\input minim-doc.sty

\manual{minim-pdf}

\ifchapter \chapter Hypertext

This chapter and the next document the support of the modern pdf features 
provided by the minim-pdf package.
Load it by saying ⟦\input minim-pdf⟧.
The next chapter concerns the creation of tagged pdf;
all other features of the package are described here.

\else \startmetadata
    author   {Esger Renkema}
    title    {minim-pdf}
    date     {2025-02-22}
    version  {2025/1.7}
    keywords {LuaTeX; Plain TeX; PDF/A; Tagged PDF; accessibility; a11y;
              hypertext; bookmarks; document outline; associated files}
    stopmetadata
\maketitle

This package adds low-level support to plain luatex for marking up the 
structure of a pdf document. The implementation is rather basic, but should 
allow you to make your pdfs fully pdf/a-compliant.
Load the package by saying ⟦\input minim-pdf⟧.

The creation of tagged pdf will be described in the second half of this manual; 
all other features will be covered first.

\fi

\section Hyperlinks %1

Hyperlinks can be made with
⟦*\hyperlink [alt {...}] [attr {...}] <action> ... \endlink⟧, where the 
⟦<action>⟧ must be one of
⟦name {...} | url {...} | name {...} | next | prev | first | last⟧
With the ⟦name⟧ action, a named destination must be used (see below), while the 
⟦user⟧ action will be passed directly to the back-end (as with the pdftex 
primitive).
After the ⟦url⟧ action, the characters \Uchar`\~, \# and \% need not be escaped.
(Of course, this does nothing for already-tokenised text; be aware of this when 
you wrap ⟦\hyperlink⟧ into another macro.)
Any spaces after the ⟦<action>⟧ will be ignored.

The ⟦*\hyperlinkstyle⟧ token list can be used so set common (pdf) link 
attributes; it defaults to ⟦/Border [0 0 0]⟧. The contents of the optional 
⟦attr⟧ parameter will be appended to these.
The ⟦alt⟧ options sets the ⟦/Contents⟧ key that is required by PDF/UA (where it 
has the purpose of an alt text).

A named destination can be created with ⟦*\nameddestination {...}⟧ (also in 
horizontal mode, unlike the backend primitive) and if you cannot think of 
a name, ⟦*\newdestinationname⟧ should generate a unique one.
If you need the latter twice, ⟦*\lastdestinationname⟧ gives the last 
generated name.


\section Bookmarks %1

Bookmarks (also known as outlines) can be added with
⟦*\outline [level n] [open|closed] [dest {name}] {title}⟧.
Add ⟦open⟧ or ⟦closed⟧ to have the bookmark appear initially open or closed 
(the default), and
say ⟦dest {name}⟧ for having it refer to a specific named destination 
(otherwise, a new one will be created where the ⟦\outline⟧ command appears).

In the absence of the ⟦level⟧ option, the bookmark is automatically associated 
with the current structure element and the hierarchy of structure elements 
determines the nesting of bookmarks. This works even if you have otherwise 
disabled tagging and is the recommended way of generating outlines.
(You can find all relevant macros in next chapter under ‘Document structure’ 
and ‘Structure element aliases’.)

As a fallback, outlines specified with the ⟦level n⟧ option will be inserted at 
the end of the current outline list at the specified level (${\tt n}≥1$ and 
need not be contiguous). Both methods can be intertwined, but please use the 
document structure if you can.


\section Page labels %1

If the page numbers of your document are not a simple sequence starting with~1, 
you can use ⟦*\setpagelabels [pre {prefix}] style nr⟧ for communicating 
this to the pdf viewer.
This command affects the page labels from the next page on: ⟦nr⟧ should be the 
numerical page number of that page.
The ⟦prefix⟧ is prepended to each number and the 
⟦style⟧ must be one of ⟦decimal⟧, ⟦roman⟧, ⟦Roman⟧, ⟦alphabetic⟧, ⟦Alphabetic⟧ 
or ⟦none⟧. In the last case, only the prefix is used.


\section PDF/A %1

You can declare pdf/a conformance with ⟦*\pdfalevel xy⟧, with version $\tt 
x ∈ \{1,2,3\}$ and conformance level $\tt y ∈ \{a,b,u\}$. This will set the 
correct pdf version and ⟦pdfaid⟧ metadata. If the conformance level is ‘⟦a⟧’, 
tagging will be enabled (see the next chapter). Finally, a default RGB colour 
profile will be included.
The conformance level can be queried from the ⟦*\pdfaconformancelevel⟧ register.

Note that merely declaring conformance will not make your document pdf/a 
compliant, and that minim will not warn you if it is not.
However, the features described in this chapter and the next should be enough 
to make pdf/a compliance possible.

Also note that there currently is no documented way of choosing a different 
colour profile from the default (i.e. the default rgb profile provided by the 
colorprofiles package).
Should you need do that, you will have to do so manually, after
disabling the automatic inclusion by saying
⟦\expandafter\let \csname minim:default:rgb:profile\endcsname = \relax⟧.

Finally, note that pdf/a requires that spaces are represented by actual space 
characters and that discretionary hyphens are marked as soft hyphens 
(⟦U+00AD⟧).
Since both features benefit accessibility and text extraction in general, they 
are enabled by default.
You can disable them by setting ⟦*\writehyphensandspaces⟧ to a nonpositive value.


\section PDF/UA %1

You can claim pdf/ua conformance with ⟦*\pdfualevel 1⟧.
By itself, this will do very little:

\startlist
\tagattribute List ListNumbering /Decimal
\item 1. The ⟦pdfuaid:part⟧ metadata key will be set.
\item 2. A conforming ⟦ViewerPreferences⟧ dictionary will be added to the document catalog.
\item 3. The ⟦/Suspects⟧ key of the ⟦MarkInfo⟧ dictionary will be set to ⟦false⟧.
\item 4. ⟦/Tab /S⟧ will be added to the page attributes.
\stoplist

Also making your document pdf/a-compliant, however, will relieve you of a few 
additional worries:

\startlist
\tagattribute List ListNumbering /Decimal
\tagattribute List ContinuedList true
\item 5. Fonts will be included properly.
\item 6. The (natural) language of every element will be known.
\item 7. Headings will be strongly-structured.
\item 8. Table headers will have their ⟦Scope⟧ set properly.
\item 9. A document outline will be generated automatically.
\stoplist

This leaves the following for you to provide before your document can be 
pdf/ua-compliant:

\startlist
\tagattribute List ListNumbering /Decimal
\tagattribute List ContinuedList true
\item 10. Figure and Formula structure elements must have alt texts.
\item 11. Hyperlinks must have alternate descriptions.
\item 12. Lists must have the ⟦ListNumbering⟧ attribute set.
\item 13. Tables must have headers that are tagged as such.
\item 14. Page headers and footers must be marked as header or footer artifacts.
\item 15. Document section structure elements should have their ⟦title⟧ set.
\item 16. All embedded files must have a description.
\stoplist


\section Embedded files %1

You can attach (associate) files with ⟦*\embedfile <options>⟧.
The file will be attached to the current structure element (see the next 
chapter) unless the ⟦global⟧ option is given: then it will be added to the 
document catalog.
Arguments consisting of a single word can be given without braces and
exactly one of the options ⟦file⟧ or ⟦string⟧ must be present.

\smallskip\marktableaslist
\tagattribute List ListNumbering /Description
\halign{#\quad\hfil&#\hfil\cr
⟦file ⟧\hfill⟦{...}⟧ & The file to embed.\cr
⟦string ⟧\hfill⟦{...}⟧ & The string to embed.\cr
⟦global⟧ & Attach to the document catalog.\cr
⟦uncompressed⟧ & Do not compress the file stream.\cr
⟦mimetype ⟧\hfill⟦{...}⟧ & The file’s mime type.\cr
⟦moddate ⟧\hfill⟦{...}⟧\rlap{*} & The modification date (see * below).\cr
⟦desc ⟧\hfill⟦{...}⟧ & A description (the ⟦/Desc⟧ key).\cr
⟦relation ⟧\hfill⟦{...}⟧ & The ⟦/AFRelationship⟧ value as defined in pdf/a-3.\cr
⟦name ⟧\hfill⟦{...}⟧ & The file name (only required when writing a ⟦string⟧).\cr }

* The modification date must be of the form ⟦yyyy[-m[m][-d[d]]]⟧. A default 
moddate can be set with ⟦*\setembeddedfilesmoddate {default}⟧.
The ⟦default⟧ date will be expanded fully at the time of embedding.
With the minim-xmp package, a useful setting is
⟦\setembeddedfilesmoddate {\getmetadata date}⟧.

\section Lua module %1

The interface of the lua module (available via
⟦local M = require('minim-pdf')⟧) should be stable by now.
Though it contains lua equivalents for most tex commands described here, using 
them directly is not very ergonomical and not recommended.
Please consult the source if you do want to use them anyway.

% 

\chapter Tagged PDF

This chapter is a continuation of the previous and describes the parts of 
minim-pdf that concern the creation of tagged pdf. All features in this chapter 
must be explicitly enabled by setting ⟦*\writedocumentstructure⟧ to a positive 
value. This will be done automatically if you declare pdf/a conformance (see 
above).

This part of the package is rather low-level and this chapter rather technical. 
For a more general introduction to and discussion of tagged pdf, please read 
the (excellent) manual of latex’s tagpdf package.


\section Quick-start guide % 1

The minimal setup for producing tagged pdf from plain tex documents is 
something like the following:

⟦% first update all fonts to TrueType (ttf) or OpenType (otf)
\input luaotfload.sty
% ... font redefinitions omitted ...
\input minim-mp
\pdfalevel 2a % declare pdf/a conformance, enable tagging
\autotagplainoutput % update the output routine
% create section markers and counters
\sectionstructure { subsection, section, chapter }⟧

You can then update your sectioning macros to look like this:

⟦\def\section#1\par{%
    % space above
    \bigskip \goodbreak
    % structure and outline (this is the new part)
    \marksection \outline open {#1}
    % section header
    \noindent {\bf \the\chapternr.\the\sectionnr. #1}
    % space below
    \smallskip \noindent}⟧

Other macros you might have for laying out structural elements, such as tables 
or lists, should of course also be updated. The rest of this chapter describes 
the tools you can use.

Please be advised that producing tagged pdf will likely forever remain 
a fragile and error-prone process.
You should always validate the resulting pdf.
An easy-to-use and free validator is \hyperlink alt {The veraPDF website.}
url {https://verapdf.org/home}veraPDF\endlink.
For inspecting the document structure, you can use the ⟦pdfinfo⟧ utility that 
comes with the \hyperlink alt {The Poppler website.}
url {https://poppler.freedesktop.org/}Poppler\endlink\ pdf library.


\section Purpose, limitations and pitfalls %1

The main purpose of this package is semi-automatically marking up the 
(hierarchical) structure of your document, thereby creating so-called tagged pdf.
The mechanism presented here is not quite as versatile as the pdf format 
allows. The most important restriction is that all content of the document must 
be seen by tex’s stomach in the \emph{logical} order.

Furthermore, while the macros in this package are sophisticated enough that 
tagging can be done without any manual intervantion, it is quite possible and 
rather easy to generate the wrong document structure, or even cause syntax 
errors in the resulting pdf code.
You should always inspect and validate the result.

This is the full list of limitations, pitfalls and shortcomings:
\startlist
\tagattribute List ListNumbering /Decimal
\item1. Document content must be seen by tex in its logical order (although you 
can mark out-of-order content explicitly if you know what you are doing; see 
below).
\item2. The contents of ⟦\localleftbox⟧ and ⟦\localrightbox⟧ must be marked 
manually, probably as artifact.
\item3. There currently is no way of marking xforms or other pdf objects as 
content items of themselves.
\item4. The content of xforms (i.e. pdf objects created by  ⟦\useboxresource⟧) 
should not contain tagging commands.
\item5. Likewise, you should be careful with box reuse: it might work, but you 
should check.
\item6. This package currently only supports pdf~1.7 tagging and is not yet 
ready for use with pdf~2.0.
\stoplist

In order to help you debugging, some errors will refer you to the resulting pdf 
file. If you get such errors, decompress the pdf and search for the string 
‘Warning:’. It will appear in the pdf stream at the exact spot the problem 
occurs.


\section General overview %1

When speaking about tagging, we have to do with two (or perhaps three) separate 
and orthogonal tagging processes.
The first is the creation of a hierarchical \emph{document structure}, made up 
of \emph{structure elements} (SEs). The document structure describes the 
logical structure of a document, made up of chapters, paragraphs, references 
etc.
The second tagging process is the tagging of \emph{marked content items} 
(MCIs): this is the partition of the actual page contents into (disjoint) 
blocks that can be assigned to the proper structure element.
Finally, as a separate process, some parts of the page can be marked as 
\emph{artifacts}, excluding their content from both content and structure 
tagging.

When using this package, artifacts and structure elements (excluding 
paragraphs; see below) must be marked explicitly, while marked content items 
will be created, marked and assigned automatically.
There is some (partial and optional) logic for automatically arranging 
structure elements in their correct hierarchical relation.

The mechanism through which this is achieved uses attributes and whatsits for 
marking the contents and borders of SEs, MCIs and artifacts. At the end of the 
output routine, just before the pdf page is assembled, this information will be 
converted into markers inserted in the pdf stream.


\section Marked content items %1

Content items are automatically delineated at page, artifact and structure 
element boundaries and terminated at paragraph or display skips.
This should relieve you from any manual intervention. However, if you run into 
problems, the commands below might be helpful.

Use of ⟦ActualText⟧, ⟦Alt⟧ or ⟦Lang⟧ attribute on MCIs, while allowed by the 
pdf standard, is not supported by this package. You should set these on the 
structure element instead.

The beginning and ending of a content item can be forced with 
⟦*\startcontentitem⟧ and ⟦*\stopcontentitem⟧, while ⟦*\ensurecontentitem⟧ will 
only open a new content item if you are currently outside any.
If you need some part to be a single content item, you can use
⟦*\startsinglecontentitem ... \stopsinglecontentitem⟧.
This will disable all SE and MCI tagging inside.

Tagging (both of MCIs and SEs) can be disabled and re-enabled locally with 
⟦*\stoptagging⟧ and ⟦*\starttagging⟧.


\section Artifacts %1

Artifacts can be marked in two ways: with ⟦*\markartifact {type} {...}⟧ or with
⟦*\startartifact {type} ... \stopartifact⟧.
The ⟦type⟧ is written to the pdf attribute dictionary directly, so that if you 
need a subtype, you can write e.g.
⟦\startartifact {Pagination /Subtype/Header} etc⟧.

Inside artifacts, other structure content markers will be ignored. Furthermore, 
this package makes sure artifacts are never part of marked content items, 
automatically closing and re-opening content items before and after the 
artifact.
While the pdf standard does not require the latter, not enforcing this seems to 
confuse some pdf software.


\section Document structure %1

Like artifacts, structure elements can be given as ⟦*\markelement {Tag} {...}⟧ 
or ⟦*\startelement {Tag} ... \stopelement {Tag}⟧.
Here, in many cases the ⟦\stopelement⟧ is optional: whenever opening an element 
would cause a nesting of incompatible ⟦Tag⟧s, the current element will be 
closed until such a nesting is possible.
Thus, opening a ⟦TR⟧ will close the previous ⟦TR⟧, opening an ⟦H1⟧ will 
automatically close any open inline or block structure elements, opening 
a ⟦TOCI⟧ will close all elements up until the current ⟦TOC⟧ etc. etc.

As a special case, the tags ⟦Document⟧, ⟦Part⟧, ⟦Art⟧, ⟦Sect⟧ and ⟦Div⟧ (and 
their aliases) will try and close all open structure elements up to and 
including the last structure element with the same tag.
(An alias will of course only match the same alias.)

While the above can greatly reduce the effort of tagging, the logic is neither 
perfect nor complete. You should always check the results in an external 
application.
Particular care should be taken when ‘skipping’ structure levels: the sequence 
chapter – subsection – section will result in the section beneath the subsection.
If you are in doubt whether an element has been closed already, you can use
⟦*\ensurestopelement {Tag}⟧ instead of ⟦\stopelement⟧ 
to prevent an error being raised.

All these helpful features can be disabled by setting 
⟦*\strictstructuretagging⟧ to a positive value. Then, every structure element 
will have to be closed by an explicit closing tag, as in xml.
In this case, ⟦\stopelement⟧ and ⟦\ensurestopelement⟧ will be equivalent.

You can query the place in the document structure of any point with 
⟦*\showdocumentstructure⟧.


\section Structure element aliases %1

New structure element tags can be created with
⟦*\addstructuretype [options] Existing Alias⟧.
This will create a new structure tag named ⟦Alias⟧ with the same properties as 
⟦Existing⟧.
The properties can be modified by specifying ⟦options⟧: these will set values 
of the corresponding entry in the ⟦structure_types⟧ table (see the lua source 
file for this package).
Any aliases you declare will be written to the pdf’s ⟦RoleMap⟧ only if they 
have actually been used.


\section Automatic tagging of paragraphs %1

By default, ⟦P⟧ structure elements are inserted automatically at the start of 
every paragraph. The tag can be changed with ⟦*\nextpartag {Tag}⟧; leaving the 
argument empty will prevent marking the next paragraph.
Keep in mind that the (internal) reassignment is local: if a paragraph marked 
with ⟦\nextpartag⟧ starts inside a group, it will not reset.
Hence, to avoid surprises, you should have ⟦\nextpartag⟧ and the start of your 
paragraph at the same grouping level.

Useful structure elements for ⟦\nextpartag⟧ include ⟦H⟧ for headings and ⟦LI⟧ 
for list items. Since minim-pdf produces strongly-structured documents, the 
tags ⟦H1⟧, ⟦H2⟧, ⟦H3⟧ etc. should not be used.

Please also note that if you add ⟦\hbox⟧es directly to a vertical list
(this includes ⟦\line⟧, ⟦\centerline⟧ and the like),
the ⟦\everypar⟧ token list is not inserted and no new paragraph structure element 
will be opened.
The contents of the ⟦\hbox⟧ will be added to the current structure element, and 
this may result in an invalid structure hierarchy (and an error messsage 
reading ‘Structure type mismatch’).
You can make your intentions clear by inserting ⟦\startelement{P}⟧ at the 
appropriate place (see above).

Auto-marking paragraphs can be (locally) disabled or enabled by saying 
⟦*\markparagraphsfalse⟧ or ⟦*\markparagraphstrue⟧.

\section Manipulating the logical order %1

With the process outlined above, the logical order of structure elements has to 
coincide with the order in which the SEs are ‘digested’ by tex.
This, together with the marked content items being assigned to structure 
elements in their order of appearance, lies behind the restriction that logical 
and processing orders should match.

With manual intervention, this restriction can be relaxed somewhat. 
Issuing the pair ⟦*\savecurrentelement ... \continueelement⟧ will append the 
MCIs following ⟦\continueelement⟧ to the SE containing ⟦\savecurrentelement⟧.
Since the assignments made here are global, this process cannot be nested;
in more complicated situations you should therefore use
⟦*\savecurrentelementto\name ... \continueelementfrom\name⟧ which restores the 
current SE from a named identifier ⟦\name⟧.

\section Structure element options %1

The ⟦\startelement⟧ command allows a few options that are not mentioned above: 
its full syntax is ⟦*\startelement <options> {Tag}⟧.
The three most useful options are ⟦alt⟧ for setting an alt-text (the ⟦/Alt⟧ 
entry in the structure element dicionary), ⟦actual⟧ for a text replacement 
(⟦/ActualText⟧) and ⟦lang⟧ for the language (⟦/Lang⟧; see the next section).
The alternative and actual texts can also be given after the fact with
⟦*\setalttext {...}⟧ and ⟦*\setactualtext {...}⟧; these apply to the current 
structure element.

Structure element attributes can be given with
⟦attr <owner> <key> <value>⟧, e.g. ⟦attr Layout Placement /Inline⟧ or added 
later with ⟦*\tagattribute⟧.
Note that for the ⟦owner⟧ and ⟦key⟧ the initial slash must be omitted; the 
⟦value⟧ on the other hand will be written to the pdf verbatim.
Any number of attributes can be added.

An identifier can be set with the ⟦id {...}⟧ option, or after the fact with
⟦*\settagid {...}⟧. This identifier will be added to the ⟦IDTree⟧ and is 
entirely optional; you will probably already know when you need it.
The ⟦ref {...}⟧ option lets a structure element refer to another (the ⟦/Ref⟧ 
option in the structure element dictionary). Its argument should be the ⟦id⟧ of 
the other structure element.

The title of the structure element (corresponding to the ⟦/T⟧ entry in the 
structure element dictionary) can be set with the ⟦title {...}⟧ option. The 
pdf/ua standard requires this key for all document sections.

Finally, structure element classes can be given with the
⟦class <classname>⟧ keyword, which can be repeated.
Classes can be defined with
⟦*\newattributeclass classname <attributes>⟧ where ⟦<attributes>⟧ can be any 
number of ⟦attr⟧ statements as above.


\section Languages %1

If you do not specify a language code for a structure element, its language 
will be determined automatically.
In order for this to work, you must associate a language code to every used 
language; you can do so with ⟦*\setlanguagecode name code⟧, where ⟦name⟧ must 
be an identifier used with ⟦\uselanguage {name}⟧ and ⟦code⟧ must be a two or 
three-letter language code, optionally followed by a dialect specification, 
a country code, and/or some other tag.
Note that the language code is associated to a language \emph{name}, not to the 
numerical value of the ⟦\language⟧ parameter. This allows you to assign 
separate codes to dialects.

There is a small set of default language code associations, which can be found 
in the file ⟦minim-languagecodes.lua⟧.
It covers most languages defined by the hyph-utf8 package, as well as (due to 
their ubiquitous use) some ancient languages.

An actual language change introduced by ⟦\uselanguage⟧ will not otherwise be 
acted upon by this package. Therefore, you will probably want to add 
⟦\startelement{Span}⟧ after every in-line invocation of ⟦\uselanguage⟧.

You can set the document language with ⟦*\setdocumentlanguage language-code⟧. 
If unset, the language code associated with the first ⟦\uselanguage⟧ statement 
will be used, or else ⟦und⟧ (undetermined). The only function of the document 
language is that it is mentioned in the pdf catalog: it has no other influence.

New languages can be declared with ⟦*\newnamedlanguage {name} {lhm} {rhm}⟧ and 
new dialects with ⟦*\newnameddialect {language name} {dialect name}⟧.
Dialects will use the same hyphenation patterns (and will indeed have the same 
⟦\language⟧ value) as their parent languages; newly declared languages will 
start with no hyphenation patterns.
Do note that you will probably also have to specify language codes for new 
languages or dialects.

This package ensures the existence of the ⟦nohyph⟧, ⟦nolang⟧, ⟦uncoded⟧ and 
⟦undetermined⟧ dummy languages, all without hyphenation.


\section Mathematics %1

You can auto-tag equations as formulas by specifing ⟦*\autotagformulas⟧.
After this command, auto-tagging can be switched off and on with
⟦*\stopformulatagging⟧ and ⟦*\startformulatagging⟧.
Auto-tagging formulas is dangerous, because sometimes equations are used for 
lay-out and should not be marked as such.
It is also somewhat fragile, as it requires equations to end with dollar 
signs (and not with ⟦\Ustopmath⟧ or ⟦\Ustopdisplaymath⟧).

The tex source of an equation can be associated with the ⟦Formula⟧ structure 
element in various ways, which can be configured with ⟦*\includeformulasources 
{options}⟧, where the ⟦options⟧ must be a comma-separated list of
⟦alttext⟧, ⟦actualtext⟧ or ⟦attachment⟧.
The ⟦alttext⟧ and ⟦actualtext⟧ option will set the ⟦/Alt⟧ or ⟦/ActualText⟧ 
attributes to the unexpanded source code of the equation, surrounded by the 
appropriate number of dollar signs.
The ⟦attachment⟧ option attaches the source of the formula as an embedded file 
with its ⟦/AFRelation⟧ set to ⟦Source⟧;
this will only work if ⟦\pdfaconformancelevel⟧ equals three.
The name of this file can be changed by redefining ⟦*\formulafilename⟧ inside 
the equation.
The default value is ⟦{actualtext,attachment}⟧.

Note that the contents of the equation will be expanded fully (as in ⟦\xdef⟧) 
before their inclusion as the equation source. This may place restrictions on 
the macros you want to use (those in minim-math should be safe).
Any occurrence of ⟦\alttext⟧ or ⟦\actualtext⟧ overrides the 
automatically-assigned value and will be stripped from the equation source.


\section Tables %1

For marking up tables, a whole array of helper macros is available. First, 
⟦*\marktable⟧ should be given \emph{before} the ⟦\halign⟧.
Then, in the template, the first cell should start with ⟦*\marktablerow⟧
⟦*\marktablecell⟧ and 
each subsequent cell with ⟦\marktablecell⟧.
If your table starts with a header, insert ⟦*\marktableheader⟧ before it and 
⟦*\marktablebody⟧ after.
Before a table footer, insert ⟦*\marktablefooter⟧.

For greater convenience, insert just ⟦*\automarktable⟧ before the ⟦\halign⟧.
Then you can leave out all the above commands (unless you ⟦\omit⟧ a template of 
course).
This assumes the table has a single header row and more than one column.
If you use a table for typesetting a list, you can use ⟦*\marktableaslist⟧ 
instead, which marks the first column as list label and the second column as 
list item. Of course, this only works with two-column tables.

Cells spanning multiple cells or rows can be marked with 
⟦*\markcolumnspan {width}⟧ and ⟦*\markrowspan {height}⟧; these statements may not 
occur before ⟦\marktablecell⟧. Note that while ⟦\markcolumnspan⟧ properly 
increases the (internal) column number, ⟦\markrowspan⟧ does nothing of the 
sort (and indeed, no general logic can be given in the latter case). Always proceed 
with caution when using cells spanning multiple rows, and inspect the resulting 
structure carefully.

Marking up a table header (except if done through ⟦\automarktable⟧)
will not connect normal table cells with their headers; you will have to 
connect these manually by including ⟦*\markcolumnhead⟧ or ⟦*\markrowhead⟧ in 
the appropriate header cells. This must be done \emph{after} ⟦\markcolumnspan⟧ 
if the latter appplies. If properly setup like this, other cells of the table 
(including header cells) will be assigned to matching row or column headers 
automatically.


\section Other helper macros %1

For marking up an entry in a table of contents, you can use the macro
⟦*\marktocentry {dest} {lbl} {title} {filler} {pageno}⟧, which should insert all 
tags in the correct way. (The ⟦dest⟧ is a link destination and can be empty; 
the ⟦lbl⟧ is a section number and can also be empty.)

For tagging (foot)notes, ⟦*\marknoteref{*}⟧ and ⟦*\marknotelbl{*}⟧, when placed 
around the footnote markers, will insert the proper ⟦Ref⟧, ⟦Note⟧ and ⟦Lbl⟧ 
tags.

Helper macros for tagging sections can be setup quickly with
⟦*\sectionstructure { <section list> }⟧. The ⟦<section list>⟧ should be an 
increasing comma-separated list of section types, e.g.
⟦{subsection, section, chapter}⟧. This will first reserve the ⟦\count⟧ 
registers ⟦\subsectionnr⟧ etc, then create the structure aliases ⟦/Subsection⟧ 
etc. and finally define the helper macros ⟦\marksubsection⟧ etc, which will do 
the following:
\startlist
\item 1. Call ⟦\ensurestopelement⟧ on all lower section types.
\item 2. Set all lower section number counts to zero.
\item 3. Increase the current section type number by one.
\item 4. Call ⟦\startelement⟧ for the current section type.
\item 5. Set the ⟦\nextpartag⟧ to ⟦H⟧.
\stoplist
The proper place for these helper macros is just before the section heading; 
inbetween those two may come an ⟦\outline⟧ statement (see the previous 
chapter). You can set the ⟦title⟧ option to the internal ⟦\startelement⟧ 
statement with an optional argument (e.g.
⟦\marksection [Section \the\sectionnr]⟧).

\section Tagging the output routine %1

The command ⟦*\autotagplainoutput⟧ will try and update plain tex’s output routine 
to produce tagged page artifacts and footnotes.
It redefines ⟦\makeheadline⟧, ⟦\makefootline⟧, 
⟦\footnoterule⟧, ⟦\footnote⟧ and ⟦\vfootnote⟧.
Headline and footline will not be marked as artifacts if their contents equal 
⟦\hfil⟧; the footnote macros are edited to include the ⟦\marknoteref⟧ and 
⟦\marknotelbl⟧ macros described above.
Note that the ⟦\topinsert⟧, ⟦\midinsert⟧ and ⟦\pageinsert⟧ macros are left 
untouched; you will have to mark those explicitly.

You can make some changes to the affected macros before calling 
⟦\autotagplainoutput⟧, as it tries to be smart about it.
Though the redefinitions involve a full expansion,
most conditionals and common typesetting instructions (⟦\line⟧, ⟦\quad⟧, 
⟦\strut⟧ etc.) are safe-to-use and will not be expanded.
If you include custom macros of your own, however, it is wise to have those 
⟦\protected⟧.

% 

\endmanual