OMML should write dPr in the order (begChr, sepChr, endChr), not (begChr, endChr, sepChr)

[refashioned]

When I generate a docx file containing math with Pandoc (v3.5) and open it in LibreOffice Writer (v7.4.7.2), every right-hand delimiters turns into right parenthesis, even when it could be something else, such as a square bracket or vertical bar. I tracked the problem down to the order in which TeXMath writes the subelements of the dPr element. TeXMath uses the order (begChr, endChr, sepChr), while LibreOffice is expecting (begChr, sepChr, endChr). The right parenthesis is a default fallback in the LibreOffice code.

This is the code in TeXMath that writes the elements in the order (begChr, endChr, sepChr):
/~https://github.com/jgm/texmath/blob/0.12.8.11/src/Text/TeXMath/Writers/OMML.hs#L209-L216

   EDelimited start end xs ->
                  [ mnode "d" $ mnode "dPr"
                               [ mnodeA "begChr" (T.unpack start) ()
                               , mnodeA "endChr" (T.unpack end) ()
                               , mnodeA "sepChr" (T.unpack sepchr) ()
                               , mnode "grow" () ]
                              : map (mnode "e" . concatMap (showExp props)) es
                  ]

This is the code in LibreOffice that expects the order (begChr, sepChr, endChr) and supplies ")" as a default:
https://git.libreoffice.org/core/+/refs/tags/libreoffice-24.8.3.2/starmath/source/ooxmlimport.cxx#301

    OUString opening = u"("_ustr;
    OUString closing = u")"_ustr;
    OUString separator = u"|"_ustr;
    if( XmlStream::Tag dPr = m_rStream.checkOpeningTag( M_TOKEN( dPr )))
    {
        if( XmlStream::Tag begChr = m_rStream.checkOpeningTag( M_TOKEN( begChr )))
        {
            opening = begChr.attribute( M_TOKEN( val ), opening );
            m_rStream.ensureClosingTag( M_TOKEN( begChr ));
        }
        if( XmlStream::Tag sepChr = m_rStream.checkOpeningTag( M_TOKEN( sepChr )))
        {
            separator = sepChr.attribute( M_TOKEN( val ), separator );
            m_rStream.ensureClosingTag( M_TOKEN( sepChr ));
        }
        if( XmlStream::Tag endChr = m_rStream.checkOpeningTag( M_TOKEN( endChr )))
        {
            closing = endChr.attribute( M_TOKEN( val ), closing );
            m_rStream.ensureClosingTag( M_TOKEN( endChr ));
        }
        m_rStream.ensureClosingTag( M_TOKEN( dPr ));
    }

I'm not an expert at OMML or XML Schemas, but I found this page that puts the contents of dPr in a sequence, which I take to mean that sepChr should precede endChr:

<complexType name="CT_DPr">
	<sequence>
	<element name="begChr" type="CT_Char" minOccurs="0"/>
	<element name="sepChr" type="CT_Char" minOccurs="0"/>
	<element name="endChr" type="CT_Char" minOccurs="0"/>
	<element name="grow" type="CT_OnOff" minOccurs="0"/>
	<element name="shp" type="CT_Shp" minOccurs="0"/>
	<element name="ctrlPr" type="CT_CtrlPr" minOccurs="0"/>
	</sequence>
</complexType>

This is a small reproduction recipe:

echo '$|x| = \max[x, -x]$' | pandoc -t docx -o x.docx

In my LibreOffice Writer, it renders like this:

When I double-click the formula, this is how it's serialized in the editor:

left lline x right ) = max left [x , − x right )

OMML output

<m:oMath>
  <m:d>
    <m:dPr>
      <m:begChr m:val="|" />
      <m:endChr m:val="|" />
      <m:sepChr m:val="" />
      <m:grow />
    </m:dPr>
    <m:e>
      <m:r>
        <m:t>x</m:t>
      </m:r>
    </m:e>
  </m:d>
  <m:r>
    <m:rPr>
      <m:sty m:val="p" />
    </m:rPr>
    <m:t>=</m:t>
  </m:r>
  <m:r>
    <m:rPr>
      <m:sty m:val="p" />
    </m:rPr>
    <m:t>max</m:t>
  </m:r>
  <m:d>
    <m:dPr>
      <m:begChr m:val="[" />
      <m:endChr m:val="]" />
      <m:sepChr m:val="" />
      <m:grow />
    </m:dPr>
    <m:e>
      <m:r>
        <m:t>x</m:t>
      </m:r>
      <m:r>
        <m:rPr>
          <m:sty m:val="p" />
        </m:rPr>
        <m:t>,</m:t>
      </m:r>
      <m:r>
        <m:rPr>
          <m:sty m:val="p" />
        </m:rPr>
        <m:t>−</m:t>
      </m:r>
      <m:r>
        <m:t>x</m:t>
      </m:r>
    </m:e>
  </m:d>
</m:oMath>

If I hack word/document.xml inside the .docx file and swap the order of endChr and sepChr, then the delimiters render correctly.

[jgm]

Thanks for the excellent bug report!