What is the correct tagging for table cells that span multiple rows?

[mbertucci47]

Here's an example that compares the tagging structure tree of the kernel \multicolumn with the multirow package's \multirow.

\DocumentMetadata
  {
    lang=en-US,
    pdfversion=2.0,
    pdfstandard=ua-2,
    testphase={phase-III,math,title,table,firstaid}
  }
\documentclass{article}
\usepackage{multirow}

\begin{document}

\begin{tabular}{cc}
\multicolumn{2}{c}{AAA} \\
C & D
\end{tabular}

\begin{tabular}{cc}
\multirow{2}{*}{AAA} & B \\
& D
\end{tabular}

\end{document}

As you can see, \multicolumn gives a single cell for the first row, while \multirow produces two cells: in the first row, the content, and in the second row, an empty cell.

For a properly tagged table, what is the correct tagging for a cell than spans multiple rows? Should it match the behavior of \multicolumn but "transposed", or is the current tagging correct? My guess is that the current behavior is correct because otherwise there is no way to tell which column the "D" in the example should appear in.

P.S. I'm aware no attempt has been made thus far to make \multirow tagging correct. I am interested in whether the "out of the box" behavior is correct.

[zauguin]

The table model in PDF tagging mostly follows HTML here, so the correct behavior is similar to how this would work in HTML: The \multirow only produces one cell in the first row, this cell contains a RowSpan attribute specifying that it spans two rows. The second row contains only the D column. The placement of D in the second column is implicitly given based on the context that the first column is already occupied (by the column from the first row).

[mbertucci47]

Thanks @zauguin. So should the \multicolumn also have an analogous ColSpan attribute? (Also, how does one inspect such attributes? It's not clear to me how to do this from the Acrobat structure tree.)