A selection of tagged tables

Normal Table

Each non-blank⟨TD⟩ data cell has a well-defined row- and column-head cell,⟨TH⟩, as indicated by the blue lines in the following annotated image.

Normal Table with 1st column header: super row-header

When the top-left cell is non-empty, it is usually a column-header for the 1st column, which consists of row-headers. This is an example of a ‘ super header ’; that is, a header cell for other header cells . The /Scope value for this header cell is /Column , since it is a descriptor for the cells in the column placed below it. We call this a ‘row super-header’ cell, as it is located from a data-cell by tracing the headers of its row-header. Typically a row super-header cell will occur within the 1st column, though need not always be in the 1st row of the table.

When the rows would be too wide for a page, the table can be turned sideways. See the next A page for the table resulting from the following LT X source. It is rather unfortunate that EAcrobat Pro cannot correctly find the bounding boxes of the cells when a table is rotated. Although the PDF is correctly tagged, screen-reading and navigation using Assistive Technology may be affected.

Source code for a sideways table with all row and column headers

\begin{sidewaystable}[htp] \caption{Stocks reviewed at September 2021 Management Track Assessment Peer Review meeting} \centering \Cut{columnRowComment}{\columnRowTypes{1}{2}{1}{2}}% \begin[\Paste{columnRowComment}]{tabular}{p{3.75cm} p{4.5cm} p{7.75cm}}% \toprule Stock & Lead Analyst/Presenter & Peer review conclusion \\\midrule \relax\hyperlink{AtlanticCod}{Gulf of Maine cod}\newline~\enDash\,Enhanced review & \newline Charles Perretti & Under retro-adjusted $\M\!=\!0.2$ model stock is overfished and overfishing is occurring. \newline Under \Mramp\ model stock is overfished but overfishing is not occurring. \\[28pt] \relax\hyperlink{AtlanticCod}{Georges Bank cod}\newline~\enDash\,Expedited review & \newline Kathy Sosebee & \newline Stock is overfished but overfishing unknown \\\bottomrule \end{tabular} \end{sidewaystable}

Table 2: Stocks reviewed at September 2021 Management Track Assessment Peer Review meeting

Stock	Lead Analyst/Presenter	Peer review conclusion
Gulf of Maine cod – Enhanced review	Charles Perretti	Under retro-adjusted M=0.2model stock is overfishe d and overfishing is occurring. UnderM-rampmodel stock is overfished but overfishing is not occurring
Georges Bank cod – Expedited review	Kathy Sosebee	Stock is overfished but overfishing unknown

Acrobat seems to have difficulty correctly indicating the tagged material with rotation in effect.

There is no problem with the actual tagging tree; so not with text extraction for accessibility, etc.

Multiple Row Blocks

Here the data rows are divided into blocks (3 of them) each with a header cell that spans a whole row, using the \multicolumn{9}{...} command, which includes the 1st column. This creates a header cell which is essentially the column header for the 1st column (consisting of row-headers). Having a blank 1st cell in the top row is necessary for this interpretation.

For greater emphasis of the different data blocks, a \midrule can precede the super-head cell, as shown below. Notice also how the /Scope of a super row-header is /Column , as its content is a qualifier for the row-header cells in the same (that is, the 1st) column, at least for the rows within the following⟨TBody⟩ block.

Multiple Header Blocks

This example has multiple data blocks, where there is a header block explicitly specified for each. For comparison of different model results, groups of columns are associated with the different models. This gives ‘super’ column header cells, with textM =0.2 andM-ramp, spanning multiple columns using \multicolumn commands. Notice that the 1st column entry is blank on the rows where these occur, and that a \midrule comes afterwards, rather than before (as in the previous example).

A ‘super header’ row occurs within the⟨THead⟩ block, having a \midrule beneath. The 2nd header block is tagged as⟨TBody⟩, with a \toprule above and \midrule below. As these ‘super header’ cells come after their column-header cells, their presence must be recorded in the history/auxiliary file. This enables them to be specified as a header for the real column headers on the next run.

Hierarchical column headers

This next example is taken from T X StackExchange E . It has a binary tree of column-header cells.

Use of \multicolumn within the⟨THead⟩ in such a (binary) tree-like fashion allows the whole structure to be deduced in a single pass. Subsets of columns for one row are refined into smaller subsets on the next row. Each larger subset establishes a ‘super-column header’ for all the header cells below it. Singleton sets define the column-header for the data cells in that column. Thus each cell, head or data, has a well-defined column header at the time it is processed.

Multiple row-header columns

A calendar, or Meeting schedule, also has structured row headers, but only needs a single⟨TBody⟩ block. Row headers occur within the 2nd column, with the 1st column having the occasional ‘super’ row-header amongst mostly blank cells.

Redundant/weak column headers

It is poor style to have long-winded headers, perhaps using multiple cells instead of just a single cell. However, this is not a good reason for failing to correctly identify what is a header for what other cell(s).

Here is such an example, based upon one of the earlier tables (now set in portrait).

The figure below shows how there are really 3 rows of headers, having content that looks like a single phrase when read vertically. With each new successive row, each non-empty cell inherits the rôle of being a column header, from the cell immediately above.

A drawback with this approach is that the information from these (weak) header cells is collected in reverse of the natural reading order. Thus it would be conceptually better to use single p-style cells, within a single row of header cells.