% \iffalse meta-comment
%
% Copyright (C) 2016-2026 by Pierre Senellart
%
% This work may be distributed and/or modified under the
% conditions of the LaTeX Project Public License, either version 1.3
% of this license or (at your option) any later version.
% The latest version of this license is in
%   http://www.latex-project.org/lppl.txt
% and version 1.3 or later is part of all distributions of LaTeX
% version 2005/12/01 or later.
%
% This work has the LPPL maintenance status `maintained'.
%
% The Current Maintainer of this work is Pierre Senellart
% <pierre@senellart.com> and a version control system for this work
% is available at http://github.com/PierreSenellart/apxproof
%
% This work consists of the files apxproof.dtx, apxproof.ins,
% apxproof-synctex and apxproof.sty.ltxml, and the derived file
% apxproof.sty.
%
% \fi
%
% \iffalse
%<package>\NeedsTeXFormat{LaTeX2e}[2005/12/01]
%<package>\ProvidesPackage{apxproof}
%<package>  [2026/08/25 v1.5.0 Automatic proofs in appendix]
%
%<*driver>
\documentclass{ltxdoc}
\usepackage{apxproof}
\usepackage{hypdoc}
\usepackage{textcomp}
\usepackage[TS1,T1]{fontenc}
\usepackage{lmodern}
\usepackage{microtype}
\newtheorem{example}{Example}
\newtheoremrep{foobar}{Foobar}
\newtheoremrep*{foobarstar}{Foobarstar}
\EnableCrossrefs
\CodelineIndex
\RecordChanges
\begin{document}
  \DocInput{apxproof.dtx}
\end{document}
%</driver>
% \fi
%
% \CheckSum{2586}
%
% \CharacterTable
%  {Upper-case    \A\B\C\D\E\F\G\H\I\J\K\L\M\N\O\P\Q\R\S\T\U\V\W\X\Y\Z
%   Lower-case    \a\b\c\d\e\f\g\h\i\j\k\l\m\n\o\p\q\r\s\t\u\v\w\x\y\z
%   Digits        \0\1\2\3\4\5\6\7\8\9
%   Exclamation   \!     Double quote  \"     Hash (number) \#
%   Dollar        \$     Percent       \%     Ampersand     \&
%   Acute accent  \'     Left paren    \(     Right paren   \)
%   Asterisk      \*     Plus          \+     Comma         \,
%   Minus         \-     Point         \.     Solidus       \/
%   Colon         \:     Semicolon     \;     Less than     \<
%   Equals        \=     Greater than  \>     Question mark \?
%   Commercial at \@     Left bracket  \[     Backslash     \\
%   Right bracket \]     Circumflex    \^     Underscore    \_
%   Grave accent  \`     Left brace    \{     Vertical bar  \|
%   Right brace   \}     Tilde         \~}
%
% \changes{v1.0.4}{2017/03/02}{Show options commented on in margin and index}
%\iffalse
% Taken from xkeyval.dtx
%\fi
%\makeatletter
%\def\DescribeOption#1{\leavevmode\@bsphack\marginpar
%              {\raggedleft\PrintDescribeOption{#1}}\SpecialOptionIndex
%              {#1}\@esphack\ignorespaces}
%\def\PrintDescribeOption#1{\strut\emph{option}\\\MacroFont #1\ }
%\def\SpecialOptionIndex#1{\@bsphack
%    \index{#1\actualchar{\protect\ttfamily#1}
%           (option)\encapchar usage}\index
%          {options:\levelchar#1\actualchar{\protect\ttfamily#1}\encapchar
%           usage}\@esphack}
%\def\DescribeOptions#1{\leavevmode\@bsphack
%  \marginpar{\raggedleft\strut\emph{options}%
%  \@for\@tempa:=#1\do{%
%    \\\strut\MacroFont\@tempa\SpecialOptionIndex\@tempa
%  }}\@esphack\ignorespaces}
%\makeatother
%
% \changes{v1.0.0}{2016/10/31}{Initial released version}
% \changes{v1.0.1}{2016/11/07}{Prevent empty bibliography environment;
% fix typos}
%
% \GetFileInfo{apxproof.sty}
%
% \makeatletter\c@IndexColumns=2\makeatother
% \DoNotIndex{
%  \newcommand,\newenvironment,\end,\begin,\edef,\if,\else,\fi,\def,
%  \begingroup,\endgroup,\csname,\let,\noexpand,\protect,\expandafter,
%  \ifthenelse,\equal,\endcsname,\@empty,\@ifnotempty,\immediate,
%  \addtocounter,\newcounter,\newtoggle,\global,\ifdefined,\iftoggle,
%  \ifx,\jobname,\makeatletter,\makeatother,\newwrite,\NewEnviron,
%  \patchcmd,\relax,\renewcommand,\renewenvironment,\RequirePackage,
%  \roman,\space,\undefined,\unexpanded
% }
%
% \title{The \textsf{apxproof} package}
%
% \author{Pierre Senellart \\ \texttt{pierre@senellart.com} \\
% \url{http://github.com/PierreSenellart/apxproof}}
% \date{\filedate \quad \fileversion}
%
% \maketitle
%
% \begin{abstract}
% This package makes it easier to write articles where proofs and other material
% are deferred to the appendix. The appendix material is written in the \LaTeX{}
% code along with the main text which it naturally complements, and it is
% automatically deferred. The package can automatically send proofs to the
% appendix, can repeat in the appendix the theorem environments stated in the
% main text, can section the appendix automatically based on the sectioning of
% the main text, and supports a separate bibliography for the appendix material.
% \end{abstract}
%
% \section{Usage}
% The \textsf{apxproof} package is intended to simplify the writing of articles where some of
% the content needs to be deferred to an appendix. This is in particular
% useful for the submission of scientific articles to conferences or
% journals that limit the number of pages in the main text but allow an
% extra appendix, where proofs of theorems and other material can be added.
%
% \subsection{Basics}
% To use \textsf{apxproof}, first load it in the header of your document:
% \begin{quote}|\usepackage{apxproof}|\end{quote}
% On its own, this does not do anything and should not change the
% appearance of your document. To add an appendix with some material from
% your document, use the |toappendix| environment:
% \begin{quote}
% |\begin{toappendix}|\DescribeEnv{toappendix}\\
% \hspace*{1em}\dots\\
% |\end{toappendix}|
% \end{quote}
% The content will appear at the end of your document, in an
% automatically generated section that refers to the current section in the main
% text.
% \begin{example}
% Throughout this documentation, all examples produce content
% deferred to the appendix, at the very end of this document.
% \begin{verbatim}
% \begin{toappendix}
% This content is in the appendix.
% \end{toappendix}
% \end{verbatim}
% \begin{toappendix}
% This content is in the appendix.
% \end{toappendix}
% \end{example}
% \changes{v1.0.3}{2017/01/10}{Note on entire sections in appendix}
% When the content to put in appendix is an entire section, make sure
% that \verb|\section| is the very first command that appears within the
% \verb|toappendix| environment. It will disable the automatic production
% of a section heading.
% \subsection{Repeated Theorems and Proofs}
% In some scientific papers that include proofs, it is common to defer
% proofs to the appendix. This can easily be achieved using the
% |appendixproof| environment:
% \begin{quote}
% |\begin{appendixproof}|\DescribeEnv{appendixproof}\\
% \hspace*{1em}\dots\\
% |\end{appendixproof}|
% \end{quote}
% This behaves like the |toappendix| environment, except that
% a proof environment is generated.
% \begin{example}We now send a proof to the appendix:
% \begin{verbatim}
% \begin{appendixproof}
% This proof is in the appendix.
% \end{appendixproof}
% \end{verbatim}
% \begin{appendixproof}
% This proof is in the appendix.
% \end{appendixproof}
% \end{example}
%
% When deferring proofs to the appendix, an annoying problem is that the
% statement of the theorem remains in the main text; it is hard to read a
% proof that is far away from the statement it proves. \textsf{apxproof}
% offers two ways to address this. With
% |\newtheoremrep|\DescribeMacro{\newtheoremrep}, the statement
% of the theorem is \emph{repeated} in the appendix, immediately above
% the deferred proof. With the starred variant
% |\newtheoremrep*|\DescribeMacro{\newtheoremrep*}, the
% theorem is \emph{not} repeated, but the proof is still automatically
% deferred to the appendix; this is convenient when the proofs are short
% enough or close enough to the main text that a restatement would be
% redundant, especially when combined with the
% |\appendixproofname|\DescribeMacro{\appendixproofname}{} hook (see
% Section~\ref{sec:customization}) to title proofs with the type and
% number of the theorem they refer to (e.g., ``Proof of Theorem~1''). In
% summary, three behaviors are available for a theorem environment:
% \begin{description}
% \item[\textbackslash|newtheorem|] (\LaTeX's standard command):
%   the theorem appears in the main text only; the |proof| environment is
%   left unchanged by \textsf{apxproof}.
% \item[\textbackslash|newtheoremrep|]: the theorem appears in the main
%   text \emph{and} is restated in the appendix above its proof; the
%   |proof| that follows in the source is automatically deferred to the
%   appendix.
% \item[\textbackslash|newtheoremrep*|]: the theorem appears in the
%   main text only; the |proof| that follows is automatically deferred to
%   the appendix, without restating the theorem there.
% \end{description}
% In all cases, theorems can share counters and a single document can
% mix the three behaviors freely.
%
% To use the repeated-theorem feature, you can define a new
% \emph{repeated theorem} environment using the |\newtheoremrep| command:
% \changes{v1.0.5}{2017/05/31}{Ability to specify a sectioning counter in newtheoremrep}
% \begin{quote}
% |\newtheoremrep|\marg{name}\oarg{counter}\marg{title}\oarg{countersec}\DescribeMacro{\newtheoremrep}
% \end{quote}
% or, equivalently, its starred form to obtain a \emph{proof-deferred}
% (but not repeated) theorem:
% \begin{quote}
% |\newtheoremrep*|\marg{name}\oarg{counter}\marg{title}\oarg{countersec}\DescribeMacro{\newtheoremrep*}
% \end{quote}
% Usage of either form is exactly the same as that of AMS \LaTeX{}'s
% |\newtheorem| macro:
% \begin{itemize}
% \item \meta{name} (e.g., |theorem|) is the name of an environment that is created for this kind of
% theorem;
% \item \meta{counter} (e.g., |definition|) is an optional counter describing
% from which kind of environment the numbering of these environments should be
% inherited;
% \item \meta{title}
% (e.g., |Theorem|) is
% the title that will be used to display this theorem environment;
% \item \meta{countersec} (e.g., |section|) is an optional counter of a
% sectioning command indicating that counters for this theorem should
% be prefixed by this counter (and reset at each occurrence of the
% sectioning command).
% \end{itemize}
% \meta{counter} and \meta{countersec} should not be used together.
% What
% differs from |\newtheorem| is that, when the following is written:
% \begin{quote}
% |\newtheoremrep{foobar}{Foobar}|
% \end{quote}
% then \emph{two} environments are defined: the \verb|foobar|
% environment, which behaves as if |\newtheorem| had been used, and the
% |foobarrep| environment, which results in the statement of this
% environment being repeated in the appendix.
% The starred form |\newtheoremrep*{foobarstar}{Foobarstar}| similarly
% defines two environments |foobarstar| and |foobarstarrep|, except that
% |\begin{foobarstarrep}| does not produce any restatement of the
% theorem in the appendix; only the following |proof| is deferred.
%
% One interesting feature of \textsf{apxproof} is that in most
% situations, there is no need to use the |appendixproof| environment.
% Indeed, the |proof| \DescribeEnv{proof} environment is redefined by \textsf{apxproof} to
% automatically put the proof either in the main text (if it follows a
% regular theorem) or in the appendix (if it follows a repeated
% theorem).
% \begin{example}
% Assume we have first defined a repeated theorem environment |foobar| as
% above.
% We can now use this theorem environment, first for a regular theorem in
% the main text, then for a theorem repeated in the main text and in the
% appendix:
% \begin{quote}
% \begin{verbatim}
% \begin{foobar}
% This foobar is a regular one, in the main text.
% \end{foobar}
% \begin{proof}
% This is the proof of the regular foobar.
% \end{proof}
% \end{verbatim}
% \end{quote}
% We obtain:
% \begin{foobar}
% This foobar is a regular one, in the main text.
% \end{foobar}
% \begin{proof}
% This is the proof of the regular foobar.
% \end{proof}
% Now, if we use a repeated theorem:
% \begin{quote}
% \begin{verbatim}
% \begin{foobarrep}
% This foobar is repeated in the appendix.
% \end{foobarrep}
% \begin{proof}
% This is the proof of the repeated foobar.
% \end{proof}
% \end{verbatim}
% \end{quote}
% We now obtain:
% \begin{foobarrep}
% This foobar is repeated in the appendix.
% \end{foobarrep}
% \begin{proof}
% This is the proof of the repeated foobar.
% \end{proof}
% Note that, since |hyperref| is loaded, there are hyperlinks
% created between the statements of the theorems in the main text and in
% the appendix.
% \end{example}
% \begin{example}
% In contrast, the starred form of |\newtheoremrep| produces a theorem
% whose proof is deferred to the appendix \emph{without} restating the
% theorem there. Assuming the declaration
% |\newtheoremrep*{foobarstar}{Foobarstar}| at the top of the document,
% writing:
% \begin{quote}
% \begin{verbatim}
% \begin{foobarstarrep}
% This foobar will not be repeated in the appendix,
% but the proof below will be deferred there.
% \end{foobarstarrep}
% \begin{proof}
% This is the proof of the (non-repeated) foobar.
% \end{proof}
% \end{verbatim}
% \end{quote}
% yields:
% \begin{foobarstarrep}
% This foobar will not be repeated in the appendix,
% but the proof below will be deferred there.
% \end{foobarstarrep}
% \begin{proof}
% This is the proof of the (non-repeated) foobar.
% \end{proof}
% In this mode, the title of the deferred proof in the appendix is
% controlled by the |\appendixproofname| hook documented in
% Section~\ref{sec:customization}; by default, this hook is empty, so
% proofs are titled simply ``Proof.'' as usual. The forward hyperlink
% from the theorem number in the main text (when \textsf{hyperref} is
% loaded) points to the location of the proof in the appendix.
% \end{example}
%
% When the proof is deferred to the appendix, it is common practice to
% add a proof sketch in the main text. \textsf{apxproof} defines a simple
% |proofsketch| environment for this purpose:
% \changes{v1.0.5}{2017/05/31}{Fix compilation of proofsketch environment in inline mode}
% \begin{quote}
% |\begin{proofsketch}|\DescribeEnv{proofsketch}\\
% \hspace*{1em}\dots\\
% |\end{proofsketch}|
% \end{quote}
% The proof sketch is typeset similarly to a proof, but is always in the
% main text. Similarly, an |inlineproof| \DescribeEnv{inlineproof}
% environment is provided so as to
% be able to have both a proof in the appendix (using the regular |proof|
% environment, or alternatively the |appendixproof| environment) and a
% different proof in the main text (using the |inlineproof| environment).
% \begin{example} Here are simple examples of proof sketches and inline
% proofs:
% \begin{verbatim}
% \begin{proofsketch}
% This is a proof sketch.
% \end{proofsketch}
% \end{verbatim}
% \begin{proofsketch}
% This is a proof sketch.
% \end{proofsketch}
% \begin{verbatim}
% \begin{inlineproof}
% This is an inline proof.
% \end{inlineproof}
% \end{verbatim}
% \begin{inlineproof}
% This is an inline proof.
% \end{inlineproof}
% \end{example}
%
% \subsection{Bibliography}
% By default, \textsf{apxproof} automatically adds a bibliography in the appendix
% with only the references cited in the appendix material. This allows for a
% clean separation of references used solely in the main text, and those
% used in the appendix.
% \begin{example}
% Assume we have citations both in the main text and in the appendix.
% \begin{verbatim}
% This is a citation in the main text~\cite{lamport86}.
% \begin{toappendix}
% This is a citation in the appendix~\cite{proofsAreHard}.
% \end{toappendix}
% \end{verbatim}
% This is a citation in the main text~\cite{lamport86}.
% \begin{toappendix}
% This is a citation in the appendix~\cite{proofsAreHard}.
% \end{toappendix}
% \end{example}
% The bibliography in the appendix can use
% a different style and heading than the bibliography in the main text (and, by
% default, it does). See
% Section~\ref{sec:customization} for how to configure the appearance of
% that bibliography.
%
% \DescribeOption{bibliography}
% In order to use a single appendix for the main text and the
% bibliography, one can specify the
% value |common| to the |bibliography| option when loading the package.
%
% \DescribeOption{bibengine}
% By default, \textsf{apxproof} relies on the \textsf{bibunits} package
% to typeset a separate bibliography for the appendix material; this is
% the |bibengine=bibtex| mode (the default), suitable for documents
% that use \textsf{bibtex} (with or without \textsf{natbib}) for
% citations and references. Documents using \textsf{biblatex} can
% instead set |bibengine=biblatex|, in which case \textsf{apxproof}
% wraps the appendix material in a \textsf{biblatex} |refsection|
% environment and emits the appendix bibliography with
% |\printbibliography|. In this mode the user is responsible for
% loading \textsf{biblatex} and declaring bibliography resources with
% |\addbibresource| in the preamble; resources are shared between the
% main document and the appendix |refsection|, but citations and the
% printed bibliography are scoped independently. The |bibengine| option
% only has an effect when |bibliography=separate|.
% (By default this option is set to |separate|.)
% \subsection{Mode}
% \DescribeOption{appendix}
% An optional \meta{mode} can be specified when loading the package:
% \begin{quote}|\usepackage[appendix=|\meta{mode}|]{apxproof}|\end{quote}
% \meta{mode} can take one of the following four values:
% \begin{description}
% \item[|append|] This is the default.
% Appendix material gathered by \textsf{apxproof} is appended to
% the main text.
% \item[|inline|] In this mode, \textsf{apxproof} simply inlines the
% content along with the main text.
% \item[|strip|] This mode functions similarly to |append| except that
% the appendix is not appended at the end of the document. All appendix
% material is therefore removed.
% \item[|chapterend|] In this mode, appendix material for each chapter
% is gathered separately and emitted at the end of that chapter
% (i.e., just before the next |\chapter|, |\part|, |\frontmatter|,
% |\mainmatter|, or |\backmatter| command, or at the end of the
% document), using the |subappendices| environment from the
% \textsf{appendix} package. Only |bibliography=common| is supported
% in this mode. See Section~\ref{sec:customization} for the
% |\chapterappendixprelim| and |\flushchapterappendix| customization
% macros.
% \end{description}
%
% \subsection{Customization}
% \label{sec:customization}
% \textsf{apxproof} provides a few macros that can be redefined (using
% |\renewcommand|) to customize the appearance of the appendix:
% \begin{description}
% \item[\textbackslash|mainbodyrepeatedtheorem|]\DescribeMacro{\mainbodyrepeatedtheorem}
% is a macro that is executed at the beginning of the body of every
% repeated theorem. This can be used to notify the reader that the
% theorem is repeated in appendix in some way, e.g., with a margin note.
% \item[\textbackslash|appendixsectionformat|\marg{number}\marg{title}]\DescribeMacro{\appendixsectionformat} is a macro
% that indicates how to format the section titles in the Appendix, given
% the number and title of the section in the main text. By
% default, they appear as ``Proofs for Section~\meta{number}
% (\meta{title})''.
% \item[\textbackslash|appendixproofname|\marg{type}\marg{ref}]\DescribeMacro{\appendixproofname}
% is a macro that determines the title of a proof deferred to the
% appendix when it follows a |\newtheoremrep| or |\newtheoremrep*|
% environment. By default, this macro is empty, which means proofs are
% titled ``Proof.'' as usual. When redefined to a non-empty value, the
% expansion is used as the title of the appendix proof; \meta{type} is
% the title of the theorem environment (e.g., |Theorem|) and \meta{ref}
% is a reference to the theorem statement in the main text. To title
% all such proofs ``Proof of \meta{type}~\meta{number}'', one would
% typically write:
% \begin{quote}
% \begin{verbatim}
% \renewcommand{\appendixproofname}[2]{Proof of #1~#2}
% \end{verbatim}
% \end{quote}
% This hook applies whether the theorem is restated in the appendix
% (|\newtheoremrep|) or not (|\newtheoremrep*|). It has no effect on
% manually written |appendixproof| environments or on proofs given an
% explicit optional title.
% \item[\textbackslash|appendixrefname|]\DescribeMacro{\appendixrefname} contains the heading that is displayed before
% the bibliography. By default, this is ``References for the
% Appendix''. (Note that this command is also defined and used by the
% |memoir| document class.)
% \item[\textbackslash|appendixbibliographystyle|]\DescribeMacro{\appendixbibliographystyle} contains the |.bst| bibliography
% style that is used in the bibliography in appendix. By default, this is
% |alpha|.
% \item[\textbackslash|appendixbibliographyprelim|]\DescribeMacro{\appendixbibliographyprelim} contains arbitrary code that is executed
% just before the production of the bibliography in appendix, which can
% be used to configure the way it is displayed.
% \item[\textbackslash|appendixprelim|]\DescribeMacro{\appendixprelim} contains arbitrary code that is executed
% just before the production of the appendix, which can
% be used to configure the way it is displayed. By default, this command
% contains |\clearpage\onecolumn| (the appendix is typeset on a new page
% in single-column mode) but redefining this option allows changing this
% behavior. This macro has no effect in |chapterend| mode; use
% |\chapterappendixprelim| instead.
% \item[\textbackslash|chapterappendixprelim|]\DescribeMacro{\chapterappendixprelim}
% (|chapterend| mode only) contains arbitrary code that is executed
% just before each chapter's appendix block. By default, this command
% contains |\clearpage| (each chapter appendix starts on a new page).
% This macro has no effect in |append| or |strip| mode; use
% |\appendixprelim| instead.
% \item[\textbackslash|flushchapterappendix|]\DescribeMacro{\flushchapterappendix}
% (|chapterend| mode only) manually emits the current chapter's
% accumulated appendix material at the point of invocation, and starts
% a fresh per-chapter buffer. The package automatically inserts an
% equivalent flush before each |\chapter|, |\part|, |\frontmatter|,
% |\mainmatter|, |\backmatter|, and at the end of the document, so
% this macro is only useful in unusual document structures where an
% appendix needs to be emitted at a location that is not a standard
% structural boundary.
% \end{description}
% \DescribeOption{repeqn}
% Another customization capability concerns \emph{numbered equations} that are present
% within repeated theorems. An optional |repeqn| option can be specified
% when loading the package, which controls whether
% equation numbers should be as in the main text (by setting this option
% to |same|, the default) or independently
% numbered (by setting this option to |independent|). In the latter case,
% whenever a referenceable counter is set with |\label{|\meta{counter}|}|,
% |\ref{|\meta{counter}|}| references the counter in the main text, while
% |\ref{|\meta{counter}|-apx}| references the counter in the appendix
% (except in |inline| mode, where both have the same effect).
%
% Enhanced cross-referencing packages are supported: the optional argument
% that \textsf{cleveref} adds to |\label| (as in
% |\label[equation]{|\meta{counter}|}|) may be used within repeated material,
% and so may the |\zlabel| command of \textsf{zref-clever}, both giving rise
% to the same |-apx| variants as |\label| does. Note that \textsf{cleveref}
% loses that optional argument inside \textsf{amsmath} display environments
% (|equation|, |align|, etc.), where the label ends up unreferenceable; this
% happens with or without \textsf{apxproof}, and affects neither plain
% |\label| nor \textsf{zref-clever}.
%
% \DescribeOption{forwardlinking}
% Another customization option concerns hyperlinking.
% Usually, when \textsf{hyperref} is loaded, |foobarrep|
% environments in the main text have their number link to their repetition in the
% appendix. To suppress this behavior and have |foobarrep| environments
% treated as if \textsf{hyperref} were not loaded, one can specify the
% value |no| to the |forwardlinking| option when loading the package. (By
% default this option is set to |yes|.)
%
% \subsection{Sync\TeX{} Support}
% \label{sec:synctex}
% \changes{v1.5.0}{2026/08/19}{\texttt{synctex} option, \texttt{auto} by default}
% \DescribeOption{synctex}
% Sync\TeX{} lets editors and PDF viewers jump between a position in the
% PDF and the source line that produced it. \textsf{apxproof} supports it
% in the appendix as well as in the main text, and nothing has to be done
% for that: rather than copying appendix material and repeated theorems to
% an auxiliary file, which would attribute them to that file, the package
% typesets them from the source itself.
%
% The |synctex| option says when it does so. With |auto| (the default),
% exactly when Sync\TeX{} output was requested (i.e., when the engine was
% run with |-synctex=1| or a similar option, which is what editors and
% \textsf{latexmk} do); with |yes|, always; with |no|, never, and
% Sync\TeX{} is then not supported. The typeset output is the same in all
% cases. Should the re-reading cause trouble in a document, |synctex=no|
% restores the previous behavior.
%
% Jumping from the PDF to the source (inverse search) works with every
% engine. Jumping from the source to the PDF (forward search) finds
% appendix material with Lua\TeX{}; with pdf\TeX{} and Xe\TeX{}, it does so
% only if the |apxproof-synctex| script distributed with the package (it
% requires Python~3) is run after each compilation:
% \begin{quote}|apxproof-synctex |\meta{jobname}\end{quote}
% With \textsf{latexmk}, adding
% \begin{quote}|$success_cmd = 'apxproof-synctex %R';|\end{quote}
% to a |latexmkrc| file does this automatically.
%
% Section~\ref{sec:issues} lists what re-reading the source asks of a
% document, and Section~\ref{sec:synctex-impl} explains how it is done.
%
% \subsection{Advanced Features}
% We now describe a few advanced macros and environments, the usage of
% which is limited to special cases:
% \begin{description}
% \item[|nestedproof|]\DescribeEnv{nestedproof} is an environment that
% can be used within a |proof| environment deferred in the appendix; this
% is required because, for technical reasons, no |proof| environment can
% be nested within a deferred |proof| environment.
% \item[\textbackslash|noproofinappendix|]\DescribeMacro{\noproofinappendix}
% can be used inside repeated theorems that are not followed by a
% |proof| or |appendixproof| environment; the point is to ensure that a further
% |proof| environment cannot be mistakenly understood as a proof of the repeated
% theorem. It should not be needed in most situations as
% \textsf{apxproof} tries figuring out when a proof follows a repeated
% theorem automatically, but may occasionally be needed in complex
% scenarios.
% \item[\textbackslash|nosectionappendix|]\DescribeMacro{\nosectionappendix}
% is to be used inside a section that \emph{does} contain appendix material, but
% for which a section in the appendix should not be created. This should
% be rarely needed. When this command is present, appendix material is
% appended to the end of the previously created section.
% \item[\textbackslash|apxproofhook|]\DescribeMacro{\apxproofhook}
% is a one-argument hook, doing nothing by default, that
% \textsf{apxproof} runs inside every proof it typesets in the appendix,
% and inside an inline proof whose title names its result through a
% cross-reference (|\ref|, |\cref|, |\Cref|, |\autoref|, |\vref| or
% |\zcref|). Its argument is a label of the result the proof establishes:
% the label of the preceding repeated theorem in the first case, the target
% of the title's cross-reference in the second. Redefining it is how a
% companion package, such as \textsf{proofgraph}, attributes each proof to
% the result it proves; documents have no reason to.
% \end{description}
% \subsection{Grouping Appendix Material Under a Single Section}
% \changes{v1.4.0}{2026/05/14}{Documented workarounds for grouping
%   per-section appendices under a single wrapping section. See \#56.}
% \label{sec:appendix-grouping}
% By default, \textsf{apxproof} produces one section in the appendix for
% each section of the main text that contains appendix material (with
% the format controlled by |\appendixsectionformat|). If you would
% rather have a single section in the appendix (e.g., titled
% ``Proofs'') with the per-section entries appearing as
% \emph{subsections} of that single section, two recipes are available.
%
% The first is fully explicit: disable the automatic per-section
% heading with |\nosectionappendix|, and write the desired structure
% directly inside |toappendix| environments.
% \begin{quote}
% \begin{verbatim}
% \begin{toappendix}
%   \section{Proofs}
% \end{toappendix}
%
% \section{First}\label{sec:first}
% \nosectionappendix
% \begin{toappendix}
%   \subsection{Proofs of Section~\ref{sec:first}}
% \end{toappendix}
%
% % ... theorems and proofs of section ``First'' ...
%
% \section{Second}\label{sec:second}
% \nosectionappendix
% \begin{toappendix}
%   \subsection{Proofs of Section~\ref{sec:second}}
% \end{toappendix}
%
% % ... theorems and proofs of section ``Second'' ...
% \end{verbatim}
% \end{quote}
% The reader of the source can then see exactly what will be produced
% in the appendix, at the cost of some verbosity.
%
% The second recipe is more concise. The section level used by
% \textsf{apxproof} for the auto-generated appendix headings is swapped
% from |\section| to |\subsection| for the duration of an explicit
% |\begingroup|/|\endgroup| group:
% \begin{quote}
% \begin{verbatim}
% \begin{toappendix}
%   \section{Proofs}
%   \begingroup\makeatletter
%     \let\axp@oldsection\subsection
%   \makeatother
% \end{toappendix}
%
% \section{First}\label{sec:first}
% % ... theorems and proofs of section ``First'' ...
%
% \section{Second}\label{sec:second}
% % ... theorems and proofs of section ``Second'' ...
%
% \begin{toappendix}
%   \endgroup
% \end{toappendix}
% \end{verbatim}
% \end{quote}
% The wrapping |\section{Proofs}| is written once, and every subsequent
% auto-generated appendix section heading is demoted to a |\subsection|
% for the scope of the group. The trade-off is that this recipe relies
% on an internal macro (|\axp@oldsection|) and on a manually balanced
% |\begingroup|/|\endgroup| pair across two separate |toappendix|
% environments.
% \section{Supported Document Classes}
% Because \textsf{apxproof} modifies sectioning commands, bibliographies,
% and proofs, it may not work straight away with arbitrary document classes.
% It has currently been tested with and is supported for the following
% document classes:
% \begin{itemize}
% \item \LaTeX{} standard document classes (|article.cls|, |book.cls|)
% \item \href{https://www.ctan.org/pkg/koma-script}{KOMA-Script}
% (|scrartcl.cls|, |scrbook.cls|)
% \item \href{https://ctan.org/pkg/memoir}{|memoir.cls|}
% \item \href{https://www.ctan.org/pkg/acmart}{ACM journals and
% proceedings} (|acmart.cls|)
% \item \href{https://www.ctan.org/pkg/ieeetran}{IEEE transactions and
% conferences} (|IEEEtran.cls|)
% \item \href{https://www.springer.com/computer/lncs/lncs+authors}{
% Springer's Lecture Notes in Computer Science} (|llncs.cls|)
% \item \sloppy
% \href{https://www.dagstuhl.de/en/publications/lipics}{Schlo\ss{}
% Dagstuhl's Leibniz International Proceedings in Informatics}
% (|lipics.cls|, |lipics-v2016.cls|, |lipics-v2019.cls|,
% |lipics-v2021.cls|)
% \end{itemize}
% Other classes may work out of the box. Adding support for specific
% classes is possible and can be requested from the
% author of this package.
% \section{HTML Conversion with \textsf{LaTeXML}}
% \label{sec:latexml}
% \changes{v1.5.0}{2026/08/25}{Document the \textsf{LaTeXML} binding}
% \textsf{LaTeXML} (\url{https://math.nist.gov/~BMiller/LaTeXML/}) converts
% \LaTeX{} documents to HTML. It cannot read |apxproof.sty| itself, so the
% package comes with a binding, |apxproof.sty.ltxml|, which reproduces its
% behavior in the conversion: proofs deferred to the appendix, repeated
% theorem statements with their original numbering, automatic appendix
% sectioning, forward links from theorems to their proofs, and a separate
% appendix bibliography.
%
% The binding is installed with the package, next to |apxproof.sty|.
% \textsf{LaTeXML} looks for it there, and in the directory of the document;
% should it be kept anywhere else, point \textsf{LaTeXML} at that directory:
% \begin{quote}|latexmlc --path=|\meta{dir}| --dest=paper.html paper.tex|\end{quote}
% Do \emph{not} pass |--includestyles|: the binding is self-contained, and
% that option makes \textsf{LaTeXML} process the dependencies of the real
% |apxproof.sty|, which fails. \textsf{LaTeXML}~0.8.8 or later is required.
% \section{Known Issues and Limitations}
% \label{sec:issues}
% We report here some issues we are currently aware of:
% \begin{itemize}
% \item When using \textsf{hyperref}, the appendix in the bibliography is
% not hyperlinked. This is to avoid possible issues with multiply defined
% bibliography entries.
% \item |appendixproof|, |proof|, |toappendix| environments cannot be nested. This is a limitation of
% the \textsf{fancyvrb} package that \textsf{apxproof} relies on. Note
% the existence of the |nestedproof| environment for nested proofs.
% \item Sync\TeX{} support (Section~\ref{sec:synctex}) works by re-reading
% the source when the appendix is typeset, which requires each captured
% environment to appear literally in a source file: |\begin{foobarrep}|,
% |\begin{proof}|, |\begin{toappendix}| and their |\end| counterparts
% cannot come from a macro (nor could they be captured before, except for
% repeated theorems). Conversely, wrapping a |foobarrep| environment in a
% user-defined environment is supported only with this option.
% \item Sync\TeX{} support also requires the source files to be readable,
% under the same names, when the appendix is typeset, which is the case in
% any normal compilation. The main file is |\jobname.tex|, or, when \TeX{} was
% given a path to it, the name recorded in the |.fls| file written by the
% |-recorder| option (\textsf{latexmk} and most editors pass it). When the
% main file cannot be located, \textsf{apxproof} says so and copies
% appendix material to the |.axp| file as with |synctex=no|, which leaves
% the document complete but without Sync\TeX{} support for that material;
% when a single piece cannot be re-read, an error message explains why and
% that piece is missing from the appendix.
% \item Sync\TeX{} forward search, from the source to the PDF, reaches
% appendix material with Lua\TeX{} only: with pdf\TeX{} and Xe\TeX{}, a line
% of a proof or of a |toappendix| environment lands on the nearest line of
% the main text unless the |apxproof-synctex| script is run
% (Section~\ref{sec:synctex}), the way Sync\TeX{} identifies files making
% it impossible to do better without breaking forward search from the main
% text. Sync\TeX{} support also has a cost: each piece of appendix material
% is reached by reading its source file from the beginning, which a document
% with hundreds of proofs in a very long file will notice.
% \item By default, when |bibliography=separate|, \textsf{apxproof}
% uses the \textsf{bibunits} package to generate a second bibliography.
% Packages incompatible with \textsf{bibunits} (notably
% \textsf{biblatex}) will therefore not work in that default
% configuration. For \textsf{biblatex} users, set |bibengine=biblatex|
% (in addition to |bibliography=separate|) to use \textsf{biblatex}'s
% own |refsection| machinery instead. The |bibliography=common| mode
% remains a third option that works with most bibliography packages.
% \end{itemize}
%
% Issues not listed here should be reported to the author.
%
% \section{License}
% Copyright \textcopyright{} 2016--2026 by Pierre Senellart.
%
% This work may be distributed and/or modified under the conditions of the
% \LaTeX{} Project Public License, either version 1.3 of this license or (at
% your option) any later version. The latest version of this license is in
% \url{http://www.latex-project.org/lppl.txt} and version 1.3 or later is part of
% all distributions of \LaTeX{} version 2005/12/01 or later.
%
% \section{Contact}
% \begin{itemize}
% \item \url{https://github.com/PierreSenellart/apxproof}
% \item
% Pierre Senellart
% \href{mailto:pierre@senellart.com}{<pierre@senellart.com>}
% \end{itemize}
% Bug reports and feature requests should
% preferably be submitted through the \emph{Issues} feature of GitHub.
%
% \section{Acknowledgments}
% Thanks to Antoine Amarilli for feedback and proofreading. Thanks to K.~D.
% Bauer for the implementation of the forward-linking mechanism, and
% for various bugfixes. Thanks to  Leonid Kostrykin for an initial
% implementation of the |forwardlinking| option.
%
% \StopEventually{
%   \PrintChanges
%   \PrintIndex
% }
%
% \section{Implementation}
% We now describe the entire code of the package, in a literate programming
% fashion. Throughout the package, we use the |axp@| prefix to identify
% local macros and environment names, which are not meant to be used by the final
% user.
% \subsection{Dependencies}
% We first load a few package dependencies:
% \begin{itemize}
% \item \textsf{environ} to easily define the repeated theorem
% environments.
%    \begin{macrocode}
\RequirePackage{environ}
%    \end{macrocode}
% \item \textsf{etoolbox} to define simple toggles.
%    \begin{macrocode}
\RequirePackage{etoolbox}
%    \end{macrocode}
% \item \textsf{fancyvrb} for the bulk of the work of exporting appendix
% material in an auxiliary file.
%    \begin{macrocode}
\RequirePackage{fancyvrb}
%    \end{macrocode}
% \item \textsf{ifthen} for easier comparison of character strings.
%    \begin{macrocode}
\RequirePackage{ifthen}
%    \end{macrocode}
% \item \textsf{kvoptions} to manage options passed to the package.
%    \begin{macrocode}
\RequirePackage{kvoptions}
%    \end{macrocode}
% \item \textsf{catchfile} to be able to check the content of files
% |\input| within appendix content.
%    \begin{macrocode}
\RequirePackage{catchfile}
%    \end{macrocode}
% \item \textsf{amsthm} for its |\newteorem| macro.
% Some document classes (e.g., \textsf{lipics})
% preload \textsf{amsthm}: this is fine, |\RequirePackage{amsthm}|
% will simply have no effect. On the other hand, some other document
% classes (e.g., \textsf{llncs} or \textsf{sig-alternate}) define a
% |proof| environment that conflicts with \textsf{amsthm}, so we have to
% undefine this environment before loading \textsf{amsthm}. In that case,
% we reestablish the existing proof environments, in case they had been
% customized (e.g., \textsf{sig-alternate})
% \changes{v1.0.4}{2017/03/08}{Re-establish custom proof environments}
%    \begin{macrocode}
\@ifpackageloaded{amsthm}{
  }{
    \let\apx@oldamsthmproof\proof
    \let\apx@oldamsthmendproof\endproof
    \let\proof\undefined
    \let\endproof\undefined
  }
\RequirePackage{amsthm}
\ifdefined\apx@oldamsthmproof
  \let\proof\apx@oldamsthmproof
  \let\endproof\apx@oldamsthmendproof
\fi
%    \end{macrocode}
% \end{itemize}
% \subsection{Option Processing}
% Many names throughout the package use an arobase (|@|) to avoid name
% conflict with user-defined names. To simplify the compilation of the
% documentation, we simply make it a regular character in all the rest.
%    \begin{macrocode}
\makeatletter
%    \end{macrocode}
% We setup the processing of options using \textsf{keyval} facilities.
%    \begin{macrocode}
\SetupKeyvalOptions{
  family=axp,
  prefix=axp@
}
%    \end{macrocode}
% We declare the following options:
% \begin{itemize}
%   \item |appendix|, with a default value of |append| (other possible
%     values: |strip|, |inline|);
%   \item |bibliography|, with a default value of |separate| (other
%     possible value: |common|);
%   \item |repeqn|, with a default value of |same| (other possible value:
%   |independent|).
% \end{itemize}
% \begin{macro}{\axp@appendix}
%    \begin{macrocode}
\DeclareStringOption[append]{appendix}
%    \end{macrocode}
% \end{macro}
% \begin{macro}{\axp@bibliography}
% \changes{v1.0.4}{2017/03/02}{\texttt{bibliography} option}
%    \begin{macrocode}
\DeclareStringOption[separate]{bibliography}
%    \end{macrocode}
% \end{macro}
% \begin{macro}{\axp@bibengine}
% \changes{v1.4.0}{2026/05/14}{\texttt{bibengine} option to select
%   the engine used for separate bibliographies: \textsf{bibunits}
%   (default) or \textsf{biblatex}. See \#29.}
%    \begin{macrocode}
\DeclareStringOption[bibtex]{bibengine}
%    \end{macrocode}
% \end{macro}
% \begin{macro}{\axp@repeqn}
% \changes{v1.1.0}{2018/07/20}{\texttt{repeqn} option}
%    \begin{macrocode}
\DeclareStringOption[same]{repeqn}
%    \end{macrocode}
% \end{macro}
% \begin{macro}{\axp@forwardlinking}
% \changes{v1.2.3}{2021/07/12}{\texttt{forwardlinking} option}
%    \begin{macrocode}
\DeclareStringOption[yes]{forwardlinking}
%    \end{macrocode}
% \end{macro}
% \begin{macro}{\axp@synctex}
% \changes{v1.5.0}{2026/08/19}{\texttt{synctex} option, \texttt{auto} by default}
%    \begin{macrocode}
\DeclareStringOption[auto]{synctex}
%    \end{macrocode}
% \end{macro}
%    \begin{macrocode}
\ProcessLocalKeyvalOptions*
%    \end{macrocode}
% We check that the value of the options are valid, and add a
% message to the compilation log.
%    \begin{macrocode}
\ifthenelse{\equal{\axp@appendix}{append}}{
  \message{apxproof: Appendix material appended to the document}
}{\ifthenelse{\equal{\axp@appendix}{strip}}{
  \message{apxproof: Appendix material stripped}
}{\ifthenelse{\equal{\axp@appendix}{inline}}{
  \message{apxproof: Appendix material inlined within the document}
}{\ifthenelse{\equal{\axp@appendix}{chapterend}}{
  \message{apxproof: Appendix material appended at the end of each chapter}
}{
  \errmessage{Error: unsupported option appendix=\axp@appendix\ for
  package apxproof}
}}}}
\ifthenelse{\equal{\axp@bibengine}{bibtex}}{
}{\ifthenelse{\equal{\axp@bibengine}{biblatex}}{
}{
  \errmessage{Error: unsupported option bibengine=\axp@bibengine\ for
  package apxproof}
}}
\ifthenelse{\equal{\axp@bibliography}{separate}}{
%    \end{macrocode}
% In |bibengine=bibtex| mode (the default), the external
% \textsf{bibunits} package is used to add a second bibliography for
% the appendix material. In |bibengine=biblatex| mode, no extra
% package is loaded here: \textsf{biblatex} is expected to have been
% loaded by the user, and its |refsection| environment is used at
% appendix typeset time to scope citations.
% \changes{v1.2.0}{2019/04/17}{Do not load \textsf{bibunits} if
% \texttt{bibliography} is set to \texttt{common}}
%    \begin{macrocode}
  \ifthenelse{\equal{\axp@bibengine}{bibtex}}{
    \RequirePackage{bibunits}
    \message{apxproof: Separate bibliography for appendix material (bibunits)}
  }{
    \message{apxproof: Separate bibliography for appendix material (biblatex)}
  }
}{\ifthenelse{\equal{\axp@bibliography}{common}}{
  \message{apxproof: Common bibliography for appendix and main text}
}{
  \errmessage{Error: unsupported option bibliography=\axp@bibliography\ for
  package apxproof}
}}
\ifthenelse{\equal{\axp@appendix}{chapterend}
            \AND\equal{\axp@bibliography}{separate}}{
  \PackageError{apxproof}{%
    bibliography=separate is not supported with appendix=chapterend}{%
    Use bibliography=common instead.}
}{}
\ifthenelse{\equal{\axp@repeqn}{same}}{
  \message{apxproof: Repeated equations keep the same numbering}
}{\ifthenelse{\equal{\axp@repeqn}{independent}}{
  \message{apxproof: Repeated equations are independently numbered}
}{
  \errmessage{Error: unsupported option repeqn=\axp@repeqn\ for
  package apxproof}
}}
%    \end{macrocode}
% \begin{macro}{\axp@forward@suppress}
%    \begin{macrocode}
\newbool{axp@forward@suppress}
\ifthenelse{\equal{\axp@forwardlinking}{yes}}{
}{\ifthenelse{\equal{\axp@forwardlinking}{no}}{
  \message{apxproof: Disable forward linking}
  \global\booltrue{axp@forward@suppress}%
}{
  \errmessage{Error: unsupported option forwardlinking=\axp@repeqn\ for
  package apxproof}
}}
%    \end{macrocode}
% \end{macro}
% \begin{macro}{axp@synctex}
% The |axp@synctex| Boolean records whether appendix material is to be
% re-read from the source files rather than copied to the |.axp| file (see
% Section~\ref{sec:synctex-impl}). With |synctex=auto|, this is decided by
% the |\synctex| primitive of the engine, which is nonzero exactly when
% Sync\TeX{} output was requested (positive for compressed, negative for
% uncompressed |.synctex| files); the primitive is only tested if it exists.
%    \begin{macrocode}
\newbool{axp@synctex}
\ifthenelse{\equal{\axp@synctex}{no}}{
}{\ifthenelse{\equal{\axp@synctex}{yes}}{
  \booltrue{axp@synctex}
}{\ifthenelse{\equal{\axp@synctex}{auto}}{
  \ifdefined\synctex
    \ifnum\synctex=\z@\else
      \booltrue{axp@synctex}
    \fi
  \fi
}{
  \errmessage{Error: unsupported option synctex=\axp@synctex\ for
  package apxproof}
}}}
%    \end{macrocode}
% Whether re-reading is actually possible further depends on the main source
% file being locatable, which is settled at the end of
% Section~\ref{sec:synctex-impl}; the Boolean may still be cleared there.
% \end{macro}
% \subsection{Macros Common to All Compilation Modes}
% \begin{macro}{\axp@newtheoremrep@definetheorem}
% \changes{v1.2.0}{2019/10/01}{Restore predefined theorem counters}
% \changes{v1.2.6}{2025/08/04}{Undefine counter printing command to
% avoid warning}
% Common to all compilation modes, we define
% |\axp@newtheoremrep@definetheorem|.
% When called with first argument |foobar|, we first undefine the existing
% |foobar| environment (and its counter) if it has already been defined (e.g., by the
% document class), then invoke |\axp@newtheorem| for the regular version
% of the theorem |foobar|, saving and restoring any existing theorem counter unless the
% |\newtheoremdep| redefines the base counter.
%    \begin{macrocode}
\def\axp@newtheoremrep@definetheorem#1#2#3#4{%
  \expandafter\let\csname #1\endcsname\undefined
  \expandafter\let\csname the#1\endcsname\undefined
  \ifcsname c@#1\endcsname
    \expandafter\expandafter\expandafter\let\expandafter\expandafter
      \csname c@axp@#1\endcsname\csname c@#1\endcsname
    \expandafter\let\csname c@#1\endcsname\undefined
  \fi
  \axp@newtheorem{#1}{#2}{#3}{#4}%
  \ifcsname c@axp@#1\endcsname
    \ifx\relax#2\relax
      \expandafter\expandafter\expandafter\let\expandafter\expandafter
        \csname c@#1\endcsname\csname c@axp@#1\endcsname
    \else
    \fi
  \fi
}
%    \end{macrocode}
% \end{macro}
% \begin{macro}{\axp@newtheorem}
% \changes{v1.0.6}{2018/05/10}{Introduce intermediary command for theorem
% macro}
% \begin{macro}{\@axp@newtheorem}
% \begin{macro}{\@@axp@newtheorem}
% We introduce an intermediate |\axp@newtheorem| command to define a new
% theorem, differently depending on whether there is a section counter or
% not. This will be useful, in particular to allow changing this
% definition depending on the document class. This command uses two
% intermediary commands, |\@axp@newtheorem| and |\@@axp@newtheorem|, for
% the non-starred and starred versions.
%    \begin{macrocode}
\def\axp@newtheorem{\@ifstar\@@axp@newtheorem\@axp@newtheorem}
\def\@axp@newtheorem#1#2#3#4{%
  \ifx\relax#4\relax
    \newtheorem{#1}[#2]{#3}%
  \else
    \newtheorem{#1}{#3}[#4]%
  \fi
}
\def\@@axp@newtheorem#1#2{%
  \newtheorem*{#1}{#2}%
}
%    \end{macrocode}
% \end{macro}
% \end{macro}
% \end{macro}
% \begin{macro}{\newtheoremrep}
% \begin{macro}{\axp@newtheoremreptmp}
% We define the high-level |\newtheoremrep| to have the same syntax as
% \textsf{amsthm}'s |\newtheorem|, plus an optional |*| variant that
% defers the proof to the appendix without restating the theorem.
% For this purpose, we need a little
% trick to deal with the second and fourth optional arguments, which is what
% |\@oparg| and |\axp@newtheoremreptmp| are
% used for. |\axp@newtheoremrep| is defined differently
% depending on the compilation mode.
% \changes{v1.4.0}{2026/05/14}{Added starred form of \cs{newtheoremrep}
% for proof-deferred theorems that are not restated in the appendix}
%    \begin{macrocode}
\newcommand\newtheoremrep{%
  \@ifstar{\axp@newtheoremrep@star}{\axp@newtheoremrep@nostar}%
}
\newcommand\axp@newtheoremrep@star[1]{%
  \global\csdef{axp@proofof@#1}{}%
  \@oparg{\axp@newtheoremreptmp{#1}}[]%
}
\newcommand\axp@newtheoremrep@nostar[1]{%
  \global\csundef{axp@proofof@#1}%
  \@oparg{\axp@newtheoremreptmp{#1}}[]%
}
\def\axp@newtheoremreptmp#1[#2]#3{%
  \@oparg{\axp@newtheoremrep{#1}[#2]{#3}}[]%
}
%    \end{macrocode}
% \end{macro}
% \end{macro}
% \begin{environment}{proofsketch}
% Simple |proofsketch| environment.
% \changes{v1.0.3}{2016/12/16}{Ignore spaces after beginning of Proof sketch}
%    \begin{macrocode}
  \newenvironment{proofsketch}{\begin{axp@oldproof}[Proof sketch]}{\end{axp@oldproof}}
%    \end{macrocode}
% \end{environment}
% \begin{macro}{\mainbodyrepeatedtheorem}
% \changes{v1.2.0}{2019/09/21}{Configurable repeated theorem command}
% \begin{macro}{\appendixrefname}
% \changes{v1.2.1}{2020/10/09}{Fix compatibility with
% \texttt{memoir} document class}
% \begin{macro}{\appendixbibliographystyle}
% \begin{macro}{\appendixbibliographyprelim}
% \begin{macro}{\appendixprelim}
% \changes{v1.0.4}{2017/03/03}{Configurable appendix style}
% \begin{macro}{\appendixsectionformat}
% \changes{v1.1.0}{2018/07/17}{Fix missing space in default
% \cs{appendixsectionformat}}
% \begin{macro}{\appendixproofname}
% \changes{v1.4.0}{2026/05/14}{Added \cs{appendixproofname} hook for
% customizing the title of a proof deferred to the appendix}
% We provide sensible defaults for these user-customizable macros.
% Even though they are not all useful in all modes, we define them for
% all modes so that a |\renewcommand| works in all cases.
%    \begin{macrocode}
\newcommand{\mainbodyrepeatedtheorem}{}
\providecommand{\appendixrefname}{References for the Appendix}
\newcommand{\appendixbibliographystyle}{alpha}
\newcommand{\appendixbibliographyprelim}{}
\newcommand{\appendixprelim}{\clearpage\onecolumn}
\newcommand{\appendixsectionformat}[2]{Proofs for Section~#1\ (#2)}
\newcommand{\appendixproofname}[2]{}
%    \end{macrocode}
% \end{macro}
% \end{macro}
% \end{macro}
% \end{macro}
% \end{macro}
% \end{macro}
% \end{macro}
% \begin{environment}{axp@oldproof}
%   We save the definition of the existing |proof| environment.
%    \begin{macrocode}
  \let\axp@oldproof\proof
  \let\endaxp@oldproof\endproof
%    \end{macrocode}
% \end{environment}
% We define a utility macro that will be used to properly set the
% |\label| command (and its \textsf{amsmath} counterpart,
% |\label@in@display|) for equations within repeated theorems, depending on the
% compilation mode.
%
% Some packages, \textsf{cleveref} in particular, give |\label| (and
% |\label@in@display|) an optional argument specifying the type of the
% counter being labeled. Our replacements must therefore look for such an
% optional argument and pass it on unchanged, or the opening bracket would
% be taken as the label name and the rest of the argument typeset as text.
% We do \emph{not} pass the optional argument on to the extra |-apx| label
% of the repeated material: that label is set within an unnumbered copy of
% the theorem, where overriding the counter type would apply to state that
% \textsf{cleveref} did not set up, and the type of the counter (for
% equations, the only case where the |-apx| label is really of use) is
% correctly inferred from the context anyway.
%
% \textsf{zref-clever}, the maintained successor of \textsf{cleveref}, does
% not touch |\label| at all: it sets its own labels through |\zlabel|, and,
% with its |labelhook| option (set by default), also through the kernel's
% |label| hook, so that a plain |\label| sets a \textsf{zref} label as well.
% The latter needs nothing from us, since our replacement of |\label| calls
% the original one. A |\zlabel| used directly, however, must be given the
% same treatment as |\label|; it has no optional argument.
% \begin{macro}{\axp@label}
% \begin{macro}{\axp@label@in@display}
% \changes{v1.5.0}{2026/08/03}{Support the optional argument that
% \textsf{cleveref} adds to \cs{label} (R. Rowe)}
% \changes{v1.5.0}{2026/08/03}{Support \cs{zlabel} of \textsf{zref-clever}}
%    \begin{macrocode}
  \def\axp@label{\@ifnextchar[\axp@label@opt\axp@label@noopt}%]
  \def\axp@label@opt[#1]#2{\axp@label@do{[#1]}{#2}}
  \def\axp@label@noopt#1{\axp@label@do{}{#1}}
  \def\axp@label@in@display{%
    \@ifnextchar[\axp@label@in@display@opt\axp@label@in@display@noopt}%]
  \def\axp@label@in@display@opt[#1]#2{\axp@label@in@display@do{[#1]}{#2}}
  \def\axp@label@in@display@noopt#1{\axp@label@in@display@do{}{#1}}
%    \end{macrocode}
% \end{macro}
% \end{macro}
% \begin{macro}{\axp@redefinelabels}
% \changes{v1.1.0}{2018/07/04}{Fix \cs{label} not being disabled
% in \textsf{amsmath} environments, where \cs{label@in@display} is
% used instead (K. D. Bauer)}
% \changes{v1.2.0}{2019/02/15}{Fix extra spacing erroneously
% introduced within the \texttt{\textbackslash axp@redefinelabels}
% macro}
%    \begin{macrocode}
  \newcommand{\axp@redefinelabels}{%
    \providecommand\label@in@display{}%
    \let\axp@oldlabel\label
    \let\axp@oldlabel@in@display\label@in@display
    \ifdefined\zlabel\let\axp@oldzlabel\zlabel\fi
    \ifthenelse{\equal{\axp@appendix}{inline}}{%
      \def\axp@label@do##1##2{%
        \axp@oldlabel##1{##2}%
        \axp@oldlabel{##2-apx}%
      }%
      \def\axp@label@in@display@do##1##2{%
        \axp@oldlabel@in@display##1{##2}%
        \axp@oldlabel{##2-apx}%
      }%
      \def\axp@zlabel##1{\axp@oldzlabel{##1}\axp@oldzlabel{##1-apx}}%
    }{%
      \def\axp@label@do##1##2{\axp@oldlabel{##2-apx}}%
      \def\axp@label@in@display@do##1##2{\axp@oldlabel@in@display{##2-apx}}%
      \def\axp@zlabel##1{\axp@oldzlabel{##1-apx}}%
    }%
    \let\label\axp@label
    \let\label@in@display\axp@label@in@display
    \ifdefined\zlabel\let\zlabel\axp@zlabel\fi
  }
%    \end{macrocode}
% \end{macro}
% \subsubsection{Class-Specific Behavior}
% Finally, some class-specific behavior common to all compilation modes.
% \paragraph{\textsf{llncs} and other Springer document classes}
% \changes{v1.0.6}{2018/05/10}{Better support of Springer document
% classes}
%     \begin{macrocode}
   \ifdefined\spnewtheorem
%    \end{macrocode}
% \begin{macro}{\@axp@newtheorem}
% \begin{macro}{\@@axp@newtheorem}
% It is necessary to use |\spnewtheorem| instead of |\newtheorem| in
% Springer document classes to obtain standard formatting.
%    \begin{macrocode}
\def\@axp@newtheorem#1#2#3#4{%
  \ifx\relax#4\relax
    \ifx\relax#2\relax
      \spnewtheorem{#1}{#3}{\bfseries}{\itshape}%
    \else
      \spnewtheorem{#1}[#2]{#3}{\bfseries}{\itshape}%
    \fi
  \else
    \spnewtheorem{#1}{#3}[#4]{\bfseries}{\itshape}%
  \fi
}
\def\@@axp@newtheorem#1#2{%
  \spnewtheorem*{#1}{#2}{\upshape\bfseries}{\itshape}%
}
%    \end{macrocode}
% \end{macro}
% \end{macro}
% \begin{macro}{proofsketch}
% \changes{v1.2.0}{2019/02/15}{Fix proof sketches in inline
% compilation mode for Springer document classes}
% We redefine the |proofsketch| environment, which is used
% differently in the base class.
%    \begin{macrocode}
\renewenvironment{proofsketch}{\begin{axp@oldproof}[sketch]}{\end{axp@oldproof}}
%    \end{macrocode}
% \end{macro}
% We have to redefine the macro |\@thmcountersep| for proper sectioned
% counters.
%    \begin{macrocode}
\def\@thmcountersep{.}
  \fi
%    \end{macrocode}
% \paragraph{\textsf{acmart}}
% \changes{v1.2.6}{2025/09/29}{Fix compilation issues in recent versions of \textsf{acmart}}
% Some versions of \textsf{acmart} define a |\ACM@origsection| macro and
% verify that |\section| has the same content. We need to update
% |\ACM@origsection| accordingly.
%    \begin{macrocode}
\ifdefined\ACM@origsection
  \AtBeginDocument{\let\ACM@origsection\section}
\fi
%    \end{macrocode}
% \subsection{Sync\TeX{} Support}
% \label{sec:synctex-impl}
% \changes{v1.5.0}{2026/08/19}{Sync\TeX{} support: appendix material can
% be re-read from the source files instead of being copied to the
% \texttt{.axp} file}
% Sync\TeX{} records, for every box in the output, the file and line
% from which the tokens that produced it were read. Two things defeat it
% in \textsf{apxproof}: the body of a |foobarrep| environment collected by
% \textsf{environ} is typeset from a macro, so all of it is attributed to
% the line of the |\end{foobarrep}|; and appendix material copied
% verbatim to the |.axp| file is attributed to that file when it is
% typeset. There is no way to tell Sync\TeX{} that some tokens ``really''
% come from elsewhere; the only tokens it attributes correctly are those
% read directly from a file. When the |synctex| option is enabled, we
% therefore do just that: repeated theorems are typeset directly from the
% source instead of being collected, and instead of copying appendix
% material to the |.axp| file, we write there the name of the source
% file, the line where the environment begins, the line where it ends
% and its name; when the appendix is typeset, the source file is opened
% again with the primitive |\input|, all lines up to the beginning of the
% environment are skipped, and the environment is executed anew, with its
% begin and end code replaced by what the appendix needs. All of this
% happens with the source file open, so Sync\TeX{} attributes every box
% to the right file and line.
%
% We store, at the beginning of each captured environment, where it was
% found. The line comes from |\@currenvline|, which |\begin| sets before
% reading any optional argument (|\inputlineno| would be off when the
% argument or the body starts on a later line). The file comes from the
% kernel's |\CurrentFileUsed| and |\CurrentFilePathUsed|, which are empty
% for the main file, in which case we use |\axp@region@mainfile|, the name
% under which the main file can be read (see |\axp@region@findmain| below);
% older kernels lack these macros, and we then have to assume that every
% environment is in the main file.
%
% We also need to know how Sync\TeX{} identifies the file, see
% |\axp@region@input| below: with Lua\TeX{}, we record the tag it gave
% the file (|\axp@region@tag| is 0 elsewhere).
% \begin{macro}{\axp@region@file}
% \begin{macro}{\axp@region@line}
% \begin{macro}{\axp@region@env}
% \begin{macro}{\axp@region@tag}
% \begin{macro}{\axp@region@mainfile}
% \begin{macro}{\axp@extract@line}
% \begin{macro}{\axp@region@record}
%    \begin{macrocode}
\def\axp@region@file{}
\def\axp@region@line{}
\def\axp@region@env{}
\def\axp@region@tag{0}
\edef\axp@region@mainfile{\jobname.tex}
\def\axp@extract@line#1line #2\@nil#3\@nnil{#2}
\def\axp@region@record{%
  \ifdefined\CurrentFileUsed
    \ifx\CurrentFileUsed\@empty
      \let\axp@region@file\axp@region@mainfile
    \else
      \ifx\CurrentFilePathUsed\@empty
        \edef\axp@region@file{\CurrentFileUsed}%
      \else
        \edef\axp@region@file{\CurrentFilePathUsed/\CurrentFileUsed}%
      \fi
    \fi
  \else
    \let\axp@region@file\axp@region@mainfile
  \fi
  \edef\axp@region@line{%
    \expandafter\axp@extract@line\@currenvline\@nil line 0\@nil\@nnil}%
  \edef\axp@region@env{\@currenvir}%
  \ifdefined\directlua
    \edef\axp@region@tag{\directlua{tex.sprint(tex.get_synctex_tag())}}%
  \else
    \def\axp@region@tag{0}%
  \fi
}
\ifbool{axp@synctex}{
  \ifdefined\CurrentFileUsed\else
    \PackageWarning{apxproof}{%
      This LaTeX kernel does not provide \string\CurrentFileUsed;
      the synctex option assumes\MessageBreak
      that all appendix material is in \jobname.tex}
  \fi
}{}
%    \end{macrocode}
% \end{macro}
% \end{macro}
% \end{macro}
% \end{macro}
% \end{macro}
% \end{macro}
% \end{macro}
% \begin{macro}{\axp@region@write}
% At the end of the environment, this writes to the |.axp| file the
% |\axp@replaybody| command replaying it, with the recorded position, the
% current line (that of the |\end|), the two arguments passed by
% the caller: the tokens that will replace the begin code of the
% environment in the appendix, and those that will replace its end code,
% and the Sync\TeX{} tag.
% The trailing |\relax| ensures that the end of the line in the |.axp|
% file, which is only tokenized once the replay is over, does not
% produce a space, as the copied material never ended with one.
%    \begin{macrocode}
\def\axp@region@write#1#2{%
  \immediate\write\axp@proofsfile{%
    \noexpand\axp@replaybody{\axp@region@file}{\axp@region@line}%
      {\the\inputlineno}{\axp@region@env}{#1}{#2}{\axp@region@tag}\relax
  }%
}
%    \end{macrocode}
% \end{macro}
% \begin{macro}{\axp@replaybody}
% |\axp@replaybody|\marg{file}\marg{start}\marg{end}\marg{env}\marg{begin}\marg{end code}\marg{tag}
% is executed in the appendix. It first checks, by reading the file with
% |\readline| (which does not affect Sync\TeX{}), that the file exists and
% that its \meta{start} line does contain |\begin{|\meta{env}|}| and its
% \meta{end} line an |\end{|; if not, an error is raised and nothing is
% typeset. Otherwise the file is replayed. The state of the replay is kept
% in |\axp@replay@...| macros, distinct from the |\axp@region@...| ones
% set by |\axp@region@record|: a |foobarrep| environment nested in a
% |toappendix| environment runs its main-text begin code when replayed,
% and must not disturb the replay of the enclosing environment.
%    \begin{macrocode}
\newif\ifaxp@region@ok
\newif\ifaxp@region@wrapped
\def\axp@replaybody#1#2#3#4#5#6#7{%
  \def\axp@replay@file{#1}%
  \def\axp@replay@line{#2}%
  \def\axp@replay@endline{#3}%
  \def\axp@replay@env{#4}%
  \def\axp@replay@begin{#5}%
  \def\axp@replay@end{#6}%
  \def\axp@replay@tag{#7}%
  \axp@region@check
  \ifaxp@region@ok
    \axp@region@replay
  \else
    \axp@region@flushsection
  \fi
}
%    \end{macrocode}
% \end{macro}
% \begin{macro}{\axp@region@section}
% \begin{macro}{\axp@region@flushsection}
% For |toappendix| environments, the appendix section heading is written
% to the |.axp| file as |\axp@region@section|, which stores it; it is
% typeset by |\axp@region@toappendix| below with the same look-ahead as
% |\axp@section|, once the content of the environment is available, or
% by |\axp@region@flushsection| without look-ahead if the environment
% cannot be replayed. The assignments are global, as the heading is
% typeset from within the group of the environment.
%    \begin{macrocode}
\let\axp@region@pendingsection\relax
\def\axp@region@section#1{\gdef\axp@region@pendingsection{#1}}
\def\axp@region@flushsection{%
  \ifx\axp@region@pendingsection\relax\else
    \expandafter\axp@oldsection\expandafter
      {\axp@region@pendingsection}%
    \global\let\axp@region@pendingsection\relax
  \fi
}
%    \end{macrocode}
% \end{macro}
% \end{macro}
% \begin{macro}{\axp@region@check}
% \begin{macro}{\axp@region@readto}
% \begin{macro}{\axp@region@readone}
% Lines are read one at a time into |\axp@line|, with |\endlinechar|
% disabled so that no end-of-line character is appended, until the
% desired line number is reached; |\axp@region@readcount| counts the lines
% read so far. We never read from a stream on which |\ifeof| is true, since
% \TeX{} would then read from the terminal. As the environments of a
% given file are replayed in the order in which they appear, we keep the
% stream open from one replay to the next and only reopen the file when
% we need an earlier line or a different file (|\axp@region@openfile|
% records the file currently open); each source file is thus read at
% most a few times overall. This state is global, as |\axp@replaybody|
% may be called within a group.
%    \begin{macrocode}
\newread\axp@readfile
\newcount\axp@region@readcount
\def\axp@region@openfile{}
\def\axp@line{}
\def\axp@region@readto#1{%
  \loop
    \ifnum\axp@region@readcount<#1\relax
      \axp@region@readone{#1}%
  \repeat
}
\def\axp@region@readone#1{%
  \ifeof\axp@readfile
    \def\axp@line{}%
    \global\axp@region@readcount=#1\relax
  \else
    \global\advance\axp@region@readcount\@ne
    \readline\axp@readfile to\axp@line
  \fi
}
\def\axp@region@check{%
  \axp@region@okfalse
  \edef\axp@region@saveendlinechar{\the\endlinechar}%
  \endlinechar\m@ne
  \ifx\axp@region@openfile\axp@replay@file
    \ifnum\axp@region@readcount<\axp@replay@line\relax
    \else
      \gdef\axp@region@openfile{}%
    \fi
  \else
    \gdef\axp@region@openfile{}%
  \fi
  \ifx\axp@region@openfile\@empty
    \closein\axp@readfile
    \openin\axp@readfile=\axp@replay@file\relax
    \global\axp@region@readcount=\z@
    \global\let\axp@region@openfile\axp@replay@file
  \fi
  \ifeof\axp@readfile
    \ifnum\axp@region@readcount=\z@
      \PackageError{apxproof}{%
        Cannot open `\axp@replay@file' to re-read the\MessageBreak
        \axp@replay@env\space environment starting on line
        \axp@replay@line}{%
        The synctex option of apxproof requires the source files of the
        document\MessageBreak
        to be readable under the same names when the appendix is
        typeset.\MessageBreak
        This appendix material will be missing.}%
      \gdef\axp@region@openfile{}%
    \else
      \axp@region@toofar
    \fi
  \else
    \axp@region@readto{\axp@replay@line}%
    \ifeof\axp@readfile
      \axp@region@toofar
    \else
      \axp@region@analyze
      \ifaxp@region@ok
        \axp@region@readto{\axp@replay@endline}%
        \axp@region@checkend
      \fi
    \fi
  \fi
  \endlinechar\axp@region@saveendlinechar\relax
}
\def\axp@region@toofar{%
  \axp@region@okfalse
  \PackageError{apxproof}{%
    File `\axp@replay@file' has fewer than \axp@replay@line\space
    lines, where\MessageBreak
    the \axp@replay@env\space environment to re-read was expected}{%
    The source file has probably changed since the main text was
    processed,\MessageBreak
    or the main file is not `\jobname.tex'.
    This appendix material will be missing.}%
}
%    \end{macrocode}
% \end{macro}
% \end{macro}
% \end{macro}
% Several macros need |\begin{|, |\end{|, |}| or an end-of-line character
% as delimiters, with the catcode 12 that these characters have when
% read from |\readline| or with the catcodes we set to skip lines. We define
% them with |\lowercase|, mapping placeholder characters to the ones we need,
% which preserves catcodes.
% \begin{macro}{\axp@region@scanline}
% \begin{macro}{\axp@region@scan}
% \begin{macro}{\axp@region@getname}
% \begin{macro}{\axp@region@scanendline}
% \begin{macro}{\axp@region@scanend}
% \begin{macro}{\axp@gobbleline}
% \begin{macro}{\axp@gobbletobegin}
% \begin{itemize}
% \item |\axp@region@scanline| splits its argument at the first |\begin{|;
% \item |\axp@region@getname| splits its argument at the first |}|;
% \item |\axp@region@scanendline| splits its argument at the first |\end{|;
% \item |\axp@gobbleline| discards the input up to and including the next
% end of line;
% \item |\axp@gobbletobegin| discards the input up to and including the
% next |\begin{|.
% \end{itemize}
% In the first three, a sentinel is appended so that a match is always
% found; the last two then continue with the macros indicated.
%    \begin{macrocode}
\begingroup
  \lccode`\!=`\\ \lccode`\?=`\{ \lccode`\|=`\} \lccode`\"=`\^^M
  \lccode`\+=`\b \lccode`\-=`\e \lccode`\*=`\g \lccode`\/=`\i
  \lccode`\==`\n \lccode`\,=`\d
  \catcode`\!=12 \catcode`\?=12 \catcode`\|=12 \catcode`\"=12
  \catcode`\+=12 \catcode`\-=12 \catcode`\*=12 \catcode`\/=12
  \catcode`\==12 \catcode`\,=12
  \lowercase{\endgroup
    \long\def\axp@region@scanline#1{%
      \axp@region@scan#1!+-*/=?\axp@region@marker\@nil}
    \long\def\axp@region@scan#1!+-*/=?#2\@nil{\axp@region@scan@i{#2}}
    \long\def\axp@region@getname#1{\axp@region@getname@i#1|\@nil}
    \long\def\axp@region@getname@i#1|#2\@nil{\def\axp@region@name{#1}}
    \long\def\axp@region@scanendline#1{%
      \axp@region@scanend#1!-=,?\axp@region@marker\@nil}
    \long\def\axp@region@scanend#1!-=,?#2\@nil{%
      \def\axp@region@rest{#2}}
    \long\def\axp@gobbleline#1"{\axp@skiplines}
    \long\def\axp@gobbletobegin#1!+-*/=?{\axp@region@nextbegin}
  }
%    \end{macrocode}
% \end{macro}
% \end{macro}
% \end{macro}
% \end{macro}
% \end{macro}
% \end{macro}
% \end{macro}
% \begin{macro}{\axp@region@analyze}
% \begin{macro}{\axp@region@scan@i}
% We look on the start line for |\begin{|\meta{env}|}|. As the
% environment may be wrapped in a user-defined environment (in which
% case the line contains, say, |\begin{mythm}|), we settle for the first
% |\begin{| on the line if the exact one is absent. |\axp@region@begins|
% is set to the number of |\begin{| to skip on the line before the one
% to use.
%    \begin{macrocode}
\newcount\axp@region@begins
\newcount\axp@skipcount
\newif\ifaxp@region@found
\def\axp@region@marker{\axp@region@marker}
\def\axp@region@analyze{%
  \axp@region@begins=\z@
  \axp@region@foundfalse
  \edef\axp@region@target{%
    \expandafter\detokenize\expandafter{\axp@replay@env}}%
  \expandafter\axp@region@scanline\expandafter{\axp@line}%
  \ifaxp@region@found
    \axp@region@oktrue
  \else
    \ifnum\axp@region@begins>\z@
      \axp@region@begins=\z@
      \axp@region@oktrue
    \else
      \PackageError{apxproof}{%
        Cannot find \string\begin\@charlb\axp@replay@env\@charrb\space
        on line \axp@replay@line\space of\MessageBreak
        `\axp@replay@file' to re-read it}{%
        The synctex option of apxproof requires
        \string\begin\@charlb\axp@replay@env\@charrb\space
        to appear\MessageBreak
        literally in the source file, on the line where the environment
        starts.\MessageBreak
        This appendix material will be missing.}%
    \fi
  \fi
}
%    \end{macrocode}
% |\axp@region@scan@i| receives what follows a |\begin{|. If this is only
% the sentinel, there is no further |\begin{| on the line. Otherwise
% we extract the environment name and compare it with the target,
% moving on to the next |\begin{| if it differs.
%    \begin{macrocode}
\long\def\axp@region@scan@i#1{%
  \def\axp@region@rest{#1}%
  \ifx\axp@region@rest\axp@region@marker
  \else
    \axp@region@getname{#1}%
    \ifx\axp@region@name\axp@region@target
      \axp@region@foundtrue
    \else
      \advance\axp@region@begins\@ne
      \axp@region@scan#1\@nil
    \fi
  \fi
}
%    \end{macrocode}
% \end{macro}
% \end{macro}
% \begin{macro}{\axp@region@checkend}
% Similarly, we check that the end line contains an |\end{|.
%    \begin{macrocode}
\def\axp@region@checkend{%
  \expandafter\axp@region@scanendline\expandafter{\axp@line}%
  \ifx\axp@region@rest\axp@region@marker
    \axp@region@okfalse
    \PackageError{apxproof}{%
      Cannot find the \string\end\space of the \axp@replay@env\space
      environment\MessageBreak
      starting on line \axp@replay@line\space of `\axp@replay@file'
      (expected on line \axp@replay@endline)}{%
      The synctex option of apxproof requires the \string\end\space
      of the environment\MessageBreak
      to appear literally in the source file. This appendix material
      will be missing.}%
  \fi
}
%    \end{macrocode}
% \end{macro}
% \begin{macro}{\axp@region@replay}
% \begin{macro}{\axp@region@cleanup}
% \begin{macro}{\axp@region@catcodes}
% Now the actual replay. Within the appendix, |\begin{|\meta{env}|}| must
% run the begin code we were given and |\end{|\meta{env}|}| the end code;
% we save the current meaning of the corresponding macros, and of |\end|
% itself (see below), to restore them afterwards. We also save the catcodes of all characters, as they will
% be modified at the end of the replay (see below): |\axp@region@catcodes|
% expands to the assignments restoring them. We deliberately do not
% wrap the replay in a group, so that whatever the |\end| of the
% environment does to the paragraph that follows (as |\@doendpe| does)
% survives it, exactly as when the material is copied to the |.axp| file.
% The file is then opened by the primitive |\input| in a group where every
% character is ignored (catcode 9) except the end-of-line character,
% which is an ordinary one: lines are skipped without \TeX{} building any
% token, until the start line, which is read with every character an
% ordinary one (so that none is special, no control sequence is formed
% and no line can be mistaken for a paragraph break); the group ends when
% we hand over to the environment.
%    \begin{macrocode}
\def\axp@region@catcodes#1{%
  \catcode#1=\the\catcode#1\relax
  \ifnum#1<255
    \expandafter\axp@region@catcodes\expandafter
      {\number\numexpr#1+1\expandafter}%
  \fi}
\def\axp@region@setcatcodes#1{%
  \count@=\z@
  \loop
    \catcode\count@=#1\relax
    \ifnum\count@<255
    \advance\count@\@ne
  \repeat}
\def\axp@region@replay{%
  \edef\axp@region@catcoderestore{\axp@region@catcodes{0}}%
  \expandafter\let\expandafter\axp@region@savedbegin
    \csname\axp@replay@env\endcsname
  \expandafter\let\expandafter\axp@region@savedend
    \csname end\axp@replay@env\endcsname
  \expandafter\let\csname\axp@replay@env\endcsname\axp@replay@begin
  \expandafter\let\csname end\axp@replay@env\endcsname\axp@replay@end
  \let\axp@region@savedendcmd\end
  \axp@region@wrappedfalse
  \begingroup
    \axp@region@setcatcodes{9}%
    \catcode13=12\relax
    \endlinechar=13\relax
    \axp@skipcount=\axp@replay@line\relax
    \advance\axp@skipcount\m@ne
    \axp@region@input
  \axp@region@cleanup
}
\def\axp@region@cleanup{%
  \axp@region@input@done
  \expandafter\let\csname\axp@replay@env\endcsname\axp@region@savedbegin
  \expandafter\let\csname end\axp@replay@env\endcsname\axp@region@savedend
  \let\end\axp@region@savedendcmd
  \expandafter\let\csname axp@region@isregion@\axp@replay@env\endcsname
    \undefined
  \ifaxp@region@wrapped
    \expandafter\let\csname end\axp@region@name\endcsname
      \axp@region@savedwrappedend
  \fi
  \axp@region@catcoderestore
}
%    \end{macrocode}
% \end{macro}
% \end{macro}
% \end{macro}
% \begin{macro}{\axp@region@input}
% \begin{macro}{\axp@region@input@done}
% \begin{macro}{\axp@region@altfile}
% Opening the file is where we have to care about the way Sync\TeX{}
% names files. \TeX{} gives every file it opens a new tag, and records
% the tag and name in the |.synctex| file, so re-reading a file
% registers it a second time under the same name; and the parser used by
% viewers and editors, when asked (forward search, from the editor to
% the PDF) for a source line of some file, resolves the name to a single
% tag, the last one registered under that name. With the re-read
% registered as is, forward search from the main text would then land in
% the appendix, or nowhere. Two remedies:
% \begin{itemize}
% \item With Lua\TeX{}, we can have the nodes created while re-reading
% carry the file's original tag, with the Lua function
% |tex.force_synctex_tag| (which requires |tex.set_synctex_mode(2)|;
% line numbers remain the actual ones), and
% not register the re-read at all (|tex.set_synctex_no_files|). Both
% directions then work everywhere. Files |\input| from within the
% material are handled normally again during their reading, thanks to
% the kernel's file hooks. All this is done in Lua, see below.
% \item With other engines, we open the file under a name that the parser
% does not consider equivalent to the recorded one, but which designates
% the same file and which editors normalize:
% \meta{cwd}|/../|\meta{cwd name}|/|\meta{file}. Forward search then
% behaves as when the material was copied (main text only), while
% inverse search from the appendix gives the right file and line. The
% working directory is read from the |.fls| file written by the
% |-recorder| option (used by \textsf{latexmk} and most editors);
% failing that, the file is opened under its plain name.
% \end{itemize}
%    \begin{macrocode}
\ifdefined\directlua
  \def\axp@region@input{%
    \ifnum\axp@replay@tag>\z@
      \directlua{apxproof.synctex_force(\axp@replay@tag)}%
      \global\let\axp@region@forcedtag\axp@replay@tag
    \fi
    \expandafter\axp@skiplines\@@input\axp@replay@file\relax}
  \def\axp@region@input@done{%
    \ifnum\axp@replay@tag>\z@
      \directlua{apxproof.synctex_unforce()}%
      \global\let\axp@region@forcedtag\relax
    \fi}
  \let\axp@region@forcedtag\relax
  \ifdefined\AddToHook
    \AddToHook{file/before}{%
      \ifx\axp@region@forcedtag\relax\else
        \directlua{apxproof.synctex_unforce()}%
      \fi}
    \AddToHook{file/after}{%
      \ifx\axp@region@forcedtag\relax\else
        \directlua{apxproof.synctex_force(\axp@region@forcedtag)}%
      \fi}
  \fi
%    \end{macrocode}
% The Lua side. Forcing the tag does not reach nodes created from Lua
% (such as the glyphs that \textsf{luaotfload} inserts for ligatures),
% which in this mode get no tag and line at all and would be attributed,
% when the page is shipped out, to wherever \TeX{} is then. A filter run
% before line breaking and box packing therefore gives such nodes, while
% a tag is forced, the tag and the line of the nearest preceding node
% (or following one, at the start of a list). (Font kerns are a
% different matter: Lua\TeX{} records them with the position of the
% enclosing |\input| rather than with the position of the glyphs they
% sit between, in the main text just as well, and does not use the
% fields we could set; this is a Lua\TeX{} peculiarity that we leave
% alone.)
%    \begin{macrocode}
  \directlua{
    apxproof = apxproof or {}
    apxproof.forced_tag = 0
    local get, set = node.direct.get_synctex_fields,
                     node.direct.set_synctex_fields
    local traverse, todirect = node.direct.traverse, node.direct.todirect
    function apxproof.synctex_force(tag)
      apxproof.saved_tag, apxproof.saved_line =
        tex.get_synctex_tag(), tex.get_synctex_line()
      tex.set_synctex_no_files(1)
      tex.set_synctex_mode(2)
      tex.force_synctex_tag(tag)
      tex.set_synctex_tag(tag)
      tex.set_synctex_line(0)
      apxproof.forced_tag = tag
    end
    function apxproof.synctex_unforce()
      apxproof.forced_tag = 0
      tex.force_synctex_tag(0)
      tex.set_synctex_tag(apxproof.saved_tag or 0)
      tex.set_synctex_line(apxproof.saved_line or 0)
      tex.set_synctex_mode(0)
      tex.set_synctex_no_files(0)
    end
    local getprev, tail = node.direct.getprev, node.direct.tail
    function apxproof.synctex_fix(head)
      local tag = apxproof.forced_tag
      if tag == 0 then return true end
      local last_l = 0
      local pending = false
      for n in traverse(todirect(head)) do
        local t, l = get(n)
        if t == 0 or (t == tag and l == 0) then
          if last_l > 0 then
            set(n, tag, last_l)
          else
            pending = true
          end
        elseif t == tag then
          last_l = l
        end
      end
      if pending then
        local n = tail(todirect(head))
        last_l = 0
        while n do
          local t, l = get(n)
          if t == 0 or (t == tag and l == 0) then
            if last_l > 0 then set(n, tag, last_l) end
          elseif t == tag then
            last_l = l
          end
          n = getprev(n)
        end
      end
      return true
    end
    luatexbase.add_to_callback("pre_linebreak_filter",
      apxproof.synctex_fix, "apxproof.synctex")
    luatexbase.add_to_callback("hpack_filter",
      apxproof.synctex_fix, "apxproof.synctex")
  }
\else
  \def\axp@region@input{%
    \axp@region@altfile
    \expandafter\axp@skiplines\@@input\axp@replay@altfile\relax}
  \let\axp@region@input@done\relax
\fi
%    \end{macrocode}
% |\axp@region@altfile| defines |\axp@replay@altfile| from
% |\axp@replay@file|. The working directory is looked up once, in the
% first line of the |.fls| file, which reads |PWD |\meta{directory}; its
% last component is extracted by |\axp@region@lastcomponent|. An
% absolute file name (an unusual case) is prefixed with a second |/|
% instead, which has the same effect. As the same read stream is used, we
% forget the file the check pass may have kept open.
%    \begin{macrocode}
\let\axp@region@pwd\relax
\def\axp@region@altfile{%
  \ifx\axp@region@pwd\relax
    \axp@region@findpwd
  \fi
  \ifx\axp@region@pwd\@empty
    \let\axp@replay@altfile\axp@replay@file
  \else
    \expandafter\axp@region@altfile@i\axp@replay@file\@nil
  \fi}
\def\axp@region@altfile@i#1#2\@nil{%
  \if/#1%
    \edef\axp@replay@altfile{/\axp@replay@file}%
  \else
    \edef\axp@replay@altfile{%
      \axp@region@pwd/../\axp@region@pwdbase/\axp@replay@file}%
  \fi}
\def\axp@region@findpwd{%
  \gdef\axp@region@pwd{}%
  \IfFileExists{\jobname.fls}{%
    \begingroup
      \endlinechar\m@ne
      \openin\axp@readfile=\jobname.fls\relax
      \ifeof\axp@readfile\else
        \readline\axp@readfile to\axp@line
        \expandafter\axp@region@findpwd@i\axp@line
          \relax\relax\relax\relax\@nil
      \fi
      \closein\axp@readfile
    \endgroup
    \gdef\axp@region@openfile{}%
  }{}%
}
\def\axp@region@findpwd@i#1#2#3#4\@nil{%
  \def\axp@tmp{#1#2#3}%
  \edef\axp@tmpa{\detokenize{PWD}}%
  \ifx\axp@tmp\axp@tmpa
    \axp@region@findpwd@ii#4\@nil
  \fi}
\def\axp@region@findpwd@ii#1\relax#2\@nil{%
  \axp@region@findpwd@iii#1\@nnil\@nnil}
\def\axp@region@findpwd@iii#1#2\@nnil#3\@nnil{%
  \ifx\@nnil#1\else
    \xdef\axp@region@pwd{#1#2}%
    \xdef\axp@region@pwdbase{\axp@region@lastcomponent#1#2/\@nil}%
    \ifx\axp@region@pwdbase\@empty
      \gdef\axp@region@pwd{}%
    \fi
  \fi}
\def\axp@region@lastcomponent#1/#2\@nil{%
  \ifx\@empty#2\@empty
    #1%
  \else
    \axp@region@lastcomponent#2\@nil
  \fi}
%    \end{macrocode}
% \end{macro}
% \end{macro}
% \end{macro}
% \begin{macro}{\axp@region@findmain}
% \changes{v1.5.0}{2026/08/25}{Locate the main source file through the
% \texttt{.fls} file, and fall back to copying when it cannot be read}
% The kernel leaves |\CurrentFileUsed| empty for the main file and tells us
% nothing about the name \TeX{} opened it under, so |\jobname.tex| is the
% obvious guess; it is wrong whenever the document was compiled by giving
% \TeX{} a path, as in |pdflatex sub/doc.tex|, and the material would then
% all be missing from the appendix. The |.fls| file records the name that
% was actually used, in the first |INPUT| line whose last component is
% |\jobname.tex| (the entries before it are the configuration files and the
% format). We look it up once, here, rather than at replay time: the
% Boolean is consulted while theorems are being defined in the preamble,
% and the |.fls| is already complete that early, since \TeX{} opens the
% main file before reading anything of it. The lines are parsed as in
% |\axp@region@findpwd| above, with |INPUT| in place of |PWD|, and
% |\axp@region@stopmain| ends the scan on the first match or at the end of
% the file.
%    \begin{macrocode}
\newif\ifaxp@region@stopmain
\def\axp@region@findmain{%
  \IfFileExists{\jobname.fls}{%
    \begingroup
      \endlinechar\m@ne
      \edef\axp@tmp{\jobname.tex}%
      \edef\axp@region@jobnametex{%
        \expandafter\detokenize\expandafter{\axp@tmp}}%
      \openin\axp@readfile=\jobname.fls\relax
      \axp@region@stopmainfalse
      \loop
        \ifaxp@region@stopmain\else
          \ifeof\axp@readfile
            \axp@region@stopmaintrue
          \else
            \readline\axp@readfile to\axp@line
            \expandafter\axp@region@findmain@i\axp@line
              \relax\relax\relax\relax\relax\relax\@nil
          \fi
      \repeat
      \closein\axp@readfile
    \endgroup
    \gdef\axp@region@openfile{}%
  }{}%
}
\def\axp@region@findmain@i#1#2#3#4#5#6\@nil{%
  \def\axp@tmp{#1#2#3#4#5}%
  \edef\axp@tmpa{\detokenize{INPUT}}%
  \ifx\axp@tmp\axp@tmpa
    \axp@region@findmain@ii#6\@nil
  \fi}
\def\axp@region@findmain@ii#1\relax#2\@nil{%
  \axp@region@findmain@iii#1\@nnil\@nnil}
\def\axp@region@findmain@iii#1#2\@nnil#3\@nnil{%
  \ifx\@nnil#1\else
    \edef\axp@tmpa{\axp@region@lastcomponent#1#2/\@nil}%
    \ifx\axp@tmpa\axp@region@jobnametex
      \xdef\axp@region@mainfile{#1#2}%
      \axp@region@stopmaintrue
    \fi
  \fi}
%    \end{macrocode}
% \end{macro}
% \begin{macro}{\axp@region@setup}
% We can now settle whether re-reading is possible at all. If the main file
% cannot be read under the name we have, nothing of it could be replayed,
% and every piece of appendix material coming from it would be lost; we
% therefore clear the Boolean, so that the whole package behaves exactly as
% with |synctex=no|, and say so. The document is then complete, only
% without Sync\TeX{} support for appendix material. This has to happen
% before any |\newtheoremrep|, which consults the Boolean to decide how to
% define the |foobarrep| environment.
%    \begin{macrocode}
\ifbool{axp@synctex}{
  \axp@region@findmain
  \IfFileExists{\axp@region@mainfile}{%
    \message{apxproof: SyncTeX support enabled, appendix material re-read
      from source files}%
  }{%
    \boolfalse{axp@synctex}%
    \PackageWarning{apxproof}{%
      Cannot read the main source file `\axp@region@mainfile';
      the synctex\MessageBreak
      option cannot re-read appendix material from it. Falling back
      to\MessageBreak
      copying it to the .axp file, as with synctex=no: the document
      is\MessageBreak
      complete, but SyncTeX does not cover appendix material.
      This\MessageBreak
      happens when TeX is given a path to the main file and no .fls
      file\MessageBreak
      is available to resolve it (option -recorder)}%
  }%
}{}
%    \end{macrocode}
% \end{macro}
% \begin{macro}{\axp@skiplines}
% \begin{macro}{\axp@region@nextbegin}
% \begin{macro}{\axp@region@afterbegin}
% Note that the tokens following |\@@input| in |\axp@region@replay|
% are only read once the file has been closed, and that when
% |\axp@skiplines| starts, the first line of the file has been read into
% \TeX's buffer but nothing has been tokenized yet. We gobble whole lines
% until the start line, which is read with all characters ordinary,
% then |\begin{| as many times as needed, and
% close the catcode group: the rest of the line, still in the buffer,
% will be tokenized with the normal catcodes, but with |@| an ordinary
% character as in the rest of the document.
%    \begin{macrocode}
\def\axp@skiplines{%
  \ifnum\axp@skipcount>\z@
    \advance\axp@skipcount\m@ne
    \expandafter\axp@gobbleline
  \else
    \axp@region@setcatcodes{12}%
    \axp@skipcount=\axp@region@begins
    \expandafter\axp@gobbletobegin
  \fi}
\def\axp@region@nextbegin{%
  \ifnum\axp@skipcount>\z@
    \advance\axp@skipcount\m@ne
    \expandafter\axp@gobbletobegin
  \else
    \expandafter\axp@region@afterbegin
  \fi}
\def\axp@region@afterbegin{%
  \endgroup
  \catcode`\@=12
  \def\axp@region@name{}%
  \axp@region@readname}
%    \end{macrocode}
% \end{macro}
% \end{macro}
% \end{macro}
% \begin{macro}{\axp@region@readname}
% \begin{macro}{\axp@region@readname@go}
% The name of the environment is then read character by character up to
% the closing brace, which is consumed by a |\let|. (A macro cannot insert
% an unbalanced |{| into the input, and a delimited argument cannot be
% delimited by |}|.)
%
% If the name is the expected one, we do not execute |\begin| at all,
% as material copied to the |.axp| file was not enclosed in a group, and
% we want the same here (a group would, e.g., undo what a
% |\subsection| in a |toappendix| environment does to the paragraph that
% follows it). We run the begin code directly, and let |\end| be
% |\axp@region@endcmd|, which runs the end code and |\axp@region@finish|
% for the expected environment, and delegates every other name to the
% saved |\end|; nested environments in the material thus behave as usual.
% The test on the name is done with |\ifcsname| on a flag set for the
% expected environment, so that the wrapper is expandable, as the kernel's
% |\end| is, up to the point where the original |\end| takes over: this
% matters when |\end{tabular}| is met while \TeX{} scans the beginning of
% an alignment row, where anything unexpandable would open a new row.
% |\axp@region@finish| makes every character ignored, so that whatever
% remains of the |\end| line is dropped by \TeX's tokenizer, and stops
% reading the file with |\endinput|. Doing this with catcodes rather than
% by grabbing the rest of the line is necessary, because pending tokens
% of the macros being executed would otherwise be grabbed as well.
%
% If the name is different (the environment is wrapped in a user
% environment), we execute |\begin| with that name; the environment's
% end code (the user's) is extended with |\aftergroup\axp@region@finish|
% so that the file is closed right after the group of the environment
% ends.
%    \begin{macrocode}
\def\axp@region@readname{%
  \futurelet\axp@region@next\axp@region@readname@i}
\def\axp@region@readname@i{%
  \ifx\axp@region@next\egroup
    \expandafter\axp@region@readname@done
  \else
    \expandafter\axp@region@readname@ii
  \fi}
\def\axp@region@readname@ii#1{%
  \edef\axp@region@name{\axp@region@name#1}\axp@region@readname}
\def\axp@region@readname@done{%
  \afterassignment\axp@region@readname@go\let\axp@region@next= }
\def\axp@region@readname@go{%
  \ifx\axp@region@name\axp@replay@env
    \expandafter\axp@region@go@direct
  \else
    \expandafter\axp@region@go@wrapped
  \fi}
\def\axp@region@go@direct{%
  \expandafter\let\csname axp@region@isregion@\axp@replay@env\endcsname
    \@empty
  \let\end\axp@region@endcmd
  \axp@replay@begin}
\def\axp@region@go@wrapped{%
  \axp@region@wrappedtrue
  \expandafter\let\expandafter\axp@region@savedwrappedend
    \csname end\axp@region@name\endcsname
  \expandafter\def\csname end\axp@region@name\endcsname{%
    \axp@region@savedwrappedend\aftergroup\axp@region@finish}%
  \expandafter\begin\expandafter{\axp@region@name}}
\def\axp@region@endcmd#1{%
  \ifcsname axp@region@isregion@#1\endcsname
    \expandafter\axp@region@endregion
  \else
    \expandafter\axp@region@savedendcmd
  \fi
  {#1}}
\def\axp@region@endregion#1{\axp@replay@end\axp@region@finish}
\def\axp@region@finish{\axp@region@setcatcodes{9}\endinput}
%    \end{macrocode}
% \end{macro}
% \end{macro}
% \begin{macro}{\axp@region@proof}
% \begin{macro}{\axp@region@rp}
% Finally, the begin codes used in the appendix, which read the optional
% argument of the environment from the source. For proofs, the first argument
% is the proof environment to use and the second is |1| if the proof is
% to be titled by |\axp@apxproofstart| (this is decided in the main text,
% and thus passed along). The optional argument is looked for exactly as
% in the main text, without skipping spaces. The end of the
% |\begin{proof}| line was not copied to the |.axp| file, and we
% |\ignorespaces| it, which also prevents the proof environment from
% mistaking a |[| at the start of the body for an optional argument.
%    \begin{macrocode}
\def\axp@region@proof#1#2{%
  \ltx@ifnextchar@nospace[%
    {\axp@region@proof@opt{#1}}{\axp@region@proof@noopt{#1}{#2}}}
\def\axp@region@proof@opt#1[#2]{\begin{#1}[#2]\ignorespaces}
\def\axp@region@proof@noopt#1#2{%
  \ifnum#2=\@ne
    \expandafter\axp@apxproofstart
  \else
    \expandafter\begin
  \fi
  {#1}\ignorespaces}
%    \end{macrocode}
% For |toappendix| environments, we typeset the pending appendix section
% heading, if any, using |\axp@section| which looks at what follows in the
% source, and otherwise skip the end of the |\begin{toappendix}| line, which
% was not copied to the |.axp| file.
%    \begin{macrocode}
\def\axp@region@toappendix{%
  \ifx\axp@region@pendingsection\relax
    \expandafter\ignorespaces
  \else
    \expandafter\axp@region@toappendix@section
  \fi}
\def\axp@region@toappendix@section{%
  \edef\axp@tmp{\expandonce\axp@region@pendingsection}%
  \global\let\axp@region@pendingsection\relax
  \expandafter\axp@section\expandafter{\axp@tmp}%
}
%    \end{macrocode}
% For repeated theorems, the arguments are the theorem name and the value of
% |axp@rpcounter|; the heading is built as in |\axp@newtheoremrep| below.
%    \begin{macrocode}
\def\axp@region@rp#1#2{%
  \@ifnextchar[{\axp@region@rp@i{#1}{#2}}{\axp@region@rp@i{#1}{#2}[]}}
\def\axp@region@rp@i#1#2[#3]{%
  \begin{axp@#1rp}[%
    \axp@refstar{axp@r#2}%
    \@ifnotempty{#3}{%
      \ifdefined\theopargself\else
        \ifdefined\@acmplainnotefont
          \@acmplainnotefont
        \else
          \ifdefined\thm@notefont\the\thm@notefont\fi
        \fi
      \fi
      {} ({#3})%
    }%
  ]%
  \axp@forward@target{axp@fw@r#2}{}%
  \axp@redefinelabels
}
%    \end{macrocode}
% \end{macro}
% \end{macro}
% \subsection{Inline Compilation Mode}
%    \begin{macrocode}
\ifthenelse{\equal{\axp@appendix}{inline}}{
%    \end{macrocode}
% \begin{macro}{\axp@newtheoremrep}
% In inline mode, |\axp@newtheoremrep| uses
% |\axp@newtheoremrep@definetheorem| to define the regular theorem
% environment and creates a repeated theorem
% environment that behaves exactly as the regular theorem environment,
% while calling |\axp@redefinelabels| to make sure that |-axp| variants
% of equation counters are defined.
% \changes{v1.2.0}{2019/02/15}{Fix formatting of theorems without
% notes in some document classes in inline mode}
% With the |synctex| option, the environment is defined with
% |\newenvironment| so that its body is typeset directly from the source
% file (see Section~\ref{sec:synctex-impl}).
%    \begin{macrocode}
  \def\axp@newtheoremrep#1[#2]#3[#4]{%
    \axp@newtheoremrep@definetheorem{#1}{#2}{#3}{#4}%
    \ifbool{axp@synctex}{%
      \newenvironment{#1rep}[1][]{%
        \ifx\relax##1\relax
          \begin{#1}\axp@redefinelabels
        \else
          \begin{#1}[##1]\axp@redefinelabels
        \fi
      }{\end{#1}}%
    }{%
      \NewEnviron{#1rep}[1][]{%
        \ifx\relax##1\relax
          \begin{#1}\axp@redefinelabels\BODY\end{#1}%
        \else
          \begin{#1}[##1]\axp@redefinelabels\BODY\end{#1}%
        \fi
      }%
    }%
  }
%    \end{macrocode}
% \end{macro}
% \begin{environment}{inlineproof}
% \begin{environment}{nestedproof}
% \begin{environment}{appendixproof}
% In inline mode, these environments behave like the regular |proof|
% environment.
%    \begin{macrocode}
  \let\inlineproof\proof
  \let\endinlineproof\endproof
  \let\nestedproof\proof
  \let\endnestedproof\endproof
  \let\appendixproof\proof
  \let\endappendixproof\endproof
%    \end{macrocode}
% \end{environment}
% \end{environment}
% \end{environment}
% \begin{environment}{toappendix}
% \begin{macro}{\noproofinappendix}
% \begin{macro}{\nosectionappendix}
% In inline mode, this environment and these macros are no-ops.
%    \begin{macrocode}
  \newenvironment{toappendix}{}{}
  \let\noproofinappendix\relax
  \let\nosectionappendix\relax
%    \end{macrocode}
% \end{macro}
% \end{macro}
% \end{environment}
%    \begin{macrocode}
}
%    \end{macrocode}
% \subsection{Append, Strip, or Chapter End Compilation Modes}
%    \begin{macrocode}
{
%    \end{macrocode}
% We now deal with the case where \textsf{apxproof} really does something
% useful: either append the appendix material to the document, or strip
% it entirely.
% \bigskip
%    \subsubsection{Auxiliary File for the Appendix}
% \begin{macro}{\axp@proofsfile}
% We open a new auxiliary file, with extension |.axp|, where the appendix
% material will be dumped.
% \changes{v1.1.0}{2018/07/11}{Initialization deferred to \cs{AtBeginDocument}
% for compatibility with \cs{dump}ed precompiled preambles (K. D.
% Bauer)}
%    \begin{macrocode}
  \AtBeginDocument{
    \newwrite\axp@proofsfile
    \ifthenelse{\equal{\axp@appendix}{chapterend}}{%
      \axp@openchapterfile
    }{%
      \immediate\openout\axp@proofsfile=\jobname.axp
    }%
  }
%    \end{macrocode}
% \end{macro}
% \begin{environment}{proof}
% \begin{macro}{\section}
% At the beginning of this file, we make |@| a regular character (since
% it will be used in several places for internal names) and reestablish the original definition of
% the |proof| environment and the |\section| macro.
%    \begin{macrocode}
  \newcommand{\axp@writeproofsfilepreamble}{%
    \immediate\write\axp@proofsfile{%
      \noexpand\makeatletter
      \noexpand\let\noexpand\proof\noexpand\axp@replayproof
      \noexpand\let\noexpand\endproof\noexpand\endaxp@oldproof
      \noexpand\let\noexpand\claimproof\noexpand\axp@oldclaimproof
      \noexpand\let\noexpand\endclaimproof\noexpand\endaxp@oldclaimproof
      \noexpand\let\noexpand\section\noexpand\axp@oldsection
    }%
  }
  \AtBeginDocument{
    \ifthenelse{\equal{\axp@appendix}{chapterend}}{}{%
      \axp@writeproofsfilepreamble
    }%
  }
%    \end{macrocode}
% \end{macro}
% \end{environment}
% \begin{macro}{\axp@unactivateeightbit}
% \changes{v1.1.0}{2019/01/28}{Fix compilation of non-ASCII
% characters with \texttt{\textbackslash usepackage[utf8]\{inputenc\}}}
% We need an auxiliary macro to disable active characters that have the
% high bit set when writing to the |.axp| file. See
% \url{https://tex.stackexchange.com/a/145361/166858}
%    \begin{macrocode}
  \def\axp@unactivateeightbit{%
    \count@=128%
    \loop
      \catcode\count@=12%
      \ifnum\count@<255%
      \advance\count@\@ne
    \repeat}
%    \end{macrocode}
% \end{macro}
% \begin{environment}{axp@VerbatimOut}
% \begin{macro}{\FVB@axp@VerbatimOut}
% \begin{macro}{\FVE@axp@VerbatimOut}
% \changes{v1.1.0}{2018/10/02}{Make \textsf{apxproof} compatible with
% independent use of \textsf{fancyvrb}}
% Using the functionalities of the \textsf{fancyvrb} package, we define a
% custom verbatim environment |axp@VerbatimOut| that writes every line
% to the |\axp@proofsfile|. We also use the previous macro to disable active
% characters with the eighth bit set, and we make sure the catcode of |@|
% is reset for every verbatim environment, in case it is used by the user (e.g.,
% as in the \textsf{xypic} package). Finally, as an additional
% precaution, we reset |\FV@CatCodesHook| that is for example set by the
% |commandchars| or |commentchar| option of |\fvset|.
% \changes{v1.2.1}{2020/01/01}{Fix compatibility with
% \textsf{xypic} package}
% \changes{v1.2.1}{2020/10/15}{Fix compatibility with other uses of \textsf{fancyvrb} that set
% \texttt{\textbackslash FV@CatCodesHook}}
% With the |synctex| option, the environment is still scanned line by line by
% \textsf{fancyvrb} (this is how we skip its content in the main text and
% find its end), but the lines are not written to the |.axp| file, which
% instead receives an |\axp@replaybody| command
% (Section~\ref{sec:synctex-impl}).
%    \begin{macrocode}
  \DefineVerbatimEnvironment{axp@VerbatimOut}{axp@VerbatimOut}{}
  \def\FVB@axp@VerbatimOut{%
    \@bsphack
    \begingroup
      \axp@unactivateeightbit
      \FV@DefineWhiteSpace
      \def\FV@Space{\space}%
      \FV@DefineTabOut
      \ifbool{axp@synctex}{%
        \let\FV@ProcessLine\@gobble
      }{%
        \def\FV@ProcessLine{\immediate\write\axp@proofsfile}%
      }%
      \let\FV@FontScanPrep\relax
      \let\@noligs\relax
      \def\FV@CatCodesHook{}%
      \FV@Scan}
  \def\FVE@axp@VerbatimOut{%
      \immediate\write\axp@proofsfile{\noexpand\makeatletter}%
      \endgroup\@esphack}
%    \end{macrocode}
% \end{macro}
% \end{macro}
% \end{environment}
% \begin{macro}{\axp@markchapterhasappendix}
% No-op by default; overridden in |chapterend| mode to set a flag
% indicating that the current chapter has appendix content.
%    \begin{macrocode}
  \newcommand{\axp@markchapterhasappendix}{}
%    \end{macrocode}
% \end{macro}
% \begin{environment}{toappendix}
% The entire content of this environment is put in appendix,
% starting a new appendix section beforehand if needed. With the
% |synctex| option, we record where the environment starts and, at its
% end, write the replay command; in the appendix, the begin code is
% |\axp@region@toappendix| and there is no end code. As
% |\axp@section| looks at what follows it to decide whether to
% typeset a heading, and the content is now only available once the
% environment is replayed, we have the section command written as
% |\axp@region@section|, which merely stores the heading for
% |\axp@region@toappendix| to handle.
%    \begin{macrocode}
  \newenvironment{toappendix}
    {\axp@markchapterhasappendix
     \ifbool{axp@synctex}{%
       \def\axp@sectioncmd{\axp@region@section}%
       \axp@writesection
       \axp@region@record
     }{%
       \axp@writesection
     }%
     \axp@VerbatimOut}
    {\endaxp@VerbatimOut
     \ifbool{axp@synctex}{%
       \axp@region@write{\noexpand\axp@region@toappendix}{\relax}%
     }{}}
%    \end{macrocode}
% \end{environment}
%    \subsubsection{Definition of New Theorems}
% \begin{macro}{axp@seenreptheorem}
% Used to indicate whether a repeated theorem was just typeset, without
% its proof.
%    \begin{macrocode}
  \newtoggle{axp@seenreptheorem}
%    \end{macrocode}
% \end{macro}
% \begin{macro}{axp@rpcounter}
% Sequentially incremented for every repeated theorem, used to create labels.
%    \begin{macrocode}
  \newcounter{axp@rpcounter}
%    \end{macrocode}
% \end{macro}
% \begin{macro}{axp@equation}
% \begin{macro}{axp@equationx}
% Used to save the value of the |equation| counter, when |repeqn| is set
% to |same|.
%    \begin{macrocode}
  \newcounter{axp@equation}
  \newcounter{axp@equationx}
%    \end{macrocode}
% \end{macro}
% \end{macro}
% \begin{macro}{axp@newtheoremrep}
% With first argument |foobar|, we use
% |\axp@newtheoremrep@definetheorem| to define the regular version of the
% theorem |foobar|. We then patch |\begin{foobar}| so as not to
% expect a proof in the appendix and define an internal theorem |axp@foobarrp| that
% will be used in the appendix to restate the existing theorem.
% In the starred mode, the theorem is not restated in the appendix,
% so we do not define the |axp@foobarrp| environment.
%    \begin{macrocode}
  \def\axp@newtheoremrep#1[#2]#3[#4]{%
    \axp@newtheoremrep@definetheorem{#1}{#2}{#3}{#4}%
    \expandafter\pretocmd\csname #1\endcsname{\noproofinappendix}{}{}%
    \ifcsdef{axp@proofof@#1}{}{%
      \axp@newtheorem*{axp@#1rp}{#3}%
    }%
    \axp@forward@setup{#1}{#2}{#3}{#4}%
%    \end{macrocode}
% We then define a |foobarrep| environment, whose begin code and code
% writing the appendix material are the two macros described below,
% |\axp@newtheoremrep@begin| and |\axp@newtheoremrep@end|. Normally, the
% body of the environment is collected by \textsf{environ}, typeset, and
% stored in a macro for the appendix. With the |synctex| option, the environment
% is instead an ordinary one, so that its body is typeset directly from the
% source file, and the appendix will re-read the body from there
% (Section~\ref{sec:synctex-impl}); we first record where the environment
% starts.
% \changes{v1.5.0}{2026/08/19}{Repeated theorems typeset directly from
% the source with the \texttt{synctex} option}
%    \begin{macrocode}
    \ifbool{axp@synctex}{%
      \newenvironment{#1rep}[1][]{%
        \axp@region@record
        \axp@newtheoremrep@begin{#1}{##1}%
      }{%
        \end{#1}%
        \axp@newtheoremrep@end{#1}{#3}{}%
      }%
    }{%
      \NewEnviron{#1rep}[1][]{%
        \axp@newtheoremrep@begin{#1}{##1}%
        \BODY\end{#1}%
        \global\expandafter\let\csname rplet\roman{axp@rpcounter}%
                               \endcsname
        \BODY
        \axp@newtheoremrep@end{#1}{#3}{##1}%
      }%
    }%
  }
%    \end{macrocode}
% \end{macro}
% \begin{macro}{\axp@newtheoremrep@begin}
% The begin code of a |foobarrep| environment increments the
% |axp@rpcounter| and typesets the regular |foobar| theorem with a label
% derived from the counter, along with a possible custom command to
% identify repeated theorems. We distinguish the case when the theorem
% argument has a note and when it does not.
% \changes{v1.0.6}{2018/05/10}{Better handling of note-free theorems in
% document classes that treat theorems differently when they have an
% empty note}
% We save the equation counter
% before typesetting the theorem environment, to reset it to the same
% value in the repeated environment when |repeqn| is set to |same|.
% \changes{v1.2.2}{2020/11/25}{Fix handling of optional arguments of
% repeated theorems containing optional arguments}
%    \begin{macrocode}
  \def\axp@newtheoremrep@begin#1#2{%
    \ifthenelse{\equal{\axp@repeqn}{same}}{%
      \setcounter{axp@equation}{\value{equation}}%
    }{}%
    \addtocounter{axp@rpcounter}{1}%
    \ifx\relax#2\relax
      \axp@with@forward{#1}{\begin{#1}}\label{axp@r\roman{axp@rpcounter}}%
    \else
      \axp@with@forward{#1}{\begin{#1}[{#2}]}\label{axp@r\roman{axp@rpcounter}}%
    \fi
    \mainbodyrepeatedtheorem
  }
%    \end{macrocode}
% \end{macro}
% \begin{macro}{\axp@newtheoremrep@end}
% Once the theorem has been typeset, with arguments the theorem name, its
% title and its note (the latter is only needed when the body was
% collected), we set the |axp@seenreptheorem| toggle to indicate that we
% are looking for the proof of the theorem.
%    \begin{macrocode}
  \def\axp@newtheoremrep@end#1#2#3{%
    \global\toggletrue{axp@seenreptheorem}%
%    \end{macrocode}
% Possibly after starting a new appendix section if needed, we typeset a
% repeated version of the theorem using the |axp@foobarrp| environment
% and a reference to the previously defined label.
% We use |\axp@redefinelabels| in this environment to avoid multiply
% defined labels. We have to deal in a careful way with theorem notes: we
% want to use a theorem note to display the number of the repeated
% theorem, but theorem notes are usually typeset in a much
% different way (different font, parentheses) than theorem
% headings. In the case of the Springer document
% classes, we use the  |\theopargself| macro to disable parentheses. For
% other document classes, we need to manually patch the |\thmhead|
% command at the right time. We also specially cover the case of
% the ACM document class where |\@acmplainnotefont| is used instead
% of |\thm@notefont|.
% \changes{v1.0.2}{2016/12/13}{Fix missing space between repeated theorem
% counter and theorem note}
% \changes{v1.0.2}{2016/12/13}{Fix display of repeated theorem counter in
% some document classes}
% \changes{v1.0.6}{2018/05/10}{Fix incorrect use of \texttt{\textbackslash noexpand}
% in optional argument of macro environment}
% \changes{v1.2.0}{2019/10/07}{Fix display of theorem notes}
%    \begin{macrocode}
    \axp@markchapterhasappendix
    \axp@writesection%
    \immediate\write\axp@proofsfile{\noexpand\makeatletter}%
    \ifthenelse{\equal{\axp@repeqn}{same}}{%
      \immediate\write\axp@proofsfile{%
        \noexpand\setcounter{axp@equationx}{\value{equation}}%
        \noexpand\setcounter{equation}{\theaxp@equation}%
      }%
    }{}%
    \ifbool{axp@forward@suppress}{%
      \global\def\axp@refstar{\ref*}
    }{%
      \global\def\axp@refstar{\ref}
    }
%    \end{macrocode}
% We write the auto-title state to the |.axp| file so that the following
% deferred proof can title itself via |\appendixproofname|. In starred
% mode, we also emit a forward-link target into the appendix so that
% hyperlinks from the inline theorem number always land somewhere
% defined, even if the user does not provide a proof.
% \changes{v1.4.0}{2026/05/14}{Write auto-title state for
% \cs{appendixproofname} hook}
%    \begin{macrocode}
    \immediate\write\axp@proofsfile{%
      \noexpand\gdef\noexpand\axp@nextproofof@type{#2}%
      \noexpand\gdef\noexpand\axp@nextproofof@label{axp@r\roman{axp@rpcounter}}%
    }%
    \ifcsdef{axp@proofof@#1}{%
      \immediate\write\axp@proofsfile{%
        \noexpand\axp@forward@target{axp@fw@r\roman{axp@rpcounter}}{}%
      }%
    }{%
%    \end{macrocode}
% In the default (unstarred) mode, we emit the repeated version of the
% theorem to the appendix; the forward-link target sits inside the
% restated theorem. In starred mode, we skip this block entirely.
% With the |synctex| option, the restated theorem is produced by
% replaying the source, and the heading is built by |\axp@region@rp|;
% otherwise we write it out here, followed by the stored body.
%    \begin{macrocode}
      \immediate\write\axp@proofsfile{{%
        \ifdefined\theopargself
          \noexpand\theopargself
        \else
          \noexpand\pretocmd{\noexpand\@begintheorem}{%
            \noexpand\patchcmd{\noexpand\thmhead}{\noexpand\@acmplainnotefont}{}{}{}%
            \noexpand\patchcmd{\noexpand\thmhead}{\noexpand\the\noexpand\thm@notefont}{}{}{}%
            \noexpand\patchcmd{\noexpand\thmhead}{(}{}{}{}%
            \noexpand\patchcmd{\noexpand\thmhead}{)}{}{}{}%
          }{}{}
        \fi
        \ifbool{axp@synctex}{%
          \noexpand\axp@replaybody{\axp@region@file}{\axp@region@line}%
            {\the\inputlineno}{\axp@region@env}%
            {\noexpand\axp@region@rp{#1}{\roman{axp@rpcounter}}}%
            {\noexpand\end{axp@#1rp}}{\axp@region@tag}%
        }{%
          \noexpand\begin{axp@#1rp}
            [%
              \noexpand\axp@refstar{axp@r\roman{axp@rpcounter}}%
              \@ifnotempty{#3}{%
                \ifdefined\theopargself
                \else
                  \ifdefined\@acmplainnotefont
                    \noexpand\@acmplainnotefont
                  \else
                    \noexpand\ifdefined\noexpand\thm@notefont
                      \noexpand\the\noexpand\thm@notefont
                    \noexpand\fi
                  \fi
                \fi
                {} (\unexpanded{{#3}})%
              }%
            ]%
            \noexpand\axp@forward@target{axp@fw@r\roman{axp@rpcounter}}{}%
            \noexpand\axp@redefinelabels
            \expandafter\noexpand\csname rplet\roman{axp@rpcounter}%
                                 \endcsname
          \noexpand\end{axp@#1rp}
        }%
      }}%
    }%
    \ifthenelse{\equal{\axp@repeqn}{same}}{%
      \immediate\write\axp@proofsfile{%
        \noexpand\setcounter{equation}{\value{axp@equationx}}%
      }%
    }{}%
  }
%    \end{macrocode}
% \end{macro}
%
%    \subsubsection{Forward-Linking Mechanism}
% \changes{v1.1.0}{2018/07/02}{Added forward-link mechanism (K. D. Bauer)}
%
% When hyperref is loaded, |foobarrep| environments
% in the main text have their number link to their
% repetition in the appendix.
%
%

%
% \begin{macro}{\axp@with@forward}
% In order to make the number of the |foobarrep| theorem a link
% to its repeated version, we temporarily redefine the
% |\thefoobar| command, or, if we inherited the counter
% from a |bazbar| environment, the |\thebazbar| command.
% This seems to be the only robust way to make the number
% a |\hyperlink|, without adding extensive dependence
% on internals of |amsthm|, the builtin |\newtheorem|
% and possibly document-class specific definitions.
%
% In order to allow users to redefine |\thefoobar| without
% breaking this feature, we redefine |\thefoobar| only
% for the duration of the |\begin{foobar}| form,
% resetting it to the old value as soon as possible.
%
% Redefining |\thefoobar| has the side effect of changing |\newlabel| entries
% in the |.aux| file, so we need to to be able to disable
% addition of the hyperlink, which is why we use an
% intermediate |\axp@forward@link|\marg{target}\marg{text} macro,
% We also redefine |\theHfoobar| which is used by |hyperref| but not
% defined if |hyperref| was loaded after |\newtheoremrep| was used
% and |\protect| it to output it verbatim into the |.aux| file.
%
% These hyperlinks are of course disabled in the |strip| compilation
% mode.
%    \begin{macrocode}
   \newcommand{\axp@with@forward}[2]{%
     \ifthenelse{\equal{\axp@appendix}{strip}}{#2}{
       \global\booltrue{axp@forward}%
       \ifcsundef{axp@old@the\csname axp@cn@#1\endcsname}{%
         \csletcs{axp@old@the\csname axp@cn@#1\endcsname}{the\csname axp@cn@#1\endcsname}%
         \csletcs{theH\csname axp@cn@#1\endcsname}{the\csname axp@cn@#1\endcsname}%
         \csdef{the\csname axp@cn@#1\endcsname}{%
           \protect\axp@forward@link{axp@fw@r\roman{axp@rpcounter}}%
             {\csname axp@old@the\csname axp@cn@#1\endcsname\endcsname}%
         }%
       }{}%
       #2%
       \ifcsdef{axp@old@the\csname axp@cn@#1\endcsname}{%
         \csletcs{the\csname axp@cn@#1\endcsname}{axp@old@the\csname axp@cn@#1\endcsname}%
       }{}%
       \global\boolfalse{axp@forward}
     }}%
%    \end{macrocode}
% \end{macro}
%
% \begin{macro}{\axp@forward@link}
% \begin{macro}{axp@forward}
% Dummy macro, for handling the unwanted change of
% the |\newlabel| entry in the |.aux| file caused by
% changing the definition of |\thefoobar|. We also use |\texorpdfstring|
% to detect whether the reference appears within a PDF bookmark, in which
% case we skip the test altogether.
% \changes{v1.2.5}{2025/07/25}{Remove forward linking command from PDF bookmarks}
%
%    \begin{macrocode}
  \newbool{axp@forward}
  \newcommand{\axp@forward@link}[2]{%
    \ifdefined\texorpdfstring\else\newcommand\texorpdfstring[2]{#2}\fi
    \texorpdfstring{%
      \ifboolexpr{ bool {axp@forward} and not bool {axp@forward@suppress} }{%
        \ifcsdef{hyperlink}{%
          \hyperlink{#1}{#2}%
        }{%
          #2%
        }%
      }{%
        #2%
      }%
    }{%
      #2%
    }%
  }%
%    \end{macrocode}
% \end{macro}
% \end{macro}
%
% \begin{macro}{\axp@forward@target}
% Provides the needed |\hypertarget|.
% Intended to be written to the |.axp| file.
%    \begin{macrocode}
  \newcommand{\axp@forward@target}[2]{%
    \ifcsname hypertarget\endcsname
      \hypertarget{#1}{#2}%
    \else
      #2%
    \fi
  }
%    \end{macrocode}
% \end{macro}
%
% \begin{macro}{\axp@forward@setup}
% \changes{v1.2.6}{2025/08/04}{Support for \texttt{aliascnt} counters}
% In order to support counter inheritance with the first
% optional argument of |\newtheoremrep|,
% we need access to the name of the counter.
% For compliance with the behavior of |\@axp@newtheorem|,
% the first optional argument (|#2|) is ignored
% if the second optional argument (|#4|) is given.
% We also check whether |aliascnt| was used to define this counter;
% if so, we use the original name as the counter name to handle
% the way such aliases work.
%    \begin{macrocode}
  \newcommand{\axp@forward@setup}[4]{%
    \csedef{axp@cn@#1}{\ifblank{#4}{\ifblank{#2}{#1}{\ifcsdef{AC@cnt@#1}{#1}{#2}}}{#1}}%
  }
%    \end{macrocode}
% \end{macro}
%
%
%    \subsubsection{Proof Environments}
% \begin{macro}{\noproofinappendix}
%   Utility macro that toggles |axp@seenreptheorem| to false.
%    \begin{macrocode}
  \newcommand\noproofinappendix{%
    \global\togglefalse{axp@seenreptheorem}%
  }
%    \end{macrocode}
% \end{macro}
% \begin{macro}{\axp@apxproofstart}
% \changes{v1.4.0}{2026/05/14}{Helper that consults \cs{appendixproofname}
% at appendix typeset time to auto-title deferred proofs}
% Helper invoked from the |.axp| file at appendix typeset time to open
% a deferred proof. Reads the |\axp@nextproofof@*| state (set by the
% preceding |theoremrep|) and consults |\appendixproofname| to decide
% whether to pass an optional title argument.
%    \begin{macrocode}
  \gdef\axp@nextproofof@type{}
  \gdef\axp@nextproofof@label{}
%    \end{macrocode}
% \begin{macro}{\apxproofhook}
% \changes{v1.4.1}{2026/05/31}{Hook fired when a deferred proof is
% typeset, or for an inline proof whose title names its result through a
% \cs{ref}; used by the \textsf{proofgraph} package}
% A hook fired, inside the proof environment, each time a deferred proof is
% typeset in the appendix (and also for an inline proof whose title names its
% result through a |\ref|, see |\axp@inlineproofhook| below). Its argument is
% the label of the theorem the proof establishes: the automatic |axp@r|\meta{n}
% label of the preceding |theoremrep| for a deferred proof, or the target of
% the title's |\ref| for a titled inline proof. A companion package (such as
% \textsf{proofgraph}) can thereby attribute, to the right result, the
% references the proof makes (which are only executed at typeset time). It is a
% no-op by default.
%    \begin{macrocode}
  \providecommand\apxproofhook[1]{}
  \def\axp@apxproofstart#1{%
    \protected@xdef\axp@apxproofstart@title{%
      \appendixproofname{\axp@nextproofof@type}%
                        {\noexpand\ref{\axp@nextproofof@label}}%
    }%
    \ifx\axp@apxproofstart@title\@empty
      \begin{#1}%
    \else
      \expandafter\axp@apxproofstart@withtitle
        \expandafter{\axp@apxproofstart@title}{#1}%
    \fi
    \apxproofhook{\axp@nextproofof@label}%
  }
  \def\axp@apxproofstart@withtitle#1#2{\begin{#2}[#1]}
%    \end{macrocode}
% \end{macro}
% \end{macro}
% \begin{macro}{\axp@inlineproofhook}
% A proof that is \emph{not} deferred (no preceding |theoremrep|, so it is
% typeset inline) never goes through |\axp@apxproofstart|, hence never fires
% |\apxproofhook|. When such a proof carries an explicit title that names
% its result through a |\ref| (the idiom |\begin{proof}[of Theorem~\ref{thm:x}]|),
% we fire the hook here, inside the proof, with that target, so a companion
% package can attribute the proof to it. We only fire when a cross-reference is
% found in the title (the result is then unambiguous). The title is scanned for
% the first |\ref| (or |\cref|, |\Cref|, |\autoref|, |\vref|, |\zcref|) by pure
% token matching, never typesetting it: a title typically contains an
% active~|~| and may carry other fragile material, which an auxiliary typeset
% pass would perturb or, with |\protect| neutralised, send into an infinite
% loop.
% \changes{v1.5.0}{2026/08/03}{Recognize \cs{zcref} of \textsf{zref-clever} in
% the title of an inline proof}
% \textsf{zref-clever}'s |\zcref| is the only one of these that is not simply
% \meta{cs}\marg{label}: it is |\zcref*[|\meta{options}|]|\marg{labels}, whose
% star and options are to be skipped and whose mandatory argument is a
% comma-separated list, of which we keep the first label. Each grabber is
% handed the whole remainder of the title and swallows it up to
% |\axp@refstop|, so that nothing of it is left to be typeset.
%
% The matching is |\in@|, hence \TeX{} delimited-argument matching, which only
% sees tokens at brace level~0. A cross-reference buried in a brace group is
% therefore not found, and the hook does not fire for such a title. This is the
% case when the cross-reference takes an optional argument of its own, since
% \LaTeX{} then requires it to be braced within the optional title, as in
% |\begin{proof}[Proof of {\zcref[noname]{thm:x}}]|. Nothing else is affected:
% the title is typeset as given, and |\apxproofhook| is a no-op unless a
% companion package has defined it.
%    \begin{macrocode}
  \newif\ifaxp@inlineproofof
  \long\def\axp@inlineproofhook#1{%
    \global\axp@inlineproofoffalse
    \axp@tryrefcmd\ref{#1}%
    \ifaxp@inlineproofof\else\axp@tryrefcmd\cref{#1}\fi
    \ifaxp@inlineproofof\else\axp@tryrefcmd\Cref{#1}\fi
    \ifaxp@inlineproofof\else\axp@tryrefcmd\autoref{#1}\fi
    \ifaxp@inlineproofof\else\axp@tryrefcmd\vref{#1}\fi
    \ifaxp@inlineproofof\else
      \axp@tryrefcmd@\zcref\axp@grabzcarg{#1}%
    \fi
    \ifaxp@inlineproofof\apxproofhook{\axp@inlineproofof@label}\fi
  }
  \let\axp@refmark\relax
  \long\def\axp@tryrefcmd#1#2{\axp@tryrefcmd@#1\axp@grabarg{#2}}
  \long\def\axp@tryrefcmd@#1#2#3{%
    \in@{#1}{#3}%
    \ifin@
      \long\def\axp@grabref##1#1##2\axp@refstop{#2##2\axp@refstop}%
      \axp@grabref#3\axp@refmark\axp@refstop
    \fi
  }
  \long\def\axp@grabarg#1#2\axp@refstop{%
    \global\axp@inlineproofoftrue\xdef\axp@inlineproofof@label{#1}}
  \def\axp@grabzcarg{\@ifstar\axp@grabzcarg@i\axp@grabzcarg@i}
  \def\axp@grabzcarg@i{\@ifnextchar[\axp@grabzcarg@opt\axp@grabzcarg@ii}%]
  \def\axp@grabzcarg@opt[#1]{\axp@grabzcarg@ii}
  \long\def\axp@grabzcarg@ii#1{%
    \axp@grabzcarg@iii#1,\@nil
    \axp@gobbletorefstop}
  \long\def\axp@grabzcarg@iii#1,#2\@nil{%
    \global\axp@inlineproofoftrue\xdef\axp@inlineproofof@label{#1}}
  \long\def\axp@gobbletorefstop#1\axp@refstop{}
%    \end{macrocode}
% Proofs placed by hand in a |toappendix| environment are written verbatim to
% the |.axp| file and replayed there with |\proof| let to |\axp@oldproof| (see
% |\axp@writeproofsfilepreamble|). Such proofs therefore never reach
% |\axp@@proof|. So that a titled one still fires |\apxproofhook|, we let, at
% replay time, |\proof| to |\axp@replayproof|, which reproduces the inline
% behaviour: open the |amsthm| proof, then fire the hook for the title's |\ref|.
%    \begin{macrocode}
  \def\axp@replayproof{\@ifnextchar[{\axp@replayproof@opt}{\axp@oldproof}}
  \def\axp@replayproof@opt[#1]{\axp@oldproof[#1]\axp@inlineproofhook{#1}}
%    \end{macrocode}
% \end{macro}
% \begin{environment}{appendixproof}
% \changes{v1.2.4}{2021/11/03}{Support optional arguments in proofs}
% We dump the content of this in appendix, within an original |proof|
% environment, possibly after creating a new appendix section.
% We support
% optional arguments in proofs, but we need to be careful not to gobble
% any newline character, that would make it impossible to process the
% first line with \textsf{fancyvrb}.
% \changes{v1.4.0}{2026/05/14}{Route deferred proofs through
% \cs{axp@apxproofstart} to support \cs{appendixproofname}}
% When the proof is auto-routed from a |theoremrep| (the toggle
% |axp@seenreptheorem| is true at this point and no explicit optional
% argument was given), we delegate the proof opening to
% |\axp@apxproofstart| so that |\appendixproofname| is consulted at
% appendix typeset time. A user-supplied optional argument or a manual
% |\appendixproof| call (toggle false) bypasses this and uses
% the standard ``Proof.'' heading.
% With the |synctex| option, nothing is written when the proof starts:
% we record its position and whether it is to be titled by
% |\axp@apxproofstart|, and the replay command written at the end
% (by |\axp@appendixproof@write|, with the name of the proof environment
% to use in the appendix) carries this information to |\axp@region@proof|,
% which reads the optional argument, if any, from the source.
%    \begin{macrocode}
  \def\appendixproof{\catcode`\^^M=\active\@ifnextchar[{\catcode`\^^M=5\@@appendixproof[axp@oldproof]}{\catcode`\^^M=5\@appendixproof[axp@oldproof]}}
  \def\@appendixproof[#1]%
    {%
      \axp@markchapterhasappendix
      \axp@writesection
      \iftoggle{axp@seenreptheorem}{%
        \ifbool{axp@synctex}{%
          \axp@region@record
          \def\axp@region@flag{1}%
        }{%
          \immediate\write\axp@proofsfile{%
            \noexpand\makeatletter\noexpand\axp@apxproofstart{#1}\noexpand\makeatother%
          }%
        }%
      }{%
        \ifbool{axp@synctex}{%
          \axp@region@record
          \def\axp@region@flag{0}%
        }{%
          \immediate\write\axp@proofsfile{%
            \noexpand\makeatletter\noexpand\begin{#1}\noexpand\makeatother%
          }%
        }%
      }%
      \axp@VerbatimOut
    }
  \def\@@appendixproof[#1][#2]%
    {%
      \axp@markchapterhasappendix
      \axp@writesection
      \ifbool{axp@synctex}{%
        \axp@region@record
        \def\axp@region@flag{0}%
      }{%
        \immediate\write\axp@proofsfile{%
          \noexpand\makeatletter\noexpand\begin{#1}[\unexpanded{#2}]\noexpand\makeatother%
        }%
      }%
      \axp@VerbatimOut
    }
  \def\axp@appendixproof@write#1{%
    \ifbool{axp@synctex}{%
      \axp@region@write
        {\noexpand\axp@region@proof{#1}{\axp@region@flag}}%
        {\noexpand\end{#1}}%
    }{%
      \immediate\write\axp@proofsfile{%
        \noexpand\end{#1}%
      }%
    }%
  }
  \def\endappendixproof
    {%
      \endaxp@VerbatimOut
      \axp@appendixproof@write{axp@oldproof}%
      \noproofinappendix
    }
%    \end{macrocode}
% \end{environment}
% \begin{environment}{proof}
% This environment either puts the proof in appendix, if we are after a repeated
% theorem without its proof, or inlines it otherwise. We support
% optional arguments in proofs, but we need to be careful not to gobble
% any newline character, that would make it impossible to process the
% first line with \textsf{fancyvrb}.
%    \begin{macrocode}
  \def\proof{\catcode`\^^M=\active\ltx@ifnextchar@nospace[{\catcode`\^^M=5\axp@@proof}{\catcode`\^^M=5\axp@proof}}
  \def\axp@proof
    {%
      \iftoggle{axp@seenreptheorem}{%
        \appendixproof
      }{%
        \axp@oldproof
      }%
    }
  \def\axp@@proof[#1]%
    {%
      \iftoggle{axp@seenreptheorem}{%
        \appendixproof[#1]
      }{%
        \axp@oldproof[#1]%
        \axp@inlineproofhook{#1}%
      }%
    }
  \def\endproof
    {%
      \iftoggle{axp@seenreptheorem}{%
        \endappendixproof
      }{%
        \endaxp@oldproof
      }%
    }
%    \end{macrocode}
% \end{environment}
% \begin{environment}{claimproof}
% \changes{v1.2.4}{2021/11/03}{Support for \texttt{claimproof} environment
% from \textsf{lipics}}
% \changes{v1.2.5}{2024/01/26}{Support for locally defined \texttt{claimproof} environment}
% If the |claimproof| environment exists (\textsf{lipics} document
% class or user-defined), we also redefine it, in the same way.
%    \begin{macrocode}
  \AtBeginDocument{
    \ifdefined\claimproof
    \let\axp@oldclaimproof\claimproof
    \let\endaxp@oldclaimproof\endclaimproof
    \def\claimproof{\catcode`\^^M=\active\ltx@ifnextchar@nospace[{\catcode`\^^M=5\axp@@claimproof}{\catcode`\^^M=5\axp@claimproof}}
    \def\appendixclaimproof{\catcode`\^^M=\active\@ifnextchar[{\catcode`\^^M=5\@@appendixproof[axp@oldclaimproof]}{\catcode`\^^M=5\expandafter\@appendixproof[axp@oldclaimproof]}}
    \def\axp@claimproof
      {%
        \iftoggle{axp@seenreptheorem}{%
          \appendixclaimproof
        }{%
          \axp@oldclaimproof
        }%
      }
    \def\axp@@claimproof[#1]%
      {%
        \iftoggle{axp@seenreptheorem}{%
          \appendixclaimproof[#1]%
        }{%
          \axp@oldclaimproof[#1]%
        }%
      }
    \def\endclaimproof
      {%
        \iftoggle{axp@seenreptheorem}{%
          \endappendixclaimproof
        }{%
          \endaxp@oldclaimproof
        }%
      }
    \def\endappendixclaimproof
    {%
      \endaxp@VerbatimOut
      \axp@appendixproof@write{axp@oldclaimproof}%
      \noproofinappendix
    }
    \fi
  }
%    \end{macrocode}
% \end{environment}

% \begin{environment}{inlineproof}
% \begin{environment}{nestedproof}
% These two environments are synonyms for the original |proof|
% environment.
%    \begin{macrocode}
  \let\inlineproof\axp@oldproof
  \let\endinlineproof\endaxp@oldproof
  \let\nestedproof\axp@oldproof
  \let\endnestedproof\endaxp@oldproof
%    \end{macrocode}
% \end{environment}
% \end{environment}
%    \subsubsection{Section Management}
% \begin{macro}{axp@seccounter}
% Sequentially incremented for every section, used to create labels.
%    \begin{macrocode}
  \newcounter{axp@seccounter}
%    \end{macrocode}
% \end{macro}
% \begin{macro}{\axp@sectitle}
% Saves the title of the last encountered section.
%    \begin{macrocode}
  \def\axp@sectitle{}
%    \end{macrocode}
% \end{macro}
% \begin{macro}{\axp@section}
% \changes{v1.2.2}{2021/06/23}{Detect a section within an included
% file to avoid produced useless sections}
% \changes{v1.3.0}{2026/04/09}{Skip blank lines before detecting
% a section command}
% \changes{v1.3.0}{2026/04/09}{Detect a section within an
% \cs{include}'d file}
% \begin{macro}{\axp@@sectionskippar}
% \begin{macro}{\axp@@sectiontestinput}
% \begin{macro}{\axp@@sectiontestinclude}
% \begin{macro}{\axp@@sectionfinishinclude}
% \begin{macro}{\axp@@sectiontestsection}
% This command |\axp@section| behaves similarly to |\axp@oldsection|,
% except that it first tests whether a |\section| follows, and if so,
% does not produce anything. This is useful to avoid producing empty
% sections in the appendix. Using the \textsf{catchfile} package, we
% also check whether a |\section| is within an |\input| or |\include|
% that immediately follows. We also skip any |\par| tokens (produced by
% blank lines) before performing the check.
%    \begin{macrocode}
  \def\axp@section#1{%
    \@ifnextchar\par
      {\axp@@sectionskippar{#1}}%
      {\@ifnextchar\input
        {\axp@@sectiontestinput{#1}}%
        {\@ifnextchar\include
          {\axp@@sectiontestinclude{#1}}%
          {\axp@@sectiontestsection{#1}}}}%
  }
  \def\axp@@sectionskippar#1\par{\axp@section{#1}}
  \def\axp@@sectiontestinput#1\input#2{%
    \CatchFileDef{\axp@tmp}{#2}{}%
    \def\axp@tmptmp{\axp@@sectiontestsection{#1}}%
    \expandafter\axp@tmptmp\axp@tmp%
  }
  \def\axp@@sectiontestinclude#1\include#2{%
    \set@curr@file{#2}%
    \edef\axp@tmp{\@strip@tex@ext\@curr@file.tex}%
    \expandafter\CatchFileDef\expandafter\axp@tmp\expandafter{\axp@tmp}{}%
    \def\axp@tmptmp{\axp@@sectionfinishinclude{#1}{#2}}%
    \expandafter\axp@tmptmp\axp@tmp\@nil
  }
  \long\def\axp@@sectionfinishinclude#1#2#3#4\@nil{%
    \ifx#3\section
      \makeatother
    \else
      \axp@oldsection{#1}\makeatother
    \fi
    \include{#2}%
  }
  \def\axp@@sectiontestsection#1{\@ifnextchar\section{\makeatother}{\axp@oldsection{#1}\makeatother}}
%    \end{macrocode}
% \end{macro}
% \end{macro}
% \end{macro}
% \end{macro}
% \end{macro}
% \end{macro}
% \begin{macro}{\axp@oldsection}
% \begin{macro}{\section}
% \changes{v1.1.0}{2018/07/25}{Rewrote definition of \cs{section}
% to enable optional argument. See \#23. (K. D. Bauer)}
% \changes{v1.1.0}{2018/07/25}{Fix handling of fragile macros within
% section headings. See \#22.}
% \begin{macro}{\@section}
% \begin{macro}{\@@section}
% We redefine the |\section| command to create a label based on
% |axp@seccounter| and to store its title in |\axp@sectitle|.
% In order to support starred and unstarred versions, as well
% as the optional short-title argument, the intermediate macros
% \cs{@section} and \cs{@@section} are needed.
%    \begin{macrocode}
  \let\axp@oldsection\section
  \def\section{\@ifstar\@section\@@section}
  \newcommand{\@section}[2][\relax]{\axp@@@section{*}{#1}{#2}}%
  \newcommand{\@@section}[2][\relax]{\axp@@@section{}{#1}{#2}}%
  \newcommand{\axp@@@section}[3]{%
    \global\def\axp@sectitle{#3}%
    \ifx\relax#2\relax
      \axp@oldsection#1{#3}%
    \else
      \axp@oldsection#1[{#2}]{#3}%
    \fi
    \addtocounter{axp@seccounter}{1}%
    \label{axp@s\roman{axp@seccounter}}%
  }
%    \end{macrocode}
% \end{macro}
% \end{macro}
% \end{macro}
% \end{macro}
% \begin{macro}{\nosectionappendix}
% We remove the current section title, to indicate no section should be
% created in the appendix.
%    \begin{macrocode}
  \newcommand{\nosectionappendix}{
    \global\def\axp@sectitle{}%
  }
%    \end{macrocode}
% \end{macro}
% \begin{macro}{\axp@writesection}
% If |\axp@sectitle| is not empty, we create
% a new section in the appendix, referring to the main text section.
%
% Here, we wrap |\ref{axp@si}| into |\axp@protectref@i|, in order
% to protect the label name from wrongly being converted to uppercase,
% e.g., in \textsf{fancyhdr} with |\pagestyle{fancy}|.
%
% This macro is defined both in the |.aux| file (in order to ensure availability
% when typesetting the |\tableofcontents|), and immediately before typesetting
% the appendix section (to ensure availability in the |\section| command).
%
% \changes{v1.0.6}{2018/05/10}{Fix extraneous space after section number
% in appendix titles}
% \changes{v1.1.0}{2018/07/09}{Make \cs{axp@tmp} wrapper more robust.
% Resolves issues from use of section title in \textsf{fancyhdr},
% and in \cs{tableofcontents} (K. D. Bauer).}
%
% The command written is normally |\axp@section|; the |toappendix|
% environment substitutes |\axp@region@section| for it with the |synctex|
% option, so that the heading is decided when the material is replayed
% (Section~\ref{sec:synctex-impl}).
%    \begin{macrocode}
  \def\axp@sectioncmd{\axp@section}
  \newcommand\axp@writesection{%
    \ifx\axp@sectitle\@empty
    \else
      \edef\axp@tmp{%
        \noexpand\global\noexpand\def
        \expandonce{\csname axp@protectref@\roman{axp@seccounter}\endcsname}{%
          \noexpand\ref{axp@s\roman{axp@seccounter}}%
        }%
      }%
      \immediate\write\@auxout{\expandonce\axp@tmp}
      \immediate\write\axp@proofsfile{%
        \expandonce\axp@tmp^^J%
        \expandafter\noexpand\axp@sectioncmd{%
          \noexpand\appendixsectionformat{%
            \protect
            \expandonce{\csname axp@protectref@\roman{axp@seccounter}\endcsname}%
          }{\expandonce\axp@sectitle}%
        }%
      }%
      \nosectionappendix
    \fi
  }
%    \end{macrocode}
% \end{macro}
% Finally, in a somewhat ad hoc manner, we disable the whole section management
% for \cs{tableofcontents}, which may be typeset using a section heading,
% but for which automatic section management does not make sense.
%
% \begin{macro}{\axp@oldtableofcontents}
% \begin{macro}{\tableofcontents}
% \changes{v1.1.0}{2018/07/09}{Disable section management for table
% of contents}
%    \begin{macrocode}
  \let\axp@oldtableofcontents\tableofcontents
  \def\tableofcontents{{\let\section\axp@oldsection\axp@oldtableofcontents}}
%    \end{macrocode}
% \end{macro}
% \end{macro}
%    \subsubsection{Append Compilation Mode}
%    \begin{macrocode}
  \ifthenelse{\equal{\axp@appendix}{append}}{
%    \end{macrocode}
% \begin{macro}{\axp@oldbibliography}
% \begin{macro}{\bibliography}
% In |bibengine=bibtex| mode, when the |bibliography| option is set to
% |separate|, we need to set the appendix bibliography source to be
% the same as that of the main text, thanks to \textsf{bibunits}'s
% |\defaultbibliography| macro. In |bibengine=biblatex| mode, no such
% rebinding is needed: \textsf{biblatex} resources declared with
% |\addbibresource| are visible to all |refsection|s globally.
% \changes{v1.4.0}{2026/05/14}{Only rebind for bibunits engine.
%   See \#29.}
%    \begin{macrocode}
    \ifthenelse{\equal{\axp@bibliography}{separate}
                \AND\equal{\axp@bibengine}{bibtex}}{
      \let\axp@oldbibliography\bibliography
      \renewcommand\bibliography[1]{%
        \defaultbibliography{#1}%
        \axp@oldbibliography{#1}%
      }
    }{}
%    \end{macrocode}
% \end{macro}
% \end{macro}
% After the end of the main text, we add the appendix (after the command
% |\appendixprelim| is issued)
% within a |bibunit| environment so as to typeset
% a separate bibliography for the appendix (unless the |bibliography|
% option is set to |common|). This is done using |\pretocmd| on
% |\@enddocumenthook| instead of |\AtEndDocument| because we want the
% code to be run before any code in the |\@enddocumenthook| that has
% been set in the document class, as in the \textsf{amsart} document
% class.
% \changes{v1.2.2}{2021/03/23}{Compatibility with AMS document classes: do not
% use \texttt{\textbackslash AtEndDocument}}
% There is an extra test to
% ensure an empty bibliography environment is not produced.
% The name of the bibliography is changed to |\appendixrefname|; in most
% document classes, it is called |\refname| but it is occasionally
% (|scrartcl|, |scrreprt|) called |\bibname|. An ad-hoc test is added
% to fix a conflict with the \textsf{natbib} package which redefines
% |bibcite| at the end of the document.
% \changes{v1.0.6}{2018/05/10}{Deal with document classes where the
% bibliography is called \texttt{\textbackslash bibname}}
% \changes{v1.2.1}{2020/02/27}{Ad hoc fix for \textsf{natbib}
% package conflict}
% \changes{v1.2.3}{2021/08/03}{More robust redefinition of
% \texttt{thebibliography} environment, for compatibility with
% \textsf{tocbibind}}
%    \begin{macrocode}
    \pretocmd{\@enddocumenthook}{%
      \ifdefined\NAT@testdef
        \renewcommand\bibcite[2]{%
          \global\@namedef{b@#1\@extra@binfo}{#2}%
        }
      \fi
      \appendixprelim
      \appendix
      \ifthenelse{\equal{\axp@bibliography}{separate}}{
        \ifthenelse{\equal{\axp@bibengine}{bibtex}}{
          \begin{bibunit}[\appendixbibliographystyle]
        }{
          \begin{refsection}
        }
      }{}
        \immediate\closeout\axp@proofsfile
        \input{\jobname.axp}
      \ifthenelse{\equal{\axp@bibliography}{separate}}{
        \ifthenelse{\equal{\axp@bibengine}{bibtex}}{
          \ifdefined\refname
            \renewcommand{\refname}{\appendixrefname}
          \else\ifdefined\bibname
            \renewcommand{\bibname}{\appendixrefname}
          \fi\fi
          \let\axp@oldthebibliography\thebibliography
          \let\endaxp@oldthebibliography\endthebibliography
          \renewenvironment{thebibliography}[1]{%
            \def\axp@tmp{#1}%
            \ifx\axp@tmp\empty
              \gdef\axp@noappendixbibliography1\relax
            \else
              \begin{axp@oldthebibliography}{#1}%
            \fi
          }{%
            \ifdefined\axp@noappendixbibliography\relax\else\end{axp@oldthebibliography}%
          \fi}
          \appendixbibliographyprelim
          \putbib
          \end{bibunit}
        }{
          \appendixbibliographyprelim
          \printbibliography[title=\appendixrefname]
          \end{refsection}
        }
        \ifdefined\NAT@testdef
          \let\bibcite\NAT@testdef
        \fi
      }{}
    }{}{}
%    \end{macrocode}
%    \begin{macrocode}
  }{}
%    \end{macrocode}
%    \subsubsection{Chapter End Compilation Mode}
% \changes{v1.3.0}{2026/04/14}{Added \texttt{chapterend} appendix mode}
% \changes{v1.4.0}{2026/04/29}{Flush \texttt{chapterend} appendices at
%   \cs{part}, \cs{frontmatter}, \cs{mainmatter}, and \cs{backmatter}
%   boundaries so the last chapter of a part/matter is not emitted under
%   the next one}
% \changes{v1.4.0}{2026/04/29}{Public \cs{flushchapterappendix} macro
%   for manually emitting the current chapter's accumulated appendix in
%   \texttt{chapterend} mode}
% In |chapterend| mode, appendix material for each chapter is gathered
% in a per-chapter auxiliary file and emitted immediately after that
% chapter's content (and bibliography). This requires the
% \textsf{appendix} package for its |subappendices| environment.
%    \begin{macrocode}
  \ifthenelse{\equal{\axp@appendix}{chapterend}}{
%    \end{macrocode}
%    \begin{macrocode}
    \RequirePackage{appendix}
%    \end{macrocode}
% \begin{macro}{\chapterappendixprelim}
% \begin{macro}{\axp@beginchapterappendix}
% \begin{macro}{\axp@endchapterappendix}
% User-customizable hooks surrounding each chapter's appendix block.
% By default they open and close a |subappendices| environment.
%    \begin{macrocode}
    \newcommand{\chapterappendixprelim}{\clearpage}
    \newcommand{\axp@beginchapterappendix}{%
      \chapterappendixprelim
      \begin{subappendices}%
    }
    \newcommand{\axp@endchapterappendix}{%
      \end{subappendices}%
    }
%    \end{macrocode}
% \end{macro}
% \end{macro}
% \end{macro}
% \begin{macro}{axp@chaptercounter}
% \begin{macro}{axp@chapterhasappendix}
% \begin{macro}{\axp@currentaxpfile}
% Counter for unique per-chapter file names, a boolean tracking whether
% the current chapter has any appendix content, and the current file name.
%    \begin{macrocode}
    \newcounter{axp@chaptercounter}
    \newbool{axp@chapterhasappendix}
    \def\axp@currentaxpfile{}
%    \end{macrocode}
% \end{macro}
% \end{macro}
% \end{macro}
% \begin{macro}{\axp@markchapterhasappendix}
% Override the no-op to set the boolean when content is written.
%    \begin{macrocode}
    \renewcommand{\axp@markchapterhasappendix}{%
      \global\booltrue{axp@chapterhasappendix}%
    }
%    \end{macrocode}
% \end{macro}
% \begin{macro}{\axp@openchapterfile}
% Open a fresh auxiliary file for the current chapter.
% We write |\begingroup| first so that all the preamble |\let|
% assignments (and any catcode changes from |\makeatletter|) are
% scoped to this chapter's appendix and do not leak into the rest of
% the document when the file is |\input|'d. We also reset the
% section-title tracker and the repeated-theorem toggle so that state
% from the previous chapter cannot bleed into the new file.
%    \begin{macrocode}
    \newcommand{\axp@openchapterfile}{%
      \edef\axp@currentaxpfile{\jobname-\arabic{axp@chaptercounter}.axp}%
      \immediate\openout\axp@proofsfile=\axp@currentaxpfile
      \immediate\write\axp@proofsfile{\noexpand\begingroup}%
      \axp@writeproofsfilepreamble
      \global\boolfalse{axp@chapterhasappendix}%
      \global\def\axp@sectitle{}%
      \global\togglefalse{axp@seenreptheorem}%
    }
%    \end{macrocode}
% \end{macro}
% \begin{macro}{\axp@flushchapterappendix}
% Close the current chapter's auxiliary file (appending |\endgroup|
% to close the group opened in |\axp@openchapterfile|) and, if it
% contains any content, emit it inside a |subappendices| block.
%    \begin{macrocode}
    \newcommand{\axp@flushchapterappendix}{%
      \immediate\write\axp@proofsfile{\noexpand\endgroup}%
      \immediate\closeout\axp@proofsfile
      \ifbool{axp@chapterhasappendix}{%
        \axp@beginchapterappendix
        \input{\axp@currentaxpfile}%
        \axp@endchapterappendix
      }{}%
    }
%    \end{macrocode}
% \end{macro}
% \begin{macro}{\axp@chapterboundary}
% \begin{macro}{\flushchapterappendix}
% |\axp@chapterboundary| flushes the previous chapter's auxiliary file,
% increments the counter, and opens the next one. It is used as a hook
% for |\chapter|, |\part|, and the |book|-class matter-switching
% commands |\frontmatter|, |\mainmatter|, |\backmatter|, so that
% per-chapter appendices are emitted before any of these structural
% transitions (otherwise the appendix of the last chapter of a
% part/matter would end up under the next part/matter).
% |\flushchapterappendix| is a public alias for the same operation,
% which can be invoked manually wherever an appendix flush boundary is
% needed (e.g., before a bibliography that does not produce its own
% chapter heading).
%    \begin{macrocode}
    \newcommand{\axp@chapterboundary}{%
      \axp@flushchapterappendix
      \stepcounter{axp@chaptercounter}%
      \axp@openchapterfile
    }
    \let\flushchapterappendix\axp@chapterboundary
%    \end{macrocode}
% \end{macro}
% \end{macro}
% \begin{macro}{\axp@hookboundary}
% Helper to install |\axp@chapterboundary| as a |\pretocmd| hook on a
% structural command (|\chapter|, |\part|, |\frontmatter|, etc.),
% issuing a package warning if the hook fails.
%    \begin{macrocode}
    \newcommand{\axp@hookboundary}[1]{%
      \pretocmd{#1}{\axp@chapterboundary}{}{%
        \PackageWarning{apxproof}{%
          Could not hook into \string#1;^^J%
          chapter appendices may not be emitted correctly}}%
    }
%    \end{macrocode}
% \end{macro}
% At the start of each |\chapter|, |\part|, or matter-switching
% command, run the boundary hook. At |\end{document}|, flush the
% last chapter's file.
%    \begin{macrocode}
    \AtBeginDocument{%
      \ifdef{\chapter}{%
        \axp@hookboundary{\chapter}%
      }{%
        \PackageWarning{apxproof}{%
          appendix=chapterend used but \string\chapter\space
          is not defined}%
      }%
      \ifdef{\part}{\axp@hookboundary{\part}}{}%
      \ifdef{\frontmatter}{\axp@hookboundary{\frontmatter}}{}%
      \ifdef{\mainmatter}{\axp@hookboundary{\mainmatter}}{}%
      \ifdef{\backmatter}{\axp@hookboundary{\backmatter}}{}%
    }
    \pretocmd{\@enddocumenthook}{%
      \axp@flushchapterappendix
    }{}{}
%    \end{macrocode}
%    \begin{macrocode}
  }{}
%    \end{macrocode}
% \subsubsection{Class-Specific Behavior}
% We conclude with some class-specific behavior.
% \paragraph{ACM Document Classes (old versions, till 2017)}
% \changes{v1.0.4}{2017/03/08}{More faithful theorem style for ACM templates}
% \changes{v1.0.4}{2017/03/08}{More robust coherent styling of proof sketches}
%    \begin{macrocode}
  \ifdefined\@acmtitlebox
%    \end{macrocode}
% We first redefine the |proofsketch| environment, which is used
% differently in the base class.
%    \begin{macrocode}
  \renewenvironment{proofsketch}{\begin{axp@oldproof}[sketch]}{\end{axp@oldproof}}
%    \end{macrocode}
% We adjust the styling of theorems for the needs of \textsf{apxproof}.
%    \begin{macrocode}
  \newtheoremstyle{mystyle}
    {6pt}
    {6pt}
    {\itshape}
    {10pt}
    {\scshape}
    {.}
    {.5em}
    {}
  \theoremstyle{mystyle}
%    \end{macrocode}
% \begin{macro}{\thebibliography}
% \begin{macro}{\refname}
% \begin{macro}{\appendixrefname}
% The section title of the bibliography is in uppercase in these document
% classes. In addition, the |\thebibliography| macro hard-codes twice
% the section title, so we un-hardcode it so that it can be modified
% in the appendix.
%    \begin{macrocode}
    \patchcmd{\thebibliography}{References}{\protect\refname}{}{}
    \patchcmd{\thebibliography}{References}{\protect\refname}{}{}
    \newcommand{\refname}{REFERENCES}
    \renewcommand{\appendixrefname}{REFERENCES FOR THE APPENDIX}
%    \end{macrocode}
% \end{macro}
% \end{macro}
% \end{macro}
%
%    \begin{macrocode}
  \fi
%    \end{macrocode}
% \changes{v1.0.6}{2018/05/10}{Support of new ACM document class
% (\texttt{acmart.cls})}
% \paragraph{\textsf{lipcs}}
%    \begin{macrocode}
   \ifdefined\lipics@opterrshort
%    \end{macrocode}
% \begin{macro}{\appendixbibliographyprelim}
% \changes{v1.0.3}{2017/02/06}{Support for lipics-v2016}
% The default bibliography in the \textsf{lipics} document class
% formatting is not compatible with the
% \textsf{alpha} bibliography style. We fix this here.
%    \begin{macrocode}
    \renewcommand{\appendixbibliographyprelim}{%
      \global\let\@oldbiblabel\@biblabel
      \def\@biblabel{\hspace*{-2em}\small\@oldbiblabel}%
    }
%    \end{macrocode}
% \end{macro}
%    \begin{macrocode}
  \fi
}
%    \end{macrocode}
% \Finale
% \bibliographystyle{plain}
% \bibliography{apxproof}