminus                 }{\mathbin}{reverse solidus operator}%
%\UnicodeMathSymbol{"029F5}{\smallsetminus            }{\mathbin}{small set minus (cf. reverse solidus)}%
\iffontchar\textfont2 "02216 \else
  \let\setminus \backslash
\fi
\iffontchar\textfont2 "029F5 \else
  \let\smallsetminus\setminus 
\fi
}

%\setmathfont{NewCMMath-Book.otf}
%\setmathfont{FiraMath-Regular.otf}
\setmathfont{Stix Two Math}

\begin{document}

$[\backslash][\setminus][\smallsetminus]$
\end{document}

[wspr]

Many thanks for the contributions and input here. I've taken this on board but implemented it slightly differently — update to come.