Skip to content

Commit

Permalink
Fix for tagging36+tagging641 (#1455)
Browse files Browse the repository at this point in the history 
* WIP

* looks about right by now

* docu and change log

* make sure we get a \@noitemerr early if there is material before the first item

* take  \@noitemerr logic out of the tagging code
  • Loading branch information
@FrankMittelbach
FrankMittelbach authored Sep 4, 2024
1 parent 7a9e646 commit fc85575
Show file tree
Hide file tree
Showing 12 changed files with 2,424 additions and 44 deletions.
22 changes: 19 additions & 3 deletions required/latex-lab/changes.txt
View file
Original file line number Diff line number Diff line change
@@ -1,9 +1,25 @@

2024-09-03 Ulrike Fischer <Ulrike.Fischer@latex-project.org>
* documentmetadata-support.dtx: switch to T1 encoding by default for non-Unicode
engines.

2024-09-02 Frank Mittelbach <Frank.Mittelbach@latex-project.org>

* latex-lab-block.dtx (subsubsection{List tags}):
Do not close LI and LBody if they were never opened because of a
missing \item (issue tagging/641)

(subsubsection{Implementation of \cs{item} template(s)}):
Set @newlist to false after the first \item (issue tagging/36)

Call \@noitemerr in para/begin hook if hmode is started before
the first item

(subsubsection{Implementation of block templates \ldots}):
Drop \@noitemerr from \addvspace and \addpenalty

2024-08-29 Ulrike Fischer <Ulrike.Fischer@latex-project.org>
* latex-lab-math.dtx: avoid loosing math if tagging is suspended,
* latex-lab-math.dtx: avoid loosing math if tagging is suspended,
tagging-project issue #661

2024-08-22 Joseph Wright <Joseph.Wright@latex-project.org>
Expand All @@ -29,7 +45,7 @@
* latex-lab-block.dtx, latex-lab-table.dtx: reset flattened level in para/restore, issue #544

2024-08-08 Ulrike Fischer <Ulrike.Fischer@latex-project.org>
* latex-lab-minipage.dtx, latex-lab-table.dtx: improve support for minipage in tables.
* latex-lab-minipage.dtx, latex-lab-table.dtx: improve support for minipage in tables.
handle issue #37

2024-08-06 Ulrike Fischer <Ulrike.Fischer@latex-project.org>
Expand All @@ -49,7 +65,7 @@
(subsection{Misc stuff}): drop unused socket

2024-07-11 Ulrike Fischer <Ulrike.Fischer@latex-project.org>
* latex-lab-toc.dtx, latex-lab-toc-kernel-changes: switch from configuration points
* latex-lab-toc.dtx, latex-lab-toc-kernel-changes: switch from configuration points
to sockets.

2024-05-25 Ulrike Fischer <Ulrike.Fischer@latex-project.org>
Expand Down
192 changes: 168 additions & 24 deletions required/latex-lab/latex-lab-block.dtx
View file
Original file line number Diff line number Diff line change
Expand Up @@ -9,8 +9,8 @@
%
% https://www.latex-project.org/lppl.txt
%
\def\ltlabblockdate{2024-08-11}
\def\ltlabblockversion{0.8p}
\def\ltlabblockdate{2024-09-03}
\def\ltlabblockversion{0.8q}
%<*driver>
\documentclass[kernel]{l3doc}
\usepackage{amstext}
Expand Down Expand Up @@ -1774,15 +1774,18 @@
\legacy_if_gset_false:n { @inlabel }
}
% \end{macrocode}
% In a pure ``displayblock'' scenario \texttt{@newlist} will be
% always false and the code bypassed, but we may have an outer list
% followed immediately by a displayblock (with the \cs{item} missing)
% In we are ending a list environment and haven't seen any
% \cs{item}, i.e., \texttt{@newlist} is still true, we raise an
% error. In pure a ``displayblock'' scenario \texttt{@newlist} will be
% always false, but if such an environment appears inside an outer
% list the code could still be triggered and that is undesirable
% (as it will be detected at the wrong point and later again). We
% therefore run it only if the current environment is a list.
% \changes{v0.8q}{2024/09/03}{Raise a \cs{@noitemerr} if appropriate}
% \begin{macrocode}
\@@_if_list:T { \legacy_if:nT { @newlist } { \@noitemerr } }
% \end{macrocode}
% \begin{macrocode}
\legacy_if:nT { @newlist }
{
\@noitemerr
\legacy_if_gset_false:n { @newlist }
}
\mode_if_horizontal:TF
{ \@@_skip_remove_last: \@@_skip_remove_last: \par }
{ \@inmatherr{\end{\@currenvir}} }
Expand All @@ -1792,6 +1795,13 @@
% \begin{macrocode}
\__kernel_displayblock_end:
% \end{macrocode}
% Resetting the \texttt{@newlist} switch is also only done if the
% current enviornment is a list and not unconditionally.
% \changes{v0.8q}{2024/09/03}{Setting \texttt{@newlist} to false
% moved after tagging code if in a list}
% \begin{macrocode}
\@@_if_list:T { \legacy_if_gset_false:n { @newlist } }
% \end{macrocode}
% What to do in terms of vertical spacing in different situations
% is still somewhat open to debate, right now this is more or less
% implementing what \LaTeXe{} list environment have been
Expand Down Expand Up @@ -1838,6 +1848,19 @@
% \end{macro}
%
%
% \begin{macro}{\@@_if_list:T}
% This needs some redesigning, as there is no good test for \enquote{is
% this environment a \enquote{list} that has \cs{item}s}. For now
% this here does the trick well enough.\fmi{revisit}
% \changes{v0.8q}{2024/09/03}{Provide a test for: Am I in a list?}
% \begin{macrocode}
\cs_new:Npn \@@_if_list:T
{ \tl_if_eq:NnT \l_@@_block_instance_tl {list} }
% \end{macrocode}
% \end{macro}
%
%
%
% \begin{macro}{\__kernel_displayblock_end:}
% The kernel hook for tagging at the end of the block.
% \begin{macrocode}
Expand Down Expand Up @@ -2081,6 +2104,71 @@
% \end{template}
%
%
% \begin{macro}{\addpenalty}
% The 2e implementation of \cs{addpenalty} (and \cs{addvspace})
% have a strange \cs{@noitemerr} if they are called in
% hmode. Generating this error in this case is definitely not
% necessary and in fact wrong, so below it is commented
% out. However, one could consider changing the logic and simply
% issue a \cs{par} in that case rather than omitting the penalty
% (or the vspace).\fmi{Evaluate if that change makes sense.}
% \changes{v0.8q}{2024/09/02}{Drop \cs{@noitemerr}}
% \begin{macrocode}
\def\addpenalty#1{%
\ifvmode
\if@minipage
\else
\if@nobreak
\else
\ifdim\lastskip=\z@
\penalty#1\relax
\else
\@tempskipb\lastskip
\begingroup
\@tempskipa\@tempskipb
\advance \@tempskipb
\ifdim\prevdepth>\maxdepth\maxdepth\else
\ifdim \prevdepth = -\@m\p@ \z@ \else \prevdepth \fi
\fi
\vskip -\@tempskipb
\penalty#1%
\ifdim\@tempskipa=\@tempskipb
\else
\advance\@tempskipb -\@tempskipa
\vskip \@tempskipb
\fi
\vskip \@tempskipa
\endgroup
\fi
\fi
\fi
% \else
% \@noitemerr
\fi
}%
% \end{macrocode}
% \end{macro}
%
%
% \begin{macro}{\addvspace}
% See \cs{addpenalty}
% \changes{v0.8q}{2024/09/02}{Drop \cs{@noitemerr}}
% \begin{macrocode}
\def\addvspace#1{%
\ifvmode
\if@minipage\else
\ifdim \lastskip =\z@
\@vspace@calcify{#1}%
\else
\setlength\@tempskipb{#1}%
\@xaddvskip
\fi
\fi
% \else
% \@noitemerr
\fi}
% \end{macrocode}
% \end{macro}
%
% \begin{macro}{\__kernel_displayblock_begin:,
% \__kernel_displayblock_beginpar_hmode:w,
Expand Down Expand Up @@ -2210,12 +2298,26 @@
\tl_set_eq:NN \@itemlabel \l_@@_item_label_tl
}
% \end{macrocode}
% finally, we signal that we are at the start of a new list (which
% Finally, we signal that we are at the start of a new list (which
% effects how the first \cs{item} is handled and how \cs{par}
% commands are interpreted.
% \begin{macrocode}
\legacy_if_gset_true:n { @newlist }
% \end{macrocode}
% If we encounter horizontal material before the first \cs{item} we
% do want a \cs{@noitemerr} straight away, because afterwards we
% end up with tagging structure faults and the cause is the
% missing \cs{item}. So we setup up \cs{@@_item_everypar:} to test
% for this. When the first \cs{item} is encountered this is then
% reset. This is only relevant for vertical lists, if we have
% inline lists one would need to test for somethingg else to
% identify that there is horizontal material between list start and
% the first \cs{item}, maybe some \cs{spacefactor} trick could be
% used then.\fmi{Think about a better implementation at some point.}
% \changes{v0.8q}{2024/09/02}{}
% \begin{macrocode}
\cs_set_eq:NN \@@_item_everypar: \@@_item_everypar_first:
% \end{macrocode}
%
% \begin{macrocode}
\@@_debug_typeout:n{template:list:std~end}
Expand Down Expand Up @@ -2306,6 +2408,7 @@
\tl_set_eq:NN \l_@@_label_given_tl \c_novalue_tl
\tl_if_empty:nF{#1}{ \SetTemplateKeys{item}{std}{#1} }
% \end{macrocode}
%
% If no optional argument was given then \cs{l_@@_label_given_tl}
% is still equal to \cs{c_novalue_tl} and so we can distinuish
% that from \verb=\item[]=.
Expand Down Expand Up @@ -2457,26 +2560,26 @@
%
%
%
% \begin{macro}{\@@_item_everypar:, \@@_item_everypar_std:}
% \begin{macro}{\@@_item_everypar:, \@@_item_everypar_std:, \@@_item_everypar_first:}
% The \cs{@@_item_everypar:} command is executed as part of \hook{para/begin}
% but most of the time does nothing, i.e., it has the following
% default definition.
% default definition outside of lists and most of the time within lists.
% \begin{macrocode}
\cs_new_eq:NN \@@_item_everypar: \prg_do_nothing:
% \end{macrocode}
%
%
% \begin{macrocode}
\AddToHook{para/begin}[lists]{\@@_item_everypar:}
\AddToHook{para/begin}[items]{\@@_item_everypar:}
% \end{macrocode}
%
% Note that we have to make sure that the above code is executed
% after the hook chunk from \texttt{tagpdf} because the latter uses
% \texttt{@inlabel} to make a decision.
%
% By the end of the day both should probably move into the kernel
% hook instead!
% hook instead or into sockets really.
% \begin{macrocode}
\DeclareHookRule{para/begin}{lists}{after}{tagpdf}
\DeclareHookRule{para/begin}{items}{after}{tagpdf}
% \end{macrocode}
%
%
Expand Down Expand Up @@ -2528,6 +2631,16 @@
}
% \end{macrocode}
%
% This is the setting for \cs{@@_item_everypar:} before the first
% \cs{item} is encountered.
% \changes{v0.8q}{2024/09/02}{Call \cs{@noitemerr} if hmode is
% started before the first item}
% \begin{macrocode}
\cs_new:Npn \@@_item_everypar_first: {
\legacy_if:nT { @newlist } { \@noitemerr }
}
% \end{macrocode}
%
% \end{macro}
%
%
Expand Down Expand Up @@ -2594,8 +2707,16 @@
}
{
\legacy_if:nTF { @newlist }
{ \__kernel_list_item_begin: }
{ \@@_inter_item: }
{
\__kernel_list_item_begin:
% \end{macrocode}
% The first item of a list also has to change the \texttt{@newlist} switch.
% \changes{v0.8q}{2024/09/02}{Set \texttt{@newlist} to false after
% the first \cs{item}}
% \begin{macrocode}
\legacy_if_gset_false:n { @newlist }
}
{ \@@_inter_item: }
% \end{macrocode}
% To avoid unnecessary key/val processing we make a quick check if
% there was an optional argument.
Expand Down Expand Up @@ -3971,15 +4092,38 @@
% Finally, at the list end we have to close the open
% \texttt{LBody}, \texttt{LI}, \texttt{L}, and possibly a
% \struct{text} if the last item ends with a list.
% However, if the user forgot to add an \cs{item} there is no
% \texttt{LI} and \texttt{LBody} open, so we check for the status
% of \texttt{@newlist}. The corresponding no-item error was
% generated earlier outside the tagging code.
%
% One could argue that it doesn't matter if the tagging is wrong
% after a \cs{@noitemerr} was issued. However, there is one case
% where it isn't an error: In the \texttt{thebibliography}
% environment (which is internally a list) it is often the case
% that documents start out with an empty environment, not
% containing any \cs{bibitem}s. For that reason \cs{@noitemerr} is
% redefined inside that environment to only produce a warning;
% hence we have to produce correct tag structures in that case.
% \changes{v0.8q}{2024/09/02}{Do not close LI and LBody if they
% never were opened.}
% \begin{macrocode}
\cs_set:Npn \@@_list_end: {
\legacy_if:nT { @endpe }
% \end{macrocode}
% If \texttt{@newlist} is true (ie when we have an error or warning
% situation) there is not much to close. If \verb=<L></L>= is not
% valid syntax we could add an empty \verb=<LI>= at this point instead.\fmi{check}
% \begin{macrocode}
\legacy_if:nF { @newlist }
{
\__tag_gincr_para_main_end_int:
\tagstructend % text-unit
\@@_debug_typeout:n{Structure-end~ P~ at~ list-end \on@line }
\legacy_if:nT { @endpe }
{
\__tag_gincr_para_main_end_int:
\tagstructend % text-unit
\@@_debug_typeout:n{Structure-end~ P~ at~ list-end \on@line }
}
\tagstructend\tagstructend % end LBody, LI
}
\tagstructend\tagstructend % end LBody, LI
\tagstructend % end L
}
% \end{macrocode}
Expand Down
Loading

0 comments on commit fc85575

Please sign in to comment.